L.S. Pontryagin ORDINARY DIFFERENTIAL EQUATIONS ORDINARY DIFFERENTIAL EQUATIONS By L. S. PONTRYAGIN Translated from the Russian by LEONAS KACINSKAS and WALTER B. COUNTS Library of Congress A T T ADD IS ON-WES LEY PUBLISHING COMPANY, INC. HEADING, MASSACHUSETTS PALO ALTO LONDON ADIWES INTERNATIONAL SERIES IN MATHEMATICS A. J. L ohw ater Consulting Editor Copyright © 1962 ADDISON-WESLEY PUBLISHING COMPANY, INC. Printed in the United States of America ALL RIGHTS RESERVED. THIS BOOK, OR PARTS THERE­ OF, MAY NOT BE REPRODUCED IN ANY FORM WITH­ OUT WRITTEN PERMISSION OF THE PUBLISHERS. Library of Congress Catalog Card No. 62-17075 PREFACE This book has been written on the basis of lectures which I delivered at the department of mathematics and mechanics of Moscow State Uni­ versity. In drawing up the program for my lectures, I proceeded on the belief that the selection of material must not be random nor must it rest exclusively on established tradition. The most important and interesting applications of ordinary differential equations to engineering are found in the theory of oscillations and in the theory of automatic control. These applications were chosen to serve as guides in the selection of material. Since oscillation theory and automatic control theory without doubt also play a very important role in the development of our contemporary tech­ nical culture, my approach to the selection of material for the lecture course is, if not the only possible one, in any case a reasonable one. In attempting to give the students not only a purely mathematical tool suitable for engineering applications, but also to demonstrate the appli­ cations themselves, I included certain engineering problems in the lectures. In the book they are presented in §13, 27, and 29. I consider that these problems constitute an integral organic part of the lecture course and, accordingly, of this book. In addition to the material presented in the lectures, I have included in the book more difficult problems which were investigated in student seminars. They are contained in §19 and 31. The material contained in §24, 25, and 30 was only partially presented in the lectures. For the convenience of the reader, in the last chapter, the sixth, are presented certain facts from linear algebra in the form in which they are used in this book. In closing, I wish to express my gratitude to my students and to my closest co-workers V. G. Boltyanskiy, R. V. Gamkrelidze, and E. F. Mishchenko, who helped me in the preparation and delivery of the lectures and in writing and editing this book. I want also to note the decisive in­ fluence upon my scientific interests exerted by the outstanding Soviet specialist in the field of oscillation theory and automatic control theory, Aleksandr Aleksandrovich Andronov, with whom for many years I have had a friendly relationship. His influence has substantially affected the character and direction of this book. Moscow L. S. Pontryagin 16 July 1960 CONTENTS Chapter 1. Introduction.................................................................................. 1 1. First-order differential equations..................................................... 1 2. Some elementary integration methods.............................................. 6 3. Formulation of the existence and uniqueness theorem . . . . 18 4. Reduction of a general system of differential equations to a normal system........................................................................................................25 5. Complex differential equations.......................................................33 6. Some properties of linear differential equations..........................39 Chapter 2. Linear Equations with Constant Coefficients . . 41 7. Linear homogeneous equation with constant coefficients. The case of simple ro o ts ....................................................................................42 8. The linear homogeneous equation with constant coefficients. Case of multiple roots..............................................................................50 9. Stable polynomials........................................................................57 10. The linear nonhomogeneous equation with constant coefficients. . 62 11. Method of elimination...................................................................67 12. The method of complex amplitudes.................................................76 13. Electrical circuits..............................................................................80 14. The normal linear homogeneous system with constant coefficients . 94 15. Autonomous systems of differential equations and their phase spaces 103 16. The phase plane of a linear homogeneous system with constant coefficients........................................................................................115 Chapter 3. Linear Equations with Variable Coefficients . . 127 17. The normal system of linear equations...........................................127 18. The linear equation of nth order.......................................................137 19. The normal linear homogeneous system with periodic coefficients . 144 Chapter 4. Existence Theorems........................................................150 20. Proof of the existence and uniqueness theorem for one equation . 150 21. Proof of the existence and uniqueness theorem for a normal system of equations.........................................................................................159 22. Local theorems of continuity and differentiability of solutions. . 170 23. First integrals...................................................................................181 24. Behavior of the trajectories on large time intervals....................189 25. Global theorems of continuity and differentiability....................192 Chapter 5. Stability...............................................................................200 26. Lyapunov’s theorem........................................................................201 27. The centrifugal governor and the analysis of Vyshnegradskiy . . 213 v VI CONTENTS 28. Limit cycles...........................................................................................220 29. The vacuum-tube oscillator....................................................................236 30. The states of equilibrium of a second-order autonomous system . 244 31. Stability of periodic solutions..............................................................261 Chapter 6. Linear Alg ebra....................................................................277 32. The minimal annihilating polynomial...................................................277 33. Matrix functions.....................................................................................284 34. The Jordan form of a matrix....................................................................291 Index..................................................................................................................296 CHAPTER 1 INTRODUCTION This chapter is devoted primarily to the definition of those concepts which will be studied subsequently. What is a system of ordinary differ­ ential equations, what do we mean by a solution of it, and how many of these solutions exist? These are the basic questions which we shall attempt to answer in this chapter. The number of solutions is determined by theorems of existence and uniqueness, which will not be proved here, but only formulated. The proofs of these and of a number of other theorems of the same type are given in the fourth chapter, but theorems previously formulated in the first chapter are repeatedly used, so that their meaning is thus clarified. In addition to such basic information, solutions of differen­ tial equations of several of the simplest types are given in the first chapter. At the end of the chapter complex differential equations and their com­ plex solutions are studied, and elementary facts concerning systems of linear differential equations are given. 1. First-order differential equations. Equations in which the unknowns are functions of one or several variables and which contain not only the functions themselves, but also their derivatives, are called differential equations. If the unknown functions are functions of several variables, then the equations are called 'partial differential equations; in the opposite case, i.e., for the case of functions of only one independent variable, the equations are called ordinary differential equations. In this book we shall deal only with the latter. In applications to physics the time is taken as the independent variable, which is conventionally designated by the letter t; throughout this book the independent variable will be designated by t. Unknown functions will be designated by x, y, z, and so on. Derivatives of functions with respect to t will as a rule be designated by dots: x = dx/dt, x = d2x/dt2, and so on. When this is inconvenient or impossible, we shall denote the order of a derivative by an upper index in parentheses; for example, x n) = dnx/dtn. First we shall study the first-order differential equation. This equation may be written in the form Fit, x, x) = 0. (1) Here t is the independent variable, x the unknown function, x = dx/dt the derivative, and F a given function of three variables. The function F need not be defined for all values of its arguments; therefore we speak of 1 2 INTRODUCTION [CHAP. 1 the domain of definition B of the function F or simply, the domain B of F; here we have in mind a domain in the space of the three variables x, y, z. Equation (1) is called a first-order equation because it contains only the first derivative x of the unknown function x. A function x = tp(t) of the independent variable t, defined on a certain interval rx < t < r2 (the cases rx = ~cc,r2 = +oo are not excluded), which, when substituted for x in equation (1), reduces (1) to an identity on the entire interval rx < t < r2, is called a solution of equation (1). The interval rq < t < r2 is called the interval of definition of the solution <p(t). It is evident that sub­ stitution of x = ip(t) in (1) is possible only when the function <p(t) has a first derivative (and, in particular, is continuous) on the entire interval rq < I < r2. For the substitution of x = <p(t) into equation (1) to be possible, it is also necessary that the point with coordinates (t, <p(t), <p(t)) belong to the domain B of the function F for any value of t in the interval rq < t < r2. Relation (1) connects the three variables t, x, x. In certain cases it deter­ mines x as a single-valued, implicit function of the independent variables t, x. In this case (1) is equivalent to a differential equation of the form x = f(t, x). (2) Equation (2) is said to be solved explicitly for the derivative; in certain re­ spects it is more amenable to study than the general differential equa­ tion (I). It is such explicit equations which we shall now study. We shall not assume that (2) has been obtained as a result of solving (1) for x, but shall proceed from the function /((, x) as a given function of the two independent variables I, x. In order to visualize the situation geometrically, we introduce for study the fz-plane P. We shall plot t, as an independent variable, along the axis of abscissas, and x, as a dependent variable, along the axis of ordinates. The function / appearing in (2) need not be defined for all values of t and x, or, in geometric language, need not be defined at all points of the plane P, but only at points of a certain set r of P (Fig. 1). We shall assume that the set T is a domain. This means that for every point p in T there is some circle of positive radius with center at p also contained in I\ Con­ cerning the function /, it will be assumed that both the function itself and its partial derivative, df/dx, are continuous functions of t and x in T. A solution x = ip{t) of equation (2) may be thought of geometrically in P as a curve with the equation x = <p(l). This curve has a tangent at every point and lies entirely in the domain V; it is called an integral curve of the differential equation (2). Existence and uniqueness theorem. In algebra it is known that a large role is played by theorems which give the number of solutions to a given equation or system of equations. One such example is the fundamental