Differential Equations in Applications

1
VVAMEtKIN
M!
8
SCI
E
NCE
FOR EV
E
RyNE
This book is an easy-reading introduction to ordinary
differential equations and their use in the study of real
phenomena and processes. Problems taken 'rom
various fields of knowledge illustrate the tools used i
setting up differential equations and the methods
employed in their qualitative investigation. The book
should be useful to high-school students, teachers of
science courses, college students, and specialists of
nonmathematical professions who use mathematics i
their w.-:k.
ISBN 5-03-0005218 Science
f or Everyone
Science
for Everyone B, I3. Amenmmn
J.[[/I(I}q)BpBH]1H3JIbHLI YPBBHGHI/IH
B [Ip]Z[.TIO}P{9HI/IHX
HBILHTGJIBGTBO (<H&yK3>) MOCKB&
B. 13. AMeJI&KHH
}.J; HifliPepeH1:J;II8JibHhle ypa BHeHIIH
B IIpliJIO meHIIHX
HaAaTeJI&CTBO << HayKa)> MocKBa v V Amelkin [ Z
Differential Equations
in Applications
\"
Mir
Publishers
Moscow
V V Amel'kin ( :
4
Differential Equations
in Applications
Mir
Publishers
Moscow Translated from the Russian by Eugene Yankovsky
First published 1990
Revised from the 1987 Russian edition
Ha anenuiicxom xsane
Printed in the Union of Soviet Socialist Republics
ISBN 5-03-000521-8
I/Iairaremsorno <<Hay1<a. Fnannaa penarmna
|i}H3HI{0MlTM&TH'ICKOi nnreparypu, 1987
English translation, Eugene Yankovsky, 1990
Translated from the Russian by Eugene Yankovsky
First published 1990
Revised from the 1987 Russian edition
Printed in the Union of Soviet Socialist Republics
ISBN 5-03-000521-8
@ llap;aTeJILCTBO HayKa. fJiaBHaJI pep;aKJ.J;IIJI
c}III311KD-MaTeMaTHqecKQH JIBTepaTypLI, 1987
English translation, Eugene Yankovsky, 1990 Contents
Preface 7
Chapter 1. Construction of Differen-
tial Models and Their Solutions 11
1.1 Whose Coffee Was Hotter? 11
1.2 Steady-State Heat Flow . . . 14
1.3 An Incident in a National Park 18
1.4 Liquid Flow Out of Vessels. The
Water Clock . . . . 26
1.5 Effectiveness of Advertising 30
1.6 Supply and Demand 32
1.7 Chemical Reactions . . . 34
1.8 Differential Models in Ecology 38
1.9 A Problem from the Mathematical
Theory of Epidemics 44
1.10 The Pursuit Curve 51
1.11 Combat Models . . . 55
1.12 Why Are Pendulum Clocks Inaccu-
rate? ....... 69
1.13 The Cycloidal Clock . . . 73
1.14 The Brachistochrone Problem 81
1.15 The Arithmetic Mean, the Geo-
metric Mean, and the Associated
Differential Equation .... 88
1.16 On the Flight of an Object Thrown
at an Ang e to the Horizon 93
1.17 Weightlessness ......... 96
1.18 Kep ers Laws of Planetary Motion 100
1.19 Beam Deflection . . . 112
1.20 Transportation of Logs 119
Chapter 2. Qualitative Methods of
Studying Differential Models . . 136
2.1 Curves of Constant Direction of
Magnetic Needle ........ 136
Contents
Preface
Chapter 1. Construction of Differen-
tial Models and Their Solutions
1.1 Whose Coffee Was Hotter?
1.2 Steady-5tate Heat Flow . . .
1.3 An Incident in a National Park
1.4 Liquid Flow Out of Vessels. The
Water Clock . . .
1.5 Effectiveness of Advertising
1.6 Supply and Demand
1. 7 Chemical Reactions .
1.8 Differential Models in Ecology
1.9 A Problem from the Mathematical
Theory of Epidemics
1.10 The Pursuit Curve
1.11 Combat Models .
1.12 Why Are Pendulum Clocks Inaccu-
rate? . . . . . . .
1.13 The Cycloidal Clock . . .
1.14 The Brachistochro11e Problem
1.15 The Arithmetic Mean, the Geo-
metric Mean, and the Associated
Differential Equation .
1.16 On the Flight of an Object Thrown
at an Angle to the Horizon
1.17 Weightlessness . .
1.18 Kepler's Laws of Planetary Motion
1.19 Beam Defiection .
1.20 Transportation of Logs
Chapter 2. Qualitative Methods of
Studying Differential Models
2.1 Curves of Constant Direction of
Magnetic Needl& ,
7
11
11
14
18
26
30
32
34
38
44
51
55
69
73
81
88
93
96
100
112
119
136
136 6 Contents
2.2 Why Must an Engineer Know
Existence and Uniqueness Theo-
rems? . . . . . . . 143
2.3 A Dynamical Interpretation of
Second-Order Differential Equa-
tions... . ... 155
2.4 Conservative Systems in Mechanics 162
2.5 Stability of Equilibrium Points
and of Periodic Motion 176
2.6 Lyapunov Functions . . . 184
2.7 Simple States of Equilibrium . 189
2.8 Motion of a UnitMass Object
Under the Action of Linear
Springs in a Medium with Linear
Drag . . . . . . . 195
2.9 Adiabatic Flow of a Perfect Gas
Through a Nozzle of Varying
Cross Section . . ..... 203
2.10 Higher-Order Points of Equilib-
rium . . . . . . . 211
2.11 Inversion with Respect to a Circle
and Homogeneous Coordinates 218
2.12 Flow of a Perfect Gas Through a
Rotating Tube of Uniform Cross
Section . ..... 223
2.13 Isolated Closed Trajectories . 236
2.14 Periodic Modes in Electric Circuits 250
2.15 Curves Without Contact . . 260
2.16 The Topographical System of
Curves. The Contact Curve . . . 263
2.17 The Divorgence of a Vector Field
and Limit Cycles 270
Selected Readings 274
Appendices 275
Appendix 1. Derivatives of Elemen-
tary Functions . . 275
Appendix 2. Basic Integrals .,.. . 278
6 Contents
Existence and Uniqueness Theo-
rems? . 143
2.3 A Dynamical Interpretation of
Second-Order Differential Equa-
tions . . . . . 155
2.4 Conservative Systems in Mechanics 162
and of Periodic Motion 176
2.6 Lyapunov Functions . . . 184
2. 7 Simple States of Equilibrium . 189
2.8 Motion of a Unit-Mass Object
Under the Action of Linear
Springs in a Medium with Linear
Drng . 1 ~
2.9 Adiabatic Flow of a Perfect Gas
Through a Nozzle of Varying
Cross Section . . . . . 203
2.10 Higher-Order Points of Equilib-
rium . . 211
2. H Inversion with Hespect to a Circle
and Homogeneous Coordinates 218
2.12 Flow of a Perfect Gas Through a
Rotating Tube of Uniform Cross
Section . . . . . . 223
2.13 Isolated Closed Trajectories . 236
2.14 Periodic Modes in Electric Circuits 250
2.15 Curves Without Contact . . 260
2.16 The Topographical System of
Curves. The Contact Curve . . . 263
2.17 The Divergence of a Vector Field
and Limit Cycles 270
Selected Readings 274
Appendices 275
Appendix 1. Derivatives of Elemen-
tary Functions . . 275
Appendix 2. Basic Integrals , , . , 278 Preface
Differential equations belong to one of the
main mathematical concepts. They are
equations for finding functions whose deriv-
atives (or differentials) satisfy"given condi-
tions. The differential equations arrived at
in the process of studying a real phenomenon
or process are called the differential model
of this phenomenon or process. It is clear
that differential models constitute a particu-
lar case of the numerous mathematical models
that can be built as a result of studies of
the world that surrounds us. It must be
emphasized that there are different types of
differential models. This book considers none
but models described by what is known
as ordinary differential equations, one char-
acteristic of which is that the unknown
functions in these equations depend on
a single variable.
In constructing ordinary differential mo-
dels it is important to know the laws of
the branch of science relating to the nature
of the problem being studied. For instance,
in mechanics these may be Newtons laws,
in the theory of electric circuits Kirchhoffs
laws, in the theory of chemical reaction
rates the law of mass action.
Of course, in practical life we often have
to deal with cases where the flaws that en-
Preface
Differential equations belong to one of the
main mathematical concepts. They are
equations for f:tnding functions whose deriv-
atives (or differentials) condi-
tions. The differential equations arrived at
in the process of studying a real phenomenon
or process are called the differential model
of this phenomenon or process. It is clear
that differential models constitute a particu-
lar case of the numerous mathematical models
that can be built as a result of studies of
the world that surrounds us. It must be
emphasized that there are different types of
differential models. This book considers none
but models described by what is known
as_ ordinaru, dU'fprential em;ations. ,one_ charO'
acteristic of which is that the unknown
functions in these equations depend on
a single variable.
In constructing ordinary differential mo-
dels it is important to know the laws of
the branch of science relating to the nature
of the problem being studied. For instance,
in mechanics these may be Newton's laws,
in the theory of electric circuits Kirchhoff's
laws, in the theory of chemical reactian
rates the law of mass action.
Of course, in practical life we oflen have
to deal with cases where the laws that en-8 Preface
able building a differentia lequation (or sev-
eral differential equations) are not known,
and we must resort to various assumptions
(hypotheses) concerning the course of the
process at small variations of the parame-
ters, the variables. Passage to the limit
will then lead to a differential equation,
and if it so happens that the results of
investigation of the differential equation as
the mathematical model agree with the ex-
perimental data, this will mean that the
hypothesis underlying the model reflects
the true situation.*
When working on this book, I had two
goals in mind. The first was to use examples,
rich in content rather than purely illustra-
tive, taken from various fields of knowledge
so as to demonstrate the possibilities of
using ordinary differential equations in
gaining an understanding of the world about
us. Of course, the examples far from ex-
haust the scope of problems solvable by or-
dinary differential equations. They give
an idea of the role that ordinary differential
equations play in solving practical problems.
The second goal was to acquaint the read-
* If the reader wishes to know more about mathe-
matical models, he can turn to the fascinating books
by A.N. Tikhonov and D.P. Kostomarov, Stories
About Applied Mathematics (Moscow: Nauka,
1979) (in Russian), and N .N. Moiseev, Mathemat-
igs Stages an Experiment (Moscow: Nauka, 1979)
(nn Russian).
8 Preface
able building a differentia !equation (or sev-
eral differential equations) are not known,
and we must resort to various assumptions
(hypotheses) concerning the course of the
process at small variations of the parame-
ters, the variables. Passage to the limit
will then lead to a differential equation,
and if it so happens that the results of
investigation of the differential equation as
the mathematical model agree with the ex-
perimental data, this will mean that the
hypothesis underlying the model reflects
the true situation.*
When working on this book, I had two
goals in mind. The first was to use examples,
rich in content rather than purely illustra-
tive, taken from various fields of knowledge
so as to demonstrate the possibilities of
using ordinary differential equations in
gaining an understanding of the world about
us. Of course, the examples far from ex-
haust the scope of problems solvable by or-
dinary differential equations. They give
an idea of the role that ordinary differential
equations play in solving practical problems.
The second goal was to acquaint the read-
If the reader wishes to know more about mathe-
matical models, he can turn to the fascinating books
by A.N. Tikhonov and D.P. Kostomarov, Stories
About Applied Mathematics (Moscow: Nauka,
1979) (in Russian), and N.N. Moiseev, Mathemat-
ics Stages an Ezpertmeflt Naqka, 1979)
(in Preface 9
er with the simplest tools and methods
used in studying ordinary differential equa-
tions and characteristic of the qualitative
theory of diyjerential equations. The fact is
that only in rare cases are we able to solve
a differential equation in the so-called
closed form, that is, represent the solution
as a formula that employs a finite number
of the simplest operations involving ele-
mentary functions, even when it is known
that the differential equation has a solu-
tion. In other words, we can say that the
great variety of solutions to differential
equations is such that for their representa-
tion in closed form a finite number of ana-
lytical operations is insufficient. A simi-
lar situation exists in the theory of algebraic
equations: while for first- and second-order
algebraic equations the solutions can always
be easily expressed in terms of radicals
and for third- and fourth-order equations
the solutions can still be expressed in terms
of radicals (although the formulas become
very complicated), for a general algebraic
equation of an order higher than the fourth
the solution cannot be expressed in radicals.
To return to differential equations. If
an infinite series of this or that form is used
to represent the solutions, then the scope
of solvable equations broadens considerab-
ly. Unfortunately, it often happens that
the most essential and interesting proper-
ties of the solutions cannot be revealed by
Preface
9
er with the simplest tools and methods
used in studying ordinary differential equa-
tions and characteristic of the qualitative
theory of differential equations. The fact is
that only in rare cases are we able to solve
a differential equation in the so-called
closed form, that is, represent the solution
as a formula that employs a finite number
of the simplest operations involving ele-
mentary functions, even when it is known
that the differential equation has a solu-
tion. In other words, we can say that the
great variety of solutions to differential
equations is such that for their representa-
tion in closed form a finite number of ana-
lytical operations is insufficient. A simi-
lar situation exists in the theory of algebraic
equations: while for first- and second-order
algebraic equations the solutions can always
be easily expressed in terms of radicals
and for third- and fourth-order equations
the solutions can still be expressed in terms
of radicals (although the formulas become
very complicated), for a general algebraic
equation of an order higher than the fourth
the solution cannot be expressed in radicals.
To return to differential equations. If
an infinite series of this or that form is used
to represent the solutions, then the scope
of solvable equations broadens considerab-
ly. Unfortunately, it often happens that
the most essential and interesting proper-
ties of the solutions cannot be revealed by 10 Preface
studying the form of the series. More than
that, even if a differential equation can be
solved in closed form, more often than not
it is impossible to analyze such a solution
since the relationship between the various
parameters of the solution often proves
to be extremely complicated.
This shows how important it is to develop
methods that make it possible to acquire
the data on the various properties of the
solutions without solving the differential
equations themselves. And indeed, such
methods do exist. They constitute the es-
sence of the qualitative theory of differen-
tial equations, which is based on the gener-
al theorems regarding the existence and
uniqueness of solutions and the continuous
dependence of solutions on the initial data
and parameters. The role of existence and
uniqueness theorems is partially discussed
in Section 2.2. As for the general qualitative
theory of ordinary differential equations,
ever since J.H. Poincar and A.M. Lya-
punov laid the foundations at the end of
the 19th century, the theory has been inten-
sively developing and its methods are wide-
ly used when studying the world about us.
I am indebted to Professors Yu.S. Bog-
danov and M.V. Fedoryuk for the construc-
tive remarks and comments expressed in
the process of preparing the book for publi-
cation.
. . l _ V.V, Amefkin
10 Preface
studying the form of the series. More than
that, even if a differential equation can be
solved in closed form, more often than not
it is impossible to analyze such a solution
since the relationship between the various
parameters of the solution often proves
to be extremely complicated.
This shows how important it is to develop
methods that make it possible to acquire
the data on the various properties of the
solutions without solving the differential
equations themselves. And indeed, such
methods do exist. They constitute the es-
sence of the qualitative theory of differen-
tial equations, which is based on the gener-
al theorems regarding the existence and
uniqueness of solutions and the continuous
dependence of solutions on the initial data
and parameters. The role of existence and
uniqueness theorems is partially discussed
in Section 2.2. As for the general qualitative
theory of ordinary differential equations,
ever since J .H. Poincare and A.M. Lya-
punov laid the foundations at the end of
the 19th century, the theory has been inten-
sively developing and its methods are wide-
ly used when studying the world about us.
I am indebted to Professors Yu.S. Bog-
danov and M.V. Fedoryuk for the construc-
tive remarks and comments expressed in
the process of preparing the book for publi-
cation.
V.V. Chapter 1
Construct1on
of D1fferent1al Models
and Their Solut1ons
When Tom and Dick ordered coffee and
cream in a lunch room, they were given both
simultaneously and proceeded as follows.
Tom poured some of his cream into the cof-
fee, covered the cup with a paper napkin,
and went to make a phone call. Dick cov-
ered his cup with a napkin and poured the
same amount of cream into his coffee only
after 10 minutes, when Tom returned. The
two started drinking their coffee at tl1e same
time. Whose was hotter?
We will solve this problem on the natural
assumption that according to the laws of
physics heat transfer through the surface of
the table and the paper napkin is much less
than through the sides of the cups and that
the temperature of the vapor above the sur-
face of the coffee in the cups equals the tem-
perature of the coffee.
We start by deriving a relationship indi-
cating the time dependence of the tempera-
ture of the coffee in Dicks cup before the
cream is added.
In accordance with our assumption on
the basis of a law of physics, the amount
Chapter 1
Construction
of Differential Models
and Their Solutions
When Tom and Dick ordered coffee and
cream in a lunch room, they were given both
simultaneously and proceeded as follows.
Tom poured some of his cream into the cof-
fee, covered the cup with a paper napkin,
and went to make a phone call. Dick cov-
ered his cup with a napkin and poured the
same amount of cream into his coffee only
after 10 minutes, when Tom returned. The
two started drinking their coffee at the same
time. Whose was hotter?
We will solve this problem on the natUl'al
assumption that according to the laws of
physics heat transfer through the surface of
the table and the paper napkin is much less
than through the sides of the cups and that
the temperature of the vapor above the sur-
face of the coffee in the cups equals the tem-
perature of the coffee.
We start by deriving a relationship indi-
cating the time dependence of the tempera-
ture of the coffee in Dick's cup before the
cream is added.
In accordance with our assumption on
the basis of a law of physics, the amount 12 Differential Equations in Applications
of heat transferred to the air from Dicks
cup is determined by the formula
dQ Z qui}?-Sat, (1)
where T is the coffees temperature at time
t, 6 the temperature of the air in the lunch
room, 1] the thermal conductivity of the
material of the cup, l the thickness of the
cup, and s the area of the cups lateral sur-
face. The amount of heat given off by the
coffee is
dQ = -cm dT, (2)
where c is the specific heat capacity of the
coffee, and m the mass (or amount) of coffee
in the cup. If we now consider Eqs. (1)
and (2) together, we arrive at the following
equation:
q J-`{it-dr: -cmdr,
which, after variable separation, can be re-
written as follows:
dT __ qs
-T_6 T:;- dt. (3)
Denoting the initial temperature of the coffee
by T0 and integrating the differential equa-
tion (3), we find that
T=6|(T09)exp(T;l%t), (4)
This formula is the analytical expression
of the law whereby the temperature of the
12 Differential Equations in Applications
of heat transferred to the air from Dick's
cup is determined by the formula
T-9
dQ='I'}-
1
-sdt, (1)
where T is the coffee's temperature at time
t, e the temperature of the air in the lunch
room, 'I'} the thermal conductivity of the
material of the cup, l the thickness of the
cup, and s the area of the cup's lateral sur-
face. The amount of heat given off by the
coffee is
dQ = -cmdT, (2)
where c is the specific heat capacity of the
coffee, and m the mass (or amount) of coffee
in the cup. If we now consider Eqs. (1)
and (2) together, we arrive at the following
equation:
T-9
'I'} -
1
-sdt= -cmdT,
which, after variable separation, can be re-
written as follows:
dT fjS
T-9 =- lcm dt. (3)
Denotingtheinitial temperature of the coffee
by T
0
and integrating the differential equa-
tion (3), we find that
T=9+(T
0
- 9) exp (- ~ ~ ~ t). (4)
This formula is the analytical expression
of the law whereby the temperature of the Ch. 1. Construction of Differential Models 13
coffee in Dicks cup varied prior to the addi-
tion of the cream.
[Now let us establish the law whereby the
temperature of the coffee in Dicks cup
changed after Dick poured in cream. We
use the heat balance equation, which here
can be written as
cm (T Bo) = 1mi(6D Ti). (5)
where BD is the temperature of the Dicks
coffee with cream at time t, T1 the temper-
ature of the cream, C1 the specific heat ca-
pacity of the cream, and ml the mass of
cream added to the coffee.
Equation (5) yields
i Clml Cm
9D cm-|c1m1 Tl-l_ cm|c1m1 T` (6)
Bearing in mind formula (4), we can re-
write (6) as follows:
HD CTn+C1m1 1 + cm+c1m1
- ....I|f.
>< [a+<T.. >xp( ,,,,, )]. <7>
which constitutes the law whereby the tem-
perature of the coffee in Dicks cup varies
after cream is added.
To derive the law for temperature varia-
tion of the coffee in Toms cup we again
employ the heat balance equation, which
now assumes the form
Cm (T0 " 90) = cimi (90 "" T1): (8)
Ch. 1. Construction of Differential Models 13
coffee in Dick's cup varied prior to the addi-
tion of the cream .
Now let us establish the law whereby the
temperature of the coffee in Dick's cup
changed after Dick poured in cream. We
use the heat balance equation, which here
can be written as
em (T - Bn) = e
1
m
1
(8n - T
1
}, (5)
where 8
0
is the temperature of the Dick's
coffee with cream at time t, T
1
the temper-
ature of the cream, e
1
the specific heat ca-
pacity of the cream, and m
1
the mass of
cream added to the coffee.
Equation (5) yields
8n= Ctml Tt + em T. (6)
cm+c
1
m
1
cm+c
1
m
1
Bearing in mind formula (4), we can re-
write (6) as follows:
x[e+(T
0
-8)exp(-
1
~ t)J, (7)
which constitutes the law whereby the tem-
perature of the coffee in Dick's cup varies
after cream is added.
To derive the law for temperature varia-
tion of the coffee in Tom's cup we again
employ the heat balance equation, which
now assumes the form
em (T
0
- 6
0
) = e
1
m
1
(6
0
- T 1), (8) 14 Differential Equations in Applications
where 00 is the temperature of the mixture.
lf we solve (8) for 00, we get
c m em
69: cmil-dlml 11*+ cm|c1m1 TO'
Then, using Eq. (4) with 0,, serving as the
initial temperature and cm + clml substi-
tuted for cm., we arrive at the law for the
temperature (0T) variation of the coffee
in Toms cup as analytically given by the
following formula:
"_ Clml - Cm L
9T`"`9+li m|c1m1 Ti T m|c1m1 To 9]
.. _.._..._..*lS
X Xp( l(m+ Cimi) tl (9)
Thus, to answer the question posed in
the problem we need only turn to formulas
(7) and (9) and carry out the necessary cal-
culations, bearing in mind that C1 z 3.9 ><
103 J/kgK, cx.1 ><103 J/kgK, and
nz0.6 V/mK and assuming, for the sake
of definiteness, that ml : 2 >< 10*2 kg,
m=8 >< 10*2 kg, T1 =20C, 0 :20C,
T0 =80C, s = 11 ><10*3 mz, and Z:
2 X 10*3 m. The calculations show that
Toms coffee was hotter.
1.2 Steady-State Heat Flow
The reader will recall that a steady-state
heat flow is- one in which the objects tem-
where eo is the temperature of the mixture.
If we solve (8) for e
0
, we get
Then, using Eq. (4) with eo serving as the
initial temperature and em + e
1
m
1
substi-
tuted for em, we arrive at the law for the
temperature (eT) variation of the coffee
in Tom's cup as analytically given by the
following formula:
Thus, to answer the question posed in
the problem we need only turn to formulas
(7) and (9) and carry out the necessary cal-
culations, bearing in mind that e
1
~ 3.9 X
10
3
J/kgK, e ~ 4.1 X 10
3
J/kgK, and
'11 ~ 0. 6 V /m K and assuming, for the sake
of definiteness, that m
1
= 2 X 10-
2
kg,
m = 8 X 10-
2
kg, T
1
= 20 oc, e = 20 C,
T
0
= 80 C, s = 11 X 10-
3
m
2
, and l =
2 X 10-
3
m. The calculations show that
Tom's coffee was hotter.
1.2 Steady-State Heat Flow
The reader will recall that a steady-state
heat flow is one in which the object's tem-Ch. 1. Construction of Differential Models 15
l \
\ w
\ u \
% n \
it N
a
x
n 2 is
ni n in
20
Fig. 1 Fig. 2
perature at each point does not vary with
time.
II'1 pI'Ol'JlI1'lS Wll0SB physical COI`ll,Ilt is
related to the effects of heat flows, an impor-
tant role is played by the socalled isother-
mal surfaces. To clarify this statement, let
US COI1SldI 3 h3lCOIldllCtlI1g (FlgÙ.I'
1) 20 cm in diameter, made of a homogene-
ous material, and protected by a layer of
magnesium oxide 10 cm thick. We assume
that the temperature of the pipe is 160 C
and the outer surface of the protective cov-
ering has a temperature of 30 C. It is
intuitively clear then that there is a surface,
designated by a dashed curve in Figure
2, at each point of which the temperature
is the same, say 95 C. The dashed curve in
Figure 2 is known as an isotherm, while the
20
Fig. 1 Fig. 2
perature at each point does not vary with
time.
In problems whose physical content is
related to the effects of heat flows, an impor-
tant role is played by the so-called isother-
mal surfaces. To clarify this statement, let
us consider a heat-conducting pipe (Figure
1) 20 em in diameter, made of a homogene-
ous material, and protected by a layer of
magnesium oxide 10 em thick. We assume
that the temperature of the pipe is 160 oc
and the outer surface of the protective cov-
ering has a temperature of 30 C. It is
intuitively clear then that there is a surface,
designated by a dashed curve in Figure
2, at each point of which the temperature
is the same, say 95 C. The dashed curve in
Figure 2 is known as an isotherm, while the 16 Differential Equations in Applications
surface corresponding to this curveiis known
as an isothermal surface. In general, iso-
therms may have various shapes, depend-
ing, for one, on the nonsteady-state nature
of the heatiflow and on the nonhomogeneity
of the material. In the case at hand the
isotherms (isothermal surfaces) are represent-
ed by concentric circles (cylinders).
We wish to derive the law of temperature
distribution inside the protective coating
and find the amount of heat released by the
pipe over a section 1 m long in the course
of 24 hours, assuming that the thermal con-
ductivity coefficient lc is equal to 1.7 ><
10*4.
To this end we turn to the Fourier law,
according to which the amount of heat re-
leased per unit time by an object that is in sta-
ble thermal state and whose temperature T at
each point is solely a function of coordinate x
can be found according to the formula
dT
Q = -kF (ac) -8-5:- == const, (IO)
where F (x) is the cross-sectional area normal
to the direction of heat flow, and lc the thermal
conductivity coefficient.
The statement of the problem implies
that F (:2:) = 2;r;xl, where l is the length of
the pipe (cm), and sc the radius of the base
of the cylindrical surface lying inside the
outer cylinder. Then on the basis of (IO)
surface corresponding to this curve:is known
as an isothermal surface. In general, iso-
therms may have various shapes, depend-
ing, for one, on the nonsteady-state nature
of the heat[flow and on the nonhomogeneity
of the material. In the case at hand the
isotherms (isothermal surfaces) are represent-
ed by concentric circles (cylinders).
We wish to derive the law of temperature
distribution inside the protective coating
and find the amount of heat released by the
pipe over a section 1 m long in the course
of 24 hours, assuming that the thermal con-
ductivity coefficient k is equal to 1.7 X
10-4.
To this end we turn to the Fourier law,
according to which the amount of heat re-
leased per unit time by an object that is in sta-
ble thermal state and whose temperature T at
each point is solely a function of coordinate x
can be found according to the formula
dT
Q = -kF (x) dX = const, (10)
where F (x) is the cross-sectional area normal
to the direction of heat flow, and k the thermal
conductivity coefficient.
The statement of the problem implies
that F (x) = 2nxl, where l is the length of
the pipe (em), and x the radius of the base
of the cylindrical surface lying inside the
outer cylinder. Then on the basis of (10) Ch. 1. Construction of Differential Models 17
we get
s0 Q 20 d
I
S dT: _ 0.00017><2nz S T (M)
160 10
T Q x dg
S dt: " 0.00017><2nz S (12)
160 10
Integrating (11) and (12) yields
160-T _ ln 0.1x __ log 0.1.1:
130 " 1n2 "' Iog2
Hence
T = 591.8 431.8 log .7:.
This formula expresses the law of tempera-
ture distribution inside the protective
coating. We see that the length of the pipe
plays no role in this law.
To answer the second question, we turn
to Eq. (11). Then for l = 100 cm we find
that
Q ___ 130><0.00017><2 ><100
1n2
_ 200n><130><0.00017
" 0.69315
and, hence, the amount of heat released by
the pipe in the course of 24 hours is 24 X
60 X GOQ -= 726 852 J.
2-0770
we get
30 20
) dT = - 0 . 0 0 0 1 ~ X 2:tl ~ dxx '
160 10
T x
~ d-r = - o.oooS x 2:tl ~ ~ ~
160 10
Integrating (11) and (12) yields
160-T
130
Hence
ln0.1x logO.tx
In 2 - log 2
T = 591.8 - 431.8 log x.
(11)
(12)
This formula expresses the law of tempera-
ture distribution inside the protective
coating. We see that the length of the pipe
plays no role in this law.
to Eq. (11). Then for l = 100 em we find
that
Q
_ 130 X 0.00017 X 2 X 100
- ln2
200:t X 130 X 0.00017
0.69315
and, hence, the amount of heat released by
the pipe in the course of 24 hours is 24 X
60 X 60Q = 726 852 J.
2-0770 18 Differential Equations in Applications
While patrolling a national park,two for-
est rangers found the carcass- of a slain wild
boar. Inspection showed that the poachers
shot had been precise and the boar had died
on the spot. Reasoning that the poacher
should return for the kill, the rangers de-
cided to wait for him and hid nearby. Soon
two men appeared, heading directly for the
dead boar. When confronted by the rangers,
the men denied having anything to do with
the poaching. But by this time the rangers
had collected indirect evidence of their
guilt. It only remained to establish the
exact time of the kill. This they did by
using the law of heat emission. Let us see
what reasoning this involved.
According to the law of heat emission,
the rate at which an object cools off in air
is proportional to the difference between
the temperature of the object and that of
the air,
2%.: -k (x-G), (13)
where x is the temperature of the object at
time t, a the temperature of the air, and
k a positive proportionality coefficient.
Solution of the problem lies in an analy-
sis of the relationship that results from in-
tegrating the differential equation (13).
Here one must bear in mind that after the
While patrolling a national park, two for-
est rangers found the carcass- of a slain wild
boar. Inspection showed that the poacher's
shot had been precise and the boar had died
on the spot. Reasoning that the poacher
should return for the kill, the rangers de-
cided to wait for him and hid nearby. Soon
two men appeared, heading directly for the
dead boar. When confronted by the rangers,
the men denied having anything to do with
the poaching. But by this time the rangers
had collected indirect evidence of their
guilt. It only remained to establish the
exact time of the kill. This they did by
using the law of heat emission. Let us see
what reasoning this involved.
According to the law of heat emission,
the rate at which an object cools off in air
is proportional to the difference between
the temperature of the object and that of
the air,
dz
-=-k(x-a)
dt '
(13)
where x is the temperature of the object at
time t, a the temperature of the air, and
k a positive proportionality coefficient.
Solution of the problem lies in an analy-
sis of the relationship that results from in-
tegrating the differential equation (13).
Here one must bear in mind that after the Ch. 1. Construction of Differential Models 19
boar was killed, the temperature of the air
could have remained constant but also
could have varied with time. In the first case
integration of the differential equation (13)
with variables separable leads to
1n%-5%: kt, xqa, (14)
where xo is the temperature of the object
at time t = O. Then, if at the time when
the strangers were confronted by the rangers
the temperature of the carcass, x, was 31 C
and after an hour 29 C and if when the shot
was fired the temperature of the boar was
:1; = 37 C and the temperature of the air
a = 21 C, we can establish when the shot
was fired by putting t = O as the time when
the strangers were detained. Using these
data and Eq. (14), we find that
31--21
Now, substituting this value of lc and .2: =
37 into (14), we get
t 14.; In
" 0.22314 31-21
1
- --OE-@ln 1.6: -2.10630.
In other words, roughly 2 hours and 6 min-
utes passed between the time the boar was
killed and the time the strangers were de-
tained.
2*
boar was killed, the temperature of the air
could have remained constant but also
could have varied with time. In the first case
integration of the differential equation (13)
with variables separable leads to
I n ~ = -kt, x=l=a, (14)
x
0
-a
where x
0
is the temperature of the object
at time t = 0. Then, if at the time when
the strangers were confronted by the rangers
the temperature of the carcass, x, was 31 C
and after an hour 29 C and if when the shot
was fired the temperature of the boar was
x = 37 oc and the temperature of the air
a = 21 oc, we can establish when the shot
was fired by putting t = 0 as the time when
the strangers were detained. Using these
data and Eq. (14), we find that
31-21
k=ln
29
_
21
=ln1.25=0.22314. (15)
Now, substituting this value of k and x =
37 into (14), we get
1 37-21
t = - 0.22314 ln 31-21
1
=-
0
.
22314
ln 1.6= -2.10630.
In other words, roughly 2 hours and 6 min-
utes passed between the time the boar was
killed and the time the strangers were de-
tained.
2* 20 Diierential Equations in Applications
In the second case, that is, when the tem-
perature of the air varies in time, the cool-
ing off of the carcass is expressed by the
following nonhomogeneous linear differen-
tial equation
$35- }kx: lca(t), (16)
where 0, (t) is the temperature of the air at
time 13.
To illustrate one of the methods for deter-
mining the time when the boar was killed,
let us assume that the temperature of the
carcass was 30 C when the strangers were
detained. Let us also suppose that it is
known that on the day of the kill the tem-
perature of the air dropped by 1 C every
hour in the afternoon and was 0 C when
the carcass was discovered. We will also
assume that after an hour had passed after
the discovery the temperature of the carcass
was 25 C and that of the air was down
to 1 C. If we now assume that the shot
was fired at t=0 and that :1:0 = 37 C
at t = 0, we get a (t) = t* t, where
t-= 15* is the time when the rangers dis-
covered the carcass.
Integrating Eq. (16), we get
x=(37tlc1)e"|t* - t {- k1.
If we 11ow bear in mind that sc = 30 C
at t=t* and :c=25C at t-==t*-{-1,
In the second case, that is, when the tem-
perature of the air varies in time, the cool-
ing off of the carcass is expressed by the
following nonhomogeneous linear differen-
tial equation
: ~ +kx=ka(t),
(16)
where a (t) is the temperature of the air at
time t.
To illustrate one of the methods for deter-
mining the time when the boar was killed,
let us assume that the temperature of the
carcass was 30 oc when the strangers were
detained. Let us also suppose that it is
known that on the day of the kill the tem-
perature of the air dropped by 1 C every
hour in the afternoon and was 0 C when
the carcass was discovered. We will also
assume that after an hour had passed after
the discovery the temperature of the carcass
was 25 C and that of the air was down
to -1 C. If we now assume that the shot
was fired at t = 0 and that x
0
= 37 C
at t = 0, we get a (t) = t* - t, where
t = t* is the time when the rangers dis-
covered the carcass.
Integrating Eq. (16}, we get
x = (37 - t - k-
1
) e-kt + t* - t + k-1.
If we now bear in mind that x = 30 C
at t = t* and x = 25 C at t = t* + 1, Ch. 1. Construction of Differential Models 21
the last formula yields
(37 - t* - Irl) exp (-Ict*) -l- Irl == 30,
(37 t* - Irl) exp [k (t* -}- 1)]
"" kl-1 7-' 26.
These two equations can be used to derive
an equation for Ic, namely,
(30 -Irl)e* 26 + Irl = 0. (17)
We can arrive at the same equation start-
ing from different assumptions. Indeed,
suppose that at if: 0 the carcass of the slain
boar was found. Then a (t) == t and
we arrive at the differential equation
13%+ kx: -kz (18)
(with the initial data 2:,, = 30 at t= 0),
from which we must {ind as as an explicit
function of t.
Solving Eq. (18), we get
x = (30 Irl) exp (Ict) - t + Irl. (19)
Setting sc = 25 and t = 1 in this relation-
ship, we again arrive at Eq. (17), which ena-
bles solving the initial problem numerical-
ly.
Indeed, as is known, Eq. (17) cannot be
solved algebraically for Ic. But it is easily
solved by numerical methods for finding the
roots of transcendental equations, for one,
the last formula yields
(37 - t* - k-
1
) exp (-kt*) + k-
1
= 30,
(37 - t* - k-
1
) exp [- k (t* + 1)]
+k-
1
= 26.
These two equations can be used to derive
an equation for k, namely,
(30- k-
1
)e-h- 26 + k-
1
= 0. (17)
We can arrive at the same equation start-
ing from different assumptions. Indeed,
suppose that at t = 0 the carcass of the slain
boar was found. Then a (t) = -t and
we arrive at the differential equation
dx +
dt kx= -kt
(18)
(with the initial data x
0
= 30 at t = 0),
from which we must find x as an explicit
function of t.
Solving Eq. (18), we get
x = (30- k-
1
) exp (-kt) - t + k-
1
(19)
Setting x = 25 and t = 1 in this relation-
ship, we again arrive at Eq. (17), which ena-
bles solving the initial problem numerical-
ly.
Indeed, as is known, Eq. (17) cannot be
solved algebraically for k. But it is easily
solved by numerical methods for finding the
roots of transcendental equations, for one, 22 Differential Equations in Applications
by Newtons method of approximation. This
method, as well as other methods of succes-
sive approximations, is a way of using a
rough estimate of the true value of a root
used to obtain more exact values of the
root. The process can be continued until
the required accuracy is achieved.
To show how Newtons method is used,
we transform Eq. (17) to the form
30k 1 |- (1 26k) exp (lc) = 0, (20)
and Eq. (19), setting :1: = 37, to the form
(37k 1 l- kt) exp (kt) 30k l 1 = 0.
(21)
Both equations, (20) and (21), are of the
type
(asc |b) exp (kx) -icx {cl = 0. (22)
If we denote the left-hand side of Eq. (22)
by qa (sc), differentiation with respect to .2:
yields
cp' (sc) = (lax [ Ptb -|- a) exp (M:) + c,
cp" (x) = (Wax {- Vb -1- 2M) exp (kx).
Then according to Newtons method for
finding a root of Eq. (22), if for the ith
approximation xi we have the inequality
(P ($2) (P" (xi) > 0
tht? DXlL &pPI`0XlIIl&l&l0U, -76;+1, C3I1 130 f0U.Hd
by Newton's method of approximation. This
method, as well as other methods of succes-
sive approximations, is a way of using a
rough estimate of the true value of a root
used to obtain more exact values of the
root. The process can be continued until
the required accuracy is achieved.
To show how Newton's method is used,
we transform Eq. (17) to the form
30k - 1 + (1 - 26k) exp (k) = 0, (20)
and Eq. (19), setting x = 37, to the form
(37k- 1 + kt) exp (kt) - 30k + 1 = 0.
(21)
Both equations, (20) and (21), are of the
type
(ax + b) exp (A.x) + ex + d = 0. (22)
If we denote the left-hand side of Eq. (22)
by <p (x), differentiation with respect to x
yields
<p' (x) = (/.ax + Ab + a) exp (A.x) + c,
<p" (x) = (/.
2
ax + 'A
2
b + 2'J..a) exp (A.x).
Then according to Newton's method for
finding a root of Eq. (22), if for the ith
approximation Xi we have the inequality
<p (xi) <p" (xi) > 0,
the xi+l can be found Ch. 1. Construction of Differential Models 23
via the formula
x
xi+1 = xi
To directly calculate the root (to within,
say, one part in a million), we compile
the following program using BASIC*:
10 CLS:PRINT "Solution of equation by New-
tons method"
15 INPUT "lambda=", L;
20 INPUT "a==", A: INPUT "b=", B;
30 INPUT "c=", C: INPUT "d=", D
40 INPUT "approximate value of root-.=", X
50 PRINT "X", "f", "f", "f""
100 E = EXP (L*X)
110 F = (A*X -1- B)*E + C*X + D
120 F1 = (L*(A*X -1- B) -1- A)*E -1- C
130 F2 = L* (L* (A*X -1- B) -1- 2*A)*E
150 PRINT X, F, F1, F2
i51 IF F = 0 THEN END
155 IF F1 = 0 THEN PRINT "Newtons method
is divergent": END
170 IF F*F2<0 THEN PRINT "Newtons
method is divergent": END
190 X = X - F/F1
200 GOTO 100
In this program X, f, f' and f" stand for
kn, cp (kn), cp' (kn), and cp" (kn), respective-
ly.
* Edit0rs note. Some BASIC lines, e.g. line
10, have been split due to the printed format;
they should be input as one line.
via the formula
q> (zt)
Xi+t = Xf- q>' (xi)
To directly calculate the root (to within,
say, one part in a million), we compile
the following program using BASIC*:
10 CLS:PRINT "Solution of equation by New-
ton's method"
15 INPUT "lambda=", L;
20 INPUT "a=", A: INPUT "b=", B;
30 INPUT "c=", C: INPUT "d=", D
40 INPUT "approximate value of root=", X
50 PRINT "X", "f", "f"', "f'"'
100 E = EXP (L*X)
110 F = (A*X + B)*E + C*X + D
120 F1 = (L*(A*X +B)+ A)*E + C
130 F2 = L* (L* (A*X +B)+ 2*A)*E
150 PRINT X, F, F1, F2
151 IF F = 0 THEN END
155 IF F1 = 0 THEN PRINT "Newton's method
is divergent": END
170 IF F*F2 < 0 THEN PHINT "Newton's
method is divergent": END
190 X = X - F/F1
200 GOTO 100
In this program X, f, f' and f" stand for
kn, cp (kn), cp' (kn), and cp" (kn), respective-
ly.
" Edttor's note. Some BASIC lines, e.g. line
10, have been split due to the printed format;
they should be input as one line. 24 Differential Equations in Applications
After starting the program enter the re-
quested values of the coefficients of the equa-
tion and the initial value of the root. The
results can be listed in a table of the approx-
imate values of the root and the respective
values of the function and its iirst and
second derivatives.
_ Employing this general procedure, let us
turn to Eq. (20).Difierentiating its left-hand
side cp (k) with respect to lc, we arrive at
cp' (lc) = 30 - (25 -| 26}.:) exp (k).
It can the11 easily be verified that tp (0) =
0, cp (1) < 0, and q> (0) > 0. Thus, the
function cp increases in a small neighborhood
of the origin and then dec1eases to a nega-
tive value at lc = 1. This implies that in the
interval (0, 1) there is a root of the equation
cp (k) : 0. To hnd this root we run the pro-
gram. Below we give the protocol of the
program:
Solution of equation by Newtons method
lambda: 1
a=26 b=1c=30 d=1
Approximate value of root = 0.5
X f f' f"
.5 -5.784655 -32.65141 -105.5182
.322836 -1.525956 -16.11805 -82.02506
.2281622 -.3514252 -8.859806 -71.52333
.1884971 -5.519438E-02 -6.103379 -67.49665
.1794539 ~2.748013E-03 -5.497021 -66.60768
After starting the program enter the re-
quested values of the coefficients of the equa-
tion and the initial value of the root. The
results can be listed in a table of the approx-
imate values of the root and the respective
values of the function and its first and
second derivatives.
_ Employing this general procedure, let us
turn to Eq. (20). Differentiating its left-hand
side cp (k) with respect to k, we arrive at
cp' (k) = 30 - (25 + 26k) exp (k).
It can then easily be verified that 'P (0) =
0, cp (1) < 0, and cp' (0) > 0. Thus, the
function cp increases in a small neighborhood
of the origin and then decreases to a nega-
tive value at k = 1. This implies that in the
interval (0, 1) there is a root of the equation
cp (k) = 0. To find this root we run the pro-
gram. Below we give the protocol of the
program:
Solution of equation by Newton's method
lambda= 1
a = -26 b = 1 c = 30 d = - 1
Approximate value of root = 0.5
X f f' f"
.5 -5.784655 -32.65141 -105.5182
.322836 -1.525956 -16.11805 -82.02506
.2281622 -.3514252 -8.859806 -71.52333
.1884971 -5.519438E-02 -6.103379 -67.49665
-66,60768 Ch. 1. Construction of Differential Models 25
.178954 -7.629395E-06 -5.46373 -66.55884
.1789526 -4.768372E-07 -5.463642 -66.5587
.1789525 0 -5.463635 -66.5587
OK
In this protocol X, f, f', and f stand for
kn, rp (kn), cp' (kn), and cp" (kn), respective-
ly.
The final step in solving the problem con-
sists in substituting the calculated value
kn z k z 0.178952 into Eq. (21) and solv-
ing the latter for t (the time when the wild
boar was killed). To employ the above
scheme, we denote the lefthand side of Eq.
(21) by g (t). Then, selecting -1 for the
value of to and bearing in mind that in
this case a =k,b : 37k-1,c =0,d=
-30k-l1, and ft:-k, we can find the time
when the boar was killed. To this end
we find the coefficients b and d, which prove
to be equal to 5.621243 and -4.368575, re-
spectively, and then running the above pro-
gram, find the sought time t. The proto-
col of the program is given below:
Solution of equation by Newtons method
lambda = 0.1789525
a =-- 0.1789525 b = 5.621243 c = 0
d = -4.368575
Approximate value of root = -1
X f f' f"
-1 .181972 .9639622 .1992802
--1.188775 3.506184E-03 .9270549 .1917861
.178954 -7 .629395E-06 -5.46373
.1789526 -4.768372E-07 -5.463642
.1789525 0 -5.463635
OK
-66.55884
-66.5587
-66.5587
In this protocol X, f, f', and f" stand for
k
11
, cp (kn), P' (k
11
), and fP" (k
11
), respective-
ly.
The final step in solving the problem con-
sists in substituting the calculated value
k
6
~ k ~ 0.178952 into Eq. (21) and solv-
ing the latter for t (the time when the wild
boar was killed). To employ the above
scheme, we denote the left-hand side of Eq.
(21) by g (t). Then, selecting -1 for the
value of t
0
and bearing in mind that in
this case a = k, b = 37k- 1, c = 0, d =
-30k + 1, and j., = k, we can find the time
when the boar was killed. To this end
we find the coefficients b and d, which prove
to be equal to 5.621243 and -4.368575, re-
spectively, and then running the above pro-
gram, find the sought time t. The proto-
col of the program is given below:
Solution of equation by Newton's method
lambda= 0.1789525
a= 0.1789525 b = 5.621243 c = 0
d = -4.368575
Approximate value of root = -1
X f f' f"
-1 .181972 .9639622 .1992802
-1.188775 3.506184E-03 .9270549 .1917861 26 Differential Equations in Applications
`cs
H
Fig. 3
-1.192557 1.430512E-06 .9263298 .1916388
-1.192559 0 .9263295 .1916387
OK
tn, g (tn), g' (tn), and g" (tn), respectively.
These results imply that the boar was
killed approximately l hour and 12 minutes
before the rangers discovered the carcass.
1.4 Liquid Flow Out of Vessels.
The Water Clock
The two problems that we now discuss illu-
strate the relationship between the physical
content of a problem and geometry. But first
let us examine some general theoretical con-
clusions.
We take a vessel (Figure 3) whose hori-
zontal cross section has an area that is a
function of the distance from the bottom
of the vessel to the cross Section. Suppose
Fig. 3
-1.192557 1.430512E-06 .9263298 .1916388
-1.192559 0 .9263295 .1916387
OK
tn, g (tn), g' (tn), and g" (tn), respectively.
These results imply that the boar was
killed approximately 1 hour and 12 minutes
before the rangers discovered the carcass.
1.4 Liquid Flow Out of Vessels.
The Water Clock
The two problems that we now discuss illu-
strate the relationship between the physical
content of a problem and geometry. But first
let us examine some general theoretical con-
clusions.
We take a vessel (Figure 3) whose hori-
zontal cross section has an area that is a
function of the distance from the bottom
of the vessel to the cross section. S u p p o ~ Ch. 1. Construction of Differential Models 27
that initially at time t = O the level of the
liquid in the vessel is at a height of h
meters. We will also suppose that the area
of the vessels cross section at height x
is denoted by S (:2:) and that the area of the
opening in the bottom of the vessel is s.
As is known, the rate v at which the liquid
flows out of the vessel at the moment when the
liquids level is aiheight .2: is given by the for-
mula v == kl/2g:c, where g = 9.8 m/sz,
and k is the rate constant of the outflow pro-
cess.
In the course of an infinitesimal time
interval dt the outflow of the liquid can be
assumed uniform, whereby during dt a
column of liquid with a height of v dt
and a cross-sectional area of s will flow out
of the vessel, which causes the level of the
liquid to change by -dx (the "minus" be-
cause the level lowers).
The above reasoning leads us to the fol-
lowing differential equation
ks dt == S (:1:) dx,
which can be rewritten as
at T.- ...`-LY-$*9..;.-. d . 23
ks \/2gx at ( )
Let us now solve the following problem.
A cylindrical vessel with a vertical axis
six meters high and four meters in diameter
has a circular opening in the bottom. The
that initially at time t = 0 the level of the
liquid in the vessel is at a height of h
meters. We will also suppose that the area
of the vessel's cross section at height x
is denoted by S (x) and that the area of the
opening in the bottom of the vessel is s.
As is known, the rate vat which the liquid
flows out of the vessel at the moment when the
liquid's level is at height x is given by the for-
mula v = k V2gx, where g = 9.8 mfs'l,
and k is the rate constant of the outflow pro-
cess.
In the course of an infinitesimal time
interval dt the outflow of the liquid can be
assumed uniform, whereby during dt a
column of liquid with a height of v dt
and a cross-sectional area of s will flow out
of the vessel, which causes the level of the
liquid to change by -dx (the "minus" be-
cause the level lowers).
The above reasoning leads us to the fol-
ks V2gx dt = -S (x) dx,
dt =- S(x) dx.
ks y2gx
(23)
Let us now solve the following problem.
A cylindrical vessel with a vertical axis
six meters high and four meters in diameter
has a circular opening in the bottom. The 28 Differential Equations in Applications
radius of this opening is 1/12 m. Find how
the level of water in the vessel depends on
time t and the time it takes all the water to
flow out. T
By hypothesis, S (.1:) = 4at and s =
1/144. Since for water k = 0.6, Eq. (23)
assumes the form
_ __ 217.152 dx
dt V;
Integrating this differential equation yields
t= 434.304 [ l/6 l/EL Ogxgfi,
which is the sought dependence of the level
of water in the vessel on time t. If we put
x = 0 in the last formula, we find that it
takes approximately 18 minutes for all the
water to flow out of the vessel.
Now a second problem. An ancient water
clock consists of a bowl with a small hole
in the bottom through which water flows
out of the bowl (Figure 4). Such clocks were
used in ancient Greek and Roman courts to
time the lawyers speeches, so as to avoid
prolonged speeches. We wish to determine
the shape of the water clock that would en-
sure that the water level lowered at a con-
stant rate.
This problem can easily be solved via
Eq. (23). We rewrite this equation in the
form
_ __ S (=> Q;
radius of this opening is 1/12 m. Find how
the level of water in the vessel depends on
time t and the time it takes all the water to
flow out. T
By hypothesis, S (x) = and s =
11144. Since for water k = 0.6, Eq. (23)
assumes the form
dt
= - 217.152 d
-,fx X.
Integrating this differential equation yields
t=434.304[ll6- VxJ,
which is the sought dependence of the level
of water in the vessel on time t. If we put
x = 0 in the last formula, we find that it
takes approximately 18 minutes for all the
water to flow out of the vessel.
Now a second problem. An ancient water
clock consists of a bowl with a small hole
in the bottom through which water flows
out of the bowl (Figure 4). Such clocks were
used in ancient Greek and Roman courts to
time the lawyers' speeches, so as to avoid
prolonged speeches. We wish to determine
the shape of the water clock that would en-
sure that the water level lowered at a con-
stant rate.
This problem can easily be solved via
Eq. (23). We rewrite this equation in the
form
Vi=-
8
(x) (24)
ks -,f2g dt Ch. 1. Construction of Differential Models 29
.z
Y
Fig. 4
Precisely, if we suppose that the bowl has
the shape of a surface of revolution, in ac-
cordance with the notations used in Figure
4, Eq. (24) yields the following result:
1/ 2% Z --2 a, <25>
ks ]/2g
where a = vx -= da:/dt is the projection on
the :1: axis of the rate of motion of the
waters free surface, which is constant by
hypothesis. Squaring both sides of Eq.
(25), we arrive at the equation
x = cr", (26)
with c = aznz/2gk2s. The latter means that
the sought shape of the water clock is ob-
tained by rotating curve (26) about the 2:
axis.
Fig. 4
Precisely, if we suppose that the bowl has
the shape of a surface of revolution, in ac-
cordance with the notations used in Figure
4, Eq. (24) yields the following result:
V
- :n:rl
x=- a,
ks y2g
(25)
where a = Vx = dx/dt is the projection on
the x axis of the rate of motion of the
water's free surface, which is constant by
hypothesis. Squaring both sides of Eq.
(25), we arrive at the equation
x = cr\ (26)
with c = a
2
:rt
2
/2gk
2
s
2
The latter means that
the sought shape of the water clock is ob-
tained by rotating curve (26) about the x
axis. 30 Differential Equations in Applications
Suppose that a retail chain is selling com-
modities of a certain type, say B, about
which only rz: buyers out of N potential buy-
ers know at time t. Let us also assume that
to speed up sales the chain has placed pro-
motion materials in the local TV and radio
network. All further information about the
commodities is distributed among the buy-
ers via personal contact between them. We
may assume with a high probability that
after the TV and radio network have re-
leased the information about B, the rate of
change in the number of persons knowing
about B is proportional both to the number
of buyers knowing about B and to the num-
ber of buyers not knowing about B.
If we suppose that time is reckoned from
the moment when the promotion materials
and advertisements were released and
N /y persons have learned about the commo-
dities, we arrive at the following differen-
tial equation:
%:kx (N- 2:), (27)
with the initial condition that sc = N /y
at t = O. In this equation lc is a positive
proportionality factor. Integrating, we find
that
% ln + C.
Suppose that a retail chain is selling com-
modities of a certain type, say B, about
which only x buyers out of N potential buy-
ers know at time t. Let us also assume that
to speed up sales the chain has placed pro-
motion rna terials in the local TV and radio
network. All further information about the
commodities is distributed among the buy-
ers via personal contact between them. We
may assume with a high probability that
after the TV and radio network have re-
leased the information about B, the rate of
change in the number of persons knowing
about B is proportional both to the number
of buyers knowing about B and to the num-
ber of buyers not knowing about B.
If we suppose that time is reckoned from
the moment when the promotion materials
and advertisements were released and
N /y persons have learned about the commo-
dities, we arrive at the following dif-feren-
tial equation:
dx
dt=kx(N-x),
(27)
with the initial condition that x = N ly
at t = 0. In this equation k is a positive
proportionality factor. Integrating, we find
that
1 X
Nln N-x =kt+C. Ch. 1. Construction of Differential Models 31
sc
xv
0 t
Fig. 5
Assuming that NC = C1, we arrive at the
equation
;..... Nm
N_x -Ae ,
with A = eC. When solved for sr, this
equation yields
Nm
"=N %.r=_Lvwr (28)
Ae +1 1+1>a *
with P = Cl/A.
In economics, Eq. (28) is commonly
known as the logistic-curve equation.
If we allow for the initial condition, Eq.
(28) becomes
N
35 = --.
1+<v4>e"'*
Figure 5 provides a schematic ofa logis-
tic curve for y = 2. In conclusion we note
t
Fig. 5
Assuming that NC = C
1
, we arrive at the
equation
_x __ =AeNkt
N-x '
with A = e
0
. When solved for x, this
equation yields
N
(28)
with P = 1/A.
In economics, Eq. (28) is commonly
known as the logistic-curve equation.
If we allow for the initial condition, Eq.
(28) becomes
N
X = - - - - - ~ ~
t+(y-f) e-Nkt
Figure 5 provides a schematic of a logis-
tic curve for y = 2. In conclusion we note 32 Differential Equations in Applications
that the problem of dissemination of tech-
nological innovations can also be reduced
to Eq. (27).
As is known, supply and demand constitute
economic categories in a commodity econo-
my, categories that emerge and function on
the market, that is, in the sphere of trade.
The first category represents the commodi-
ties that exist on the market or can be deliv-
ered to it, while the second represents the
demand for commodities on the market.
One of the main laws of a market economy
is the law of supply and demand, which
can be formulated simply by saying that
for each commodity some price must exist
that will cause the supply and the demand
to be just equal. Such a price establishes an
"equilibrium" on the market.
Let us consider the following problem.
Suppose that in the course of a (relatively
long) time interval a farmer sells his produce
(say, apples) on the market, and that he
does this immediately after the apple har-
vest has been taken in and then once each
following week (i.e. with a weeks interval).
The farmers stock of apples being fixed af-
ter he has collected his harvest, the weeks
supply will depend on both the expected
price of apples in the coming week and the
expected change in price in the following
that the problem of dissemination of tech-
nological innovations can also be reduced
to Eq. (27).
As is known, supply and demand constitute
economic categories in a commodity econo-
my, categories that emerge and function on
the market, that is, in the sphere of trade.
The first category represents the commodi-
ties that exist on the market or can be deliv-
ered to it, while the second represents the
demand for commodities on the market.
One of the main laws of a market economy
is the law of supply and demand, which
can be formulated simply by saying that
for each commodity some price must exist
that will cause the supply and the demand
to be just equal. Such a price establishes an
"equilibrium" on the market.
Suppose that in the course of a (relatively
long) time interval a farmer sells his produce
(say, apples) on the market, and that he
does this immediately after the apple har-
vest has been taken in and then once each
following week (i.e. with a week's interval).
The farmer's stock of apples being fixed af-
ter he has collected his harvest, the week's
supply will depend on both the expected
price of apples in the coming week and the
expected change in price in the following Ch. 1. Construction of Differential Models 33
weeks. If it is assumed that in the coming
week the price of apples will fall and in the
following weeks it will grow, then supply
will be restrained provided that the ex-
pected rise in price exceeds storage costs. In
these conditions the greater the expected rise
in price in the following weeks, the lower
the supply of apples on the market. On the
other hand, if the price of apples in the
coming week is high and then a fall is ex-
pected in the following weeks, then the
greater the expected fall in price in the fol-
lowing weeks, the higher the supply of ap-
ples on the market in the coming week.
If we denote the price of apples in the
coming week by p and the time derivative
of price (the tendency of price formation)
by p', both supply and demand are func-
tions of these quantities. As practice has
shown, depending on various factors the
supply and demand of a commodity may be
represented by different functions of price
and tendency of price formation. For exam-
ple, one function is given by a linear de-
pendence expressed mathematically in the
form y = ap' + bp l- c, where a, b, and
c are real numbers (constants). Say, in
our example, the price of apples in the
coming week is taken at 1 rouble for 1 kg,
in t weeks it was p (t) roubles for 1 kg,
and the supply s and demand q are given
by the functions
s=44p }2p1, q==4p-2p|39.
3-0770
weeks. If it is assumed that in the coming
week the price of apples will fall and in the
following weeks it will grow, then supply
will be restrained provided that the ex-
pected rise in price exceeds storage costs. In
these conditions the greater the expected rise
in price in the following weeks, the lower
the supply of apples on the market. On the
other hand, if the price of apples in the
coming week is high and then a fall is ex-
pected in the following weeks, then the
greater the expected fall in price in the fol-
lowing weeks, the higher the supply of ap-
ples on the market in the coming week.
If we denote the price of apples in the
coming week by p and the time derivative
of price (the tendency of price formation)
by p', both supply and demand are func-
tions of these quantities. As practice has
shown, depending on various factors the
supply and demand of a commodity may be
represented by different functions of price
and tendency of price formation. For exam-
ple, one function is given by a linear de-
pendence expressed mathematically in the
form y = ap' + bp + c, where a, b, and
c are real numbers (constants). Say, in
our example, the price of apples in the
coming week is taken at 1 rouble for 1 kg,
in t weeks it was p (t) roubles for 1 kg,
and the supply s and demand q are given
by the functions
s = 44p' + 2p- 1, q = 4p'- 2p + 39.
Then, for the supply and demand to be in
equilibrium, we must require that
44p l-2p1=4p' 2p39.
This condition leads us to the differential
equation
dp __
Integration yields p = Cermt -} 10. If we
allow for the initial condition that p = 1
at t = 0, the equilibrium price is given by
the formula
p = 9e1 + 10. (29)
Thus, if we want the supply and demand to
be in equilibrium all the time, the price
must vary according to formula (29).
1.7 Chemical Reactions
A chemical equation shows how the inter-
action of substances produces a new sub-
stance. Take, for instance, the equation
2H2 {O2 ->2H2O.
It shows how as a result of the interaction
of two molecules of hydrogen and one mo-
lecule of oxygen two molecules of water are
formed.
Then, for the supply and demand to be in
equilibrium, we must require that
44p' + 2p - 1 = 4p' - 2p + 39.
This condition leads us to the differential
equation
dp
p-10 = -10dt.
Integration yields p = Ce-
10
t + 10. If we
allow for the initial condition that p = 1
at t = 0, the equilibrium price is given by
the formula
p = -9e-
10
t + 10. (29)
Thus, if we want the supply and demand to
be in equilibrium all the time, the price
must vary according to formula (29).
1. 7 Chemical Reactions
A chemical equation shows how the inter-
action of substances produces a new sub-
stance. Take, for instance, the equation
2H
2
+ 0
2
-+ 2H
2
0.
It shows how as a result of the interaction
of two molecules of hydrogen and one mo-
lecule of oxygen two molecutes of water are
formed. Ch. 1. Construction of Differential Models 35
Generally a chemical reaction can be writ-
ten in the form
aA i- bB l- cC l- >mM + nN
+ pP +
where A, B, C, are molecules of the in-
teracting substances, M, N, P, . . mole-
cules of the reaction products, and the con-
stants a, b, c, . ., m, n, p, . positive
integers that stand for the number of mole-
cules participating in the reaction.
The rate at which a new substance is
formed in a reaction is called the reaction
rate, and the active mass or concentration
of a reacting substance is given by the num-
ber of moles of this substance per unit vol-
ume.
One of the basic laws of the theory of
chemical reaction rates is the law of mass
action, according to which the rate of a chem-
ical reaction proceeding at a constant tem-
perature is proportional to the product of the
concentrations of the substances taking part
in the reaction at a given moment.
Let us solve the following problem. Two
liquid chemical substances A and B occu-
pying a volume of 10 and 20 liters, respec-
tively, form as a result of a chemical reaction
a new liquid chemical substance C. Assum-
ing that the temperature does not change
in the process of the reaction and that every
two volumes of substance A and one vol-
3* r Ch. 1. Construction of Differential Models 35
Generally a chemical reaction can be writ-
ten in the form
aA + bB + cC +
+pP+
~ m l l f + nN
where A, B, C, are molecules of the in-
teracting substances, llf, N, P, .. mole-
cules of the reaction products, and the con-
stants a, b, c, .. , m, n, p, . positive
integers that stand for the number of mole-
cules participating in the reaction.
The rate at which a new substance is
formed in a reaction is called the reaction
rate, and the active mass or concentration
of a reacting substance is given by the num-
ber of moles of this substance per unit vol-
ume.
One of the basic laws of the theory of
chemical reaction rates is the law of mass
action, according to which the rate of a chem-
ical reaction proceeding at a constant tem-
perature is proportional to the product of the
concentrations of the substances taking part
in the reaction at a given moment.
Let us solve the following problem. Two
liquid chemical substances A and B occu-
pying a volume of 10 and 20 liters, respec-
tively, form as a result of a chemical reaction
a new liquid chemical substance C. Assum-
ing that the temperature does not change
in the process of the reaction and that every
two volumes of substance A and one vol-
3* 36 Differential Equations in Applications
ume of substanceBform three volumes of
substance C, {ind the amount of substance
C at an arbitrary moment of time t if it
takes 20 min to form 6 liters of C.
Let us denote by x the volume (in liters)
of substance C that has formed by the time
t(in hours). By hypothesis, by that time
2.1:/3 liters of substance A and x/3 liters of
substance B will have reacted. This means
that we are left with 10 - 2x/3 liters of A
and 20 x/3 liters of B. Thus, in accord-
ance with the law of mass action, we arrive
at the following differential equation
dx 2::: rc
w:K("T) (Z0-?)
dx
where k is the proportionality factor (lc =
2K/9). We must also bear in mind that since
initially (t = 0) there was no substance
C, we may assume that x = 0 at t = 0. As
for the moment t = 1/3, we have sc = 6.
Thus, the solution of the initial problem
has been reduced to the solution of a so-
called boundary value problem:
ft;-:k (15-x) (60-:c), x (0): 0,
ume of substance B form three volumes of
substance C, find the amount of substance
C at an arbitrary moment of time t if it
takes 20 min to form 6 liters of C.
Let us denote by x the volume (in liters)
of substance C that has formed by the time
t (in hours). By hypothesis, by that time
2x/3 liters of substance A and x/3 liters of
substance B will have reacted. This means
that we are left with 10 - 2x/3 liters of A
and 20 - x/3 liters of B. Thus, in accord-
ance with the law of mass action, we arrive
at the following differential equation
~ ; = K ( 10-
2
; ) ( 20- ~ )
dx
5
dt = k (1 -x) (60- x),
where k is the proportionality factor (k =
2K/9). We must also bear in mind that since
initially (t = 0) there was no substance
C, we may assume that x = 0 at t = 0. As
for the moment t = 1/3, we have x = 6.
has been reduced to the solution of a so-
called boundary value problem:
dx
dt=k(15-x)(60-x), x(O)=O,
X (1/3) = 6. Ch. 1. Construction of Differential Models 37
To solve it, we iirst integrate the differen-
tial equation and allow for the initial con-
dition x(0) :0. As a result we arrive
at the relationship
_?_g_;_;_ Z 445m_
Since :1: = 6 at t = 1/3, substituting these
values into the relationship yields emh =
3/2. Hence,
_T4(15k)3t:;4(3/2)3t,
that is,
$:15 i-(2/3)***
i(i/@(2/3)3 l
This gives the amount of substance C
that has been formed in the reaction by
time t.
A remark is in order. From practical
considerations it is clear that only a finite
volume of substance C can be formed when
10 liters of A interact chemically with 20
liters of B. However, a formal study of
the above relationship between x and t
shows that for a finite t, namely at (2/3);** =
4, the variable .1: becomes iniinite. But this
fact does not contradict practical consid-
erations because it is realized only for a
negative value of t, while the chemical reac-
tion is considered only for t nonnegative.
To solve it, we first integrate the differen-
tial equation and allow for the initial con-
dition x (0) = 0. As a result we arrive
at the relationship
60-x -- 4e45kt
15-x - '
Since x = 6 at t = 113, substituting these
values into the relationship yields e
15
k =
3/2. Hence,
60-x
15
_x =4 (el5h)31 = 4 (3/2)31,
that is,
X = 15 __ 1 -----'-(2-'--/3-'-)3_1 ----,--
1- (1/4) (2/3)
3
1
This gives the amount of substance C
that has been formed in the reaction by
time t.
A remark is in order. From practical
considerations it is clear that only a finite
volume of substance C can be formed when
10 liters of A interact chemically with 20
liters of B. However, a formal study of
the above relationship between x and t
shows that for a finite t, namely at (2/3)3t =
4, the variable x becomes infinite. But this
fact does not contradict practical consid-
erations because it is realized only for a
negative value oft, while the chemical reac-
tion is considered only for t nonnegative. 38 Differential Equations in Applications
1.8 Differential Models in Ecology
Ecology studies the interaction of man and,
in general, living organisms with the envi-
ronment. The basic object in ecology is the
evolution of populations. Below we describe
differential models of populations that deal
with their reproduction or extinction and
with the coexistence of various species of
animals in the predator vs. prey situation.*
Let x (t) be the number of individuals
in a population at time t. If A is the number
of individuals in the population that are
born per unit time and B the number of
individuals that die off per unit time, then
there is reason to say that the rate at which
x varies in time is given by the formula
dx
(F..AB. (30)
The problem consists in finding the depen-
dence of A and B on :1:. The simplest situa-
tion is the one in which
A = ax, B = bx, (31)
where a and b are the coefficients of births
and deaths of individuals per unit time,
respectively. Allowing for (31), we can re-
write the differential equation (30) as
dx
Ti-T -. (cz-- b) as. (32)
* For more detail see J.D. Murray, "Some simple
mathematical models in ecology," Math. Spectrum
16, No. 2: 48-54 (1983/84).
t.8 Differential Models in Ecology
Ecology studies the interaction of man and,
in general, living organisms with the envi-
ronment. The basic object in ecology is the
evolution of populations. Below we describe
differential models of populations that deal
with their reproduction or extinction and
with the coexistence of various species of
animals in the predator vs. prey situation.*
Let x (t) be the number of individuals
in a population at time t. If A is the number
of individuals in the population that are
born per unit time and B the number of
individuals that die off per unit time, then
there is reason to say that the rate at which
x varies in time is given by the formula
(30)
The problem consists in finding the depen-
dence of A and B on x. The simplest situa-
tion is the one in which
A = ax, B = bx, (31)
where a and b are the coefficients of births
and deaths of individuals per unit time,
respectively. Allowing for (31), we can re-
write the differential equation (30) as
dx
dt=(a-b)x. (32)
* For more detail see J.D. Murray, "Some simple
mathematical models in ecology," Math. Spectrum
16, No. 2: 48-54 (1983/84). Ch. 1. Construction of Differential Models 30
Assuming that at t = to the number of in-
dividuals in the population is sc = :1:,,
and solving Eq. (32), we get
:1: (t) = .2:0 exp [(a b) (t t,,)l.
This implies that if a > b, then the number
of individuals sc tends to infinity as t> oo,
but if a<b, then a:>O as t-+00 and
the population becomes extinct.
Although the above model is a simplified
one, it still reflects the actual situation in
some cases. However, practically all mo-
dels describing real phenomena and proces-
ses are nonlinear, and instead of the differ-
ential equation (32) we are forced to con-
sider an equation of the type
dx __
jg--f (iv),
where f (.2:) is a nonlinear function, say the
equation
%:f(x) =ax-bx2,
where
a > 0, b > O.
Assuming that x = xo at t = to and solving
the last equation, we get
; : _ 33
ll =0+</bx0>exp1<0>1 ( )
We see that as t > oo the number of indiv-
iduals in the population, x (t), tends to
Ch. 1. Construction of Differential Models 3!l
Assuming that at t = t
0
the number of in-
dividuals in the population is x = x
0
and solving Eq. (32), we get
x (t) = x
0
exp [(a- b) (t- t
0
)1.
This implies that if a > b, then the number
of individuals x tends to infmity as t--+ oo,
but if a< b, then x--+ 0 as t--+ oo and
the population becomes extinct.
Although the above model is a simplified
one, it still reflects the actual situation in
some cases. However, practically all mo-
dels describing real phenomena and proces-
ses are nonlinear, and instead of the differ-
ential equation (32) we are forced to con-
sider an equation of the type
dx
dt=f(x),
where f (x) is a nonlinear function, say the
equation
dx
"ii't= f (x) =ax- bx
2
,
where
a> 0, b > 0.
Assuming that x = x
0
at t = t
0
and solving
the last equation, we get
X (t) = x
0
a/b . (
33
)
x
0
+ (afb- x
0
) exp [-a (t- t
0
)}
We see that as t--+ oo the number of indiv-
iduals in the population, x (t), tends to 40 Differential Equations in Applications
cz
J? 5>.120
1-zz--1***-@""'*
0 b <.Z0 6
Fig. 6
a/b. Two cases are possible here: a/b > xo
and a/b < xo. The difference between the
two is clearly seen in Figure 6. Note that
formula (33) describes, for one thing, the
populations of fruit pests and some types
of bacteria.
If we consider several coexisting species,
say, big and small {ish, where the small fish
serve as prey for the big, then by setting up
differential equations for each species we
arrive at a system of equations
%?=f, (1:,, x,,), i=1, 2, n.
Let us study in greater detail the two-spe-
cies predator vs. prey model, first intro-
duced by the Italian mathematician Vito
Volterra (1860-1940) to explain the oscilla-
tions in catch volumes in the Adriatic Sea,
which have the same period but differ in
phase.
a
7i> :Co
- - ~ - - - - - - - - - - - - - - -
Fig. 6
alb. Two cases are possible here: alb > x
0
and alb < x
0
The difference between the
two is clearly seen in Figure 6. Note that
formula (33) describes, for one thing, the
populations of fruit pests and some types
of bacteria.
If we consider several coexisting species,
say, big and small fish, where the small fish
serve as prey for the big, then by setting up
differential equations for each species we
arrive at a system of equations
dxi (
(ft = /; x
1
, Xn), i = 1, 2, n.
Let us study in greater detail the two-spe-
cies predator vs. prey model, fust intro-
duced by the Italian mathematician Vito
Volterra (1860-1940) to explain the oscilla-
tions in catch volumes in the Adriatic Sea,
which have the same period but differ in
phase. Ch. 1. Construction of Differential Models 41
Let ac be the number of big fish (predators)
that feed on small fish (prey), whose num-
ber we denote by y. Then the number of
predator fish will grow as long as they have
sufficient food, that is, prey fish. Finally, a
situation will emerge in which there will
not be enough food and the number of big
Hsh will diminish. As a result, starting from
a certain moment the number of small {ish
begins to increase. This, in turn, assists a
new growth in the number of the big spe-
cies, and the cycle is repeated. The model
constructed by Volterra has the form
da:
Y: a:c|b:cy, (34)
dy - as
TH-.-cxdxy, ( )
where ez, b, c, and d are positive constants.
In Eq. (34) for the big iish the term bxy
reflects the dependence of the increase in
the number of big fish on the number of
small {ish, while in Eq. (35) the term
d;z:y expresses the decrease in the number
of small fish as a function of the number of
big {ish.
To make the study of these two equations
more convenient we introduce dimension-
less variables:
d b
u(t)=-Fx, v(:)=&y, 1:=ct, ot=a/c.
Let x be the number of big fish (predators)
that feed on small fish (prey), whose num-
ber we denote by y. Then the number of
predator fish will grow as long as they have
sufficient food, that is, prey fish. Finally, a
situation will emerge in which there will
not be enough food and the number of big
fish will diminish. As a result, starting from
a certain moment the number of small fish
begins to increase. This, in turn, assists a
new growth in the number of the big spe-
cies, and the cycle is repeated. The model
constructed by Volterra has the form
dx
dt= -ax+bxy,
(34)
dy d
dt=cx- xy,
(35)
where a, b, c, and d are positive constants.
In Eq. (34) for the big fish the term bxy
reflects the dependence of the increase in
the number of big fish on the number of
small fish, while in Eq. (35) the term
-dxy expresses the decrease in the number
of small fish as a function of the number of
big fish.
To make the study of these two equations
more convenient we introduce dimension-
less variables:
d b
u(T)=-x, v()=-y, 't=ct, a.=alc.
c a 42 Differential Equations in Applications
As a result the differential equations (34)
and (35) assume the form
u,' : ow. (v 1), v' : v (1 -- u), (36)
with oc positive and the prime standing for
differentiation with respect to 1.
Let us assume that at time rc = to the
number of individuals of both species is
known, that is,
u (T0) = wo, v (10) = va (37)
Note that we are interested only in positive
solutions. Let us establish the relation-
ship between u and v. To this end we divide
the first equation in system (36) by the sec-
ond and integrate the resulting differential
equation. We get
ow+u-lnvu =ow0i-un-lnvgu0
EH,
where H is a constant determined by the
initial conditions (37) and parameter oc.
Figure 7 depicts the dependence of u
on v for different values of H. We see that
the (u, v)plane contains only closed curves.
Let us now assume that the initial val-
ues uo and vt, are specified by point A
on the trajectory that corresponds to the
value H = H3. Since un > 1 and vo < 1,
the first equation in (36) shows that at first
variable u, decreases. The same is true of
variable v. Then, as u approaches a value
close to unity, v' vanishes, which is fol-
lowed by a prolonged period of time 1: in the
As a result the differential equations (34)
and (35) assume the form
u' = au (v - 1), v' = v (1 - u), (36)
with a positive and the prime standing for
differentiation with respect to 't'.
Let us assume that at time 't' = To the
number of individuals of both species is
known, that is,
u ('t'
0
) = U
0
, v (T
0
) = V
0
(37)
Note that we are interested only in positive
solutions. Let us establish the relation-
ship between u and v. To this end we divide
the first equation in system (36) by the sec-
ond and integrate the resulting differential
equation. We get
av + u - ln vau = av
0
+ u
0
- ln l{'u
0
=H,
where H is a constant determined by the
initial conditions (37) and parameter a.
Figure 7 depicts the dependence of u
on v for different values of H. We see that
the (u, v)-plane contains only closed curves.
Let us now assume that the initial val-
ues u
0
and v
0
are specified by point A
on the trajectory that corresponds to the
value H = Ha. Since u
0
> 1 and v
0
< 1,
the first equation in (36) shows that at first
variable u decreases. The same is true of
variable v. Then, as u approaches a value
close to unity, v' vanishes, which is fol-
lowed by a prolonged period of time 't' in the Ch. 1. Construction of Differential Models 43
u
A
. H5
Hz
{D"'
1
I
0 1 v
Fig. 7
u v
v
00 M
do
0 z
Fig. 8
course of which variable v increases. When
v becomes equal to unity, uf vanishes and
variable u begins to increase. Thus, both
u and v traverse a closed trajectory. This
means that the solutions are functions peri-
odic in time. The variable u does not achieve
u
v
Fig. 7
u v
0
v
Fig. 8
course of which variable v increases. When
v becomes equal to unity, u' vanishes and
variable u begins to increase. Thus, both
u and v traverse a closed trajectory. This
means that the solutions are functions peri-
odic in time. The variable u does not achieve 44 Differential Equations in Applications
its maximum when v does, that is, oscil-
lations in the populations occur in differ-
ent phases. A typical graph of u and v as
functions of time is shown in Figure 8 (for
the case where v0>1 and u0<1).
In conclusion we note that the study of
communities that interact in a more com-
plex manner provides results that are more
interesting from the practical standpoint
than those obtained above. For example, if
two populations fight for the same source of
food (the third population), it can be demon-
strated that one of them will become
extinct. Clearly, if this is the third popula-
tion (the source of food), then the other two
will also become extinct.
1.9 A Problem from the
Mathematical Theory of Epidemics
Let us consider a differential model encoun-
tered in the theory of epidemics. Suppose
that a population consisting of N individ-
uals is split into three groups. The first in-
clude individuals that are susceptible to a
certain disease but are healthy. We denote
the number of such individuals at time
t by S (t). The second group incorporates
individuals that are infected, that is, they
are ill and serve as a source of infection.
We denote the number of such idividuals
in the population at time t by I (t). Final-
ly, the third group consists of individuals
its maximum when v does, that is, oscil-
lations in the populations occur in differ-
ent phases. A typical graph of u and v as
functions of time is shown in Figure 8 (for
the case where v
0
> 1 and u
0
< 1).
In conclusion we note that the study of
communities that interact in a more com-
plex manner provides results that are more
interesting from the practical standpoint
than those obtained above. For example, if
two populations fight for the same source of
food (the third population}, it can be demon-
strated that one of them will become
extinct. Clearly, if this is the third popula-
tion (the source of food), then the other two
will also become extinct.
1.9 A Problem from the
Mathematical Theory of Epidemics
Let us consider a differential model encoun-
tered in the theory of epidemics. Suppose
that a population consisting of N individ-
uals is split into three groups. The first in-
clude individuals that are susceptible to a
certain disease but are healthy. We denote
the number of such individuals at time
t by S (t). The second group incorporates
individuals that are infected, that is, they
are ill and serve as a source of infection.
We denote the number of such idividuals
in the population at time t by I (t). Final-
ly, the third group consists of individuals Ch. 1. Construction of Differential Models 45
who are healthy and immune to the di-
sease. We denote the number of such indivi-
duals at time t by R (t). Thus,
S (t) l I (t) l R (t) = N. (38)
Let us also assume that when the number
of infected individuals exceeds a certain
fixed number I *, the rate of change in the
number of individuals susceptible to the
disease is proportional to the number of
such individuals. As for the rate of change
in the number of infected individuals that
eventually recover, we assume that it is
proportional to the number of infected in-
dividuals. Clearly, these assumptions sim-
plify matters and in a number of cases they
reflect the real situation. Because of the
first assumption, we suppose that when the
number of infected individuals I (t) is
greater than I *, they can infect the individ-
uals susceptible to the disease. This means
that we have allowed for the fact that the
infected individuals have been isolated for
a certain time interval (as a result of quar-
antine or because they have been far from
individuals susceptible to the disease).
We, therefore, arrive at the following dif-
ferential equation
-ocS if I t > I*,
%:l 0 `fI()" * (39)
I I .
Now, since each individual susceptible
to the disease eventually falls ill and be-
who are healthy and immune to the di-
sease. We denote the number of such indivi-
duals at time t by R (t). Thus,
S (t) + I (t) + R (t) = N. ( ~ ~ 8 )
Let us also assume that when the number
of infected individuals exceeds a certain
fixed number I*, the rate of change in the
number of individuals susceptible to the
disease is proportional to the number of
such individuals. As for the rate of change
in the number of infected individuals that
eventually recover, we assume that it is
proportional to the number of infected in-
dividuals. Clearly, these assumptions sim-
plify matters and in a number of cases they
reflect the real situation. Because of the
first assumption, we suppose that when the
number of infected individuals I (t) is
greater than I*, they can infect the individ-
uals susceptible to the disease. This means
that we have allowed for the fact that the
infected individuals have been isolated for
a certain time interval (as a result of quar-
antine or because they have been far from
individuals susceptible to the disease).
We, therefore, arrive at the following dif-
ferential equation
dS --f -a..S if I(t)>I*,
lit_ ~ ~ :[. ' . ' ( ~ , ~ ~ '.* .
(39)
Now, since each individual susceptible
to the disease eventually falls ill and be-46 Differential Equations in Applications
comes a source of infection, the rate of
change in the number of infected individuals
is the difference, per unit time, between the
newly infected individuals and those that
are getting well. Hence,
dj {ocSB1 if 1(t) > 1*, 40
dt --BI if 1(t) { 1*. ( )
We will call the proportionality factors
ot and B the coefyicients of illness and recovery.
Finally, the rate of change of the number
of recovering individuals is given by the
equation dR/dt -= BI.
For the solutions of the respective equa-
tions to be unique, we must hx the initial
conditions. For the sake of simplicity we
assume that at time t = O no individuals
in the population are immune to the disease,
that is, R (O) = O, and that initially the
number of infected individuals was I (0).
Next we assume that the illness and recov-
ery coefficients are equal, that is, on = B
(the reader is advised to study the case
where ot =;& B). Hence, we find ourselves in
a situation in which we must consider two
cases.
Case 1. 1 (O)< 1*. With the passage of
time the individuals in the population will
not become infected because in this case
dS/dt = O, and, hence, in accordance with
Eq. (38) and the condition R (O) == O,
we have an equation valid for all values
comes a source of infection, the rate of
change in the number of infected individuals
is the difference, per unit time, between the
newly infected individuals and those that
are getting well. Hence,
dl = { aS- if I (t) > I*, (
4
0)
dt
We will call the proportionality factors
a and the coefficients of illness and recovery.
Finally, the rate of change of the number
of recovering individuals is given by the
equation dR/dt =
For the solutions of the respective equa-
tions to be unique, we must fix the initial
conditions. For the sake of simplicity we
assume that at time t = 0 no individuals
in the population are immune to the disease,
that is, R (0) = 0, and that initially the
number of infected individuals was I (0).
Next we assume that the illness and recov-
ery coefficients are equal, that is, a =
(the reader is advised to study the case
where a =F Hence, we find ourselves in
a situation in which we must consider two
cases.
Case 1. I (0)::;;;; I*. With the passage of
time the individuals in the population will
not become infected because in this case
dS/dt = 0, and, hence, in accordance with
Eq. (38) and the condition R (0) = 0,
we have an equation valid for all values Ch. 1. Construction of Differential Models 47
N . ---------- -
S()
If
--"""""" *"' """'/?()
IO5)
0 it
Fig. 9
of tz
S(t)==S(O)=NI(0).
The case considered here corresponds to the
situation when a fairly large number of in-
fected individuals are placed in quaran-
tine. Equation. (40) then leads us to the fol-
dI
7]-;--- otI.
This means that I (t) = I (0) e* and,
hence,
R(t)=NS(t)I(t)
= I (O) [1 e*].
Figure 9 provides diagrams that illustrate
the changes with the passage of time in the
Ch. 1. Construction of Differential Models 4 7
N ----------------------
1------------S(t)
Fig. 9
of t:
S (t) = S (0) = N - I (0).
The case considered here corresponds to the
situation when a fairly large number of in-
fected individuals are placed in quaran-
tine. Equation (40) then leads us to the fol-
dl
dt=-a.I.
This means that I (t) = I (0) e-at and,
hence,
R (t) = N - S (t) - I (t)
= I (0) [1 - e-at],
Figure 9 provides diagrams that illustrate
the changes with the passage of time in the 48 Differential Equations in Applications
number of individuals in each of the three
groups.
Case 2. I (0) > I*. In this case there must
exist a time interval OQ t < T in which
I (t) > I* for all values of t, since by its
very meaning the function I = I (t) is
continuous. This implies that for all ts
belonging to the interval [O, T] the disease
spreads to the individuals susceptible to it.
Equation (39), therefore, implies that
S (t) = S (0)e'
for OQ t < T. Substituting this into Eq.
(40), we arrive at the differential equation
ij;--}-0tI =otS (O) e. (41)
If we now multiply both sides of Eq. (41)
by em, we get the equation
. (lem) = as (0).
Hence, Ie = otS (0) t i- C and, there-
fore, the set of all solutions to Eq. (41) is
given by the formula
I (t) = Ce"* l cxS (0) te. (42)
Assuming that t = O, we get C = I (0) and,
hence, Eq. (42) takes the form
I (t) = [I (0) + 0tS (O) tl e'", (43)
with Og t < T.
48 Differential Equation11 in Applications
number of individuals in each of the three
groups.
Case 2. I (0) > I*. In this case there must
exist a time interval 0 t < 1' in which
I (t) > I* for all values of t, since by its
very meaning the function I = I (t) is
continuous. This implies that for all t's
belonging to the interval [0, T] the disease
spreads to the individuals susceptible to it.
Equation (39), therefore, implies that
S (t) = S (O)e-at
for 0 ~ t < T. Substituting this into Eq.
(40}, we arrive at the differential equation
(41)
If we now multiply both sides of Eq. (41)
by ea.t, we get the equation
d
Cit (feat) =aS (0).
Hence, [eat = aS (0) t + C and, there-
fore, the set of all solutions to Eq. (41) is
I (t) = Ce-at + aS (0) te-at. (42)
Assuming that t = 0, we get C = I (0) and,
hence, Eq. (42) takes the form
I (t) = [/ (0) + aS (0) t) e-at, (43)
with 0 ~ t < T. Ch. 1. Construction of Differential Models 49
We devote our further investigations to
the problem of finding the specific value of
T and the moment of time tmax at which the
number of infected individuals proves
maximal.
The answer to the first question is impor-
tant because starting from T the individuals
susceptible to the disease cease to become
infected. If we turn to Eq. (43), the afore-
said implies that at t= T its right-hand
side assumes the value 1*, that is,
I* = [I (O) -1- ocS (O) TleT. (44)
But
S (T) = 1imS (t) =S (oo)
f->00
is the number of individuals that are suscep-
tible to the disease and yet avoid falling
ill; for such individuals the following chain
of equalities holds true:
S (T) = S (oo) = S (0) eT.
Hence,
__ 1 S (0)
Thus, if we can point to a definite value
of S (oo), we can use Eq. (45) to predict the
time when the epidemic will stop. Substitut-
ing T given by (45) into Eq. (44), we arrive
at the equation
* = S (0) S ()
I [I (O)|-S(O) ln Sho) il Sw) ,
4-0770
We devote our further investigations to
the problem of finding the specific value of
T and the moment of time tmax at which the
number of infected individuals proves
maximal.
The answer to the first question is impor-
tant because starting from T the individuals
susceptible to the disease cease to become
infected. If we turn to Eq. (43), the afore-
said implies that at t = T its right-han.d
side assumes the value I*, that is,
I* = [I (0) + aS (0) T]e-aT,
But
S (T) = lim8 (t) =8 (oo)
f-+00
(44)
is the number of individuals that are suscep-
tible to the disease and yet avoid fallin.g
ill; for such individuals the following chain
of equalities holds true:
8 (T) = 8 (oo) = 8 (0) e-aT,
Hence,
T- ..!..1 s (0)
-a nS(oo)'
(45)
Thus, if we can point to a definite value
of 8 (oo), we can use Eq. (45) to predict the
time when the epidemic will stop. Substitut-
ing T given by (45) into Eq. (44), we arrive
at the equation
I*= [I (0) +8 (0) ln % / ~ ) ] ~ \ ~ l ,
or
.iL.. = i& ifi
S(0) $(0) +1H S(00)
which can be rewritten in the form
1* __ I (0)
S(cO) 1nS(0o) --S(0) -}-1nS (0). (46)
Since I * and all the terms on the right-hand
side of (46) are known, we can use this equa-
tion to determine S (oo).
to Eq. (43). In accordance with the posed
question, this equation yields
ig-=(as (0) - a1 (0) - azs (0) q e< :0.
The moment in time at which I attains its
maximal value is given by the following for-
mula:
... _i_ _ I (0)
tmu = li S0)
If we now substitute this value into Eq. (43),
we get
Im..x= S(0)X1>{li -- (0)/S (0>l}
= S (tmax)
This relationship shows, for one, that at
time tmax the number of individuals sus-
ceptible to the disease coincides with the
number of infected individuals.
or
I* I (0) S (0)
S(oo) = S(O) +In S(oo)'
which can be rewritten in the form
I* I (0)
S(oo) +InS(oo)= S(O) +InS(O). (46)
Since I* and all the terms on the right-hand
side of (46) are known, we can use this equa-
tion to determine 8 ( oo ).
to Eq. (43). In accordance with the posed
question, this equation yields
~ =[aS (0)-ai (0) -a
2
8 (0) t] e-at= 0.
The moment in time at which I attains its
maximal value is given by the following for-
mula:
tmax = ! [ 1 - ~ ~ ~ ~ J.
If we now substitute this value into Eq. (43),
we get
I max= 8 (0) exp {-[1- I (0)/8 (0)1}
=8 (tmax)
This relationship shows, for one, that at
time tmax the number of individuals sus-
ceptible to the disease coincides with the
number of infected individuals. Ch. 1. Construction of Differential Models 51
N **""""****""' **** """""""
' " ''`"""'``'' R(t}
[ __-ml _ mv- sm
I
0 Ht)
tH7U.Z' T Z
Fig. 10
But if t > T, the individuals susceptible
to the disease cease to be infected and
I (t) =I*e*(*T).
Figure 10 gives a rough sketch of the dia-
grams that reflect the changes with the pas-
sage of time of the number of individuals
in each of the three groups considered.
1.10 The Pursuit Curve
Let us examine an example in which differ-
ential equations are used to choose a cor
rect strategy in pursuit problems.
Qi

J(t)
T t
Fig. 10
But if t > T, the individuals susceptible
to the disease cease to be infected and
'}, Ut.),= __
('
Figure 10 gives a rough sketch of the dia-
grams that reflect the changes with the pas-
sage of time of the number of individuals
in each of the three groups considered.
f .10 The Pursuit Curve
Let us examine an example in which differ-
ential equations are used to choose a cor-
rect strategy in pursuit problems. 52 Differential Equat.ions in Applications
d8 /I x
fi \\
\\
\
x
dr
r if
0
Fig. 11
We assume that a destroyer is in pursuit
of a submarine in a dense fog. At a certain
moment in time the fog lifts and the sub-
marine is spotted floating on the surface
three miles from the destroyer. The destroy-
ers speed is twice the submarines speed.
We wish to {ind the trajectory (the pursuit
curve) that the destroyer must follow to
pass exactly over the submarine if the latter
immediately dives after being detected
and proceeds at top speed and in a straight
line in an unknown direction.
To solve the problem we first introduce
polar coordinates, r and 9, in such a manner
that the pole O is located at the point at
which the submarine was located when dis-
covered and that the point at which the
destroyer was located when the submarine
was discovered lies on the polar axis r
(Figure 11). Further reasoning is based on
Fig. 11
We assume that a destroyer is in pursuit
of a submarine in a dense fog. At a certain
moment in time the fog lifts and the sub-
marine is spotted floating on the surface
three miles from the destroyer. The destroy-
er's speed is twice the submarine's speed.
We wish to find the trajectory (the pursuit
curve) that the destroyer must follow to
pass exactly over the submarine if the latter
immediately dives after being detected
and proceeds at top speed and in a straight
line in an unknown direction.
To solve the problem we first introduce
polar coordinates, r and e, in such a manner
that the pole 0 is located at the point at
which the submarine was located when dis-
covered and that the point at which the
destroyer was located when the submarine
was discovered lies on the polar axis r
(Figure 11). Further reasoning is based on Ch. 1. Construction of Differential Models 53
the following considerations. First, the de-
stroyer must take up a position in which it
will be at the same distance from pole O
as the submarine. Then it must move about
O in such a way that both moving objects
remain at the same distance from point O all
the time. Only in this case will the destroyer
eventually pass over the submarine when
circling pole O. The aforesaid implies that
the destroyer must first go straight to point
O until it finds itself at the same distance
x from O as the submarine.
Obviously, distance x can be found either
from the equation
27 _ 3--:1:
72-- 2v
or from the equation
x _ 3-]-::
T `"` -5-
where v is the submarines speed a11d 20
the destr0yers speed. Solving these equa-
tions, we find that this distance is either one
mile or three miles.
Now, if the two still have not met, the
destroyer must circle pole O (clockwise or
counterclockwise) and head away from the
pole at a speed equal to that of the subma-
rine, v. Let us decompose the destroyers ve-
locity vector (of length 2v) into two compo-
nents, the radial component U,. and the tan-
gential component vt (Figure 11).
Ch. t. Construction of Differential Models 53
the following considerations. First, the de-
stroyer must take up a position in which it
will be at the same distance from pole 0
as the submarine. Then it must move about
0 in such a way that both moving objects
remain at the same distance from point 0 all
the time. Only in this case will the destroyer
eventually pass over the submarine when
circling pole 0. The aforesaid implies that
the destroyer must first go straight to point
0 until it finds itself at the same distance
x from 0 as the submarine.
Obviously, distance x can be found either
from the equation
:e 3-z
-v="""2rJ
or from the equation
:e 3+z
v = ~
where v is the submarine's speed and 2v
the destroyer's speed. Solving these equa-
tions, we find that this distance is either one
mile or three miles.
Now, if the two still have not met, the
destroyer must circle pole 0 (clockwise or
counterclockwise) and head away from the
pole at a speed equal to that of the subma-
rine, v. Let us decompose the destroyer's ve-
locity vector (of length 2v) into two compo-
nents, the radial component Vr and the tan-
gential component Vt (Figure 11). 54 Diflerential Equations in Applications
The radial component is the speed at
which the destroyer moves away from pole
O, that is
,, - 2;
I. _
while the tangential component is the li-
near speed at which the destroyer circles
the pole. The latter component, as is known,
is equal to the product of the angular ve
locity dt)/dt and radius r, that is,
d0
"t=*a;
But since v, must be equal to v, we have
vt:]/(2v)2-02:]/-3:0.
is reduced to the solution of a system of two
differential equations,
dr d9
-52- ..v, r W- V 3 v,
which, in turn, can be reduced to a single
equation, dr/r = d9/ V3, by excluding the
variable t.
Solving the last differential equation, we
find that
r=Ce/Vg,
with C an arbitrary constant. o
The radial component is the speed at
which the destroyer moves away from pole
0, that is
dr
Vr = dt'
while the tangential component is the li-
near speed at which the destroyer circles
the pole. The latter component, as is known,
is equal to the product of the angular ve-
locity d9/dt and radius r, that is,
d6
Vt=rdt.
But since Vr must be equal to v, we have
Vt= V(2v)
2
-V
2
= y'3v.
is reduced to the solution of a system of two
differential equations,
dr de
113
dt=v, rdt= v v,
which, in turn, can be reduced to a single
equation, drlr = del 'Jf3, by excluding the
variable t.
Solving the last differential equation, we
find that
r = Ce6tl'"3,
with C an arbitrary constant. Ch. 1. Construction of Differential Models 55
If we now allow for the fact that destroyer
beginsits motion about pole O starting from
the polar axis r at a distance of sz: miles away
fromO, thatis,r = 1 at 0 = 0 and r = 3 at
9 =:n;, we conclude that in the first case
ii C = 1 and in the second C = 3e/V; Thus,
to fulfill its mission, the destroyer must
move two or six miles along a straight line
toward the point where the submarine was
discovered and then move in the spiral
r = e9/VT or the spiral r = 3e(9+)/V?
1.11 Combat Models
During the First World War the British
engineer and mathematician Frederick
W. Lanchester (1868-1946) constructed sev-
eral mathematical models of air battles.
Later these models were generalized so as to
describe battles involving regular troops or
guerilla forces or the two simultaneously.
Below we consider these three models.
Suppose that two opposing forces, nv
and y, are in combat. We denote the num-
ber of personnel at time t measured in days
starting from the first day of combat opera-
tions by sz: (t) and y (t), respectively. It is
the number of personnel that will play a
decisive part in the construction of these
models, since it is difficult, practically
speaking, to specify the criteria that would
allow taking into account, when comparing
the opposing sides, not only the number of
If we now allow for the fact that destroyer
hegins its motion about pole 0 starting from
the polar axis r at a distance of x miles away
from 0, that is, r = 1 at e = 0 and r = 3 at
9 = -n:, we conclude that in the first case
C = 1 and in the second C = Thus,
to fulfill its mission, the destroyer must
move two or six miles along a straight line
toward the point where the submarine was
discovered and then move in the spiral
r = e9tV"3 or the spiral r =
1. 1 1 Combat Models
During the First World War the British
engineer and mathematician Frederick
W. Lanchester (1868-1946) constructed sev-
eral mathematical models of air battles.
Later these models were generalized so as to
describe battles involving regular troops or
guerilla forces or the two simultaneously.
Below we consider these three models.
Suppose that two opposing forces, x
andy, are in combat. We denote the num-
ber of personnel at time t measured in days
starting from the first day of combat opera-
tions by x (t) and y (t), respectively. It is
the number of personnel that will play a
decisive part in the construction of these
models, since it is difficult, practically
speaking, to specify the criteria that would
allow taking into account, when comparing
the opposing sides, not only the number of 56 Differential Equations in Applications
personnel but also combat readiness, level
of military equipment, preparedness and
experience of the officers, morale, and a
great many other factors.
We also assume that x(t) and y(t) change
continuously and, more than that, are
differentiable as functions of time. Of course,
these assumptions simplify the real sit-
uation because as (t) and y (t) are integers.
But at the same time it is clear that if the
number of personnel on each side is great,
an increase by one or two persons will have
an infinitesimal effect from the practical
standpoint if compared to the effect pro-
duced by the entire force. Therefore, we may
assume that during small time intervals
the change in the number of personnel is
also small (and does not constitute an inte-
gral number). The above reasoning is, of
course, insufficient to specify concrete for-
mulas for ;z: (t) and y (t) as functions of t, but
we can already point to a number of factors
that enable describing the rate of change
in the number of personnel on the two
sides. Precisely, we denote by OLR the rate
at which side x suffers losses from disease
and other factors not related directly to
combat operations, and by CLR the rate
at which it suffers losses directly in combat
operations involving side y. Finally by RR
we denote the rate at which reinforcements
are supplied to side x. It is then clear that
the total rate of change of x (t) is given by
personnel but also combat readiness, level
of military equipment, preparedness and
experience of the officers, morale, and a
great many other factors.
We also assume that x(t) and y{t)change
continuously and, more than that, are
differentiable as functions of time. Of course,
these assumptions simplify the real sit-
uation because x {t) and y (t) are integers.
But at the same time it is clear that if the
number of personnel on each side is great,
an increase by one or two persons will have
an infinitesimal effect from the practical
standpoint if compared to the effect pro-
duced by the entire force. Therefore, we may
assume that during small time intervals
the change in the number of personnel is
also small (and does not constitute an inte-
gral number). The above reasoning is, of
course, insufficient to specify concrete for-
mulas for x (t) and y (t) as functions oft, but
we can already point to a number of factors
that enable describing the rate of change
in the number of personnel on the two
sides. Precisely, we denote byOLR the rate
at which side x suffers losses from disease
and other factors not related directly to
combat operations, and by CLR the rate
at which it suffers losses directly in combat
operations involving side y. Finally by RR
we denote the rate at which reinforcements
are supplied to side x. It is then clear that
the total rate of change of x (t) is given by Ch. 1. Construction of Differential Models 57
the equation
.<l=1l = - (OLR +01,12) + RR. (47)
A similar equation holds true for y (t).
The problem now is to {ind the appropriate
formulas for the quantities OLR, CLR,
and RR and then examine the resulting
differential equations. The results should
indicate the probable winner.
We use the following notation: a, b,
c, d, g, and h are nonnegative constants
characterizing the effect of various factors
on personnel losses on both sides (x and y),
P (t) and O (t) are terms that allow for the
possibility of reinforcements being supplied
to :2: and y in the course of one day, and
xo and yo are the personnel in :1: and y prior
to combat. We are now ready to set up the
three combat models suggested by Lan-
chester.* The iirst refers to combat opera-
tions involving regular troops:
rw: a$ (t)""b!/ U) +P (t)
9%-gl:-cx(:)-dy()+Q(z).
* Examples of combat operations are given in the
article by C.S. Coleman, "Combat models" (see
Dijjercntial Equation Models, eds. M. Braun,
C.S. Coleman, and D.A. Drew, New York: Sprin-
ger, 1983, pp. 109-131). The article shows how
the examples agree with the models considered.
the equation
d ~ ~ t ) =-(OLR+CLR)+RR. (47)
A similar equation holds true for y (t).
The problem now is to find the appropriate
formulas for the quantities OLR, CLR,
and RR and then examine the resulting
differential equations. The results should
indicate the probable winner.
We use the following notation: a, b,
c, d, g, and h are nonnegative constants
characterizing the effect of various factors
on personnel losses on both sides (x and y),
P (t) and Q (t) are terms that allow for the
possibility of reinforcements being supplied
to x and y in the course of one day, and
x
0
and Yo are the personnel in x and y prior
to combat. We are now ready to set up the
three combat models suggested by Lan-
chester. * The first refers to combat opera-
tions involving regular troops:
d ~ ~ t ) =-ax (t)-by (t) +P (t),
d ~ ~ t ) =-cx(t)-dy(t)+Q(t).
Examples of combat operations are given in the
article by C.S. Coleman, "Combat models" (see
Differential Equation Models, eds. M. Braun,
C.S. Coleman, and D.A. Drew, New York: Sprin-
ger, f983, pp. f09-f31). The article shows how
the examples ajO"ee with thll models conaidered. 58 Differential Equations in Applications
In what follows we call this system the A-
type differential system (or simply the A-
type system).
The second model, specified by the equa-
tions
d<r(> ..
q;$()()y()+P()
d i
!jg%=dy(*)h()y()+Q()
describes combat operations involving only
guerilla forces. We will call this system the
B-type system. Finally, the third model,
which we call the C-type system, has the
form
d
%%Q= * (t)-.()y()+P()
d
%L= - cw ()dy(> +Q ()
and describes combat operations involving
both regular troops and guerilla forces.
Each of these differential equations
describes the rate of change in the number
of personnel on the opposing sides as a func-
tion of various factors and has the form
(47). Losses in personnel that are not direct-
ly related to combat operations and are de-
termined by the terms a.2: (t) and
dy (t) make it possible to describe the
constant fractional loss rates (in the absence
of combat operations and reinforcements)
In what follows we call this system the A-
type differential system (or simply the A-
type system).
The second model, specified by the equa-
tions
d.r (t)
(It= -ax (t) -gx (t) y (t) +P (t),
d y ~ : =-dy(t)-hx(t)y(t)+Q(t)
describes combat op-erations involving only
guerilla forces. We will call this system the
B-type system. Finally, the third model,
which we call the C-type system, has the
form
d x d ~ t ) = -ax (t)- gx (t) y (t) + P (t),
dyd(:) = --ex (t)- dy (t) + Q (t)
and describes combat operations involving
both regular troops and guerilla forces.
Each of these differential equations
describes the rate of change in the number
of personnel on the opposing sides as a func-
tion of various factors and has the form
(47). Losses in personnel that are not direct-
ly related to combat operations and are de-
termined by the terms -ax (t) and
-dy (t) make it possible to describe the
constant fractional loss rates (in the absence
of combat operations and reinforcements) Ch. 1. Construction of Differential Models 59
via the equations
1 dx ___ 1 Q_ _
YHT" a Y dz "' d'
If the Lanchester models contain only
terms corresponding to reinforcements and
losses not associated with combat opera-
tions, this means that there are no com-
bat operations, while the presence of the
terms by (t), c;z: (t), -gx (t) y (t), and
-hx (t) y (t) means that combat operations
take place.
In considering the A-type system, we
assume, first, that each side is within the
range of the fire weapons of the other and,
second, that only the personnel directly
involved in combat come under fire. Under
such assumptions Lanchester introduced the
term -by (t) for the regular troops of side
zz: to reflect combat losses. The factor b
characterizes the effectiveness of side y in
combat. Thus, the equation
1 dx
2wb
shows that constant b measures the average
effectiveness of each man on side y. The
same interpretation can be given to the term
cx (t). Of course, there is no simple way
in which we can calculate the effectiveness
coefficients b and c. One way is to write
these coefficients in the form
b == I',/py, C == Txpx, (48)
via the equations
1 d% !. dy = - d
"i" Tt = -a, y dt
If the Lanchester models contain only
terms corresponding to reinforcements and
losses not associated with combat opera-
tions, this means that there are no com-
bat operations, while the presence of the
terms -by (t), -ex (t), -gx (t) y (t), and
-hx (t) y (t) means that combat operations
take place.
In considering the A-type system, we
assume, first, that each side is within the
range of the fire weapons of the other and,
second, that only the personnel directly
involved in combat come under fire. Under
such assumptions Lanchester introduced the
term -by (t) for the regular troops of side
x to reflect combat losses. The factor b
characterizes the effectiveness of side y in
combat. Thus, the equation
1 dx
--=-b
y dt
shows that constant b measures the average
effectiveness of each man on side y. The
same interpretation can be given to the term
-ex (t). Of course, there is no simple way
in which we can calculate the effectiveness
coefficients b and c. One way is to write
these coefficients in the form
(48) 60 Differential Equations in Applications
where ry and rx are the coefficients of fire-
power of sides y and x, respectively, and py
and px are the probabilities that each shot
fired by sides y and x, respectively, proves
accurate.
We note further that the terms correspond-
ing to combat losses in the A-type system
are linear, whereas in the B-type system
they are nonlinear. The explanation lies in
the following. Suppose that the guerilla
forces amounting to x (t) personnel occupy a
certain territory R and remain undetected
by the opposing side. And although the lat-
ter controls this territory by firepower, it
cannot know the effectiveness of its ac-
tions. lt is also highly probable that the
losses suffered by the guerilla forces x are,
on the one hand, proportional to the num-
ber of personnel sc (t) on R and, on the
other, to the number y (t) of personnel of
the opposing side. Hence, the term corre-
sponding to the losses suffered by the gueril-
la forces x has the form ga; (t) y (t), where
the coefficient reflecting the effectiveness of
combat operations of side y is, in general,
more difficult to estimate than the coef-
fiient b in the first relationship in (48).
Nevertheless, to find g we can use the fire-
power coefficient ry and also allow for the
ideas expounded by Lanchester, according
to which the probability of an accurate
shot fired by side y is directly proportional
to the so-called territorial effectiveness A,-y
where r y and r x are the coefficients of fire-
power of sides y and x, respectively, and Pu
and Px are the probabilities that each shot
fired by sides y and x, respectively, proves
accurate.
We note further that the terms correspond-
ing to combat losses in the A-type system
are linear, whereas in the B-type system
they are nonlinear. The explanation lies in
the following. Suppose that the guerilla
forces amounting to x (t) personnel occupy a
certain territory R and remain undetected
by the opposing side. And although the lat-
ter controls this territory by firepower, it
cannot know the effectiveness of its ac-
tions. It is also highly probable that the
losses suffered by the guerilla forces x are,
on the one hand, proportional to the num-
ber of personnel x (t) on R and, on the
other, to the number y (t) of personnel of
the opposing side. Hence, the term corre-
sponding to the losses suffered by the gueril-
la forces x has the form -gx (t) y (t), where
the coeffi.cient reflecting the effectiveness of
combat operations of side y is, in general,
more difficult to estimate than the coef-
fiient b in the first relationship in (48).
Nevertheless, to find g we can use the fire-
power coefficient r y and also allow for the
ideas expounded by Lanchester, according
to which the probability of an accurate
shot fired by side y is directly proportional
to the so-called territorial effectiveness A ry Ch. 1. Construction of Differential Models 61
of a single shot Bred by y and inversely
proportional to the area Ax of the territory
R occupied by side x. Here by ATU we
denote the area occupied by a single gueril-
la frghter. Thus, with a high probability
we can assume that the formulas for
finding g and h are
A A
g=ry-f, h=rxx-{Ji-. (49)
Below we discuss in greater detail each of
the three differential models.
Case A (differential systems of the A-type
and the quadratic law). Let us assume that
the regular troops of the two opposing
sides are in combat in the simple situation
in which the losses not associated directly
with combat operations are nil. If, in addi-
tion, neither side receives any reinforce-
ment, the mathematical model is reduced
to the following differential system:
d d
z-. -by, : -cx. (50)
Dividing the second equation by the first,
we find that
dy __ isi
-5 -- by . (51)
Integrating, we arrive at the following re-
lationship:
b [y2 (t) -y:] = c [x2 (t) xg]. (52)
Ch. 1. Construction of Differential Models 61.
of a single shot fired by y and inversely
proportional to the area Ax of the territory
R occupied by side x. Here by A ry we
denote the area occupied by a single gueril-
la f1ghter. Thus, with a high probability
we can assume that the formulas for
finding g and h are
h
Ar:x:
= r x ~
II
(49)
Below we discuss in greater detail each of
the three differential models.
Case A (differential systems of the A-type
and the quadratic law). Let us assume that
the regular troops of the two opposing
sides are in combat in the simple situation
in which the losses not associated directly
with combat operations are nil. If, in addi-
tion, neither side receives any reinforce-
ment, the mathematical model is reduced
to the following differential system:
~ b ~ 50
y,=- y, dt= -ex. ( )
Dividing the second equation by the first,
we find that
:: = ~ = . (51)
Integrating, we arrive at the fallowing re-
lationship:
b [y
2
{t) - y:J = e [x
2
(t) - x:J. (52) 62 Differential Equations in Applications
ya,) A>0, y wins
K = 0) even
Wl//b K < Q .z* wins
0 r
{/-/(/c $(5)
Fig. 12
This relationship explains why system (50)
corresponds to a model with a quadratic
law. If by K we denote the constant quanti-
ty by c:c, the equation
byz - cxz = K (53)
obtained from Eq. (52) specifies a hyperbola
(or a pair of straight lines if K = 0) and
we can classify system (50) with great pre-
cision. We can call it a differential system
with a hyperbolic law.
Figure 12 depicts the hyperbolas for differ-
ent values of K; for obvious reasons we
consider only the first quadrant (:1: > 0,
y > 0). The arrows on the curves point in
Fig. f2
This relationship explains why system (50)
corresponds to a model with a quadratic
law. If by K we denote the constant quanti-
ty b y ~ - ex:, the equation
by
2
- cx
2
= K (53)
obtained from Eq. (52) specifies a hyperbola
(or a pair of straight lines if K = 0) and
we can classify system (50) with great pre-
cision. We can call it a differential system
with a hyperbolic law.
Figure 12 depicts the hyperbolas for differ-
ent values of K; for obvious reasons we
consider only the first quadrant (x ~ 0,
y ~ 0). The arrows on the curves point in Ch. 1. Construction of Differential Models 63
the direction in which the number of per-
sonnel changes with passage of time.
To answer the question of who wins in
the constructed model (50), we agree to say
that side y (or ac) wins if it is the first to
wipe out the other side x (or y). For exam-
ple, in our case the winner is side y if
K > 0 since in accordance with Eq. (53)
the variable y can never vanish, while the
variable sc vanishes at y (t) = l/K/b. Thus,
for y to win a victory, it must strive for a
situation in which K is positive, that is,
byg > c:1:. (54)
Using (48), we can rewrite (54) in the form
yo 2 fx Px
<..> > ry sa
The left-hand side of (55) demonstrates that
changes in the personnel ratio yo/xo give
an advantage to one side, in accordance
with the quadratic law. For example, a
change in the yo/.2:0 from one to two gives y
a fourfold advantage. Note also that
Eq. (53) denotes the relation between the
personnel numbers of the opposing sides but
does not depend explicitly on time. To de-
rive formulas that would provide us with a
time dependence, we proceed as follows.
We differentiate the first equation in (50)
with respect to time and then use the second
equation in this system. As a result we
the direction in which the number of per-
sonnel changes with passage of time.
To answer the question of who wins in
the constructed model (50), we agree to say
that side y (or x) wins if it is the first to
wipe out the other side x (or y). For exam-
ple, in our case the winner is side y if
K > 0 since in accordance with Eq. (53)
the variable y can never vanish, while the
variable x vanishes at y (t) = V Klb. Thus,
for y to win a victory, it must strive for a
~ i t u a t i o n in which K is _positive, that is,
(54)
Using (48), we can rewrite (54) in the form
(55)
The left-hand side of (55) demonstrates that
changes in the personnel ratio y
0
/x
0
give
an advantage to one side, in accordance
with the quadratic law. For example, a
change in the y
0
/x
0
from one to two gives y
a four-fold advantage. Note also that
Eq. (53) denotes the relation between the
personnel numbers of the opposing sides but
does not depend explicitly on time. To de-
rive formulas that would provide us with a
time dependence, we proceed as follows.
We differentiate the first equation in (50)
with respect to time and then use the second
equation in this system. As a result we 64 Differential Equations in Applications
.9
%
(U
i.
M2
---1-----Zv- *I1lQ
0 ( .
T
Fig. 13
arrive at the following differential equation:
dh:
EF " bC.IZ T-" O.
If for the initial conditions we take
dx
$(0)=% g;L=0= b1/0,
then the solution to Eq. (56) can be ob-
tained in the form
sc (t) = :1:0 cos [St yyo sin Bt, (57)
where B = VE and Y =
In a similar manner we can show that
y (t) L-. yo cos Bt gg- sin Bt. (58)
Figure 13 depicts the graphs of the func-
tions specihed by Eqs. (57) and (58) in
:c !I
Fig. 13
dx
dt -bcx=O.
(56)
If for the initial conditions we take
X (0) = x0, :; lf=O = - by0,
then the solution to Eq. (56) can be ob-
tained in the form
x (t) = x
0
cos ~ t - 'VYo sin ~ t , (57)
where ~ = Vbc and 'V = Vblc.
In a similar manner we can show that
y (t) =Yo cos Pt- ~ s i n ~ t . (58)
'V
Figure 13 depicts the graphs of the func-
tions specified by Eqs. (57) and (58) in Ch. 1. Construction of Differential Models 65
the case where K >O (i.e. by > cxg, or
YU 0 > xo)- _ _
We note in conclusion that for side y to
win a victory it is not necessary that yu be
greater than xo. The only requirement is
that vyo be greater than xo.
Case B (difierential systems of t.he B-type
and the linear law). The dynamical equa-
tions that model combat operations between
two opposing sides can easily be solved,
as in the previous case, if we exclude the
possibility of losses not associated with
combat and if neither side receives rein-
forcements. Under these limitations the
Btype differential system assumes the form
d d
f-: gy, -5%: hy- (59)
Dividing the second equation in (59) by
the first, we get the equation
i!!...
dx -_ g
which after being integrated yields
g ly () - ytl = h lx () - wo]- (60)
This linear relationship explains why the
nonlinear system (59) corresponds to a mo-
del with a linear law for conducting combat
operations. Equation (60) can be rewritten
in the form
gy hx = L, (61)
50770
the case where K > 0 (i.e. b y ~ > e x ~ , or
YYo > Xo)
We note in conclusion that for side y to
win a victory it is not necessary that y
0
be
greater than x
0
The only requirement is
that yy
0
be greater than x
0

Case B (differential systems of the B-type
and the linear law). The dynamical equa-
tions that model combat operations between
two opposing sides can easily be solved,
as in the previous case, if we exclude the
possibility of losses not associated with
combat and if neither side receives rein-
forcements. Under these limitations the
B-type differential system assumes the form
dx dy h
dt= -gxy, dt=- xy.
(59)
Dividing the second equation in (59) by
the first, we get the equation
dy h
dx=g
which after being integrated yields
g [y (t) - y
0
] = h [x (t) - x
0
1. (60)
This linear relationship explains why the
nonlinear system (59) corresponds to a mo-
del with a linear law for conducting combat
operations. Equation (60) can be rewritten
in the form
gy - hx = L, (61)
.90} DQ _ywz'ns
l=0, EUC/Z
4/y L<Q, .z* wirzs
0 .
;//7 mw)
Fig. 14
with L = gy,) -hay,. This implies, for one,
that if L is positive, side y wins as a result
of combat operations, while if L is nega-
tive, side sc wins.
Figure 14 provides a geometrical inter-
pretation of the linear law (61) for different
values of L.
Let us now study in greater detail the
situation where one of the sides wins. We
assume that this is side y. Then, as we al-
ready know, gyo hxo must be positive, or
L l
$0 > 8
If we now turn to (49), we see that side y
wins if
_/L "xArxAx
xo > ryA,.yAy ' ( )
!f(t) L> 0, !I wins
tc(t)
Fig. 14
with L = gy
0
- hx
0
This implies, for one,
that if L is positive, side y wins as a result
of combat operations, while if L is nega-
tive, side x wins.
Figure 14 provides a geometrical inter-
pretation of the linear law (61) for different
values of L.
Let us now study in greater detail the
situation where one of the sides wins. We
assume that this is side y. Then, as we al-
ready know, gy
0
- hx
0
must be positive, or
..J!.g_ >.!!...
zo g
If we now turn to (49}, we see that side y
wins if
(62) Ch. 1. Construction of Differential Models 67
Thus, the strategy of y is to make the ratio
yo/xo as large as possible and the ratio
Ax/A y as small as possible. From the practi-
cal standpoint it is more convenient to
write inequality (62) in the form
Ay!/0 xArx
Axxo > "yAry
We see that in a certain sense the products
A yyo and Axxo constitute critical values.
We note, finally, that by combining
(61) and (59) we can easily derive formulas
that give the time dependence of changes
in the personnel numbers of both sides.
Case C (differential systems of the Ctype
and the parabolic law). In the C-model the
guerilla forces face regular troops. We in-
troduce the simplifying assumptions that
neither is supplied with reinforcements and
that the losses associated directly with com-
bat operations are nil. In this case we have
the following system of differential equa-
tions:
where :2: (t) is the number of personnel in
the guerilla forces, and y (t) the number of
personnel in the regular troops. Dividing
the second equation in (63) by the first, we
arrive at the differential equation
12 -.;
dx " gy '
5*
Ch. f. Construction ot Differential Models 67
Thus, the strategy of y is to make the ratio
y
0
/x
0
as large as possible and the ratio
A ~ / A y as small as possible. From the practi-
cal standpoint it is more convenient to
write inequality (62) in the form
AyYo > rxArx
AxXo r
11
Ary '
We see that in a certain sense the products
AyYo and Axxo constitute critical values.
We note, finally, that by combining
(61) and (59) we can easily derive formulas
that give the time dependence of changes
in the personnel numbers of both sides.
Case C (differential systems of the C-type
and the parabolic law). In the C-model the
guerilla forces face regular troops. We in-
troduce the simplifying assumptions that
neither is supplied with reinforcements and
that the losses associated directly with com-
bat operations are nil. In this case we have
the following system of differential equa-
tions:
dx dy
dt=-gxy, dt=-cx, (63)
where x (t) is the number of personnel in
the guerilla forces, and y (t) the number of
personnel in the regular troops. Dividing
the second equation in (63) by the first, we
arrive at the differential equation
dy c
dx = gy 68 Differential Equations in Applications
_y() /V/>Q ywi/ZS //=Q eve/z
1//Qi, /*/<Q .z wins
M/20 .z(t')
Fig. 15
Integrating with the appropriate limits
yields the following relationship:
gyz (t) =.= 20:1: (t) -M, (64)
where M : gy -- 2c;zv0. Thus, the system
of differential equations (63) corresponds to
a model with a parabolic law for conducting
combat operations. The guerilla forces win
if M is negative; they are defeated if M is
positive.
Figure 15 depicts the parabolas defined
via Eq. (64) for different values of lll.
Experience shows that regular troops can
defeat guerilla forces only if the ratio yo/:::0
considerably exceeds unity. Basing our rea-
soning on the parabolic law for conducting
combat operations and assuming M posi-
tive, we conclude that the victory of regular
HB Differential Equations in Applications
!J(l) H>O, !I wins #=4 even.
-M/2c :c{t)
Fig. 15
Integrating with the appropriate limits
yields the following relationship:
gy2 (t) =-= 2cx (t) +M, (64)
where M = gy: - 2cx
0
. Thus, the system
of differential equations (63) corresponds to
a model with a parabolic law for conducting
combat operations. The guerilla forces win
if M is negative; they are defeated if M is
positive.
Figure 15 depicts the parabolas defined
via Eq. (64) for different values of M.
Experience shows that regular troops can
defeat guerilla forces only if the ratio y
0
/x
0
considerably exceeds unity. Basing our rea-
soning on the parabolic law for conducting
combat operations and assuming M posi-
tive, we conclude that the victory of regular Ch. 1. Construction of Differential Models 69
troops is guaranteed if (yo/x0)2 is greater
than (2c/g) .1:;*. If we allow for (48) and
(49), we can rewrite this condition in the
form
y 2 T A p_ 1
> 2 i yt .;;_
1.12 Why Are Pendulum Clocks
Inaccurate?
To answer this question, let us consder an
idealized model of a pendulum clock con-
sisting of a rod of length Z and a load of mass
m attached to the lower end of the rod
(the rods mass is assumed so small that it
can be ignored in comparison to m) (Fig-
ure 16). Iilthe load is;def1eoted by an angle ot
7
{ `X
CX\\E
v \\
I 8 \\\
l x>
| /
t r'
nz
L=-rz
Fig. 16
Ch. 1. Construction of Differential Models 6\J
troops is guaranteed if (y
0
/x
0
)
2
is greater
than (2c/g) x;. If we allow for (48) and
(49), we can rewrite this condition in the
form
1.12 Why Are Pendulum Clocks
Inaccurate?
To answer this question, let us consder an
idealized model of a pendulum clock con-
sisting of a rod of length l and a load of mass
m attached to the lower end of the rod
(the rod's mass is assumed so small that it
can be ignored in comparison to m) (Fig-
ure 16). If the load i(deflected by an angle a
I
I
I
I
I
I
~ - - - -
s
Fig. 16 70 Differential Equations in Applications
and then released, in accordance with
energy conservation we obtain
ili=mg(lcos9lcosu), (65)
where v is the speed at which the load
moves, and g is the acceleration of gravity.
Since s == l9, we have v = ds/dt ==
l(d9/dt) and (65) leads us to tl1e following
differential equation:
-g.(-)2=g(e0Se0Sa). (ee)
If we now allow for the fact that -6 decreases
with the passage of time t (for small fs),
we can rewrite Eq_. (66) in the form `
dt: I .
1/ 23 ]/cos6-cosot
If we denote the period of the pendulum
by T, we have
_ o
J;- .. LS ...19...
4 1/ 29 ]/cost)-cosoc
(I
or
T Q d6
I/28 0 ]/cosBcosoz ( )
The last formula shows that the period of
the pendulum depends on ot. This is the
and then released, in accordance with
energy conservation we obtain
mvs
-
2
-=mg (l cosO-lcosa), (65)
where v is the speed at which the load
moves, and g is the acceleration of gravity.
Since s = za, we have v = dsldt =
l (d9/dt) and (65) leads us to the following
l ( dO )2
2
dt =g(cosO-cosa). (66)
If we now allow for the fact that a decreases
with the passage of time t (for small t's),
we can rewrite Eq. (66) in the form
dt = - .. / l
v 2g v cos 0-cos a.
If we denote the period of the pendulum
by T, we have
.. ;-- 0
T V l dO
T =- 2g ycos9-cosa. '
0\
or
T=
4
.. /=l rJ=:::::;<=d=a =
V 2g y cos9-cosa. '
0
(67)
The last formula shows that the period of
the pendulum depends on a. This is the Ch. 1. Construction of Differential Models 71
main reason why pendulum clocks are inac-
curate since, practically speaking, every
time that the load is deflected to the ex-
treme position the deflection angle differs
from cx.
We note that formula (67) can be written
in a simpler form. Indeed, since
cos9=l2sin2g, cosoc:-1--2 sin2E2 ,
we have
___ G
T- 2 I/L S ...........@._.......
8 _ on _ 6
0 I/-sin? -2s1n2 -5-
"' d9
0 I/k2si112 -5
with k = sin (cx/2).
Now instead of variable 9 we introduce
a new variable cp via the formula
sin (6/2) = lc sin cp. This implies that when
6 increases from O to ct, the variable cp
grows from O to rt/2, with
goes-;-d6=kcosq>dcp,
or
2 I/ksin
2k cos cp dcp 2
COS? }/1-k s1nq>
Ch. 1. Censtruetion of Differential Models 71
main reason why pendulum clocks are inac-
curate since, practically speaking, every
time that the load is deflected to the ex-
treme position the deflection angle differs
from a.
We note that formula (67) can be written
in a simpler form. Indeed, since
cos8=1-2sin
2
~ , cosa=1-2 sin
2
~ ,
we have
(68)
with k = sin (a/2).
Now instead of variable 8 we introduce
a new variable q> via the formula
sin (8/2) = k sin q>. This implies that when
e increases from 0 to a, the variable q>
g_rows from 0 to nt2, with
1 6
2
cos
2
d8 =k cos <pdq>,
or
da = 2k cos ~ dcp
cos2 72 Differential Equations in Applications
The last relationship enables us to rewrite
formula (68) in the form
[Ta/2 d
.. ___ ....._L.....
T"4l g ]/1k2sinq>
:.4 1/..F(k, it/2),
where the function
w dw
F k, = S -
( 0 1/1k2 sinzcp
is known as the elliptic integral of the first
kind, differing from the elliptic integral of
the second kind
wb
E'(k, ip)-: S l/1k2sin2q> dqn.
0
Elliptic integrals cannot be expressed in
terms of elementary functions. Therefore,
all further discussion of the pendulum prob
lem will be related to an approach consid-
ered when conservative systems are studied
in mechanics. Here we only note that the
starting point in our studies will be the
differential equation
-gag--f-kSiH9=O, k=
which can be obtained from Eq. (60) by
differentiating with respect to t.
The last relationship enables us to rewrite
formula (68) in the form
_n/2
T=4l;_z r dcp
g ~ y1-k2sin2cp
0
=4 y ; F (k, n/2),
where the function
1b
r dcp
F(k, 1p)= ~ yi-k=sin2cp
0
is known as the elliptic integral of the first
kind, differing from the elliptic integral of
the second kind
1b
E(k,'lj1)= ~ Y1-k
2
sin
2
cpdcp.
0
Elliptic integrals cannot be expressed in
terms of elementary functions. Therefore,
an further discussion of the pendulum prob-
lem will be related to an approach consid-
ered when conservative systems are studied
in mechanics. Here we only note that the
starting point in our studies will be the
d26 . -v-
dt2 +ksm9=0, k= gil,
which can be obtained from Eq. (60) by
differentiating with respect to t. Ch. 1. Construction of Differential Models 73
1.13 The Cycloidal Clock
We have established that clocks with ordi-
nary (circular) pendulums are inaccurate.
Is there any pendulum whose period is in-
dependent of the swing? This problem was
first formulated and solved as early as the
17th century. Below we give its solution,
but first let us turn to the derivation of the
equation of a remarkable curve, which
Galileo Galilei called the cycloid (from the
Greek word for "circular"). This is a plane
curve generated by a point on the circum-
ference of a circle (called the generating
circle) as it rolls along a straight line with-
out slippage.
Suppose that the x axis is the straight
line along which the generating circle rolls
and that the radius of this circle is r (Fig-
ure 17). Let us also assume that initially
the point that traces the cycloid is at the
origin and that after the circle turns through
an angle 6, the point occupies position M.
Then, on the basis of geometrical reasoning
we obtain
x = OS -= OP - SP,
y-=MS=CPCN.
But
OP =MP ==r9, SP =MN=rsin9,
CP = r, CN = rcostl.
1.13 The Cycloidal Clock
We have established that clocks with ordi-
nary (circular) pendulums are inaccurate.
Is there any pendulum whose period is in-
dependeRt of the swing? This problem was
first formulated and solved as early as the
17th century. Below we give its solution,
but first let us turn to the derivation of the
equation of a remarkable curve, which
Galileo Galilei ca lied the cycloid (from the
Greek word for "circular"). This is a plane
curve generated by a point on the circum-
ference of a circle (called the generating
circle) as it rolls along a straight line with-
out slippage.
Suppose that the x axis is the straight
line along which the generating circle rolls
and that the radius of this circle is r (Fig-
ure 17). Let us also assume that initially
the poiDt that traces the cycloid is at the
origin and that after the circle turns through
an angle e, the point occupies position M.
Then, on the basis of geometrical reasoning
we obtain
x =OS = OP - SP,
y = .MS = CP - CN.
But
OP = MP = re, SP = MN = r sine,
CP = r, CN = r cos e. 74 Differential Equations in Applications
.9
M L 1
0 A .z
_ 3 P ZJz'r ` _
Fig. 17
Hence, parametrically the cycloid is spec_i-
fied by the following equations:
x=r(9-sin9), y=r(1cos6). (69)
If we exclude parameter 9 from these equa-
tions, we arrive at the following equation
of a cycloid in a rectangular Cartesian co-
ordinate system Oxy:
33 =.rcos1(-K;-i)l/2ry-y2.
The very method of constructing a cy-
cloid implies that the cycloid consists of con-
gruent arcs each of which corresponds to a
full revolution of the generating circle.*
* The reader may find many interesting facts about
the cycloid and related curves in G.N. Bermans
book The Cycloid (Moscow: Nauka, 1980) (in
Russian).
II
Fig. 17
Hence, the cycloid is speci-
fied by the following equations:
X = r (a - sin a), y = r (1 - COS a). (69)
If we exclude parameter a from these equa-
tions, we arrive at the following equation
of a cycloid in a rectangular Cartesian co-
ordinate system Oxy:
x = r cos-
1
( r-; Y } - V 2ry- y2
The very method of constructing a cy-
cloid implies that the cycloid consists of con-
gruent arcs each of which corresponds to a
full revolution of the generating circle.*
The reader may find many interesting facts about
the cycloid and related curves in G.N. Berman's
book The Cycloid (Moscow: Nauka, 1980) (in
Russian). Ch. 1. Construction of Differential Models 75
The s,epara_te arcs are linked at points
where they have a common vertical tan-
gent. These points, known as cusps, corre-
spond to the lowest possible positions occu-
pied by the point on the generating circle
that describes the cycloid. The highest
possible positions occupied by the same
point lie exactly in the middle of each are
and are known as the vertices. The distance
along the straight line between successive
cusps is 2:nr, and the segment of that
straight line between successive cusps is
known as the base of an arc of the cycloid.
The cycloid possesses the following prop-
erties:
(a) the area bounded by an are of a cycloid
and the respective base is thrice the area of
the generating circle (Galileos theorem);
(b) the length of one cycloid arc is four
times the length of the diameter of the gen-
erating circle (Wrens theorem).
The last result is quite unexpected, since
to calculate the length of such a simple
curve as the circumference of a circle it is
necessary to introduce the irrational number
at, whose calculation is not very simple,
while the length of an arc of a cycloid is
expressed as an integral multiple of the ra-
dius (or diameter) of the generating circle.
The cycloid has many other interesting
properties, which have proved to be extreme-
ly important for physics and engineering.
For example, the proiile of pinion teeth
The separa.te a,:"cs 11.re lip,ked at points
where they have a common vertical tan-
gent. These points, known as cusps, corre-
spond to the lowest possible positions occu-
pied by the point on the generating circle
that describes the cycloid. The highest
possible positions occupied by the same
point lie exactly in the middle of each arc
and are known as the vertices. The distance
along the straight line between s u c c e s s ~ v e
cusps is 2nr, and the segment of that
straight line between successive cusps is
known as the base of an arc of the cycloid.
The cycloid possesses the following prop-
erties:
(a) the area bounded by an arc of a cycloid
and the respective base is thrice the area of
the generating circle (Galileo's theorem);
(b) the length of one cycloid arc is four
times the length of the diameter of the gen-
erating circle (Wren's theorem).
The last result is quite unexpected, since
to calculate the length of such a simple
curve as the circumference of a circle it is
necessary to introduce the irrational number
n, whose calculation is not very simple,
while the length of an arc of a cycloid is
expressed as an integral multiple of the ra-
dius (or diameter) of the generating circle.
The cycloid has many other interesting
properties, which have proved to be extreme-
ly important for physics and engineering.
For example, the profile of pinion teeth 76 Differential Equations in Applications
:4 /1
m
lei
V
Fig. 18
and of many typos of eccentrics, cams, and
other mecha11ical parts of devices have the
shape of a cycloid.
Let us now turn to the problem that
enabled the Dutch physical scientist, astro-
nomer, and mathematician Christian Huy-
gens (1629-1695) to build an accurate clock
in 1673. The problem consists of building
in the vertical plane a curve for which the
time of descent of a heavy particle sliding
without friction to a fixed horizontal line
is the same wherever the particle starts on
the curve (it is assumed that initially, at
time t = to, the particle is at rest). Huy
gens found that the cycloid possesses such
an isochronous (from the Greek words for
"equal" and "time") or tautochronous
property (from tautos for "identical").
From the practical standpoint the prob-
lem can be solved in the following manner.
Let us assume that a trough in the form of a
cycloid is cut out of a piece of wood as
shown in Figure 18. We take a small metal
ball and send it rolling down the slope.
Ignoring friction, let us try to determine
Fig. 18
and of many types of eccentrics, cams, and
other mechanical parts of devices have the
shape of a cycloid.
Let us now turn to the problem that
enabled the Dutch physical scientist, astro-
nomer, and mathematician Christian Huy-
gens (1629-1695) to build an accurate clock
in 1673. The problem consists of building
in the vertical plane a curve for which the
time of descent of a heavy particle sliding
without friction to a fixed horizontal line
is the same wherever the particle starts on
the curve (it is assumed that initially, at
time t = t
0
, the particle is at rest). Huy-
gens found that the cycloid possesses such
an isochronous (from the Greek words for
"equal" and "time") or tautochronous
property (from to.utos for "identir.al").
From the practical standpoint the prob-
lem can be solved in the following manner.
Let us assume that a trough in the form of a
cycloid is cut out of a piece of wood as
shown in Figure 18. We take a small metal
ball and send it rolling down the slope.
Ignoring friction, let us try to determine Ch. 1. Construction of Differential Models 77
the time it takes the ball to reach the low-
est possible point, K, starting from point
M, say.
Let :1:0 and y0 be the coordinates of the
initial position of the ball, i.e. point M,
and 90 the corresponding value of para-
meter 9. When the ball reaches a pointN (9),
the distance of descent along the vertical
will be h, which in view of Eq. (69) can be
found in the following manner:
h=yy0=r(1cos6)
r (1 cos 60) = r (cos 90 cos 9).
We know that the speed of a falling object
is given by the formula
U= l/EEE,
where g is the acceleration of gravity. In
our case the last formula assumes the form
v.-: l/2gr (cos 90- cos 9)
On the other hand, since speed is the deriv-
ative of distance s with respect to time T,
we arrive at the formula
Tg-: V 2gr (cos 90 cos 6)
Since for a cycloid ds = 2r sin (9/2) dB, we
can rewrite this formula (which is actually
a differential equation) in the form
dT :2 2r sin (9/2) df}
]/2gr (cos 90 cos 6)
Ch. 1. Construction of Differential Models 7 7
the time it takes the ball to reach the low-
est possible point, K, starting from point
M, say.
Let x
0
and Yo be the coordinates of the
initial position of the ball, i.e. point M,
and 8
0
the corresponding value of para-
meter e. When the ball reaches a point N (8),
the distance of descent along the vertical
will be h, which in view of Eq. (69) can be
found in the following manner:
h = y - Yo = r (1 - cos 8)
- r (1 - cos 8
0
) = r (cos 8
0
- cos 8).
We know that the speed of a falling object
is given by the formula
V= V2gh,
where g is the acceleration of gravity. In
our case the last formula assumes the form
v = V2gr (cos eo- cos 8)
On the other hand, since speed is the deriv-
ative of distances with respect to time T,
we arrive at the formula
:; = lf2gr(cos8
0
-cos8)
Since for a cycloid ds = 2r sin (8/2) d8, we
can rewrite this formula (which is actually
a differential equation) in the form
dT = 2r sin (6/2) d6
V 2gr (cos 6
0
- cos 6) 78 Diiierential Equations in Applications
Integrating this equation within approp-
riate limits, we obtain
JT
T__ S 2r sin (B/2) dB
U `|/2gr(cos00-cos0)
___ TIS
___2l/LS dc0s(9/2)
g 0 ./ cos? -%-cos2g
== JT
Thus, the time interval T in the course
of which the ball rolls down from point M
to point K is given by the formula
T = TL
which shows that period T is independent
of 9,,, that is, the period does not depend
on the initial position of the ball, M.
Obviously, two balls that begin their mo-
tion simultaneously from points M and N
will roll down and find themselves at point
K at the same moment of time.
Since we agreed to ignore friction, in its
motion down the slope the ball passes point
K and, by inertia, continues its motion
up the slope to point M1 lying at the same
level as point M. Proceeding then in the
opposite direction and traversing its path,
the ball completes a full cycle. This con-
stitutes the motion of a cycloidal pendulum
78 bifferential Equations in Applications
Integrating this equation within approp-
riate limits, we obtain
"'
T = r 2r sin (9/2) rl9
J Jl 2gr (cos 0
0
-cos 0)
u
_:rt
= _
2
.. / !_ r _ d cos (9;2)
V g J 9 9
o ~ / cos2 --f"-cos
2
2
;-
=n l rig.
Thus, the time interval T in the course
of which the ball rolls down from point M
to point K is given by the formula
T=n"Vrlg,
which shows that period T is independent
of eo, that is, the period does not depend
on the initial position of the ball, M.
Obviously, two balls that begin their mo-
tion simultaneously from points M and N
will roll down and find themselves at point
K at the same moment of time.
Since we agreed to ignore friction, in its
motion down the slope the ball passes point
K and, by inertia, continues its motion
up the slope to point M
1
lying at the same
level as point M. Proceeding then in the
opposite direction and traversing its path,
the ball completes a full cycle. This con-
stitutes the motion of a cycloidal pendulum Cli. 1. Const.ruction of Differential Models 79
\
\\ \
\ ~`
\\ \\ `
\\\` ///)
Fig. 19
with a period of oscillations
T0 = lm Vr/g. (70)
A peculiar feature of the cycloidal pen-
dulum, which sets it apart from the simple
(circular) pendulum, is that its period does
not depend on the amplitude (or swing).
Let us now show how an ordinary circu-
lar pendulum can be made to move in a
tautochronous manner without resorting to
trough and similar devices with consider-
able friction. To this end it is sufficient
to make a template (say, out of wood) con-
sisting of two semiarcs of a cycloid witha
common cusp (Figure 19). The template
is fixed in the vertical plane and a string
with a ball is suspended from the cusp O.
The length of the string must be twice the
diameter of the generating circle of the
cycloid.
If the ball on the string is deflected to an
arbitrary point M, it begins to swing back
g and forth with a period that is independ-
Fig. 19
with a period of oscillations
T
0
= 4nVr!g. (70)
A peculiar feature of the cycloidal pen-
dulum, which sets it apart from the simple
(circular) pendulum, is that its period does
not depend on the amplitude (or swing).
Let us now show how an ordinary circu-
lar pendulum can be made to move in a
tautochronous manner without resorting to
trough and similar devices with consider-
able friction. To this end it is sufficient
to make a template (say, out of wood) con-
sisting of two semiarcs of a cycloid with a
common cusp (Figure 19). The template
is fixed in the vertical plane and a string
with a ball is suspended from the cusp 0.
The length of the string must be twice the
diameter of the generating circle of the
cycloid.
If the ball on the string is deflected to an
arbitrary point M, it begins to swing back
\. and forth with a period that is independ-80 Differential Equations in Applications
y ent of the position
of point M. Even
if due to friction
a11d air drag the
amplitude of the
\_ oscillations dimin-
u ishes, the period
' remains unchan-
A ged. For a circu-
B lar pendulum,
C if- which moves along
an arc of a cir-
Fig 20 cumference,the iso-
chronous proper-
ty is satisfied approximately only for
small amplitudes, when the arc differs
little from the arc of a cycloid.
By way of an example let us study the
small oscillations that a pendulum executes
along the arc AB of a cycloid (Figure 20).
lf these oscillations are very small, the
effect of the guiding template is practically
nil and the pendulum oscillates almost like
an ordinary circular pendulum with a string
(the "rod") whose length is 4r. The path AB
of a cycloidal pendulum will differ little
from the path CE of a circular pendulum
with a string length equal to [ir. Hence, the
period of small oscillations of a circular
pendulum with a string length Z: 4r is
practically the same as that of a cycloidal
pendulum of the same length.
A
c
1...
'lo
ll
Fig. 20
ent of the position
of point M. Even
if due to friction
and air drag the
amplitude of the
oscillations dimin-
ishes, the period
remains unchan-
ged. For a circu-
S lar pendulum,
E which moves along
an arc of a cir-
cumference, the iso-
chronous proper-
ty is satisfied approximately only for
small amplitudes, when the arc differs
little from the arc of a cycloid.
By way of an example let us study the
small oscillations that a pendulum executes
-
along the arc AB of a cycloid (Figure 20).
If these oscillations are very small, the
effect of the guiding template is practically
nil and the pendulum oscillates almost like
an ordinary circular pendulum with a string
(the "rod") whose length is 4r. The path AB
of a cycloidal pendulum will differ little
from the path CE of a circular pendulum
with a string length equal to 4r. Hence, the
period of small oscillations of a circular
pendulum with a string length l = 4r is
practically the same as that of a cycloidal
pendulum of the same length. Ch. 1. Construction of Differential Models 81
Now, if in (70) we put r == Z/4, the period
of small oscillations of a circular pendu-
lum can be expressed in terms of the length
of the string:
T = 2% l/{E
This formula is derived (in a different way)
in high school physics.
In conclusion we note that the cycloid
is the only curve for which a particle mov-
ing along it performs isochronous oscilla-
tions.
1.14 The Brachistochrone Problem
The problem concerning the brachistochrone
(from the Greek words for "shortest" and
"time"), a curve of fastest descent, was pro-
posed by the Swiss mathematician John
Bernoulli (1667-1748) in 1696 as a challenge
to mathematicians and consists of the fol-
lowing.
Take two points A and B lying in a ver-
tical plane but not on a single vertical line
(Figure 21). Among the various curves pass-
ing through these two points we must {ind
A
Fig. 21
6-0170 .
Now, if in (70) we put r = l/4, the period
of small oscillations of a circular pendu-
lum can be expressed in terms of the length
of the string:
T = 2:rt Vllg.
This formula is derived (in a different way)
in high school physics.
In conclusion we note that the cycloid
is the only curve for which a particle mov-
ing along it performs isochronous oscilla-
tions.
1.14 The Brachistochrone Problem
The problem concerning the brachistochrone
(from the Greek words for "shortest" and
"time"}, a curve of fastest descent, was pro-
posed by the Swiss mathematician John
Bernoulli (1667-1748) in 1696 as a challenge
to mathematicians and consists of the fol-
lowing.
Take two points A and B lying in a ver-
tical plane but not on a single vertical line
(Figure21). Among the various curves pass-
ing through these two points we must find
Fig. 21
A U.
al OQ
| .2* n c-.z
I O | I
1 ar 1
i 1 Z
I % :
{ n
C 1
5
Fig. 22
the one for which the time required for a
particle to fall from point A to point B
along the curve under the force of gravity is
minimal.
The problem was tackled by the best
mathematicians. It was solved by John
Bernoulli himself and also by Gottfried W.
von Leibniz (1646-1716), Sir Isaac Newton
(1642-1727), Guillaume LHospital (1661-
1704), and Jakob Bernoulli (1654-1705).
The problem is famous not only from the
general scientific viewpoint but also for
being the source of ideas in a completely
new field of mathematics, the calculus of
variations.
Solution of the brachistochrone problem
can be linked with that of another problem
originating in optics. Let us turn to Fig-
ure 22, in which a ray of light is depicted as
propagating from point A to point P with
Fig. 22
the one for which the time required for a
particle to fall from point A to point B
along the curve under the force of gravity is
minimal.
The problem was tackled by the best
mathematicians. It was solved by John
Bernoulli himself and also by Gottfried W.
von Leibniz (1646-1716), Sir Isaac Newton
(1642-1727), Guillaume L'Hospital (1661-
1704), and Jakob Bernoulli (1654-1705).
The problem is famous not only from the
general scientific viewpoint but also for
being the source of ideas in a completely
new field of mathematics, the calculus of
variations.
Solution of the brachistochrone problem
can be linked with that of another problem
originating in optics. Let us turn to Fig-
ure 22, in which a ray of light is depicted as
propagating from point A to point P with Ch. 1. Construction of Differential Models 83
a velocity U1 and then from point P to
point B in a denser medium with a lower ve-
locity v2. The total time Trequired for the
ray to propagate from point A to point B
can, obviously, be found from the following
formula:
T: 1/av+x _,r]/z+;c-x) l
1 2
If we assume that the ray of light propa-
gates from pointA to point B along this path
in the shortest possible time T, the deriva-
tive dT/dx must vanish. But then
I __ C--.23
vi VPKF? vi 7
or
sincil __ sinoiz
Z""?5I"
The last formula expresses Snells well-
known law of refraction, which initially
was discovered in experiments in the form
sin oil/sin oiz = a, with a constant.
The above assumption that light chooses
a path from A to B that would take the
shortest possible time to travel is known as
Fermats principle, or the principle of least
time. The importance of this principle lies
not only in the fact that it can be taken as
a rational basis for deriving Snells law but
also, for one, in that it can be applied to
finding the path of a ray of light in a me-
dium of variable density, where generally
60 Ch. 1. Construction of Differential Models S3
a velocity v
1
and then from point P to
point B in a denser medium with a lower ve-
locity v
2
The total time T required for the
ray to propagate from point A to point B
can, obviously, be found from the following
formula:
T= yaq:=xi +
vl Va
If we assume that the ray of light propa-
gates from point A to point B along this path
in the shortest possible time T, the deriva-
tive dT/dx must vanish. But then
vl va+zl
or
sin a
1
sina
2
vl Va
e-x
The last formula expresses Snell's well-
known law of refraction, which initially
was discovered in experiments in the form
sin a
1
/sin a
2
= a, with a constant.
The above assumption that light chooses
a path from A to B that would take the
shortest possible time to travel is known as
Fermat's principle, or the principle of least
time. The importance of this principle lies
not only in the fact that it can be taken as
a rational basis for deriving Snell's law but
also, for one, in that it can be applied to
finding the path of a ray of light in a me-
dium of variable density, where generally
II* 84 Differential Equations in Applications
Z4 lil ````"``` "
vg O, OE"` `````' "
........ .1...2..- __ ____
v GT
..2... .... .. .... .1.05 _____
v4 I
1%
Fig. 23
the light travels not along straightline
segments.
For the sake of an example let us con-
sider Figure 23, which depicts a ray of light
propagating through a layered medium. In
each layer the speed of light is constant,
but it decreases when we pass from an up
per layer to a lower layer. The incident ray
is refracted more and more strongly as it
passes from layer to layer and moves closer
and closer to the vertical line. Applying
Snells law to the interfaces between the
layers, we get
sin cz.1 _ sinotz __ sin ons _ sin oa,
U1 _ V2 _ va _ V4
Now let us assume that the layer thick-
nesses decrease without limit while the num-
ber of layers increases without limit. Then,
in the limit, the velocity of light changes
(decreases) continuously and we conclude
84 Differential Equaiions in Applications
Fig. 23
the light travels not along straight-line
segments.
For the sake of an example let us con-
sider Figure 23, which depicts a ray of light
propagating through a layered medium. In
each layer the speed of light is constant,
but it decreases when we pass from an up-
per layer to a lower layer. The incident ray
is refracted more and more strongly as it
passes from layer to layer and moves closer
and closer to the vertical line. Applying
Snell's law to the interfaces between the
layers, we get
Now let us assume that the layer thick-
nesses decrease without limit while the num-
ber of layers increases without limit. Then,
in the limit, the velocity of light changes
(decreases) continuously and we conclude Ch. 1. Construction of Differential Models 85
\ I
\ OC
~/1
",.-.... -..-..
i \
| x
\
Fig. 24
(soe Figure 24) that
@2:% (7;)
U
with a = const. A similar situation is ob-
served (with certain reservations) when a
ray of Sun light falls on Earth. As the ray
travels through Eartl1s atmosphere of in-
creasing density, its velocity decreases and
the ray bends.
Let us go back to the brachistochrono
problem. We introduce a system of coor-
dinates in the vertical plane in the way
shown in Figure 25. We imagine that a
ball (like a ray of light propagating in me-
dia) is capable of choosing the path of de-
scent from point/1 to pointB with the short-
est possible time of descent. Then, as fol-
Fig. 24
(see Figure 24) that
sin a
-v-=a,
(71)
with a = const. A similar situation is ob-
served (with certain reservations) when a
ray of Sun light falls on Earth. As the ray
travels through Earth's atmosphere of in-
creasing density, its velocity decreases and
the ray bends.
Let us go back to the brachistochrone
problem. We introduce a system of coor-
dinates in the vertical plane in the way
shown in Figure 25. We imagine that a
ball (like a ray of light propagating in me-
dia) is capable of choosing the path of de-
scent from point A to pointE with the short-
(lllt possible time of dellcent. Then, as fol-86 Differential Equations in Applications
A
I w
n
1
I
{H
x I
\ n
n
I Ot \\\
I \ B
Fig. 25
lows from the above reasoning, formula (71)
is valid.
The energy conservation principle implies
that the speed gained by the ball at a
given level depends only on the loss of po-
tential energy as the ball reaches the level
and not on the shape of the trajectory fol-
lowed. This means that v= l/2gy.
On the other hand, geometric construc-
tion enables us to show that
SlHG;COSB Z
SGGB 1/1+ta|$
....__L___
1/4+<y*>2
Combining the last two relationships with
(71) yields
E/[1 +(y')21== C (72)
Fig. 25
lows from the above reasoning, formula (71)
is valid.
The energy conservation principle implies
that the speed gained by the ball at a
given level depends only on the loss of po-
tential energy as the ball reaches the level
and not on the shape of the trajectory fol-
lowed. This means that v = lf2gy.
On the other hand, geometric construc-
tion enables us to show that
A 1 1
SlniX=COSp =--A-=,;
sect' v 1 +tanl ~
1
- J/t + (y')2
Combining the last two relationships with
(71) yields
y [1 + (y')
2
) :;:: c. (72) Ch. i. Construction of Differential Models 87
This constitutes the brachistochrone equa-
tion. We wish to demonstrate that only a
cycloid can represent a brachistochrone.
Indeed, since y' = dy/dx, by dividing the
variables in Eq. (72) we arrive at the equa-
tion
_ y l/2
dx .- (C__y ) dy.
We introduce a new variable cp via the
following relationship:
y */2...
( C__y ) -tan cp.
Then
y=.-Csinz cp, dy=2C sinqpcoscpdcp,
dx:-.tancpdy=2Csin2cpdcp
= C (1 cos 2cp) dep.
Integration of the last equation yields
sc == g-(2cpsin 2<p)lC,,
where, in view of the initial conditions,
:c=y=0 at cp:-0 and C1:-O. Thus,
sz == g (2cp sin 2q>),
y=-Csin2cp=-g(1cos 2q>).
Assuming that C/2 = r and 2rp = 9, we
arrive at the standard parametric equations
Ch. f. Construction of Differential Models 87
This constitutes the brachistochrone equa-
tion. We wish to demonstrate that only a
cycloid can represent a brachistochrone.
Indeed, since y' = dyldx, by dividing the
variables in Eq. (72) we arrive at the equa-
tion
dx = ( C Y Y ) 1/2 dy.
We introduce a new variable cp via the
(
y ) 1/2
C-y =tan cp.
Then
y = C sin2 cp, dy = 2C sin cp cos cp dcp,
dx =tan cp dy = 2C sin
2
cp dcp
= c (1- cos 2cp) dcp.
Integration of the last equation yields
X= ; (2cp- sin 2cp) + C1,
where, in view of the initial conditions,
X = y = 0 at cp = 0 and cl = 0. Thus,
x = ; (2cp- sin 2cp),
y=Csin2cp= ~ (1-cos2cp).
Assuming that C/2 = r and 2p = a, we
arrive at the standard parametric equations 88 Differential Equations in Applications
of the cycloid (69). The cycloid is, indeed, a
remarkable curve: it is not only isochro-
nousit is brachistochronous.
1.15 The Arithmetic Mean, the
Geometric Mean, and the
Associated Differential Equation
Let us consider the following curious prob-
lem first suggested by the famous German
mathematician Carl F. Gauss (1777-1855).
Let ml, and nl, be two arbitrary positive
numbers (ml, > 77.0). Out of ml, and nl, we
construct two new numbers ml and nl that
are, respectively, the arithmetic mean and
the geometric mean of ml, and nl,. ln other
words, we put
mii , ni; Vm0n0
Treating ml and nl in the same manner as
ml, and nl,, we put
mz :2%, nz: l/mlnl.
Continuing this process indefinitely, we ar-
rive at two sequences of real numbers, {mh}
and {nh} (lc = O, 1, 2, .), which, as can
easily be demonstrated, are convergent. We
wish to know the difference of the limits of
these two sequences.
An elegant solution of this problem
amounting to setting up a second-order linear
differential equation is given below. It be-
of the cycloid (69}. The cycloid is, indeed, a
remarkable curve: it is not only isochro-
nous-it is brachistochronous.
1.15 The Arithmetic Mean, the
Geometric Mean, and the
Associated Differential Equation
Let us consider the following curious prob-
lem first suggested by the famous German
mathematician Carl F. Gauss (1777-1855}.
Let m
0
and n
0
be two arbitrary positive
numbers (m
0
> n
0
}. Out of m
0
and n
0
we
construct two new numbers m
1
and n
1
that
are, respectively, the arithmetic mean and
the geometric mean of m
0
and n
0
In other
words, we put
mi _ mo-;t-n; , _ n, V rp,.,nn,; y
Treating m
1
and n
1
in the same manner as
m
0
and n
0
, we put
m1+n1 -v--
mz = 2 ' nz= mini.
Continuing this process indefinitely, we ar-
rive at two sequences of real numbers, {m,.}
and {n,.} (k = 0, 1, 2, .}, which, as can
easily be demonstrated, are convergent. We
wish to know the difference of the limits of
these two sequences.
An elegant solution of this problem
amounting to setting up a second-order linear
differential equation is given pclow. It be-Ch. 1. Construction of Differential Models 89
longs to the German mathematician Carl
W. Borchardt (1817-1880). Let a be the
sought difference. It obviously depends on
ml, and nl,, a fact expressed analytically as
follows: cz =f (ml,, nl,),where f is a func-
tion. The definition of a also implies that
a =f(ml, nl). Now, if we multiply ml,
and nl, by the same number k, each of the
numbers ml, 721, ml, nl, introduced
above, including a, will be multiplied by
k. This means that a is a firstorder homo-
geneous function in ml, and nl, and, hence,
a 2 mo}: (is no/mo) : m1]((1 nl/m1)'
Introducing tlie notation x = nl,/ml,, xl =
ni/my U Z 1/lc (L no/m0) yl =
1/f (1, nl/ml), we find that
__ m __ 2yl
y-- y1"`""""-`-`.1_l_x (73)
Since xl is related to sc via the equation
__ 2 i/5
$1* "W? i
we find that
dm _ 1= __<r=>(1+)
dn: _` (1+x)2l/Q _ 2(.rx3) '
On the other hand, Eq. (73) leads to the
dy __ ___ 2 2 dyl dxl
:-_` (1l:c)2 yl-l- 1-}-.2: dxl dw '
longs to the German mathematician Carl
W. Borchardt (1817 -1880). Let a be the
sought difference. It obviously depends on
m
0
and n
0
, a fact expressed analytically as
follows: a = f (m
0
, n
0
), where f is a fune-
tion. The definition of a also implies that
a = f (m
1
, n
1
). Now, if we multiply m
0
and n
0
by the same number k, each of the
numbers m
1
, n
1
, m
2
, n
2
, introduced
above, including a, will be multiplied by
k. This means that a is a first-order homo-
geneous function in m
0
and n
0
and, hence,
a = m
0
/ (1, n
0
1m
0
) = md (1, n
1
1m
1
).
Introducing the notation x = n
0
1m
0
, x
1
=
n
1
1m
1
, ., y = 1/f (1, n
0
fm
0
), Y
1
=
1/f (1, n
1
1m
1
), we find that
(73)
Since x
1
is related to x via the equation
2 vx
xt = t+x '
we find that
dx
1
1-x

(IX= (1+x)2 Jfx = 2 (:r--x
3
)
On the other hand, Eq. (73) leads to the
dy 2 + 2 dy1 dx1
=- (1+x)2 Y1 t+x dxl di"' 90 Differential Equations in Applications
Replacing dxl/dx with its value given by
the previous relationship and factoring out
(x x3), we {ind that
dy_2:v(xi)
<-x> zswr;ry1
d
+<4+x><=l$>,.
Difierentiating both sides of this equation
with respect to x, we {ind that
d [<r>%]
dx
[ x(a:1) ]
_2 d 1+x +2** ($*1) dl/1 d1
* yi dx 1+x dxl dx
d
+<xi~i>$
d d d
dx]. (1:;:1 '
By elementary transformation this equa-
tion can be reduced to
d dy
x[<xx3>aal*y
_ ix _d_ ___ _3 dy; _
_(1+x) {dx1[(xi 11) dxl] xiyi}
If we now replace x with xl, then xl will
become xg. If we then replace xl with xg,
Replacing dx
1
/dx with its value given by
the previous relationship and factoring out
(x- r}, we find that
(
S) dy_2x(x-1)
x-x dx- t+x Yt
+(1+x}
ddy
1

xl
Differentiating both sides of this equation
with respect to x, we find that
d [<x-x3)* J
dx
[
x(x-1) J
_
2
d i+x +2x(x-1) dy
1
dx1
- Yt dx 1+x dx1 dx
+ (xt- ddyl
xl
+<1+x} d:
1
J
By elementary transformation this equa-
tion can be reduced to
_i_[(x-x
3
} dy] -xy
dx dx
1-x { d [ { s} dy1 J }
(1+xrV.X dxl dxt -XtYt
If we now replace x with x
1
, then x
1
will
become x
2
If we then replace xl with x
2
, Ch. 1. Construction of Differential Models 91
we find that :::2 is transformed into x3, etc.
Hence, assuming that
d dy ..
g;,,[(-wa) ;,;]<y-a* (y>
we arrive at the following formula:
as, (y): 1--as 1--.1: 1x2
<1+=>1/5Z<<+=i> V?{(i+2> Vi?
1-:1:,, *
>< >< -; a y ).
<1+xn>1/xi (
If we now send rz to infinity, we find that
1 - x,, tends to zero and, hence,
* (y) = 0-
This means that y satisfies the differential
equation
(:1: -.2:3) -3.752) %Q;;xy=O. (74)
If we note that
2
=r<m., mi)-=y ---- ,; " .
0
we can easily find the value of this num-
ber. Indeed, since y must be the constant
solution to Eq. (74), we find that only
y E O is such a solution. Thus, the differ-
ence of the limits of the sequences {mn} and
{nk} is zero.
The differential equation (74) is remark-
able not only because it enabled reducing
we find that x
2
is transformed into x
3
, etc.
Hence, assuming that
( (x- x
3
) xy =a* (y),
we arrive at the following formula:
a* (y) = 1-x 1-x 1-x2
(i+x) Vx (1+xl) Vx1 (1+x
2
) Vi;"
X 1-Xn * ( )
X (i+xn) J.fxn a Yn.
If we now send n to infinity, we find that
1 - Xn tends to zero and, hence,
a* (y) = 0.
This means that y satisfies the differential
equation
d2y dy
(x-x
3
) dxt +(1-3x
2
) dx -xy=O. (74)
If we note that
we can easily find the value of this num-
ber. Indeed, since y must be the constant
solution to Eq. (74), we find that only
y == 0 is such a solution. Thus, the differ-
ence of the limits of the sequences and
{n
11
} is zero.
The differential equation (74) is remark-
not only because it enabled reducing 92 Differential Equations in Applications
the initial problem to an obvious one but
also because it is linked directly to the so-
lution of the problem of calculating the pe-
riod of small oscillations of a circular pen-
dulum.
As demonstrated earlier, the period of
small oscillations of a circular pendulum
can be found from the formula
T = 4]/UQF (lc, at/2),
with
arr/2 d
F ky 2 -T S ,
( n/) 0 l/1k2sin2cp
It has been found that if 0 { lc < 1, then
sr/2
S `
0 1/1-kzsinzqn
__:m OO i2><32><5><...><(2n1)2
?l1l-E1 2X42><6X ...><(2f1) k2n)
1],:*.
where
_ OO 12><32><52><...><(2n1)2 n
y-1+21 22><4><62>< X(2)2 Z2
7],:*.
is a solution to the differential equation (74).
the initial problem to an obvious one but
also because it is linked directly to the so-
lution of the problem of calculating the pe-
riod of small oscillations of a circular pen-
dulum.
As demonstrated earlier, the period of
small oscillations of a circular pendulum
can be found from the formula
T = 4 "Vllg F (k, n/2),
with
n/2
F (k, n/2) =
d!p
0
It has been found that if 0 =::;;; k < 1, then
where
00
y=1+ x2n
n=t
is a ;solutiofl to the 1iiffe:re11tial equation (74). Ch. 1. Construction of Differential Models 93
1.16 On the Flight of an Object
Thrown at an Angle to the Horizon
Suppose that an object is thrown at an an-
gle cx to the horizon with an initial velocity
vo. We wish to derive the equation of the
objects motion that ignores forces of
friction (air drag).
Let us select the coordinate axes as shown
in Figure 26. At an arbitrary point of the
trajectory only the force of gravity P
equal to mg, with m the mass of the object
and g the acceleration of gravity, acts on
the object. Hence, in accordance with
Newtons second law, we can write the
differential equations of the motion of the
object as projected on the :4: and y axes as
follows:
d2 d2
mq;-:0, m%:mg.
Factoring out m yields
daz ___ day _
wO W- *5* (75)
y __.,:=;.Et="
m
v
va tm.?
0 A .7.
Fig. 26
Ch. L Construction of Differential Models 93
1.16 On the Flight of an Object
Thrown at an Angle to the Horizon
Suppose that an object is thrown at an an-
gle a to the horizon with an initial velocity
v
0
We wish to derive the equation of the
object's motion that ignores forces of
friction (air drag).
Let us select the coordinate axes as shown
in Figure 26. At an arbitrary point of the
trajectory only the force of gravity P
equal to mg, with m the mass of the object
and g the acceleration of gravity, acts on
the object. Hence, in accordance with
Newton's second law, we can write the
differential equations of the motion of the
object as projected on the x and y axes as
follows:
d2y
m'"""'dti"=-mg.
Factoring out m yields
d
1
x dy
dt1 =O, dt1 =-g.
Fig. 26
The initial conditions imposed on the ob-
jects motion are
.CC=O, y=-0, ?;:-UOCOSOC,
t
Tgzvosinoz. at t-=O.
Integrating Eqs. (75) and allowing for the
initial conditions (76), we find that the
equations of the objects motion are
sc = (vo cos ot) 23, y = (vo sin cx) t gtz/2.
(77)
A number of conclusions concerning the
character of the objects motion can be
drawn from Eqs. (77). For example, we can
find the time of the objects flight up to the
moment when the object hits Earth, the range
of the flight, the maximum height that the
object reaches in its flight, and the shape
of the trajectory.
The first problem can be solved by finding
the value of time t at which y = 0. The
second equation in (77) implies that this
happens when
t[v0 since--]=O,
that is, either t = O or t = (2120 sin oc)/g.
The second value provides the solution.
The second problem can be solved by
calculating the value of x at a value of t
equal to the time of flight. The first equa-
The initial conditions imposed on the ob-
ject's motion are
dx
x=O, y=O, dt=v
0
cosa,
~ ~ =v
0
sinct at t=O.
(76)
Integrating Eqs. (75) and allowing for the
initial conditions (76), we find that the
equations of the object's motion are
x = (v
0
cos ct} t, y = (v
0
sin ct} t - gt
2
12.
(77)
A number of conclusions concerning the
character of the object's motion can be
drawn from Eqs. (77). For example, we can
find the time of the object's flight up to the
moment when the object hits Earth, the range
of the flight, the maximum height that the
object reaches in its flight, and the shape
of the trajectory.
The first problem can be solved by finding
the value of time t at which y = 0. The
second equation in (77) implies that this
happens when
t [ v
0
sin a- g ~ J = 0,
that is, either t = 0 or t = (2v
0
sin a)/g.
The second value provides the solution.
The second problem can be solved by
calculating the value of x at a value of t
equal to the time of flight. The first equa-Ch. 1. Construction of Differential Models 95
tion in (77) implies that the range of the
flight is given by the formula
(vo cos ct) (2v,, sin oc) _ 1:% sin 20:.
8* ___ E
This implies, for one thing, that the range
is greatest when 2ot == 90, or ot = !i5 In
this case the range is vg/g.
The solution to the third problem can be
obtained immediately by formulating the
maximum condition for y. But this means
that at the point where y is maximal the
derivative dy/dt vanishes. Noting that
fg-: gt|v0 sin oc,
we arrive at the equation -gt + vo sin oc =O,
which yields
t = (vo sin ot)/y.
Substituting this value of t into the second
equation in (77), we find that the maximum
height reached by the object is v sinz on/2g.
The solution to the fourth problem has
already been found. Namely, the trajectory
is represented by a parabola since Eqs. (77)
represent parametrically a parabola, which
in rectangular Cartesian coordinates can
be written as follows:
y= x tan oc- gi? seczct.
tion in (77) implies that the range of the
flight is given by the formula
(v
0
cos a) (2v
0
sin a)
g
v ~ sin 2a
g
This implies, for one thing, that the range
is greatest when 2a = 90, or a = 45 In
this case the range is v:lg.
The solution to the third problem can be
obtained immediately by formulating the
maximum condition for y. But this means
that at the point where y is maximal the
derivative dy/dt vanishes. Noting that
dy t+ .
dt= -g V
0
sma,
we arrive at the equation -gt + v
0
sin a =0,
which yields
t = (v
0
sin a)ly.
Substituting this value of t into the second
equation in (77), we find that the maximum
height reached by the object is v ~ sin
2
a/2g.
The solution to the fourth problem has
already been found. Namely, the trajectory
is represented by a parabola since Eqs. (77)
represent parametrically a parabola, which
in rectangular Cartesian coordinates can
be written as follows:
ll
y=xtana- P
2
, sec2a.
Vo 96 Differential Equations in Applications
1.17 Weightlessness
The state of weightlessness (zero g) can be
achieved in various ways, although it is
associated (consciously or subconsciously)
with the floating" of astronauts in the
cabin of a spacecraft.
Suppose that a person of weight P is stand-
ing in an elevator that is moving down-
ward with an acceleration co = osg, with
O<ot<1 and g the acceleration of grav-
ity. Let us determine the pressure that
the person exerts on the cabins floor and
the acceleration that the elevator must
undergo so that this pressure will vanish.
Two forces act on the person in the eleva-
tor (Figure 27): the force of gravity P and
the force Q that the floor exerts on the per-
Q
P
.2*
Fig. 27
1.17 Weightlessness
The state of weightlessness (zero g) can be
achieved in various ways, although it is
associated (consciously or subconsciously)
with the "floating" of astronauts in the
cabin of a spacecraft.
Suppose that a person of weight P is stand-
ing in an elevator that is moving down-
ward with an acceleration (J) = a.g, with
0<a.<1 and g the acceleration of grav-
ity. Let us determine the pressure that
the person exerts on the cabin's floor and
the acceleration that the elevator must
undergo so that this pressure will vanish.
Two forces act on the person in the eleva-
tor (Figure 27): the force of gravity P and
the force Q that the floor exerts on the per-
a.
p
a:
Fig. 27 Ch. 1. Construction of Differential Models 97
son (equal numerically to the force of pres-
sure of the person on the floor). The differ-
ential equation of the persons motion can
be written in the form
dzx
mq?--PQ. (78)
Since d20:/dtz =-to = ug and m = P/g, we
can rewrite Eq. (78) as follows:
d2
Q:.-PmTi-:P(1ot). (79)
Since 0 < ot < 1, we conclude that Q < P.
Thus, the pressure that the person exerts on
the floor of the cabin of an elevator mov-
ing downward is determined by the force
Q = P (l OL).
On the other hand, when the elevator is mov-
ing upward with an acceleration to = osg,
O < oc < 1, the pressure that the person
exerts on the floor of the cabin is deter-
mined by the force Q = P (1 -{ ot). Let us
now establish at what acceleration the pres-
sure vanishes. For this it is sufficient to put
Q = O in (79). We conclude that in this
case ot = 1, that is, for Q to vanish the
acceleration of the elevator must be equal
to the acceleration of gravity.
Thus, when the cabin is falling freely
with an acceleration equal to g, the pres-
sure that the person exerts on the floor is
nil. lt is this state that is called weightless-
7-0770
son (equal numerically to the force of pres-
sure of the person on the floor). The differ-
ential equation of the person's motion can
d2x
mdt2=P-Q.
(78)
Since d
2
x/dt
2
= oo = a.g and m = Pig, we
can rewrite Eq. (78) as follows:
d2x
Q=P-mdt2=P(1-a.). (79)
Since 0 <a.< 1, we conclude that Q < P.
Thus, the pressure that the person exerts on
the floor of the cabin of an elevator mov-
ing downward is determined by the force
Q = p (1- a.).
On the other hand, when the elevator is mov-
ing upward with an acceleration oo = a.g,
0 < a. < 1, the pressure that the person
exerts on the floor of the cabin is deter-
mined by the force Q = P (1 + a.). Let us
now establish at what acceleration the pres-
sure vanishes. For this it is sufficient to put
Q = 0 in (79). We conclude that in this
case a. = 1, that is, for Q to vanish the
acceleration of the elevator must be equal
to the acceleration of gravity.
Thus, when the cabin is falling freely
with an acceleration equal to g, the pres-
sure that the person exerts on the floor is
nil. It is this state that is called zveightless-
ness. In it the various parts of a persons
body exert no pressure on each other, so
that the person experiences extraordinary
sensations. In the state of weightlessness
all points of an object experience the same
acceleration.
Of course, weightlessness is experienced
not only during a free fall in an elevator.
For illustration let us consider the following
problem.
What must be the speed of a spacecraft
moving around the Earth as an artificial
satellite for a person inside it to be in the
state of weightlessness?
One assumption in this problem is that
the spacecraft follows a circular orbit of
radius r + h, where r is Earths radius, and
h is the altitude at which the spacecraft
travels (reckoned from Earths surface). The
previous problem implies that in the state
of weightlessness the pressure on the walls
of the spacecraft is zero, so that the force Q
acting on an object inside the spacecraft
is zero, too. Hence, Q == O. Let us now
turn our attention to Figure 28. The as axis
is directed along the principal normal rz
to the circular trajectory of the spacecraft.
We use the differential equation of the mo-
tion of a particle as projected on the prin-
cipal normal:
TL
mvz
T: E Frm,
h=l
ness. In it the various parts of a person's
body exert no pressure on each other, so
that the person experiences extraordinary
sensations. In the state of weightlessness
all points of an object experience the same
acceleration.
Of course, weightlessness is experienced
not only during a free fall in an elevator.
For illustration let us consider the following
problem.
What must be the speed of a spacecraft
moving around the Earth as an artificial
satellite for a person inside it to be in the
state of weightlessness?
One assumption in this problem is that
the spacecraft follows a circular orbit of
radius r + h, where r is Earth's radius, and
h is the altitude at which the spacecraft
travels (reckoned from Earth's surface). The
previous problem implies that in the state
of weightlessness the pressure on the walls
of the spacecraft is zero, so that the force Q
acting on an object inside the spacecraft
is zero, too. Hence, Q = 0. Let us now
turn our attention to Figure 28. The x axis
is directed along the principal normal n
to the circular trajectory of the spacecraft.
We use the differential equation of the mo-
tion of a particle as projected on the prin-
cipal normal:
n
mv2 "
-p-= LJ F ~ ~ . n ,
1<=1 Ch. 1. Construction of Differential Models 99
Q
(
F
H $
K
.2:
Fig. 28
Tl.
where p :r\h, E FM =-F, and F is
1;:1
directed along the principal normal to the
trajectory. We get
mv? dz.1:
T;@F : mrmn
or the equation
dzx __ v2
dig r-|h
Substituting this value of dzx/dig into
Eq. (78), we find that
mv?
7_-{-:1)-Q. (80)
n
X
Fig. 28
n
where p = r + h, ~ Fkn = F, and F is
k=i
directed along the principal normal to the
trajectory. We get
~ = F = ' m d2x
r+h dt2 '
or the equation
d2x v2
""'dt2- r+h
Substituting this value of d
2
x/dt
2
into
Here force P is equal to the force F of
attraction to Earth, which, in accordance
with Newtons law of gravitation, is inverse-
ly proportional to the square of the dis-
tance r -\~h from the center of Earth,
that is,
km
F " (r+h)2
where m is the mass of the spacecraft, and
constant lc can be determined from the
following considerations. At Earths sur-
face, where h = 0, the force of gravity F
is equal to mg. The above formula then
yields lc = gr? Hence,
p Z F: Jeri.
(r+h>
where g is the acceleration of gravity at
Earths surface.
lf we now substitute the obtained value
of P into (80) and note that Q = 0, we find
that the required speed is given by the
formula
-- 8
U " " l/ r+h
1.18 Keplers Laws of Planetary
Motion
In accordance with Newtons law of gravi-
tation, any two objects separated by a dis-
tance r and having masses rn and M are at-
Here force P is equal to the force F of
attraction to Earth, which, in accordance
with Newton's law of gravitation, is inverse-
ly proportional to the square of the dis-
tance r + h from the center of Earth,
that is,
F
where m is the mass of the spacecraft, and
constant k can be determined from the
following considerations. At Earth's sur-
face, where h = 0, the force of gravity F
is equal to mg. The above formula then
yields k = gr
2
Hence,
mgr2
p = F= (r+h)2 '
where g is the acceleration of gravity at
Earth's surface.
If we now substitute the obtained value
of P into (80) and note that Q = 0, we find
that the required speed is given by the
formula
v = rv ~ h
1.18 Kepler's Laws of Planetary
Motion
In accordance with Newton's law of gravi-
tation, any two objects separated by a dis-
tance r and having masses m and M are at-Ch. 1. Construction of Differential Models 101
y \\V
/*005 gv m
I \ _
F ___ FSL/70
// I
w'
/ ry
/ I
M / I
7
0 .2* P ar
Fig. 29
tracted with a force
GmM
F :71 (81)
where G is the gravitational constant.
Basing ourselves on this law, let us de-
scribe the motion ofthe planets in the solar
system assuming that m is the mass of a
planet orbiting the Sun and M is the
Suns mass. The effect of other planets on
this motion will be ignored.
Let the Sun be at the origin of the coor-
dinate system depicted in Figure 29 and
the planet be at time t at a point with
running coordinates x and y. The attractive
force F acting on the planet can be decom-
posed into two components: one parallel to
the :1: axis and equal to F cos cp and the
other parallel to the y axis and equal to
y
./
f/
/
I
./
ly
./
I
./
!!'
I
.z;
p .z;
Fig. 29
tracted with a force
F = G ~ M , (81)
where G is the gravitational constant.
Basing ourselves on this law, let us de-
scribe the motion of the planets in the solar
system assuming that m is the mass of a
planet orbiting the Sun and M is the
Sun's mass. The effect of other planets on
this motion will be ignored.
Let the Sun be at the origin of the coor-
dinate system depicted in Figure 29 and
the planet be at time t at a point with
running coordinates x and y. The attractive
force F acting on the planet can be dtlcom-
poscd into two components: one parallel to
the x axis and equal to F cos cp and the
other parallel to the y axis and equal to 102 Differential Equations in Applications
F sin rp. Using formula (81) and Newtons
second law, we arrive at the following
equations:
most: -Fcos cp = ?cos cp, (82)
my: Fsinq>:I}%sin cp. (83)
Bearing in mind that sin cp = y/r and
cos cp = .2:/r, we can rewrite Eqs. (82) and
(83) as follows:
k k
x Z ;E;;` U Z _ jig/
where constant It is equal to GM.
Finally, allowing for the fact that r =
l/x2 -Iy2, we arrive at the differential
equations
T- i
x-_ (x2_|_y2)3/2 9 y __' l-(x2_*_y2)3/2
(84)
Without loss of generality we can assume
that
x:a,y=-O,x=O,g)=v0att=O. (85)
We see that the problem has been re-
duced to the study of Eq. (84) under the
initial conditions (85). The special fea-
tures of Eqs. (84) suggest that the most con-
venient coordinate system here is the one
using polar coordinates: x = r cos cp and
F sin cp. Using formula (81) and Newton's
second law, we arrive at the following
equations:
GmM
mx = - F cos (p = - -r-
2
- cos cp, (82)
. GmM
my= -Fsmcp=--r
2
-smcp. (83)
Bearing in mind that sin cp = ylr and
cos cp = xlr, we can rewrite Eqs. (82) and
(83) as follows:
kx ky
X=--,:a, y=-ra,
where constant k is equal to GM.
Finally, allowing for the fact that r =
V x
2
+ y
2
, we arrive at the differential
equations
kx ky
Y = - (x2+ y2)3f2
(84)
Without loss of generality we can assume
that
. .
x =a, y = 0, x = 0, y = v
0
at t = 0. (85)
We see that the problem has been re-
duced to the study of Eq. (84) under the
initial conditions (85). The special fea-
tures of Eqs. (84) suggest that the most con-
venient coordinate system here is the one
using polar coordinates: x = r cos q> and Ch. 1. Construction of Diiierential Models 103
y = r sin q>. Then
dzzrooscp (rsinrp) tp,
Q: rsincp + (rcoscp)q>,
QJ: rooos cp - 2 ($ sm tp) gb (se)
(r sin cp) q; (r oos cp) qlz,
[;=I:SlI1 cp | 2 cos cp)
+ (r cos cp) (r sin cp) qaz.
Hence,
dz.: rlpz) eos qu (2rtp { sin cp,
Q.; (lTq32)SlI1 (P + + COS (P.
Using the last two relationships, we can
rewrite the differential equations (84) in
the form
(or.-- rep?) cos cp - (Zrrp + sin cp
= <87>
(7:rcl>2) sin cp -|- (Zrco-I- cos rp
Z .- . (88)
Multiplying both sides of Eq. (87) by
cos cp, both sides of Eq. (88) by sin ep, and
y = r sin cp. Then
. . .
x = r cos cp - (r sin fP) cp,
. . .
y = r sin cp + (r cos cp) cp,
. .
x = r cos cp - 2 (r sin cp) cp
(86)
- (r sin cp) fP - (r cos cp) ~
2
,
. .
y = r sin cp + 2 (r cos cp) cp
.. .
+ (r cos cp) {p - (r sin cp) cp
2

Hence,
. . . . . . .
x = (r - rcp
2
) cos cp - (2r cp + r cp) sin cp,
. . . . . . . . .
y = (r - rcp
2
) sin cp + (2r cp + r cp) cos cp.
Using the last two relationships, we can
rewrite the differential equations (84) in
the form
. . . ..
(r- np
2
) cos cp- (2rcp + r cp) sin cp
k cos cp
r2
( r- np2) sin cp + (2rcp + r cp) cos cp
k sin <p
=- r2
(87)
(88)
cos cp, both sides of Eq. (88) by sin cp, and 104 Differential Equations in Applications
adding the products, we find that
ro- rqlz = -k/rz. (89)
sin cp, both sides of Eq. (88) by cos cp, and
subtracting the second product from the
first, we arrive at the equation
2rqi + riii Z 0. (eo)
As for the initial conditions (85), in polar
coordinates they assume the form
(91)
Thus, we have reduced the problem of
studying Eqs. (84) under the initial condi-
tions (85) to that of studying Eqs. (89) and
(90) under the initial conditions (91). We
also note that Eq. (90) can be rewritten in
the form
d
-H-(rzqn) = 0. (92)
But Eq. (92) yields
rah - C., <93>
where C1 is a constant possessing an in-
teresting geometric interpretation. Precise-
ly, suppose that an object moves from
point P to point Q along the arc PQ (Fig-
ure 30). Let S be the area of the sector limit-
ed by the segments OP and OQ and the
adding the products, we find that
;_ ~
2
= -klr
2
(89)
sin p, both sides of Eq. (88) by cos p, and
subtracting the second product from the
first, we arrive at the equation
..
2rp + rp = 0. (90)
As for the initial conditions (85), in polar
coordinates they assume the form
r = a, p = 0, ; = 0, ~ = v
0
/a at t = 0.
(91)
Thus, we have reduced the problem of
studying Eqs. (84) under the initial condi-
tions (85) to that of studying Eqs. (89) and
(90) under the initial conditions (91). We
also note that Eq. (90) can be rewritten in
the form
d
dt r 2 ~ p ) = 0. (92)
But Eq. (92) yields
r
2
p = cl, (93)
where C
1
is a constant possessing an in-
teresting geometric interpretation. Precise-
ly, suppose that an object moves from
point P to point Q along the arc PQ (Fig-
ure 30). LetS be the area of the sector limit-
ed by the segments 0 P and OQ and the Ch. 1. Construction of Differential Models 105
3
Cl
{
0
P .2
Fig. 30
arc From the calculus course we know
tl1at
cp
S--i rzd
... 2 (P,
0
or dS = (1/2) rz dcp. Hence,
dS _1 2 dcp _ 1 2
7F* 2 " at ?r " (94)
The derivative dS/dt constitutes what is
known as the areal velocity, and since in
view of (93) rzqn is a constant, we conclude
that the areal velocity is a constant, too.
But this, in turn, means that the object
moves in such a manner that the radius
vector describes equal areas in equal time
intervals. This law of areas constitutes one
of the three Kepler laws. In full it can be
formulated as follows: each planet moves
along a plane curve around the Sun in such
a manner that the radius vector connecting
!I
a
Fig. 30
arc PQ. From the calculus course we lmow
that
1 r
8=
2
J r
2
dqJ,
0
or dS = (1/2) r
2
dqJ. Hence,
dS 1 d ~ 1
dt=T r2 dt =2 r2qJ
(94)
The derivative dS/dt constitutes what is
known as the areal velocity, and since in
view of (93) r
2
qJ is a constant, we conclude
that the areal velocity is a constant, too.
But this, in turn, means that the object
moves in such a manner that the radius
vector describes equal areas in equal time
intervals. This law of areas constitutes one
of the three Kepler laws. In full it can be
formulated as follows: each planet moves
along a plane curve around the Sun in such
a manner that the radius rector connecting 106 Differential Equations in Applications
the S un with the planet describes equal areas
in equal time intervals.
To derive the next Kepler law, which
deals with the shape of the planetary tra-
jectories, we return to Eqs. (89) and (90)
with the initial conditions (91) imposed on
them. The initial conditions imply, for
one, that r = a and cp : vo/a at t = O.
But then condition (93) implies that C1 =
avo. Hence,
rzcp = avo, or rp = avg/rz. (95)
This transforms Eq. (89) into
a2v2 lc
':%"Tr
Assuming that : p, we can rewrite this
equation in the form
-rE;- Q- evt _ k
dt_dr dt _p dr_ r3 F
or
dp __ a2v _ k
Pw"_r W
Separating the variables in the last diffe-
rential equation and integrating, we arrive
at the following relationship:
TM r w+C2
Since p = : O at r : a, we {ind that
v2 lc
C-ijT
the Sun with the planet describes equal areas
in equal time intervals.
To derive the next Kepler law, which
deals with the shape of the planetary tra-
jectories, we return to Eqs. (89) and (90)
with the initial conditions (91) imposed on
them. The initial conditions imply, for
one, that r = a and ~ = v
0
/a at t = 0.
But then condition (93) implies that C
1
=
av
0
Hence,
r
2
~ = av
0
, or ;p = av
0
/r
2

This transforms Eq. (89) into
a
2
v ~ k
r = ---,:3 - ---;:2
(95)
Assuming that r = p, we can rewrite this
equation in the form
ie_ _ iE._ ~ _ .!!E._ _ a 2 v ~ _ ...!!____
dt - dr dt - p dr - ,.a ,.z '
or
~ _ a 2 v ~ _ _!:__
p dr - ,.a r2
Separating the variables in the last diffe-
rential equation and integrating, we arrive
at the following relationship:
p2 k a2v2
2=7- 2r2o +Cz.
Since p = r = 0 at r = a, we fmd that
C
_ 3 _ _ _ ~
z- 2 a Ch. 1. Construction of Differential Models 107
Thus, we arrive at the equation
Lz ____ k azvz v lc
T- T r 22%+*: tv
or, if we consider only the positive value
of the square root,
dr ___ l/ 2k 2]:: a2v2
ar- ("<T)+T; (99
Dividing Eq. (96) by Eq. (95), we hud that
%-"L`7 -I/.(XJ2l2Br1,
where
__ 1 ___ 2k B_ lc
Q--? a3v a2v,
The last equation can be integrated by sub-
stituting 1/u for r. The result is
T: a,2v/lc
1+<>S(<>+63>
where e = il/cx + [32/B = avg/lc - 1. The
constant C3 can be determined from the
condition that r = a at cp = O. It is easy
to verify that C3 = O. Thus, we finally have
_ a2v/lc
rv 1+ecos<p (97)
From analytic geometry we know that this
is the equation of a conic in terms of polar
coordinates, with e the eccentricity of the
conic. The following cases are possible
Thus, we arrive at the equation
r
2
k a2v2 v
2
k
- "+ 0
y--,:-- 2r2 2-a,
or, if we consider only Lhe positive value
of the square root,
= -. / ( V _ 2k ) + 2k _
dt V o a ,. r2
(96)
Dividing Eq. (96) by Eq. (95), we fmd that
=r V o:rLi-2Pr-1,
where
1 2k k
0: = i2 - a3 v2 ' p = a2v1
0 0
The last equation can be integrated by sub-
stituting 1/u for r. The result is
a
2
v
2
/k
r- o
- 1+ecos(cp+C
3
) '
where e = V o: +
= - 1. The
constant C
3
can be determined from the
condition that r = a at cp = 0. It is easy
to verify that C
3
= 0. Thus, we finally have
a
2
v
2
/k
r = 1 + e rp (97)
From analytic geometry we know that this
is the equation of a conic in terms of polar
coordinates, with e the eccentricity of the
conic. The following cases are possible 108 Differential Equations in Applications
here:
(1) an ellipse if e<1, or vg<2k/cz,
(2) a hyperbola if e > 1, or vg > 2k/a,
(3) a parabola if e : 1, or vg = 2k/a,
(4) a circle if e = O, or vg = lc/o.
Astronomical observations have shown
that for all the planets belonging to the
solar system the value of vg is smaller
han 2l.c/a. We, therefore, arrive at another
of Keplers laws: the planets describe el-
lipses with the Sun at one focus.
Note that the orbits of the Moon and
the artificial satellites of Earth are also
ellipses, but in the majority of cases these
ellipses are close to circles, that is, e differs
little from zero.
As for recurrent comets, like, say, Hal-
leys comet, their orbits resemble "prolate"
ellipses whose eccentricity is smaller than
unity but very close to it. Say, Halleys
comet appears in Earths neighborhood
approximately every 76 years. Its latest
apparition was in the period between the
end of 1985 and the beginning of 1986.
Celestial bodies that move along para-
bolic and hyperbolic orbits may be observed
only once, since they never return to
the same place.
Let us now establish the physical mean-
ing of eccentricity e. First we note that
the components x and y of a planets vel-
ocity vector Valong the as and y axes,_respec-
here:
(1) an ellipse if e < 1, or v ~ < 2kla,
(2) a hyperbola if e > 1, or v ~ > 2kla,
(3) a parabola if e = 1, or v ~ = 2kla,
(4) a circle if e = 0, or v ~ = lela.
Astronomical observations have shown
that for all the planets belonging to the
solar system the value of v: is smaller
han 2kla. We, therefore, arrive at another
of Kepler's laws: the planets describe el-
lipses with the Sun at one focus.
Note that the orbits of the Moon and
the artificial satellites of Earth are also
ellipses, but in the majority of cases these
ellipses are close to circles, that is, e differs
little from zero.
As for recurrent comets, like, say, Hal-
ley's comet, their orbits resemble "prolate"
ellipses whose eccentricity is smaller than
unity but very close to it. Say, Halley's
comet appears in Earth's neighborhood
approximately every 76 years. Its latest
apparition was in the period between the
end of 1985 and the beginning of 1986.
Celestial bodies that move along para-
bolic and hyperbolic orbits may be observed
only once, since they never return to
the same place.
Let us now establish the physical mean-
ing of eccentricity e. First we note that
the components ~ and y of a planet's vel-
ocity vector V along the x andy axes,_respec-Ch. 1. Construction of Differential Models 109
tively, and the size v of vector V satisfy
the relationship
U2 I WL Q2,
which when combined with (86) yields
w:#&+h
From this it follows that the kinetic energy
of a planet of mass m is given by the for-
mula
% mvz = - m (rzcpz + rz). (98)
The potential energy of a system is minus
one times the amount of work needed to
move the planet to infinity (where the po-
tential energy is zero by convention).
Hence,
OO km km <> km
T
lf by E we denote the total energy of the
system, which in view of energy conserva-
tion must be constant, then formulas (98)
and (99) yield
1 2.2 2 km
Assuming that cp = 0 and combining (97)
with (100), we get
__ azvz/lc mr2a2v2 _ km __
' E
tively, and the size v of vector V satisfy
the relationship
vz = ;2 + y2,
which when combined with (86) yields
. .
v2 = r2<p2 + r2.
From this it follows that the kinetic energy
of a planet of mass m is given by the for-
mula
1 1 .
2
mv
2
=
2
m (r21Jl
2
+ r2). (98)
The potential energy of a system is minus
one times the amount of work needed to
move the planet to infinity (where the po-
tential energy is zero by convention).
Hence,
00
_ \ km dr = km j"" = _ km
J r
2
r r r
(99)
r
If by E we denote the total energy of the
system, which in view of energy conserva-
tion must be constant, then formulas (98)
and (99) yield
1 km
2
m [r
2
c:p
2
+ r
2
] --r-=E. (100)
Assuming that <p = 0 and combining (97)
with (100), we get
a2v2Jk mr2a2v2 km
r= 1-t.e ' 2r'
0
- -r-=E. 110 Differential Equations in Applications
Excluding r from the last two relationships,
we {ind that
P 2a2v
" Z l/1 +L nm
Equation (97) for the shape of the orbit
finally assumes the form
T: azvg/lc
1+ ]/1{E (2a2v/mkz) cos cp
This formula implies that the orbit is an
ellipse, hyperboba, parabola, or circle if,
respectively, E < O, E >O, E = O, or
E = mk2/2a,2v. Thus, the shape of a
planets orbit is completely determined by
the value of E. Say, if we could impart
such a "blow" to Earth that it would increase
Earths total energy to a positive value,
Earth would go over to a hyperbolic orbit
and leave the solar system forever.
Let us now turn to Keplers third law.
This law deals with the period of revolu-
tion of planets around the Sun. Taking
into account the results obtained in deriv-
ing the previous Kepler law, we naturally
restrict our discussion to the case of ellip-
tic orbits, whose equation in terms of Carte-
sian coordinates, as is known, is
:1:2 12
@+%:1*
Excluding r from the last two relationships,
we find that
Equation (97) for the shape of the orbit
finally assumes the form

r =
This formula implies that the orbit is an
ellipse, hyperboba, parabola, or circle if,
respectively, E < 0, E > 0, E = 0, or
E =
Thus, the shape of a

planet's orbit is completely determined by
the value of E. Say, if we could impart
such a "blow" to Earth that it would increase
Earth's total energy to a positive value,
Earth would go over to a hyperbolic orbit
and leave the solar system forever.
Let us now turn to Kepler's third law.
This law deals with the period of revolu-
tion of planets around the Sun. Taking
into account the results obtained in deriv-
ing the previous Kepler law, we naturally
restrict our discussion to the case of ellip-
tic orbits, whose equation in terms of Carte-
sian coordinates, as is known, is Ch. 1. Construction of Differential Models 111
.7
A
C F J2
g I
Y' {
Fig. 31
where (Figure 31) the eccentricity e = C/E,
with C2 = E,2 n2, so that
e2 = (E2 n2)/E2,
or
n2 = E2 (1 B2)- (101)
Combining this with (97) and allowing for
the properties of an ellipse, we conclude
that
___ 1 a2v2/k a2v2/k __ a2v2
1-lge + 1-3:2 )_lc(1-0422)
2v3E
I M12 v
or
nz: 1%%.*; _ (102)
Ch. 1. Construction of Differential Models Hi
!J
Fig. 31
where (Figure 31) the eccentricity e = Cis
with C
2
= '6
2
- fj
2
, so that
e2 = (;2 - 'Yj2)fs2'
or
'Yj2 = s2 (1 - e2).
(101)
Combining this with (97) and allowing for
the properties of an ellipse, we conclude
that
s=_!_ { + ) =
2 1 + e 1 - e lc ( 1-e2)

or
2

'Yj=-k-.
Let us denote the period of revolution of a
planet by T. By definition, T is the time
it takes the planet to complete one full or-
bit about the Sun. Then, since the area lim-
ited by an ellipse is Jtgn, we conclude, on
the basis of (94) and (95), that atn ==
av0T/2. Finally, taking into account (102),
we arrive at the following result:
4;-{,22n2 43-[2
T2: .a.:g:T
This constitutes the analytical description
of Keplers third law: the squares of the
periods of revolution of the planets are pro
portional to the cubes of the major axes of the
planets orbits.
1.19 Beam Deflection
Let us consider a horizontal beam AB
(Figure 32) of a constant cross section and
made of a homogeneous material. The
symmetry axis of the beam is indicated in
Figure 32 by a dashed line. Suppose that
forces acting on the beam in the vertical
plane containing the symmetry axis bend
the beam as shown in Figure 33. These forces
may be the weight of the beam itself or
an external force or the two forces acting
simultaneously. Clearly, the symmetry axis
will also bend due to the action of these
forces. Usually a bent symmetry axis is
called the elastic line. The problem of deter-
H2 Differential Equations in Applications
Let us denote the period of revolution of a
planet by T. By definition, T is the time
it takes the planet to complete one full or-
bit about the Sun. Then, since the area lim-
ited by an ellipse is we conclude, on
the basis of (94) and (95), that =
av
0
T /2. Finally, taking into account (102),
we arrive at the following result:
T2 - 4n2s2TJ2 - 4n2
- - k
This constitutes the analytical description
of Kepler's third law: the squares of the
periods of revolution of the planets are pro-
portional to the cubes of the major axes of the
planets' orbits.
1.19 Beam Deflection
Let us consider a horizontal beam AB
(Figure 32) of a constant cross section and
made of a homogeneous material. The
symmetry axis of the beam is indicated in
Figure 32 by a dashed line. Suppose that
forces acting on the beam in the vertical
plane containing the symmetry axis bend
the beam as shown in Figure 33. These forces
may be the weight of the beam itself or
an external force or the two forces acting
simultaneously. Clearly, the symmetry axis
will also bend due to the action of these
forces. Usually a bent symmetry axis is
called the elastic line. The problem of deter-Ch. 1. Construction of Differential Models 113
mining the shape of this line plays an im-
portant role in elasticity theory.
Note that there can be various types of
beam depending on the way in which beams
are fixed or supported. For example, Fig-
ure 34 depicts a beam whose end A is rigidly
fixed and end B is free. Such a beam is said
to be a cantilever. Figure 35 depicts a beam
lying freely on supports A and B. Another
type of beam with supports is shown in
Figure 36. There are also various ways in
which loads can be applied to beams. For
example, a uniformly distributed load is
shown in Figure 34. Of course, the load can
vary along the entire length of the beam
or only a part of this length (Figure 35).
Figure 36 illustrates the case of a concen-
trated load.
Let us consider a horizontal beam OA
(Figure 37). Suppose that its symmetry
axis (the dashed line in Figure 37) lies on
the :1: axis, with the positive direction
being to the right of point O, the origin.
The positive direction of the y axis is
downward from point O. External forces
F1, F2, (and the weight of the beam
if this is great) bend the symmetry axis,
which becomes the elastic line (also depict-
ed in Figure 38 by a curved dashed line).
The deflection y of the elastic line from the
III axis is known as the sag of the beam at
point x. Thus, if we know the equation of
the elastic line, we can always find the sag
8-0770
mining the shape of this line plays an im-
portant role in elasticity theory.
Nate that there can be various types of
beam depending on the way in which beams
are fi.xed or supported. For example, Fig-
ure 34 depicts a beam whose end A is rigidly
fi.xed and end B is free. Such a beam is said
to be a cantilever. Figure 35 depicts a beam
lying freely on supports A and B. Another
type of beam with supports is shown in
Figure 36. There are also various ways in
which loads can be applied to beams. For
example, a uniformly distributed load is
shown in Figure 34. Of course, the load can
vary along the entire length of the beam
or only a part of this length (Figure 35).
Figure 36 illustrates the case of a concen-
trated load.
Let us consider a horizontal beam OA
(Figure 37). Suppose that its symmetry
axis (the dashed line in Figure 37) lies on
the x axis, with the positive direction
being to the right of point 0, the origin.
The positive direction of the y axis is
downward from point 0. External forces
F
1
, F
2
, (and the weight of the beam
if this is great) bend the symmetry axis,
which becomes the elastic line (also depict-
ed in Figure 38 by a curved dashed line).
The deflection y of the elastic line from the
x axis is known as the sag of the beam at
point x. Thus, if we know the equation of
the elastic line, we can always fi.nd the sag
A m/# m#//1///0/////100// B
Fig. 32
A B
%""Ill!.."/M
Fig. 33
Al B
Fig. 34
of the beam. Below we show how this can
be done in practical terms.
Let us denote by M (x) the bending mo-
ment in the cross section of the beam at
coordinate .2:. The bending moment is de-
fined as the algebraic sum of the moments
of the forces that act from one side of the
beam at point x. In calculating the mo-
ments we assume that the forces acting on
the beam upward result in negative mo-
ments while those acting downward result
in positive moments.
Fig. 32
A
B
Fig. 33
A ~ HHI:HH*\8
~
Fig. 34
of the beam. Below we show how this can
be done in practical terms.
Let us denote by M (x) the bending mo-
ment in the cross section of the beam at
coordinate x. The bending moment is de-
fined as the algebraic sum of the moments
of the forces that act from one side of the
beam at point x. In calculating the mo-
ments we assume that the forces acting on
the beam upward result in negative mo-
ments while those acting downward result
in positive moments. Ch, 1. Construction of Diherential Models 145
P cz
Fig. 35
_ P
A B
s
Fig. 36
F F M
1 fg 5 (Z`) G
0 A
a:
JI}
.9
Fig. 37
The strength-ofmaterials course proves
that the bending moment at point x is re-
lated to the curvature radius of the elastic
HDB via HIE 9ql1&tiOIl
EJ .4.:M , 103
u+<y >=1=x= (x) ( )
where E is Youngs modulus, which de-
8*
Ch. 1. Construction of Differential Models U:i
i t i l !
Al 18
f'
a
Fig. 35
Ac
!p
Ia
~
2S
Fig. 36
F,
'Z
~ 11(x) F.,
t l
~ iC
J
i
..
-- I
:c
:c
Fig. 37
The strength-of-materials course proves
that the bending moment at point x is re-
lated to the curvature radius of the elastic
line via the equation
y"
EJ
11
+ (y')BJSfB M (x), (103)
where E is Y oung'.s modulus, which de-
5
F
0 3 fl;
wl F5 W
`4
y (D
Fig. 38
pends on the type of material of the beam,
J is the moment of inertia of the cross sec-
tion of the beam at point as about the hori-
zonta-1 straight line passing through the
center of mass of the cross section. The
product EJ is commonly known as the
flexural rigidity, in what follows we assume
this product constant.
Now, if we suppose that the sag of the
beam is small, which is usually the case in
practice, the slope y' of the elastic line is
extremely small and, therefore, instead of
Eq. (103) we can take the approximate
equation
EJ y" = M (:1:). (104)
To illustrate how Eq. (104) is used in
practice, we consider the following prob-
lem. A horizontal homogeneous steel beam
of length l lies freely on two supports and
H6 Differential Equations in Applications
A
9
Fig. 38
pends on the type of material of the beam,
J is the moment of inertia of the cross sec-
tion of the beam at point x about the hori-
zonta-l straight line passing through the
center of mass of the cross section. The
product EJ is commonly known as the
flexural rigidity, in what follows we assume
this product constant.
Now, if we suppose that the sag of the
beam is small, which is usually the case in
practice, the slope y' of the elastic line is
extremely small and, therefore, instead of
Eq. (103) we can take the approximate
equation
Ely"= M (x). (104)
To illustrate how Eq. (104) is used in
practice, we consider the following prob-
lem. A horizontal homogeneous steel beam
of length l lies freely on two supports and Ch. 1. Construction of Differential Models 117
pi/2 pz/2
L-a:
0 X A
` __________ ,....-- .2:
1 Q I
9 *` ,<&~v}
Fig. 39
sags under its own weight, which is p kgf
per unit length. We wish to determine the
equation of the elastic line and the maxi-
mal sag.
In Figure 39 the elastic line is depicted
by the dashed curve. Since we are dealing
with a two-support beam, each support acts
on the beam with an upward reaction force
equal to half the weight of the beam (or
pl/2). The bending moment M (sc) is the
algebraic sum of the moments of these forces
acting on one side from point Q (Fig-
ure 39). Let us first consider the action of
forces to the left of point Q. At a distance x
from point Q a force equal to pl/2 acts on
the beam upward and generates a negative
moment. On the other hand, a force equal
to px acts on the beam downward at a
distance .1/2 from point Q and generates a
positive moment. Thus, the net bending
pl/2
:;
Fig. 39
sags under its own weight, which is p kgf
per unit length. We wish to determine the
equation of the elastic line and the maxi-
mal sag.
In Figure 39 the elastic line is depicted
by the dashed curve. Since we are dealing
with a two-support beam, each support acts
on the beam with an upward reaction force
equal to half the weight of the beam (or
pl/2). The bending moment M (x) is the
algebraic sum of the moments of these forces
acting on one side from point Q (Fig-
ure 39). Let us fi.rst consider the action of
forces to the left of point Q. At a distance x
from point Q a force equal to pl/2 acts on
the beam upward and generates a negative
moment. On the other hand, a force equal
to px acts on the beam downward at a
distance x/2 from point Q and generates a
positive moment. Thus, the net bending 118 Differential Equations in Applications
moment at point Q is given by the formula
z 2 1
Mw: Lx+Px(%)=f%%
(105)
If we consider the forces acting to the
right of point Q, we find that a force equal
to p (l :1;) acts on the beam downward at
a distance (l -- sz:)/2 from point Q and gen-
erates a positive moment, while a force
equal to pl/2 acts on the beam upward at a
distance l az from point Q and generates
a negative moment. The net bending mo-
ment in this case is calculated by the for-
mula
M(x>=p (i>i?L<ix>
..r.2=i-E2
... 2 .2 (106)
Formulas (105) and (106) show that the
bending- moments prove to be equal. Now,
knowing how to find a bending moment, we
can easily write the basic equation (104),
which in our case assumes the form
Ely": Jl? -%E (107)
Since the beam does not bend at points O
and A, we write the boundary conditions
for Eq. (107) so as to {ind y:
y=Oatx=Oand.py=O,atx:l. ..
1t8 Differential Equations in Applications
moment at point Q is given by the formula
M (
pl ( x ) px2 plx
x)= -yx+px 2 =-2---2-
(105)
If we consider the forces acting to the
right of point Q, we find that a force equal
top (l- x) acts on the beam downward. at
a distance (l - x)/2 from point Q and gen-
erates a positive moment, while a force
equal to pl/2 acts on the beam upward at a
distance l - x from point Q and generates
a negative moment. The net bending mo-
ment in this case is calculated by the for-
mula
l-x pl
M(x)=p (l-x)-
2
--
2
(l-x)
px2 plx
=-
2
--T (106)
Formulas (105) and (106) show that the
bending moments prove to be equal. Now,
knowing how to find a bending moment, we
can easily write the basic equation (104),
which in our case assumes the form
Ely"= px2 - plx
2 2
(107)
Since the beam does not bend at points 0
and A, we write the boundary conditions
for Eq. (107) so as to find y:
y = 0 at x = 0 and y = 0 at x -:- l. Ch. 1. Construction of Differential Models 119
Integration of Eq. (107) then yields
y = @1% (xi -2Zx3| Pcs). (108)
This constitutes the equation of the elastic
line. It is used to calculate the maximal
sag. For instance, in this example, basing
our reasoning on symmetry considerations
(the same can be done via direct calcula-
tions), we conclude that the maximal sag
will occur at :2: = Z/2 and is equal to
5pl*/384E./`, with E =-21 >< 105 kgf/cm2
and J = 3 >< 104 cm4.
In transporting logs to saw mills, logging
trucks move along forest roads some of the
time. The width of the forest road is usual-
ly such that only one truck can travel along
the road.Sections of the road are made wid-
er so that trucks can pass each other. Ignor-
ing the question of how traffic should be
arranged that loaded and empty traffic
trucks meet only at such sections, let us
establish how wide the turns in the road
must be and what trajectory the driver
must try to follow so that, say, thirty-me-
ter logs can be transported. It is assumed
that the truck is sufficiently maneuverable
to cope with a limited section of the road.
Usually a logging truck consists of a
tractor unit and a trailer connected freely
Ch. 1. Construction of Differential Models H9
Integration of Eq. (107) then yields
y= ;EI (x
4
-2lx3+Px). (108)
This constitutes the equation of the elastic
line. It is used to calculate the maximal
sag. For instance, in this example, basing
our reasoning on symmetry considerations
(the same can be done via direct calcula-
tions), we conclude that the maximal sag
will occur at x = l/2 and is equal to
5pl
4
/384EJ, with E = 21 x 10
6
kgf/cm
2
and J = 3 X 10
4
cm
4

In transporting logs to saw mills, logging
trucks move along forest roads some of the
time. The width of the forest road is usual-
ly such that only one truck can travel along
the road. Sections of the road are made wid-
er so that trucks can pass each other. Ignor-
ing the question of how traffic should be
arrangsd that loaded and empty traffic
trucks meet only at such sections, let us
establish how wide the turns in the road
must be and what trajectory the driver
must try to follow so that, say, thirty-me-
ter logs can be transported. It is assumed
that the truck is suffiCiently maneuverable
to cope with a limited section of the road.
Usually a logging truck consists of a
tractor unit and a trailer connected freely 120 Differential Equations in Applications
S ,9 C Q P
X B A Y
s R Q P A. F
Fig. 40
to each other. The tractor unit has a front
(driving) axle and two back axles above
which a round platform carrying two posts
and a rotating beam are fastened. This
platform can rotate in the horizontal plane
about a symmetrically positioned vertical
axis. One end of each log is placed on this
platform. The trailer has only two back
axles, but also has a platform with two
posts. The other ends of the logs are put on
this platform. The trailers chassis consists
of two metal cylinders one of which can
partly move inside the other. The chassis
connects the back platform with the axis
that links the trailer with the tractor unit.
Thus, the length of the chassis can change
during motion, which enables the tractor
unit and the trailer to move independently
to a certain extent. Schematically the log-
ging truck is depicted in Figure 40. The
points A and B correspond to the axes of
the platforms a distance h apart. By XY
we denote a log for which AX : kh. Point
C corresponds to a small axis that connects
X
R c a
s ~ P
: V ~ ~
s R P F
F
Fig. 40
to each other. The tractor unit has a front
(driving) axle and two back axles above
which a round platform carrying two posts
and a rotating beam are fastened. This
platform can rotate in the horizontal plane
about a symmetrically positioned vertical
axis. One end of each log is placed on this
platform. The trailer has only two back
axles, but also has a platform with two
posts. The other ends of the logs are put on
this platform. The trailer's chassis consists
of two metal cylinders one of which can
partly move inside the other. The chassis
connects the back platform with the axis
that links the trailer with the tractor unit.
Thus, the length of the chassis can change
during motion, which enables the tractor
unit and the trailer to move independently
to a certain extent. Schematically the log-
ging truck is depicted in Figure 40. The
points A and B correspond to the axes of
the platforms a distance h apart. By XY
we denote a log for which AX =')..h. Point
C corresponds to a small axis that connects Ch. 1. Construction of Differential Models 121
the tractor unit with the trailer, with
AC=ah. Usually a-:0.3, but in the
simplest case of log transportation a = O.
Next, FF is the front axle of the tractor
unit, and PP and QQ are the back axles of
the tractor unit, while RR and SS are
the trailer axles. All axles have the same
length 2L, so that for the sake of simplic-
ity we assume that the width of the log-
ging truck is 2L. As for the width of the
load in its rear, we put it equal to 2W.
In what follows we will need the concept
of the sweep of the logging truck, which is
commonly understood to be the maximum
deflection of the rear part of the logging
truck (for the sake of simplicity we assume
this part to be point X) from the trajectory
along which the logging truck moves. Let
us suppose that the road has a width of
2[iih and that usually a turn in the road is
simply an arc of a circle of radius h/cz
centered at point O (Figure 41). For the
sake of simplicity we assume that a logging
truck enters a turn in the road in such a
way that the tractor unit and the trailer
are positioned along a single straight line,
the driver operates the truck in such a way
that point A, corresponding to the axis of
the front platform, is exactly above the
roads center line. Point A in Figure 41
is determined by the angle X that the truck
AC makes with the initial direction. Here
it is convenient to fix a coordinate system
the tractor unit with the trailer, with
AC = ah. Usually a = 0.3, but in the
simplest case of log transportation a = 0.
Next, FF is the front axle of the tractor
unit, and P P and QQ are the back axles of
the tractor unit, while RR and SS are
the trailer axles. All axles have the same
length 2, so that for the sake of simplic-
ity we assume that the width of the log-
ging truck is 2. As for the width of the
load in its rear, we put it equal to 2W.
In what follows we will need the concept
of the sweep of the logging truck, which is
commonly understood to be the maximum
deflection of the rear part of the logging
truck (for the sake of simplicity we assume
this part to be point X) from the trajectory
along which the logging truck moves. Let
us suppose that the road has a width of
2 ~ h and that usually a turn in the road is
simply an arc of a circle of radius h/a.
centered at point 0 (Figure 41). For the
sake of simplicity we assume that a logging
truck enters a turn in the road in such a
way that the tractor unit and the trailer
are positioned along a single straight line,
the driver operates the truck in such a way
that point A, corresponding to the axis of
the front platform, is exactly above the
road's center line. Point A in Figure 41
is determined by the angle 'X. that the truck
AC makes with the initial direction. Here
it is convenient to tix a coordinate system 122 Differential Equations in Applications
X
\\\\
Zrzifial __ g_8 \_\____
dirccfian '/ \\
// \
/ 0
x
x
y // // A
/// //// \
// /`/// \
/ /.// \
x /
0 .z
Fig. 41
Oxy in such a way that the horizontal axis
points in the initial direction and the ver-
tical axis is perpendicular to it. In a general
situation the load carried by the truck
will make an angle 6 with the initial direc-
tion. As for angle BAC in Figure 41,
denoting it by u we {ind that u, = X 9.
Usually this angle is the logging trucks
angle of lag. The required halfwidth h of
the road, which determines the sweep of the
logging truck at a turn and is known as the
halfwidth of the road at the outer curb of a
Initial
y
0
Fig. 41
Oxy in such a way that the horizontal axis
points in the initial direction and the ver-
tical axis is perpendicular to it. In a general
situation the load carried by the trucl{
will make an angle 8 with the initial direc-
tion. As for angle BAG in Figure 41,
denoting it by u we find that u = 'X - e.
Usually this angle is the logging truck's
angle of lag. The required halfwidth h of
the road, which determines the sweep of the
logging truck at a turn and is known as the
halfwidth of the road at the outer curb of a Ch. 1. Construction of Differential Models 123
turn, is defined as the algebraic sum
OX OA l W, while the halfwidth of the
road at the inner curb of a turn is defined
as the algebraic sum OA + L OP, where
OP is the perpendicular dropped from
point O onto AB.
We stipulate that in its motion a logging
trucks wheel either experiences no lateral
skidding at all or the skidding is small.
This requirement, for one thing, means that
the center line AC of the tractor unit con-
stitutes a tangent to the arc of a circle at
point A, so that OA is perpendicular to AC,
and angle X is determined by the motion of
point A along the arc of the circle. Note,
further, that in building a road the curva-
ture of a turn in the road is determined by
an angle N that corresponds to an arc
length of a turn of approximately 30 m.
In our notation,
O i80 30u
N ;.:TT- , (109)
where h is measured in meters. Thus, at
h = 9 m and ot -:0.1 we have N z 19,
while at h = 12 m and cx = 1.0 we have
N := 142. For practical considerations we
must consider only such cxs that lie be-
tween 0 and 1, and the greater the value of
oc, the greater the maneuverability of the
logging truck.
The log length 7th will be greater than h,
but again, reasoning on practical grounds,
turn, is defined as the algebraic sum
OX- OA + W, while the halfwidth of the
road at the inner curb of a turn is defined
as the algebraic sum OA +L-OP, where
OP is the perpendicular dropped from
point 0 onto AB.
We stipulate that in its motion a logging
truck's wheel either experiences no lateral
skidding at all or the skidding is small.
This requirement, for one thing, means that
the center line AC of the tractor unit con-
stitutes a tangent to the arc of a circle at
point A, so that OA is perpendicular to AC,
and angle r., is determined by the motion of
point A along the arc of the circle. Note,
further, that in building a road the curva-
ture of a turn in the road is determined by
an angle N that corresponds to an arc
length of a turn of approximately 30 m.
In our notation,
N"- = 180 30a.
n h '
(109)
where h is measured in meters. Thus, at
h = 9 m and a= 0.1 we have N ~ 19,
while at h = 12 m and a = 1.0 we have
N ~ 142. For practical considerations we
must consider only such a's that lie be-
tween 0 and 1, and the greater the value of
a, the greater the maneuverability of the
logging truck.
The log length 'Ah will be greater than h,
but again, reasoning on practical grounds, 124 Differential Equations in Applications
it must not be greater than 3h. Thus, the
value of 2. varies between 1 and 3. As for
constant a, it is assumed that 0 { a < 0.5.
Finally, we note that in each case the
value of h is chosen differently, but it varies
between 9 and 12 m.
Since the wheels of the tractor unit do
not skid laterally, the coordinates of point
A in Figure 41 are
x= EsinX, y:-.cosX.
a ex
The coordinates of point B are
X=;sin Xhcos 9,
Y: cosX +hsin0. (110)
Since the trailers wheels do not skid either,
point B moves in the direction BC and
dY
-ai-..:tanxp, (111)
where mp is the angle which BC makes with
the initial direction. Next, employing the
fact that X = u. + 0 and studying triangle
ABC, we arrive at the following chain of
equalities:
sin (X1|>) ___ sin (01|>) __ sin u.
r"*;T*-Th (**2)
where 0<b<1, and xp, 0, and u are
it must not be greater than 3h. Thus, the
value of A. varies between 1 and 3. As for
constant a, it is assumed that 0 a < 0.5.
Finally, we note that in each case the
value of h is chosen differently, but it varies
between 9 and 12 m.
Since the wheels of the tractor unit do
not skid laterally, the coordinates of point
A in Figure 41 are
h . X h X
x = a sm , y =a cos .
The coordinates of point B are
X-hcosa,
a;
Y= +hsin6.
a;
(110)
Since the trailer's wheels do not skid either,
point B moves in the direction BC and
dY
dX =-tan lJl, (111)
where lJl is the angle which BC makes with
the initial direction. Next, employing the
fact that X = u + 6 and studying triangle
ABC, we arrive at the following chain of
equalities:
sin (X-'!') _ sin (6-'IJ) _ sin u (112)
h ah
where 0 < b < 1, and lJl, 6, and u are Ch. 1. Construction of Differential Models 125
functions of X. If we combine (111) with
(110), we find that
(-Q-S1nx+z, %3cos9)cos1p
It dB . .
{-(-Ex-cosX+hT1T(-s1n9)s1n1p-O.
Carrying out the necessary calculations, W9
arrive at
sin(X1p)== on gcos(91p), (113)
If we exclude variable 142 from this relation-
ship fora fixed value of a by employing (112)
and the fact that X =- u l B, we arrive
(since 9 : 0 at X = 0) at the differential
equation
2.- -...22;... 114
dx _1 0t(1ac0Su) ( )
with the initial condition u (0) :. O, where
the angle of lag plays the role of the sought
function.* _
By substituting v for tan (u/2) in the
differential equation (114), we can mtegrate
the equation in closed form. However, the
resulting u vs. X dependence proves to be
extremely complicated, which, of course,
hinders an effective study. Nevertheless,
Eq. (114) can easily be integrated and
studied numerically. To this end we can
* See A.B. Tayler, "The sweep of a logging truck",
Math. Spectrum 7, No. 1: 19-26 (1974/75).
functions of X If we combine (111) with
(110), we find that
( - ~ sin X+ h : ~ cos fl) cos 'IJl
+ ( cos X + h : ~ sin fl) sin 'IJl = 0.
Carrying out the necessary calculations, we
arrive at
sin(X-'IJl) =ex ~ ~ cos(fi-'IJl). (113)
If we exclude variable 'IJl from this relation-
ship for a fixed value of a by employing (112)
and the fact that x = u + fl, we arrive
(since fl = 0 at x = 0) at the differential
equation
du sin u 4
df=
1
- a.(i-acosu) (
11
)
with the initial condition u (0) = 0, where
the angle of lag plays the role of the sought
function.*
By substituting v for tan (u/2) in the
differential equation (114), we can integrate
the equation in closed form. However, the
resulting u vs. x dependence proves to be
extremely complicated, which, of course,
hinders an effective study. Nevertheless,
Eq. (114) can easily be integrated and
studied numerically. To this end we can
See A.B. Tayler, "The sweep of a logging truck",
Math. Spectrum 1, No. 1: 19-26 (1974/75). 126 Differential Equations in Applications
employ, for example, the RungeKutta sec-
ondorder numerical method the essence of
which is the following.
Suppose that an integral curve u = cp (X)
of a differential equation du/dX = j (X, u)
passes through a point (X0, uc). At equidis-
tant points xm xp xw (xm ~ xi =
AX >0) we select values uc, ul, u2, .
such that ui x cp, (X), where the successive
values ul, uz, are specified by the
formulas
ui+1 Z ui l (ki l` k2)/2
with
ki = f (xi, ui) AX
k2 : ft (Xi JF AX ui l ki) AX-
To solve Eq. (114) numerically with the
initial condition u (0) == 0, we compile the
following program using BASIC:
10 REM Runge-Kutta second-order method
20 DEF FNF (X, U) =
1 SIN (U)/(AL=(1 A=COS(U)))
30 CLS: PRINT "Soluti0n of differential equa-
tion"
40 PRINT "RungeKutta second-order method"
100 PRINT "Parameters:"
110 INPUT "Alpha=", AL
120 INPUT "a=", A
130 INPUT "Step of independent V&I'I&bI=, DX
140 PRINT "Initial values:"
150 INPUT "of independent V3.II&bI6==, X
employ, for example, the Runge-Kutta sec-
ond-order numerical method the essence of
which is the following.
Suppose that an integral curve u = q> (x,)
of a differential equation du/dx, = f (x,, u)
passes through a point (x,
0
, u
0
). At equidis-
tant points X.o 'X.11 'X.2 ., (X.;+I- ')(.; =
A.x, > 0) we select values u
0
, u
11
u
2
,
such that u; ~ q>; (x,), where the successive
values u
1
, u
2
, are specifted by the
formulas
U;+I = U; + (kl + k2)/2,
with
kl = f (')(.11 U;) fl.')(.,
k2 = f (')(.; + fl.')(., U; + k1) fl.')(..
To solve Eq. (114) numerically with the
initial condition u (0) = 0, we compile the
following program using BASIC:
10 REM Runge-Kutta second-order method
20 DEF FNF (X, U) =
1 - SIN(U)/(AL(1 - ACOS(U)))
30 CLS: PRINT "Solution of differential equa-
tion"
40 PRINT "Runge-Kutta second-order method"
100 PRINT "Parameters:"
110 INPUT "Alpha=", AL
120 INPUT "a=", A
130 INPUT "Step of independent variable=", DX
140 PRINT "Initial values:"
150 INPUT "of independent variable=", X Ch. 1. Construction of Differential Models 127
160 INPUT of function=", U
200 BEM Next 20 values of function
210 CLS: PRINT "X", U"
220 FOR I :1 TO 20
230 PRINT X, U
240 K1 : DXFNF (X,U)
250 X = X "l DX
260 K2 = DX*FNF(X, U -{ K1)
270 U = U | (K1 + K2)/2
280 NEXT I
290 INPUT "Continue (Yes = 1, No = 0); I
300 IF I ( )0 GOTO 210
310 END
Note that to compile the table of values
of the solution for concrete values of on and
a, we must know the limiting value C of
this solution. The number C can be found
from the condition that
cx(1 acosC) =sinC,
which leads to the formula
C__Sin_1 ( Oi [1-a 1/1-q2.|.a2q,2])
1|a2cz.2
Here, if ex = 1, we have C = at/2-
2 tan1a, but if cx < 1, then C z ot (1 a).
The limiting value C can be approximated
(i by an exponential, namely
C u z eV%,
where y = oc (cos C - a)/sin2 C. For small
cxs the value of y is fairly great and is
Ch. 1. Construction of Differential Models 1.27
1.60 INPUT "of function=", U
200 REM Next 20 values of function
210 CLS: PRINT "X", "U"
220 FOR I = 1 TO 20
230 PRINT X, U
240 K1 = DXFNF (X,U)
250 X= X+ DX
260 K2 = DX*FNF(X, U + K1)
270 U = U + (K1 + K2)/2
280 NEXT I
290 INPUT "Continue (Yes= 1, No= 0)"; I
300 IF I ( ) 0 GOTO 210
31.0 END
Note that to compile the table of values
of the solution for concrete values of a and
a, we must know the limiting value C of
this solution. The number C can be found
from the condition that
a (1 - a cos C) = sin C,
which leads to the formula
C
- . -1 ( a. [1-a yf-a.B+asa.t) )
-Sin 1+a2a.2
Here, if a = 1, we have C = 'it/2-
2 tan-
1
a, but if a ~ 1, then C ~ a (1 - a).
The limiting value C can be approximated
by an exponential, namely
C- u ~ e--v:x.,
where y = a (cos C - a)/sin
2
C. For small
a's the value of y is fairly great and is 128 Differential Equations in Applications
approximately given by the relationship
/\J
YN @(1-a)
But if ot :1, we have
(1+2>
Y: 1a
For the step AX in the variation of the
independent variable X we must take a
number that does not exceed C. This require-
ment is especially important for small oas
and in the neighborhood of X = O. Below
we give the protocol for calculating the
values of the function u. = u (X).
Solution of differential equation by
Runge-Kutta second-order method
Parameters:
alpha = 1.0
a = 0.3
Step of independent variable = 0.5
Initial values:
of independent variable = 0
of function = 0
X u
0 0
.2 .1437181
_4 .2299778
.6 .2824244
.8 .3146894
1 .3347111
approximately given by the relationship
1
v ~ - - - : ~ " " " " " " 7 " "
ex (1-a)
But if a = 1, we have
a(1+a2)
"/= 1-aB
For the step 11'1. in the variation of the
independent variable '1. we must take a
number that does not exceed C. This require-
ment is especially important for small a's
and in the neighborhood of '1. = 0. Below
we give the protocol for calculating the
values of the function u = u ("f.).
Solution of differential equation by
Runge-Kutta second-order method
Parameters:
alpha= 1.0
a= 0.3
Step of independent variable = 0.5
Initial values:
of independent variable = 0
of function = 0
X
0
.2
,4
.6
.8
1
u
0
.1437181
.2299778
.2824244
.3146894
.3347111 Ch. 1. Construction of Differential Models 129
1.2 .3472088
1.4 .3550403
1.6 .35996
1.8 .3630556
2 .3650054
2.2 .3662343
2.4 .3670091
2.600001 .3674978
2.800001 .367806
3.000001 .3680005
3.200001 .3681232
3.400001 .3682006
3.600001 .3682494
3.800001 .3682802
Continue (Yes = 1, No = O)? 1
X u
4.000001 .3682997
4.200001 .368312
4.4 .3683197
4.6 .3683246
4.8 .3683276
5 .3683296
5.2 .3683308
5.399999 .3683316
5.599999 .3683321
5.799999 .3683324
5.999999 .3683326
6.199999 .3683327
6.399998 .3683328
6.599998 .3683328
6.799998 .3683329
6.999998 .3683329
9-0770
Gh. 1. Construction of Differential Models 129
1.2
1.4
1.6
1.8
2
2.2
2.4
2.600001
2.800001
3.000001
3.200001
3.400001
3.600001
3.800001
Continue (Yes = 1, No = 0)? 1
'X
4.000001
4.200001
4.4
4.6
4.8
5
5.2
5.399999
5.599999
5.799999
5.999999
6.199999
6.399998
6.599998
6.799998
6.999998
9-0770
.3472088
.3550403
.35996
.3630556
.3650054
.3662343
.3670091
.3674978
.367806
.3680005
.3681232
.3682006
.3682494
.3682802
u
.3682997
.368312
.3683197
.3683246
.3683276
.3683296
.3683308
.3683316
.3683321
.3683324
.3683326
.3683327
.3683328
.3683328
.3683329
.3683329 130 Differential Equations in Applications
7.199998 .3683329
7.399997 .3683329
7.599997 .3683329
7.799997 .3683329
Continue (Yes == 1, No = O)? 1
gg u
7.999997 .3683329
8.199997 .3683329
8.399997 .3683329
8.599997 .3683329
8.799996 .3683329
8.999996 .3683329
9.199996 .3683329
9.399996 .3683329
9.599996 .3683329
9.799996 .3683329
9.999995 .3683329
10.2 .3683329
10.4 .3683329
10.6 .3683329
10.8 .3683329
10.99999 .3683329
11 .19999 .3683329
11.39999 .3683329
11.59999 .3683329
1 1 . 79999 .3683329
Continue (Yes = 1, No = O)? 0
OK
The results of a numerical calculation at
a = O.3 are presented graphically in Fig-
ure 42. They show how an increase in X
7.199998
7.399997
7 . 5 ~ 9 9 9 7
7.799997
Continue (Yes= 1, No= 0)? 1
;.:
7.999997
8.199997
8.399997
8.599997
8.799996
8.999996
9.199996
9.399996
9.599996
9.799996
9.999995
10.2
10.4
10.6
10.8
10.99999
11.19999
11.39999
11.59999
11.79999
Continue (Yes= 1, No= 0)? 0
OK
.3683329
.3683329
.3683329
.3683329
u
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
.3683329
The results of a numerical calculation at
a = 0.3 are presented graphically in Fig-
ure 42. They show how an increase in 'X. Ch. 1. Construction of Differential Models 131
influences the angle of lag of the logging
truck. For clarity of exposition the scales
on the u and X axes are chosen to be differ-
ent.
Now let us determine the sweep of the
logging truck by using the concept of half-
width of the road at the outer curb of a turn
(this quantity was defined earlier as the al-
gebraic sum OX OA }W(Figure 41)).
First we note that
OX2 = (-gsin X9h cos 9)2
}(c0sX-|7tn sin6)2
=h2 (2%-1-}.2--2-g;-sin u) . -
This shows that the sweep decreases as X
grows since the angle of lag, u, increases.
Thus, the maximal halfwidth Bh of the
road can be determined from the following
formula for B:
1 1 W
B = I/2+w"tr+v7
In Figure 43 the solid curves reflect the
relationship that exists between B-- W/h
and oc for different values of Z., while the
dashed curves show what must be the
value of B - L/h to guarantee the necessary
"margin" at the inner curb of the turn. For
Qi
influences the angle of lag of the logging
truck. For clarity of exposition the scales
on the u and x axes are chosen to be diffet-
ent.
Now let us determine the sweep of the
logging truck by using the concept of half-
width of the road at the outer curb of a turu
(this quantity was defined earlier as the al-
gebraic sum OX - OA + W (Figure 41)).
First we note that
oxz = ( ~ s i n X-'J..hcos0)2
+ ( ; cos X + 'J..n sin 0)2
==.h2 { _!_+J..2-2 s i n u)
a
2
a
This shows that the sweep decreases as x
grows since the angle of lag, u, increases.
Thus, the maximal halfwidth ~ h of the
road can be determined from the following
formula for ~ :
V
1 1 w
R= 'J..2+---+-
t' a a h
In Figure 43 the solid curves reflect the
relationship that exists between ~ - W!h
and ex for different values of ').., while the
dashed curves show what must be the
value of ~ - Llh to guarantee the necessary
"mar.gin" at the inner curb of the turn. For
II* 132 Differential Equations in Applications
u
7 C`=CZ987?826
CX = 7
6'=(Z56853Z9
Of = Q 5
a=g_; C=QO70/372
0 50 X
Fig. 42
WL
/ 7 #" F
7 I /
AEJ ;d0
I
I
A =2 / {cc-= Q3
/ I
1
/ -
// /,z
,//
//'
,#
0 ` 7 Oc 0
Fig. 43
u
1
0
Fig. 42
Fig. 43
C=D.9878826
C=0.3683329
CmQOJOfJf2
50 X
L
,ft--
h
1
0 Ch. 1. Construction of Differential Models 133
every position of the logging truck on the
road we must have
[5:.- %-(1cosu)-|%
1 L
<;(1c0s C) +-5- (115)
Next, since the value of C decreases with a,
the angle of lag u proves to be the greatest
when a, = 0, which corresponds to the simpl-
est case of logging. Then (115) yields
_. .... 2
[1--2-<;(1-cos C)|a=0-.:2K6%2,
Now let us dwell on the results that follow
from the above line of reasoning. Here is a
typical example that illustrates these re-
sults.
If a logging truck in which the distance
between the front and back platforms is
12 m follows a turn along an arc of a circle
of radius 60 m, then oc = 0.2 and, in accor-
dance with (109) N is approximately 28.
If the width of the logging truck is 2.4 m
and width of the load in the rear is 1.2 m,
then for logs of approximately 24 m in
length (reckoning from the axis of the
front platform) we have X = 2, W/h =
0.05, and L/h = 0.1. Thus, from Figure 43
it follows that for all values of cz we have
B = 0.45 for the outer curb of the turn and
B = 0.2 for the inner curb. Theoretically
the necessary halfwidth of the road at the
every position of the logging truck on the
road we must have
P= ~ (1-cosu) +
< ~ (1-cos C)+ ~
(115)
Next, since the value of C decreases with a,
the angle of lag u proves to be the greatest
when a= 0, which corresponds to the simpl-
est case of logging. Then (115) yields
P-..<-
1
(1-cosC)I =t-vr=a:s.
h a. a=O a.
Now let us dwell on the results that follow
from the above line of reasoning. Here is a
typical example that illustrates these re-
sults.
If a logging truck in which the distance
between the front and back platforms is
12 m follows a turn along an arc of a circle
of radius 60 m, then a = 0.2 and, in accor-
dance with (109) N is approximately 28.
If the width of the logging truck is 2.4 m
and width of the load in the rear is 1.2 m,
then for logs of approximately 24 m in
length (reckoning from the axis of the
front platform) we have ). = 2, Wlh =
0.05, and Llh = 0.1. Thus, from Figure 43
it follows that for all values of a we have
P = 0.45 for the outer curb of the turn and
P = 0.2 for the inner curb. Theoretically
the necessary halfwidth of the road at the 134 Differential Equations in Applications
outer curb of the turn is equal to 5.4 m
and that at the inner curb to 2.4 m. If the
logging truck transports logs whose length
is 14.4 m (reckoned from the axis of the
front platform) and whose bunch width in
the rear is 1.8 m, then in the case at hand
Pt = 1.2. The value of [5 then proves to be
the same for the inner and outer curbs of
the turn and equal to 0.22. Hence, as can
easily be seen, the necessary halfwidth of
the road at the outer curb of the turn is
2.64 m, while the same quantity at the
inner curb, as in the previous case, is equal
to 2.4 mi. This reasoning shows that the
longer the logs transported the wider the
road at a turn must be. For one, if we com-
pare the two cases considered here, we see
that an increase in the length of the logs by
9.6 m requires widening the road at a turn
by 2.76 m so that the driver can drive the
truck along a curve whose length at the turn
is approximately equal to the length of the
roads center line. Practice has shown that
an inexperienced driver is not able to
drive his truck along such a curve and needs
a road whose total width at a turn is at
least 10.8 m (if the load transported is 24 m
long) for a truck width of 2.4 m.
The theory developed above shows that
the sweep of the logging truck is the great-
est when the truck enters the turn, since
in this case the angle of lag grows. This
conclusion also holds true in the situation
outer curb of the turn is equal to 5.4 m
and that at the inner curb to 2.4 m. If the
logging truck transports logs whose length
is 14.4 m (reckoned from the axis of the
front platform) and whose bunch width in
the rear is 1.8 m, then in the case at hand
f.. = 1. 2. The value of ~ then proves to be
the same for the inner and outer curbs of
the turn and equal to 0.22. Hence, as can
easily be seen, the necessary halfwidth of
the road at the outer c.urb of the turn is
2.64 m, while the same quantity at the
inner curb, as in the previous case, is equal
to 2.4 m. This reasoning shows that tho
longer the logs transported the wider the
road at a t.urn must be. For one, if we com-
pare the two cases considered here, we see
that an increase in the length of the logs by
9.6 m requires widening the road at a turn
by 2. 76 m so that the driver can drive the
truck along a curve whose length at the turn
is approximately equal to the length of the
road's center line. Practice has shown that
an inexperienced driver is not able to
drive his truck along such a curve and needs
a road whose total width at a turn is at
least 10.8 m (if the load transported is 24m
long) for a truck width of 2.4 m.
The theory developed above shows that
the sweep of the logging truck is the great-
est when the truck enters the turn, since
in this case the angle of lag grows. This
conclusion also holds true in the situation Ch. 1. Construction of Differential Models 135
when the truck enters one section of a zigzag
turn after completing the previous one at
the point of inflection. The results represent-
ed graphically in Figure 43 correspond to
the case where prior to entering a turn the
tractor unit and the trailer are positioned
on a single straight line. But if there exists
a nonzero initial angle of lag C0 caused by
the zigzag nature of the turn, we must select
the initial condition in the differential
equation (114) in the form u (0) == C0.
Then the required width of the road is
determined from the following formula for B:
rs: 1/ xw, $,-+2 gmc, --27.+ Q".
We note that in the case of simple log
transportation, that is, a-:-0, it is impossible
to pass a turn with on greater than unity.
However, for fairly large values of a the
values on >1 become possible, they must
obey the following relationship: ot (1
a cos C) = sin C. Thus, the maximal value
of on is (1 a2)**/2, with the practical
extremal value of ot being 1.25 at a == 0.5.
We note in conclusion that for ot > 0.5
considerable economy in the width of a road
is achieved by increasing a (see Figure 43).
If the load is such that 2. is not much greater
than unity, the value of oc is chosen such
that the required halfwidth of the road at
the inner curb of any turn is always smaller
than at the outer curb.
when the truck enters one section of a zigzag
turn after completing the previous one at
the point of inflection. The results represent-
ed graphically in Figure 43 correspond to
the case where prior to entering a turn the
tractor unit and the trailer are positioned
on a single straight line. But if there exists
a nonzero initial angle of lag C
0
caused by
the zigzag nature of the turn, we must select
the initial condition in the differential
equation (114) in the form u (0) = -C
0

Then the required width of the road is
determined from the following formula for ~ :
A= .. / ;..z+ -
1
- +2 !:_sinC
0
- _!_ +
t' V a;Z <X <X h
We note that in the case of simple log
transportation, that is, a=O, it is impossible
to pass a turn with a greater than unity.
However, for fairly large values of a the
values a > 1 become possible, they must
obey the following relationship: a (1 -
a cos C) = sin C. Thus, the maximal value
of a is (1 - a
2
)-
1
12, with the practical
extremal value of a being 1.25 at a = 0.5.
We note in conclusion that for a > 0.5
considerable economy in the width of a road
is achieved by increasing a (see Figure 43).
If the load is such that A. is not much greater
than unity, the value of a is chosen such
that the required halfwidth of the road at
the inner curb of any turn is always smaller
than at the outer curb. Chapter 2
Qualitative Methods
of Studying Differential
Models
In solving the problems discussed in Chap-
ter 1 we constructed differential models
and then sought answers by integrating the
resulting differential equations. However,
as noted in the Preface, the overwhelming
majority of differential equations are not
integrable in closed (analytical) form.
Hence, to study differential models of real
phenomena and processes we need methods
that will enable us to extract the necessary
information from the properties of the
differential equation proper. Below with
concrete examples we show how in solving
practical problems one can use the simplest
approaches and methods of the qualitative
theory of ordinary differential equations.
2.1 Curves of Constant Direction
of Magnetic Needle
Let us see how in qualitative integration,
that is, the process of establishing the gen-
eral nature of solutions to ordinary differen-
tial equations, one can use a general proper-
ty of such equations whose analogue, for
Chapter 2
Qualitative Methods
of Studying Differential
Models
In solving the problems discussed in Chap-
ter 1 we constructed differential models
and then sought answers by integrating the
resulting differential equations. However,
as noted in the Preface, the overwhelming
majority of differential equations are not
integrable in closed (analytical) form.
Hence, to study differential models of real
phenomena and processes we need methods
that will enable us to extract the necessary
information from the properties of the
differential equation proper. Below with
concrete examples we show how in solving
practical problems one can use the simplest
approaches and methods of the qualitative
theory of ordinary differential equations.
2.1 Curves of Constant Direction
of Magnetic Needle
Let us see how in qualitative integration,
that is, the process of establishing the gen-
eral nature of solutions to ordinary differen-
tial equations, one can use a general proper-
ty of such equations whose !lllalogue, for Ch. 2. Qualitative Methods 137
example, is the property of the magnetic
field existing at Earths surface. The reader
will recall that curves at Earths surface
can be specified along which the direction
of a magnetic needle is constant.
Thus, let us consider the first-order or-
dinary differential equation
-3%-=f(, 11), (U6)
where the function f (1:, y) is assumed single-
valued and continuous over the set of
variables sc and y within a certain domain D
of the (ec, y)-plane. To each point M (x, y)
belonging to the domain D of function
f (x, y) the differential equation assigns
a value of dy/dx, the slope K of the tangent
to the integral curve at point M (:2:, y).
Bearing this in mind, we say that at each
point M (:c, y) of D the differential equation
(116) defines a direction or a line element.
The collection of all line elements in D is
said to be the field of directions or the
line element field. Graphically a linear ele-
ment is depicted by a segment for which
point M (az, y) is an interior point and which
makes an angle 9 with the positive direction
of the x axis such that K = tan 9 =
f (sc, y). From this it follows that geometri-
cally the differential equation (116) expresses
the fact that the direction ofa tangent at
every point of an integral curve coincides with
the direction of the field at this point.
Ch. 2. Qualitative Methods
f37
example, is the property of the magnetic
field existing at Earth's surface. The reader
will recall that curves at Earth's surface
can be specified along which the direction
of a magnetic needle is constant.
Thus, let us consider the first-order or-
dinary differential equation
dy (
rli"=l x, y),
(116)
where the function f (x, y) is assumed single-
valued and continuous over the set of
variables x andy within a certain domain D
of the (x, y)-plane. To each point M (x, y)
belonging to the domain D of function
f (x, y) the differential equation assigns
a value of dy/dx, the slope K of the tangent
to the integral curve at point M (x, y).
Bearing this in mind, we say that at each
point M (x, y) of D the differential equation
(116) defines a direction or a line element.
The collection of all line elements in D is
said to be the field of directions or the
line element field. Graphically a linear ele-
ment is depicted by a segment for which
point M (x, y) is an interior point and which
makes an angle a with the positive direction
of the X axis such that K = tan a =
f (x, y). From this it follows that geometri-
cally the differential equation (116) expresses
the fact that the direction of a tangent at
every point of an integral curve coincides with
the direction of the field (Lt this point. 138 Differential Equations in Applications
To construct the field of directions it has
proved expedient to use the concept of iso-
clinals (derived from the Greek words for
"equal" and "sloping"), that is, the set of
points in the (.2:, y)-plane at which the
direction of the field specified by the differ-
ential equation (116) is the same.
The isoclinals of the magnetic field at
Earths surface are curves at each point of
which a magnetic needle points in the same
direction. As for the differential equation
(116), its isoclinals are given by the equa-
tion
f (J:. y) = vi
where v is a varying real parameter.
Knowing the isoclinals, we can approx-
imately establish the behavior of the
integral curves of a given differential equa-
tion. Let us consider, for example, the
{11%-=w2l-y'*,
which cannot be integrated in closed (ana-
lytical) form. The form of this equation
suggests that the family of isoclinals is given
by the equation
x2+y2=v, v>O,
that is, the isoclinals are concentric circles
of radius I/v centered at the origin and ly-
ing in the (as, y)-plne, At each point of
To construct the field of directions it has
proved expedient to use the concept of iso-
clinals (derived from the Greek words for
"equal" and "sloping"), that is, the set of
points in the (x, y)-plane at which the
direction of the field specified by the differ-
ential equation (116) is the same.
The isoclinals of the magnetic field at
Earth's surface are curves at each point of
which a magnetic needle points in the same
direction. As for the differential equation
(116), its isoclinals are given by the equa-
tion
I (x, y) = v,
where v is a varying real parameter.
Knowing the isoclinals, we can approx-
imately establish the behavior of the
integral curves of a given differential equa-
tion. Let us consider, for example, the
: ~ =xz+ yz,
which cannot be integrated in closed (ana-
lytical) form. The form of this equation
suggests that the family of isoclinals is given
by the equation
x
2
+ y
2
= v, v > 0,
that is, the isoclinals are concentric circles
of radius Vv centered at the origin and ly-
ing in the (x, y)-plane. At each point of Ch. 2. Qualitative Methods 139
9
V gg? it
Fig. 44
such an isoclinal the slope of the tangent
to the integral curve that passes through
this point is equal to the squared radius of
the corresponding circle. This information
alone is sufficient to convey an idea of the
behavior of the integral curves of the
given differential equation (Figure 44).
We arrived at the final result quickly
because the example was fairly simple.
However, even with more complicated equa-
tions knowing the isoclinals may prove to
be expedient in solvinga specific problem.
Let us consider a geometric method of
integration of differential equations of the
139
!I
Fig. 44
such an isoclinal the slope of the tangent
to the integral curve that passes through
this point is equal to the squared radius of
the corresponding circle. This information
alone is sufficient to convey an idea of the
behavior of the integral curves of the
given differential equation (Figure 44).
We arrived at the final result quickly
because the example was fairly simple.
However, even with more complicated equa-
tions knowing the isoclinals may prove to
he expedient in &olving a specific problem.
Let us consider a geometric method of
integration of differential equations of the 140 Differential Equations in Applications
type (116). The method is based on using
the geometric properties of the curves
given by the equations
f( y)=0 (I0)
0f(=v. y) 6f(==. 11) _
3;,+*5,;*f(9? ll)- 0- (L)
Equation (I 0), that is, the "zero" isoclinal
equation, specifies curves at whose points
dy/dx = 0. This means that the points of
these curves may prove to be points of maxi-
ma or minima for the integral curves of
the initial differential equation. This ox-
plains why out of the entire set of isoclinals
we isolate the "zero" isoclinal.
For greater precision in constructing
integral curves it is common to find the
set of inflection points of these curves (pro-
vided that such points exist). As is known,
points of inflection should be sought among
the points at which y" vanishes. Employing
. _ 6f(=, y) 6f( y)
y 5;+w*y
0 . 6* ,
:;.%+Ji.!%,(_,,, y)_
The curves specified by Eq. (L) are the pos-
sible pointof-inflection curves.* Note, for
* It is assumed here that integral curves that
{ill a certain domain possess the property that only
(que integral curve passes through each point of the
omain.
type (116). The method is based on using
the geometric properties of the curves
given by the equations
f (x, y) = 0, (/
0
)
of (z, y) + of (z, y) f ( ) _
0
(L)
oz iJy X, y - .
Equation (/
0
), that is, the "zero" isoclinal
equation, specifies curves at whose points
dy/dx = 0. This means that the points of
these curves may prove to be points of maxi-
ma or minima for the integral curves of
the initial differential equation. This ex-
plains why out of the entire set of isoclinals
we isolate the "zero" isoclinal.
For greater precision in constructing
integral curves it is common to find the
set of inflection points of these curves (pro-
vided that such points exist). As is known,
points of inflection should be sought among
the points at which y" vanishes. Employing
y"
= of (z, y) + of (z, y) ,
oz oy y
of (z, y) + of (z, y) f ( )
8z 8y x, Y
The curves specified by Eq. (L) are the pos-
sible point-of-inflection curves.* Note, for
It is assumed here that integral curves that
fill a certain domain possess the property that only
one integral curve passes througli each point of the
domain. Ch. 2. Qualitative Methods 141
one, that a point of inflection of an integral
curve is a point at which the integral curve
touches an isoclinal.
Thezcurves consisting of extremum points
(maxima and minima points) and points of
inflection of integral curves break down the
domain of f into such subdomains S1,
S2, ., Sm in which the first and second
derivatives of the solution to the differ-
ential equation have definite signs. In
each specific case these subdomains should
be found. This enables giving a rough
picture of the behavior of integral curves.
As an example let us consider the differ-
ential equation y = x + y. The equation
of curve (I 0) in this case has the form
as + y = 0, or y = -:1:. A direct check
verifies that curve (I 0) is not an integral
curve. As for curve (L), whose equation in
this case is y -lx } 1 = 0, we find that
it is an integral curve and, hence, is not
a point-0finflection curve.
The straight lines (I 0) and (L) break
down the (sc, y)plane into three subdomains
(Figure 45): S1 (y' >O, y" >O), to the
right of the straight line (I 0), S 2 (y' < 0,
y" > 0), between the straight lines (I0) and
(L), and S3 (y' < 0, y" < 0), to the left ofthe
straight line (L). The points ofminima of the
integral curves lie on the straight line (I 0).
T 0 the right of (I 0) the integral curves point
upward, while to the left they point down-
ward (left to right in Figure 45). There are
f41
one, that a point of inflection of an integral
curve is a potnt at which the integral curve
touches an isoclinal.
The: curves consisting of extremum points
(maxima and minima points) and points of
inflection of integral curves break down tho
domain of f into such subdomains S
1
,
S
2
, , S m in which the first and second
derivatives of the solution to the differ-
ential equation have definite signs. In
each specific case these subdomains should
be found. This enables giving a rough
picture of the behavior of integral curves.
ential equation y' = x + y. The equation
of curve (I
0
) in this case has the form
x + y = 0, or y = -x. A direct check
verifies that curve (I
0
) is not an integral
curve. As for curve (L), whose equation in
this case is y + x + 1 = 0, we find that
it is an integral curve and, hence, is not
a point-of-inflection curve.
The straight lines (I
0
) and (L) break
down the (x, y)-plane into three subdomains
(Figure 45): S
1
(y' > 0, y" > 0), to the
right of the straight line (I
0
), S
2
(y' < 0,
y" > 0), between the straight lines (/
0
) and
(L), and S
3
(y' < 0, y" < 0), totheleftofthe
straight line (L). The points of minima of the
integral curves lie on the straight line (I
0
).
To the right of (/
0
) the integral curves point
upward, while to the left they point down-
ward (left to right in Figure 45). There are 142 Differential Equations in Applications
SI
0
Y
S5 \ K
-1
Fig. 45
no points of inflection. To the right of the
straight line (L) the curves are convex down-
ward and to the left convex upward. The
behavior of the integral curves on the whole
is shown in Figure 45.
Note that in the given case the integral
curve (L) is a kind of "dividing" line, since
it separates one family of integral curves
from another. Such a curve is commonly
known as a separatrix.
f42 Differential Equations in Applications
Fig. 45
no points of inflection. To the right of the
straight line (L) the curves are convex down-
ward and to the left convex upward. The
behavior of the integral curves on the whole
is shown in Figure 45.
Note that in the given case the integral
curve (L) is a kind of "dividing" line, since
it separates one family of integral curves
from another. Such a curve is commonly
known as a separatrix. Ch. 2. Qualitative Methods 143
Existence and Uniqueness
Theorems?
When speaking of isoclinals and pointof
inflection curves in Section 2.1, we tacitly
assumed that the differential equation in
question had a solution. The problem of
when a solution exists and of when it is
unique is solved by the socalled existence
and uniqueness theorems. These theorems
are important for both theory and practice.
Existence and uniqueness theorems are
highly important because they guarantee
the legitimacy of using the qualitative meth-
ods of the theory of differential equations
to solve problems that emerge in science
and engineering. They serve as a basis for
creating new methods and theories. Often
their proof is constructive, that is, the meth-
ods by which the theorems are proved sug-
gest methods of finding approximate solu-
tions with any degree of accuracy. Thus,
existence and uniqueness theorems lie at
the base of not only the above-noted quali-
tative theory of differential equations but
also the methods of numerical integration.
Many methods of numerical solution of
differential equations have been developed,
and although they have the common draw-
back that each provides only a concrete
solution, which narrows their practical
potential, they are widely used. It must be
Existence and Uniqueness
Theorems?
143
When speaking of isoclinals and point-of-
inflection curves in Section 2.1, we tacitly
assumed that the differential equation in
question had a solution. The problem of
when a solution exists and of when it is
unique is solved by the so-called existence
and uniqueness theorems. These theorems
are important for both theory and practice.
Existence and uniqueness theorems are
highly important because they guarantee
the legitimacy of using the qualitative meth-
ods of the theory of differential equations
to solve problems that emerge in science
and engineering. They serve as a basis for
creating new methods and theories. Often
their proof is constructive, that is, the meth-
ods by which the theorems are proved sug-
gest methods of finding approximate solu-
tions with any degree of accuracy. Thus,
existence and uniqueness theorems lie at
the base of not only the above-noted quali-
tative theory of differential equations but
also the methods of numerical integration.
Many methods of numerical solution of
differential equations have been developed,
and although they have the common draw-
back that each provides only a concrete
solution, which narrows their practical
potential, they are widely used. It must be 144 Differential Equations in Applications
noted, however, that before numerically
integrating a differential equation one must
always turn to existence and unique-
ness theorems. This is essential to avoid
misunderstandings and incorrect conclu-
sions.
To illustrate what has been said, let us
take two simple examples,* but hrst let us
formulate one variant of existence and
uniqueness theorems.
Existence theorem If in Eq. (116) func-
tion f is defined and continuous on a bounded
domain D in the (x, y)plane, then for every
point (xo, yo) E D there exists a solution y (x)
to the initial-value problem **
dy 117
E- f (ac. y). y (wo.) yo. ( >
that is defined on a certain interval con-
taining point xo.
Existence and uniqueness theorem If
in Eq. (116) function f is defined and contin-
uous on a bounded domain D in the (x, y)-
* See C.E. Roberts, J r., "Why teach existence and
uniqueness theorems in the first course of ordinary
differential equations?", Int. J. Math. Educ. Sci.
Technol. 7, No. 1: 41-44 (1976).
** If we wish to find a solution of a differential
equation satisfying a certain initial condition (in
our case the initial condition is y (xo) = yo), such
a problem issaid to be an initial-value problem.
noted, however, that before numerically
integrating a differential equation one must
always turn to existence and unique-
ness theorems. This is essential to avoid
misunderstandings and incorrect conclu-
sions.
To illustrate what has been said, let us
take two simple examples,* but first let us
formulate one variant of existence and
uniqueness theorems.
Existence theorem If in Eq. (116) func-
tion f is defined and continuous on a bounded
domain D in the (x, y)-plane, then for every
point (x
0
, y
0
) ED there exists a solution y (x)
to the initial-value problem * *
: ~ = f (x, y), Y (xo,) = Yo
(117)
that is defined on a certain interval con-
taining point x
0

Existence and uniqueness theorem If
in Eq. (116) function f is defined and contin-
uous on a bounded domain D in the (x, y)-
See C.E. Roberts, 1r., "Why teach existence and
uniqueness theorems in the first course of ordinary
differential equations?", Int. J. Math. Educ. Sci.
Technol. 7, No. 1: 41-44 (1976).
** If we wish to find a solution of a differential
equation satisfying a certain initial condition (in
our case the initial condition is y (.:r
0
) = y
0
), such
a problem is said to be an initial-value problem. Ch. 2. Qualitative Methods 145
plane and satisfies in D the Lipschitz condi-
tion in variable y, that is,
liisv, yi)-f(x, y1)|<L|y2-y1l.
with L a positive constant, then for every
point (xo, yo) E D there exists a unique solu-
tion y (x) to the initial-value problem (117)
defined on a certain interval containing
point xo.
Extension theorem If the hypotheses of
the existence theorem or the existence and
uniqueness theorem are satisfied, then every
solution to Eq. (116) with the initial data
(xo, y0) E D can be extended to a point that
lies as close to the boundary of D as desired.
In the first case the extension is not necessarily
unique while in the second it is.
Using the numerical Euler integration
method with the iteration scheme y,+1 =
yi -} hf (xi, yi) and step h = 0.1, solve
the initial-value problem
y' == -x/y, y ( 1) = 0.21 (118)
on the interval l-- 1, 3].
Note that the problem involving the
equation y' = x/y emerges, for example,
from the problem considered on p. 162 con-
cerned with a conservative system consist-
ing of an object oscillating horizontally in
10-0770
145
plane and satisfies in D the Lipschitz condi-
tion in variable y, that is,
with L a positive constant, then for every
point (x
0
, y
0
) ED there exists a unique solu-
tion y (x) to the initial-value problem (117)
defined on a certain interval containing
point x
0

Extension theorem If the hypotheses of
the existence theorem or the existence and
uniqueness theorem are satisfied, then every
solution to Eq. (116) with the initial data
(x
0
, y
0
) ED can be extended to a point that
lies as close to the boundary of D as desired.
In the first case the extension is not necessarily
unique while in the second it is.
Using the numerical Euler integration
method with the iteration scheme Yi+l =
Yr + hf (x" Yt) and step h = 0.1, solve
the initial-value problem
y' = -xly, y (- 1) = 0.21 (118)
on the interval [- 1, 31.
Note that the problem involving the
equation y' = -x/y emerges, for example;
from the problem considered on p. 162 con-
cerned with a conservative system consist-
ing of an object oscillating horizontally in
t0-0770 146 Differential Equations in Applications
a vacuum under forces exerted by linear
springs.
To numerically integrate the initialvalue
problem (117) on the l-1, 2.8] interval
we compile a program for building the
graph of the solution:
10 REM Numerical integration of differential
equation
20 DEF FNF(X, Y) :
30 GOSUB 1110: REM Coordinate axes
40 REM Next value of function
1000 LINE (FNX(X), FNY(Y)), 2
1010 IF X < 2.9 GOTO 100
1020 LOCATE 23, 1
1030 END
1100 REM Construction of coordinate axes
1110 SCREEN 1, 1, 0: KEY OFF: CLS
1120 DEF FNX(X) = 88 l 80*X
1130 DEF FNY(Y) = 96 80*Y
1140 REM Legends on coordinate axes
1150 LOCATE 1, 11, 0 PRINT "Y"
1160 LOCATE 3, 11: PRINT "1"
1170 LOCATE 13, 39: PRINT "X"
1180 LOCATE 23, 10: PRINT "-1"
1190 FOR I = -1 TO 2: LOCATE 13, 10*1 +
11: PRINT USING "4:t xt"; I;: NEXT I
1200 REM Oy
1210 DRAW"BM88, ONM -2, {8NM + 2,
|8D16"
1220 FORI = 1 TO 11: DRAW"NR2D16": NEXT
1230 REM Ox
a vacuum under forr.es exerted by linear
springs.
To numerically integrate the initial-value
problem (117) on the [-1, 2.8] interval
we compile a program for building the
graph of the solution:
tO REM Numerical integration of differential
equation
20 DEF FNF(X, Y) =
1000 LINE -(FNX(X), FNY(Y)), 2
t010 IF X < 2.9 GOTO 100
1020 LOCATE 23, 1
t030 END
1 tOO REM Construction of coordinate axes
1110 SCREEN 1, 1, 0: KEY OFF: CLS
H20 DEF FNX(X) = 88 + 80"'X
H30 DEF FNY(Y) = 96 - 80"'Y
H40 REM Legends on coordinate axes
H50 LOCATE 1, H, 0 PRINT "Y"
H60 LOCATE 3, 11: PRINT "1"
H70 LOCATE 13, 39: PRINT "X"
1180 LOCATE 23, 10: PRINT "-1"
H90 FOR I = -1 TO 2: LOCATE t3, tOI +
11: PRINT USING"# #"; 1;: NEXT I
1200 REM Oy
1210 DRAW "BM88, ONM- 2, +8NM + 2,
+8D16"
1220 FOR I= t TO 11: DRAW"NR2D16": NEXT
1230 REM Ox Ch. 2. Qualitative Methods 147
1240 DRAW"BM319, 96NM - 7, 2NM 7
|2L23"
1250 FOR I = 1 TO 19: DRAW"NU2L16": NEXT
1260 REM Input of initial data
1270 LOCATE 25, 1: PRINT STRING$(40, "),
1280 LOCATE 25, 1: INPUT, "xO =", X: INPUT;
"dx =, DX: INPUT; "yO =", Y
1290 PSET (FNX(X), FNY(Y)), 2
1300 RETURN
Here in line 20 the righthand side of the
differential equation is specified, and lines
50 to 990 must hold the program of the
numerical integration of the differential
equation.
In our case (the initial-value problem
(118)) the beginning of the program has
the following form:
10 REM Eulers method
20 DEF FNF(X, Y) = X/Y
100 Y = Y |- DX*FNF(X, Y)
110 X = X + DX
The results are presented graphically in
Figure 46.
Let us now turn to the existence theorem.
For the initial-value problem (118), the
function f (:2:, y) = x/y is defined and
continuous in the entire (x, y)plane exclud-
10*
147
1240 DRAW "BM319, 96NM- 7, -2NM- 7
+2L23''
1250 FOR I = 1 TO 19: DRAW "NU2L16": NEXT
1260 REM Input of initial data
1270 LOCATE 25, 1: PRINT STRING$(40, "),
1280 LOCATE 25, 1: INPUT, "xO =",X: INPUT;
"dx =", DX: INPUT; "yO=", Y
1290 PSET (FNX(X), FNY(Y)), 2
1300 RETURN
Here in line 20 the right-hand side of the
differential equation is specified, and lines
50 to 990 must hold the program of the
numerical integration of the differential
equation.
In our case (the initial-value problem
(118)) the beginning of the program has
the following form:
10 REM Euler's method
;w DEF FNF(X, Y) = -X/Y
100 Y = Y + DX*FNF(X, Y)
110 X= X+ DX
Figure 46.
Let us now turn to the existence theorem.
For the initial-value problem (118), the
function f (x, y) = -xly is defined and
continuous in the entire (x, y)-plane exclud-
10* H
.
N
4 Q
*<
\`
<o
BO
< N <\ <

5 " U U 5 T
N n
9
<
S
I ec
<1
'\ E
' In
<o
r::,
""
t::l
::::,..,
""
(() ....
r::,
c:::>
I
""I
I
c:::t
I
<o
c:::t
I
eo
....
....
tiD
. ..
I
""" Ch. 2. Qualitative Methods 149
ing the :1: axis. Thus, in accordance with
the existence theorem, the initial-value
problem (118) has a solution y (x) defined on
a certain interval containing point xo =
-1. This solution, according to the exten-
sion theorem, can be extended to a value
of y (cc) close to the value y (x) =- O. As
a result of numerical integration we have
arrived at a solution of (118) defined on an
interval (a, b), with a< -1 and 1.3<
b<1.4. However, allowing for the concrete
form of the differential equation, we can
specify the true interval in which the
solution to the initial-value problem (118)
exists. Indeed, since in the initial equation
the variables can be separated, we have
y I
S mm- we
0.21 -1
Integrating, we get y = I/1.0441 1:2.
Hence, a solution to the initialvalue prob-
lem (118) exists only for | :2: | < I/1.0441
,*:1 1.0218.
Thus, by resorting to the existence theo-
rem (and to the extension theorem) we were
able to "cut ofi" the segment on which there
is certain to be no solution of the initial-
value problem. If we employ only numerical
integration, we arrive at an erroneous
result. The fact is that as the solution
y = y (as) approaches the sv axis, the angle
149
ing the x axis. Thus, in accordance with
the existence theorem, the initial-value
problem (118) has a solution y (x) defined on
a certain interval containing point x
0
=
-1. This solution, according to the exten-
sion theorem, can be extended to a value
of y (x) close to the value y (x) = 0. As
a result of numerical integration we have
arrived at a solution of (118) defined on an
interval (a, b), with a < -1 and 1.3 <
b < 1.4. However, allowing for the concrete
form of the differential equation, we can
specify the true interval in which the
solution to the initial-value problem (118)
exists. Indeed, since in the initial equation
the variables can be separated, we have
y X
~ 11 dt] = - ~ ~ d ~ .
0.21 -1
Integrating, we get y = V 1.0441 - x
2
:
Hence, a solution to the initial-value prob-
lem (118) exists only for I x I< V 1.0441
~ 1.0218.
Thus, by resorting to the existence theo-
rem (and to the extension theorem) we were
able to "cut off'' the segment on which there
is certain to be no solution of the initial-
value problem. If we employ only numerical
integration, we arrive at an erroneous
result. The fact is that as the solution
y = y (x) approaches the x axis, the angle 150 Differential Equations in Applications
of slope of the curve tends to 90. Therefore,
in the time that the independent varia-
ble x changes by 0.1, the value of y is able to
"jump over" the as axis, and we find ourselves
on an integral curve that differs from the
original. This happens because the Euler
method allows for the angle of slope only
at the running point.
The following example is even more
instructive. We wish to solve the initial-
value problem
y' = 3x?/yi y (-1) == --1 (M9)
on the segment [ 1, 1]. The approach here
consists in first employing the Euler method
and then an improved Euler method with
a step h = 0.1 and an iteration scheme
E/z+1 = Hz + hl (xi+1/21 yi+1/2)? with
yi+1/2 = yi + hf ( ?z yi)/2-
We solve the initial-value problem (119)
using the above program, whose beginning
in the case at hand has the following form:
10 REM Eulers method
20 DEF FNF(X, Y) =3=X=SGN(Y)ABS(Y) A
(1/3)
30 GOSUB 1100: BEM Coordinate axes
100 Y = Y l DX*FNF(X, Y)
110 X = X { DX
Figure 47.
of slope of the curve tends to 90. Therefore,
in the time that the independent varia-
ble x changes by 0.1, the value of y is able to
"jump over" the x axis, and we find ourselves
on an integral curve that differs from the
original. This happens because the Euler
method allows for the angle of slope only
at the running point.
The following example is even more
instructive. We wish to solve the initial-
value problem
3f-
y' = 3xv y, y (-1) = -1 (119)
on the segment[- 1, 1]. The approach here
consists in first employing the Euler method
and then an improved Euler method with
a step h = 0.1 and an iteration scheme
Yi-l-t = Yi + hj (xi+l/2 Yt+t/2), with
Yi+t/2 = Yi + hf ( Xj, Yi)/2.
We solve the initial-value problem (119)
using the above program, whose beginning
in the case at hand has the following form:
10 REM Euler's method
20 DEF FNF(X, Y) = 3XSGN(Y)ABS(Y) _...,
(1/3)
100 Y = Y + DX*FNF(X, Y)
110 X= X+ DX
Figure 47. Ch. 2. Qualitative Methods 151
.16/*
7
Q6
Q2
*7 l26 Q2 0 QZ Q5 1 as
-Q2
*625
I
Fig. 47
As for the improved Euler method, the
beginning of the program has the following
form:
10 REM Improved Eulers method
20 DEF FNF(X, Y) == 3*X*SGN(Y)*ABS(Y) A
(1/3)
50 D2 = DX/2
151
0.8
0.2
-o.z
0
az
-az
-ao
- 1
Fig. 47
As for the improved Euler method, the
beginning of the program has the following
form:
10 REM Improved Euler's method
20 DEF FNF(X, Y) = 3*X*SGN(Y)*ABS(Y) ........
(1/3)
50 D2 = DX/2 152 Differential Equations in Applications
9
1
Q6
622
-7 -05 QZ
0 az as 1 .7;
QZ
-06
-7
Fig. 48
100 Y1 = Y + D2FNF(X, Y)
110 Y = Y -IDX=FNF(X l D2, Y1)
120 X = X + DX
Figure 48, and the diagram differs from
that depicted in Figure 47.
To look into the reason for such striking
discrepancy in the results, we integrate the
-1
-0.6
Fig. 48
!I
1
Q6
0.2
-0.2
fOO Y1 = Y + D2FNF(X, Y)
110 Y = Y + DXFNF(X + D2, Y1)
120 X= X+ DX
Figure 48, and the diagram differs from
that depicted in Figure 47.
To look into the reason for such striking
discrepancy in the results, we integrate the Ch. 2. Qualitative Methods 153
initial-value problem. Separating the vari-
ables, we get
Q! x
S 1]**/3 dn:-3S d,
-1 -1
or, finally, y = ;i;x". This result already
suggests that the solution via the Euler
method gives the function yl (sc) == 1:3 while
the solution via the improved Euler method
gives
.2:3 if :1:<O,
y2(x) Zl -.1:3 if ;1:>O
Both yl and yl are solutions to the initial-
value problem (119), which means that the
solution of the initialvalue problem con-
sidered on the segment [-1, 1] is not
unique.
Let us now turn to the existence and
uniqueness theorem in connection with this
problem. First, we note that since the func-
tion f (.2:, y) = 3;::%/y` is continuous in the
entire (1:, y)-plane, the existence theorem
implies that the initialvalue problem (119)
has a solution defined on a segment con-
taining point xl, = --1, and, according to
the extension theorem, this solution can
be extended to any segment. Further, since
6f (ac, y)/0y = :cy2/3, the functionf (sv, y)=
153
initial-value problem. Separating the vari-
ables, we get
y X
~ 11-1/3 d11 = 3 ~ ~ d ~ ,
-1 -1
or, finally, y = x
3
This result already
suggests that the solution via the Euler
method gives the function y
1
(x) = x
3
while
the solution via the improved Euler method
gives
f
x
3
if x ~ O ,
Yz (x) = ' - x3 if x > 0.
Both y
1
and y
2
are solutions to the initial-
value problem (119), which means that the
solution of the initial-value problem con-
sidered on the segment [-1, 1) is not
unique.
Let us now turn to the existence and
uniqueness theorem in connection with this
problem. First, we note that since the func-
tion f (x, y) = 3xVy is continuous in the
entire (x, y)-plane, the existence theorem
implies that the initial-value problem (119)
has a solution defined on a segment con-
taining point x
0
= -1, and, according to
the extension theorem, this solution can
be extended to any segment. Further, since
at (x, y)lay = xy-
2
1
3
, the function I (x, y)= 154 Differential Equations in Applications
3x:}/y satisfies the Lipschitz condition in
variable y in any domain not containing
the 1: axis. If, however, a domain does
contain points belonging to the x axis, it
is easy to show that the function does
not satisfy the Lipschitz condition. Hence,
from the existence and uniqueness theorem
(and the extension theorem) it follows that
in this case the solution to the initialvalue
problem can be extended in a unique man-
ner at least to the x axis. But since the
straight line y = O constitutes a singular
integral curve of the differential equation
y' = Bx:}/y, we already know that as soon
as y vanishes, there is no way in which we
can extend the solution to the initial-value
problem (119) beyond point O(O, 0) in
a unique manner.
Thus, by resorting to the existence and
uniqueness theorem (and the extension
theorem) we were able to understand the
results of numerical integration, that is, if
we are speaking of the uniqueness of the
solution of the initialvalue problem (119)
on the [-1, 1] segment, the solution exists
and is defined only on the segment [-1, O].
Generally, however, there can be several
such solutions.
3xVY' satisfies the Lipschitz condition in
variable y in any domain not containing
the x axis. If, however, a domain does
contain points belonging to the x axis, it
is easy to show that the function does
not satisfy the Lipschitz condition. Hence,
from the existence and uniqueness theorem
(and the extension theorem) it follows that
in this case the solution to the initial-value
problem can be extended in a unique man-
ner at least to the x axis. But since the
straight line y = 0 constitutes a singular
integral curve of the differential equation
y' = 3xVY.' we already know that as soon
as y vanishes, there is no way in which we
can extend the solution to the initial-value
problem (119) beyond point 0 (0, 0) in
a unique manner.
Thus, by resorting to the existence and
uniqueness theorem (and the extension
theorem) we were able to understand the
results of numerical integration, that is, if
we are speaking of the uniqueness of the
solution of the initial-value problem (119)
on the [ -1, 1] segment, the solution exists
and is defined only on the segment [-1, 0].
Generally, however, there can be several
such solutions. Ch. 2. Qualitative Methods 155
2.3 A Dynamical Interpretation
of Second-Order Differential
Equations
Let us consider the nonlinear differential
equation
dzx dx
1at:(m) (*20)
whose particular case is the second-order dif-
ferential equation obtained on p. 72 when
we considered pendulum clocks. We take
a simple dynamical system consisting of
a particle of unit mass that moves along the
:1: axis (Figure 49) and on which a force
f (sz, dx/dt) acts. Then the differential
equation (120) is the equation of motion
of the particle. The values of zz: and dx/dt
at each moment in time characterize the
state of the system and correspond to a
point in the (:1:, dx/dt)-plane (Figure 50),
which is known as the plane of states or
the phase (az, dx/dt)plane. The phase plane
depicts the set of all possible states of the
dynamical system considered. Each new state
of the system corresponds to a new point in
the phase plane. Thus, the changes in the
state of the system can be represented by
the motion of a certain point in the phase
plane. This point is called a representative
point, the trajectory of the representative
point is known as the phase trajectory,
2.3 A Dynamical Interpretation
of Second-Order Differential
Equations
155
Let us consider the nonlinear differential
equation
dtx ( dx )
des =f x,dt
(120)
whose particular case is the second-order dif-
ferential equation obtained on p. 72 when
we considered pendulum clocks. We take
a simple dynamical system consisting of
a particle of unit mass that moves along the
x axis (Figure 49) and on which a fo.rce
f (x, dx/dt) acts. Then the differential
equation (120) is the equation of motion
of the particle. The values of x and dx/dt
at each moment in time characterize the
state of the system and correspond to a
point in the (x, dx/dt)-plane (Figure 50),
which is known as the plane of states or
the phase (x, dxldt)-plane. The phase plane
depicts the set of all possible states of the
dynamical system considered. Each new state
of the system corresponds to a new point in
the phase plane. Thus, the changes in the
state of the system can be represented by
the motion of a certain point in the phase
plane. This point is called a representative
point, the trajectory of the representative
point is known as the phase trajectory, 156 Differential Equations in Applications
da:
f(">Ez/T)
O J;
Fig. 49
a'
1 B-? (J;) gl?)-plane
0 as
Fig. 50
and the rate of motion of this point as the
phase velocity.
If we introduce the variable y = dx/dt,
Eq. (120) can be reduced to a system of
two differential equations:
d
g- =y, =.f(> y) (121)
If we take t as a parameter, then the solu-
tion to system (121) consists of two func-
tions, :1: (t) and y (t), that in the phase
(:1:, y)-plane define a curve (a phase trajec-
tory).
Fig. 49
dx
dt
0

-
a;
(x,

Fig. 50
and the rate of motion of this point as the
phase velocity.
If we introduce the variable y = dxldt,
Eq. (120) can be reduced to a system of
two differential equations:
dx dy )
dt=y, dt=f(x, y.
(121)
If we take t as a parameter, then the solu-
tion to system (121) consists of two func-
tions, x (t) and y (t), that in the phase
(x, y)-plane define a curve (a phase trajec-
tory). Ch. 2. Qualitative Methods 157
It can be shown that system (121) and
even a more general system
%f=X(r y) {1]%-=Y( y) (122)
where the functions X and Y and their
partial derivatives are continuous in a do-
main D, posseses the property that if .1: (t)
and y (t) constitute a'solution to the differ-
ential system, we can write
w=w(+C),y=y(+C). (123)
where C is an arbitrary real constant, and
(123) also constitute a solution to the same
differential system. All solutions (123)
with different values of C correspond to a
single phase trajectory in the phase (.1:, y)-
plane. Further, if two phase trajectories
have at least one common point, they
coincide. Here the increase or decrease in
parameter t corresponds to a certain direc-
tion of motion of the representative point
along the trajectory. In other words, a
phase trajectory is a directed, or oriented,
curve. When we are interested in the
direction of the curve, we depict the direc-
tion of the representative point along the
trajectory by placing a small arrow on the
curve.
Systems of the (122) type belong to the
class of autonomous systems of differential
equations, that is, systems of ordinary
differential equations whose right-hand
Ch. 2. Qualitative Methods t57
It can be shown that system (121) and
even a more general system
~ ~ =X (x, y), ~ ~ = Y (x, y), (122)
where the functions X and Y and their
partial derivatives are continuous in a do-
main D, posseses the property that if x (t)
and y W constitute a"solution to the differ-
ential system, we can write
X = X (t + C), y = y (t + C), (123)
where C is an arbitrary real constant, and
(123) also constitute a solution to the same
differential system. All solutions (123)
with different values of C correspond to a
single phase trajectory in the phase (x, g)-
plane. Further, if two phase trajectories
have at least one common point, they
coincide. Here the increase or decrease in
parameter t corresponds to a certain direc-
tion of motion of the representative point
along the trajectory. In other words, a
phase trajectory is a directed, or oriented,
curve. When we are interested in the
direction of the curve, we depict the direc-
tion of the representative point along the
trajectory by placing a small arrow on the
curve.
Systems of the (122) type belong to the
class of autonomous systems of differential
equations, that is, systems of ordinary
differential equations whose right-hand 158 Differential Equations in Applications
sides do not explicitly depend on time t.
But if at least in one of the equations of
the system the right-hand side depends
explicitly on time t, then such a system is
said to be nonautonomous.
In connection with this classification of
differential equations the following remark
is in order. If a solution .2: (t) to Eq. (120)
is a nonconstant periodic solution, then
the phase trajectory of the representative
point in the phase (x, y)plane is a simple
closed curve, that is, a closed curve without
selfintersections. The converse is also true.
If differential systems of the (122) type
are specified in the entire (x, y)plane, then,
generally speaking, phase trajectories will
completely cover the phase plane without
intersections. And if it so happens that
X (xm yo) ;' Y ($01 y0) Z 0
at a point Mo (xo, yo), the trajectory degen-
erates into a point. Such points are called
singular. In what follows we consider primar-
ily only isolated singular points. A singu-
lar point Mo (xo, yo) is said to be isolated
if there exists a neighborhood of this point
which contains no other singular points
X0Pl> M0 (xm yo)-
From the viewpoint of a physical inter-
pretation of Eq. (120), the point Mo (xo, 0)
is a singular point. At this point y =0 and
f (xo, 0) = 0. Thus, in this case the isolated
singular point corresponds to the state of
sides do not explicitly depend on time t.
But if at least in one of the equations of
the system the right-hand side depends
explicitly on time t, then such a system is
said to be nonautonomous.
In connection with this classification of
differential equations the following remark
is in order. If a solution x (t) to Eq. (120)
is a nonconstant periodic solution, then
the phase trajectory of the representative
point in the phase (x, y)-plane is a simple
closed curve, that is, a closed curve without
self-intersections. The converse is also true.
If differential systems of the (122) type
are specified in the entire (x, y)-plane, then,
generally speaking, phase trajectories will
completely cover the phase plane without
intersections. And if it so happens that
X (xo, Yo) = Y (xo, Yo) = 0
at a point M
0
(x
0
, y
0
), the trajectory degen-
erates into a point. Such points are called
singular. In what follows we consider primar-
ily only isolated singular points. A singu-
lar point M
0
(x
0
, y
0
) is said to be isolated
if there exists a neighborhood of this point
which contains no other singular points
except M
0
(x
0
, y
0
).
From the viewpoint of a physical inter-
pretation of Eq. (120), the point M
0
(x
0
, 0)
is a singular point. At this point y = 0 and
f (x
0
, 0) = 0. Thus, in this case the isolated
singular point corresponds to the state of Ch. 2. Qualitative Methods 159
a particle of unit mass in which both the
speed dx/dt and the acceleration dy/di =
dzsc/dig of the particle are simultaneously ze-
ro, which simply means that the particle
is in the state of rest or in equilibrium.
In view of this, singular points are also
called points of rest or points of equilibrium.
The equilibrium states of a physical sys-
tem constitute very special states of the
system. Hence, a study of the types of
singular points occupies an important place
in the theory of differential equations.
The first to consider in detail the clas-
sification of singular points of differential
systems of the (122) type was the distin-
guished Russian scientist Nikolai E. Zhu-
kovsky (1847-1921). In his masters the-
sis "The kinematics of a liquid body",
presented in 1876, this problem emerged in
connection with the theory of velocities
and accelerations of fluids. The modern
names of various types of singular points were
suggested by the great French mathemati-
cian Jules H. Poincar (1854-1912).
Now let us try to answer the question of
the physical meaning that can be attached
to phase trajectories and singular points of
differential systems of the (122) type. For
the sake of clarity we introduce a two-
dimensional vector field (Figure 51) defined
by the function
V (rv. y) = X (wl y)i + Y (rv, y)i.
159
a particle of unit mass in which both the
speed dx/dt and the acceleration dy/dt =
d
2
x/dt
2
of the particle are simultaneously ze-
ro, which simply means that the particle
is in the state of rest or in equilibrium.
In view of this, singular points are also
called points of rest or points of equilibrium.
The equilibrium states of a physical sys-
tem constitute very special states of the
system. Hence, a study of the types of
singular points occupies an important place
in the theory of differential equations.
The first to consider in detail the clas-
sification of singular points of differential
systems of the (122) type was the distin-
guished Russian scientist Nikolai E. Zhu-
kovsky (1847-1921). In his master's the-
sis "The kinematics of a liquid body",
presented in 1876, this problem emerged in
connection with the theory of velocities
and accelerations of fluids. The modern
names of various types of singular points were
suggested by the great French mathemati-
cian Jules H. Poincare (1854-1912).
Now let us try to answer the question of
the physical meaning that can be attached
to phase trajectories and singular points of
differential systems of the (122) type. For
the sake of clarity we introduce a two-
dimensional vector field (Figure 51) defined
by the function
V (x, y) = X (x, y)i + Y (x, y)j, 160 Differential Equations in Applications
P -1
X
Y
Fig. 51
where i and j are the unit vectors directed
along the x and y axes, respectively, in
a Cartesian system of coordinates. At every
point P (x, y) the field has two components,
the horizontal X (x, y) and the vertical
Y (sc, y). Since dx/dt -= X (x, y) and dy/dt:
Y (as, y), the vector associated with each
nonsingular point P (x, y) is tangent at
this point to a phase trajectory.
lf variable t is interpreted as time, vec-
tor V can be thought of as the vector of
velocity of a representative point moving
along a trajectory. Thus, we can assume
that the entire phase plane is filled with
representative points and that each phase
trajectory constitutes the trace of a moving
representative point. As a result we arrive
at an analogy with the two-dimensional
motion of an incompressible fluid. Here,
since system (122) is autonomous, vector V
Fig. 51
where i and j are the unit vectors directed
along the x and y axes, respectively, in
a Cartesian system of coordinates. At every
point P (x, y) the field has two components,
the horizontal X (x, y) and the vertical
Y (x, y). Since dxldt = X (x, y) and dyldt=
Y (x, y), the vector associated with each
nonsingular point P (x, y) is tangent at
this point to a phase trajectory.
If variable t is interpreted as time, vec-
tor V can be thought of as the vector of
velocity of a representative point moving
along a trajectory. Thus, we can assume
that the entire phase plane is filled with
representative points and that each phase
trajectory constitutes the trace of a moving
representative point. As a result we arrive
at an analogy with the two-dimensional
motion of an incompressible fluid. Here,
since system (122) is autonomous, vector V Ch. 2. Qualitative Methods 1.61
at each fixed point P (sc, y) is time-inde-
pendent and, therefore, the motion of the
fluid is steady-state. The phase trajectories
in this;case are simply the trajectories of
the moving particles of fluid and the singu-
lar points O, O', and O" (see Figure 51) are
those where the fluid is at rest.
The most characteristic features of the
fluid motion shown in Figure 51 are (1)
the presence of singular points, (2) the
different patterns of phase trajectories near
singular points, (3) the stability or instabil-
ity at singular points (i.e. two possibili-
ties may realize themselves: the particles
that are in the vicinity of singular points
remain there with the passage of time or
they leave the vicinity for other parts of
the plane), and (4) the presence of closed
trajectories, which in the given case corre-
spondmto periodic motion.
These features constitute the main part
of the phase portrait, or the complete
qualitative behavior pattern of Qthe phase
trajectories of a generaltype system (122).
Since, as noted earlier, differential equa-
tions cannot generally be solved analytically,
the aim of the qualitative theory of ordinary
differential equations of the (122) type is to
build a phase portrait as complete as possi-
ble directly from the functions X (x, y)
and Y (sc, y).
11-0770
Ch. 2. Qualitative. Methods
t6t
at each ftxed point P (x, y) is time-inde-
pendent and, therefore, the motion of the
fluid is steady-state. The phase trajectories
in this )ase are simply the trajectories of
the moving particles of fluid and the singu-
lar points 0, 0', and 0" (see Figure 51) are
those where the fluid is at rest.
The most characteristic features of the
fluid motion shown in Figure 51 are (1)
the presence of singular points, (2) the
different patterns of phase trajectories near
singular points, (3) the stability or instabil-
ity at singular points (i.e. two possibili-
ties may realize themselves: the particles
that are in the vicinity of singular points
remain there with the passage of time or
they leave the vicinity for other parts of
the plane}, and (4) the presence of closed
trajectories, which in the given case corre-
spond .,.to periodic motion.
These features constitute the main part
of the phase portrait, or the complete
qualitative behavior pattern of )he phase
trajectories of a general-type system (122).
Since, as noted earlier, differential equa-
tions cannot generally be solved analytically,
the aim of the qualitative theory of ordinary
differential equations of the (122) type is to
build a phase portrait as complete as possi-
ble directly from the functions X (x, y)
and Y (x, y).
11-0770 162. Differential Equations in Applications
2.4 Conservative Systems
in Mechanics
Practice gives us ample examples of the
fact that any real dynamical system
dissipates energy. The dissipation usually
occurs as a result of some form of friction.
But in some cases it is so slow that it can be
neglected if the system is studied over a fair-
ly small time interval. The law of energy
conservation, namely, that the sum of kinet-
ic and potential energies remains constant,
can be assumed to hold true for such sys-
tems. Systems of this kind are called con-
servative. For example, rotating Earth may
be seen as a conservative system if we take
a time interval of several centuries. But if
we study Earths motion over several mil-
lion years, we must allow for energy dissipa-
tion related to tidal ilows of water in seas
and oceans.
A simple example of a conservative sys-
tem is one consisting of an object moving
horizontally in a vacuum under forces ex-
erted by two springs (Figure 52). lf as is the
displacement of the object (mass m) from
the state of equilibrium and the force with
which the two springs act on the object (the
restoring force) is proportional to :1:, the
equation of motion has the form
m-Q-1+kx=o, 1i>o.
162. Differential Equations in Applications
2.4 Conservative Systems
in Mechanics
Practice gives us ample examples of the
fact that. any real dynamical system
dissipates energy. The dissipation usually
occurs as a result of some form of friction.
But in some cases it is so slow that it can be
neglected if the system is studied over a fair-
ly small time interval. The law of energy
conservation, namely, that the sum of kinet-
ic and potential energies remains constant,
can be assumed to hold true for such sys-
tems. Systems of this kind are called con-
servative. For example, rotating Earth may
be seen as a conservative system if we take
a time interval of several centuries. But if
we study Earth's motion over several mil-
lion years, we must allow for energy dissipa-
tion related to tidal flows of water in seas
and oceans.
A simple example of a conservative sys-
tem is one consisting of an object moving
horizontally in a vacuum under forces ex-
erted by two springs (Figure 52). If x is the
di!!placement of the object (mass m) from
the state of equilibrium and the force with
which the two springs act on the object (the
restoring force) is proportional to x, the
equation of motion has the form
dx
m "'ifii'" + kx = 0, k > 0. Ch. 2. Qualitative Methods 163
rn \
Fig. 52
Springs of this type are known as linear,
since the restoring force exerted by them is
a linear function of x.
If an object of mass m moves in a medi-
um that exerts a drag on it (the damping
force) proportional to the objects velocity,
the equation of motion for such a nonconser-
vative system is
m!+i+kx=.0 c>o (124)
d2 dz
Here we are dealing with linear damping,
since the damping force is a linear function
of velocity dx/dt.
If f and g are such arbitrary functions that
f(0) = O and g (O) = 0, the more general
equation,
daz dx
m(W+(&T)+f(x)~0 (125)
can be interpreted as the equation of
motion of an object of mass m under
a restoring force j (x) and a damping
force g (dx/dt). Generally, these forces
11*
163
t 000000200000 0 0 0 0 ~ 0 0 0 0 0 0 j
Fig. 52
Springs of this type are known as linear,
since the restoring force exerted by them is
a linear function of x.
If an object of mass m moves in a medi-
um that exerts a drag on it (the damping
force) proportional to the object's velocity,
the equation of motion for such a nonconser-
vative system is
dlz dz
m""'"'(fti"+cdt+kx=O, c>O. (124)
Here we are dealing with linear damping,
since the damping force is a linear function
of velocity dxldt.
Iff and g are such arbitrary functions that
f (0) = 0 and g (0) = 0, the more general
equation,
dtz ( dx ) O
m dt2 +g Tt +f(x)=
(125)
can be interpreted as the equation of
motion of an object of mass m under
a restoring force -/ (x) and a damping
force -g (dxldt). Generally, these forces
are nonlinear; hence Eq. (125) can be con
sidered the basic equation of nonlinear me-
chanics.
Let us briefly examine the special case of
a nonlinear conservative system described
by the equation
d2
mg? +1* (w) =0 (126)
where the damping force is zero and, hence,
energy is not dissipated. From Eq. (126) we
can pass to the autonomous system
3 __ gg _ _ f()
in "y at m (127)
If we now exclude time t from Eqs. (127),
we arrive at a differential equation for the
trajectory of the system in the phase plane:
dy __ __ f ()
(E my . (128)
This equation can be written as
my dy = -j(x) dx. (129)
Then, assuming that as = xo at t = to and
y = yo, we can integrate Eq. (129) from to
to t. The result is
33
inzz/3---1- m 2: -S f(E) dE
2 2 'yf) - ...7
xo
are nonlinear; hence Eq. (125) can be
sidered the basic equation of nonlinear
chanics.
Let us briefly examine the special case of
a nonlinear conservative system deseribed
by the equation
dB.x
m dt
2
+ f (x) = 0,
(126)
where the damping force is zero and, hence,
energy is not dissipated. From Eq. (126) we
ean pass to the autonomous system
dx dy f (.x)
(ft=Y, dt= ---;n-
(127)
If we now exclude time t from Eqs. (127),
we arrive at a differential equation for the
trajectory of the system in the phase plane:
dy I (x)
dX=- my
This equation can be written as
mydy= -f(x) dx.
(128)
(129)
Then, assuming that x = x
0
at t = t
0
and
y = y
0
, we can integrate Eq. (129) from t
0
to t. The result is
X
my2-+ f (6)
Xo Ch. 2. Qualitative Methods 165
which may be rewritten as
1 1
my2+S f() d=*g'!/3+S 1() dw-
0 0
(130)
Note that my'!/2 := m(d:1:/dt)2/2 is the ki-
netic energy of a dynamical system and
x
V(w) = S f(E)d (131)
0
is the systems potential energy. Thus,
Eq. (130) expresses the law of energy conser-
vation:
mz/2+V() =E (132)
where E = my?/2 -}- V (.2:0) is the total ener-
gy of the system. Clearly, Eq. (132) is the
equation of the phase trajectories of sys-
tem (127), since it is obtained by integrating
Eq. (128). Thus, different values of E corre-
spond to diiierent curves of constant energy
in the phase plane. The singular points of
system (127) are the points M`, (::1,,, 0),
where :1:,, are the roots of the equation f (as) =
0. As noted earlier, the singular points are
points of equilibrium of the dynamical sys-
tem described by Eq. (126). Equation (128)
implies that the phase trajectories of the
system intersect the ac axis at right angles,
while the straight lines sc = 0:,, are parallel
165
which may be rewritten as
X Xo
fmyz_l- f(x)dx.
0 0
(130)
Note that my
2
12 = m (dx!dt)212 is tho ki-
netic energy of a dynamical system and
X
v (x) = 1 ds
(131)
0
is the system's potential energy. Thus,
Eq. (130) expresses the law of energy c.onser-
vation:
1
2
my
2
+ V(x) =E, (132)
where E = + V (x
0
) is the total ener-
gy of the system. Clearly, Eq. (132) is the
equation of the phase trajectories of sys-
tem (127), since it is obtained by integrating
Eq. (128). Thus, different values of E corre-
spond to different curves of constant energy
in the phase plane. The singular points of
system (127) are the points M v (xv, 0),
where Xv are the roots of the equation I (x) =
0. As noted earlier, the singular points are
points of equilibrium of the dynamical sys-
tem described by Eq. (126). Equation (128)
implies that the phase trajectories of the
system intersect the x axis at right angles,
While the straight lines x = xv are parallel 166 Differential Equations in Applications
to the as axis. In addition, Eq. (132) shows
that the phase trajectories are symmetric
with respect to the .2: axis.
In this case, if we write Eq. (132) in the
form
y = 1]/% [EV(v)l, (133)
we can easily plot the phase trajectories.
Indeed, let us introduce the (.2:, z)-plane,
the plane of energy balance (Figure 53),
with the z axis lying on the same vertical
line as they axis of tl1e phase plane. We then
plot the graph of the function z = V (1:) and
several straight lines z = E in the (x, z)-
plane (one such straight line is depicted in
Figure 53). We mark a value of E V (:1:)
on the graph. Then for a definite sc we multi-
ply E V (x) by 2/m and allow for formu-
la (133). This enables us to mark the respec-
tive values of y in the phase plane. Note
that since dx/dt = y, the positive direction
along any trajectory is determined by the
motion of the representative point from left
to right above the az axis and from right to
left below the x axis.
The above reasoning is fairly general and
makes it possible to investigate the equa-
tion of motion of a pendulum in a medium
without drag, which has the form (see p. 72)
(E+ksinx==0 (134)
dia 3
166 Differelltial Equations in Applications
to the x axis. In addition, Eq. (132) shows
that the phase trajectories are symmetric
with respect to the x axis.
In this case, if we write Eq. (132) in the
form
y = +y! [E-V(x)],
(133)
we can easily plot the phase trajectories.
Indeed, let us introduce the (x, z)-plane,
the plane of energy balance (Figure 53),
with the z axis lying on the same vertical
line as they axis of the phase plane. We then
plot the graph of the function z = V (x) and
several straight lines z = E in the (x, z)-
plane {one such straight line is depicted in
Figure 53). We mark a value of E - V (x)
on the graph. Then for a definite x we multi-
ply E - V (x) by 2/m and allow for formu-
la (133). This enables us to mark the respec-
tive values of y in the phase plane. Note
that since dxldt = y, the positive direction
along any trajectory is determined by the
motion of the representative point from left
to right above the x axis and from right to
left below the x axis.
The above reasoning is fairly general and
makes it possible to investigate the equa-
tion of motion of a pendulum in a medium
without drag, which has the form (seep. 72)
d
2
x
dtB + k Slll X= 0,
(134) Ch. 2. Qualitative Methods 167
Z z=V(z2)
5-1/(1;) { FE
1
1 1
1 1
0
1 ' "
{ I
I
9 1 I
1 I
: 1
1 1
1
1/-$[1* - 1/(.:2] 1
0 .2:
Fig. 53
with k a positive constant.
Since Eq. (134) constitutes a particular
case of Eq. (126), it can be interpreted as
an equation describing the frictionless moe
tion in a straight line of an object of unit
mass under a restoring force equal to
167
z
!/
Fig. 53
with k a positive constant.
Since Eq. (134) constitutes a particular
case of Eq. (126), it can be interpreted as
an equation describing the frictionless m o ~ .
tion in a straight line of an object of unit
mass under a restoring force equal to 168 Differential Equations in Applications
-k sin :1:. In this case the autonomous
system corresponding to Eq. (134) is
d d .
a?=y, =-ksinx. (135)
The singular points are (O, O), (j rt, 0),
(;|; Zn, O), . ., and the differential equa-
tion of the phase trajectories of system (135)
assumes the form
dy __ _ k sin x
YE" `_'
Separating the variables and integrating,
we arrive at an equation for the phase tra-
jectories,
-%- yz-1+k (1 -cos ac) = E.
This equation is a particular case of Eq. (132)
with m = 1, where the potential energy
determined by (131) is specified by the
relationship
3
V(x) = S f()d==k(1cos:z:).
0
In the (x, z)plane we plot the function
z = V (x) as well as several straight lines
z = E (in Figure 54 only one such line,
z = E = 2k, is shown). After determining
a value of E - V (cv) we can draw a sketch
-k sin x. In this case the autonomous
system corresponding to Eq. (134) is
d.z dy
dt=y, dt= -ksmx.
(135)
The singular points are (0, 0), ( n, 0),
( 2n, 0), .. , and the differential equa-
tion of the phase trajectories of system (135)
assumes the form
dy _ k sin .z
d%---y-
Separating the variables and integrating,
we arrive at an equation for the phase tra-
jectories,
t
2
y
2
+k (1-cos x) =E.
This equation is a particular case ofEq. (132)
with m = 1, where the potential energy
determined by (131) is specified by the
relationship
lC
V(x) = ~ /(6) d6=k(1-cosx).
0
In the (x, z)-plane we plot the function
z = V (x) as well as several straight lines
s = E (in Figure 54 only one such line,
z = E = 2k, is shown). After determining
a valu\1 of E ...-. V (x) we can draw a s ~ e t c b Ch. 2. Qualitative Methods 169
z
z=_=2/{f _Wx) Z==V{2:)=kkc0sa:
' 1
{ 1
1 It I
I I
-5.rrI -2zr wl _ m Zyr Jyrl as
1 I I
1 { I 1
1 I 3/ I I
1 , I 1
| 1
, 1 * ~
\J\J\ / 1
Fig. 54
of the trajectories in the phase plane by
employing the relationship
y=:I:l/2[E-V()I
ch. 2. Qualitative Methods
169
z
ZVfa:)=k-kcosx
Fig. 54
of the trajectories in the phase plane by
employing the relationship
y= Y2 [E- V (x)]. 170 Difierential Equations in Applications
The resulting phase portrait shows (see
Figure 54) that if the energy E varies from O
to 2k, the corresponding phase trajectories
prove to be closed and Eq. (134) acquires
periodic solutions. On the other hand, if
E >2lc, the respective phase trajectories
are not closed and Eq. (134) has no periodic
solutions. Finally, the value E = 2lc corre-
sponds to a phase trajectory in the phase
plane that separates two types of motion,
that is, is a separatrix. The wavy lines
lying outside the separatrices correspond to
rotations of a pendulum, while the closed
trajectories lying inside the regions bou_nd-
ed by separatrices correspond to oscilla-
tions of the pendulum.
Figure 54 shows that in the vicinity of
the singular points (j2rcm, O), m = 0, 1,
2, . ., the behavior of the phase trajecto-
ries differs from that of the phase trajecto-
ries in the vicinity of the singular points
(_l_:rm, 0), where rz = 1, 2, . ..
There are different types of singular points.
With some we will get acquainted shortly.
As for the above example, the singular
points (5; 2:1tm, O), m = O, 1, 2, . . ., be-
long to the vorteaypoint type, while the
singular points (jam, 0), rz. == 1, 2, . .,
belong to the saddle-point type. A singular
point of an autonomous differential system
of the (122) type is said to be a vortex
point if there exists a neighborhood of this
point completely filled with nonintersecting
t 70 Differential Equations in Applications
The resulting phase portrait shows (see
Figure 54) that if the energy E varies from 0
to 2k, the corresponding phase trajectories
prove to be closed and Eq. (134) acquires
_periodic solutions. On the other hand, if
E > 2k, the respective phase trajectories
are not closed and Eq. (134) has no periodic
solutions. Finally, the value E = 2k corre-
sponds to a phase trajectory in the phase
plane that separates two types of motion,
that is, is a separatrix. The wavy lines
lying outside the separatrices correspond to
rotations of a pendulum, while the closed
trajectories lying inside the regions bound-
ed by separatrices correspond to oscilla-
tions of the pendulum.
Figure 54 shows that in the vicinity of
the singular points (2nm, 0), m = 0, 1,
2, .. , the behavior of the phase trajecto-
ries differs from that of the phase trajecto-
ries in the vicinity of the singtilar points
(nn, 0), where n = 1, 2, .
There are different types of singular points.
With some we will get acquainted shortly.
As for the above example, the singular
points ( 2nm, 0), m = 0, 1, 2, ... , be-
long to the vortex-point type, while the
singular points (+nn, 0), n = 1, 2, .. ,
belong to the saddle-point type. A singular
point of an autonomous differential system
of the (122) type is said to be a vortex
point if there exists a neighborhood of this
point completely filled with nonintersecting Ch. 2. Qualitative Methods 171
phase trajectories surrounding the point.
A saddle point is a singular point adjoined
by a finite number of phase trajectories
("whiskers") separating a neighborhood of
the singular point into regions where the
trajectories behave like a family of hyper-
bolas defined by the equation xy = const.
Now let us establish the effect of linear
friction on the behavior of the phase tra-
jectories of a conservative system. The
equation is
%%+cgli+ksinx=O, c>O.
lf friction is low, that is, the pendulum is
able to oscillate about the position of
equilibrium, it can be shown that the phase
trajectories are such as shown in Figure 55.
But if friction is so high that oscillations
become impossible, the pattern of phase
trajectories resembles the one depicted in
Figure 56.
If we now compare the phase portrait of
a conservative system with the last two por-
traits of nonconservative systems, we see
that saddle points have not changed (we
consider only small neighborhoods of singu-
lar points), while in the neighborhood of
the singular points (j2tm, O), m = O, 1,
2, . ., the closed phase trajectories have
transformed into spirals (for low friction)
or into trajectories that "enter" the singular
points in certain directions (for high fric-
171
phase trajectories surrounding the point.
A saddle point is a singular point adjoined
by a finite number of phase trajectories
("whiskers") separating a neighborhood of
the singular point into regions where the
trajectories behave like a family of hyper-
bolas defined by the equation xy = const.
Now let us establish the effect of linear
friction on the behavior of the phase tra-
jectories of a conservative syStem. The
equation is
dlz dz . O
dt' +cdt+ksmx= , c>O.
If friction is low, that is, the pendulum is
able to oscillate about the p o s i t i o ~ of
equilibrium, it can be shown tha:t th"Ei _pliase
trajectories arc such as shown in Figure 55.
But if friction is so high that oscillations
become impossible, the pattern of phase
trajectories resembles the one depicted in
Figure 56.
If we now compare the phase portrait of
a conservative system with the last two por-
traits of nonconservative systems, we see
that saddle points have not changed (we
consider only small neighborhoods of singu-
lar points), while in the neighborhood of
the singular points (2nm, 0), m = 0, 1,
2, .. , the closed phase trajectories have
transformed into spirals (for low friction)
or into trajectories that "enter" the singular
points in certain directions (for high fric-h
gy
5
l174 Differential Equations in Applications
tion). In the first case (spirals) we have
singular points of the focal-point type and
in the second, of the nodal-point type.
A singular point of a two-dimensional
autonomous differential system of the gen-
eral type (122) (if such a point exists)
is said to be a focal point if there exists
a neighborhood of this point that is com-
pletely filled with nonintersecting phase
trajectories resembling spirals that "wind"
onto the singular point either as t- +<x>
or as t-> -0o. A nodal point is a singular
point in whose neighborhood each phase
trajectory behaves like a branch of a parabo-
la or a half-line adjoining the point along
a certain direction.
Note that if a conservative system has
a periodic solution, the solution cannot be
isolated. More than that, if l` is a closed
phase trajectory corresponding to a periodic
solution of the conservative system, there
exists a certain neighborhood of l` that is
completely filled with closed phase tra-
jectories.
Note, in addition, that the above defi-
nitions of types of singular points have
a purely qualitative, descriptive nature.
As for the analytical features of these types,
there are no criteria, unfortunately, in
the general case of systems of the (122)
types, but for some classes of differential
equations such criteria can be formulated.
tion). In the first case (spirals) we have
singular points of the focal-point type and
in the second, of the nodal-point type.
A singular point of a two-dimensional
autonomous differential system of the gen-
eral type (122) (if such a point exists)
is said to be a focal point if there exists
a neighborhood of this point that is com-
pletely filled with nonintersecting phase
trajectories resembling spirals that "wind"
onto the singular point either as t-+ +oo
or as t-+ -oo. A nodal point is a singular
point in whose neighborhood each phase
trajectory behaves like a branch of a parabo-
la or a half-line adjoining the point along
a certain direction.
Note that if a conservative system has
a periodic solution, the solution cannot be
isolated. More than that, if r is a closed
phase trajectory corresponding to a periodic
solution of the conservative system, there
exists a certain neighborhood of r that is
completely filled with closed phase tra-
jectories.
Note, in addition, that the above defi-
nitions of types of singular points have
a purely qualitative, descriptive nature.
As for the analytical features of these types,
there are no criteria, unfortunately, in
the general case of systems of the (122)
types, but for some classes of differential
equations such criteria can be formulated. Ch. 2. Qualitative Methods 175
A simple example is the linear system
i%:a1$+b1!/ g'%=a2$+bzy
where al, bl, az, and bz are real constants.
If the coefficient matrix of this sytem is
nonsingular, that is, the determinant
at b1`
07
Iaz bz lf
the origin O (O, O) of the phase plane is
the only singular point of the differential
system.
Assuming the last inequality valid, we de-
note the eigenvalues of the coefficient ma-
trix by X1 and X2. It can then be demonstrat-
ed that
(1) if K1 and K2 are real and of the same
sign, the singular point is a nodal point,
(2) if X1 and K2 are real and of opposite
sign, the singular point is a saddle point,
(3) if X], and A2 are not real and are not
pure imaginary, the singular point is a fo-
cal point, and
(4) if kl and K2 are pure imaginary, the
singular point is a vortex point.
Note that the first three types of singular
points belong to the socalled coarse singu-
lar points, that is, singular points whose
nature is not affected by small perturba-
tions of the righthand sides of the initial
differential system. On the other hand,
175
A simple example is the linear system
dx dy
dt=a
1
x+b
1
y, Tt=a
2
x+b
2
y,
where a
1
, b
1
, a
2
, and b
2
are real constants.
If the coefficient matrix of this sytem is
nonsingular, that is, the determinant
the origin 0 (0, 0) of the phase plane is
the only singular point of the differential
system.
Assuming the last inequality valid, we de-
note the eigenvalues of the coefficient ma-
trix by A
1
and A
2
It can then be demonstrat-
ed that
(1) if A
1
and A
2
are real and of the same
sign, the singular point is a nodal point,
(2) if A
1
and A
2
are real and of opposite
sign, the singular point is a saddle point,
(3) if A
1
and A
2
are not real and are not
pure imaginary, the singular point is a fo-
cal point, and
(4) if A
1
and All are pure imaginary, the
singular point is a vortex point.
Note that the first three types of singular
points belong to the so-called coarse singu-
lar points, that is, singular points whose
nature is not affected by small perturba-
tions of the right-hand sides of the initial
differential system. On the other hand, 176 Differential Equations in Applications
a vortex point is a fine singular point; its
nature changes even under small perturba-
tions of the right-hand sides.
and of Periodic Motion
As we already know, singular points of
different types are characterized by different
patterns of the phase trajectories in sufficient-
ly small neighborhoods of these points. There
is also another characteristic, the stability
of an equilibrium point, which provides
additional information on the behavior of
phase trajectories in the neighborhood of
singular points. Consider the pendulum de-
picted in Figure 57. Two states of equilib-
rium are shown: (a) an object of mass m is
in a state of equilibrium at the uppermost
point, and (b) the same object is in a state
of equilibrium at its lowest point. The
first state is unstable and the second, stable.
And this is why. If the object is in its
uppermost state of equilibrium, a slight
push is enough to start it moving with an
ever increasing speed away from the equilib-
rium position and, hence, away from the
initial position. But if the object is in the
lowest possible state, a push makes it move
away from the position of equilibrium
with a decreasing speed, and the weaker the
push the smaller the distance by which the
a vortex point is a nne singular point; its
nature changes even under small perturba-
tions of the right-hand sides.
2.5 Stability of Equilibriwn Points
and of Periodic Motion
As we already know, singular points of
different types are characterized by different
patterns of the phase trajectories in sufficient-
ly small neighborhoods of these points. There
is also another characteristic, the stability
of an equilibrium point, which provides
additional information on the behavior of
phase trajectories in the neighborhood of
singular points. Consider the pendulum de-
picted in Figure 57. Two states of equilib-
rium are shown: (a) an object of mass m is
in a state of equilibrium at the uppermost
point, and (b) the same object is in a state
of equilibrium at its lowest point. The
ftrst state is unstable and the second, stable.
And this is why. If the object is in its
uppermost state of equilibrium, a slight
push is enough to start it moving with an
ever increasing speed away from the equilib-
rium position and, hence, away from the
initial position. But if the object is in the
lowest possible state, a push makes it move
away from the position of equilibrium
with a decreasing speed, and the weaker the
push the smaller the distance by which the Ch. 2. Qualitative Methods {77
I
I
I
I
I
I
I
Q;
Fig. 57
object is displaced from the initial posi-
tion.
The state of equilibrium of a physical
system corresponds to a singular point in
the phase plane. Small perturbations at an
unstable point of equilibrium lead to large
displacements from this point, while at
a stable point of equilibrium small pertur-
bations lead to small displacements. Start-
ing from these pictorial ideas, let us
consider an isolated singular point of sys-
tem (122), assuming for the sake of simplici-
ty that the point is at the origin O (O, O)
of the phase plane. We will say that this
singular point is stable if for every positive R
there exists a positive rg}? such that
12-0770
177
Fig. 57
object is displaced from the initial posi-
tion.
The state of equilibrium of a physical
system corresponds to a singular point in
the phase plane. Small perturbations at an
unstable point of equilibrium lead to large
displacements from this point, while at
a stable point of equilibrium small pertur-
bations lead to small displacements. Start-
ing from these pictorial ideas, let us
consider an isolated singular point of sys-
tem (122), assuming for the sake of simplici-
ty that the point is at the origin 0 (0, 0)
of the phase plane. We will say that this
singular point is stable if for every positive R
there exists a positive r ~ R such that
Nikolai G. Chetaev (1902-1959) wrote: *
...If a passenger plane is being designed, a
certain degree of stability must be provided
for in the future movements of the plane so
that it will be stable in {light and accident-
free during take-off and landing. The crank-
shaft must be so designed that it does not
break from the vibrations that appear in real
conditions of motor operation. To ensure that
an artillery gun has the highest possible ac-
curacy of aim and the smallest possible spread,
the gun and the projectiles must be con-
structed in such a manner that the trajectories
of projectile flight are stable and the projec-
tiles fly correctly.
Numerous examples can be added to this
list, and all will prove that real movements
require selecting out of the possible solutions
of the equations of motion only those that
correspond to stable states. Moreover, if we
wish to avoid a certain solution, it is advisable
to change the design of the corresponding
device in such a way that the state of motion
corresponding to this solution becomes un-
stable.
Returning to the pendulum depicted in
Figure 57, we note the following curious
and somewhat unexpected fact. Research
has shown that the upper (unstable) posi-
tion of equilibrium can be made stable by
* See N.G. Chetaev, Stability of Motion (Moscow:
Nauka, 1965: pp. 8-9 (in Russian).
i80 Differential Equations in Applications
Nikolai G. Chetaev (1902-1959) wrote: *
.. .If a passenger plane is being designed, a
certain degree of stability must be provided
for in the future movements of the plane so
that it will be stable in flight and accident-
free during take-off and landing. The crank-
shaft must be so designed that it does not
break from the vibrations that appear in real
conditions of motor operation. To ensure that
an artillery gun has the highest possible ac-
curacy of aim and the smallest possible spread,
the gun and the projectiles must be con-
structed in such a manner that the trajectories
of projectile flight are stable and the projec-
tiles fly correctly.
Numerous examples can be added to this
list, and all will prove that real movements
require selecting out of the possible solutions
of the equations of motion only those that
correspond to stable states. Moreover, if we
wish to avoid a certain solution, it is advisable
to change the design of the corresponding
device in such a way that the state of motion
corresponding to this solution becomes un-
stable.
Returning to the pendulum depicted in
Figure 57, we note the following curious
and somewhat unexpected fact. Research
has shown that the upper (unstable) posi-
tion of equilibrium can be made stable by
See N.G. Chetaev, Stability of Motton (Moscow:
Nauka, f965: pp. 8-9 (in Russian). Ch. 2. Qualitative Methods 181
introducing vertical oscillations of the
point of suspension. More than that,
not only the upper (vertical) position of
equilibrium can be made stable but also any
other of the pendulums positions (for one,
a horizontal position) by properly vibrating
the point of suspension. *
Now let us turn to a concept no less im-
portant than the stability of an equilibrium
point, the concept of stability of periodic
movements (solutions). Let us assume that
we are studying a conservative system that
has periodic solutions. In the phase plane
these solutions are represented by closed
trajectories that completely fill a certain
region. Thus, to each periodic motion of
a conservative system there corresponds
a motion of the representative point along
a fixed closed trajectory in the phase plane.
Generally, the period of traversal of
different trajectories by representative
points is different. In other words, the period
of oscillations in a conservative system
depends on the initial data. Geometrically
this means that two closely spaced repre-
sentative points that begin moving at
a certain moment t = to (say, at the x axis)
will move apart to a certain finite distance
* The reader can find many cxancplts cf stabili-
zation of different 15 pcs of pcndulums in the book
by T.G. Strizhak, Methods of Investigating Dynam-
ical Systems of the Pendulum Type (Alma-Ata:
Nauka, 1981) (in Russian).
181
introducing vertical oscillations of the
point of suspension. More than that,
not only the upper (vertical) position of
equilibrium can be made stable but also any
other of the pendulum's positions (for one,
a horizontal position) by properly vibrating
the point of suspension. *
Now let us turn to a concept no less im-
portant than the stability of an equilibrium
point, the concept of stability of periodic
movements (solutions). Let us assume that
we are studying a conservative system that
has periodic solutions. In the phase plane
these solutions are represented by closed
trajectories that completely fill a certain
region. Thus, to each periodic motion of
a conservative system there corresponds
a motion of the representative point along
a fixed closed trajectory in the phase plane.
Generally, the period of traversal of
different trajectories by representative
points is different. In other words, the period
of oscillations in a conservative system
depends on the initial data. Geometrically
this means that two closely spaced repre-
sentative points that begin moving at
a certain moment t = t
0
(say, at the x axis)
will move apart to a certain finite distance
* The nader can find many rxarrplc s d elal:ili-
zation of different l) pes flf pcnd1Jh1rns in !he book
by T.G. Strizhak, Methods of Investigating Dynam-
ical Systems of the "Pendulum" Type (Alma-Ata:
Nauka, 1981) (in Hussian). 178 Differential Equations in Applications
.9
V
WI W
`J
Fig. 58
every phase trajectory originating at the
initial moment t = to at a point P lying
inside the circle :1:2 i- y2 = r2 will lie
inside the circle :::2 + y2 = R2 for all t > to
(Figure 58). Without adhering to rigorous
reasoning we can say that a singular point
is stable if all phase trajectories that are
near the point initially remain there with
the passage of time. Also, a singular point
is said to be asymptotically stable if it is
stable and if there exists a circle x2 I
y2 = rf, such that each trajectory that at
time t = to lies inside the circle converges
to the origin as t-> {oo. Finally, if
a singular point is not stable, it is said to
be unstable.
!I
Fig. 58
every phase trajectory originating at the
initial moment t = t
0
at a point P lying
inside the circle x
2
+ y
2
= r
2
will lie
inside the circle x
2
+ y
2
= R
2
for all t > t
0
(Figure 58). Without adhering to rigorous
reasoning we can say that a singular point
is stable if all phase trajectories that are
near the point initially remain there with
the passage of time. Also, a singular point
is said to be asymptotically stable if it is
stable and if th('re exists a circle x
2
+
y
2
= r: such that each trajectory that at
time t = t
0
lies inside the circle converges
to the origin as t-+ +oo. Finally, if
a singular point is not stable, it is said to
be unstable. Ch. 2. Qualitative Methods 179
A singular point of the vortex type is
always stable (but not asymptotically).
A saddle point is always unstable. In
Figure 55, which illustrates the behavior
of the phase trajectories for pendulum oscil-
lations in a medium with low drag, the sin-
gular points, focal points, are asymptotical-
ly stable; in Figure 56 the singular points,
which are nodal points, are also asymptoti-
cally stable.
The introduced concept of stability of an
equilibrium point is purely qualitative,
since no mention of properties referring to
the behavior of phase trajectories has been
made. As for the concept of asymptotic
stability, if compared with the notion of
simple stability, it is additionally neces-
sary that every phase trajectory tend to
the origin with the passage of time. How-
ever, in this case, too, no conditions are
imposed on how the phase trajectory must
approach point O (O, O).
The concepts of stability and asymptotic
stability play an important role in appli-
cations. The fact is that if a device is
designed without due regard for stability
considerations, when built it will be sensi-
tive to the very smallest external perturba-
tions, which in the iinal analysis may lead
to extremely unpleasant consequences. Em-
phasizing the importance of the concept of
stability, the well-known Soviet specialist
in the field of mathematics and mechanics
12 Ch. 2. Qualitative Methods f79
A singular point of the vortex type is
always stable (but not asymptotically).
A saddle point is always unstable. In
Figure 55, which illustrates the behavior
of the phase trajectories for pendulum oscil-
lations in a medium with low drag, the sin-
gular points, focal points, are asymptotical-
ly stable; in Figure 56 the singular points,
which are nodal points, are also asymptoti-
cally stable.
The introduced Mncept of stability of an
equilibrium point is purely qualitative,
since no mention of properties referring to
the behavior of phase trajectories has been
made. As for the concept of asymptotic
stability, if compared with the notion of
simple stability, it is additionally neces-
sary that every phase trajectory tend to
the origin with the passage of time. How-
ever, in this case, too, no conditions are
imposed on how the phase trajectory must
approach point 0 (0, 0).
The concepts of stability and asymptotic
stability play an important role in appli-
cations. The fact is that if a device is
designed without due regard for stability
considerations, when built it will be sensi-
tive to the very smallest external perturba-
tions, which in the final analysis may lead
to extremely unpleasant consequences. Em-
phasizing the importance of the concept of
stability, the well-known Soviet specialist
in the field of mathematics and mechanics
12 182 Differential Equations in Applications
y 8
(:*2
P5 8
Y} g
lìg. 59
with the passage of time. But it also may
happen that these points do not separate.
To distinguish between these two possibil-
ities, the concept of stability in the sense
of Lyapunov is introduced for periodic
solutions. The essence of this concept lies
in the following. If knowing an a-neighbor-
hood (with a as small as desired) of a point
M moving along a closed trajectory I`
(Figure 59) * ensures that we: know a mov-
ing 6 (s)-neighborhood of the same point M
such that every representative point that
initially lies in the 6 (s)-neighborhood will
never leave the e-neighborhood with the
passage of time, then the periodic solution
* An sneighborhood of a point M is understood
to be a disk of radius 6 centered at point M,
f82 Equations in Applications
!I
l'ig. 59
with the passage of time. But it also may
happen that these points do not separate.
To distinguish between these two possibil-
ities, the concept of stability in the sense
of Lyapunov is introduced for periodic
solutions. The essence of this concept lies
in the following. If knowing an e-neighbor-
hood (with e as small as desired) of a point
M moving along a closed trajectory r
(Figure 59) * ensures that we: know a mov-
ing B (e)-neighborhood of the same point M
such that every representative point that
initially lies in the B (e)-neighborhood will
never leave the with the
passage of time, then the periodic solution
An -neighborhood of a point M is understood
to be a disk of radius e centered at point M. Ch. 2. Qualitative Methods 183
corresponding to l` is said to be stable in
the sense of Lyapunov. If a periodic solution
is not stable in the sense of Lyapunov, it is
said to be unstable in the sense of Lyapunov.
When it comes to periodic solutions that
are unstable in the sense of Lyapunov, we
must bear in mind that they still possess
some sort of stability, orbital stability,
which means that under small variations of
initial data the representative point trans-
fers from one phase trajectory to another
lying as close as desired to the initially
considered trajectory.
Examples of periodic solutions that are
stable in the sense of Lyapunov are those
that emerge, for instance, when we consider
the differential equation that describes the
horizontal movements of an object of mass
ni in a vacuum with two linear springs acting
on the object (Figure 52). An example of
periodic solutions that are unstable in the
sense of Lyapunov but are orbitally stable
is the solutions of the differential equa-
tion (134), which describes the motion of
a circular pendulum in a medium without
drag.
In the first case the oscillation period
does not depend on the initial data and is
found by using the formula T = 2nl/m/lc.
In the second the oscillation period depends
on the initial data and is expressed, as we
know, in terms of an elliptic integral of the
fnrst kind taken from 0 to at/2.
f83
corresponding to r is said to be stable in
the sense of Lyapunov. If a periodic solution
is not stable in the sense of Lyapunov, it is
said to be unstable in the sense of Lyapunov.
When it comes to periodic solutions that
are unstable in the sense of Lyapunov, we
must bear in mind that they still possess
some sort of stability, orbital stability
1
which means that under small variations of
initial data the representative point trans-
fers from one phase trajectory to another
lying as close as desired to the initially
considered trajectory.
Examples of periodic solutions that are
stable in the sense of Lyapunov are those
that emerge, for instance, when we consider
the differential equation that describes the
horizontal movements of an object of mass
m in a vacuum with two linear springs acting
on the object (Figure 52). An example of
periodic solutions that are unstable in the
sense of Lyapunov but are orbitally stable
is the solutions of the differential equa-
tion (134), which describes the motion of
a circular pendulum in a medium without
drag.
In the first case the oscillation period
does not depend on the initial data and is
found by using the formula T = 2nV mlk.
In the second the oscillation period depends
on the initial data and is expressed, as we
know, in terms ofan elliptic integral of the
first kind taken from 0 to 1(:/2, 184 Differential Equations in Applications
Finally, we note that the question of
whether periodic movements are stable in
the sense of Lyapunov is directly linked
to the question of isochronous vibrations. *
2.6 Lyapunov Functions
Intuitively it is clear that if the total
energy of a physical system is at its mini-
mum at a point of equilibrium, the point
is one of stable equilibrium. This idea lies
at the base of one of two methods used in
studying stability problems, both suggested
by the famous Russian mathematician Alek-
sandr M. Lyapunov (1857-1918). This meth-
od is known as Lyapunovs direct, or
second, method for stability investigations. **
We illustrate Lyapunovs direct method
using the (122) type of system when the
origin is a singular point.
Suppose that l` is a phase trajectory of
system (122). We consider a function V =
V (x, y) that is continuous together with its
first partial derivatives 0V/6x and 6V/6y
* See, for example, the book by V. V. Amelkin,
N. A. Lukashevich, and A.P. Sadovskii, Non-
linear Vibrations in Second-Order Systems (Minsk:
Belorussian Univ. Press, 1982) (in Russian).
** The reader will End many interesting examples
of stability investigations involving differential
models in the book by N. Rouche, P. Habets, and
M. Laloy, Stability Theory by Lyapunov`s Direct
Method (New York: Springer, 1977).
Finally, we note that the question of
whether periodic movements are stable in
the sense of Lyapunov is directly linked
to the question of isochronous vibrations. *
2.6 Lyapunov Functions
Intuitively it is clear that if the total
energy of a physical system is at its mini-
mum at a point of equilibrium, the point
is one of stable equilibrium. This idea lies
at the base of one of two methods used in
studying stability problems, both suggested
by the famous Russian mathematician Alek-
sandr M. Lyapunov (1857-1918). This meth-
od is known as Lyapunov's direct, or
second, method for stability investigations.**
We illustrate Lyapunov's direct method
using the (122) type of system when the
origin is a singular point.
Suppose that r is a phase trajectory of
system (122). We consider a function V =
V (x, y) that is continuous together with its
first partial derivatives 8V/8x and aVJay
See, for example, the book by V. V. Amel' kin,
N. A. Lukashevich, and A.P. Sadovskii, Non-
linear Vibrations tn Second-Order Systems (Minsk:
Belorussian Univ. Press, 1982) (in Russian).
The reader will find many interesting examples
of stability investigations involving differential
models in the book by N. Rouche, P. Habets, and
M. Laloy, Stability Theory by Lyapunov's Direct
Method (New York: Springer, 1977). Ch. 2. Qualitative Methods 185
in a domain containing I` in the phase plane.
If the representative point (sc (t), y (1:))
moves along curve I`, then on this curve
the function V (x, y) may be considered
a function of t, with the result that the
rate at which V (:1:, y) varies along l` is
dV GV dx OV dy _ OV
g;=g jgl"5,; gX($ H)
where X (.2:, y) and Y (sc, y) are the right-
hand sides of system (122).
Formula (136) is essential in the realiza-
tion of Lyapunovs direct method. The
following concepts are important for the
practical application of the method.
Suppose that V = V(x, y) is continu-
ous together with its first partial deriva-
tives 6V/Ox and 0V/dy in a domain G con-
taining the origin in the phase plane, with
V (0, O) = O. This function is said to be
positive (negative) definite if at all points
of G except the origin V (sc, y) it is positive
(negative). But if at points of G we have
V (sc, y)/> 0 (Q O), the function V = V (1:,
y) is said to be nonnegative (nonpositive).
For example, the function V defined by the
formula V (sc, y) = x2 + y2 and considered
in the (az, y)-plane is positive definite, while
the function V (as, y) = $2 is nonnegative
since it vanishes on the entire y axis.
185
in a domain containing r in the phase plane.
If the point (x (t), y (t))
moves along curve r, then on this curve
the function V (x, y) may be considered
a function of t, with the result that the
rate at which V (x, y) varies along r is
y)
dt - iJx dt iJy dt - iJx '
ov
+ay Y (x, y), (136)
where X (x, y) and Y (x, y) are the right-
hand sides of system (122).
Formula (136) is essential in the realiza-
tion of Lyapunov's direct method. The
following concepts are important for the
practical application of the method.
Suppose that V = V (x, y) is continu-
ous together with its first partial deriva-
tives avtax and avtay in a domain G con-
taining the origin in the phase plane, with
V (0, 0) = 0. This function is said to be
positive (negative) definite if at all points
of G except the origin V (x, y) it is positive
(negative). But if at points of G we have
V (x, y) 0 (::::;;; 0), the function V = V (x,
y) is said to be nonnegative (nonpositive).
For example, the function V defined by the
formula V (x, y) = x
2
+ y
2
and considered
in the (x, y)-plane is positive definite, while
the function V (x, y) = x
2
is nonnegative
since it vanishes on the entire y axis. 186 Differential Equations in Applications
If V (x, y) is positive definite, we can
require that at all points of the domain
G\O the following inequality hold true:
OV 2 0V 2
(H) HE) *
This means that the equation z = V (x, y)
can be interpreted as the equation of a sur-
face resembling a paraboloid that touches
the (x,y)-plane at point O (0,0) (Figure
60). Generally, the equation z = V (ay, y)
with a positive definite V may specify
a surface of a more complex structure. One
such surface is shown in Figure 61, where
the section of the surface with the plane
z = C results not in a curve but in a ring.
lf a positive definite function V (0:, y) is
such that
Wo, y>=2-,%x<x, y>
+ii-ii Y as y> we
is nonpositive, V is said to be the Lyapunov
function of system (122). We note here that
in view of (136) the requirement that Wbe
nonpositive means that dV/dtg 0 and,
hence, the function V = V (zz, y) does not
increase along the trajectory I` in the
neighborhood of the origin.
Here is a result arrived at by Lyapunov:
if for system (122) there exists a Lyapunov
function V (x, y), then the origin, which is
If V (x, y) is positive definite, we can
require that at all points of the domain
G"'-0 the following inequality hold true:
( ~ ~ ) 2 + ( : ) 2 * o.
This means that the equation z = V (x, y)
can be interpreted as the equation of a sur-
face resembling a paraboloid that touches
the (x, y)-plane at point 0 (0, 0) (Figure
60). Generally, the equation z = V (x, y)
with a positive definite V may specify
a surface of a more complex structure. One
such surface is shown in Figure 61, where
the section of the surface with the plane
z = C results not in a curve but in a ring.
If a positive definite function V (x, y) is
such that
W (x, y) = iJV ~ ~ y) X (x, y)
+ av ~ y) y (x, y)
(137)
is nonpositive, V is said to be the Lyapunov
function of system (122). We note here that
in view of (136) the requirement that W be
non positive means that d Vldt 0 and,
hence, the function V = V (x, y) does not
increase along the trajectory r in the
neighborhood of the origin.
Here is a result arrived at by Lyapunov:
if for system (122) there exists a Lyapunov
functton- V (x, l/)
1
then the origin, wMc4 ts Ch. 2. Qualitative Methods {87
z
cv 0 5/
Fig. 60
z .
W
lI\\& y
Fig. 61
a singular point, is stable. If the positive
definite function V = V (sc, y) is such that
function Wdefined via (137) is negative defi-
nite, then the origin is asymptotically stable.
We will shew with an example how to
187
z
Fig. 60
Fig. 61
a singular point, is stable. If the positive
definite function V = V (x, y) is such that
function W defined via (137) is negative defi-
nite, then the origin is stable.
We will with an example .hQw tQ 188 DiHerential Equations in Applications
apply the above result. Let us consider the
equation of motion of an object of unit
mass under a force exerted by a spring,
which in view of (124) can be written in the
form
-(3;;;+clg-?+kx=0, c>0. (138)
The reader will recall that in this equation
c > 0 characterizes the drag of the medium
in which the object moves and k > 0
characterizes the properties of the spring
(the spring constant). The autonomous sys-
tem corresponding to Eq. (138) has the
form
-3-?=y, $4:-kx-cy. (139)
For this system the origin of the phase
(x, y)-plane is the only singular point. The
kinetic energy of the object (of unit mass)
is y2/2 and the potential energy (i.e. the
energy stored by the spring) is
S kg dg Z TQ- azz.
0
This implies that the total energy of the
system is
VMS, y)=% y2+ M2- (140)
It is easy to See that V specified by (140)
apply the above result. Let us consider the
equation of motion of an object of unit
mass under a force exerted by a spring,
which in view of (124) can be written in the
form
d
8
z dz
dt +cdt+kx=O, c> 0. (138)
The reader will recall that in this equation
c > 0 characterizes the drag of the medium
in which the object moves and k > 0
characterizes the properties of the spring
(the spring constant). The autonomous sys-
tem corresponding to Eq. (138) has the
form
dz dy
Tt=y, dt= -kx-cy.
(139)
For this system the origin of the phase
(x, y)-plane is the only singular point. The
kinetic energy of the object (of unit mass)
is y
2
/2 and the potential energy (i.e. the
energy stored by the spring) is
X
~ ks ds = +kx
2

0
This implies that the total energy of the
system is
1 1
V (x, y) = T y
2
+ T kx
2
(140)
It is easy to see that V specified by (140) Ch. 2. Qualitative Methods 189
is a positive definite function. And since
in the given case
0
-g-X( y>+;,g-Y< y)
=/~y+y(k>-vy>= cy2<0.
this function is the Lyapunov function of
system (139), which means that the singu-
lar point O (O, O) is stable.
In the above example the result was
obtained fairly quickly. This is not always
the case, however. The formulated Lyapu-
nov criterion is purely qualitative and
this does not provide a procedure for
Ending the Lyapunov function even if we
know that such a function exists. This
makes it much more difficult to determine
whether a concrete system is stable or not.
The reader must bear in mind that the
above criterion of Lyapunov must be seen
as a device for finding effective indications
of equilibrium. Many studies have been
devoted to this problem and a number of
interesting results have been obtained in
recent yea1s.*
2.7 Simple States of Equilibrium
The dynamical interpretation of second-
order differential equations already implies
" The interested reader can refer to the book by
E.A. Barbashin, Lyapunov Functions (Moscow:
Nauka, 1970) (in Russian).
1S9
is a positive definite function. And since
in the given case
OV 8V
TxX (x, y) + ay Y (x, y)
= kxy + y (-kx- cy) = - cy2 0,
this function is the Lyapunov function of
system (139), which means that the singu-
lar point 0 (0, 0) is stable.
In the above example the result was
obtained fairly quickly. This is not always
the case, however. The formulated Lyapu-
nov criterion is purely qualitative and
this does not provide a procedure for
finding the Lyapunov function even if we
know that such a function exists. This
makes it much more difficult to determine
whether a concrete system is stable or not.
The reader must bear in mind that the
above criterion of Lyapunov must be seen
as a device for finding effective indications
of equilibrium. Many studies have been
devoted to this problem and a number of
interesting results have been obtained in
recent years.*
2. 7 Simple States of Equilibrium
The dynamical interpretation of second-
order differential equations already implies
The interested reader can refer to the book by
E.A. Barbashin, Lyapunov Functions (Moscow:
Nauka, 1970) (in Russian). 190 Differential Equations in Applications
that investigation of the nature of equilibri-
um states or, which is the same, the singular
points provides a key for establishing the
behavior of integral curves.
It is also clear that, generally speaking,
differential equations cannot be integrated
in closed form. We need criteria that will
enable us to determine the type of a sin-
gular point from the form of the initial
differential equation. Unfortunately, as
a rule it is extremely difficult to {ind such
criteria, but it is possible to isolate certain
classes of differential equations for which
this can be done fairly easily. Below we
show, using the example of an object of unit
mass subjected to the action of linear
springs and moving in a medium with
linear drag, how some results of the quali-
tative theory of differential equations can
be used to this end. But first let us discuss
a system of the (122) type. It so happens
that the simplest case in establishing the
type of a singular point is when the Jacobian
or functional determinant
6X 6X
T9? 75]
J ("* y): ay ar
"`6T T9?
is nonzero at the point.
If (x* , y*) is a singular point of system (122)
and if J* = J (x*, y*) ak 0, then the type
of the singular point, which in the given
that investigation of the nature of equilibri-
um states or, which is the same, the singular
points provides a key for establishing the
behavior of integral curves.
It is also clear that, generally speaking,
differential equations cannot be integrated
in closed form. We need criteria that will
enable us to determine the type of a sin-
gular point from the form of the initial
differential equation. Unfortunately, as
a rule it is extremely difficult to find such
criteria, but it is possible to isolate certain
classes of differential equations for which
this can be done fairly easily. Below we
show, using the example of an object of unit
mass subjected to the action of linear
springs and moving in a medium with
linear drag, how some results of the quali-
tative theory of differential equations can
be used to this end. But first let us discuss
a system of the (122) type. It so happens
that the simplest case in establishing the
type of a singular point is when the Jacobian
or functional determinant
ax ax
OZ"Ty
J (x, Y) = ay ay
8Z 7iii
is nonzero at the point.
If (x*, y*) is a singular point of system (122)
and if J* = J (x*, y*) =I= 0, then the type
of the singular point, which in the given Ch. 2. Qualitative Methods 191
case is called a simple singular point,
depends largely on the sign of constant J *.
For instance, if J * is negative, the singular
point (x*, y*) is a saddle point, and if .I`* is
positive, the singular point may be a vor-
tex point, a nodal point, or a focal point.
The singular point may be a vortex point
only if the divergence
6X GY
D(w, y) =q;+g,-7
vanishes at the singular point, that is, only
if D* = D (;z:*, y*) = O. Note, however,
that the condition D* = 0 is generally
insufficient for the singular point (x*, y*) to
be a vortex point. For a vortex point to
be present certain additional conditions
must be met, conditions that include high-
er partial derivatives. And, generally,
there can be an infinite number of such
conditions. But if functions X and Y are
linear in variables as and y, the condition
D* = 0 becomes sufficient for the singular
point (.2:*, y*) to be a vortex point.
If J * > 0 but the singular point (42:*, y*)
is not a vortex point, a sufficiently small
neighborhood of this point is filled with
trajectories that either spiral into this point
or converge to it in certain directions. Here,
if D* > 0, the singular point is reached as
i> --o and is unstable while if D* < 0,
the singular point is reached as i-> +o
and proves to be stable. If the phase tra-
191
case is called a simple singular point,
depends largely on the sign of constant J*.
For instance, if J* is negative, the singular
point (x*, y*) is a saddle point, and if J* is
positive, the singular point may be a vor-
tex point, a nodal point, or a focal point.
The singular point may be a vortex point
only if the divergence
ax ay
D (x, y) = az+ay
vanishes at the singular point, that is, only
if D* = D (x*, y*) = 0. Note, however,
that the condition D* = 0 is generally
insufficient for the singular point (x*, y*) to
be a vortex point. For a vortex point to
be present certain additional conditions
must be met, conditions that include high-
er partial derivatives. And, generally,
there can be an infinite number of such
conditions. But if functions X and Y are
linear in variables x and y, the condition
D* = 0 becomes sufficient for the singular
point (x*, y*) to be a vortex point.
If J* > 0 but the singular point (x*, y*)
is not a vortex point, a sufficiently small
neighborhood of this point is filled with
trajectories that either spiral into this point
or converge to it in certain directions. Here,
if D* > 0, the singular point is reached as
t-+ -oo and is unstable while if D* < 0,
the singular point is reached as t-+ +oo
and proves to be stable. If the phase tra-192 Differential Equations in Applications
jectories that reach the singular point are
spirals, we are dealing with a focal point,
but if the integral curves converge to a sin-
gular point along a certain tangent, the
point is a nodal point (Figure 62).
Irrespective of the sign of Jacobian .}*,
the tangents to the trajectories of the differ-
ential system (122) at a singular point (x*,
y*) can be found from what is known as the
characteristic equation
0Y(=*. y*) ~ 6Y(>*. y*)~
tg; Bx x + Oy y
141)
Q; 6X (=* y*) ~ 6X (=* y*) ~ (
0:1: w+ 6y y
where
,.;-:;::-.10*, ;7=yy*. (142)
If X and Y contain linear terms, the
partial derivatives in Eq. (141) act as co-
efficients of x and y in the system obtained
from system (122) after introducing the
substitution (142).
Equation (141) is homogeneous. Hence,
if we introduce the slope 2. =. of what is
known as exceptional directions, we have
the following quadratic equation for find-
ing K:
XQK2 + (X; Y;) A. Y; = O. (143)
The discriminant of this equation is
A = (X; lY;)2 4.}* == D*2 4.}*.
t92 Differential Equations in Applications
jectories that reach the singular point are
spirals, we are dealing with a focal point,
but if the integral curves converge to a sin-
gular point along a certain tangent, the
point is a nodal point (Figure 62).
Irrespective of the sign of Jacobian J*,
the tangents to the trajectories of the differ-
ential system (122) at a singular point (x*,
y*) can be found from what is known as the
characteristic equation
ay (.z*, y"'),.., ay (.z*, y"'),..,
y a.z z + ay Y
; =ax (z"', y"'),.., ax (.z"', y*) ,..,
az z+ ay y
(141)
where
; = x - x*, y = y - y*. (142)
If X and Y contain linear terms, the
partial derivatives in Eq. (141) act as co-
efficients of x and y in the system obtained
from system (122) after introducing the
substitution (142).
Equation (141) is homogeneous. Hence,
if we introduce the slope A = Y/; of what is
known as exceptional directions, we have
the following quadratic equation for find-
ing A:
XZA
2
+ (X: - YZ) A - Y: = 0. (143)
The discriminant of this equation is
a= (X:+ YZ)
2
- 4J* = D*
2
- 4J*. Q
Ik
% Hg
I
I
R
' $-3
CQ __ * x
I QS"
I
I ~>`
I
*35
Q
va __ IL]
I Q
I
I
I ~m
~ .2%
Ln
13-0770
~ - -
1
..., ...
~ - -
1
13-0770 194 Differential lriquations in Applications
Hence, if J* <O, which corresponds to
a saddle point, Eq. (143) always gives two
real exceptional directions. But if J * > 0,
there are no real exceptional directions,
or there are two, or there is one 2fold
direction. In the iirst case the singular
point is either a vortex point or a focal point.
The existence of real exceptional direc-
tions means (provided that J * is positive)
that there is a singular point of the nodal
type. For one thing, if there are two real
exceptional directions, it can be proved
that there are exactly two trajectories (one
on each side) whose tangent at the singular
point is one of the exceptional straight lines
(directions) while all the other trajecto-
ries "enter" the singular point touching the
other exceptional straight line (Figure 62a).
If A = O and Eq. (143) is not an identity,
we have only one exceptional straight line.
The pattern of the trajectories for this case
is illustrated by Figure 62b. It can be
obtained from the previous case when the
two exceptional directions coincide. The
singular point divides the exceptional
straight line into two halflines, ll and Z2,
while the neighborhood of the singular point
is divided into two sectors, one of which is
completely filled with trajectories that
"enter" the singular point and touch ll and
the other is completely filled with trajecto-
ries that "enter" the singular point and
touch Z2. The boundary between the sectors
194 Differential bquations in Applications
Hence, if J* < 0, which corresponds to
a saddle point, Eq. (143) always gives two
real exceptional directions. But if J* > 0,
there are no real exceptional directions,
or there are two, or there is one 2-fold
direction. In the first case the singular
point is either a vortex point or a focal point.
The existence of real exceptional direc-
tions means (provided that J* is positive)
that there is a singular point of the nodal
type. For one thing, if there are two real
exceptional direetions, it can be proved
that there are exactly two trajectories (one
on each side) whose tangent at the singular
point is one of the exceptional straight lines
(directions) while all the other trajecto-
ries "enter" the singular point touehing the
other exceptional straight line (Figure 62a).
I f ~ = 0 and Eq. (143) is not an identity,
we have only one exceptional straight line.
The pattern of the trajectories for this case
is illustrated by Figure 62b. It can be
obtained from the previous case when the
two exceptional directions coincide. The
singular point divides the exceptional
straight line into two half-lines, 1
1
and l
2
,
while the neighborhood of the singular point
is divided into two sectors, one of which is
completely filled with trajectories that
"enter" the singular point and touch l
1
and
the other is completely filled with trajecto-
ries thnt "enter" the singular point and
to11ch l
2
The boundary between the sectors Ch. 2. Qualitative Methods 195
consists of two trajectories, one of which
touches I1 at the singular point and the other
touches L2 at this point.
lf in Eq. (143) all coeflicients vanish, we
arrive at an identity, and then all the
straight lines passing through the singular
point are exceptional and there are exactly
two trajectories (one on each side) that
touch each of these straight lines at the
singular point. This point (Figure 62c) is
similar to the point with one 2-fold real
exceptional direction.
2.8 Motion of a UnitMass Object
Under the Action of Linear Springs
in a Medium with Linear Drag
As demonstrated earlier, the differential
equation describing the motion of a unit-
mass object under the action of linear
springs in a medium with linear drag has
the form
2
.}E;i+c%+kx=o. (144)
So as not to restrict the differential model
(144) to particular cases we will not tix the
directions in which the forces c (dx/dt)
and -kx act. As shown earlier, with
Eq. (144) we can associate an autonomous
system of the form
-=y, -%=kxcy. (145)
13*
f95
consists of two trajectories, one of which

L -an'cttne 6tner
touches l
2
at this point.
If in Eq. (143) all coeflicients vanish, we
arrive at an identity, and then all the
straight lines passing through the singular
point are exceptional and there are exactly
two trajectories (one on each side) that
touch each of these straight lines at the
singular point. This point (Figure 62c) is
similar to the point with one 2-fold real
exceptional direction.
2.8 Motion of a Unit-Mass Object
Under l:he Action of Linear Springs
in a Medium with Linear Drag
As demonstrated earlier, the differential
equation describing the motion of a unit-
mass object under the action of linear
springs in a medium with linear drag has
the form
dllz dz
dtz +cdt+kx=O.
(144)
So as not to restrict the differential model
(144) to particular cases we will not fix the
directions in which the forces -c (dx/dt)
and -kx act. As shown earlier, with
Eq. (144) we can associate an autonomous
system of the form
dz dy k
dt=y, dt=- x-cy.
(145)
If we now exclude the trivial case with
lc = O, which we assume is true for the
meantime, the differential system (145) has
an isolated singular point at the origin.
System (145) is a particular case of the
general system (122). In our concrete exam-
ple
the Jacobian J (:1:, y) = lc, and the divergence
D (xr, y) : c. The characteristic equa-
tion assumes the form
X2-{-cli.-|-k=0,
where A = cg 4k is the discriminant of
this equation. In accordance with the results
obtained in Section 2.7 we arrive at the
following cases.
(1) lf lc is negative, the singular point is
a saddle point with one positive and one
negative exceptional direction. The phase
trajectory pattern is illustrated in Fig-
ure 63, where we can distinguish between
three different types of motion. When the
initial conditions correspond to point a,
at which the velocity vector is directed to
the origin and the velocity is sufficiently
great, the representative point moves along
a trajectory toward the singular point at
a decreasing speed; after passing the origin
the representative point moves away from
it at an increasing speed. If the initial
velocity decreases to a critical value, which
i96 DifferPnlial in Applications
If we now exclude the trivial case with
k = 0, which we assume is true for the
meantime, the differential system (145) has
an isolated singular point at the origin.
System (145) is a particular case of the
general system (122). In our concrete exam-
ple
X (x, y) = y, Y (x, y) = -kx - cy,
the Jacobian J (x, y) = k, and the divergence
D (x, y) = -c. The characteristic equa-
tion assumes the form
1..
2
+cA. +k= 0,
where !J. = c
2
- 4k is the discriminant of
this equation. In accordance with the results
obtained in Section 2. 7 we arrive at the
following cases.
(1) If k is negative, the singular point is
a saddle point with one positive and one
negative exceptional direction. The phase
trajectory pattern is illustrated in Fig-
ure 63, where we can distinguish between
three different types of motion. When the
initial conditions correspond to point a,
at which the velocity vector is directed to
the origin and the velocity is sufficiently
great, the representative point moves along
a trajectory toward the singular point at
a decreasing speed; after passing the origin
the representative point moves away from
it at an increasing speed. If the initial
velocity decreases to a critical value, which Ch. 2. Qualitative Methods 197
y=.i:
0
E`
` C
a
Fig. 63
corresponds to point b, the representative
point approaches the singular point at
a decreasing speed and "reaches" the origin
in an infinitely long time interval. Finally,
if the initial velocity is lower than the
critical value and corresponds, say, to point
C, the representative point approaches the
origin at a decreasing speed, which vanishes
at a certain distance xl from the origin.
At point (xl, O) the velocity vector reverses
its direction and the representative point
moves away from the origin.
If the phase point corresponding to the
initial state of the dynamical system lies
in either one of the other three quadrants,
the interpretation of the motion is obvious.
197
Fig. 63
corresponds to point b, the representative
point approaches the singular point at
a decreasing speed and "reaches" the origin
in an infinitely long time interval. Finally,
if the initial velocity is lower than the
critical value and corresponds, say, to point
C, the representative point approaches the
origin at a decreasing speed, which vanishes
at a certain distance x
1
from the origin.
At point (x
1
, 0) the velocity vector reverses
its direction and the representative point
moves away from the origin.
If the phase point corresponding to the
initial state of the dynamical system lies
in either one of the other three quadrants,
the interpretation of the motion is obvious. 198 Differential Equations in Applications
yar:
x
Fig. 64
fj
Fig. 65
(2) If lc > O, then J * is positive and the
type of singular point depends on the
value of c. This leaves us with the fol-
lowing possibilities:
(2a) lf c = O, that is, drag is nil, the
singular point is a vortex point (Figure 64).
The movements are periodic and their
amplitude depends on the initial condi-
tions.
Fig. 64
Fig. 65
(2) If k > 0, then J* is positive and the
type of singular point depends on the
value of c. This leaves ns with the fol-
lowing possibilities:
(2a) If c = 0, that is, drag is nil, the
singular point is a vortex point (Figure 64).
The movements are periodic and their
amplitude depends on the initial condi-
tions. Ch. 2. Qualitative Methods 199
(2b) If c > O, that is, damping is posi-
tive, the divergence D : OX/0.2; + GY/fiy is
negative and, hence, the representative
point moves along a trajectory toward the
origi11 and reaches it in an infinitely long
time interval.
More precisely:
(2b,) If A <O, that is, cg < 4k, the sin--
gular point proves to be a focal point
(Figure 65) and, hence, the dynamical
system performs damped oscillations about
the state of equilibrium with a decaying
amplitude.
(2b2) If A :0, that is, c2 :4k, the
singular point is a nodal point with a single
negative exceptional direction (Figure 66).
The motion in this case is aperiodic and cor-
responds to the socalled critical damping.
(2b,,) If A > O, that is, cg > 4k, the
singular point is a nodal point with two
negative exceptional directions (Figure 67).
Qualitatively the motion of the dynamical
system is the same as in the previous case
and corresponds to damped oscillations.
From the above results it follows that
when c >O and k>O, that is, drag is pos-
itive and the restoring force is attractive,
the dynamical system tends to a state of
equilibrium and its motion is stable.
(2c) lf c < O, that is, damping is nega-
tive, the qualitative pattern of the phase
trajectories is the same as in the case (2b),
the only difference being that here the
199
(2b) If c > 0, that is, damping is posi-
tive, the divergence D = aX/8:-c + aY!oy is
negative and, hence, the representative
origin and reaches it in an infmitely long
time interval.
More l}recisel y:
(2b
1
) If !1 < 0, that is, c
2
< 4k, the sin-
gular point proves to he a focal point
(Figure 65) and, hence, the dynamical
system performs damped oscillations about
the state of equilibrium with a deeaying
amplitude.
(2b
2
) If !1 = 0, that is, c
2
= 4k, the
singular point is a nodal point with a single
negative exceptional direction (Figure 66).
The motion in this case is aperiodic and cor-
responds to the so-called critical damping.
( 2 b ~ ) If !1 > 0, that is. c ~ > 4k, the
singular point is a nodal point with two
negative exceptional direetions (Figure 67).
Qualitatively the motion of the dynamical
system is the same as in the previous case
and corresponds to damped oscillations.
From the above results it follows that
when c >0 and k >0, that is, drag is pos-
itive and the restoring force is attractive,
the dynamical system tends to a state of
equilibrium and its motion is stable.
(2c) If c < 0, that is, damping is nega-
tive, the qualitative pattern of the phase
trajectories is the same as in the case (2b),
the only difference being that here the 200 Differential Equations in Applications
y=.z
0 Z'
Fig. 66
_;/=.i~
.... \
` I _Z.
Fig. 67
200 Differential Eqnations in Applications
Fig. 66
Fig. 67 Ch. 2. Qualitative Methods 201
k .1*
.-0*
V %
Eg #*~
Fig. 63
y=.2
` i Lv
C > O
Fig. 69
dynamical system ceases to be stable.
Figure 68 contains all the above results
and the dependence of the type of singular
201
Fig. 68
Fig. 69
dynamical system ceases to be stable.
Figure 68 contains all the above results
and the dependence of the type of singular
c 202 Diflerential Equations in Applications
point on the values of parameters c and k.
Note that the diagrams can also be inter-
preted as a summary of the results of studies
of the types of singular points of system (122)
when J* #0 at c = D* and lc = J*.
However, the fact that c = O does not
generally mean that system (122) possesses
a vortex point, and the fact that k = 0 does
not mean that the system of a general type
has no singular point. These cases belong
to those of complex singular points, which
we consider below.
Returning to the dynamical system con-
sidered in this section, we note that if k = 0
(c q O), the autonomous system (145) cor-
responding to Eq. (144) assumes the form
dx dy
Wz ya `(T? = ""cy
This implies that the straight line y = 0 is
densely populated by singular points, with
the phase trajectory pattern shown in
Figure 69.
Finally, if k = c = 0, Eq. (144) assumes
the form
dzx
az? *-0-
The respective autonomous system is
da: dy
point on the values of parameters c and k.
Note that the diagrams can also be inter-
preted as a summary of the results of studies
of the types of singular points of system (122)
--ivnen r-:1 ~ '#u -lit. c = ~ t : r " aha' If, =r:J .
However, the fact that c = 0 does not
generally mean that system (122) possesses
a vortex point, and the fact that k = 0 does
not mean that the system of a general typo
has no singular point. These cases belong
to those of complex singular points, which
we consider below.
Returning to the dynamical system con-
sidered in this section, we note that if k = 0
(c =1= 0), the autonomous system (145) cor-
responding to Eq. (144) assumes the form
dx dy
dt=y, Cit= -cy.
This implies that the straight line y = 0 is
densely populated by singular points, with
the phase trajectory pattern shown in
Figure 69.
Finally, if k = c = 0, Eq. (144) assumes
the form
The respective autonomous system is
dx dy
Tt=y, Tt=
0
Ch. 2. Qualitative Methods 203
Here, as in the previous case, the .1: axis is
densely populated by singular points. The
respective phase trajectory pattern is
shown in Figure 70.
2.9 Adiabatic Flow
of a Perfect Gas Through a Noazle
of Varying Cross Section
A study of the flow of compressible viscous
media is highly important from the practi-
cal viewpoint. For one thing, such flow
emerges in the vicinity of a wing and fuse-
lage of an airplane; it also influences the
operation of steam and gas turbines, jet
engines, the nuclear reactors.
Below we discuss the flow of. a perfect
gas through a nozzle with a varying cross
section (Figure 7-1); the specific heat capaci-
ty of the gas is cp and the nozzlesvarying
cross-sectional area is denoted by A. The
flow is interpreted as one-dimensional, that
is, all its properties are assumed to be uni-
form in a single cross section of the nozzle.
Friction in the boundary layer is caused
by the tangential stress 1: given by the for-
mula
t = qpvz/2, (146)
where q is the friction coefficient depend_ing
basically on the Reynolds number but
assumed constant along the nozzle, p the
203
Here, as in the previous case, the .1: axis is
densely populated by singular points. The
respective phase trajectory pattern is
shown in Figure 70.
2.9 Adiabatic Flow
of a Perfect Gas Through a
of Varying Cross Section
A study of the flow of compressible viscous
media is highly important from the practi-
cal viewpoint. For one thing, such flow
emerges in the vicinity of a wing and fuse-
lage of an airplane; it also influences the
operation of steam and gas turbines, jet
engines, the reactors.
Below we discuss the flow of. a perfect
gas through a nozzle with a varying cross
section (Figure 71); the specific heat capaci-
ty of the gas is Cp and the nozzle's varying
cross-sectional area is denoted by A. Tho
flow is interpreted as one-dimensional, that
is, all its properties are assumed to be uni-
form in a single cross section of the nozzle.
Friction in the boundary layer is caused
by the tangential stress 't' given by the for-
mula
't' = qpv
2
12, (146)
where q is the friction coefficient depending
basically on the Reynolds number but
assumed constant along the noule, p the 204 Differential Equations in Applications
_;/-.5*
.z
Fig. 70
as
/
Fig. 71
flux density, and v the flow velocity. Final-
ly, we assume that adiabaticity conditions
are met, that is, drag, combustion, chemi-
cal transformations, evaporation, and con-
densation are excluded.
One of the basic equations describing
this type of flow is the well-known conti-
Fig. 70
..
Fig. 71
!J=i
flux density, and v the flow velocity. Final-
ly, we assume that adiabaticity conditions
are met, that is, drag, combustion, chemi-
cal transformations, evaporation, anrl con-
densation are excluded.
One of the basic equations describing
this type of Oow is the well-known conti-Ch. 2. Qualitative Methods 205
nuity equation, which in the given cse
is written in the form
w = pA v, (147)
where the flux variation rate w is assumed
constant. From this equation it follows
that
dp dA dv _
7+7-}--7-0. (148)
Let us now turn to the equation for the
energy of steadystate flow. We note that
generally such an equation links the exter-
nal work done on the system and the action
of external heat sources with the increase
in enthalpy (heat content) in the flow and
the kinetic and potential energies. In our
case the flow is adiabatic; hence, the energy-
balance equation can be written in the form
0=w(h-ldh)-wh
{ w [v2/2 + d (122/2)] - wvz/2,
or
dh i d (02/2) : O, (149)
where h is the enthalpy of the flow (the
thermodynamic potential) at absolute tem-
perature T. But in Eq. (149) dh = c,,dT
and, therefore, we can write the equation
for the flow energy as
c,,dT + d (v2/2) = O. (150)
205
nuity equation, which in the given case
is written in the form
w = pAv,
(147)
where the flux variation rate w is assumed
constant. From this equation it follows
that
dA
p ' A v
(148)
Let us now turn to the equation for the
energy of steady-state flow. We note that
generally such an equation links the exter-
nal work done on the system and the action
of external heat sources with the increase
in enthalpy (heat content) in the flow and
the kinetic and potential energies. In our
case the flow is adiabatic; hence, the energy-
balance equation can be written in the form
0 = w (h + dh)- wh
+ w [v
2
/2 + d (v
2
/2)1 - wv
2
12,
or
dh + d (v
2
/2) = 0,
(149)
where h is the enthalpy of the flow (the
thermodynamic potential) at absolute tem-
perature T. But in Eq. (149) dh = cpdT
and, therefore, we can write the equation
for the flow energy as
+ d (v
2
/2) = 0.
Now let us derive the momentum equa-
tion for the flow. Note that here the common
approach to problems involving a steady-
state flow is to use Newtons second law.
Assuming that the divergence angle of the
nozzle wall is small, we can write the
momentum equation in the form
pA+pdA(p+dp)(A-ldA)1:dA
= w dv,
or
A dp - dA dp 1 dA = w dv, (151)
where p is the static pressure.
The term dA dp in Eq. (151) is of a
higher order than the other terms and,
therefore, we can always assume that the
momentum equation for the flow has the
form
-A dp -- 1: dA = w dv. (152)
If we denote by D the hydraulic diameter,
we note that its variation along the nozzles
axis is determined by a function F such that
D = F (:1:), where x is the coordinate along
the nozzles axis. From the definition of the
hydraulic diameter it follows that
dA 4da:
-1-I-zi . (153)
Noting that pvz/2 = ypM2, where Y is the
specific heat ratio of the medium, and M
Now let us derive the momentum equa-
tion for the flow. Note that here the common
approach to problems involving a steady-
state flow is to use Newton's second law.
Assuming that the divergenr.e angle of the
nozzle wall is small, we can write the
momentum equation in the form
pA + p dA - (p + dp) (A + dA)- T dA
= w dv,
or
-A dp - dA dp - -r dA = w dv, (151)
where p is the static pressure.
The term dA dp in Eq. (151) is of a
higher order than the other terms and,
therefore, we can always assume that the
momentum equation for the flow has the
form
-A dp - -r dA = w dv. (152)
If we denote by D the hydraulic diameter,
we note that its variation along the nozzle's
axis is determined by a function F such that
D = F (x), where x is the coordinate along
the nozzle's axis. From the definition of the
hydraulic diameter it follows that
(153)
Noting that pv
2
/2 = ypM
2
, where y is the
specific heat ratio of the medium, and M Ch. 2. Qualitative Methods 207
is the Mach number, we can write formula
(146) thus:
1 := qypM2. (154)
Combining this with Eqs. (147) and (153),
we arrive at the following represen
tation for the momentum equation (152):
dp da: dv? __
-1;+ YM-'*(4qj-|;;)-O. (155)
Denoting the square of the Mach number by
y, employing Eqs. (148), (150), and (155),
and performing the necessary algebraic
transformations, we arrive at the following
..1 I
@(1+-Ly) (wry-F (=>>
-k-:---..-.-.- (156)
dx (i-y) F ()
where the prime stands for derivative of
the respective quantity.*
The denominator of the right-hand side of
Eq. (156) vanishes at y = 1, that is, when
the Mach number becomes equal to unity.
This means that the integral curves of the
last equation intersect what is known as the
sonic line and have vertical tangents at the
intersection points. Since the right-hand
* See J. Kestin and S.K. Zaremba, "One-dimen-
sional highspeed flows. Flow patterns derived
for the flow of gases through nozzles, including
compressibility and viscosity elfects, Aircraft
Engm. 25, No. 292: 172-175, 179 (1953).
207
is the Mach number, we can write formula
(146) thus:
"t = qypM
2
(154)
Combining this with Eqs. (147) and (153),
we arrive at the following represen-
tation for the momentum equation (152):
d: + yMz ( 4q ~ + d:-
2
) = 0. (155)
Denoting the square of the Mach number by
y, employing Eqs. (148), (150), and (155),
and performing the necessary algebraic
transformations, we arrive at the following
d 4y { 1+ ";
1
y) (yqy-F' (x}}
d ~ = (1-y)F(x) (
156
)
where the prime stands for derivative of
the respective quantity.*
The denominator of the right-hand side of
Eq. (156) vanishes at y = 1, that is, when
the Mach number becomes equal to unity.
This means that the integral curves of the
last equation intersect what is known as the
sonic line and have vertical tangents at the
intersection points. Since the right-hand
* See J. Kestin and S.K. Zaremba, "One-dimen-
sional high-speed flows. Flow patterns derived
for the flow of gases through nozzles, including
compressibility and viscosity effects," Aircraft
Engin. 25, No. 292: 172-175, 179 (1953). 208 Differential Equations in Applications
side of Eq. (156) reverses its sign in the
process of intersection, the integral curves
ilip over" and the possibility of inflection
points is excluded. The physical meaning
of this phenomenon implies that along
integral curves the value of a: must increase
continuously. Hence, the section on which
the integral curves intersect the sonic line
with vertical tangents must be the exit
section of the nozzle. Thus, the transition
from subsonic flow to supersonic (and back)
can occur inside the nozzle only through
a singular point with real exceptional di-
rections, that is, through a saddle point or
a nodal point.
The coordinates of the singular points of
Eq. (156) are specified by the equations
y* :1. F(*> = v<1
which imply that these points are situated
in the divergiug part of the nozzle. A saddle
point appears if J* is negative, that is,
F(:1:*) > O. Since q is a sufficiently small
constant, a saddle point appears nea1 the
throat of the nozzle. A nodal point, on the
other hand, appears only if F" (x*) is neg-
ative. Thus, a nodal point emerges in the
part of the nozzle that lies behind an in-
flection point of the nozzles profile or, in
practical terms, at a certain distance from
the throat of the nozzle, provided that the
proiile contains an inflection point.
side of Eq. (156) reverses its sign in the
process of intersection, the integral curves
"flip over" and the possibility of inflection
points is excluded. The physical meaning
of this phenomenon implies that along
integral curves the value of x must increase
continuously. Hence, the section on which
the integral curves intersect the sonic line
with vertical tangents must be the exit
section of the nozzle. Thus, the transition
from subsonic flow to supersonic (and back)
can occur inside the nozzle only through
a singular point with real exceptional di-
rections, that is, through a saddle point ot'
a nodal point.
The coordinates of the singular points of
Eq. (1!16) are specifted by the equations
y* = 1, F' (x*) = yq,
which imply that these points are situated
in the diverging part of the nozzle. A saddle
point appears if J* is negative, that is,
F" (x*) > 0. Since q is a sufficiently small
constant, a saddle point appears near the
throat of the nozzle. A nodal point, on the
other hand, appears only if F" (x*) is neg-
ative. Thus, a nodal point emerges in the
part of the nozzle that lies behind an in-
flection point of the nozzle's profile or, in
practical terms, at a certain distance from
the throat of the nozzle, provided that the
profile contains an inflection point. Ch. 2. Qualitative Methods 209
From the characteristic equation
F(*) K2 + 2qv (V +1)%
-2(Y +i)F(*) =0
we can see that the slopes of the two excep-
tional directions are opposite in sign in the
case of a saddle point and have the same
sign (are negative) in the case of a nodal
point. This means that only a saddle point
allows for a transition from supersonic to
subsonic velocities and from subsonic to
supersonic velocities (Figure 72). The case
of a nodal point (Figure 73) allows for a
continuous transition only from supersonic
to subsonic flow.
Since Eq. (156) cannot be integrated in
closed form, we must employ numerical
methods of integration in any further dis-
cussion. It is advisable in this connection
to begin the construction of the four sepa-
ratrices of a saddle point as integral curves
by allowing for the fact that the singular
point itself is a point, so to say, from which
these integral curves emerge. Such a con-
struction is indeed possible since the char-
acteristic equation provides us with the
direction of the two tangents at the singular
point S (:1:*, l). lf this is ignored and the
motion monitored as beginning at points
a and b in Figure 72, which lie on different
sides of a separatrix, the corresponding
points move along curves on and [5 that
strongly diverge and, hence, provide no in-
14-0770 Ch. 2. Qualitative Methods
From the characteristic equation
F(x*) A.
2
+ 2qy ('\' + 1) A.
- 2 (y + 1) F" (x*) = 0
209
we can see that the slopes of the two excep-
tional directions are opposite in sign in the
case of a saddle point and have the same
sign (are negative) in the case of a nodal
point. This means that only a saddle point
allows for a transition from supersonic to
subsonic velocities and from subsonic to
supersonic velocities (Figure 72). The case
of a nodal point (Figure 73) allows for a
continuous transition only from supersonic
to subsonic flow.
Since Eq. (156) cannot be integrated in
closed form, we must employ numerical
methods of integration in any further dis-
cussion. It is advisable in this connection
to begin the construction of the four sepa-
ratrices of a saddle point as integral curves
by allowing for the fact that the singular
point itself is a point, so to say, from which
these integral curves emerge. Such a con-
struction is indeed possible since the char-
acteristic equation provides us with the
direction of the two tangents at the singular
point S (x*, 1). If this is ignored and the
motion monitored as beginning at points
a and b in Figure 72, which lie on different
sides of a separatrix, the corresponding
points move along curves a and ~ that
strongly diverge and, hence, provide no in-
U.-0770 210 Differential Equations in Applications
:;~1Z F Z22)
ty Ly:.-F
/
cv
/
7 I S
(2I/ / I
b { I /3
/ I
{Ct .Z* ,9
Fig. 72
!=z
.S
7
I
I
I
I
0 1** .2*
Fig. 73
formation about the integral curve (the
separatrix) that "enters" point S. On the
!f=11
2
F?.x)
a ~
I
Fig. 72
Fig. 73
formation about the integral curve (the
separatrix) that "enters" point S. On the Ch. 2. Qualitative Methods 211
other hand, if we move along an integral
curve that "emerges" from point S and we
assume that the initial segment of the curve
coincides with a segment of an exceptional
straight line, the error may be minimized
if we allow for the convergence of integral
curves in the direction in which the values
of sc diminish.
Figure 72 illustrates the pattern of inte-
gral curves in the vicinity of a singular
point. The straight line passing through
point xt (the throat of the nozzle) corre-
sponds to values at which the numerator
in the right-hand side of Eq. (156) vanishes,
which points to the presence of extrema.
2.10 Higher-Order Points
of Equilibrium
In previous sections we studied the types
of singular points that emerge when the
Jacobian J * is nonzero. But suppose that
all the partial derivatives of the functions X
and Y in the righthand sides of system (122)
vanish up to the nth order inclusive. Then
in the vicinity of a singular point there
may be an iniinitude of phasetrajectory
patterns. However, if we exclude the pos-
sibility of equilibrium points of the vortex
and focal types emerging in this picture,
then it appears that the neighborhood of
a singular point can be broken down into
14*
211
other hand, if we move along an integral
curve that "emerges" from point S and we
assume that the initial segment of the curve
coincides with a segment of an exceptional
straight line, the error may be minimized
if we allow for the convergence of integral
curves in the direction in which the values
of x diminish.
Figure 72 illustrates the pattern of inte-
gral curves in the vicinity of a singular
point. The straight line passing through
point Xt (the throat of the nozzle) corre-
sponds to values at which the numerator
in the right-hand side of Eq. (156) vanishes,
which points to the presence of extrema.
2.10 Higher-Order Points
of Equilibrium
In previous sections we studied the types
of singular points that emerge when the
Jacobian J* is nonzero. But suppose that
all the partial derivatives of the functions X
and Y in the right-hand sides of system (122)
vanish up to the nth order inclusive. Then
in the vicinity of a singular point there
may be an infinitude of phase-trajectory
patterns. However, if we exclude the pos-
sibility of equilibrium points of the vortex
and focal types emerging in this picture,
then it appears that the neighborhood of
a singular point can be broken down into 212 Differential Equations in Applications
a finite number of sectors belonging to
three standard types. These are the hyper-
bolic, parabolic, and elliptic. Below we de-
scribe these sectors in detail, but first let us
make several assumptions to simplify mat-
ters.
We assume that the origin is shifted to
the singular point, that is, :1:* = y* = O;
the right-hand sides of system (122) can
X (v y) = X n (rv, y) + (D (re y) (157)
Y (rm y) = Ya (w, y) + I (rn y>
where X ,, and Y,, are polynomials of de-
gree n homogeneous in variables as and y
(one of these polynomials may be identically
zero), and the functions (D and Whave in
the neighborhood of the origin continuous
iirst partial derivatives. In addition, we
assume that the functions
<D(== y) (Dx ( y) (Dy (=, y)
(x2___y2)(7l+1)/2 (x2,_|_y2)7l/2 1
I<==, y) Ix(, y) Wy (rv, y)
(x2_)_y2)('Tl+1)/2 (x2__y2)71/P
are bounded in the neighborhood of the
origin. Under these assumptions the follow-
ing assertions hold true.
(1) Every trajectory of the system of
equations (122) with right-hand sides of the
(157) form that "enters the origin along
a certain tangent touches one of the ex-
a finite number of sectors belonging to
three standard types. These are the hyper-
bolic, parabolic, and elliptic. Below we de-
scribe these sectors in detail, but first let us
make several assumptions to simplify mat-
ters.
We assume that the origin is shifted to
the singular point, that is, x* = y* = 0;
the right-hand sides of system (122) can
X (x, y) = X n (x, y) + ci> (x, y),
Y (x, y) = Y n (x, y) + ' (x, y),
(157)
where X n and Y n are polynomials of de-
gree n homogeneous in variables x and y
(one of these polynomials may be identically
zero), and the functions ci> and 'I" have in
the neighborhood of the origin continuous
first partial derivatives. In addition, we
assume that the functions
II> (x, y) ll>x (x, y) ll>y (x, y)
(z2+ y2)(n+l)f2 (zll+ yB)n/2 ' (zZ+ yZ)ntz
'I' (x, y) 'I' x (x, y) 'I' y (x, y)
(z2+ yll)(n+I)/2 (x+ yZ)n/P
1
(x+ yZ)n/2
are bounded in the neighborhood of the
origin. Under these assumptions the follow-
ing assertions hold true.
(1) Every trajectory of the system of
equations (122) with right-hand sides of the
(157) form that "enters" the origin along
a certain tangent touches one of the ex-Ch. 2. Qualitative Methods 213
ceptional straight lines specified by the
equation
wYn (an y) !/Xu (fe y) = 0- (158)
Since the functions X ,, and Y,, are homo-
geneous, we can rewrite Eq. (158) as an
equation for the slope A = y/x. Then the
exceptional straight line is said to be sin-
gular if
Xa (vs y) = Yum y) =0
on this straight line. Some examples of
such straight lines are shown in Figure 62.
The straight lines defined by Eq. (158) but
not singular are said to be regular.
(2) The pattern of the phase trajectories
of (122) in the vicinity of one of the two
rays that "emerge" from the origin and
together form an exceptional straight line
can be studied by considering a small disk
(centered at the origin) from which we
select a sector limited by two radii lying
sufficiently close to the ray on both sides
of it. Such a sector is commonly known as
a standard domain.
More than that, in the case of a regular
exceptional straight line, which corre-
sponds to a linear factor in Eq. (158), the
standard domain considered in a disk of
a sufficiently small radius belongs t.o either
one of two types: attractive or repulsive.
(2a) The attractive standard domain is
characterized by the fact that each tra-
213
ceptional straight lines specified by the
equation
xY n (x, y) - yX n (x, y) = 0. (158)
Since the functions X n and Y n are homo-
geneous, we can rewrite Eq. (158) as an
equation for the slope A. = y/x. Then the
exceptional straight line is said to be sin-
gular if
X n (x, y) = Y n (x, y) = 0
on this straight line. Some examples of
such straight lines are shown in Figure 62.
The straight lines defined by Eq. (158) but
not singular are said to be regular.
(2) The pattern of the phase trajectories
of (122) in the vicinity of one of the two
rays that "emerge" from the origin and
together form an exceptional straight line
can be studied by considering a small disk
(centered at the origin} from which we
select a sector limited by two radii lying
sufficiently close to the ray on both sides
of it. Such a sector is commonly known as
a standard domain.
More than that, in the case of a regular
exceptional straight line, which corre-
sponds to a linear factor in Eq. (158), the
standard domain considered in a disk of
a sufficiently small radius belongs to either
one of two types: attractive or repulsive.
(2a} The attractive standard domain is
characterized by the fact that each tra-214 Diflerential Equations in Applications
Fig. 74
Fig. 75
jectory passing through it reaches the
origin along the tangent that coincides
with the exceptional straight line (Fig-
ure 74).
(2b) The repulsive standard domain is
characterized by the fact that only one
phase trajectory passing through it reaches
the singular point along the tangent that
coincides with the exceptional straight line.
All ot.her phase trajectories of (122) that
enter the standard domain through the
boundary of the disk leave the disk by
crossing one of the radii that limit the
domain (Figure 75).
Fig. 74
Fig. 75
jectory passing through it reaches the
origin along the tangent that coincides
with the exceptional straight line (Fig-
ure 74).
(2b) The repulsive standard domain is
characterized by the fact that only one
phase trajectory passing through it reaches
the singular point along the tangent that
coincides with the exceptional straight line.
All other phase trajectories of (1.22) that
enter the standard domain through the
boundary of the disk leave the disk by
crossing one of the radii that limit the
domain (Figure 75). Ch. 2. Qualitative Methods 215
Let us turn our attention to the following
fact. If the disk centered at the origin is
small enough, the two types of standard
domains can be classified according to the
behavior of the vector (X, Y) on the
boundary of the domain. [`he behavior of
vector (X, Y) can be identified here with
the behavior of vector (X ny Yn). More
than that, it can be demonstrated that if,
as assumed, a fixed exceptional direction
does not correspond to a multiple root of
the characteristic equation, the vector con-
sidered on one of the radii that limit the
domain is directed either inward or out-
ward. Then if in the first case the vector
considered on the part of the boundary
of the standard domain that is the arc of
the circle is also directed inward, and in
the second case outward, the standard
domain is attractive. But if the opposite
situation is true, the standard domain is
repulsive. It must be noted that in any case
the vector considered on the part of the
boundary that is the arc of the circle is
always directed either inward or outward
since it is almost parallel to the radius.
Standard domains corresponding to sin-
gular exceptional directions or multiple
roots of the characteristic equation have
a more complicated nature, but since to
some extent they constitute a highly rare
phenomenon, we will not describe them here.
lf we now turn, for example, to a saddle
215
Let us turn our attention to the following
fact. If the disk centered at the origin is
small enough, the two types of standard
domains can be classified according to the
behavior of the vector (X, Y) on the
boundary of the domain. The behavior of
vector (X, Y) can be identified here with
the behavior of vector (X n Y n) More
than that, it can be demonstrated that if,
as assumed, a fixed exceptional direction
does not correspond to a multiple root of
the characteristic equation, the vector con-
sidered on one of the radii that limit the
domain is directed either inward or out-
ward. Then if in the first case the vector
considered on the part of the boundary
of the standard domain that is the arc of
the circle is also directed inward, and in
the second case outward, the standard
domain is attractive. But if the opposite
situation is true, the standard domain is
repulsive. It must be noted that in any case
the vector considered on the part of the
boundary that is the arc of the circle is
always directed either inward or outward
since it is almost parallel to the radius.
Standard domains corresponding to sin-
gular exceptional directions or multiple
roots of the characteristic equation have
a more complicated nature, but since to
some extent they constitute a highly rare
phenomenon, we will not describe them here.
If we now turn, for example, to a saddle 216 Differential Equations in Applications
point, we note that it allows for four
repulsive standard domains. In the neigh-
borhood of a singular point of the nodal
type there are two attractive standard
domains and two repulsive.
(3) If there exist real-valued exceptional-
direction straight lines, the neighborhood
of a singular point can be divided into a
finite number of sectors each of which is
bounded by the two phase trajectories of
(122) that "enter" the origin along definite
tangents. Each of such sectors belongs to
one of the following three types.
(3a) The elliptic sector (Figure 76) con-
tains an infinitude of phase trajectories
in the form of loops passing through the
origin and touching on each side of the
boundary of the sector.
(3b) The parabolic sector (Figure 77)
is filled with phase trajectories that connect
the singular point with the boundary of the
neighborhood.
(3c) The hyperbolic sector (Figure 78)
is filled with phase trajectories that ap-
proach the boundary of the neighborhood
in both directions.
More precisely:
(Zia) Elliptic sectors are formed between
two phase trajectories belonging to two
successive standard domains, both of which
are attractive.
(4b) Parabolic sectors are formed between
two phase trajectories belonging to" two
point, we note that it allows for four
repulsive standard domains. In the neigh-
borhood of a singular point of the nodal
type there are two attractive standard
domains and two repulsive.
(3) If there exist real-valued exceptional-
direction straight lines, the neighborhood
of a singular point can be divided into a
finite number of sectors each of which is
bounded by the two phase trajectories of
(122) that "enter" the origin along definite
tangents. Each of such sectors belongs to
one of the following three types.
(3a) The elliptic sector (Figure 76) con-
tains an infinitude of phase trajectories
in the form of loops passing through the
origin and touching on each side of the
boundary of the sector.
(3b) The parabolic sector (Figure 77)
is filled with phase trajectories that connect
the singular point with the boundary of the
neighborhood.
(3c) The hyperbolic sector (Figure 78)
is filled with phase trajectories that ap-
proach the boundary of the neighborhood
in both directions.
More precisely:
(4a) Elliptic sectors are formed between
two phase trajectories belonging to two
successive standard domains, both of which
are attractive.
(4b) Parabolic sectors are formed between
two phase trajectories belonging to.,. two Ch. 2. Qualitative Methods gn
Fig. 76 Fig. 77
Fig. 78
successive standard domains one of which
is attractive and the other repulsive. All
phase trajectories that pass through the
latter domain touch at the singular point
of the exceptional straight line that defines
the attractive domain.
217
s
Fig. 76 Fig. 77
$
Fig. 78
successive standard domains one of which
is attractive and the other repulsive. All
phase trajectories that pass through the
latter domain touch at the singular point
of the exceptional straight line that defines
the attractive domain. 218 Differential Equations in Applications
(4c) Hyperbolic sectors are formed be-
tween two phase trajectories belonging to
two successive repulsive standard domains.
For example-, it is easy to distinguish the
four hyperbolic sectors at a saddle point
and the four parabolic sectors at a nodal
point. Elliptic sectors do not appear in
the case of simple singular points, where
the Jacobian J * is nonzero.
lf a singular point does not allow for the
existence of real exceptional directions,
the phase trajectories in its neighborhood
always possess the vortex or focal struc-
ture.*
2.11 Inversion with Respect
to a Circle and Homogeneous
Coordinates
Above we described methods for establish-
ing the local behavior of phase trajectories
of differential systems of the (122) type in
the neighborhood of singular points. And
although in many cases all required infor-
mation can be extracted by following these
methods, there may be a need to study the
" Methods that make it possible to distinguish
between a vortex {Joint and a focal point are dis-
cussed, for examp e, in the book by V.V Ame1-
kin, N.A. Lukashevich, and A.P. Sadovskii,
Nonlinear Vibrations in Second0rder Systems
(Minsk: Belorussian Univ. Press, 1982) (in Rus-
sian .
(4c) Hyperbolic sectors are formed be-
tween two phase trajectories belonging to
two successive repulsive standard domains.
For example, it is easy to distinguish the
four hyperbolic sectors at a saddle point
and the four parabolic sectors at a nodal
point. Elliptic sectors do not appear in
the case of simple singular points, where
the Jacobian J* is nonzero.
If a singular point does not allow for the
existence of real exceptional directions,
the phase trajectories in its neighborhood
always possess the vortex or focal struc-
ture.*
2.11 Inversion with Respect
to a Circle and Homogeneous
Coordinates
Above we described methods for establish-
ing the local behavior of phase trajectories
of differential systems of the (122) type in
the neighborhood of singular points. And
although in many cases all required infor-
mation can be extracted by following these
methods, there may be a need to study the
* Methods that make it possible to distinguish
between a vortex point and a focal point are dis-
cussed, for example, in the book by V.V Amel'-
kin, N.A. Lukashevieh, and A.P. Sadovskii,
Nonlinear Vibrations in Second-Order Systems
(Minsk: Belorussian Univ. Press, 1982) (in Rus-
sian). Ch. 2. Qualitative Methods 219
trajectory behavior in infinitely distant
parts of the phase plane, as x2 1~ y2 > oo.
A simple way of studying the asymptotic
behavior of the phase trajectories of the
differential system (122) is to introduce
the point at infinity by transforming the
initial differential system in an appropriate
manner, say by inversion, which is defined
by the following formulas:
.__ E. __ n
{L- {32+*12 y- +n
$ y
(g2`_;_Fi,
Geometrically this transformation consti-
tutes what has become known as inversion
with respect to ci circle and maps the origin
into the point at infinity and vice versa.
Transformation (159) maps every finite
point M (az, y) of the phase plane into point
M (, 1]) of the same plane, with points M
and M' lying on a single ray that emerges
from the origin and obeying the condition
OM >< OM' =- rz (Figure 79). It is well
known that such a transformation maps
circles into circles (straight lines are con-
sidered circles passing through the point
at infinity). For one thing, straight lines
passing through the origin are invariant
under transformation (159). Hence, the
slopes of asymptotic directions are the
slopes of tangents at the new origin =
1] = O. Note that in the majority of
219
trajectory behavior in infinitely distant
parts of the phase plane, as x
2
+ y
2
-+ oo.
A simple way of studying the asymptotic
behavior of the phase trajectories of the
differential system (1.22) is to introduce
the point at infinity by transforming the
initial differential system in an appropriate
manner, say by inversion, which is defmed
by the following formulas:
TJ
(1.59)
Geometrically this transformation consti-
tutes what has become known as inversion
with respect to a circle and maps the origin
into the point at infinity and vice versa.
Transformation (159) maps every finite
point M (x, y) of the phase plane into point
M' ( ~ , rt) of the same plane, with points M
and M' lying on a single ray that emerges
from the origin and obeying the condition
OM X OM' = r
2
(Figure 79). It is well
known that such a transformation maps
circles into circles (straight lines are con-
sidered circles passing through the point
at infinity). For one thing, straight lines
passing through the origin are invariant
under transformation (1.59). Hence, the
slopes of asymptotic directions are the
slopes of tangents at the new origin ~ =
rt = 0. Note that in the majority of 220 Differential Equations in Applications
ml
wv, f
Fig. 79
N
Fig. 80
cases the new origin serves as a singular
point. The reasons for this are discussed
below.
Fig. 79
Fig. 80
cases the new origin serves as a singular
point. The reasons for this are discussed
below. Ch. 2. Qualitative Methods 221
As for the question of how to construct
in the old (sc, y)-plane a curve that has a
definite asymptotic direction, that is, has
a definite tangent at the origin of the new
(Q, 1])-plane, this can be started by con-
sidering the (, ij)-plane, more precisely,
by considering this curve in, say, a unit
circle in the (Q, 1])-plane. The fact is that
since a unit circle is mapped, via transfor-
mation (159), into itself, we can always estab-
lish the point where the curve intersects the
respective unit circle in the (1:, y)-plane;
any further investigations can be carried
out in the usual manner.
We also note that completion of the
(:1:, y)-plane with the point at infinity is
topologically equivalent to inversion of
the stereographic projection (Figure 80),
in which the points on a sphere are mapped
onto a plane that is tangent to the sphere
at point S. The projection center N is the
antipodal point of S. It is clear that the
projection center N corresponds to the
point at infinity in the (41:, y)-plane. Con-
versely, if we map the plane onto the
sphere, a vector field on the plane trans-
forms into a vector field on the sphere and
the point at infinity may prove to be a sin-
gular point on the sphere.
Although inversion with respect to a
circle is useful, it proves cumbersome and
inconvenient when the point at infinity has
a complicated structure. In such cases
221
As for the question of how to construct
in the old (x, y)-plane a curve that has a
definite asymptotic direction, that is, has
a definite tangent at the origin of the new
( ~ . TJ)-plane, this can he started by con-
sidering the ( ~ , TJ)-plane, more precisely,
by considering this curve in, say, a unit
circle in the ( ~ , TJ)-plane. The fact is that
since a unit circle is mapped, via transfor-
mation (159), into itself, we can always estab-
lish the point where the curve intersects the
respective unit circle in the (x, y)-plane;
any further investigations can he carried
out in the usual manner.
We also note that completion of the
(x, y)-plane with the point at infinity is
topologically equivalent to inversion of
the stereographic projection (Figure 80),
in which the points on a sphere are mapped
onto a plane that is tangent to the sphere
at point S. The projection center N is the
antipodal point of S. It is clear that the
projection center N corresponds to the
point at infinity in the (x, y)-plane. Con-
versely, if we map the plane onto the
sphere, a vector field on the plane trans-
forms into a vector field on the sphere and
the point at infinity may prove to be a sin-
gular point on the sphere.
Although inversion with respect to a
circle is useful, it proves cumbersome and
inconvenient when the point at infinity has
a complicated structure. In such cases 222 Differential Equations in Applications
another, more convenient, transformation of
the (:1:, y)plane is used by introducing
homogeneous coordinates:
as = /z, y = 1]/z.
Under this transformation, each point of
the (x, y)-plane is associated with a triple
of real numbers (, 1], z) that are not
simultaneously zero and no difference is
made between the triples (, eq, z) and
(k, lm, kz) for every real k q'-= O. If a
point (x, y) is not at infinity, z =;= 0. But
if z = 0, we have a straight line at infinity.
The (x, y)-plane completed with the straight
line at infinity is called the projective plane.
Such a straight line may carry several sin-
gular points, and the nature of these points
is usually simpler than that of a singular
point introduced by inversion with respect
to a circle.
If we now consider a pencil of lines and
describe its center with a sphere of, say,
unit radius, then each line of the pencil
intersects the sphere at antipodal points.
This implies that every point of the pro-
jective plane is mapped continuously and
in a one-to-one manner onto a pair of anti-
podal points on the unit sphere. Thus, the
projective plane may be interpreted as the
set of all pairs of antipodal points on a unit
sphere. To visualize the projective plane,
we need only consider one-half of the sphere
and assume its points to be the points of the
another, more convenient, transformation of
the (x, y)-plane is used by introducing
homogeneous coordinates:
x = ~ / z , y = r]/z.
Under this transformation, each point of
the (x, y)-plane is associated with a triple
of real numbers ( ~ , f], z) that are not
simultaneously zero and no difference is
made between the triples (s, fJ, z) and
( k ~ , k1), kz) for every real k =I= 0. If a
point (x, y) is not at infinity, z =I= 0. But
if z = 0, we have a straight line at infinity.
The (x, y)-plane completed with the straight
line at infinity is called the projective plane.
Such a straight line may carry several sin-
gular points, and the nature of these points
is usually simpler than that of a singular
point introduced by inversion with respect
to a circle.
If we now consider a pencil of lines and
describe its center with a sphere of, say,
unit radius, then each line of the pencil
intersects the sphere at antipodal points.
This implies that every point of the pro-
jective plane is mapped continuously and
in a one-to-one manner onto a pair of anti-
podal points on the unit sphere. Thus, the
projective plane may be interpreted as the
set of all pairs of antipodal points on a unit
sphere. To visualize the projective plane,
we need only consider one-half of the sphere
and assume its points to be the points of the Ch. 2. Qualitative Methods 233
=z=0
**0 AM;
we =z=0
lL i iA'
E w
.9
Fig. 81
projective plane. If we orthcgonally project
this hemisphere (say, the lower one) onto
the ot-plane, which touches it at pole S
(Figure 81), the projective plane is mapped
onto a unit disk whose antipodal points
on the boundary are assumed identical.
Each pair of the antipodal points of the
boundary corresponds to a line at infinity,
and the completion of the Euclidean plane
with this line transforms the plane into
a closed surface, the projective plane.
2.12 Flow of ra Perfect Gas
Through a Rotating Tube
of Uniform Cross Section
In some types of turboprop helicopters and
airplanes and in jet turbines, the gaseous
fuelair mixture is forced through rotating
223
Fig. 81
projective plane. If we orthogonally project
this hemisphere (say, the lower one) onto
the a.-plane, which touches it at pole S
(Figure 81), the projective plane is mapped
onto a unit disk whose antipodal points
on the boundary are assumed identical.
Each pair of the antipodal points of the
boundary corresponds to a line at infinity,
and the completion of the Euclidean plane
with this line transforms the plane into
a closed surface, the projective plane.
2.12 Flow of a Perfect Gas
Through a Rotating Tube
of Uniform Cross Section
In some types of turboprop helicopters and
airplanes and in jet turbines, the gaseous
fuel-air mixture is forced through rotating 224 Differential Equations in Applications
d
CK 5 C V
QH `
e
Fig. 82
tubes of uniform cross section installed
in the compressor blades and linked by a
hollow vertical axle. To establish the opti-
mal conditions for rotation we must analyze
the flow of the mixture through a rotating
tube and link the solution with boundary
conditions determined by the tube design.
In a blade the gaseous mixture participates
in a rotational motion with respect to the
axis with constant angular velocity co
and moves relative to the tube with an
acceleration v (dv/dr), where v is the speed
of a gas particle with respect to the tube,
and r is the coordinate measured along the
rotating compressor blades.
Figure 82 depicts schematically a single
rotating tube of a compressor blade.* It is
* See I . Kestin and S.K. Zaremba, "Adiabatic
one-dimensional flow of a perfect gas through a
rotating tube of uniform cross section," Aeronaut.
Quart. 4: 373-399 (1954).
r
e
Fig. 82
tubes of uniform cross section installed
in the compressor blades and linked by a
hollow vertical axle. To establish the opti-
mal conditions for rotation we must analyze
the flow of the mixture through a rotating
tube and link the solution with boundary
conditions determined by the tube design.
In a blade the gaseous mixture participates
in a rotational motion with respect to the
axis with constant angular velocity (J)
and moves relative to the tube with an
acceleration v (dv!dr), where v is the speed
of a gas particle with respect to the tube,
and r is the coordinate measured along the
rotating compressor blades.
Figure 82 depicts schematically a single
rotating tube of a compressor blade.* It is
See 1. Kestin and S.K. Zaremba, "Adiabatic
one-dimensional How of a perfect gas through a
rotating tube of uniform cross section," Aeronaut.
Quart. 4: 373-399 (1954). Ch. 2. Qualitative Methods 225
assumed that the fuel-air mixture, whose
initial state is known, is supplied along
the hollow axle to a cavity on the axle in
which the flow velocity may be assumed
insignificant. The boundary of the cavity
occupied by the gas is denoted by a, the
gas is assumed perfect with a speciiic heat
ratio y, and all the processes that the gas
mixture undergoes are assumed reversibly
adiabatic (exceptions are noted below).
It is assumed that the gas expands through
a nozzle with an outlet cross section b that
at the same time is the inlet of a tube of
uniform cross section. The expansion of
the gas mixture from state a. to state b
is assumed to proceed isentropically; we
denote the velocity of the gas after expan-
sion by vi and the distance from the rotation
axis O, to the cross section b by rl.
When passing through the tube, whose
uniform crosssectional area will be denoted
by A and whose hydraulic diameter by D,
the gas is accelerated thanks to the com-
bined action of the pressure drop and dy-
namical acceleration in the rotation compres-
sor blade. We ignore here the effect produced
by pressure variations in the tube (if they
exist at all) and by variations in pressure
drop acting on the cross section plane, both
of which are the result of the Coriolis force.
The last assumption, generally speaking,
requires experimental verification, since
the existence of a lateral pressure drop may
is-0770
225
assumed that the fuel-air mixture, whose
initial state is known, is supplied along
the hollow axle to a cavity on the axle in
which the flow velocity may be assumed
insignificant. The boundary of the cavity
occupied by the gas is denoted by a, the
gas is assumed perfect with a specific heat
ratio y, and all the processes that the gas
mixture undergoes are assumed reversibly
adiabatic (exceptions are noted below).
It is assumed that the gas expands through
a nozzle with an outlet cross section b that
at the same time is the inlet of a tube of
uniform cross section. The expansion of
the gas mixture from state a to state b
is assumed to proceed isentropically; we
denote the velocity of the gas after expan-
sion by v
1
and the distance from the rotation
axis 0
1
to the cross section b by r
1

When passing through the tube, whose
uniform cross-sectional area will be denoted
by A and whose hydraulic diameter by D,
the gas is accelerated thanks to the com-
bined action of the pressure drop and dy-
namical acceleration in the rotation compres-
sor blade. We ignore here the effect produced
by pressure variations in the tube (if they
exist at all) and by variations in pressure
drop acting on the cross section plane, both
of which are the result of the Coriolis force.
The last assumption, generally speaking,
requires experimental verification, since
the existence of a lateral pressure drop may
Ui-0770 226 Differential Equations in Applications
serve as a cause of secondary flows. But if
the diameter of the tube is small compared
to the tubes length, such an assumption
is justified.
It is now clear that the equations of mo-
mentum and energy balance for a compres-
sible mixture traveling along a tube of
uniform cross section must be modified so
as to allow for forces of inertia that appear
in a rotating reference frame. As for the
continuity equation, it remains the same.
Now let us suppose that starting from
cross section c at the right end of the tube
the gas is compressed isentropically and
passes through a diverging nozzle. In the
process it transfers into a state of rest with
respect to the compressor blade in the second
cavity at a distance r3 from the rotation
axis and reaches a state with pressure Pd
and temperature Td.
From the second cavity the gaseous mix-
ture expands isentropically into a converg-
ing or convergingdiverging nozzle in such
a manner that it leaves the cavity at right
angles to the tubes axis. This produces
a thrust force caused by the presence of
a torque.
Below for the sake of simplicity we as-
sume that the exit nozzle is a converging
one and has an exit (throat) of area A*.
Denoting the external (atmospheric) pres-
sure by Pa, we consider two modes of pas-
sage of the mixture through the nozzle.
serve as a cause of secondary flows. But if
the diameter of the tube is small compared
to the tube's length, such an assumption
is justified.
It is now clear that the equations of mo-
mentum and energy balance for a compres-
sible mixture traveling along a tube of
uniform cross section must be modified so
as to allow for forces of inertia that appear
in a rotating reference frame. As for the
continuity equation, it remains the same.
Now let us suppose that starting from
cross section c at the right end of the tube
the gas is compressed isentropically and
passes through a diverging nozzle. In the
process it transfers into a state of rest with
respect to the compressor blade in the second
cavity at a distance r
3
from the rotation
axis and reaches a state with pressure P d
and temperature T d
From the second cavity the gaseous mix-
ture expands isentropically into a converg-
ing or converging-diverging nozzle in such
a manner that it leaves the cavity at right
angles to the tube's axis. This produces
a thrust force caused by the presence of
a torque.
Below for the sake of simplicity we as-
sume that the exit nozzle is a converging
one and has an exit (throat) of area A*.
Denoting the external (atmospheric) pres-
sure by P
4
, we consider two modes of pas-
sage of the mixture through the nozzle. Ch. 2. Qualitative Methods 227
The first applies. to a situation in which
the P3-to-Pd ratio exceeds the critical
value, or
P.,/Pd > (2/(v + i))""""
In this case the flow at the nozzl0s throat
is subsonic and, hence, pressure P3 at the
throat is equal to the atmospheric pressure,
that is, P3 -= P3. The second case applies
to a situation in which the Pa-to-Pd ratio
is below the critical value, or
P.,/Pa < (2/(v + i)""*-
The pressure P3 at the nozzles throat has
a fixed value that depends on Pd but not
on P3. Hence,
P3 -= (2/(V + i))"""" Pa-
In the latter case the flow at the nozzles
throat has the speed of sound
Us = (2/(v + i))"2 aa
where ad depends only on temperature Td.
In further analysis the flow is assumed
adiabatic everywhere and isentropic every-
where except in the tube between cross sec-
tions b and c.
As in the case where we derive the differ-
ential equation that describes the adia-
batic flow of a perfect gas through a nozzle
of varying cross section, let us now consider
the continuity equation, the equation of
15*
227
The first applies. to a situation in which
the P a-to-P d ratio exceeds the critical
value, or
Pa!Pd > (2/(y + 1))'1'/<v-t>.
In this case the flow at the nozzle's throat
is subsonic and, hence, pressure P
3
at the
throat is equal to the atmospheric pressure,
that is, P
3
= P a The second case applies
to a situation in which the P a-to-P d ratio
is below the critical value, or
Pa!Pd < (2/(y + 1)'1'/<v-t>.
The pressure P
3
at the nozzle's throat has
a fixed value that depends on P d but not
on P a Hence,
P
3
= (2/(y + 1))'1'1<v-t> Pd.
In the latter case the flow at the nozzle's
throat has the speed of sound
v
3
= (2/(y + 1))11
2
ad,
where ad depends only on temperature T d
In further analysis the flow is assumed
adiabatic everywhere and isentropic every-
where except in the tube between cross sec-
tions b and c.
As in the -case where we derive the differ-
ential equation that describes the adia-
batic flow of a perfect gas through a nozzle
of varying cross section, let us now consider
the continuity equation, the equation of
x
U 7 U+d0
,0 P+6Z'/D
v /
[.,4
r' + dr
Fig. 83
momentum, and the equation of energy
balance. For instance, the equation of con-
tinuity in this case has the form
- .2. ..1; .4; _
1p V .. V3 A ..const, (160)
with V the speciiic volume and xp the flux
density, or
\|> = mr/A,
where m' is the flux mass.
To derive the equation of motion, or the
momentum equation, we turn to Figure 83.
We note that the dynamical effect of the
rotational motion of the compressor blade
can be used to describe the flow with re-
spect to the moving tube, where in accord-
ance with the DA1embe1t principle the
force of inertia
dI = {3- wzf dr
Fig. 83
momentum, and the equation of energy
balance. For instance, the equation of con-
tinuity in this case has the form
(160)
with V the specific volume and 'lj> the flux
density, or
"'> = m'IA,
where m' is the flux inass.
To derive the equation of motion, or the
momentum equation, we turn to Figure 83.
We note that the dynamical effect of the
rotational motion of the compressor blade
can be used to describe the flow with re-
spect to the moving tube, where in accord-
ance with the D'Alembert principle the
force of inertia
A
dl =y-w2rdr Ch. 2. Qualitative Methods 229
is assumed to act in the positive direction
of r. Hence, the element of mass dm 2
(A/V) dr moves with an acceleration
v(dv/dr) caused by the combined action
of the force of inertia dI, the force of pres-
sure A dP, and the friction force dF =
(4A/D) dr. Here t : Z. (02/2V), where 7.
depends on the Reynolds number R. As
a first approximation we can assume that
2. remains constant along the entire tube.
With this in mind, we can write the mo-
mentum equation in the form
A dv 2}Av2 A
I7dTUH;-- dTlI7ZU2Tdr,
or after certain manipulations,
V dP |v dv-}- ED?-L v2drw2r dr= 0. (161)
As for the energy-balance equation, it is
simple to derive if we use the first law of
thermodynamics for open systems and bear
in mind that the amount of work performed
by the system is wzr dr. Thus
dh|vdv-w2rdr=O,
where h is the enthalpy. If we define the
speed of sound a via the equation
h:`QlT>
we find that
ada-lY%vdv -3%-}w2rdr=O.
229
is assumed to act in the positive direction
of r. Hence, the element of mass dm =
(AIV) dr moves with an acceleration
v (dv/dr) caused by the combined action
of the force of inertia d/, the force of pres-
sure A dP, and the friction force dF =
(4A/D) dr. Here -r = i. (v
2
/2V), where A.
depends on the Reynolds number R. As
a first approximation we can assume that
A. remains constant along the entire tube.
With this in mind, we can write the mo-
mentum equation in the form
=-A dP-
2
f..Avz
V dr VD V
or after certain manipulations,
2?..
V dP +v dv+ D vzdr- wzr dr = 0. (161)
As for the energy-balance equation, it is
simple to derive if we use the first law of
thermodynamics for open systems and bear
in mind that the amount of work performed
by the system is w
2
r dr. Thus
dh + v dv- w
2
r dr = 0,
where h is the enthalpy. If we define the
speed of sound a via the equation

v-1 '
we find that
ada+ "';
1
vdv- "';
1
w2rdr=0. 230 Differential Equations in Applications
This leads us to the following energy-balance
equation:
ag Z az -1- -Y-gi U2- 3% ww. (162)
If we now introduce the dimensionless
quantities
M0 = v/ao, as = r/D, G2 = w2D/a, (163)
we arrive at the equation
2 -1 -1
%:1-JT-M-l- %G2x2. (164)
By excluding the pressure and specific
volume from the basic equations (160),
(161), and (164) we can derive an equation
that links the dimensionless quantity M0
with the dimensionless distance. This equa
tion serves as the basic equation for solving
our problem. The continuity equation (160)
implies that
V = v/wp. (165)
Then, combining
__ ___ 'yPv
G2 -- T
with Eq. (162), we arrive at the following
relationships:
_ a2 _ y1 _ y1 wz
pr (Ri W*wT"2)f
dP _ Y1 wgr __ y1 a2
YT"` y v ( 2v Jr qe;2
v1 w2r2 dv ..
+ ig" *,7) 7;; (166)
This leads us to the following energy-balance
equation:
a ~ = a 2 - l - '\';
1
v
2
- '\'
2
1
w2r2, (162)
If we now introduce the dimensionless
quantities
M
0
= vla
0
, x = riD, G
2
= w
2
D
2
1a:, (163)
we arrive at the equation
al. -1- Y-
1
M
2
+ y-t G2x2 (164)
a2 - 2 2
By excluding the pressure and specific
volume from the basic equations (160),
(161), and (164) we can derive an equation
that links the dimensionless quantity M
0
with the dimensionless distance. This equa-
tion serves as the basic equation for solving
our problem. The continuity equation (160)
implies that
V = vi'!'. (165)
Then, combining
yPv
a
2
=yPV =-
'1'
with Eq. (162), we arrive at the following
relationships:
(
a ~ y-1 y-1 w2 )
P = ----v +-- -- r2 h
yv 2y 21' I' 't
dP = y-1 w
2
r --{ y-1 + a8
d r y v , 2y yv2
+ y-1 w2r2 ) dv
2 Y ~ dr"
Substituting the value of V from Eq. (165)
and the value of dP/dr from Eq. (166) into
the momentum equation (161) and allowing
for the relationship (163), we arrive at the
djtlg Z 2M1(22,yM-Gaz;) o (167)
at 4-% Mg .!. lg.; Gzxz
Assuming that
1
M=y, m=2?v, p=y2(v+1>.
1
q : if (Y '- Us
we can rewrite Eq. (167) as
dy __ 2y (myG2x)
"ds?" 1py|qG2x2 (*68)
The differential equation (168) is the
object of our further investigation. In-
troducing the variables (159), we reduce it to
Q Z 2E*l(EP*lE+1GE,2)+2*1(*12~E2)Q
di A (EPT\E | <1G) + @.*129
(169)
where E : Q2-}- q2, Q = mq- G2, A : ,2T]2.
Clearly, the origin constitutes a singular
point for Eq. (169). The lowestorder terms
in both numerator and denominator are
quadratic. If we discard higher-order terms,
231
Substituting the value of V from Eq. (165)
and the value of dP!dr from Eq. (166) into
the momentum equation (161) and allowing
for the relationship (163), we arrive at the
dM
2
__ o=
dx
Assuming that
y, m= 2A.y,
1
q= 2(y-1),
we can rewrite Eq. (167) as
dy 2y (my-G2x)
dx 1-py+qG2x2
(168)
The differential equation (168) is the
object of our further investigation. In-
troducing the variables (159), we reduce it to
d1J Q
df A
(169)
where B=6
2
+YJ
2
, Q=mYJ-G
2
6, A=s2-YJ2.
Clearly, the origin constitutes a singular
point for Eq. (169). The lowest-order terms
in both numerator and denominator are
quadratic. If we discard higher-order terms, 232 Differential Equations in Applications
as we did on p. 211 when discussing higher-
order equilibrium points, we iind that
Y4 (gv Tl)
= 2qG2E211 + 211 (112 E2) (mn G2E),
X4 (gv
= qG2E2 (E2 112) + 4E112 (mn G2E),
and, hence, the origin is a higher-order
singular point.
The characteristic equation of the differ-
ential equation (169) assumes the following
form after certain manipulations:
En (E2 + 112) [(q + 2) G2E 2*1111] = 0-
(170)
This leads us to three realvalued excep-
tional straight lines:
() E=0 (b)11 =0
<> (q + 21 Gag 2m. = 0. (27*)
Each is a regular straight line and corre-
sponds to one of the factors in Eq. (170).
It is easy to see that for and 1] positive
the value of X4 (, iq) is positive in the
neighborhood of all three exceptional
straight lines, with the result that the
expression
Yi (E, 11) _;
X4 (gs g
Z gl} TIS) l(q"l`2) G2E"2"ll
E lqG2E2 (E2 112)+ 4En* (mn - G2E)l
as we did on p. 211 when discussing higher-
order equilibrium points, we find that
Y4 <s. '11>
= 2qG2ss'I'J + 2'1'] ('112 - s2) (m'I'J - G2s),
x4 (s. '11)
= qG2s2 (62 - '112) + 4s'l12 (m'll - G2s),
and, hence, the origin is a higher-order
singular point.
The characteristic equation of the differ-
ential equation (169) assumes the following
form after certain manipulations:
s'I'J (s
2
+ '11
2
) [(q + 2) G
2
s - 2m'1']1 = o.
(170)
This leads us to three real-valued excep-
tional straight lines:
(a) s = 0, (b) '11 = 0,
(c) (q + 2) G
2
s - 2m'I'J = o.
(171)
Each is a regular straight line and corre-
sponds to one of the factors in Eq. (170).
It is easy to see that for s and '11 positive
the value of x4 (s, 'I']) is positive in the
neighborhood of all three exceptional
straight lines, with the result that the
expression
y4 TJ) _ _!L
x, TJ)
[(q+2)
= [qGS62

has the same sign in both the first quadrant
and the neighborhood of the straight lines
as the left-hand side of (171c) and, hence,
is negative below the straight line (171c)
and positive above. The explanation lies
in the fact that (172) may change sign only
when an exceptional straight line is crossed.
Geometrically this means that the right-
hand side of the differential equation
El; Y4 (gv ll)
dg X4 (gs
determining the behavior of integral curves
in a disk of sufficiently small radius and
centered at the origin fixes an angle greater
than the angle of inclination of the radii
that lie in the first quadrant between the
straight lines (171b) and (171c). As for
the radii lying between the straight lines
(171c) and (171a), on them the righthand
side of this differential equation fixes an
angle that is smaller than the angle of
inclination of these radii (Figure 84). Thus,
the standard domain containing the straight
line (171c) must be repulsive, while the
domains containing the straight lines (17ia)
and (171b) must be attractive. In view of
the symmetry of the field specified by vector
(X4, Y4), the above facts remain valid for
standard domains obtained from those men-
tioned earlier by a rotation about the
singular point through an angle of 180.
233
has the same sign in both the first quadrant
and the neighborhood of the straight lines
as the left-hand side of (171c) and, hence,
is negative below the straight line (171c)
and positive above. The explanation lies
in the fact that (172) may change sign only
when an exceptional straight line is crossed.
Geometrically this means that the right-
hand side of the differential equation
determining the behavior of integral curves
in a disk of sufficiently small radius and
centered at the origin fixes an angle greater
than the angle of inclination of the radii
that lie in the first quadrant between the
straight lines (171b) and (171c). As for
the radii lying between the straight lines
(171c) and (171a), on them the right-hand
side of this differential equation fixes an
angle that is smaller than the angle of
inclination of these radii (Figure 84). Thus,
the standard domain containing the straight
line (171c) must be repulsive, while the
domains containing the straight lines (171a)
and (171b) must be attractive. In view of
the symmetry of the field specified by vector
(X
4
, Y
4
), the above facts remain valid for
standard domains obtained from those men-
tioned earlier by a rotation about the
singular point through an angle of 180. 234 Differential Equations in Applications
/
\
em 7 X
~ f
`\ TY //4/
/ K
// x \
jr +\
Ai `\
/7 Ev
Fig. 84
Thus, there are exactly two integral
curves that "enter" the singular point along
the tangent (171c) and an infinitude of
integral curves that touch the coordinate
axes (171a) and (171b) at the point of rest.
We see then that the second and fourth
quadrants contain elliptic sectors since
they lie between two successive attractive
standard domains (Figure 85). Each of the
first and third quadrants is divided into
two sectors by the integral curves that
touch the exceptional straight line (171c)
at the origin. These sectors are parabolic
because they lie between two successive
standard domains, one of which is attractive
and the other repulsive. More than that,
all the integral curves except those that
touch the straight line (171c) touch the
Fig. 84
Thus, there are exactly two integral
curves that "enter" the singular point along
the tangent (171c) and an infinitude of
integral curves that touch the coordinate
axes (171a) and (171b) at the point of rest.
We see then that the second and fourth
quadrants contain elliptic sectors since
they lie between two successive attractive
standard domains (Figure 85). Each of the
first and third quadrants is divided into
two sectors by the integral curves that
touch the exceptional straight line (171c)
at the origin. These sectors are parabolic
because they lie between two successive
standard domains, one of which is attractive
and the other repulsive. More than that,
all the integral curves except those that
touch the straight line (171c) touch the Ch. 2. Qualitative Methods 235
7
is
Fig. 85
coordinate axes (171a) and (17lb) at the
singular point.
Basing our reasoning on physical con-
siderations, we can analyze the differential
equation (168) solely in the first quadrant
of the (sc, y)plane. Turning to this plane,
we see that there exists exactly one integral
curve having an asymptotic direction with
a slope (q {- 2) G2/2m. All other integral
curves allow for the asymptotic direction
of one of the coordinate axes. Indeed, it is
easy to prove that all these integral curves
asymptotically approach one of the coor-
dinate axes, that is, along each of them in
the movement toward infinity not only does
y/x >0 or :1:/y ->0 but so does y -+0 or
cc ->0 (with .2: ->oo or y ->cx>, respec-
235
Fig. 85
coordinate axes (171a) and (171b) at the
singular point.
Basing our reasoning on physical con-
siderations, we can analyze the differential
equation (168) solely in the first quadrant
of the (x, y)-plane. Turning to this plane,
we see that there exists exactly one integral
curve having an asymptotic direction with
a slope (q + 2) G
2
/2m. All other integral
curves allow for the asymptotic direction
of one of the coordinate axes. Indeed, it is
easy to prove that all these integral curves
asymptotically approach one of the coor-
dinate axes, that is, along each of them in
the movement toward infmity not only does
y/x -+ 0 or x/y -+ 0 but so does y -+ 0 or
x -+ 0 (with x -+ oo or y -+ oo, res pee-236 Diflerential Equations in Applications
0
.2:
Fig. 86
tively). This is illustrated in Figure 86,
where some integral curves have been plot-
ted at great values of :1: and y. Note that the
pattern of integral curves at a finite dis-
tance from the origin depends primarily
on the position of the singular points, and
a special investigation is required to clarify
it. It can also be proved that the integral
curves that approach the coordinate axes
asymptotically (Figure 86) leave the first
quadrant at a finite distance from the
origin.
2.13 Isolated Closed Trajectories
We already know that in the case of a
singular point of the vortex type a certain
region of the phase plane is completely
filled with closed trajectories. However,
!J
Fig. 86
tively). This is illustrated in Figure 86,
where some integral curves have been plot-
ted at great values of x andy. Note that the
pattern of integral curves at a finite dis-
tance from the origin depends primarily
on the position of the singular points, and
a special investigation is required to clarify
it. It can also be proved that the integral
curves that approach the coordinate axes
asymptotically (Figure 86) leave the first
quadrant at a finite distance from the
origin.
2.13 Isolated Closed Trajectories
We already know that in the case of a
singular point of the vortex type a certain
region of the phase plane is completely
filled with closed trajectories. However, Ch. 2. Qualitative Methods 237
a more complicated situation may occur
when there is an isolated closed trajectory,
that is, a trajectory in a certain neigh-
borhood of which there are no other closed
trajectories. This case is directly linked
with the existence of isolated periodic
solutions. Interestingly, only nonlinear dif-
ferential equations and systems can have
isolated closed trajectories.
Isolated periodic solutions correspond
to a broad spectrum of phenomena and pro-
cesses occurring in biology, radiophysics,
oscillation theory, astronomy, medicine,
and the theory of device design. Such solu-
tions emerge in differential models in eco-
nomics, in various aspects of automatic
control, in airplane design, and in other
fields. Below we study the possibility of
isolated periodic solutions emerging in pro-
cesses that occur in electric circuits; we also
consider as a model the nonlinear differ-
ential system
{j=y+=v(i wy2).
dy (173)
Ei-:-:z:|y(l -0:2-y2).
To solve this system, we introduce polar
coordinates r and 9, where x = rcos 6
and y = rsin 9. Then, differentiating the
relationships x2 |y2 == rz and 9 =
237
a more complicated situation may occur
when there is an isolated closed trajectory,
that is, a trajectory in a certain neigh-
borhood of which there are no other closed
trajectories. This case is directly linked
with the existence of isolated periodic
solutions. Interestingly, only nonlinear dif-
ferential equations and systems can have
isolated closed trajectories.
Isolated periodic solutions correspond
to a broad spectrum of phenomena and pro-
cesses occurring in biology, radiophysics,
oscillation theory, astronomy, medicine,
and the theory of device design. Such solu-
tions emerge in differential models in eco-
nomics, in various aspects of automatic
control, in airplane design, and in other
fields. Below we study the possibility of
isolated periodic solutions emerging in pro-
cesses that occur in electric circuits; we also
consider as a model the nonlinear differ-
ential system
:: = -y + x (1 - x
2
- y
2
),
: ~ = x + y (1 - x
2
- y
2
).
(173)
To solve this system, we introduce polar
COOrdinates r and 6, where X = r COS 6
and y = r sin e. Then, differentiating the
relationships x
2
+ y
2
= r
2
and e = 238 Differential Equations in Applications
tan1(y/x) with respect to t, we arrive at
dx dy __ dr dy _ dx _ _ dB`
"w+ya7rw xr; ff aT2a2
(174)
Multiplying the first equation into sc and
the second into y, adding the products,
and allowing for the first relationship in
(174), we find that
dr
r -8-2-: rz (1 -1*2). (175)
Multiplying the second equation in (173)
into sc and the first into y, subtracting one
product from another, and allowing for the
second relationship in (174), we find that
rz gil; rz. (176)
System (173) has only one singular point,
O (0, 0). Since at the moment we are only
interested in constructing trajectories, we
can assume that r is positive. Then Eqs.
(175) and (176) imply that system (173)
can be reduced to the form
dr df) _
717--T(1T2), aT1.
Each of these equations can easily be inte-
grated and the entire family of solutions,
tan -
1
(y/x) with respect to t, we arrive at
dx _J__ dy dr dy dx _ .
2
de
xdt,ydt=rdt, xfft-Yfft-
1
dt'
(174)
Multiplying the first equation into x and
the second into y, adding the products,
and allowing for the first relationship in
(174), we find that
(175)
Multiplying the second equation in (173)
into x and the fust into y, subtracting one
product from another, and allowing for the
second relationship in (174), we find that
dO
r
z __ r2
dt- .
(176)
System (173) has only one singular point,
0 (0, 0). Since at the moment we are only
interested in constructing trajectories, we
can assume that r is positive. Then Eqs.
(175) and (176) imply that system (173)
can be reduced to the form
dr _ (
1
2
) de _
1
dt-r -r , (ft- .
(177)
Each of these equations can easily be inte-
grated and the entire family of solutions, Ch. 2. Qualitative Methods 239
as can easily be seen, is given by the for-
mulas
1
9:-1t+t (178)
or, in terms of the old variables x and y,
COS + tg) sin to)
x = -T I! = +*_-
V 1+C I/1+ 06**
If now in the first equation in (178) we put
C-:0, we get r=1 and 6:-t-{t0.
These two relationships define a closed
trajectory, a circle x2 l- y2 L-= 1. If C is
negative, it is clear that r is greater than
unity and tends to unity as t > +oo. But
if C is positive, it is clear that r is less than
unity and tends to unity as t - +oo. This
means that there exists only one closed
trajectory r == _1 which all other trajectories
approach along spirals with the passage of
time (Figure 87).
Closed phase trajectories possessing such
properties are known as limit cycles or,
more precisely, (orbitally) stable limit cycles.
In fact, there can be two additional types
of limit cycles. A limit cycle is said to be
(orbitally) unstable if all neighboring tra-
jectories spiral away from it as t|o<>.
And a limit cycle is said to be (orbitally)
half-stable if all neighboring trajectories
on one side (say, from the inner side) spiral
on to it and all neighboring trajectories
239
alii can easily be seen, is given by the for-
mulas
(178)
or, in terms of the old variables x and y,
cos (t+ t
0
) sin (t+ t
0
)
x- y - ~ ~ ~ ~
- lli+Ce 2t ' - vt+Ce-2t
If now in the first equation in (178) we put
c = 0, we get r = 1 and e = t + to.
These two relationships define a closed
trajectory, a circle x
2
+ y
2
1. If C is
negative, it is clear that r is greater than
unity and tends to unity as t -+ +oo. But
if C is positive, it is clear that r is less than
unity and tends to unity as t -+ +oo. This
means that there exists only one closed
trajectory r = .1 which all other trajectories
approach along spirals with the passage of
time (Figure 87).
Closed phase trajectories possessing such
properties are known as limit cycles or,
more precisely, (orbitally) stable limit cycles.
In fact, there can be two additional types
of limit cycles. A limit cycle is said to be
(orbitally) unstable if all neighboring tra-
jectories spiral away from it as t -+ +oo.
And a limit cycle is said to be (orbitally)
half-stable if all neighboring trajectories
on one side (say, from the inner side) spiral
on to it and all neighboring trajectories 240 Differential Equations in Applications
.7
7
.z*
Fig. 87
on the other side (say, on the outer side)
spiral away from it as t->-|.
In the above example we were able to
find in explicit form the equation of a closed
phase trajectory, but generally, of course,
this cannot be done. Hence the importance
in the theory of ordinary differential equa-
tions of criteria that enable at least spe-
cifying the regions where a limit cycle
may occur. Note that a closed trajectory of
(122), if such a trajectory exists, contains
within its interior at least one singular
point of the system. This, for one thing,
implies that if there are no singular points
of a differential system within a region of
the phase plane, there are no closed tra-
jectories in the region either.
Fig. 87
on the other side (say, on the outer side)
spiral away from it as t -+ +oo.
In the above example we were able to
find in explicit form the equation of a closed
phase trajectory, but generally, of course,
this cannot be done. Hence the importance
in the theory of ordinary differential equa-
tions of criteria that enable at least spe-
cifying the regions where a limit cycle
may occur. Note that a closed trajectory of
(122), if such a trajectory exists, contains
within its interior at least one singular
point of the system. This, for one thing,
implies that if there are no singular points
of a differential system within a region of
the phase plane, there are no closed tra-
jectories in the region either. Ch. 2. Qualitative Methods 241
T1
f/."""`_"-`7<\\
X 0 ` F
x
( i
I 1
l 1
\ /
\\ //
Fig. 88
Let D be a bounded domain that lies
together with its boundary in the phase
plane and does not contain any singular
points of system (122). Then the Poincar-
Bendixson criterion holds true, namely, ij
I` is a trajectory of (122) that at the initial
moment t:- to emerges from a point that
lies in D and remains in D for all t) to,
then F is either closed or approaches a closed
trajectory along a spiral with the passage of
time.
We illustrate this in Figure 88. Here D
consists of two closed curves I`, and I`2
and the circular domain between them.
With each boundary point of D we associate
a vector
V( y) = X (rv, y)i+Y(. y)i-
Then, if a trajectory T` that emerges at the
initial moment t= to from a boundary
point, enters D, and remains there at all
16-0770
Ch. 2. Qualitative Methods 24t
Fig. 88
Let D be a bounded domain that lies
together with its boundary in the phase
plane and does nut contain any singular
points of system (122). Then the Poincare-
Bendixson criterion holds true, namely, if
r is a trajectory of (122) that at the initial
moment t = t
0
emerges from a point that
lies in D and remains in D for all t ;;;.:. t
0
,
then r is either closed or approaches a closed
trajectory along a spiral with the passage of
time.
We illustrate this in Figure 88. Here D
consists of two closed curves r 1 and r 2
and the circular domain between them.
With each boundary point of D we associate
a vector
V (x, y) = X (x, y) i + Y (x, y) j.
Then, if a trajectory r that emerges at the
initial moment t = t
0
from a boundary
point, enters D, and remains there at all
t 6-0770 242 Differential Equations in Applications
moments tgto, then, according to the
above criterion, it will along a spiral
approach a closed trajectory F0 that lies
entirely i11 D. The curve I`0 must surround
a singular point of the differential system,
point P, not lying in D.
The differential system (173) provides
a simple example illustrating the appli-
cation of the above criterion in finding limit
cycles. Indeed, system (173) has only one
singular point, O (O, O), and, therefore,
the domain D lying between the circles
with radii r = 1/2 and r = 2 contains no
singular points. The first equation in (177)
implies that dr/dt is positive on the inner
circle and negative on the oute1. Vector V,
which is associated with the points on the
boundary of D, is always directed into D.
This means that the circular domain lying
between circles with 1adii r = 1/2 and
r : 2 must contain a closed trajectory of the
differential system (173). Such a closed
trajectory does indeed exist, it is a circle of
radius r = 1.
Note, however, that great difficulties are
generally encountered in a system of the
(122) type when we wish to realize prac-
tically the Poincar-Bendixson criterion,
since no general methods exist for building
the appropriate domains and, therefore,
success depends both on the type of system
and on the experience of the researcher. At
the same time we must bear in mi11d that
moments t ;;;:: t
0
, then, according to the
above criterion, it will along_ a spiral
approach a closed trajectory r 0 that lies
entirely in D. The curve r
0
must surround
a singular point of the differential system,
point P, not lying in D.
The differential system (173) provides
a simple example illustrating the appli-
cation of the above criterion in iinding limit
cycles. Indeed, system (173) has only one
singular point, 0 (0, 0), and, therefore,
the domain D lying between the circles
with radii r = 1/2 and r = 2 contains no
singular points. The first equation in (177)
implies that dr/dt is positive on the inner
circle and negative on the outer. Vector V,
which is associated with the points on the
boundary of D, is always directed into D.
This means that the circular domain lying
between circles with radii r = 1/2 and
r = 2 must contain a closed trajectory of the
differential system (173). Such a closed
trajectory docs indeed exist, it is a circle of
radius r = 1.
Note, however, that great difficulties are
generally encountered in a system of the
(122) type when we wish to realize prac-
tically the Poincare-Bendixson criterion,
since no general methods exist for building
the appropriate domains and, therefore,
success depends both on the type of system
and on the experience of the researcher. At
the same time we must bear in mind that Ch. 2. Qualitative Methods 243
finding the conditions in which no limit
cycles exist is no less important than estab-
lishing criteria for their existence. ln this
respect the most widespread condition is the
Dulac criterion: if there exists a function
B (x, y) that is continuous together with its
first partial derivatives and is such that in
a simply connected domain D of the phase
plane the sum
0 (BX) 0 (BY)
0.2: + Oy
is a function of fixed sign,* then no limit
cycles of the dijjerential system (122) can
exist in D. At B (x, y) Ei 1 the criterion
transforms into the Bendixson criterion.
lf we turn to the differential equation
(156), which constitutes a differential model
for describing an adiabatic OIl9"(llI1"l|3l1Sl(l113l
flow of a perfect gas of constant specific
heat ratio througli a nozzle with drag, then
for this equation we have
X(v. y)=(iy)F(),
...1 I
Y(v, y)=4y (1l+ y) <vqyF ())
If we put
v1 *
Btw, y)= {4yF(v) [1+ q, 0]}
* That is, positive definite or negative definite.
16*
finding the conditions in which no limit
cycles exist is no less important than estab-
lishing criteria for their existence. In this
respect the most widespread condilioit is the
Dulac criterion: if there exists a function
B (x, y) that is continuous together with its
first partial deri"atiues and is such that in
a simply connected domain D of the phase
plane the sum
a (BX) + cJ (BY)
ax Dy
is a function of fixed sign,* then no limit
cycles of the differential system (122) can
exist in D. At B (x, y) = 1 the criterion
transforms into the Bendixson criterion.
If we turn to the differential equation
(156), which constitutes a differential model
for describing an adiabatic one-dimensional
flow of a perfect gas of constant specilic
heat ratio through a nozzle with drag, then
for this equation we have
X(x, y)=(1-y)F(x),
Y(x, y)=4y(11+ ~ ' ;
1
y)(yqy-F'(x)).
If we put
B (x, y) = { 4yF (x) [ 1 + I'
2
1
y ]} -
1
* That is, positive definite or negative deliuile.
16* 244 Difierential Equations in Applications
we find that
6(BX) 6(BY) _ Yq
dx + dy = F(x) >O
and, hence, Eq. (156) has no closed integral
curves.
Let us discuss one more concept that can
be employed to establish the existence of
limit cycles. The concept is that of the
index of a singular point.
Let I` be a simple closed curve (i.e. a
curve without self~intersections) that is
not necessarily a phase trajectory of system
(122), lies in the phase plane, and does not
pass through the singular points of this
system. Then, if P (x, y) is a point of I`,
the vecto1
V(v,y)=X(v 1/)i+Y(>,y)i.
with i and j the unit vectors directed along
the Cartesian axes, is a nonzero vector and,
hence, is characterized by a certain direction
speciiied by an angle 9 (Figure 89). If
point P (x, y) moves along I`, say, counter-
clockwise a11d completes a full cycle, vector
V performs an integral number of cycles
in the process, that is, angle 9 acquires an
increment A9 = 2nn, with n a positive
integer, a negative integer, or zero. This
number n is said to be the index of the closed
curve I` (or the index of cycle I`).
lf we begin to contract I` in such a manner
that under this deformation I` does not
244 Differential Equations in Applicatiol'.;
we find that
a (BX) + a (BY)
ax By
and, hence, Eq. (156) has no closed integral
curves.
Let us discuss one more concept that can
be employed to establish the existence of
limit cycles. The concept is that of the
index of a singular point.
Let r be a simple closed curve (i.e. a
curve without self-intersections) that is
not necessarily a phase trajectory of system
(122), lies in the phase plane, and does not
pass through the singular points of this
systPm. Then, if p (x, y) is a point of r'
the vector
V (x, y) = X (x, y) i + Y (x, y) j,
with i and j the unit vectors directed along
the Cartesian axes, is a nonzero vector and,
hence, is characterized by a certain
specified by an angle e (Figure 89). If
point p (x, y) moves along r, say, counter-
clockwise and completes a full cycle, vector
V performs an integral number of cycles
in the process, that is, angle e acquires an
increment L\8 = 2nn, with n a positive
integer, a negative integer, or zero. This
number n is said to be the index of the closed
cun.:e r (or the index of cycle f).
If we begin to contract r in such a manner
that under this deformation r does not Ch. 2. Qualitative Methods 245
V
y 0
F
0 eZ`
Fig. 89
pass through singular points of the given
vector Held, the index of the cycle must,
on the one hand, vary continuously and,
on the other, remain an integer. This means
that under continuous deformation of the
curve the index of the cycle does not change.
This property leads to the notion of the
index of a singular point as the index of a
simple closed curve surrounding the sin-
gular point.
The index has the following properties:
(1) the index of a closed trajectory of the
diyjerential system (122) is equal to +1,
(2) the index of a closed curve surrounding
several singular points is equal to the sum
of the indices of these points, and
(3) the index of a closed curve encompassing
only ordinary points is zero.
245
y
r
o ~ - - - - - - - - - - - - - - ~
Fig. 89
pass through singular points of the given
vector field, the index of the cycle must,
on the one hand, vary continuously and,
on the other, remain an integer. This means
that under continuous deformation of the
curve the index of the cycle does not change.
This property leads to the notion of the
index of a singular point as the index of a
simple closed curve surrounding the sin-
gular point.
The index has the following properties:
(1) the index of a closed trajectory of the
differential system (122) is equal to +1,
(2) the index of a closed curve surrounding
several .r;ingular points is equal to the sum
of the indices of these points, and
(3) the index of a closed curve encompassing
only ordinary points is zero. 246 Differential Equations in Applications
This implies, for one thing, that since
the index of a closed trajectory of system
(122) is always +1, a closed trajectory must
enclose either one singular point with an
index -}-1 or several singular points with
a net index equal to -{-1. This fact is often
used to prove the absence of limit cycles.
The index of a singular point is calculated
by the formula
n:1+#, uw)
where e is the number of elliptic sectors,
and h is the number of hyperbolic sectors.
For practical purposes the following simple
method may be suggested. Suppose that L
is a cycle that does not pass through sin-
gular points of (122) and is such that any
trajectory of (122) has no more than a finite
number of points common to L. The tra-
jectories may intersect L or touch it. In
the latter case only exterior points of con-
tact (type A) or interior points of contact
(type B) are taken into account, while the
C-type points, points of inflection, are not
(see Figure 90). We can still use formula
(179) to calculate the index of a singular
point, but now e is the number of interior
points of contact and h the number of
exterior points of contact of trajectories
of (122) with cycle L.
Figure 91 depicts singular points with
This implies, for one thing, that since
the index of a closed trajectory of system
(122) is always +1, a closed trajectory must
enclose either one singular point with an
index +1 or several singular points with
a net index equal to +1. This fact is often
used to prove the absence of limit cycles.
The index of a singular point is calculated
by the formula
e--h
n = 1 + -
2
-, (179)
where e is the number of elliptic sectors,
and h is the number of hyperbolie sectors.
For practieal purposes the following simple
method may be suggested. Suppose that L
is a cycle that does not pass through sin-
gular points of (122) and is such that any
trajectory of (122) has no more than a fmite
number of points common to L. The tra-
jectories may intersect L or touch it. In
the latter case only exterior points of con-
tact (type A) or interior points of contact
(type B) are taken into account, while the
C-type points, points of inflection, are not
(see Figure 90). We can still use formula
(179) to calculate the index of a singular
point, but now e is the number of interior
points of contact and h the number of
exterior points of contact of trajectories
of (122) with cycle L.
Figure 91 depicts singular points with Ch. 2. Qualitative Methods 247
O
am
Q M
B'<\>L
Fig. 90
indices 0, +2, +3, +1, and -2, re-
spectively.
Earlier it was noted that constructing
a complete picture of the behavior of the
phase trajectories of the differential system
(122) is facilitated by introducing the point
at infinity via transformations (159). To-
pological considerations provide a very
general theorem which states that when
a continuous vector field with a finite num-
ber of singular points is specified on a
sphere, the not index of the points is +2.
Thus, if the net index of all the singular
points of a differential system (possessing
a finite number of such po.ints) that lie in
a finite phase-plane domain is distinct
_from +2, the point at infinity must be a
singular point with a nonzero index.
But if instead of inversion we employ
homogeneous coordinates, the net index of
all the singular points is already +1. That
247
/..
Fig. 90
indices 0, +2, +3, +1, and -2, re-
spectively.
Earlier it was noted that constructing
a complete picture of the behavior of t-he
phase trajectories of thfil differential system
(122) is facilitated by introducing the point
at infinity via transformations (159). To-
pological considerations provide a very
general theorem which states that when
a continuous vector fteld with a finite num-
ber of singular points is specified on a
sphere, the net index of the points is +2.
Thus, if the net index of all the singular
points of a differential system (possessing
a fmite number of such points) that lie i ~
a finite phase-plane domain is distinct
.from +2, the point at infmity must be a
singular point with a nonzero index.
But if instead of inversion we employ
homogeneous coordinates, the net index of
all the singular points is already +1. That Ch. 2. Qualitative Methods 249
this is so can be seen from the fact that if
the plane is projected onto a sphere with
the center of projection placed at the center
of the sphere, two points on the sphere cor-
respond to a single point on the projective
plane, and the circumference of the great
circle parallel to the plane corresponds to
a straight line at infinity.
If we turn to Eq. (168), which describes
adiabatic one-dimensional flow of a perfect
gas through a rotating tube of uniform cross
section, then, as shown in Figure 85, for the
singular point at infinity we have e = 2
and h == O. It follows from this that the
index of this point is +2 and does not
depend on the values of the constants in
Eq. (168). This implies, for one thing, that
the net index of finite singular points is
zero. It can also be shown that, depending
on whether the straight line specified by
the equation my - Ganz == 0 has two points,
one double point, or not a single point of
intersection with the parabola fixed by the
equation 1 -py + qG2x2 = O (which, in turn,
is equivalent to G being greater than G0,
equal to G0, or less than G0, with G0 =
2mq*/2/p), the following combinations of
finite singular points occur.
(a) G > G0. A saddle point and a nodal
point.
(b) G = G0. A higher-order singular point
with two hyperbolic (h .= 2) sectors and
two parabolic (e = O).
this is so can be seen from the fact that if
the plane is projected onto a sphere with
the center of projection placed at the center
of the sphere, two points on the sphere cor-
respond to a single point on the projective
plane, and the circumference of the great
circle parallel to the plane corresponds to
a straight line at infinity.
If we turn to Eq. (168), which describes
adiabatic one-dimensional flow of a perfect
gas through a rotating tube of uniform cross
section, then, as shown in Figure 85, for the
singular point at infinity we have e = 2
and h = 0. It follows from this that the
index of this point is +2 and does not
depend on the values of the constants in
Eq. (168). This implies, for one thing, that
the net index of finite singular points is
zero. It can also be shown that, depending
on whether the straight line specified by
the equation my - Gx
2
= 0 has two points,
one double point, or not a single point of
intersection with the parabola fixed by the
equation 1 - py + qG
2
x
2
= 0 (which, in turn,
is equivalent to G being greater than G
0
,
equal to G
0
, or Jess than G
0
, with G
0
=
2mq
1
1
2
/p), the following combinations of
finite singular points occur.
(a) G > G
0
A saddle point and a nodal
point.
(b) G = G
0
A higher-order singular point
with two hyperbolic (h = 2) sectors and
two parabolic (e = 0}. 250 Differential Equations in Applications
(c) G <G0. Singular points are absent.
We see that in three cases the net index
is zero, as it should be.
2.14 Periodic Modes in Electric
Circuits
We will show how limit cycles emerge in
a dynatron oscillator (Figure 92). An anal-
ysis of the operation of such an oscillator
leads to what is known as the Van der Pol
equation. Although phenomena linked with
generation of limit cycles can be illustrated
by examples from mechanics, biology, and
economics} we will show how such phe-
nomena emerge in the study of electric
circuits.
Figure 92 depicts schematically a dyna-
tron oscillator, with the ia-va characteristic
shown by a solid curve in Figure 93. Here
ia is the current and va the voltage in the
t
' r C
e '|' +
r
...L__
Fig. 92
(c) G < G
0
Singular points are absent.
We see that in three cases the net index
is zero, as it should be.
2.14 Periodic Modes in Electric
Circuits
We will show how limit cycles emerge in
a dynatron oscillator (Figure 92). An anal-
ysis of the operation of such an oscillator
leads to what is known as the Van der Pol
equation. Although phenomena linked with
generation of limit cycles can be illustrated
by examples from mechanics, biology, and
we will show how such phe-
nomena emerge in the study of electric
circuits.
Figure 92 depicts schematically a dyna-
tron oscillator, with the ia-Va characteristic
shown by a solid curve in Figure 93. Here
ia is the current and Va the voltage in the
c
Fig. 92 Ch. 2. Qualitative Methods 251
screen-grid tube, the dynatron. The circuit
consists of resistance r, inductance L,
and capacitance C connected in parallel
and known as the tank circuit, which is
connected in series with the screengrid tube.
The real circuit can be replaced in this case
with an equivalent circuit shown schematic-
ally in Figure 94. The characteristic of the
tube may be approximated with a third-
degree polynomial i = ow + YU3, which is
shown by a dashed curve in Figure 93.
Here i and v stand for the coordinates in
a system whose origin is shifted to the
point of inflection O. As follows from
Figure 93,
on > O, Y > O.
In accordance with one of Kirchhoffs laws,
ilir+iLlio=O,
with ir = v/r, L (<liL/dt) == v, and ic =
Cv. As a result of simple manipulations we
" < 1 3v 2 1-;...
" +lzr+;z+@"" lvi LC -*0-
If we now put
cz,. 1 _ 3y __ _1___ 2
E+?F" 2`b rc "<
we can write the p1evious equation in the
form
all- _(;z l- bvz) + o>v = O.
251
screen-grid tube, the dynatron. The circuit
consists of resistance r, inductance L,
and capacitance C connected in parallel
and known as the tank circuit, which is
connected in series with the screen-grid tube.
The real circuit can be replaced in this case
with an equivalent circuit shown schematic-
ally in Figure 94. The characteristic of the
tube may be approximated with a third-
degree polynomial i = av + yv
3
, which is
by a dashed curve in Figure 93.
Here i and v stand for the coordinates in
a system whose origin is shifted to the
point of inflection 0. As follows from
Figure 93,
a>O, y>O.
In accordance with one of Kitchhoff's laws,
i + i
7
+ iL + ic = 0,
with i r = vir, L (eli ddt) = v, and ic =
Cv. As a result of simple manipulations we
( a. 1 3y ) . 1
v + c+rc+c v2 v-t- LC =0.
If we now put
a. 1 - 3y- 1 - 2
LC -Wo,
we can write the previous equation in the
form
+ + bv
2
) + = 0. 252 Differential Equations in Applications
ia li /
* /
{ /
/I-\\\ 1 /
- # .... -:'--- -7/ -,.
/ 0| / U
/ l \\ 1/
n
1
I
Od va
Fig. 93
i
-(---
l #1 it rl
if R f" A C
Fig. 94
This differential equation is known as the
Van der Pol equation. If we introduce the
transformations

"=y ""yW=
we can associate with the Van der Pol
Fig. 93
i
J:--- L
-l ____ .L....--..l---____.a.____..j
c
Fig. 94
This differential equation is known as the
Van der Pol equation. If we introduce the
transformations
v=y,
dy
v=ydv,
we can as5ociate with the Van der Po] Ch. 2. Qualitative Methods 253
equation the following first-order differen-
tial equation:
dy ___ __ (+bv2)y+<3v __ Y(v. y)
TY" y "` V (v. y) (180)
The only finite singular point of this
equation is the origin, and
J* (0, 0) = m>0,
D (v, y) == -(a -4- bvz).
Since b>0, we assume that a>0 and
conclude that divergence D does not re-
verse its sign and, hence, Eq. (180) can
have no closed integral curves. We, there-
fore, will consider only the case where a
is negative, that is, ot< -1/r. This im-
plies that D (0, 0) == a > 0 and, hence,
the singular point is either a nodal point
or a focal point. lf we now consider the
differential system corresponding to the
differential equation (180), that is,
dy __ dv __
ÈZ""Y(Uv y)v `(È"I/(vv y)v
we see that as t grows, the representative
singular point. Thus, a trajectory that
emerges from the point at infinity cannot
reach the singular point at the origin no
matter what the value of t including the
case where t-= l-oo. This implies that if
we can prove that a trajectory originating
at the point at infinity resembles a spiral
253
equation the following first-order differen-
tial equation:
dy (a+bv2)y+w8v Y(v, y) (
1
SO)
dv = - y V (v, y)
The only finite singular point of this
equation is the origin, and
J* (0, 0) = ( I } ~ > 0,
D (v, y) = -(a + bv
2
).
Since b > 0, we assume that a > 0 and
conclude that divergence D does not re-
verse its sign and, hence, Eq. (180) can
have no closed integral curves. We, there-
fore, will consider only the case where a
is negative, that is, a< -1/r. This im-
plies that D (0, 0) = -a > 0 and, hence,
the singular point is either a nodal point
or a focal point. If we now consider the
differential system corresponding to the
differential equation (180), that is,
dy dv
dt= Y (v, y), dt =. V (v, y),
we s e ~ that as t grows, the representative
singular point. Thus, a trajectory that
emerges from the point at infinity cannot
reach the singular point at the origin no
matter what the value of t including the
case where t = +oo. This implies that if
we can prove that a trajectory originating
at the point at infmity resembles a spiral 254 Differential Equations in Applications
winding onto the origin as t-><x>, this
will guarantee that at least one limit cycle
exists.
The existence proof for such a cycle can
be obtained either numerically or analyti-
cally. A numerical method suggested by the
Dutch physicist and mathematician Van
der Pol (1889-1959) consists in building
a trajectory that originates at a point
positioned at a great distance from the
origin and in checking whether this tra-
jectory possesses the above-mentioned prop-
erty. Such a procedure yields an app1oxi-
mation to the limit cycle, but it can be
employed only in the case where concrete
numerical values are known.
Below we give a proof procedure* based
on analytical considerations and the in-
vestigation of the properties of singular
points at infinity. Here, in contrast to the
method by which Eq. (168) was studied,
we employ the more convenient (in the
present problem) homogeneous-coordinate
transformations
v = E,/z, y = 1]/z. (181)
The straight line at infinity is fixed by the
equation z = O. To reduce the number of
* See the paper by J. Kestin and S.K. Zaremba,
"Geometrical methods in the analysis of ordinary
differential equations," Appl. Sei. Res. B3: 144
189 (1953).
winding onto the origin as t -+ oo, this
will guarantee that at least one limit cycle
exists.
The existence proof for such a cycle can
be obtained either numerically or analyti-
cally. A numerical method suggested by the
Dutch physicist and mathematician Van
der Pol (1889-1959) consists in building
a trajectory that originates at a point
positioned at a great distance from the
origin and in checking whether this tra-
jectory possesses the above-mentioned prop-
erty. Such a procedure yields an approxi-
mation to the limit cycle, but it can be
employed only in the case where concrete
numerical values are known.
Below we give a proof procedure* based
on analytical considerations and the in-
vestiga lion of the properties of singular
points at infinity. Here, in contrast to the
method by which Eq. (168) was studied,
we employ the more convenient (in the
present problem) homogeneous-coordinate
transformations
v = ;/z, y = 11/z. (181)
The straight line at infinity is fixed by the
equation z = 0. To reduce the number of
* See the paper by J. Kestin and S.K. Zaremba,
"Geometrical methods in the analysis of ordinary
differential equations," Appl. Sci. Res. B3: 144
189 (1953). Ch. 2. Qualitative Methods 255
variables to two, we first exclude the point
Q = z = 0. This can be done if we assume
that = 1. Then
v == 1/z, y : 1]/z.
In this case the differential system asso-
ciated with Eq. (180) is transformed to
dz __ dn (az2|b)ql-0J?z2
717*- " "l" af: **2-
It is convenient to introduce a new para-
meter, 9, in the following manner:
dt == zz dt). (182)
Then this system of equations (3311 be writ-
ten as
gg: -(azz-}b)1]c0z`*1]2z2=i!_,
d (183)
HT)-- T|Z .
Note, {irst, that the straight line at infinity,.
z -= 0, constitutes a trajectory of the dif-
ferential system (183) and as t grows (hence,
as 9 grows) the representative point moves
along this trajectory toward the only sin-
gular point z = iq = O. The characteristic
equation, which in this case has the form
b1jz = O, speciijes t]1e regular exceptional
direction z = O, with two repulsive regions
corresponding to this direction, which can
be establishedby studying the appropriate
255
variables to two, we ftrst exclude the point
~ = z = 0. This can be done if we assume
that s = 1. Then
v = 1/z, y = 11/z.
In this case the differential system asso-
ciated with Eq. (180) is transformed to
It is convenient to introduce a new para-
meter, 0, in the following manner:
dt = z
2
dO. (182)
Then this system of equations can be writ-
ten as
: ~ = - ( az
2
+ b) 11- ro:z2- 1J2z2 = "//.,
dz _ _
3
_q,r
"(ffi"- 'I'JZ - ~ .
(183)
Note, first, that the straight line at infinity,
z = 0, constitutes a trajectory of the dif-
ferential system (183) and as t grows (hence,
as e grows) the representative point moves
along this trajectory toward the only sin-
gular point z = 1J = 0. The characteristic
equation, which in this case has the form
-b11z = 0, specifies the regular exceptional
direction z = 0, with two repulsive regions
corresponding to this direction, which can
be established. by studying the appropriate 256 Diflerential Equations in Applications
z
QT I E
3:0
\
112/ x
\
\
?
/
\ /
j/
4)//
"Z Jr E
Fig. 95
vector field (Figure 95). The second excep-
tional direction 1] = 0 is singular and,
therefore, additional reasoning is required.
The locus of points 9 == U is a curve that
touches the straight line iq == 0 at the
origin and passes through the second and
third quadrants (Figure 95), which makes
it possible to fix three directions, I, II,
and III, on different sides of the symmetry
axis z = O. The region lying between the
curve fixed by the equation 9 = 0 and
the axis fr] = O is topologically equivalent
to two repulsive regions. Thus, on each side
of the straight line mq = 0 there is at least
Fig. 95
vector field (Figure 95). The second excep-
tional direction 11 = 0 is singular and,
therefore, additional reasoning is required.
The locus of points 'Y = 0 is a curve that
touches the straight line 11 = 0 at the
origin and passes through the second and
third quadrants (Figure 95), which makes
it possible to fix three directions, I, II,
and I I I, on different sides of the symmetry
axis z = 0. The region lying between the
curve fixed by the equation 'Y = 0 and
the axis 11 = 0 is topologically equivalent
to two repulsive regions. Thus, on each side
of the straight line 11 = 0 there is at least Ch. 2. Qualitative Methods 257
one t1ajectory that touches the straight
line at the origin.
If we now consider the differential equa-
tion
dn __ a b wt ii ._
jy;7+;;+7,;+;f( Z),
we note that for small values of | TII
0j __ 1 m?,
ia-? (ir?) <"
in the second quadrant between curve ay =
O and axis fr] = O. Hence, if we take two
phase trajectories with the same value of 1],
it is easy to see that the representative
points moving along these trajectories will
also move apart as z decreases. This means
that on each side of the straight line 1] = O
there can be only one phase trajectory that
touches this straight line at the origin
and belongs to the region considered. lt is
also clear that the qualitative picture is
symmetric about the axis z == O. More than
that, since for small values of z and positive
tj we have
la/2 I2 wi/luz I.
t.here can be no curves that touch the axis
1] = 0 at the origin and pass through the
first or fourth quadrant. Such curves are
also absent from the region that lies to the
left of the curve fixed by the equation
iy = O since in this case 9 is positive and
l7U770
257
one trajectory that touches the straight
line at the origin.
If we now consider the differential equa-
tion
dTJ _ a b + ~ + TJ _
dz --z+"'Z3 -:qz z--1 (rJ, z),
we note that for sma1l values of I '11 j
~ I = _!_ ( 1- ( ! ) ~ ) < 0
OTJ z TJ
2
in the second quadrant between curve ':!f =
0 and axis '11 = 0. Hence, if we take two
phase trajectories with the same value of '11
it is easy to see that the representative
points moving along these trajectories will
also move apart as z decreases. This means
that on each side of the straight line '11 = 0
there can be only one phase trajectory that
touches this straight line at the origin
and belongs to the region considered. It is
also clear that the qualitative picture is
symmetric about the axis z = 0. Mol'e than
that, since for small values of z and positive
'11 we have
I ':!liZ I ~ w:11 'I'JZ I,
there can be no curves that touch the axis
'11 = 0 at the origin and pass through the
first or fourth quadrant. Such curves are
also absent from the region that lies to the
left of the curve fixed by the equation
~ = 0 since in this case ':!/ is positive and
TZ has the same sign as z (Figure 95,
arrows IV). This reasoning suggests that
the singular point rj = z == O is a saddle
point. The two phase trajectories that reach
this point as t > 5}-oo are segments of the
straight line at infinity, z = O, that link the
previous point with the point 7; = z = 0.
The other two phase trajectories reach the
saddle point as t->- oo.
In ou1 reasoning we did not discuss the
point lf; = z = O. To conclude our investi-
gation, let us put 1] = 1 in (181). Then the
differential system associated with the dif-
ferential equation (180) can be written as
follows:
.9_..Zz_{_r(aZzbz 2 2..;
do- + +w,,Z)-1,
?;=Z (Z2+b2+wZ2)=Q,
where the variable 9 is defined in (182). The
point E, == z = U proves to be singula1, and
the characteristic equation here is z3 = 0.
Hence, the exceptional direction z = O is
singular, too. The curve fixed by the equa-
tion P (E, z) = Ol (Figure 96) touches the
axis z = O at the origin and has a cuspidal
point there, while the curve fixed by the
equation Q (, z) = 0 has a multiple point
at the origin. By studying the signs of P
and Q we can establish the pattern of the
vector iield (Figure 96) and the phase-tra-
Z has the same sign as z (Figure 95,
arrows IV). This reasoning suggests that
the singular point 'I'J = z = 0 is a saddle
point. The two phase trajectories that reach
this point as t -+- + oo are segments of the
straight line at infinity, z = 0, that link the
previous point with the point 6 = z = 0.
The other two phase trajectories reac.h the
saddle point as t -+-- oo.
In our reasoning we did not discuss the
point 6 = z = 0. To conclude our investi-
gation, let us put 11 = 1 in (181). Then the
differential system associated with the dif-
ferential equation (180) can he written as
follows:
ddj = z
2
+ s ( az
2
+ b
2
+ w:6z
2
) = P,
dz
'dO= 2. ( azz + b2 + w:6z2) = Q,
where the variable 0 is defined in (182). The
point 6 = z = 0 proves to be singular, and
the characteristic equation here is z
3
= 0.
Hence, the exceptional direction z = 0 is
singular, too. The curve fixed by the equa
tion P (6, z) = 0 ~ (Figure 96) touches the
axis z = 0 at the origin and has a cuspidal
point there, while the cmve f1xed by the
equation Q (, z) = 0 has a multiple point
at the origin. By studying the signs of P
and Q we can establish the pattern of the
vector lield (Figure 96) and the phase-tra-rCh. 2. Qualitative Methods 259
0:0 Z //
E /
x
\i\ \ \ / /"
/ // \
xy // \\ \
..4.. iv-
/ \\
\\
Fig. 96
\ z
//
\\ // g
( ,` t
/ - \\
\\
\
Fig. 97
jectory pattern in the neighborhood of the
singular point ig = z = O (Figure 97).
We note, for one thing, that far from the
axis ig': O there are no phase trajectories
that enter the singular point from the right.
17*
.Ch. 2. Qualitative !Vlethods
fl=O z
'
P=O ~
\ , ~ \
"'-.J'-.... ....... _'
/r ....... -- /
/ I
l
I.
-r
Fig. 96
Fig. 9 7
z
259
jectory pattern in the neighborhood of the
singular point a = z = 0 (Figure 97).
We note, for one thing, that far from the
axis a.= 0 there are no phase trajectories
that enter the singular point from the right.
This follows from the fact that in the first
and fourth quadrants
IZ/E I> IQ/P I-
The above reasoning shows that the Van
der Pol equation has no phase trajectories
that tend to infinity as t grows but has an
infinitude of phase curves that leave infinity
as t grows. This proves the existence of at
least one limit cycle for the Van der Pol
differential equation.
2.15 Curves Without Contact
In fairly simple cases the complete pattern
of the integral curves of a given differential
equation or, which is the same, the phase-
trajectory pattern of the corresponding
differential system is determined by the
type of singular points and closed integral
curves (phase trajectories), if the latter
exist. Sometimes the qualitative picture
can be constructed if, in addition to estab-
lishing the types of singular points, we can
find the curves, separatrices, that link the
singular points. Unfortunately, no general
methods exist for doing this. Hence, it is
expedient in qualitative integration to
employ the socalled curves without con-
tact. The reader will recall that a curve or
an arc of a curve with a continuous tangent
iS said to be a curve (arc) without contact
if it touches the vector (X, Y) specified
260 Differential Equations in Application1
This follows from the fact that in the ftrst
and fourth quadrants
I / ~ I > I Ql P 1.
The above reasoning shows that the Van
der Pol equation has no phase trajectories
that tend to infinity as t grows but has an
infinitude of phase curves that leave infinity
as t grows. This proves the existence of at
least one limit cycle for the Van der Pol
differential equation.
2.15 Curves Without Contact
In fairly simple cases the complete pattern
of the integral curves of a given differential
equation or, which is the same, the phase-
trajectory pattern of the corresponding
differential system is determined by the
type of singular points and closed integral
curves (phase trajectories), if the latter
exist. Sometimes the qualitative picture
can be constructed if, in addition to estab-
lishing the types of singular points, we can
find the curves, separatrices, that link the
singular points. Unfortunately, no general
methods exist for doing this. Hence, it is
expedient in qualitative integration to
employ the so-called curves without con-
tact. The reader will recall that a curve or
an arc of a curve with a continuous tangent
is said to be a curve (arc) without contact
if it touches the vector (X, Y) specified Ch. 2. Qualitative Methods 261
by the differential system (122) nowhere.
The definition implies that vector (X, Y)
must point in one direction on the entire
curve. Thus, a curve without contact may
be intersected by the phase trajectories of
(122) only in one direction when t grows
and in the opposite direction when t di-
minishes. Therefore, knowing the respective
curve may provide information about the
pattern of a particular part of a phase tra-
jectory.
Various auxiliary inequalities can be
used in qualitative integration of differen-
tial equations. Fo1 example, if two differ-
ential equations are known, say,
g'g";f(Tv y)v 'g%`Tg(xv y)
and it is known that f(;z:, y) { g (.2:, y)
in a domain D, then, denoting by yl (x)
a solution of the first equation such that
2/1(T0) = I/os with (xm yo) D and by
y2(x) a solution of the second equation
with the same initial data, we can prove
that yl (sr) { y2 (.2:) for x Q xo in D. But
if in D we have the strict inequality
f(v, y) < s (v y). then y1()< y2 (rv) fer
x>;1:0 in D and the curve y = y2 (0:) is
a curve without contact.
ential equation (168). We have already
Ch. 2. Qualitative Methorls
2fi1
by the differential system (122) nowhere.
The definition implies that vector (X, Y)
must point in one direction on the entire
curve. Thus, a curve without contact may
be intersected by the phase trajectories of
(122) only in one direction when t grows
and in the opposite direction when t di-
minishes. Therefore, knowing the respective
curve may provide information about the
pattern of a particular part of a phase tra-
jectory.
Various auxiliary inequalities can be
used in qualitative integration of differen-
tial equations. For example, if two differ-
ential equations are known, say,
dy )
dz = f (x, y'
~ ~ =g(x, y)
and it is known that I (x, y) ~ g (x, y)
in a domain D, then, denoting by y
1
(x)
a solution of the first equation such that
Y1 (xo) = Yo with (x
0
, Yo) ED, and by
y
2
(x) a solution of the second equation
with the same initial data, we can prove
that y
1
(x) ~ y
2
(x) for x ~ x
0
in D. But
if in D we have the strict inequality
I (x, y) < g (x, y), then y
1
(x) < y
2
(x) for
x > x
0
in D and the curve y = y
2
(x) is
a curve without contact.
ential equation (168). We have already 262 Differential Equations in Applications
shown that if G > G0, the equation allows
for two finite singular points, a saddle
point and a nodal point, which are points
of intersection of the straight line my
Gzx = 0 and the parabola 1 - py {
qG2:c2 = O. The segments of the straight line
and parabola that link these two finite sin-
gular points are curves without contact and
they specify a region of the plane which we
denote by A. If we put
X (fm y) = 1 - py + 062332,
Y (rv, y) = 2y (my - 6%),
it is easy to see that vector (X, Y) is di-
rected on the boundary of A outward
except at the singular points. Hence, if the
representative point emerges from any inte-
rior point of A and moves along an integral
curve with t decreasing, it cannot leave A
without passing through one of the singular
points. But since inside A we have
X (as, y) < 0, the singular point that is
attractive is the nodal point.
Finding the slopes of the exceptional
directions for the saddle point, we can see
that one of the exceptional straight lines
passes through A. This implies that an
integral curve that is tangent to this
straight line at the saddle point must enter
A and then proceed to the nodal point.
shown that if G > G
0
, the equation allows
for two finite singular points, a saddle
point and a nodal point, which are points
of intersection of the straight line my -
G
2
x = 0 and the parabola 1 - py +
qG
2
x
2
= 0. The segments of the straight line
and parabola that link these two finite sin-
gular points are curves without contact and
they specify a region of the plane which we
tlenot.e hy A.. If we put
X (x, y) = 1 - py + qG
2
x
2
,
Y (x, y) = 2y (my - G
2
.x),
it is easy to see that vector (X, Y) is di-
rected on the boundary of A. outward
except at the singular points. Hence, if the
representative point emerges from any inte-
rior point of A. and moves along an integral
curve with t decreasing, it cannot leave A
without passing through one of the singular
points. But since inside A we have
X (x, y) < 0, the singular point that is
attractive is the nodal point.
Finding the slopes of the exceptional
directions for the saddle point, we can see
that one of the exceptional straight lines
passes through A.. This implies that an
integral curve that is tangent to this
straight line at the saddle point must enter
A. and then proeeed to the nodal point. Ch. 2. Qualitative Methods 263
2.16 The Topographical System
of Curves. The Contact Curve
Earlier we noted that in the qualitative
solution of differential equations it is expe-
dient to use curves without contact. It must
be noted, however, that there is no general
method of building such curves that would
be applicable in every case. Hence, the
importance of various particular methods
and approaches in solving this problem.
One approach is linked to the selection of
an appropriate topographical system of
curves. Here a topographical system of curves
dehned by an equation cb (x, y) = C, with
C a real-valued parameter, is understood to
be a family of nonintersecting, enveloping
each other, and continuously dijjerentiable
simple closed curves that completely jill a
doubly connected domain G in the phase
plane.*
li the topographical system is selected
in such a manner that each value of para~
meter C is associated with a unique curve,
then, assuming for the sake of deiiniteness
that the curve corresponding to a deiinite
value of C envelopes all the curves with
smaller values of C (which means that the
"size" oi the curves grows with C) and that
no singular points of the corresponding
* Other definitions of a topographical system also
exist in the mathematical literature, but essentially
they differ little from the one given here.
2.16 The Topographical System
of Curves. The Cont.act Curve
203
Earlier we noted that in the qualilative
solution of differential equations it is expe-
dient to use curves without contact. It must
be noted, however, that there is no general
method of building such curves that would
he applicable in every case. Hence, the
importance of various particular methods
and approaches in solving this problem.
One approach is linked to the selection of
an appropriate topographical system of
curves. Here a topographical system of curues
defined by an equation <I> (x, y) = C, with
C a real-valued parameter, is understood to
be a family of nonintersecting, enveloping
each other, and continuously differentiable
simple closed curves that completely fill a
doubly connected domain G in the phase
plane.*
If the topographical system is selected
in such a manner that each value of para-
meter C is associated with a unique curve,
then, assuming for the sake of definiteness
that the curve corresponding to a definite
value of C envelopes all the curves with
smaller values of C (which means that the
"size" of the curves grows with C) and that
no poinls of the corresponding
* Other definitions of a topographical system also
exist in the mathematical literature, but essentially
they differ little from the one given here. 264 Differential Equations in Applications
differential system lie on the curves be-
longing to the topographical system, we
arrive at the following conclusion. If we
consider the function (D (cp (t), ip (t)) where
x = cp (t) and y = ip (t) are the parametric
equations of the trajectories of the differen-
tial system of the (122) type and calculate
the derivative with respect to t, that is,
d<1>(<.(> t\>()) 0D(. y)
i?":".w*X ( yl
then the curves (cycles in our case) without
contact are the curves of the topographical
system on which the derivative dD/dt
is a function of fixed sign. lf, in addition,
dD/dt is positive on a certain curve be-
longing to the topographical system, then
such a curve, being a cycle without contact,
possesses the property that all phase tra-
jectories intersecting it leave, with the
passage of time i, the finite region bounded
by it. And if dD/dt is negative, all the
phase trajectories enter this region. This
implies, for one thing, that if in an annular
region completely filled with curves belong-
ing to the topographical system the deriv-
ative d<D/dt is a function of fixed sign,
the region cannot contain any closed tra-
jectories, say, the limit cycles of the dif-
ferential system. Limit cycles may exist
only in the annular regions where the
differential system lie on the curves be-
longing to the topographical system, we
arrive at the following conclusion. If we
consider the function d> (cp (t), 'IJ' (t)) where
x = cp (t) and y = 'ljl (t) are the parametric
equations of the trajectories of the differen-
tial system of the (122) type and calculate
the derivative with respect to t, that is,
dcD (q> (t), ' (t)) iJcp (x, y) X ( )
dt iJx x, Y
+
iJCIJ (x, y) Y ( . )
iJy x, y '
then the curves (cycles in our case) without
contact are the curves of the topographical
system on which the derivative dd>/dt
is a function of fixed sign. If, in addition,
dd>/dt is positive on a certain curve be-
longing to the topographical system, then
such a curve, being a cycle without contact,
possesses the property that all phase tra-
jectories intersecting it leave, with the
passage of time t, the fmite region bounded
by it. And if dd>/dt is negative, all the
phase trajectories enter this region. This
implies, for one thing, that if in an annular
region completely filled with curves belong-
ing to the topographical system the deriv-
ative d<D/dt is a function of fixed sign,
the region cannot contain any closed tra-
jectories, say, the limit cycles of the dif-
ferential system. Limit cycles may exist
only in the annular regions where the Ch. 2. Qualitative Methods 265
derivative dD/dt is a function with alter-
nating signs.
To illustrate this reasoning, we consider
the differential system
d d
gif-=y:v+w3. H%=rvy+y3, (184)
which has only one singular point in the
finite part of the plane, O (0, O), a stable
focal point. For the topographical system
of curves we select the family of concentric
circles centered at point O (O, O), that is,
the family of curves specified by the equa-
tion x2 + y2 == C, with C a positive pa-
rameter. In our case the derivative dD/dt
is given by the relationship
d<D
-5 = 2 (wz + y2) l 2 (" + 1%*)..-
which in polar coordinates x = r cos 6 and
y = r sin 6 assumes the form
-S4?)-..:. 2r2-{ 2r* (cost 9 [sin*= 9) = 2r2
\~2r* ( cos 49).
Noting that the greatest and smallest values
of the expression in parentheses are 1 and
1/2, respectively, we arrive a the_ following
conclusion. For r greater than I/2 the value
of the derivative dCD/dt is positive Wh11G
for r less than unity it is negative. On the
basis of the Poincar-Bendixson criterion
18-0770
265
derivative dct>/dt is a function with alter-
nating signs.
To illustrate this reasoning, we consider
the differential system
dx dy
3
(ft=Y-x+x
3
, dt=-x-y+y, (184)
which has only one singular point in the
finite part of the plane, 0 (0, 0), a stable
focal point. For the topographical system
of curves we select the family of concentric
circles centered at point 0 (0, 0), that is,
the family of curves spec.if1ed by the equa-
tion x
2
+ y
2
= C, with C a positive pa-
rameter. In our case the derivative dll>/dt
is given by the relationship
d<l>
dt=-2 (x2+ y2)+ 2 (x'+ y4),.
which in polar coordinates x = r cos a and
y = r sin e assumes the form
~ ; =. - 2r
2
+ 2r" (cos' a+ sini 9) = - 2r
2
+2r
4
( ~ +{- cos4a).
Noting that the greatest and smallest values
of the expression in parentheses are 1 and
1/2, respectively, we arrive a the following
conclusion. For r greater than V2 the value
of the derivative dct>/dt is positive while
for r less than unity it is negative. On the
basis of the Poincare-Bendixson criterion
we conclude (if we replace t with ~t in
system (184)) that the annular region bound-
ed by the circles x2 + y2 -:1 and x2 +
y2 == 2 contains a limit cycle of system
(184), and this limit cycle is unique.
To prove the uniqueness it is sufficient
to employ the Dulac criterion for a doubly
connected domain: if there exists a function
B (x, y) that is continuous together with
its first partial derivatives and is such that
in adoubly connected domain G belonging
to the domain of definition of system (122)
the function
6(BX) 0(BY)
01: + Oy
is of fixed sign, then in G there can be no
more than one simple closed curve consisting
of trajectories of (122) and containing within
it the interior boundary of G.
ln our case, selecting B (x, y) E 1 for
system (184), we Hind that OX/0x +
6Y/dy = 3 (x2 i y2) 2 and, as can easily
be seen, in the annular region with bound-
aries x2 | y2 :1 and x2 | y2 = 2 the
expression 3 (x2 + y2) - 2 always retains
its sign. Allowing now for the type of the
singular point O (O, O), we conclude that
the cycle is an unstable limit cycle.
To return to the problem oi building
curves without contact in the general case,
let us examine a somewhat different way
we conclude (if we replace t with -t in
system (184)) that the annular region bound-
ed by the circles x
2
+ y
2
= 1 and x
2
+
y
2
= 2 contains a limit cycle of system
(184), and this limit cycle is unique.
To prove the uniqueness it is sufficient
to employ the Dulac criterion for a doubly
connected domain: if there exists a function
B (x, y) that is continuous together with
its first partial derivatives and is such that
in a doubly connected domain G belonging
to the domain of definition of system (122)
the function
iJ (BX) + a (BY)
ax ay
is of fixed sign, then in G there can be no
more than one simple closed curve consisting
of trajectories of (122) and containing within
it the interior boundary of G.
In our case, selecting B (x, y) = 1 for
system (184}, we find that ax/ax +
aY lay = 3 (x
2
+ y
2
} - 2 and, as can easily
be seen, in the annular region with bound-
aries x
2
+ y
2
= 1 and x
2
+ y
2
= 2 the
expression 3 (x
2
+ y
2
) - 2 always retains
its sign. Allowing now for the type of the
singular point 0 (0, 0}, we conclude that
the cycle is an unstable limit cycle.
To return to the problem of building
curves without contact in the general case,
let us examine a somewhat different way Ch. 2. Qualitative Methods 267
of using the topographical system of
curves. This approach is based on the
notion of a contact curve. Introduction of
this concept can be explained by the fact
that if the derivative dd)/dt vanishes on
a set of points of the phase plane, this set
constitutes the locus of points at which the
trajectories of the differential system touch
the curves belonging to the topographical
system. Indeed, the slope of the tangent to
a trajectory of the differential system is
Y/X and the slope of the tangent to a curve
belongirg to the topographical system is
-(Gil 0:1:)/(GCD/Gy). Thus, when
0 IJ 6iD
the slopes coincide, or
L - -@/23
X I- G1! , Gy
Hence, the locus of points at which the
trajectories of the differential system of the
(122) type touch the curves belonging to the
topographical system (defined by the equa-
tion (D (1, y) =- C) is called a contact curve.
The equation of a contact curve has the
(185) form. Of particular interest here is
the case where the topographical system
can be selected in such a manner that either
the contact curve itself or a real branch of
this curve proves to be a simple closed
curve. Then the topographical system con-
of using the topographical system of
curves. This approach is based on the
notion of a contact curve. Introduction of
this concept can be explained by the fact
that if the derivative dl>/dt vanishes on
a set of points of the phase plane, this set
constitutes the locus of points at which the
trajectories of the differential system touch
system. Indeed, the slope of the tangent to
a trajectory of the differential system is
YIX and the slope of the tangent to a curve
belongir-g to the topographical system is
-(ocl iJx)l(of!>/oy). Thus, when
a l1 X+ act> Y = 0
iJx ay ,
the slopes coincide, or
..!_ _ _ act> / act>
X- ax. oy '
(185)
Hence, the locus of points at which the
trajectories of the differential system of the
(122) type touch the curves belonging to the
tnpographieal system (defined by the equa-
tion f!> (:r, y) = C) is called a contact curve.
The equation of a contact curve has the
(185) fon1. Of particular interest here is
the case where the topographical system
can be selected in such a manner that either
the contact curve itself or a real branch of
this curve proves to be a simple closed
curve. Then the topographical system con-268 Differential Equations in Applications
/ \
x
\ /
@
\ /
\ /
Fig. 98
tains the "greatest" and "smallest" curves
that are touched either by a contact curve
or by a real branch of this curve (in Fig-
ure 98 the contact curve is depicted by a
dashed curve). If the derivative dD/dt
is nonpositive (nonnegative) on the "great-
est" curve specified by the equation
CD (sc, y) = C1 and nonnegative (nonposi-
tive) on the "smallest" curve specified by
the equation (D (az, y) = C2, then the annu-
lar region bounded by these curves contains
at least one limit cycle of the differential
system studied.
Specifically, in the last example (see
(184)% the contact curve specified by the
equation :1:2 -1- y2 =- .154 -{- y4 is a closed
curve (Figure 99), which enables us to find
the gIeatest" and "smallest" curves in the
selected topographical system that are
touched by the contact curve. Indeed, if
Fig. 98
tains the "greatest" and "smallest" curves
that are touched either by a contact curve
or by a real branch of this curve (in Fig-
ure 98 the contact curve is depicted by a
dashed curve). If the derivative dll>/dt
is nonpositive (nonnegative) on the "great-
est" curve specified by the equation
11> (x, y) = cl and nonnegative (nonposi-
tive) on the "smallest" curve specified by
the equation 11> (x, y) = C
2
, then the annu-
lar region bounded by these curves contains
at least one limit cycle of the differential
system studied.
Specifically, in the last example (see
(184)), the contact curve specified by the
equation x
2
+ y
2
= x
4
+ y
4
is a closed
curve (Figure 99), which enables us to find
the "greatest" and "smallest" curves in the
selected topographical system that are
touched by the contact curve. Indeed, if Ch. 2. Qualitative Methods 269
H
Fig. 99
we note that in polar coordinates the equa-
tion of the contact curve has the form rz =
(c0s 9 + sin 6)*, we can easily {ind the
parametric representation of this curve:
x __ cos 9 y __ sin 9
l/c0s* 9|sin*9 _ }/cos49{sin*9
Allowing now for the fact that the greatest
and smallest values of r2 are, respectively,
2 and l, we conclude that the "greatest
and "smal1est" curves of the topographical
Ch. 2. Qualitative
269
!I
Fig. 99
we note that in polar coordinates the equa-
tion of the contact curve has the form r
2
=
(cos
4
8 + sin
4
8)-t, we can easily find the
parametric representation of this curve:
X= cosO
V cos
4
O+ sin' 0
sinO
y =
v cos' O+ sin' 0
Allowing now for the fact that the greatest
and smallest values of r
2
are, respectively,
2 and 1, we conclude that the "greatest"
ilnd "smallel!t" cv.rves of the topographical 270 Differential Equations in Applications
system :::2 { y2 == C are the circles :1:2 }
y2 == 1 and :::2 + y2 == 2, respectively.
These circles, as we already know, specify an
annular region containing the limit cycle
of system (184).
2.17 The Divergence of a Vector
Field and Limit Cycles
Returning once more to the general case,
we note that in building the topographical
system of curves we can sometimes employ
the right-hand side of system (122). It
has been found that the differential system
may be such that the equation
0X GY
where A is a real parameter, specifies the
topographical system of curves.
For instance, if we turn to the differential
system (184), Eq. (186) assumes the form
3 (x2 | y2) 2 = it. Then, assuming that
A = (7t -[2)/3, with A 6 (-2, |oo), we
arrive at the topographical system of curves
x2 + y2 == A used above.
A remark is in order. If the "greatest" and
"smallest" curves merge, that is, the contact
curve or a real branch of this curve coin-
cides with one of the curves belonging to
the topographical system, such a curve is
a trajectory of the differential system.
system x
2
+ y
2
= C are the circles x
2
+
y
2
= 1 and x
2
+ y
2
= 2, respectively.
These circles, as we already know, specify an
annular region containing the limit cycle
of system (184).
2.17 The Divergence of a Vector
Fie]d and Limit Cyc]es
Returning once more to the general case,
we note that in building the topographical
system of curves we can sometimes employ
the right-hand side of system (122). It
has been found that the differential system
may be such that the equation
ax+ ay =-A.
f}x f}y I
(186)
where A. is a real parameter, specifies the
topographical system of curves.
For instance, if we turn to the differential
system (184), Eq. (186) assumes the form
3 (x2 + y
2
) - 2 = A.. Then, assuming that
A = (A.+ 2)/3, with A. E (-2, +oo), we
arrive at the topographical system of curves
x
2
+ y
2
= A used above.
A remark is in order. If the "greatest" and
"smallest" curves merge, that is, the contact
curve or a real branch of this curve coin-
cides with one of the curves belonging to
the topographical system, such a cune is
a trajectory of the diffe:reptial system, C11. 2. Qualitative Methods 271
Let us consider, for example, the differ-
ential system (173):
3;%: y+x (i>2y2>,
g/= $+9 (1- $2-*92)-
For this system
GX
-5;+%;; 2**4 ($2+!!2),
and Eq. (186) assumes the form 2 -
4 (:1:2 + y2) :: X. If instead of parameter R.
we introduce a new parameter A =
(2 - K)/4, with it E (-<x>, 2), then the
latter equation, which assumes the form
azz |- y2 = A, defines the topographical
system of curves. The contact curve in this
case is given by the equation (:1:** -{y2) X
(.2:2 -| y2 1) == 0 and, as we see, its real
branch :v + y2 == 1 coincides with one of
system. This branch, as shown on pp. 237-
239, proves to be a limit cycle of system
(173).
The last two examples, of course, illus-
trate, a particular situation. At the same
time the very idea of building a topograph-
ical system of curves by employing the
concept of divergence proves to be fruitful
and leads to results of a general nature.
We will not dwell any further on these re-
sults, but only note that the following
fact may serve as justification for what
ential system (173):
dx
dt =- - y + x (1- xz- yz),
dy
dt= x+ y (1- xz- yz).
For this system
iJX iJY
iJx +ay-=-2-4(x2+yz),
and Eq. (186) assumes the form 2 -
4 (x
2
+ y
2
) = A.. If instead of parameter A.
we introduce a new parameter A =
(2- A.)/4, with A. E (-oo, 2), then the
latter equation, which assumes the form
x
11
+ y
2
= A, defines the topographical
system of curves. The contact curve in this
case is given by the equation (x
2
+ y
2
) x
(x
2
+ y
2
- 1) = 0 and, as we see, its real
branch x
2
+ y
2
= 1 coincides with one of
system. This branch, as shown on pp. 237-
239, proves to be a limit cycle of system
(173).
The last two examples, of course, illus-
trate, a particular situation. At the same
time the very idea of building a topograph-
ical system of curves by employing the
concept of divergence proves to be fruitful
and leads to results of a general nature.
We will not dwell any further on these re-
sults, but only note that the following
fact may serve as justification for what 272 Differential Equations in Applications
has been said: if a diyferential system of the
(122) type has a limit cycle L, there exists a
real constant lt and a positive and continu-
ously diyfferentiable in the phase plane func-
tion B (sc, y) such that the equation
0(BX) 0(BY) _
*a?+*a;` X
specifies a curve that has a finite real branch
coinciding with cycle L.
ential system
-%;:.-.2:1;- 2:1:3 {cozy- Bxyz {- y3,
%= v+=v2y+y2.
which has a limit cycle specified by the
equation x2 I y2 = 1. For this system the
divergence of the vector field is
0X GY
5Tz:_"5!;T 2*** 5$2 3yz.
We can easily verify that there is not a
single real lt for which the equation 2
5.1:2 3y2 = lt specifies a trajectory of the
initial differential system. But if we take
a function B (x, y) = 3:::2 - 4xy + 7y2 +
3, then
0(BX) 0 (BY) ___ .
*"a;+_a;,(2+y)
>< (- 23x2-1-16xy-25y2-20)-14,
has been said: if a differential system of the
(122) type has a limit cycle L, there exists a
real constant A and a positive and continu-
ously differentiable in the phase plane func-
tion B (x, y) such that the equation
a (BX) + a (BY) A.
a:e ay
specifies a curve that has a finite real branch
coinciding with cycle L.
ential system
!: =2x-2x
3
+xzy- 3xyz+ y3,
~ = - -xs+xzy+xyz
dt . ,
which has a limit cyde specified by the
equation x
2
+ y
2
= 1. For this system the
divergence of the vector field is
ax+ aY =2-5xz-3yz.
ax iJy
We can easily verify that there is not a
single real A for which the equation 2 -
5x
2
- 3y
2
= A specifies a trajectory of the
initial differential system. But if we take
a function B (x, y) = 3x
2
- 4xy + 7y
2
+
3, then
a(BX) + a(BY) (xZ . 2-f)
ax ay -t- Y
x ( -23x
2
+ 16xy-25v
2
-20)-14, Ch. 2. Qualitative Methods 273
and the equation
_E.Q.)Q+ ..21.iYL Z .. 14
Ox Oy
specifies a curve whose finite real branch
$2 + y2 -1 =- 0 proves to be the limit
cycle.
In conclusion we note that in studying
concrete differential models it often proves
expedient to employ methods not discussed
in this book. Everything depends on the
complexity of the differential model, on
how deep the appropriate mathematical
tools are developed, and, of course, on the
erudition and experience of the researcher.
and the equation
o(BX) + rJ(BY) = _
14
ax ay
specifies a curve whose finite real branch
xB + y
2
- 1 = 0 proves to be the limit
cycle.
In conclusion we note that in studying
concrete differential models it often proves
expedient to employ methods not discussed
in this book. Everything depends on the
complexity of the differential model, on
how deep the appropriate mathematical
tools are developed, and, of course, on the
erudition and experience of the researcher. Selected Readmgs
Amelkin, V.V., and A.P. Sadovskii, Mathematical
Models and Dijferential Equations (Minsk:
Vysheishaya Shkola, 1982) (in Russian).
Andronov, A.A., A.A. Vitt, S.E. Khaikin, and
N.A. Zheleztsov, Theory of Oscillations (Read-
ing, Mass.: Addison-Wesley, 1966).
Braun, M., C.S. Cleman, and D.A. Drew (eds.),
Dijferential Equation Models (New York:
Springer, 1983).
Derrick, W.R., and S.I. Grossman, Elementary
Dijferential Equations with Applications, 2nd
ed. (Reading, Mass.: AddisonWesley, 1981).
Erugin, N'.P., A Reader in the General Course of
Diyjerential Equations (Minsk: Nauka i Tekh-
nika, 1979) (in Russian).
Ponomarev, K.K., Constructing Dijierential Equa-
tions (Minsk: Vysheishaya Shkola, 1973) (in
Russian).
Simmons, G.F., Diyfferential Equations with Ap-
plications and Historical Notes (New York:
McGraw-Hill, 1972).
Spiegel, M.H., Applied Diyfferential Equations
(Englewood Cliffs, NJ.: PrenticeIIall, 1981).
Selected Readings
Amel'kin, V.V., and A.P. Sadovskii, Mathematical
Models and Differential Equations (Minsk:
Vysheishaya Shkola, 1982) (in Russian).
Andronov, A.A., A.A. Vitt, S.E. Khaikin, and
N .A. Zheleztsov, Theory of Oscillations (Read-
ing, Mass.: Addison-Wesley, 1966).
Braun, M., C.S. Cleman, and D.A. Drew (eds.),
Differential Equation Models (New York:
Springer, 1983).
Derrick, W.R., and S.I. Grossman, Elementary
Differential Equations with Applications, 2nd
ed. (Reading, Mass.: Addison-Wesley, 1981).
Erugin, N.P., A Reader in the General Course of
Differential Equations (Minsk: Nauka i Tekh-
nika, 1979) (in Russian).
Ponomarev, K.K., Constructing Differential Equa-
tions (Minsk: Vyshcishaya Shkola, 1973) (in
Russian).
Simmons, G.F ., Differential Equations with Ap-
plications and Historical Notes (New York:
McGraw-Hill, 1972).
Spiegel, M.R., Applied Differential Equations
(Englewood Cliffs, N .1.: Prentice-Hall, 1981). Appendices
Appendix 1.
Derivatives of Elementary
Functions
Function I Derivative
C (constant) 0
rc 1
x" nx""1
1 _ 1
.2: x2
1 zz
F " Um?
_. 1
/ `_"".:'
I J: 2 ]/J:
... 1
1;/ x rz';/x"1
cx ex
ax ax In a
In :1: -1-
x
log x L log e= L-
:2: rv ln a
10 1 1 ~ 0.4343
E10 x x 0g10 8 ~ T
sin x cos x
cos :1: - sin x
Appendices
Appendix t.
Derivatives of Elementary
Functions
Function Derivative
C (constant) 0
X 1
xn nxn-l
1 1
X
-Xi
1 n
xn
- x1Hl
1
;-x
1
2}/:X
vx-
1
Ito/ xn-1
ex eX
ax ax Ina
lnx
1
X
1 1
log ax -loga e=-
1
-
x :r: na
1 0.4343
log
10
X

X X
sin x COS X
cosx -sin x
_ s:;s, _____ .. - .--... .... __ - -- _...., ____ --. __ ..._ ::;: _ 276 Appendices
Function Derivative
1
tan :1: -52-: secz as
00s :1:
1 _ 2 _
cot as r;- - csc IL
sm x
sin x
secu: -?=tanxseca:
cos sz
c0s x
CSCIB COtIl7CSC.IJ
SID .1:
. 1
SIH"1 I1?
II, 1 ___ x2
COS .27 *
1/1- me
1
tan1 x
1~|-- x2
1
colfl J:
1l~ :1:2
1 1
SCG- Il? ***.:;*.;
IB |/ xg ~ 1
1
csc*' .77
:1: |/ :1:2- 1
sinh sz cosh x
cosh 1: sinh 1:
n x
ta h 1
c0sh2a:
1
C01}h +
x s1nh2:1:
_ 1
Sll11l`1 :12 `7"`;`_"-
1/ 1 + $2
276
Function
tan x
cot x
secx
esc x
sin-
1
x
cos-
1
x
tan-
1
x
col-
1
x
scc-
1
:r:
csc-
1
x
sinh .x
cosh x
tanh x
coth x
sinh-
1
x
Appendices
Derivative
1
---=scc2x
cos2 x
1
- sin2x = -cscx
sinx
cos
2
x = tan x sec .;,
- ~ o s x = -cot x esc x
sm
2
x
1
v 1-x2
1
V 1-x2
1
1+:r:2
1
- 1-f-x2
1
x Jl x
2
-1
1
x vx
2
-1
cosh .x
sinh x
1
cosh2x
1
- sinh2x Appendices 277
Function I Derivative
(:0511-1 /C , SU >
'|/ xg- 1
l:.mh1 .2: ..i... _
1-_-:t2 1 I 'T I < 1
1
-1 ....
coth 2: 1__x2,I33]>i
Appendices
Function
tanh-
1
x
coth-
1
x
Derivative
1
- - - ~ , x>t
]/x
1
-1
1
1-x2' I xl < 1
1
1-xz , I :r I > 1
277 Appendix 2.
Basic Intcgrals
Power-Law Functions
In xn+1
S 22 d$="7z-TI_T (TL =}$
S -E1-=1n I .2: |
x
Trigonomctric Functions
Ssin as dx: c0s:v
Sc0s:vd;z:=sinx
Stanxdx= ln | cosxl
Sc0txda:=1n l sinxl
dx
S -7* = (-HI} {B
cos as
S .;*.2;.. .. t
sinza: -_-C0 x
Fractional Rational Functions
da: _ 1 _ x
S a=+x7T"" (1:)
dx 1 _ :c
$;::.;?;7*" 1 (7)
::-2-2;-111-%;- (| :1: | <a)
Appendix 2.
Basic Integrals
Power-Law Functions

xn+1
xndx=--
n+i

(n =1= 1)
Trigonometric Functions
sin x dx = -cos x
cos x dx = sin x
tan x dx = -In I cos x I
cotxdx=ln I sinx I

dx
--
2
-=t.anx
COS X

dx
-.--=-cotx
sm2x
Fractional Rational Functions
l
.l a2+xll a a
l
J a2-x2 a a
=_!_In a+x (lxl<a)
2a a-x Appendices 279
dx 1 _ x
Smw:7"h 1 (7)
=Q%1%;% (I = I >)
Exponential Functions
S ex ex
S ax dx: i
In a
Hypcrbolic Functions
S sinh x dx: cosh x
S cosh x dx: sinh x
Stanhx dx: ln lcoshxl
Scothxdx=ln I sinhxl
dx
S = tanh {IJ
dx
S 1: coth SB
Irrational Functions
dx x
= 11-1 ...)
S l/az-3:2 sm ( a
d .
S =cosh1(-ig-)=lH|(=-I-]/2)}
Appendices
\
J z -a a a
1 z-a
=-ln-- (I xI >a)
2a x+a
Exponential Functions
eX dx=ex
dx=
1
::
Hyperbolic Functions
sinh x dx= cosh x
cosh x dx= sinh x
tanhxdx=ln I coshx I
" .
coth z dx = ln I sinh x I
\ dzh. =tanh x
J cos z
\
J smh
2
z
Irrational Functions
279 To the Reader
Mir Publishers would be grateful for
your comments on the content, transla-
tion and design of this book. We would
also be pleased to receive any other
suggestions you may wish to make.
Our address is:
Mir Publishers
2 Pervy Rizhsky Pereulok
I-110, GSP, Moscow, 129820
USSR
To the Reader
Mir Publishers would be grateful for
your comments on the content, transla-
tion and design of this book. We would
also be pleased to receive any other
suggestions you may wish to make.
Our address is:
Mir Publishers
2 Pervy Rizhsky Pereulok
1-110, GSP, Moscow, 129820
USSR {Q5?

Differential Equations in Applications

Diunggah oleh

Informasi Dokumen

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Differential Equations in Applications

Diunggah oleh

Hak Cipta:

Format Tersedia

1

Thus, the shape of a

< N <\ <

(172) Ch. 2. Qualitative Methods 233

Anda mungkin juga menyukai