Author: Edward A. Desloge  

Tags: mechanics  

ISBN: 0-471-09144-8

Year: 1982

Text
                    CLASSICAL MECHANICS
Volume 2
EDWARD A. DESLOGE
Department of Physics
Florida State University
A WILEY-INTERSCIENCE PUBLICATION
JOHN WILEY & SONS
New York • Chichester • Brisbane • Toronto ♦ Singapore


Copyright © 1982 by John Wiley & Sons, Inc. All rights reserved. Published simultaneously in Canada. Reproduction or translation of any part of this work beyond that permitted by Section 107 or 108 of the 1976 United States Copyright Act without the permission of the copyright owner is unlawful. Requests for permission or further information should be addressed to the Permissions Department, John Wiley & Sons, Inc. Library of Congress Cataloging in Publication Data: Desloge, Edward A., 1926- Classical mechanics. "A Wiley-Interscience publication." Includes index. 1. Mechanics. I. Title. QC122.D47 531 81-11402 ISBN 0-471-09144-8 (v. 1)AACR2 ISBN 0-471-09145-6 (v. 2) Printed in the United States of America 10 987654321
Preface This book is the product of teaching classical mechanics on both the undergraduate and the graduate levels intermittently over the past 20 years. It covers mechanics in a unified fashion from the foundations, through elementary and intermediate mechanics, to a moderately advanced graduate level. A knowledge of calculus is presumed. Any other mathematics needed is provided in the appendices. Volume 1 can be used as a text for an undergraduate course. Volume 2 can be used as a text for a graduate course. By judicious deletion of material, courses of almost any length can be accommodated. The breadth and detail of the coverage is such that the book can also be used by students wanting to learn mechanics on their own, or by instructors wanting to direct students through self-paced programs. All the topics customarily found in an undergraduate or graduate text are covered, though frequently, in greater depth or detail, or from a slightly different point of view. In addition there are many interesting and useful topics that are seldom found in the standard texts. Hence this book can also serve as a source book for an instructor in mechanics, or as an aid to a researcher whose need for mechanics exceeds what is provided by the usual text. Much material is covered, but no attempt has been made to write an encyclopedic text on classical mechanics; rather, the subject is arranged in such a way that additional material can be inserted easily and naturally. A knowledge of mechanics that will continue to mature beyond the termination of a formal course requires a clear and accurate grasp of the organization of the subject. Throughout this book, I have tried to stress organization, clarity, and accuracy. This emphasis may sometimes appear to result in an approach that is overly stiff, detailed, and formal. The niceties of charm, elegance, and warmth, where lacking, can be supplied by a good instructor. A weakness in organization is almost impossible to repair. One of the most crucial stages in the exposition of any subject in physics is the choice of notation. A good notation is comprehensive; that is, it carries all the information that is required to avoid misinterpretation, and yet it is simple.
Preface Slight differences are often quite significant. For example it is both more meaningful and more useful to express the components of a vector A with respect to a frame S' as Ar rather than A-. I have thought a great deal about the notation. If I have erred, it is more often than not in using notation that is somewhat overloaded. The appearance of many of the discussions and derivations could be improved by stripping the notation of some of its appendages. The loss however would generally outweigh the gain. For example, the appearance of arguments involving partial derivatives, such as (3// 3x)y or equivalently df(x, y)/dy could be improved by writing such terms simply 3//3.x. I have at times done this myself. However, anyone who has taught a course in thermodynamics will verify that the consequences can be disastrous, if one is not careful. In writing this book, I have used a number of organizational and pedagogical devices that I have found over the years to be particularly helpful. (1) The material is highly subdivided, to give emphasis to the organization and to facilitate reorganization. (2) The chapters are short, to make assimilation easier and to aid in the addition, deletion, or rearrangement of material. (3) Formal definitions, postulates, and theorems are frequently used, to emphasize and clarify important concepts and to foster exactness. (4) Most chapters contain one or more examples designed to illustrate an idea, or to provide general techniques for the solution of problems, rather than to serve as models to be slavishly imitated. (5) There are a large number of problems, since the ability to systematically solve a wide variety of problems is one of the goals of a course in mechanics. (6) All mathematical developments needed are provided in appendices. In this way the continuity of a physical argument is not broken, and yet the mathematical tools are readily available. A number of benefits, other than problem solving, can be derived from a course in classical mechanics. Three in particular have influenced me strongly in writing this book. 1. Since physics as we know it today had its origins in classical mechanics, the very structure and language of physics is permeated by ideas that come from classical mechanics; hence a knowledge of the foundations of mechanics can provide a student with a deeper grasp of all of physics. I have therefore devoted considerable effort to developing the foundations of classical mechanics. 2. A course in classical mechanics is not only a course in physics but a course in applied mathematics. The inclusion of certain topics, and the extensive mathematical appendices in this book, reflect this aspect of mechanics. As much consideration has been given to the writing of the appendices as to the text proper. Though relegated to appendices, the mathematical topics are an integral part of the text. 3. The human brain contains two hemispheres whose characters have been shown to be different but complementary. In most individuals, the right
Preface vii hemisphere, which is associated directly with the left hand, the left field of vision, and so forth, is superior in handling geometrical concepts, and the left hemisphere, which is associated directly with the right hand, the right field of vision, and so forth, is superior in handling formal analytical concepts. Learning physics involves an interesting interplay between these two abilities. The left hemisphere provides the analytical map that takes us from one point to another, while the right hemisphere provides the geometrical vision necessary to see the goal and landmarks along the way. A course in classical mechanics offers a marvelous vehicle for developing and integrating both the geometrical and the analytical powers of the brain. The Newtonian approach to mechanics is strongly geometrical, whereas the Lagrangian and Hamiltonian approaches are strongly analytical. It follows that a course designed to exercise both sides of the brain should, as this text does, include a thorough foundation in Newtonian mechanics before proceeding to Lagrangian and Hamiltonian mechanics. Too many modern courses in classical mechanics are built on a weak foundation of Newtonian mechanics, with the resultant complaint by instructors that a particular student is good at mathematics but does not seem to have any physical intuition. To make Volume 1 adequate for an undergraduate course and Volume 2 adequate for a graduate course, certain topics are covered in both volumes. Central force motion, the differential scattering cross section, and small oscillations are introduced in Volume 1, and reviewed and extended in Volume 2. Rigid body motion is covered in great detail from the Newtonian point of view in Volume 1, and briefly reviewed and treated from the Lagrangian point of view in Volume 2. An introduction to Lagrangian and Hamiltonian mechanics is given in Volume 1, to prepare the student for the full treatment in Volume 2. To conclude this preface, I point out some aspects of the book that are unique or are, in my opinion, treated better or more thoroughly here than elsewhere. In Chapters 1-4, 85, and 86 the basic principles of Newtonian and relativistic kinematics are derived starting from the assumption of the existence and equivalence of inertial frames of reference. With this assumption, the law of transformation between inertial frames arises naturally and contains only one undetermined parameter, which is identified as the upper speed with which a particle can move with respect to an inertial frame. By choosing this parameter to be infinite we obtain the Galilean transformation, and by choosing it to be finite we obtain the Lorentz transformation. In Chapters 8, 9, 88, and 89 an investigation of quantities that might be conserved in a collision leads naturally to the momentum conservation laws of Newtonian and relativistic mechanics, and from these laws to the definitions of force in both Newtonian and relativistic mechanics.
viii Preface The treatment of the foundations of Newtonian and relativistic mechanics contained in the two above-mentioned sets of chapters clearly brings out the close relation between Newtonian and relativistic mechanics, and the primacy of the law of conservation of momentum over Newton's equation of motion and its counterpart in relativistic mechanics. It is the most thorough treatment that can be found anywhere. Many of the details are new and unique. Chapter 6 presents very carefully and thoroughly the definition and properties of the angular velocity of one coordinate system with respect to another. By starting with the analytical definition of angular velocity rather than the geometrical definition, as is customarily done, many of the difficulties associated with this concept are avoided. Even though a good grasp of this concept is a prerequisite to an ability to express the laws of motion in rotating frames, and to an understanding of the kinematics and dynamics of rigid bodies, it is amazing to me how many of the standard texts are weak on this subject, and frequently contain spurious definitions of angular velocity. Chapter 23 contains one of the simplest and most organized treatments of central force motion of which I am aware. Chapters 24, 26, 60, and 61 contain a very complete treatment of the differential scattering cross section. Chapter 61 derives the relationship between the center of mass and laboratory cross sections for a collision in which both particles are moving, the collision is inelastic, and the product particles differ from the incident particles. No other text that I know of contains this complete result. The treatment of rigid body motion in Chapters 34-40, with the possible addition of Chapters 41, 63, and 64, is sufficiently complete and detailed to form a course by itself. Chapter 38 is a thorough treatment of the inertia tensor from both the analytical and the geometric points of view. Chapter 39 contains an extremely detailed treatment of Euler angles, and the kinematics of rigid body motion. The expression in Chapter 39 of the equations of transformation between different reference frames and the relative angular velocities between these frames in terms of Euler angles is a very useful aid to anyone interested in solving rigid body motion problems. Chapters 47-54 give a thorough treatment of Lagrange's equations of motion and include a detailed treatment of constraints both holonomic and anholonomic. Since most undergraduate courses do not have the time for such a complete treatment, an introductory abbreviated treatment appears in Chapters 42-44. The treatment of small vibrations in Chapters 45 and 65 contains a number of results that cannot be found elsewhere. The material in Chapter 66 together with the material in Appendices 28-30 represents a complete introduction to group theory and its application to symmetrical vibrating systems. Although the complexity of the notation makes the going a little tedious at times, the text contains none of the gaps and guesswork that seem to mar the treatments I have found elsewhere.
Preface ix Quasi-coordinates and the Gibbs-Appell equations of motion covered in Chapters 67-70 are probably unknown to most physicists. There are only a few mechanics books in which they are even mentioned. Because I think they deserve more attention and are quite useful in solving certain problems, I have included them. The treatment of Hamilton's equations of motion, canonical transformations, and Hamilton-Jacobi theory in Chapters 71-78 is quite unusual in that no use is made of the calculus of variations. Although this results in many of the proofs being a little longer than usual, I feel that it provides a more secure foundation in the subject, since most students who are encountering this subject for the first time are not completely sure of themselves in the use of the calculus of variations. This unsureness is usually compounded by the casual and sometimes erroneous use of the calculus of variations by some authors. Although I prefer not to use the calculus of variations in a student's first encounter with Hamiltonian mechanics, I certainly believe it to be an extremely useful tool, hence have devoted Part 8 and Appendix 32 to this subject. The initial separation of the calculus of variations from Hamiltonian mechanics has the added organizational advantage of allowing the instructor who so desires to pursue the subject beyond and apart from what is required in Hamiltonian mechanics. The presentation of the foundation and basic principles of relativistic mechanics in Part 9 is I believe one of the most logical and straightforward to be found anywhere. Though some of the proofs are a little ponderous, the general flow of concepts does not require an extraordinary distortion of a student's imagination. As a consequence relativistic mechanics seems almost inevitable. The nature and importance of constants of the motion is stressed time and again throughout the text, starting with Chapter 17 and proceeding through Chapters 33, 46, 56, 71, and 81. Chapter 56 considers in detail the relationship between constants of the motion and the invariance of the Lagrangian under certain transformations, and Noether's theorem is presented without making use of variational techniques as is usually done in those few sources where this theorem can be found. The subject of impulse, which usually causes students a great deal of unnecessary grief, is developed in detail in Chapters 18, 41, and 57. Many of the appendices are quite useful in themselves, apart from their value in the body of the text. However since this book is intended primarily as a mechanics text, and only secondarily as a course in applied mathematics, many theorems in the appendices are stated without proof. Proofs that are particularly pertinent or cannot easily be found elsewhere are given. In Appendices 4-6, and 23, the reader is led systematically from the geometric concept of a vector as a directed line segment to the highly analytical concept of general tensors. With a little amplification and completion of proofs, the material would make a good course in vectors and tensors.
x Preface Similarly the material in Appendices 12 and 24-26, with the possible addition of the material on quadratic forms in Appendix 27, forms a good outline for a course in matrices. Appendix 32 provides an excellent introduction to the calculus of variations. Appendices 11, 15, 19, and 22 cover a number of topics very important to classical mechanics in a manner that is both simpler and clearer than can be found elsewhere. While writing this book I have not had in mind a hypothetical audience, but rather have written as if I were to be the reader. The book is in a sense a reflection of myself. Its exposure to numerous students over the years has sharpened rather than altered this reflection. I suspect that this is how most texts are written. Interestingly, I am dominantly a right hemisphere thinker— that is, I think in terms of pictures—but the first impression one gains of the text is that it is dominantly analytical. The probable explanation of this apparent anomaly is that one tends to emphasize the things that are personally difficult while at the same time ignoring what comes easily, with the result that the material is more analogous to a photographic negative than to a positive print. In any case the success of this book will depend on how many others share the difficulties, problems, loves, and hates that I experienced in learning classical mechanics. I hope that there are many, and that through this book they will derive some of the pleasures I have found in my encounter with classical mechanics. Edward A. Desloge Tallahassee, Florida December 1981
Contents VOLUME 1 Introduction PART 1. THE NEWTONIAN MECHANICS OF PARTICLES Section 1. The Basic Principles of Newtonian Kinematics 1. Space and Time, 7 2. Inertial Frames, 11 3. Transformation Between Inertial Frames, 13 4. Absolute Space and Time, 4 Section 2. Auxiliary Principles of Newtonian Kinematics 5. Relative Motion of Particles, 27 6. Relative Motion of Frames of Reference, 32 7. The Description of Motion Using Orthogonal Curvilinear Coordinates, 43 Section 3. The Basic Principles of Newtonian Dynamics 8. Mass and Momentum, 49 9. Force, 59 10. Center of Mass, 65 Section 4. Elementary Applications of Newton's Law 11. Some Basic Forces, 73 12. Statics of a Particle, 83 13. Dynamics Problems, 88 25 47 71
Contents Section 5. Auxiliary Principles of Newtonian Dynamics 14. Torque and Angular Momentum, 101 15. Work and Kinetic Energy, 107 16. Potential Energy, 113 17. Constants of the Motion, 120 18. Impulse, 126 19. The Equations of Motion in Noninertial Reference Frames, 136 20. The Equations of Motion in Orthogonal Curvilinear Coordinate Systems, 144 99 Section 6. Applications 21. One Dimensional Motion in an Arbitrary Potential, 151 22. The Harmonic Oscillator, 158 23. Central Force Motion, 172 24. The Differential Scattering Cross Section, 187 25. Two Particle Systems, 196 26. Two Particle Collisions, 203 149 PART 2. THE NEWTONIAN MECHANICS OF SYSTEMS OF PARTICLES Section 1. Basic Principles 27. Dynamical Systems, 215 28. Force and Linear Momentum, 218 29. Torque and Angular Momentum, 226 30. Work and Kinetic Energy, 234 31. Cartesian Configuration Space, 240 32. Potential Energy, 243 33. Constants of the Motion, 247 213 Section 2. Rigid Body Motion 34. Rigid Bodies, 253 35. Equivalent Systems of Forces, 255 36. Statics of a Rigid Body, 264 37. Uniplanar Motion of a Rigid Body, 269 38. The Inertia Tensor, 286 251
Contents xiii 39. Rigid Body Kinematics, 303 40. Rigid Body Dynamics, 320 41. Impulsive Motion of a Rigid Body, 338 PART 3. AN INTRODUCTION TO LAGRANGIAN AND HAMILTONIAN MECHANICS Section 1. Lagrangian Mechanics 42. Holonomic Constraint Forces, 349 43. Generalized Coordinates for Holonomic Systems, 352 44. Lagrange's Equations of Motion for a Holonomic System, 361 347 Section 2. Applications of Lagrangian Mechanics 45. Vibrating Systems, 373 371 Section 3. Hamiltonian Mechanics 46. Hamilton's Equations of Motion, 387 385 SUPPLEMENTARY MATERIAL Appendices 1. Analytical Representation of a Sine Function, 397 2. Partial Differentiation, 398 3. Jacobians, 401 4. Vector Algebra, 402 5. Vector Calculus, 408 6. Cartesian Tensors, 417 7. Orthogonal Curvilinear Coordinates, 427 8. Ordinary Differential Equations, 432 9. Linear Differential Equations, 436 10. Differentiation of an Integral, 441 11. Exact Differentials, 442 12. Matrices, 447 13. Systems of Linear Equations, 458 14. Functional Dependence, 460 15. The Method of Lagrange Multipliers, 464 16. Elliptic Functions, 467 17. Coordinate Transformations, 470 395
xiv Contents Tables 475 1. Abbreviations of Units, 477 2. Constants, 477 3. Conversion Factors, 478 4. Centers of Mass, 479 5. Moments of Inertia, 482 6. Vector Identities, 486 Answers 489 Combined Index 1-1
VOLUME 2 PART 4. LAGRANGIAN MECHANICS Section 1. 47. 48. 49. 50. 52. 53. 54. 55. Lagrange's Equations of Motion Generalized Coordinates, 513 Lagrange's Equations of Motion for Elementary Systems, 521 Constraint Forces, 528 Lagrange's Equations of Motion for Holonomic Systems, 538 The Determination of Holonomic Constraint Forces, 549 Lagrange's Equations of Motion for Anholonomic Systems, 554 Generalized Force Functions, 558 Lagrange's Equations of Motion for Lagrangian Systems, 564 Lagrange's Equations of Motion and Tensor Analysis, 570 511 Section 2. Auxiliary Principles of Lagrangian Mechanics 56. Constants of the Motion in the Lagrangian Formulation, 575 57. Lagrange's Equations of Motion for Impulsive Forces, 588 573 PART 5. APPLICATIONS OF LAGRANGIAN MECHANICS Section 1. Central Force Motion 58. Central Force Motion, 597 59. Bertrand's Theorem, 606 60. The Differential Scattering Cross Section, 614 61. Two Particle Collisions, 622 62. The Restricted Three Body Problem, 634 595
xvi Section 2. Rigid Body Motion 63. Rigid Body Kinematics, 647 64. Rigid Body Dynamics, 653 Section 3. Small Oscillations 65. Vibrating Systems, 665 66. Symmetrical Vibrating Systems, 682 645 663 PART 6. 67. 68. 69. 70. QUASI-COORDINATES Quasi-Coordinates, 711 Lagrange's Equations for Quasi-Coordinates, 715 The Gibbs-Appell Equations of Motion, 720 The Gibbs-Appell Equations and Rigid Body Motion, 727 PART 7. HAMILTONIAN MECHANICS Section 1. Hamilton's Equations of Motion 735 71. Hamilton's Equations of Motion, 737 72. Equations of Motion of the Hamiltonian Type, 744 73. Point Transformations, 750 Section 2. Canonical Transformations 753 74. Canonical Transformations, 755 75. A Condensed Notation, 768 76. Hamilton's Canonical Equations of Motion, 773 Section 3. Hamilton-Jacobi Theory 787 77. Generating Functions, 789 78. The Hamilton-Jacobi Equations, 797 79. Action and Angle Variables, 807 Section 4. Poisson Formulation of the Equations of Motion 821 80. Poisson Brackets, 823 81. The Poisson Formulation of the Equations of Motion, 828
Contents xvii PART 8. VARIATIONAL PRINCIPLES IN CLASSICAL MECHANICS 82. D'Alembert's Principle, 835 83. Hamilton's Principle, 838 84. The Modified Hamilton's Principle, 843 PART 9. RELATIVISTIC MECHANICS Section 1. Relativistic Kinematics 85. The Basic Postulates of Relativistic Kinematics, 857 86. The Lorentz Transformation, 859 87. Some Consequences of the Lorentz Transformation, 867 855 Section 2. Relativistic Dynamics of a Particle 873 88. Mass and Momentum, 875 89. Force, 885 90. Work and Energy, 888 Section 3. Four Dimensional Formulation of Relativistic Mechanics 891 91. Four Dimensional Formulation of Relativistic Mechanics, 893 Section 4. Relativistic Lagrangian and Hainiltonian Mechanics 899 92. Relativistic Lagrangian Mechanics, 901 93. Relativistic Hamiltonian Mechanics, 903 SUPPLEMENTARY MATERIAL Appendices 18. Inverse Transformations, 909 19. Change of Variables in an Integral, 912 20. Euler's Theorem, 914 21. The Gram-Schmidt Orthogonalization Process, 915 22. Legendre Transformations, 918 23. General Tensors, 921 907
xviii Contents 24. Matrix Transformations, 928 25. Eigenvalues of Matrices, 931 26. Diagonalization of Matrices, 935 27. Quadratic Forms, 937 28. Group Theory, 945 29. Vector Spaces, 951 30. Representations of a Group, 955 31. Multiply Periodic Functions, 967 32. Calculus of Variations, 968 Answers 981 Combined Index 1-1
CLASSICAL MECHANICS Volume 2
PART LAGRANGIAN MECHANICS
SECTION Lagrange's Equations of Motion
47 Generalized Coordinates INTRODUCTION The material in this and the chapters that follow provides a complete and detailed analysis of Lagrangian dynamics. There is some danger however that, because of the detail and the generality of the results, the reader will lose sight of the essential simplicity of the Lagrangian approach. To counteract this danger, any reader who has not previously encountered Lagrange's equations of motion should read the material in Chapters 42-44 in Volume 1 before starting this section. NEWTON'S LAW The position of a particle can be specified by giving its coordinates Xi , Xy^ 3 with respect to a Cartesian coordinate system S fixed in an inertial frame. If the mass of the particle is m and the components of the force acting on the particle are/,, /2, and/3, the motion of the particle can be determined by Newton's law / = mxi i = 1,2,3 (1) The configuration of a system of N particles can be specified by the set of coordinates jc,(1),x2(l),x3(l),x,(2), ..., where xt{n) is the /th coordinate of the A th particle. If the mass of the nth particle is m(n) and the components of the force acting on the nth particle are/,(«),/2(«), and/3(«), the motion of the system can be determined by the set of equations /,(«) = m{n)xt{n) /-1,2,3 n= 1,..., N 513
514 Generalized Coordinates The discussion of systems of particles can be considerably simplified if we use the notation introduced in Chapter 31, that is, if we let Xi(n) = X3n-3 + i fi(n)=f3n-3 + i m(n) = m3n_2 = m3n_x = m3n (3) With this notation, the configuration of a system of N particles can be described by the set of coordinates x = xux2, . . . , x3N. Geometrically, the configuration x of the system can be represented by a single point in the Cartesian configuration space x, which is the space whose coordinates are the members of the set of coordinates x. The motion of this point is governed by Newton's law, which in terms of the set of coordinates x is simply fi = mixi /=1,2, ...,37V (4) Besides the set x there are many other sets of coordinates that could be used to describe the configuration of the system. Thus in this Section we consider the possibility of using sets of coordinates other than x, and we determine the modification of the equations of motion that must be made when they are used. GENERALIZED COORDINATES AND VELOCITIES Let us assume that we are given a dynamical system consisting of N particles. If there exists a set of 37V coordinates g]9 g2, . . . , g3N that uniquely specifies the configuration of the system, the coordinates gl9 g2, . . . , g3N are called generalized coordinates and their time rates of change gl9 g2, . . . , g3N are called generalized velocities. We designate the set gl9 g2, . . . , g3N as the set g, and the set g]9 gl9 . . . , g3N as the set g. Just as we can represent the configuration of a system by a point in the Cartesian configuration space x, we can more generally represent the configuration of a system by a point in a space whose coordinates are any set of generalized coordinates g. Such a space is called simply a configuration space. The Cartesian configuration space x is a particular example of a configuration space. We call the configuration space whose coordinates are the set g, the configuration space g. note: The reader who is already familiar with the use of generalized coordinates may be disturbed by the use of the symbol g{ to represent a generalized coordinate rather than the more conventional qr The symbol gi was chosen to permit us initially to distinguish between the full 37V dimensional configuration space and the reduced configuration space that is used when holonomic systems are being considered. The letters qi are reserved for the latter case.
Generalized Components of a Force 515 VIRTUAL DISPLACEMENTS AND VIRTUAL WORK Generalization of the equations of motion requires not only the generalization of what is meant by coordinates, but also generalization of what is meant by the components of a force. Before we can carry this out, the concepts of virtual displacement and virtual work must be introduced. A virtual displacement of a dynamical system is a hypothetical change in the configuration of the system at some fixed time t. The notation 8g=8gu §g . . , 8g3N designates an infinitesimal virtual displacement. Two points should be particularly noticed in the foregoing definition. First a virtual displacement occurs at a particular time t and does not involve a change in the time. We can imagine the system to be initially frozen in its configuration at time t, at which point we step in and alter its configuration. The second point to notice is that a virtual displacement does not in general correspond to the actual displacement the system undergoes as a result of the forces acting on it. It is for this reason that we use the notation Sg rather than dg, since the quantity dg usually designates the actual displacement of the system in a time dt. The virtual work 8W of a force in the virtual displacement 8g is the work that would be done by the force if the system underwent the virtual displacement 8g. Since a virtual displacement occurs at a fixed time t, the value of the force that is acting on the system at the time t is used in calculating the virtual work at time t. If the force is a velocity dependent force, the value of the velocity of the system at the time t is used in evaluating the force. GENERALIZED COMPONENTS OF A FORCE With the introduction of the concept of virtual work we are in a position to define generalized components of a force. This is done in the following theorem. Theorem 1. If a dynamical system consisting of N particles, whose configuration can be defined in terms of the set of generalized coordinates g = gx, £2* • • • , £3a/ is acted on by a force, the virtual work 8W of the force in an arbitrary infinitesimal virtual displacement Sg can be written in the form SW'-SG/Sft (5) i The quantities Gx, G2, ..., G3N are called generalized components of the force. A particular generalized component Gt is called the gt generalized component of the force or simply the gt component of the force. The set of generalized components is designated G. The force whose generalized components are the set G is called the force G.
516 Generalized Coordinates proof: At a given time the virtual work 8W depends only on the force and the virtual displacement Sg. If we expand 8W in powers of Sg1? 8gi> • • • > ^3n anc* note ^at tne higher order terms can be dropped because the 8G; are infinitesimals, we obtain Eq. (5). This completes the proof. Since Eq. (5) must be true for arbitrary infinitesimal virtual displacements Sg, the gi component of a force can be determined by calculating the work done in an infinitesimal displacement in which the generalized coordinate g is varied while all the other generalized coordinates are held fixed, then dividing the work done SW by the displacement Sg. If we choose the set of coordinates x as our generalized coordinates, the work done in an infinitesimal displacement is given by «^=S/y fix, (6) i It follows that the x( generalized component of the force, which we designate as X( or Gx, is just/, that is, G^X^f,. (7) note 1: The g component of a force does not in general have the dimensions of a force. For example if g is an angle, the g component of the force will have the dimensions of a torque, not a force. It follows that the g component of a force cannot be interpreted as the component of the force in the gi direction. The relationship between these two quantities is considered in greater detail in a later section. note 2: The value of the g component of a force depends not only on the coordinate g but also on the other coordinates in the set g, in the same sense that a partial derivative depends not only on the quantity that is varied but also on the quantities that are held constant. It would therefore be more precise to speak of "the g component in the space g" than of "the g component." However for the sake of simplicity we use the shorter terminology. THE TRANSFORMATION OF THE GENERALIZED COMPONENTS OF A FORCE If we know the generalized components of a force in one space and we wish to find the generalized components in another space we can use the following theorem. Theorem 2. If Gu G2, . . . , G3N are the components of a force in the space whose coordinates are g,, g2, . . . , g3N, the components Gv, G2, ..., G3N, in
The Physical Components of a Force 517 the space whose coordinates are gr, gT, ..., g3N, are given by proof: From Theorem 1 the Gj are defined by the condition SW^GjSgj (9) J and the Gr are defined by the condition 8W=2lGr8gi, (10) But % = 2^> (ii) Combining Eqs. (9)-(11) we obtain SMfr-SSGylrte (12) /" /' j Si' Equating the coefficients of 8gr on both sides we obtain Eq. (8). If we use Theorem 2 to transform generalized components in the space x to generalized components in the space g we obtain Gi-Zfj-ifc1 <13> Equation (13) can be and often is used as the definition of the gi generalized component of a force. The definition given in Theorem 1 is preferable for two reasons: (a) it is ordinarily easier to calculate the generalized components of a force starting with Eq. (5) in Theorem 1 than with Eq. (13); (b) using Eq. (13) as the definition elevates the Cartesian configuration space to a special status, which partially defeats our ultimate purpose, namely, to put all configuration spaces on an equal footing. THE PHYSICAL COMPONENTS OF A FORCE If the coordinate gi undergoes an infinitesimal virtual change 8gt while the remaining members of the set g are held fixed, the point x representing the configuration of the system in the Cartesian configuration space will undergo an infinitesimal virtual displacement. The direction of this displacement in the x space is referred to as the gt direction.
518 Generalized Coordinates If we represent a force f as a directed line segment in the space x, the orthogonal projection of the force onto the g- direction is called the physical component of the force in the g- direction and is designated F&. As we pointed out in the preceding section the quantity G/5 the g- generalized component of a force, and the quantity F&, the physical component of the force in the g- direction, are not in general equal. In the Cartesian space x the physical component Fx is identical with the quantity we have been calling/, and as we have seen earlier/ is also identical to the xt generalized component Xi9 that is, ^ = /5 = */ (14) To determine the general relationship between F 9 the physical component in the g- direction of a force, and Gh the g generalized component of the force, we note first that the component in the x. direction of a virtual displacement in the g- direction is ^ = ¾¾ (15) hence the component in the x- direction of n(g,), the unit vector in the g, direction, is V&)= r-,, .- 21./2 (16) If we take the scalar product of the unit vector n( g) with the force vector f we obtain the physical component in the g direction of the force f, that is, Comparing Eqs. (17) and (13) we obtain [2/3ya&)2] It follows that Gh the g component of the force, and F&, the physical component of the force in the g direction, are not equal but are related as shown above. THE ORDINARY COMPONENTS OF A FORCE There are still other quantities, besides the generalized components and the physical components, that are called the components of a force. The most common are what we shall call the ordinary components of a force. The ordinary components in the g directions of a force are defined as the set of
g2 direction Examples 519 ■ g, direction Fig.l vectors in the g directions that when added vectorially give the force vector. If we represent the force as a directed line segment in the space x, the ordinary component in the g- direction is the vector whose direction is in the g- direction and whose magnitude is equal to the distance in the g. direction between the pair of planes that are parallel to the plane of the remaining coordinate directions g,, g2, . . . , g_,, g + 1, . . . , g3yv, and pass, respectively, through the head and the tail of the line segment. The difference between a physical component and an ordinary component is illustrated in Fig. 1 for a space of two dimensions. The vector OA is the ordinary component in the g- direction of the vector f. The vector OB is the physical component in the g direction of the vector f. If the directions g are all orthogonal to one another the ordinary component in the g direction is identical with the physical component in the g direction. It is possible to determine the analytical relationship between the g generalized component of a force and the ordinary component in the g direction of a force. Since we will have no occasion to use ordinary components, we simply state the result. If we define the matrix [ gy] as the matrix whose elements are *Ek(dxk/dgi)(dXk/dgj) and [giJ] as the inverse of the matrix [gy], the ordinary component in the g direction of a force can be shown to be given by *2j(gn)]/2giJGp where Gj is the gj generalized component of the force (see Appendix 23). note: The terminological distinction between "physical components" and "ordinary components" is not usually made. The term "physical component" is used by some authors in the sense used, and by other authors to mean what we have called "ordinary components." The reader is therefore cautioned to find out how a particular author defines the term "physical component." EXAMPLES Example 1. Consider a system consisting of a single particle acted on by a force F. Choosing the spherical coordinates r, 0, <f> as generalized coordinates, determine the relationship between the generalized components Gn G0, and G^ of the force and the physical components Fn F0, and F^ of the force. Solution. The work done in a virtual displacement 8r is given by SW=F-8r (19)
520 Generalized Coordinates Expressing F and Sr in terms of the unit vectors er, eB, and e^, we obtain F=Frer+F9e9 + Ffy (20) Sr = 8rer + r80e9 + rsm08<j>e^ (21) Substituting Eqs. (20) and (21) in Eq. (19) we obtain 8W = Fr 8r + rF9 80 + r sin 0 F^8<j> (22) From the definition of the generalized components of a force we have 8W= Gr8r + G980+ G^8<j> (23) Comparing Eqs. (22) and (23) we obtain Gr-F, (24) G9 = rF9 (25) G^rsmOFt (26) PROBLEMS 1. A system consists of a particle of mass m suspended from a fixed point O by a string of length a. Let r, 0, and <f> be the spherical coordinates that determine the position of the particle with respect to the point O. Let the polar axis be vertically down. Choosing r, 0, and § as generalized coordinates, determine the generalized components of the gravitational force mg and the force T that the string exerts on the particle. 2. A system consists of two particles of masses m and M, respectively, that are in a gravitational field and interact with each other by means of a repulsive force that is directed along the line joining the two particles and is of magnitude/(r), where r is the distance between the two particles. Let X, Y, Z be the Cartesian coordinates of the center of mass of the system, and r,0,<j> the spherical coordinates that determine the position of the particle of mass m with respect to the particle of mass M. Let the Z axis and the polar axis be vertically up. Choosing X, Y, Z, r, 0, and § as generalized coordinates, determine the generalized components of the force acting on the system.
48 Lagrange's Equations of Motion for Elementary Systems INTRODUCTION If a dynamical system is acted on by a set of known forces, the motion of the system is governed by the equations f{ = m^. The aim of this chapter is to generalize these equations, that is, to find equations of motion that are valid for any set of coordinate g = gu g2, . . . , g3N, not just for the set of coordi- nates x = xi,Xy9 • • • > 3n* ELEMENTARY SYSTEMS We use the term elementary system to mean a dynamical system containing a known number of particles on which known forces are acting. Our interest in this chapter is in elementary systems. A USEFUL LEMMA We will find the following lemma useful in the proof in the next section and in a number of later proofs. Lemma 1. Let g = gl, g2, . . . , g3N and g' = gv, g2., ..., g3N, be two sets of generalized coordinates, and g = gu g2, . . . , g3N and g' = gv, g2.9 ..., g3N, be the corresponding sets of generalized velocities. Then 3&<ft8>0 _ d dgj dt 521 (1)
522 Lagrange's Equations of Motion for Elementary Systems and Hj hj (2) proof: If we are given the transformation it follows that a? (3) (4) Treating gr as a function of g, g, and t and taking the partial derivative with respect to g- we obtain 3gK&g>0 a2g,-(g>0 . 3V(g,Q 3§ r hjh« 8k Hjtt = 2 k d dt 3& ' 3&(ft0 ' [38-(8.01 . hj \ &+i ' 3^(8.0' 9§ (5) which is the first part of the lemma. To prove the second part we again start with Eq. (4). Treating gr as a function of g, g, and t and taking the partial derivative with respect to gj, we immediately obtain Eq. (2), which completes the proof of the lemma. LAGRANGE'S EQUATION OF MOTION FOR ELEMENTARY SYSTEMS We are now in a position to determine the generalized equations of motion for an elementary system. The result is stated in the following theorem. Theorem 1. If a dynamical system consisting of TV particles is moving under the action of a known force, the equations of motion of the system in terms of the generalized coordinates gl5 g2, . . . , g3N are A dt 97Xg,g,0 *T(g,g9t) 3ft = G, (6)
Lagrange's Equation of Motion for Elementary Systems 523 where T is the kinetic energy of the system and the G, are the generalized components of the force. We refer to these equations as Lagrange's equations of motion for elementary systems. proof: From Newton's law But "*-i\& "V*/=// t2>/*/ A dt 3T(x) 3x, Substituting Eq. (8) in Eq. (7) we obtain A dt 3T(i) = /' If we multiply Eq. (9) by 3x,(g, t)/dg- and sum over i we obtain dt 3T(*) dx. 3x,(&0 = 3*,(&0 (7) (8) (9) (10) 3¾ ?■" a® According to Eq. 13 in Chapter 47, the right hand side of Eq. (10) is given by (ii) 2/,^-* Now consider the left hand side of Eq. (10). If we first rearrange the terms using the relation {dA/di)B = d(AB)/dt - A{dB/dt) and then make use of Lemma 1 we obtain dt = ?l 3r(x)13x,(g,Q **i \ dSj 9T(i) 3x,(g,Q 3x, 3g; 3T(x) 3x,(g,g,Q dX; 97X*) J 3x- J/ 3*,(g,Q = ?l 3r(&g,/) 8® 9§ 3r(g,g,/) 9£ -S 3r(x) 3x,(g,g,0 3X: 9§- (12) If we substitute Eqs. (11) and (12) into Eq. (10) we obtain Eq. (6), which is the desired result.
524 Lagrange's Equations of Motion for Elementary Systems note 1: The xt component of the unit vector in the g- direction was shown in Chapter 47 to be proportional to dx^dgj. It follows that the vector whose xt component is dxjdgj is a vector in the g- direction. Multiplying Eq. (9) by dxt/dgj and summing over i as was done in the proof is equivalent to taking the scalar product of Eq. (9) with the vector whose x,- component is dx^/dgj, hence is equivalent to projecting Eq. (9) onto the gj direction and multiplying by a scalar. note 2: If F(g, g, i) is the total time derivative of any function F(g, t) then ^[3^(&ft 0/3&V* - 3A&&0/3& = 0. (See Chapter 54.) Hence if T were replaced by T + F, Eq. (6) would remain true. THE KINETIC ENERGY To use the generalized equations of motion, we must be able to express the kinetic energy 77 as a function of the variables g, g, and t. If we know the transformation equations that relate the set of Cartesian coordinates x to the set of generalized coordinates g, that is, the set of equations x = x(g, /), we can determine 7"(g,g, t) by simply substituting the transformation equations into the expression T = ^^/^V^2- The general result we obtain is contained in the following theorem. Theorem 2. Let x be a set of Cartesian coordinates and g a set of generalized coordinates describing the configuration of a dynamical system. Let x = x(g, t) be the set of transformation equations connecting the set x and the set g. Then the kinetic energy T expressed as a function of g, g, and t is given by T(g,g,t) = a(g,t) + 2«;(&04/+ S £«;*(&')&& (13) J J k where _ 1 fl(&0-o 2"1/ 2 dt 2 «y(*0s2»*-a£ 97- «,*(&') = 2 ?m'^ hT (14) If the transformation equations do not contain the time, it follows immediately that r=2 2«y*(g)M* (15) j k
Examples 525 proof: Substituting the transformation equations into the Cartesian expression for the kinetic energy we obtain •2 T=j^mixf = \ 2w, 9* + 22 dx,(g,t) . 3x,(&9 3*,(&0 9x,(g.O . 3*,(&0 9/ 3? + 2 J ,m, 3^(gv0 3x,(g,f) 3¾ 9/ Sj 1 ^ 3*/(&0 3*/(&0 §•& 7 J k which completes the proof of the theorem. (16) EXAMPLES Example 1. Obtain the equations of motion for a particle of mass m in terms of spherical coordinates r, 0, <J>. Solution. The kinetic energy of the particle in terms of the chosen generalized coordinates is T= ±mr2+ -krnr202+ ±mr2 smftQ2 (17) |mr2+ ^mr202 + }-™»2' The generalized components of the force were obtained in the Example given in Chapter 47 and are Gr = Fr (18) Ge = rFe (19) G+ = rsin0F# (20) where Fr, F9, and F^ are the physical components in the r, 9, and <j> directions. Lagrange's equations are d(dT\ 97\ A (dT\_uj_=r dt\dr ) dr r At ST]-!*!- r dt\d9) 99 ' d_( <yr\ _ 9r = r dt \ 8<j> ) 9<J> * (21) (22) (23)
526 Lagrange's Equations of Motion for Elementary Systems Substituting Eqs. (17)-(20) in Eqs. (21)-(23) we obtain ■j- (mr) - mr92 - mr sin20 <j>2 = Fr (24) j-t (mr29 ) - mr2 sin 9 cos 0 <j>2 = rF9 (25) jr (mr2 sin20 <j>) = r sin 9 F^ (26) or equivalently mr — mrO 2 — mr sin20 <j>2 = Fr (27) rar# + 2mr0 — mr sin 9 cos 0 <j>2 = F9 (28) mr sin 0 <£ + 2rar<j> sin 0 + 2mr0<j> cos 0 = F^ (29) These are the same equations we obtained in Chapter 20, directly from Newton's law. PROBLEMS 1. Obtain the equations of motion of a particle of mass m in terms of cylindrical coordinates p,<j>,z. Express the force in terms of its physical components. 2. A particle of mass m is moving in a plane under the action of a force F. Find the kinetic energy function and the equations of motion of the particle in terms of the following pairs of coordinates: a. parabolic coordinates £ = r + x and tj = r — x; b. oblique coordinates x and tj = ax + by, where b ^ 0; c. elliptic coordinates X and 9 defined by the equations x = (a2 + A2)1/2cos0 and y = X sin 0, where X > 0 and 0 < 9 < 2m. 3. Newton's equations of motion for a particle in a two dimensional conservative force field are mx = — dV/dx and my = — 3F/3j. Derive from these the equations of motion in polar coordinates r, 9, and compare these with the equations that would be obtained using Lagrange's equations of motion. 4. A point P is rotating in a vertical plane in a circle of radius a with an angular speed to = At + B. One end of a light elastic string of natural length a and modulus of elasticity X is attached to the point P, and the other end is attached to a particle of mass m. Determine the equations of motion of the particle in terms of r the length of the string and 9 the angle that the string makes with the vertical. Assume that at time / = 0 the point
Problems 527 P is at its lowest point, and that the motion of the point P and the mass m lie in the same vertical plane. 5. Use Lagrange's equations to determine the equations of motion of a particle with respect to a Cartesian frame S' that is rotating with a constant angular velocity with respect to an inertial frame S. Assume that the z axis is the axis of rotation and that the z and zr axes coincide.
49 Constraint Forces INTRODUCTION If a dynamical system is moving under the action of a force and the force is a known function of the set of generalized coordinates g, the set of generalized velocities g, and the time t, we can use Lagrange's equations for elementary systems to determine g and g as functions of the time. In many situations, however, a force is described not in terms of what it does to a system but in terms of what it prevents the system from doing. For example, if a particle is confined to remain on a particular surface, we are not ordinarily given the value of the force the surface exerts on the particle; rather we are simply told that the surface exerts whatever force is necessary to constrain the particle to remain on it. In such cases the value of the force can usually be found only after we have solved for the motion. In this chapter we develop techniques for handling certain classes of such forces. GIVEN FORCES AND CONSTRAINT FORCES A given force or an active force is a force that is a known function of the configuration and motion of the system on which it is acting and possibly also the time. A constraint force or a. passive force is the force exerted by an agent, called a constraint, that responds to the action of a dynamical system in such a way as to prevent it from assuming certain configurations or making certain displacements. In solving a particular problem the constraint forces cannot ordinarily be handled in the same fashion as a given force, since we usually cannot determine the exact value of the constraint force until after we have solved for the motion of the system. It is therefore sometimes helpful to distinguish between the given forces and the constraint forces in the equations of motion. When we wish to make this distinction we retain the notation G to designate the given forces, but use the notation R to designate the constraint forces. 528
Ideal Constraint Forces 529 GEOMETRIC AND KINEMATIC CONSTRAINT FORCES A geometric constraint force is a constraint force that limits the possible configurations of a system. A kinematic constraint force is a constraint force that limits the possible displacements of a system. Every geometric constraint force will also be a kinematic constraint force, since it is impossible to restrict the possible configurations of a system without at the same time restricting the possible displacements of the system. However it is possible to restrict the displacements without restricting the configurations; therefore a kinematic constraint force is not necessarily a geometric constraint force. Let us consider some examples of geometric and kinematic constraint forces. If a particle is enclosed in a box, the walls of the box act as a constraint, and the force the walls exert in containing the particle is a geometric constraint force. If a particle is constrained to remain on a particular surface, the normal force the surface exerts on the particle is a geometric constraint force. If a ball rolls on a two dimensional flat rough surface without slipping, the normal force the surface exerts on the object is a geometric constraint force because it limits the position of the center of mass to a particular plane. But, in addition to this constraint, the tangential force due to the roughness of the surface is a purely kinematic constraint force that couples the rotational motion of the ball to the translational motion of the center of mass. That is, because of the roughness of the surface, the center of mass of the ball cannot undergo an infinitesimal translation in a particular direction without an accompanying rotation of the ball. This constraint force is purely kinematic because despite the roughness of the surface, it is always possible to roll the ball on the surface in such a way that the final position of the center of mass and the final orientation of the ball are completely arbitrary; therefore the roughness of the surface does not impose a geometric constraint on the ball. The tangential force exerted by a rough surface is not necessarily a purely kinematic constraint. If for example a ball or cylinder rolls on a flat rough surface without slipping, and in addition the motion is constrained to be uniplanar, then in this case all the forces, including the tangential force exerted by the surface, are geometrical constraints. IDEAL CONSTRAINT FORCES If the work done by a constraint force in any virtual displacement that does not violate the constraint condition is zero, the constraint force is said to be an ideal constraint force. Some important ideal constraint forces are listed below. 1. The internal forces that maintain the interparticle distances in a rigid body fixed are ideal constraint forces. To show this consider a rigid dynamical system consisting of N particles. Let r(oi) be the position of
530 Constraint Forces particle i with respect to the origin of coordinates, r(y) the position of particle j with respect to particle /, and t(ij) the force particle i exerts on particle j. The virtual work done by the internal forces in a virtual displacement of the system is sw = 22»(i0 'WJ) (1) ' j According to the law of action and reaction and Postulate V (Chapter 27) m=-Hji) = c(uy(v) (2) where c(ij) is a scalar. If we make use of Eq. (2), then Eq. (1) can be rewritten 8^=SSc((0r(«y)-«r(i0 (3) ' j The condition that the interparticle distances remain fixed can be stated mathematically as follows: r(y) • r(y) = constant (4) If the virtual displacement 8r(ij) is to be consistent with the constraint condition, Eq. (4), it must satisfy the condition r(y) • 8r(ij) = 0 (5) If we substitute Eq. (5) in Eq. (4) we obtain 8W=0 (6) The constraint is thus an ideal constraint. 2. If a body is constrained to remain in contact with a smooth surface, the force exerted by the surface on the object is an ideal constraint force. This follows immediately from the fact that in a virtual displacement that is consistent with the constraint, the point on the body that is in contact with the surface can move only in a direction that is tangent to the surface while the reaction force of the surface on the body is normal to the surface. It should be noted that the constraint force is ideal even if the surface is moving. The fact that the surface is moving does not alter the argument above, since a virtual displacement is made at a fixed time. Although the virtual work done on the constrained object by the moving surface is zero, the actual work done is not generally zero. 3. If a body rolls without slipping on a surface, the force exerted by the surface on the body is an ideal constraint force. This follows from the fact that in a virtual displacement that is consistent with the constraint the point on the body that is in contact with the surface is at rest, and therefore the force exerted by the surface on the object does no work. 4. If two parts of the same object are smoothly hinged together, the resultant of the pair of forces acting at the hinge is an ideal constraint force. This follows from the fact that the respective forces exerted on the two parts by
Holonomic Constraint Forces 531 their mutual interaction are equal in magnitude, opposite in direction, and are applied at a common point; therefore, in a virtual displacement that is consistent with the constraint, the virtual work done by the constraint on the one part is the negative of the virtual work done on the other part. The net virtual work done on the system by the constraint force is therefore zero. HOLONOMIC CONSTRAINT FORCES A holonomic constraint force is an ideal geometric constraint force that restricts the possible configurations of a system to those that satisfy an equation of the form *(&') = ° (7) or equivalently it is an ideal kinematic constraint force that restricts the possible displacements of a system to those that satisfy an equation of the form ?lAi(g,t)dgi+Ao(g,t)dt = 0 (8) i where the quantity S/^/(g>0^& + A0(g,t)dt is an exact differential or can be made exact by multiplying it by some function of g and t. To show the equivalence of the definitions above we note first that if Eq. (7) is true, all displacements must satisfy the condition <fo = 2, 8 d&+ dt dt = 0 (9) Thus the conditions in the second definition are satisfied. Conversely if the quantity 2/^/(g>0^/ + ^o(g>0^ is an exact differential (see Appendix 11), there exists a function <j> such that d<$> = *ZAi(g,t)dSl+A0(%,t)dt (10) i If in addition Eq. (8) is true the function <j> satisfies the equation d<j> = 0 (11) From Eq. (11) it immediately follows that there exists a <J> satisfying Eq. (7). This completes the proof of the equivalence of the two definitions. In the definition above we have assumed that the constraint force limits the configurations to those that satisfy a single equation <j>(g, t) = 0. It is possible of course that a constraint might limit the configurations to those that satisfy a set of equations <$>x{g,t) = 0, <f>2(g,0 = 0, . . . . To handle this case we simply break the force the constraint is exerting into component forces and associate one force with each constraint equation. Thus if there are M constraint conditions, there will be M constraint forces.
532 Constraint Forces Examples of holonomic constraint forces are the internal forces that maintain the interparticle distances in a rigid body fixed, and the force that constrains an object to remain in contact with a smooth surface. A nonholonomic constraint force is a constraint force that is not holonomic. Some authors restrict the meaning of the term "nonholonomic" to a particular class of constraint forces that are not holonomic; we refer to this class as anholonomic. ANHOLONOMIC CONSTRAINT FORCES An anholonomic constraint force is an ideal kinematic constraint force that restricts the possible displacements of a system to those that satisfy an equation of the form IlAi(g,t)dgi+A0(g,t)dt = 0 (12) i where the quantity 2/^/(8» 0¾ + ^o(g>0 is not an exact differential nor can it be made exact by multiplying it by some function of g and t. If an object is rolling on a perfectly rough surface and the motion is not restricted to uniplanar motion, the force that prevents the object from slipping is an anholonomic constraint force. THE GENERALIZED COMPONENTS OF HOLONOMIC AND ANHOLONOMIC CONSTRAINT FORCES Although we cannot generally write down an expression for a particular constraint force acting on a system as a function of g, g, and t before we have solved for the motion of the system, we can sometimes take a step in that direction. In particular, in the case of holonomic and anholonomic constraint forces, we can determine the generalized components of a constraint force to within a common scalar multiplicative function. The following theorem states this fact precisely. Theorem 1. If a dynamical system consisting of TV particles is subjected to a holonomic or an anholonomic constraint force R, which restricts the possible displacements of the system to those that satisfy the equation 3N *ZAidgi+A0dt = 0 (13) where the quantities A0,AU . . . , A3N are known functions of g and t, the generalized components Rl9R2, . . . , R3N of the constraint force R satisfy the equations
The Generalized Components of Holonomic and Anholonomic Constraint Forces 533 or equivalently there exists a quantity X, called a Lagrange multiplier, such that */ = H (15) for all values of i from 1 to 3N. proof 1: Let Sg be a virtual displacement that is consistent with the constraint. From Eq. (13) it follows that 24*&-0 (16) i And since the constraint force is ideal, the virtual work done by the constraint force in a virtual displacement that is consistent with the constraint is zero, that is, 2*/«&=0 (17) i Since the set of 3N 8gt must satisfy both Eqs. (16) and (17), only 3 N - 1 of the 8gi are independent. Let us choose the components Sg2, ..., 8g3N as the independent components. To eliminate Sgx from Eqs. (16) and (17) we multiply Eq. (16) by R} and Eq. (17) by Ax and take the difference of the resulting equations. We obtain '2[RlA1-R1Al]Sgl=0 (18) / = 2 Since the components 8g2, . . . , 8g3N are all independent, the coefficients of each of the 8gi in Eq. (18) must be zero. It follows that T, = a~ i = 2'--->3N (19) The set of Eqs. (19) is equivalent to the set of Eqs. (14). Thus the first part of the theorem has been proved. We next note that if Eq. (14) is true, there exists a quantity X defined by the equation X~T, (20) that satisfies the second part of the theorem. This completes the proof. proof 2: We could have alternatively but equivalently proved the theorem as follows. If we multiply Eq. (16) by an arbitrary quantity X and subtract the result from Eq. (17) we obtain 2W-H)«»=o (21) / Equation (21) must be true for arbitrary values of X and any 37V — 1 of the variables gu g2, . . . , g3N. If we choose X such that R.-XA^O (22)
534 Constraint Forces then Eq. (21) becomes 2(^-^,)^/=° (23) / = 2 This can be true for arbitrary values of the 3 N - 1 variables Sg2, . . . , 8g3N only if R.-XA^O i = 2, . . . , 3N (24) Combining Eqs. (22) and (24) it follows that there exists a X such that Eq. (15) is satisfied. This completes the proof. DEGREES OF FREEDOM Our interest in the chapters that follow is primarily in systems that are acted on by given forces, and holonomic or anholonomic constraint forces. A dynamical system consisting of N particles that is subjected to L anholonomic constraint forces and M holonomic constraint forces is said to have 3N — L — M degrees of motional freedom and 3N — M degrees of configurational freedom. Thus the number of degrees of motional freedom is equal to the number of independent displacements Sg, that are required in addition to the constraint conditions, to uniquely specify a particular displacement Sg. And the number of degrees of configurational freedom is equal to the number of independent generalized coordinates g,- that are required, in addition to the constraint conditions, to uniquely specify the configuration g of the system. Most authors do not distinguish between degrees of motional freedom and degrees of configurational freedom, but simply use the phrase "degrees of freedom" to mean one or the other of these types of degrees of freedom. In reading other texts one should therefore be careful to find out what the author means by "degrees of freedom." If the only constraint forces acting on a system are holonomic constraint forces, as is most often the case, the number of degrees of motional freedom and the number of degrees of configurational freedom are equal, and in this case one can speak simply of degrees of freedom without any danger of ambiguity. EXAMPLES Example 1. A disk of radius a rolls without slipping on a horizontal plane. The plane of the disk remains vertical but is free to rotate about a vertical axis through the center of the disk. Discuss the constraints. Solution. In addition to the internal forces, which maintain the disk rigid and are holonomic constraints, there is a constraint that holds the disk vertical, a constraint that keeps the center of the disk a fixed distance above the plane, and a set of constraints that keep the disk from slipping. To discuss
Examples 535 the latter constraints we choose an inertial frame S, with origin 0 in the horizontal plane, 1 and 2 axes also in the horizontal plane, 3 axis vertically up, and a frame S fixed in the disk, with origin O at the center of the disk, I and 2 axes in the plane of the disk, 3 axis perpendicular to the plane of the disk. We then let e1? e2, e3, ey, e2, and e3 be the unit vectors in the 1,_2, 3, T, 2, and 3 directions, respectively; xl9 x2, and x3 be the coordinates of O with respect to the frame S; 0 be the angle between e3 and e3; <J> the angle between the vector e3 X e3 and the vector ej; and \j/ the angle between the vector ey and the vector e3 X e3. The angles 0, <j>, and t// as chosen are the Euler angles, which were used in the discussion of rigid body motion. The constraint that holds the disk vertical can be represented by the equation 9 = f (25) This is a holonomic constraint. The constraint that keeps the center of the disk a fixed distance above the plane can be represented by the equation x3 = a (26) This is a holonomic constraint. The constraint that keeps the disk from slipping can be obtained by noting that if the disk cannot slip, the velocity of the point P on the disk that is in contact with the surface must be at rest with respect to the point O. If we let \(OP) be the velocity of P with respect to O, \(00) be the velocity of O with respect to O, and \(0P) be the velocity of P with respect to 0, this condition becomes \(0P) = \(00 ) + \(0P) = 0 (27) But \(00 ) = e^j + e2x2 and (see Chapter 6) v(OP) = (e3t//) X (- e3a) = exa\p cos <$> + e2a\p sin <$> Substituting Eqs. (28) and (29) in (27) we obtain i:, + a\j/ cos <j> = 0 x2 + a\p sin <j> = 0 or equivalently dxx + a cos<}> d\fj = 0 dx2 + a sin <f> d\p = 0 (28) (29) (30) (31) (32) (33) The differentials above are not exact differentials. Furthermore since it is always possible to roll the disk on the surface in such a way that jc, , x2, <f>, and *P take on any arbitrary value, the constraints represented by Eqs. (32) and (33) are purely kinematic constraints; hence there are no functions that can by multiplication convert the differential expressions in Eqs. (32) and (33) into
536 Constraint Forces exact differentials. It follows that the constraints represented by Eqs. (32) and (33) are anholonomic constraints. If we enumerate all the constraints acting on the disk, we find that it is a system with four degrees of configurational freedom and two degrees of motional freedom. PROBLEMS 1. Two rigid bodies are moving in space. A rigid massless rod joins by means of smooth ball-and-socket joints a given point of one body to a given point of the other body, and one of the bodies is constrained to roll without slipping on a given fixed surface. How many degrees of motional freedom, and how many degrees of configurational freedom, does the system have? 2. A disk is constrained to roll without slipping along a prescribed curve on a horizontal surface. The plane of the disk is constrained to remain vertical and tangent to the curve on the horizontal surface. How many degrees of motional freedom does the system have? How many degrees of configurational freedom does the system have? Express the rolling constraint in the form of a differential equation, and determine whether the constraint is holonomic or anholonomic. 3. A sphere of radius a is free to roll on a perfectly rough horizontal plane. Discuss the constraints. 4. A sphere of radius a is constrained to roll without slipping on a vertical plane that is free to rotate about a vertical axis in the plane. Discuss the constraints. 5. A sharp-edged wheel of radius a is rolling on a rough horizontal surface. Its instantaneous position is determined by assigning values to: (1) the coordinates x, y of the point of contact of the wheel with the surface, referred to a rectangular coordinate system x,y,z whose x-y plane coincides with the surface, (2) the angle 0 between the axle of the wheel and the z axis, (3) the angle § between the x axis and the intersection of the plane of the wheel and the surface, and (4) the angle \p that the radius toward the instantaneous point of contact of the wheel makes with an arbitrary fixed radius. Show that the rolling constraint requires that 8x = acos<f> S\p and Sy = a sin <j>S\p. Show that these conditions cannot be reduced to equations between the coordinates themselves, hence that the constraint is anholonomic. 6. A particle of mass m is sliding on a smooth wire that is bent in the form of a parabola, the equation of which is x2 = lay. If the value of x at arbitrary time t is x(t), and the value of the x component of the constraint force is Rx(t), what is the value of thej component of the constraint force at time /?
Problems 537 7. A bead of mass m is moving on a smooth wire that is bent in the form of a helix the equations of which in cylindrical coordinates p, <j>, z are z = a<j> and p = b, where a and Z> are constants. How are the p, <j>, and 2 generalized components of the constraint force related? 8. A particle of mass m is constrained to move on the inner surface of a smooth paraboloid of revolution, the equation of which is x1 + y1 = 2az. How are the x, y, and z components of the constraint force related?
50 Lagrange's Equations of Motion for Holonomic Systems INTRODUCTION Lagrange's equations of motion for an elementary system, which we derived in Chapter 48, are not of much practical use when we are dealing with systems containing a large number of particles, since the more particles we have, the more equations we have to solve. However if the system is acted on by a large number of holonomic constraints and has as a result only a few degrees of freedom, as would be the case for example with a rigid body, then the problem can be enormously simplified, as we see in this chapter. HOLONOMIC SYSTEMS A dynamical system that is acted on by forces that are either given forces or holonomic constraint forces is called a holonomic system. DEGREES OF FREEDOM Let us consider a holonomic system in which there are N particles and M holonomic constraint conditions. Let g = gl9 g2, . . . , g3N be a set of generalized coordinates for the system of particles, and let *i(&0 = 0, <f>2(&0 = 0, . . . , <f>M(g,t) = 0 (1) be the M constraint conditions. Of the 37V coordinates g}, g2, . . . , g3N, only 3N — M are independent. The number 37V — M, which we designate by the letter /, is defined as the number of degrees of freedom. Alternatively, we can define the number of degrees of freedom as the minimum number of coordinates required in addition to the constraint conditions to determine the 538
Generalized Components of a Force 539 configuration of the system. Note that in the case of holonomic systems there is no difference between degrees of motional freedom and degrees of configu- rational freedom, and therefore there is no ambiguity involved in using the term "degrees of freedom." GENERALIZED COORDINATES Let us consider further the system discussed in the preceding section. Of the 37V coordinates g = gx, g2, . . . , g3N, only / = 3N - M are independent. Let us designate a particular set of / of the coordinates as q = quq2, . . . , qf and let us assume that the set q together with the constraint conditions uniquely determine the configuration of the system. It then follows that the coordinates gi can be written as functions of the set of coordinates q and the time t, that is, &-&(*') (2) The coordinates qt are completely independent. The coordinates gi are not. The set of Eqs. (2) implicitly contains the constraint conditions and therefore may be considered to define the constraint conditions. Any/coordinates qx,q2, • • • , q/ that together with the constraint conditions uniquely specify the configuration of a holonomic system of f degrees of freedom, are called generalized coordinates for the holonomic system, and their time rates of change quq2> • • • > ?/ are called generalized velocities for the holonomic system. CONFIGURATION SPACE The motion of a dynamical system consisting of N particles can be represented by the motion of a point in a 37V dimensional configuration space g. The motion of a holonomic system having f degrees of freedom can be represented by the motion of a point in an f dimensional subspace q of the total configuration space. We refer to this subspace as a configuration space for the holonomic system. The motion of the representative point in the space g is governed by Lagrange's equations of motion for elementary systems. In this chapter we derive the equations that govern the motion of the representative point in the space q. GENERALIZED COMPONENTS OF A FORCE In Chapter 47 it was pointed out that the generalized component of a force corresponding to a particular generalized coordinate gt depends not only on gi but on the whole set of coordinates g = gx, g2, . . . , g3N in the same sense that a partial derivative depends not only on the quantity varied but also on the
540 Lagrange's Equations of Motion for Holonomic Systems quantities that are held constant. Since the set of generalized coordinates q that are used to define the configuration of a holonomic system do not by themselves constitute a complete set of generalized coordinates for the system of particles, it follows that the generalized components of the force corresponding to the coordinates q are not uniquely determined by our previous definition, which is contained in Theorem 1 in Chapter 47. We can avoid this difficulty by simply restricting the virtual displacements of a holonomic system to those that do not violate the constraint conditions. Such virtual displacements are uniquely determined by the set of displacements 8q= 8qu 8q2, . . . , 8qf. With this restriction we can then modify Theorem 1 in Chapter 47 as follows. Theorem 1. Let us assume that we are given a holonomic system that has / degrees of freedom and is acted on by a force. The virtual work 8W of the force in an arbitrary virtual displacement Sq, which does not violate any of the constraint conditions, can be written in the form w-2a«fc (3) i where the quantities g,, g2, . . . , Qj depend only on the value of the force and the constraint conditions. The quantities g,, g2, . . . , Qj are called the generalized components of the force acting on the holonomic system. The set of such quantities is designated Q, and the force whose generalized components are the set Q is called the force Q acting on the holonomic system. proof: The proof follows the same lines as the proof of Theorem 1 in Chapter 47. The following theorems give us some useful properties of the generalized components g. Theorem 2. If g,, g2, . . . , Qj and gr, Q2, . . . , Qj are two sets of generalized components for a force Q acting on a holonomic system, corresponding to the two sets of generalized coordinates qx,q2, . . . , qj and qv,q2>, . . . , q^, and if the two sets of generalized coordinates are related by the transformation equations then the two sets of generalized components of the force are related by the transformation equations a.^¥«n^...».0 (5) proof: The proof of this theorem follows the same lines as the proof of Theorem 2 in Chapter 47.
Lagrange's Equations of Motion for Holonomic Systems 54* Theorem 3. Consider a holonomic system containing N particles and having f degrees of freedom. Let g = gl9 g2, . . . , g3N be a set of generalized coordinates for the system of particles, and q = qx,q2, . . . , <7/ a set of generalized coordinates for the constrained system, and let the transformation equations between the set q and the set g be given by & = &(*0 («) If Gl5G2, . . . , G3N are the generalized components of a force acting on the system of particles, the generalized components g,, g2, . . . , Qj of the force are given by G-2<r J H (7) proof: The virtual work done in an arbitrary virtual displacement Sg is given by SW=^GjSgj (8) j If the values of the displacements Sg, are restricted to those that are consistent with the constraints, then (9) Substituting Eq. (9) in Eq. (8) we obtain for the work done in a virtual displacement that is consistent with the constraint w=*z H (10) Equation (7) follows immediately from Eq. (10). Theorem 4. The generalized components g,, Q2, . . . , Qf of the constraint forces acting on a holonomic system are zero. proof: The constraint forces acting on a holonomic system are holonomic constraint forces. By definition the work done by such forces in a virtual displacement that is consistent with the constraint is zero. LAGRANGE'S EQUATIONS OF MOTION FOR HOLONOMIC SYSTEMS We are now in a position to determine the equations of motion for a holonomic system. The result is stated in Theorem 5. The following lemma is used in the proof of Theorem 5.
542 Lagrange's Equations of Motion for Holonomic Systems Lemma 1. Let g = gx, g2, • • • , g$N be a set of generalized coordinates for a dynamical system. If the system is a holonomic system and if q = qx, q2, . . . , qj is a set of generalized coordinates for the holonomic system, then dg/(q.<i>0 d_ dt and 3&(q,4,/) 3&(q,0 H H (11) (12) proof: The proof is perfectly analogous to the proof of Lemma 1 in Chapter 48. Theorem 5. Consider a holonomic system having / degrees of freedom, whose configuration is defined by the set of generalized coordinates q = qx, q2, . . . , qj. If the system is acted on by a given force Q, the equations of motion of the system are A dt dT(q,q,t) Hi ZT(q,q,t) H = Qi (13) where 7"(q,q, t) is the kinetic energy of the constrained system. We refer to these equations as Lagrange's equations for holonomic systems. note: We give two proofs of the theorem. Each proof brings out different aspects of the theorem. proof 1: Consider a dynamical system consisting of N particles that is acted upon by a given force G and a set of M holonomic constraint forces R,,R2, . . . , RM. Let g = gl9 g2, . . . , g3N be an arbitrary set of generalized coordinates that uniquely specifies the configuration of the system, and fl — ?i>?2> • • • > ?/ a set °f / — 3N — M generalized coordinates that together with the M constraint conditions uniquely specify the configuration of the system. The equations of motion for the set g are dt 3r(ftg,Q 3r(g,g,Q 9£ -^ +2** (14) where Rkj is the g. component of the constraint force R^. If we multiply Eq. (14) by 9§-(q,0/3?/ anc* sum overy we obtain 3r(ft*.0 2 j A dt 94/ dgj{q,t) _ 3r(g,g,0 dgj(q,t) H Zgj 3ft Vr9^0 dgj(q,t) (15)
Lagrange's Equations of Motion for Holonomic Systems 543 The quantity 2;G,[3g,(q, 0/3ft] is just Qh the q,- generalized component of the given force G, and the quantity Syi^g/q.O/Sft] is just the qt generalized component of the constraint force Rk and is zero by virtue of Theorem 4. Thus Tr9^q10_ 2GJ-!a--Q< (16) (17) Now consider the left hand side of Eq. (15). If we rearrange the first term, using the relation (dA/dt)B = d(AB)/dt — A(dB/dt), and then make use of Lemma 1 we obtain 2 j i dt *T(g,g,t) Hj HjM ar(g,g,o dgj(q,t) H aS 4 dt *T(g,g,t) 3§.(q,0 Hj 3ft 3g7(q>0 3ft -2 3ft 3r(ftg,0 3§(q,0 9§ 3ft A. dt 2 r 3r(g,g,Q 3g;(q,q,Q a* 3v"(q,q,Q 3ft 3ft 3^,4,0 3ft 2 j dT(g,g,t) dgj(q,t) hj 3ft (18) If we substitute Eqs. (16)-(18) in Eq. (15) we obtain Eq. (13), which is the desired result. proof 2: Consider a dynamical system consisting of N particles that is acted on by a given force G and a set of M holonomic constraint forces Ri,R2, ..., RM. Let g = gu g2, . . ., g3N be an arbitrary set of generalized coordinates that uniquely specify the configuration of the system. The equations of motion of the system are d_ dt 37-(8,8-0 3 7X& g, 0 = 6/ + 2¾ 3& * Now let us assume that the constraint equations are given by **(&') = ° k= 1,2,..., M (19) (20)
544 Lagrange's Equations of Motion for Holonomic Systems The constraint force R^ restricts the displacements to those that satisfy the equations <j>k(g,t) = 0. If we take the differential of this equation we see that the constraint force R^ restricts the displacements to those satisfying the equations 2j —^— d&+ —= = ° (21) It follows from Theorem 1 in Chapter 49 that the gt component of the constraint force R^ is given by Rkt - K' 3& Substituting Eq. (22) into Eq. (19) we obtain 3r(ftg.O A dt Hi (22) (23) Now suppose we choose as our generalized coordinates the set g = £l> #2> • • • >g3N = ?1>?2> • • • ><l3N-M><i>l><i>2> ■ ■ ■ > *A/ = 4 * (24) where the set q is any set of 3N — M generalized coordinates that together with the M constraint conditions uniquely determine the configuration of the system. The corresponding set of generalized components of the given force G will be designated: G = G1?G2, . . . , G3N = g„ Q2, • • • , Qsn-m, *i,*2> > ®m = Q,* (25) With this choice for the generalized coordinates the equations of motion become A. dt d dt ■ dT(q,q,<j>,<j>,t) - ' dT(q,q,<j>,j>,t) ' 3*, _ ar(q,q,<j>,<jM) dT{q,q,<$>,$,t) Hi -Q, = ¢.- + \ (26) (27) Equations (26) and (27) provide us with 3N equations in the 3N + M unknowns q{,q2, • • • , ?3/v-m> <h>4>2> • • • > 4>m> ^i>^2> • — >^m- Tne constraint equations ¢, = 0 (28) provide the necessary additional equations. If we substitute Eqs. (28) in Eqs. (26) we obtain d_ dt 9 T(q, q,0,0,Q H 9r(q,q,0,0,Q = fi,- (29) But T(q, q, 0,0, /) = T(q, q, /) is the kinetic energy the system would have if it were constrained to the surfaces <J>j = 0, <j>2 = 0, . . . , <j>M = 0. Furthermore the
Lagrange's Equations for Elementary Systems and Holonomic Systems 545 Q. are defined by the requirement that the virtual work be given by sw-2G-*ft+2***** (30) /' k It follows that 2/2/½ is tne work done in a virtual displacement in which <j>v<j>2, . . ., and <f>M are held constant. But holding 4>,,4>2, • • • , and <j>M constant is equivalent to considering only virtual displacements that are consistent with the constraints. This completes the proof. THE RELATIONSHIP BETWEEN LAGRANGE'S EQUATIONS FOR ELEMENTARY SYSTEMS AND LAGRANGE'S EQUATIONS FOR HOLONOMIC SYSTEMS The motion of a dynamical system consisting of TV particles can be represented by the motion of a point in a 37V dimensional configuration space g. The motion of the representative point in the space g is governed by Lagrange's equations for elementary systems, which we discussed in Chapter 48. If the system is a holonomic system having / degrees of freedom, where/< 37V, the motion of the system can be represented by the motion of a point in the / dimensional subspace q. The motion of the representative point in the sub- space q is governed by Lagrange's equations for holonomic systems. Formally these equations appear to be identical with Lagrange's equations for elementary systems. There are however important differences. In the first place the expression for the kinetic energy appearing in Lagrange's equations for elementary systems assumes the possibility of completely arbitrary motions of the particles in the system, whereas the expression for the kinetic energy appearing in Lagrange's equations for holonomic systems is the kinetic energy associated with the constrained motion of the system. In the second place the generalized component of the force appearing on the right hand side of Lagrange's equations for elementary systems includes contributions from the constraint forces as well as the given forces and is defined in terms of the work done in an arbitrary virtual displacement, while the generalized component of the force appearing on the right hand side of Lagrange's equations for holonomic systems includes contributions from the given forces only and is defined in terms of the work done in a virtual displacement that is consistent with the constraints. Although Lagrange's equations for holonomic systems are applicable only when the constraint forces are holonomic, they represent a tremendous simplification over Lagrange's equations for elementary systems in the cases of systems that consist of large numbers of particles that are subjected to large numbers of holonomic constraints. Consider for example a rigid body consisting of TV particles. The number 37V - M in this case is six, and thus only six equations are required to determine the motion. One might ask which set of equations is more fundamental. On the one hand, one could argue that Lagrange's equations for elementary systems are
546 Lagrange's Equations of Motion for Holonomic Systems more fundamental because there are no restrictions on the nature of the forces, and therefore the equations apply to a wider class of problems. On the other hand one can apparently derive Lagrange's equations for elementary systems from Lagrange's equations for holonomic systems by simply reclassifying all the holonomic constraint forces as given forces. In this case there would be 37V degrees of freedom, and the equations of motion obtained would be identical to the generalized equations of motion. The source of the paradox rests in our failure to note that when constraint forces are involved, Lagrange's equations of motion for elementary systems represent an incomplete set of equations, because the generalized components of the constraint forces are not completely known before solving for the motion and therefore we have too few equations for the number of unknowns. Thus when we speak of Lagrange's equations for elementary systems as being more fundamental we are implicitly including with these equations whatever constraint conditions are necessary to provide us with as many equations as we have unknowns. On the other hand if we start with Lagrange's equations for holonomic systems, we cannot reclassify a holonomic constraint force as a given force unless we also introduce the appropriate constraint equations. EXAMPLES Example 1. A bead moves on a smooth wire that is bent in the form of a vertical circle of radius a. The wire rotates about a fixed vertical diameter with a uniform angular velocity co. Find the equation of motion of the bead. Solution. Let r, 0,<j> be the spherical coordinates of the bead, and let the origin be at the center of the circle and the polar axis vertically down. The kinetic energy in terms of these coordinates is T={mr2 + {mr202 + {mrhm20^2 (31) The bead is not free but is subject to the constraints r = a (32) <j> = co (33) The two constraints are holonomic, hence the system is a holonomic system with one degree of freedom. We choose 0 as our generalized coordinate. Substituting Eqs. (32) and (33) in Eq. (31) we obtain T(0j ) = \ma202 + ±rnaWsin20 (34) The work done in a virtual displacement that is consistent with the constraints is 8W= mgd(acosO)= -rngasm686 (35) Hence Q9 = — mga sin 6 (36)
Problems 547 Lagrange's equation is A. dt dT(0,9) 3770,0) -^r2 = a (37) Substituting Eqs. (34) and (36) in Eq. (37) we obtain ma20 — ma2co2sin 0 cos 0 = — mga sin 0 (38) or equivalently 0 - co2sin 0 cos 0 + -£ sin 0 = 0 (39) The motion of the bead can be determined by solving this equation for 0 as a function of time. PROBLEMS 1. Two rods AB and BC of lengths a and b, respectively, are freely jointed at B and hang in a vertical plane from A, which is fixed. Choosing the angles 0 and <j> that the respective rods make with the vertical, determine the generalized components of a horizontal force of magnitude F acting at C. 2. Discuss the motion of a simple pendulum, taking in turn the following generalized coordinates: the angular displacement, the horizontal displacement, and the vertical displacement. 3. A bead of mass m is free to slide on a smooth helical wire the equation of which in cylindrical coordinates is p = a, z = b<$>. Gravity is acting in the positive z direction. The particle is released from rest at the point p = a, <t> = 0, z = 0. Determine z as a function of the time. 4. A bead of mass m slides along a smooth wire bent into the form of the parabola y = x2. Gravity is acting in the negative y direction. The plane of the wire rotates about the y axis with a constant angular speed w. At time t = 0 the bead is at the point x = a,y = a2, and is not moving in the y direction. Obtain equations with which one could determine x as a function of the time. 5. A particle of mass m is supported by a frictionless horizontal disk that rotates about a vertical axis through its center with a constant angular velocity <o. The particle is connected by a massless string of length a to a point located a distance b from the center of the disk. Show that the motion of the particle with respect to the disk is similar to that of a simple pendulum, and find the frequency of small oscillations of the system. 6. Two particles of masses m and M, respectively, are connected by a massless rod of length a. The particle of mass m is constrained to move
548 Lagrange's Equations of Motion for Holonomic Systems on the surface of a sphere of radius b, and the particle of mass M is constrained to move along the vertical diameter of the sphere. Choose an appropriate set of generalized coordinates and obtain the equations of motion. 7. A particle is constrained to move on a smooth surface in the form of the anchor ring x = p cos \p, y = p sin \f/, z = b sin 0, where p = a + b cos 0 and a > b. Prove that if there are no forces except the reaction of the surface, motion around the outer equatorial circle will be stable, and that if this motion is slightly disturbed the new path will cross the equator at intervals of length ir\b{a + Z>)]1/2. Show also that motion around the inner equatorial circle is unstable, and that if such motion is slightly disturbed the path of the particle will cross the outer equatorial circle at an angle 2arctan(Z?/a)1/2. 8. A uniform rod of mass m and length 2a is held with one end on a smooth floor and the other end against a smooth wall. Rectangular Cartesian axes are chosen so that the wall is the plane x = 0, the floor is the plane z = 0, and the ends of the rod are at the points (2a, — /J, 0) and (0, /J, 2y), respectively. Prove that if the rod is released from rest in this position, the initial downward acceleration of the center of mass is (3g/4)[(a2 + 4yS2)/(«2 + 4,62 + Y2)]- 9. One end of a uniform rod AB of mass m and length 2a is constrained to move on a smooth straight horizontal wire, and the rod is constrained to move in the vertical plane containing the wire. The rod is acted on by gravity and a force of magnitude F acting at B in a direction parallel to the wire. Let x be the distance of A from a fixed point on the wire, and 9 the angle the rod makes with the downward vertical. Determine the equations of motion of the rod in terms of the variables x and 0. 10. A uniform rod OA of mass m and length 2a is pivoted at O. A bead P of mass am slides on the rod and is joined to O by a light elastic string of unstretched length a and spring constant nmg/a. Let x be the distance OP, 0 the angle the rod makes with the downward vertical OZ, and <J> the angle which the plane AOZ makes with a fixed vertical plane. Write down Lagrange's equation of motion for the system in terms of the variables x, 9, and <J>. Show that if a steady motion is possible with 0 = 7r/3, and x = 4a/3, then <j>2 = 3g/2a and n = 6a. 11. The ends A and B of a uniform rod of mass m and length 2a are constrained to remain on a smooth circular wire of radius b, where b > a. The wire is made to rotate with a constant angular speed Q, about a fixed vertical axis passing through the center C of the wire. Using as a generalized coordinate 0, the angle the rod makes with the horizontal, obtain Lagrange's equation of motion.
51 The Determination of Holonomic Constraint Forces INTRODUCTION If we are given a holonomic system, we can use Lagrange's equations for holonomic systems to determine the motion of the system. The big advantage gained by using these equations rather than Lagrange's equations for elementary systems is the elimination of the constraint forces from consideration. Sometimes however we are interested in the value of a particular constraint force. In this chapter we show how one can in this case obtain the desired constraint force. THE DETERMINATION OF A HOLONOMIC CONSTRAINT FORCE If we are given a system in which there are TV particles and M holonomic constraint conditions, and we wish to determine the constraint force associated with the Mth constraint condition but are not interested in the constraint forces associated with the first M-l constraints, we simply treat the first Af-1 constraint forces as we would any set of holonomic constraint forces, but we treat the Mth constraint force as if it were a given force. If we do this and also make use of the relation developed in Chapter 49 between a constraint condition and the corresponding components of the constraint force, we will be able to solve not only for the motion of the system but also for the Mth constraint force. This technique is summarized in the following theorem. Theorem 1. Consider a system consisting of N particles, acted on by a given force, and M holonomic constraint forces. If q = ql9q2, • • • , ^3n-m+i *s a set of generalized coordinates that together with the first M — 1 holonomic constraint conditions uniquely determine the configuration of the system, then 549
550 The Determination of Holonomic Constraint Forces the A/th constraint condition can be written in the form 4>(q,t) = 0 and the generalized coordinates qt satisfy the set of equations A. dt 9r(q.q>0 H = a+A d<Kq,Q (1) (2) where A is a function of the time, T(q, q, t) the kinetic energy, and Qi the qt generalized component of the given force. Equations (1) and (2) provide us with 3N - M + 2 equations in the 3N - M + 2 unknowns quq2, . . . , <73A/-a/+i anc* K and can thus be used to find the q({t) and \{t). The generalized components Rt of the constraint force associated with the constraint condition (1) are given by 3<f>(q, t) proof: If we treat the A/th constraint force, which we designate R, as a given force, and make use of Lagrange's equations of motion for a holonomic system, we obtain A. dt 9r(q,q,Q dT(q,q,t) = 0, +#, (4) where jRf. is the qt component of R. But from the results of Chapter 49 the qf component of R is given by Eq. (3). This completes the proof of the theorem. EXAMPLES Example 1. A bead of mass m is free to slide on a smooth helical wire, the equations of which in cylindrical coordinates are p = a, z— b<j>. Gravity is acting in the positive z direction. The particle is released from rest at the point p = a, <j> = 0, z = 0. Determine as a function of § the z and § physical components of the reaction force the wire exerts on the bead. Solution. The kinetic energy of the bead in terms of the coordinates p, <f>, and z is T=imp2 + \mp24>2 + ± 2Lrap" + \mp"<$>" + \mz" (5) The bead is subject to the holonomic constraints p - a = 0 (6) z - b<f> = 0 (7) The system is a holonomic system with one degree of freedom. If we choose z as the generalized coordinate, the kinetic energy function is T(z,z) = \m a2+b2 (8)
Examples 551 and the bead is acted on by a given force whose z generalized component is Qz = mg Lagrange's equation for the system is A. dt dT(z,z) dz dT(z,z) dz -a Substituting Eqs. (8) and (9) in Eq. (10) we obtain a2 + b2 , m- ■ z = mg Solving Eq. (11) and applying the boundary conditions we obtain ,_ gbV 2( a2 + b2) Combining Eqs. (7) and (12) we obtain ,_ gbt2 2(a2 + b2) (9) (10) (11) (12) (13) This approach tells us nothing about the constraint force. Suppose that instead of proceeding as above we treat the force associated with the constraint z = b<f> as a given force; then the system is subject only to the constraint p = a and is thus a holonomic system with two degrees of freedom. Choosing z and <j> as generalized coordinates, the kinetic energy function is obtained by setting p = a in Eq. (5). Thus T(z,<j>J,<j>) = \mal$L + \mz (14) The bead is now acted on by the gravitational force, the z and <j> components of which are Q2 = mg 0., = 0 (15) (16) and by the force due to the constraint z — Z><j> = 0, the components of which are related as follows: RZ = X (17) (18) where X is an unknown parameter. Lagrange's equations in this case are A. dt A. dt 3T(z,<^,i dz dT(z,<t>,z ,<*>) ,<i>) d<j> dT(z dT(z ,<f>,z dz ,<M ,4») ,<?>) a<> -& + R* (19) (20)
552 The Determination of Holonomic Constraint Forces Substituting Eqs. (14)-(18) into Eqs. (19) and (20) we obtain mi = mg + \ (21) ma2$=-\b (22) These two equations together with the constraint equation z = b4> (23) provide us with three equations in the three unknowns z, <f>, and A. If we solve for z and <J> we obtain the same results given by Eqs. (12) and (13). If we solve for X we obtain X=-^-2 (24) a2+b2 V } Substituting Eq. (24) in Eqs. (17) and (18) and noting that the physical components Fz and F^ of the reaction force are related to the generalized components Rz and R^ by the relations Rz = Fz and R^ = pF^ we obtain PROBLEMS 1. The radius r of a soap bubble is changing with time according to the equation r = at + b, where a and b are constants. A particle of mass m is constrained to remain on the surface. At t = 0 the particle is given a tangential velocity v. Find the path of the particle and the reaction of the surface. 2. A particle is constrained to move in a smooth plane that is rotating with a constant angular velocity to about a horizontal axis in the plane. Find the distance r of the particle from the axis of rotation as a function of the time, and also the reaction of the plane on the particle as a function of time. 3. A particle of mass m is constrained to move along a rough wire bent into the form of a horizontal circle of radius a. The coefficient of kinetic friction is ju, and the particle is initially moving with a speed v0. Find the angular position of the particle as a function of the time, and the reaction of the wire on the particle. Assume gravity is not acting. 4. A particle of mass m is constrained to move along a smooth wire bent into the form of a horizontal circle of radius a and is subject to an air resistance of magnitude mkv2, where k is a constant and v is the speed of
Problems 553 the particle. The particle is initially moving with a speed v0. Find the angular position of the particle as a function of the time, and the resultant reaction of the wire on the particle. Assume gravity is not acting. 5. A smooth wire is bent into the form of a helix the equations of which, in cylindrical coordinates, are z = a<}> and p = b, where a and b are constants. The origin is the center of an attractive force that varies directly as the distance. Determine the equation of motion of a bead that is free to slide on the wire, and find the p, <f>, and z physical components of the reaction force of the wire. 6. A particle slides down a smooth stationary sphere, under the action of gravity, from a position of rest at the top. Use Lagrange's equations to find the reaction of the sphere at any value of 0, where 6 is the angle between the vertical diameter of the sphere and the normal to the sphere that passes through the particle. Find the value of 6 at which the particle falls off. 7. A uniform solid cylindrical drum of mass M and radius a is free to rotate about its axis, which is horizontal. An inextensible cable of negligible mass and length I is wound on the drum, and carries on its free end a mass m. Write down the Lagrangian function in terms of an appropriate generalized coordinate, assuming no slipping of the cable. The cable is initially fully wound up, and the system is released from rest with the mass in the same horizontal plane as the axis of the cylinder. Find the angular velocity of the drum when it is fully unwound. By treating the length I of the cable as a variable, and finding Lagrange's equations for two generalized coordinates, show that the tension in the cable is Mmg/(M + 2m). 8. A particle of mass m is at rest on top of a smooth paraboloid, the equation of which is x2 + y1 — a(a — z), where a is a constant and the z axis is vertically up. The particle is displaced slightly from its equilibrium position and slides down the surface. Choosing cylindrical coordinates p, <j>, and z as generalized coordinates, write down Lagrange's equations of motion. Determine the constraint force. At what height will the particle leave the paraboloid? 9. A solid cylinder A of radius a and mass m is at rest on top of a fixed cylinder B of radius b. The axes of the two cylinders are parallel and horizontal. The cylinder A is displaced from its equilibrium position and rolls without slipping down cylinder B. Use Lagrange's equations of motion to determine the constraint forces, and find the position at which the cylinders separate.
52 Lagrange's Equations of Motion for Anholonomic Systems INTRODUCTION If the forces acting on a dynamical system are given forces or are holonomic constraint forces, we can use Lagrange's equations for elementary systems or Lagrange's equations for holonomic systems to determine the motion. If however some of the forces are anholonomic constraint forces, we cannot directly apply the equations above. The purpose of this chapter is to determine how to handle these situations. ANHOLONOMIC SYSTEMS A dynamical system that is acted on by one or more anholonomic constraint forces together with any other given forces or holonomic constraint forces is called an anholonomic system. LAGRANGE'S EQUATIONS OF MOTION FOR ANHOLONOMIC SYSTEMS The equations of motion for anholonomic systems can be obtained by a straightforward extension of the results we have already obtained. Theorem 1. Consider an anholonomic system consisting of N particles, acted on by a given force, M holonomic constraint forces, and L anholonomic constraint forces. Let q= <7i,<72> • • • > #3n-m ^>e a set °f generalized coordinates that together with the M holonomic constraint conditions uniquely determine the configuration of the system. If the anholonomic constraint forces Rl5R2, . . . , RL restrict the displacements to those satisfying the equa- 554
Determination of Anholonomic Constraint Forces 555 tions *ZAki(q,t)dqi+AkO(q,t)dt = 0 (1) i or equivalently to those satisfying the equations ^Aki(q,t)q^Ako(q,t) = 0 (2) i where k ranges from 1 to L, then the motion of the system can be determined by the 3N - M equations A. dt dT(q,q,t) 3ft 37Yq,q,0 „ H =& + pA (3) together with the L constraint equations (2), where T{q,q,t) is the kinetic energy and Qt the qt generalized component of the given force. We refer to the set of Eqs. (3) together with the set of Eqs. (2) as Lagrange's equations for anholonomic systems. proof: If we treat the anholonomic constraint forces as if they were given forces and make use of Lagrange's equations for holonomic systems, we obtain dt 37^)1 *T(q,q,t)_ -^^\^H~-Qi + Vki (4) where Rki is the q( component of the constraint force R^. But from the results of Chapter 49, the q{ component of the constraint force R^ can be written Rki = M*/ (5) Combining Eqs. (4) and (5) we obtain the set of Eqs. (3). The set of Eqs. (3) is a set of 37V — M equations in the 3N — M + L unknowns qi,q2, • • • > ?3;v-a/> A1? A2, . . . , \L. The constraint equations (2) provide us with the necessary L additional equations. This completes the proof of the theorem. DETERMINATION OF ANHOLONOMIC CONSTRAINT FORCES If we are given an anholonomic system consisting of N particles acted on by a given force, M holonomic constraint forces, and L anholonomic constraint forces R,,R2, . . . , RL for which the constraint conditions are 2/^w^?/ + Akodt, the equations of motion consist of 3N - M + L equations in the 3N - M + L unknowns quq2, • • • , ?3;v-m>^i>^2> ---,^- These equations can be used to solve not only for the qt as functions of time but also for the \k as functions of the time. Knowing the Xk we can immediately write down the ?/ component of the constraint force R^ from the relation Rki = \Aki (6)
556 Lagrange's Equations of Motion for Anholonomic Systems Thus Lagrange's equations for anholonomic systems can be used not only to determine the motion but also to determine the anholonomic constraint forces. EXAMPLES Example 1. A disk of radius a and mass m whose plane is constrained to remain vertical rolls on a perfectly rough horizontal surface, under the action of a force f whose line of action passes through the center of the disk. Obtain equations of motion for the disk. Solution. This is the same system that was discussed in the Example given in Chapter 49, where it was shown that the system has four degrees of configurational freedom. A suitable set of generalized coordinates is the set xl9 x2, <t>, and \f/ defined in the above-mentioned example. In terms of these coordinates the kinetic energy of the system is T = i m (x] + jc|) + I ma\$2 + 2^2) (7) There are two anholonomic constraints acting on the system that restrict the displacements to those satisfying the equations xx + ai//cos<j> = 0 x2 + a\p sin <j> = 0 or equivalently satisfying the equations dxx + acos<j>d\p = 0 dx2 + a sin<J> dxp = 0 The generalized components of the force f are &,=/. QXi = fi 0«=0* = o Applying Lagrange's equations for anholonomic systems we obtain mxx =/i+Ai m*2 =/2 + ^2 \mar§ = 0 \ma2^ = acos<j>\} + asm<f>\2 (8) (9) (10) (11) (12) (13) (14) (15) (16) (17) (18) Equations (15)-(18) together with Eqs. (8) and (9) provide us with six equations in the six unknowns xl9 x2, <J>, \p, A,, and X2.
Problems 557 PROBLEMS 1. The position of a particle of mass m is described by Cartesian coordinates x,y,z. The particle is acted on by a force F and is subjected to an anholonomic constraint that constrains the velocity in such a way that y/x = z. Determine the equations of motion. 2. The position of a particle of mass m is described by Cartesian coordinates x,y,z. The particle is acted on by a force F and is subjected to an anholonomic constraint that constrains the velocity in such a way that x—ty — a, where a is a constant. Determine the equations of motion. 3. A uniform disk of radius a and mass m rolls down a flat inclined surface that makes an angle a with the horizontal. The disk is constrained so that its plane is always perpendicular to the inclined plane, but it may rotate about an axis perpendicular to the surface. Determine the motion of the center of mass of the disk. 4. A disk of mass m and radius a is constrained to roll without slipping on a horizontal plane. The disk is not constrained to remain vertical. Obtain equations of motion. 5. Two thin homogeneous disks Dx and D2, each of mass m and radius a are connected by a massless axle of length 2b. The disks are free to rotate about the axle but are constrained to remain perpendicular to the axle. The system rolls without slipping on a horizontal plane. Let x and y be the coordinates of the midpoint of the axle with respect to a Cartesian coordinate system with x and y axes fixed in the plane. Let 0 be the angle on the x-y plane that the direction of the axle makes with the x axis, <j> be the angle that a line passing from the center of Dx to a point fixed on its circumference makes with the upward vertical, and \p be the angle that a line from the center of D2 to a point fixed on its circumference makes with the upward vertical. Show that the system has four degrees of configura- tional freedom and two degrees of motional freedom. Obtain the equations of motion in terms of the set of generalized coordinates x, y, 0, <j>. 6. A homogeneous sphere of mass M and radius a rolls without slipping on a perfectly rough horizontal plane under the action of a force F whose line of action passes through the center of mass. Determine the equations of motion of the sphere.
53 Generalized Force Functions note: Unless otherwise stated, in this and the following chapters we restrict our attention to holonomic systems. Instead of constantly stressing the fact that the systems are holonomic, we shall for the sake of simplicity of terminology suppress reference to the holonomic nature of the system as much as possible. Thus we shall simply speak of a system having / degrees of freedom, whose configuration is defined by the set of generalized coordinates Q — ?i > ?2> • • • > ?/ and whose motion can be determined by solving Lagrange's equations. INTRODUCTION In many cases the generalized components for a particular force can be derived from a single scalar function. The use of such functions can often simplify the equations of motion and their solutions. In this chapter we consider some general types of functions from which the generalized components of a force can be derived. GENERALIZED POTENTIALS Theorem 1. Consider a dynamical system having / degrees of freedom. Let Q— Q\> Qi> - - - > Qf anc* Q' = Qyy Qi'> • • > Qf be two arbitrary sets of generalized components for a given force, corresponding to the two sets of generalized coordinates q = q\,q2> • • • > <7/ and q' = qv,q2', • • • , <7/- K there exists a quantity £/(q, q, t) such that O = — ^ dt 9E/(q,q,Q 9C(q,q,Q 3ft (1) 558
Generalized Potentials 559 Then the generalized components Qr will be given by ^ dt 3t/(q',qV) 9qr (2) The quantity U is called a generalized potential, and the given force is said to be derivable from a generalized potential. proof: From Theorem 2 in Chapter 50 we have If we substitute Eq. (1) in Eq. (3) we obtain (3) fl-2 i dt -25 3£/(q,q,0 3f/(q,q,0 ] 39y(q',0 3¾ dqf Hj dqt 3¾ 3* If we rearrange the first term on the right hand side of Eq. (4) using the relation (dA/dt)B = d(AB)/dt — A(dB/di), and make use of Lemma 1 in Chapter 48, we obtain &=2 1 dt W(q,q,t) dqjtf.t) H v WM.Q d ^ *" dt H H- 9qf -2 j dU(q,q,t) 3$(q\/) 3<7y 3* _ a y» dU(q,q,t) dqj(q',q',t) 3¾ 3<7,< -2 j dU(q,q,t) dqj(q',q',t) + dU(q,q,t) dqj(q',t) dt 3¾ 3f(q',q',Q dq, dqr dqj W(q',q',t) dqr H' (5) This proves the theorem. note: If F(q,q, t) is the total time derivative of any function F(q, t) then ^[3^(q,q,0/3*]/.<# - 9F(q,q, 0/9¾ = 0. (See Chapter 54.) Hence if *7 were replaced by U + F, Theorem 1 would remain true.
560 Generalized Force Functions The following corollary is an immediate consequence of Theorem 1. Corollary 1. Consider a dynamical system having/ degrees of freedom. Let Q = Q\, Qi, • • • , Qf and Q = Qy, Q2,, . . . , Qf be two arbitrary sets of generalized components for a given force, corresponding to the two sets of generalized coordinates q = qx,q2, . . . , qr and q' = qy,qr, •••>?/• If there exists a quantity V(q, t) such that the generalized components Qr will be given by 9V(qf,t) Qi = -^r (7) The quantity V is called a potential, and the given force is said to be derivable from a potential. The potential V(q, t) is simply a special case of the generalized potential U(q, q, i). We have singled it out in a corollary because it represents the most frequently encountered situation. DISSIPATION FUNCTIONS Theorem 2. Consider a dynamical system having f degrees of freedom. Let Q = 2i> £?2> • • • > £?/ and Q' = Qv QT, . . . , Qf be two arbitrary sets of generalized components for a given force, corresponding to the two sets of generalized coordinates q = quq2, • • • , qj and q' = qvqT, . . . , qj>. If there exists a quantity W(q, q, t) such that a w2 (8> the generalized components Q,, will be given by dW(q',q',t) «■ TCjr2 (9) The quantity W is called a dissipation function. proof: From Theorem 2 in Chapter 50 ^-2¾^ (10)
Examples 561 If we substitute Eq. (8) in Eq. (10) and make use of Lemma 1 in Chapter 48 we obtain = -2 7 H- a* 3^(q,q,Q 3$(q',q',0 7 H a?/' 3*F(q',q',0 (11) 3?/' This proves the theorem. EXAMPLES Example 1. A particle of charge q is moving in an electromagnetic field defined by the vector fields E and B, or equivalently by the scalar potential $ and the vector potential A, where E = — V$ — dA/dt and B = V x A. Show that the generalized components of the electromagnetic force acting on the particle can be derived from the generalized potential U = q$ — q\ • A. Solution. To prove that U is a suitable generalized potential, we must show that it leads to the correct expressions for the generalized components of the force for some set of generalized coordinates. In terms of the Cartesian coordinates xl9x2,x3 we have U(x,x,t) = q<P(x,t) - q'ZxtAfat) (12) Substituting this expression into the equation a = — *■> dt dU(x,x,t) dx dU(x,x,t) dx: (13) we obtain -.2^-,^-,¾^¾ (.4) On the other hand if we calculate the xl9 x2, and x3 components of the force using the Lorentz expression F = qE + qy x B = q{ - V<1> - M J + qy x (V X A) (15)
562 Generalized Force Functions we obtain where eiJk is the Levi-Civita symbol discussed in Appendix 6. Comparing Eqs. (14) and (16) we obtain Qj = Fi (17) Thus for Cartesian coordinates the potential U = q$> — q\ • A leads to the correct generalized components. From the results of this chapter it follows that it will lead to the correct generalized components for any set of generalized coordinates. PROBLEMS 1. Obtain the equation of motion for a particle falling vertically under the influence of gravity when frictional forces obtainable from a dissipation function kv2/2 are present. Integrate the equation to obtain the velocity as a function of the time and show that the maximum possible velocity for fall from rest is v = mg/k. 2. A pendulum of length b with a bob of mass m is suspended from a spring of unstretched length a and spring constant k. The point of support of the pendulum is restricted to vertical motion. The bob is immersed in a viscous medium that exerts a force whose direction is opposite to the direction of motion of the bob and whose magnitude is proportional to the speed of the bob. The proportionality constant is a. Obtain the equations of motion of the particle in terms of y, the distance of the point of support of the pendulum below its position when the system is in static equilibrium, and 0, the angle the pendulum makes with the vertical. 3. A particle of mass m is moving under the action of a force that is derivable from a potential V. Let xu x2, and x3 be the coordinates of the particle with respect to an inertial frame S. Let x,y,z be the coordinates of the particle with respect to a frame S' whose origin O and z axis coincide with the origin O and 3 axis, respectively, of the frame S but
Problems 563 which is rotating about the z axis with constant angular speed co. Show that the equations of motion of the particle in the frame S' can be regarded as the equations of motion of a particle in a fixed coordinate system acted upon by the force associated with the potential V and by a force derivable from a generalized potential U. Hence find a velocity dependent potential for the centrifugal and Coriolis forces.
54 Lagrange's Equations of Motion for Lagrangian Systems LAGRANGIAN SYSTEMS In most of the dynamical problems encountered in physics, the constraint forces are holonomic and the given forces are derivable from a generalized potential. We refer to systems in which the constraint forces are holonomic and the given forces are derivable from a generalized potential as Lagrangian systems. Our interest in this chapter is in Lagrangian systems. LAGRANGE'S EQUATIONS OF MOTION FOR LAGRANGIAN SYSTEMS Theorem 1. Consider a Lagrangian system of / degrees of freedom. If we define a quantity L, called the Lagrangian for the system, as follows: L=T-U (1) where T is the kinetic energy and U the generalized potential for the system, then the motion of the system, in terms of an arbitrary set of / generalized coordinates q = quq2, • • • , ft, can be determined by the equations A dt 3L(q,q,f) 3ft ^M=0 (2) 3ft These equations are called Lagrange's equations of motion for Lagrangian systems, or frequently simply Lagrange's equations. 564
The Lagrangian 565 proof: Since a Lagrangian system is a holonomic system, the motion is governed by Lagrange's equations for holonomic systems, A dt ^-» From the definition of the generalized potential U, O = — ^ dt 3£/(q,q,Q 9q, at/(q,q,/) -isr- (4) If we combine Eqs. (1), (3), and (4) we obtain Eq. (2), which completes the proof. THE LAGRANGIAN If a particular dynamical system is acted on by a set of holonomic constraint forces and a given force that is derivable from a generalized potential, and if we know the configuration and motion of the system at some instant of time, we can in principle determine the configuration and motion at some later time by solving Lagrange's equations. In setting up Lagrange's equations for a particular problem there are two major steps. The first step is to choose a set of independent generalized coordinates q, that together with the constraint conditions uniquely determine the configuration of the system. Such a set is said to define the configuration of the system. The second step is to determine the Lagrangian L as a function of q, q, and t. Once we know L(q,q, t) we can write down the equations of motion. Thus we say that the Lagrangian function L(q,q, t) defines the behavior of the system under the action of the given force. It is important to note that L must be known as a function of q, q, and t if we wish to determine the behavior of the system. Knowing the Lagrangian as a function of some other set of variables is not necessarily equivalent to knowing the Lagrangian as a function of q, q and t. It is for this reason we say that the Lagrangian function L(q,q, t) defines the behavior of the system, rather than saying that the Lagrangian L defines the behavior of the system. Instead of defining the Lagrangian L as the difference between the kinetic energy T and the generalized potential U, it is possible to generalize the definition by simply defining the Lagrangian as any quantity that when expressed as a function of q, q, and t and substituted into Lagrange's equation that is, Eq. (2)—leads to the correct equations of motion. In this spirit it is
566 Lagrange's Equations of Motion for Lagrangian Systems interesting and sometimes useful to note that one can add the total time derivative of any function F(q, t) to the Lagrangian without altering the equations of motion. To prove this we note first that and thus . <*F(*0 „ 8f(q,() . ^ 3F(q,l) dt *i da, V ' 3? 3^(¾ q,Q _ 32F(q,Q . 3^(q,Q 3¾ f 3^,3¾ *' 3¾3? 3F(q,q,/) 3F(q,/) A dt 3¾ 3F(q,q,f) 3^~ 3ft 3^0 32F(q,0 = 2J a„ a„ ft+ y 3¾-3ft dt 3F(q,q,f) 3ft 3^(q,q,Q 3ft dtdq; = 0 (5) (6) (7) (8) (9) EXAMPLES Example. A hoop of mass M, moment of inertia Mk2 about an axis through its center, and radius a, rolls without slipping on a horizontal plane. A particle of mass m is suspended from the center of the hoop by a string of length b. The motion of uniplanar. Find the equations of motion. Solution. The system has two degrees of freedom. We choose as generalized coordinates the angle 9 that a particular radius of the hoop makes with the vertical, and the angle <j> that the string makes with the vertical. The kinetic and potential energy functions are T(9,4>J,j>) = \[M(a2 + k2) + ma2]92 + \mb2$2 - mab94>cos<}> (10) V(9,4>)= -mgb cos 4> (11) The Lagrangian function is L(9,4>,9,i>) = \[M(a2 + k2) + ma2]92 + \mb2§2 - mab9j>cos<f> + mgb cos <j> (12)
Lagrange's equations are ns are A. dt A dt 30 ' 3L(0,<M,<i>) " dL(O,<j>,0,i>) do dL(0,<f>,0,i>) Problems 5ff7 (13) (14) Substituting Eq. (12) in Eqs. (13) and (14) we obtain [M(tf2+ k2) + ma2]0 -mab$cos¢ + mabj>2sm<j> = 0 (15) mb2<$> — mab9 cos § + mgb sin <J> = 0 (16) We thus have two equations in the two unknowns 0 and <j>, which can be solved to give us 0{t) and <j>(t). PROBLEMS 1. A particle of mass m is suspended by a massless rod of length b from a support P, which moves uniformly on a vertical circle of radius a with constant angular speed co. Assuming that the motion of the system is restricted to the vertical plane containing the circle, determine the equation of motion of the system. 2. A bead of mass m moves on a smooth wire bent into the form of a horizontal circle. The radius of the circle varies with time according to the equation r = at2 + b. One point P on the circle remains fixed, and the direction of the diameter through P remains fixed. Determine the equation of motion of the bead. 3. A cone of mass M and vertical angle 2a can move freely about its axis, which is vertical. The cone has a fine smooth groove cut along its surface so as to make a constant angle /? with the generating lines of the cone. A particle of mass m is free to slide in the groove. The system is initially at rest, with the particle a distance c from the vertex. Show that if 0 is the angle through which the cone has turned when the particle is a distance r from the vertex, then {Mk2 + mr2sin2a)/(Mk2 + mc2sm2a) = exp(20sinacot /?), where Mk2 is the moment of inertia of the cone about its axis. 4. A thin wire of mass M is bent in the form of the helix x = a cos \p, y = a sin ty, z = a\p tan /?, and is mounted so as to be free to rotate about the z axis, which is horizontal. The length of the wire is such that the center of mass lies on the axis. A small smooth ring of mass m slides on
568 Lagrange's Equations of Motion for Lagrangian Systems the wire. The wire is at rest when the ring is projected horizontally along the wire from the lowest position on it with velocity aco. Show that if co is not too large the plane through the axis and the ring oscillates like a simple pendulum of length a(M + rasin2/?)/(Mcos2/? + msm2/3) and that the wire turns through an angle (mo)T sm2/3 cos /?/(Mcos2/? + msin2/3) in each period T of the pendulum motion. 5. A particle of mass m is attached to a fixed point on a rough horizontal table by a massless spring of unstretched length a and spring constant k. The particle is constrained to move on a straight line passing through the fixed point. The coefficient of friction for the contact between the particle and the surface is jit. Find a Lagrangian that will give the correct equation of motion when used in Lagrange's equation. 6. A particle of mass m can slide freely on a light wire bent into the shape of the curve a3y = x4 which can rotate about the y axis, which is vertically upward. If a steady motion takes place with the particle at x = a, find the angular velocity Q, of the wire. If the wire is constrained to rotate with the steady angular velocity £2, and the particle is slightly disturbed from its steady motion at x = a, show that it will oscillate about this position with a period of oscillation 27r(\la/Sg)^2. Find the period of oscillation if the wire is not constrained but moves freely about the y axis. 7. A system is composed of three particles of equal mass m. Between any two of them there are forces derivable from a potential V = — a exp( — br\ where r is the distance between the two particles and a and b are constants. In addition, each of two of the particles exerts a force on the third that can be obtained from a generalized potential of the form U = — c\ • r, where v is the relative velocity of the interacting particles and c is a constant. Set up the Lagrangian for the system using as coordinates the radius vector R of the center of mass, and the two vectors px = rx — r2, and p2 = r2 — r3. Is the total angular momentum of the system conserved? 8. Show that the equations of motion of a charged particle in an electromagnetic field are unaffected by a gauge transformation. 9. A uniform rod OA of mass m and length la is pivoted at O. The rod makes the angle 0 with the downward vertical OZ and the plane AOZ makes the angle <J> with a fixed vertical plane. A bead P of mass am slides on the rod and is joined to O by a light elastic string of natural length a and modulus of elasticity nmg. Choosing x = OP, 6, and <j> as generalized coordinates, obtain Lagrange's equations of motion and show that if steady motion is possible with 6 = it/3 and x = 4a/3, then <f>2 = 3g/2a and n = 6a.
Problems 569 10. Two uniform rods AB and BC, each of mass m and length 2a, are smoothly joined at B. The rod AB is free to rotate about A, which is fixed; C can slide freely along a horizontal line through A, and the plane of the rods is vertical and is free to rotate about a vertical axis through A. Let 9 be the angle the rod AB makes with the horizontal (where 9 is positive when the rod is below the horizontal), and \f/ the angle the plane of the rods makes with a fixed vertical plane through A. Obtain the equations of motion of the system in terms of the coordinates 9 and \p. If at time t = 0, 9 = it/4, 9 = 0, and \p = Q, show that at any later time 4a92(\ + 3sin20) = 4att2(2 - sec20) - 3g[V2 - 2sin0] and that B initially rises if 22 > 3(2)]/2g/\6a. 11. A solid homogeneous cylinder of mass m and radius a rolls without slipping down the inclined plane face of a wedge of mass M, which in turn rests on a perfectly smooth horizontal floor. The angle between the inclined surface and the horizontal is a, and the entire motion takes place in a plane perpendicular to the floor and containing the normal to the inclined plane. Use Lagrange's equations of motion to find the acceleration of the wedge. 12. A circular cylinder of radius a and mass m rolls without slipping inside a horizontal semicircular groove of radius b cut in a block of mass M. The block is constrained to move without friction in a vertical guide, and is supported by a spring of spring constant k. Taking as coordinates the vertical displacement x of the block and the angular displacement 9 of the center of the cylinder, both measured from the position of static equilibrium, determine by means of Lagrange's equations the equations of motion of the system.
55 Lagrange's Equations of Motion and Tensor Analysis INTRODUCTION In this chapter we show how the techniques of tensor analysis can be used to obtain the equations of motion of a particle in terms of an arbitrary set of coordinates, and the relationship between these equations and Lagrange's equations. The reader who is not familiar with general tensors should read Appendix 23 at this point because the concepts and notation of Appendix 23 are used throughout this chapter. Although for simplicity our analysis is restricted to a single particle, the results can be readily extended to a system of particles. NEWTON'S LAW IN TENSOR FORM If we can express a physical law in terms of a particular set of coordinates, and if we know the law of transformation between this set of coordinates and a second set of coordinates, we can use this transformation to express the given law in terms of the second set of coordinates. If we can express a physical law by an equation, each of whose sides is a tensor of the same kind, the form of the equation will be invariant when we transform from one coordinate system to another; hence it becomes a simple job to change coordinates. It is therefore highly desirable though not necessary to be able to express a physical law in tensor form. How does one go about expressing a law in tensor form? One way is to find an equation written in tensor form that for a particular set of coordinates reduces to a form that is known to be correct in that coordinate system. This is the technique we use to write Newton's law in tensor form. 570
Lagrange's Equations of Motion 571 Newton's law in Cartesian coordinates for a single particle can be written The quantity xl is not a general tensor. However, the quantity dxi/dt is a contravariant vector, since dxl _ s? dxl dxi n. dt f 3jc; dt W or equivalently (using the notation of Appendix 23) UX n V UXJ ,1 s. dT-9JdT (3) which is the law of transformation for a contravariant vector. Although dxl I dt is a tensor, d2xl /dt2 is not a tensor, since the ordinary derivative of a tensor with respect to a scalar is not necessarily a tensor. If, however, we take the absolute derivative of dxl/dt with respect to t we will obtain a tensor. Hence 8 ( dxl\ 8t\ dt ) u X ■ -p/ uXJ uX /a\ ~^ + lJkdTir w is a contravariant vector. Furthermore for Cartesian coordinates the Tl-k vanish and 8(dxi/dt)/8t reduces to d2xl/dt2. It follows that we can write Eq. (1) dt2 J dt dt J v J If we now define the vector /' as the contravariant vector whose components in Cartesian coordinates are the usual components of the force, the right hand side of Eq. (5) is also a contravariant vector. Each side of Eq. (5) is now a tensor of the same kind. Furthermore Eq. (5) gives the correct law of motion for Cartesian coordinates. It follows that Eq. (5) is the correct law of motion for arbitrary coordinates xl. If we multiply Eq. (5) by gin sum over /, and relabel indices, we obtain ™<r d2xj , wV dxj dxk f ,rx m^^+mT^drir=fi (6) which is the covariant form of Eq. (5). LAGRANGE'S EQUATIONS OF MOTION If we define T = \mgJk ^- ^- ^\mgjkxJxk (7)
572 Lagrange's Equations of Motion and Tensor Analysis and note that ^=^¼^ (8) ox 1¾="%/*/ (9) OX = mgijXj + { m [ djgkl + dkgij]xJxk (10) we obtain i( g ) " J~ = mgyxJ + tm[djgu + dkgiJ - digjk]x^ = mgijV + {mTjKixJxk (11) Substituting Eq. (11) in Eq. (6) we obtain d_ dt im-*L=fi (12) \dxl) dxl J V } Equation (12) is an alternate form of the equation of motion for arbitrary coordinates. The equation of motion in this form is simply Lagrange's equation of motion. THE GENERALIZED COMPONENTS OF A FORCE The covariant vector / is the covariant vector whose components in Cartesian coordinates are the ordinary components of the force. From the analysis in the preceding section it follows that the ft are equivalent to what we have been referring to as the generalized components of the force. It follows that the generalized components of a force are simply the components of the covariant vector whose components in Cartesian coordinates are the ordinary components of the force.
SECTION Auxiliary Principles of Lagrangian Mechanics
J
56 Constants of the Motion in the Lagrangian Formulation INTRODUCTION In Chapter 33 we saw that solving the equations of motion for a dynamical system is equivalent to finding the independent constants of motion. Lagrange's equations can frequently be a great help in finding constants of motion for they allow us to determine the equations of motion in terms of different sets of coordinates, and the use of one set of coordinates can frequently reveal a constant of motion that would be quite hidden if viewed from some other set of coordinates. In this chapter we consider the problem of finding constants of the motion from the Lagrangian point of view. We restrict our attention to systems whose behavior is defined by a Lagrangian. CONSTANTS OF THE MOTION Consider a dynamical system having f degrees of freedom. Let q = q\,q2, • • . ,gy be a set of generalized coordinates for the system, and q= q\,q2> • • . , qj the corresponding set of generalized velocities. Any function ^(q, q, t) whose value remains constant during the motion of the system is called a constant of the motion. Theorem 1. There are exactly 2/ independent constants of motion for a dynamical system having / degrees of freedom. proof: The proof of this theorem follows the same lines as the proof of Theorem 1 in Chapter 17. 575
576 Constants of the Motion in the Lagrangian Formulation Theorem 2. If the behavior of a dynamical system having f degrees of freedom is defined by a Lagrangian function L(q, q), which is not an explicit function of the time, there are 2/ - 1 independent constants of the motion that are not explicit functions of the time t. proof: The proof of this theorem follows the same lines as the proof of Theorem 2 in Chapter 17. ENERGY Theorem 3. Consider a dynamical system having f degrees of freedom whose configuration is defined by the set of generalized coordinates q = qu q2, . . . , qj. If the behavior of the system is defined by a Lagrangian function ^(q>q), which is not an explicit function of the time, the quantity is a constant of the motion. If furthermore the generalized potential U is a function only of q, and either the equations of transformation between the set of Cartesian coordinates x and the set of generalized coordinates q are not functions of the time or the kinetic energy T is a function of the set of generalized coordinates q, and a homogeneous quadratic function of the set of generalized velocities q, then H(q,q)=T(q,q)+U(q) (2) and therefore in this case T + U is a constant of the motion. proof: The time rate of change of the quantity H is given by dH _ v d dt 41 dt a£(q,q) Hi ^, 9£(q>q).. ^, 3i(q,q) . = 2I| 3^(q,q) d^(q,q) H (3) But from Lagrange's equations the term in the braces is zero and therefore f =0 (4) It follows that H is a constant of the motion. We have therefore proved the first part of the theorem. To prove the second part we note first, from the
Generalized Momentum 577 definition of the Lagrangian, that L=T-U (5) and if U = C/(q), then BL^q). ^ ar(q,q) . We next note that if x = x(q) then T is a homogeneous quadratic function of the generalized velocities qx,q2, • • • , ?/ (see Theorem 2 in Chapter 48). But if T is a homogeneous quadratic function of the velocities qx,q2, . • • , <7/, then from Euler's theorem, which is proved in Appendix 20, or by substituting T= i2/SA?/?7 into the expression 2/[9T(q>q)/9?i]?/> we obtain v9^q). S-a^-*=2r (7) Combining Eqs. (1), (5), (6), and (7) we obtain Eq. (2). This completes the proof of the theorem. GENERALIZED MOMENTUM Consider a system of / degrees of freedom, whose behavior is defined by the Lagrangian function L(q, q, t). The generalized momentum conjugate to the generalized coordinate q( is defined as Pl = ~W~ (8) Note that if there are no constraints, and we choose the set of Cartesian coordinates x as our generalized coordinates, and the generalized potential is not a function of the velocity, the generalized momentum conjugate to the coordinate x( ispi = m^. In this case the generalized momentum conjugate to the coordinate xi is the same as the x( component of the ordinary linear momentum. However it should be noted carefully that the generalized momentum conjugate to a Cartesian coordinate xi is not in general equal to the x( component of the linear momentum. Consider for example a particle of mass w that is constrained to remain on the surface xx + x2 + x3 = 0 but is otherwise free. If we choose xx and x2 as our generalized coordinates then ^== T= \mx\ + \mx\ + {m( — xx - x2)2 and thus px = 2mxx + mx2. As a second example consider an unconstrained particle of mass m that is moving in the potential U = Cxxxx, where C is a constant. The Lagrangian in this case ls L = ±mx\ + \mx\ + \mx\ — Cxxxx and thus/?! = mxx — Cxx. If the Lagrangian function is not an explicit function of a particular generalized coordinate q{ the coordinate is said to be cyclic or ignorable.
578 Constants of the Motion in the Lagrangian Formulation Theorem 4. The generalized momentum conjugate to a cyclic coordinate is a constant of the motion. proof: Starting with Lagrange's equations A dt 3L(q,q,0 we note that if L(q, q, t) is not a function of a particular coordinate qh then 3L(q,q,0 Combining Eqs. (8)-(10) we obtain = 0 (10) = 0 (11) And thus the generalized momentum/?, conjugate to the cyclic variable qt is a constant of the motion. SYMMETRY AND CONSTANTS OF THE MOTION If the Lagrangian function L(q,q,/) is not a function of a particular coordinate, the behavior of the system does not depend on the value of that coordinate, and the conjugate momentum is a constant of the motion. Frequently we can determine a constant of motion for a system subject to a given force without explicitly writing down the Lagrangian by noting from symmetry arguments that the behavior of the system is unaffected by the change of a particular coordinate. We shall consider two important cases. In the following theorems we use the notation 3//3z, where z = e,Zj + e2z2 + e3z3 is an arbitrary vector, to represent the quantity e! 3//3^1 + e23//3z2 + e33//3z3. Theorem 5. Consider a dynamical system consisting of TV particles. Let m(i) be the mass of particle i. Let r(/) and v(/) be the position and velocity, respectively, of particle i with respect to the origin of coordinates. If the behavior of the system is defined by a Lagrangian function L(r, v, t) = L[r(l), ..., r(N),v(l), ... , \(N),t] and if the Lagrangian function is invariant under an arbitrary translation in a particular direction in space, the component of the quantity 2/d£(r,v,/)/3v(/) in the direction of the translation is a constant of the motion. If furthermore the generalized potential is not a function of the velocities v(0, then ]T/3£(r,v,0/3v(0 = S/w0'M0 = P> where P is the total linear momentum of the system with respect to the origin of coordinates.
Symmetry and Constants of the Motion 579 proof: Consider a virtual displacement in which the system undergoes an infinitesimal uniform translation c. Then 6r(l) = Sr(2)= • • • =8r(N) = e (12) Sv(l) = 8v(2)= • • • =8v(N) = 0 (13) If the Lagrangian is invariant under such a displacement then „ dL(r,v,t) ^ 3L(r,v,/) 3.-(/) BL^VM) = e-2,^^ o 3v(i) 3r(/) But from Lagrange's equations dt 3L(r,v,/) 3v(0 and therefore 3L(r1v10=^ 3r(/) i/ 3L(r,v,f) 3£(r,v,Q 3v(/) = 0 Substituting Eq. (16) in Eq. (14) we obtain A. dt « -^2 This result must be true for arbitrary values of the magnitude of c. It follows that £ A. V or 3L(r,v,f) 3v(0 values 3£(r,v,Q 3v(/) 3L(r,v,0 = 0 (14) (15) (16) (17) -y e f 3v(/) = 0 = 0 (18) (19) And thus the component of '£lidL(r,\,t)/d\(i) along the direction c/e is a constant of the motion. If furthermore the generalized potential U is not a function of the velocities v(z'), then 3L(r,y,f) 3r(r,v,Q t 3v(/) f 3v(/) = 2^(^2-(^(7)-(7)} = 2W(0V(0 = P (20) This completes the proof of the theorem.
580 Constants of the Motion in the Lagrangian Formulation Theorem 6. Consider a dynamical system consisting of N particles. Let m(i) be the mass of particle /. Let r(/) and v(/) be the position and velocity, respectively, of particle i with respect to the origin of coordinates. If the behavior of the system is defined by a Lagrangian function L(r,v, t)~ L[r(l), ..., r(N),v(l), ..., \(N\t] and if the Lagrangian is invariant under an arbitrary rotation in space about a particular axis passing through the origin of coordinates, the component of the quantity 2/r(0 X [3L(r,v,/)/ 3v(/)] in the direction of the axis of rotation is a constant of the motion. If furthermore the generalized potential is not a function of the velocities v(z') then 2>(0 X [3L(r,v,0/3v(0] = 2>(0 X m(i)\(i) = H, where H is the total angular momentum of the system with respect to the origin of coordinates. proof: Consider a virtual displacement in which the system undergoes an infinitesimal rotation c about a particular axis through the origin of coordinates, where c is a vector whose direction is along the axis of rotation and whose magnitude is equal to the angle of rotation. Then (21) (22) 8r(i) = c x r(/) «*(/) = ex v(/)=4[ ex r(/)] dt If the Lagrangian is invariant under such a rotation, then dL(r,\,t) „ dL(r,\,t) = 2 8r(0 3L(r,v,/) av(o , „ dL(r,v,t) j _ ■[«x-(0]+ 2-^4^0] [exr(0]+2 [exr(0] 3v(/) dt 3L(r,v,0 d 3v(0 dt [exr(/)] = 0 (23) This result must be true for arbitrary values of the magnitude of c, hence 3L(r,v,f) A dt € y v ' 3v(/) = 0 (24)
Noether's Theorem 581 It follows that the component of 2/r(0 X [3L(r, v, t)/d\(i)] along the direction €/e is a constant of the motion. The rest of the proof is the same as in the proof of Theorem 5. NOETHER'S THEOREM All the results obtained in the preceding sections could have been obtained by making use of the following theorem. Noether's Theorem. Consider a system whose dynamical state at a given instant of time t can be described by a set of generalized coordinates q= <7p<72> • • • > ?/ and a set °f generalized velocities q = q\,q2> • • • > ?/ and for which there exists a Lagrangian function L(t, q,q) that when substituted into Lagrange's equations of motion determines the dynamical behavior of the system. If it is possible to find a set of functions T=T(t,q,q,e) (25) Q = Q(*,q,q,£) (26) where e is a time independent parameter, and (^),.0= ' (27) (Q)«-o-<i (28) and if when the arguments /, q, and q in the Lagrangian function L(t, q, q) are replaced, respectively, by r(/,q,q,e), Q(/,q,q,€) and Q(/,q,q,<M) where ■ __ dQ dQ/dt Q _Q(/,q,q,q,€) V dT dT/dt f T(/,q,q,q,€) l ] the resulting function satisfies the relation £W^4q>*)>Q(^*q^ (30) where F is the total time derivative of some function F(/,q,q), then the quantity ^ 3L(/,q,q) ^ + S \? to-ifi-F (31) where € = 9e (32) Je = 0
582 Constants of the Motion in the Lagrangian Formulation and 9<2,-(',q.q><0 ■ni = 3e (33) is a constant of the motion. proof: Carrying out the derivative in Eq. (30) we obtain ^ 3L(r,Q,Q) 3&(*,q,q,€) + 2 3ft 3e 3(r,Q,Q) 3<2,(^,q,q,q,€) 3& d€ 3L(r,Q,Q) 3T(;,q,q,£) dT 3e ^ . 3f(/,q,q,q,€) + L(T,Q,Q)—K—e '- = F (34) e = 0 To evaluate the terms in Eq. (34), we first expand T and Qt in powers of €, obtaining ^=(^ = 0 + a-(&)«_(>+ 3r(f,q,q,€) 3ft(',q,q,0 e + £ = 0 € + (35) (36) e = 0 Substituting Eqs. (27), (28), (32), and (33) in Eqs. (35) and (36) we obtain T=t+ &+■■• (37) ft = q, + 1,,6+--- (38) From Eqs. (37) and (38) we obtain 7=1+^+--- (39) ft = q, + 7j,€ + • • • (40) From Eqs. (37)-(40) it follows that (7-),=0=1 (41) (2,),=0=¾ (42) (&)--(tL-* (43)
dT(t,q,q,e) 3e r 3ft(',q,q, df(t, 3ft(' 3ftC 3e qq.q> 3« q,qq> 3e -q.i<i «) 0 *) if) = « = u le=0 = 1 Je = 0 = *?/ Noether's Theorem 583 (44) (45) (46) (47) : = 0 3e _3_ 3« ft(f,q,q,q,e) f(t,q,q,q,e) e = 0 J_ 3g;(^q>q>q>e) _ ft 3r(/,q,q,q,e) r 3€ f2 3« J£ = 0 = U-4,I (48) Using Eqs. (27), (28), (41)-(48) in Eq. (34) and noting also that 3L(r,Q,Q) 3ft 3L(T,Q,Q) dT 3L(/,q,q) H 3L(*,q,q) 37 [L(r,Q,0)]€_o = L(/,q,q) (49) (50) (51) we obtain y ^LJt,q,q) L^3L(?,q,q) 3L(f,q,q) (52) In the remainder of the proof we suppress the argument of functions appearing in a partial derivative, since in all cases the argument would be (/, q, q) and there is very little chance of confusion. We now note dL H L.> dldL-A d ( dL\-c dL~t (53) (54) (55)
584 Constants of the Motion in the Lagrangian Formulation Substituting Eqs. (53)-(55) in Eq. (52) we obtain ?st*+? -2 + AhV dt(dqir\ A*ttq*r Am)9*'nq* dt f v dL ■ 3L - i H 3¾ which can be simplified to 9L H . _d_(dL\ '/ ^ { dqt ) ?H From Lagrange's equations H dt { dqt ) Using Eq. (58) in Eq. (57) we obtain ^ + 2^0?,-^)-^ d_ dt ^ + 2ffu-^)-^ = 0 (56) = 0 (57) (58) (59) It follows that the quantity in square brackets in Eq. (59) is a constant of the motion, which is what we set out to prove. EXAMPLES Example 1. A hoop of mass M, moment of inertia Mk2 about an axis through its center, and radius a, rolls without slipping on a horizontal plane. A particle of mass m is suspended from the center of the hoop by a string of length b. The motion is uniplanar. Find the equations of motion. Solution. The system has two degrees of freedom. If we choose as generalized coordinates the angle 9 that a particular radius of the hoop makes with the vertical, and the angle <j> that the string makes with the vertical, then from Ex. 1 in Chapter 54 L = ±[M(a2+ k2) + ma2}62 + \mb2$2 - mab9<j>cos<j> + mgbcos<f> (60) Since L is not a function of the time, V is a function of <J> only, and T is a homogeneous quadratic function of 9 and <#>, the quantity T + V is a constant of the motion, that is, ±[M(a2+ k2) + ma2]92 + ±rnb2<p2 - rnab9<j>cos<J> - mgbcos<j>= E (61) where E is a constant, and is just the total energy of the system. Since L is not a function of 9, the quantity dL/d9 is a constant of the motion, that is, [M(a2 + k2) + ma2~\9 -rnab$ cos $ = A (62)
Examples 585 where A is a constant. Equations (61) and (62) provide two first order differential equations in the two unknowns 9 and <j>, rather than two second order equations as would be obtained from a direct application of Lagrange's equations. Example 2. Use Noether's theorem to prove Theorem 3. Solution. If L is not an explicit function of the time, it is obvious that if we let T = t + € (63) Q = q (64) F=0 (65) in Noether's theorem, condition (30) is satisfied. For this choice of T and Q € = 1 (66) Vi = 0 (67) hence L - 2 |4 qt = constant (68) i Hi Example 3. Use Noether's theorem to prove Theorem 4. Solution. If L is not an explicit function of a particular coordinate, which for example we let be q{9 then if we let T = t (69) Qx = <7i + € (70) ft = 0 / = 2,3,... (71) F = 0 (72) in Noether's theorem, condition (30) is satisfied. For this choice of T and Q I = 0 (73) Th = 1 (74) tj,. = 0 i = 2,3, . . . (75) hence dL dq = constant (76) Example 4. Use Noether's theorem to prove Theorem 6. Solution. If the Lagrangian L(r, v, t) is invariant under a rotation about some axis, which for convenience we choose to be the z axis, then condition
586 Constants of the Motion in the Lagrangian Formulation (30) in Noether's theorem will be satisfied if we replace r(/), v(/), and t in the Lagrangian by T = t (77) X(i) = x(/)cose -j(/)sine (78) 7(/) = jc(/)sinc + j(/)cose (79) Z(i) = z(i) (80) and let F = 0. For this choice £ = 0 (81) vx(i)=-y(i) (82) i,(9=+*(0 (83) t,z(/) = 0 (84) hence -2 9¾ *0 + ? a^*(0 constant (85) The quantity above is just the component of 2/r(0 X [3L/3v(/)] in the z direction. This is the same result we obtained in Theorem 6. PROBLEMS 1. A particle of mass m is free to move on a smooth horizontal wire. A second particle of mass M is suspended from the first particle by a massless rod of length 2a. The motion is restricted to the vertical plane containing the wire. Let x be the displacement of the particle of mass m from a reference point on the wire, and 0 the angle the rod makes with the vertical. Obtain 0 and x as functions of 0. 2. Two particles of masses m and M, respectively, are connected by an inextensible string passing through a hole in a smooth table so that m rests on the table surface and M hangs suspended. At time t = 0 the particle of mass m is a distance r0 from the hole and is moving with a speed u0, and the particle of mass M is at rest. Find the radial velocity r and angular velocity <#> of the mass m as a function of r, the distance of mass m from the hole. 3. A uniform rod OA of mass m and length 2a is pivoted at O. A bead P of mass am slides on the rod and is joined to O by a light elastic string of natural length a and spring constant nmg/a. Let 0 be the angle that OA makes with the vertical, <j> be the angle that the vertical plane containing OA makes with a fixed vertical plane, and r be the distance OP. Choosing r, 0, and <f> as generalized coordinates, write down the Lagrangian function, and obtain two constants of the motion.
Problems 587 4# A hoop of mass m and radius a rolls without slipping down an inclined plane of mass M that makes an angle a with the horizontal and is free to slide without friction on a horizontal surface. The motion is uniplanar. Choosing as generalized coordinates x9 the horizontal distance of the plane from some reference position, and s, the distance the hoop has traveled down the inclined plane from some reference point on the inclined plane, find four independent constants of the motion.
57 Lagrange's Equations of Motion for Impulsive Forces INTRODUCTION In Chapters 18 and 41 we discussed the effect that an instantaneous impulse produces on the motion of a particle, and on the motion of a rigid body. In this chapter we show how to handle impulsive forces from the Lagrangian point of view. GENERALIZED IMPULSE If a holonomic system, whose configuration is described by the set of generalized coordinates q= q},q2, . . . 9 qj9 is acted on between the time t and t + At by a force, and if Qt is the qt generalized component of the force, the qt generalized component of the impulse applied to the system between time t and t + At is defined as &(/,/ +A0 = j['+A'ft(/)<ft (1) If the duration A/ of the force is so short that the change in the configuration of the system, during the time the force is acting, is negligible, we can assume that the impulse is applied instantaneously. In this case we refer to the impulse as an instantaneous impulse and we designate the #, generalized component of the instantaneous impulse as QM or simply (¾. The qt generalized component of the instantaneous impulse associated with an impulse Qt(t9t + At) is defined mathematically as follows: &(/) = Lft(/,/ + A/) (2) where L = limit as A/-» 0, Qt-» oo and (5,.(/, t + At) is held constant (3)
Lagrange's Equations for Impulsive Forces 589 LAGRANGE'S EQUATIONS FOR IMPULSIVE FORCES The following theorem summarizes the effect that an instantaneous impulse produces on the motion of a holonomic system. Theorem 1. If a holonomic system whose configuration is described by the set of generalized coordinates q= ql9q2, . . . , qj is acted on by an instantaneous impulse whose qi generalized component is (¾ the change in the dynamical state of the system as a result of the impulse is determined by the equations H = 0 (4) A*, = G, (5) where st is defined by proof: Lagrange's equations for a holonomic system are given by d_ldT\_dT = Q (7) A If we integrate Eq. (7) from t to t + A/ and take the limit L given by Eq. (3) we obtain The first term can be written Since the integrand in the second term in Eq. (8) remains finite during the impulse, L[ fI*-0 (10) From the definition of an instantaneous impulse A ft + At A Ljt Qidt=Qi (11) Substituting Eqs. (9)-(11) in Eq. (8) we obtain As,. = ft (12)
590 Lagrange's Equations of Motion for Impulsive Forces The change in the coordinate q{ is given by H = L[qi(t + M)-qi(t)] a rt + At a rt + At. = L\ qfdt (13) Since the integrand qt remains finite during the impulse, the integral vanishes in the limit as A/ -»0. We thus obtain A9/ = 0 This completes the proof of the theorem. (14) EXAMPLES Example 1. Four identical uniform rods each of mass m and length 2 a are freely joined at their ends to form a rhombus A BCD. The rhombus is initially sliding on a smooth horizontal plane with a uniform speed v in the direction AC, and with the angle between BC and AC equal to a constant a. The point C encounters an inelastic wall that is perpendicular to the direction AC. Determine the motion of the system (Fig. 1) immediately after the collision. Solution. We choose as generalized coordinates the distance x of the center of mass O from the wall, and the angle 9 that the side BC makes with the diagonal A C. The kinetic energy is given by Sma202 T = 2mx2 + Thus Sr = sa = dT dx dT 30 = Amx \6ma20 (15) (16) (17) Fig. 1
Examples 591 When the rhombus collides with the wall it is acted on at the point C by a force / in the x direction. The work done by this force in a virtual displacement consistent with the constraints is 8W = f8(x - 2acos0) = f8x + 2afsin080 (18) Hence fix =/ (19) Q9 = 2afsin0 (20) A For an impulsive force f in the x direction 4=/ (21) Qg = 2a/ sin 9 (22) The impulsive equations of motion are *sx = 4 (23) A*e = 4 (24) Substituting Eqs. (16), (17), (21), and (22) into Eqs. (23) and (24) we obtain A(4mi)=/ (25) A(l6n^9_\2aJsin0 (26) or *"-*'=/- (27) Am v J ^_^,= 3/sin« 8ratf v ' where the primes indicate precolHsion quantities and the double primes postcollision quantities. From the given initial conditions x' = - v (29) 0' = 0 (30) And since the point C is fixed immediately after the collision, x" = -2asina0" (31) Substituting Eqs. (29)-(31) into Eqs. (27) and (28) and solving for 9" we obtain Q" = 3d sina ,~~. 2a(l+3sin2a) ^ ' which tells us how the system is moving immediately after the collision.
592 Lagrange's Equations of Motion for Impulsive Forces PROBLEMS 1. A dumbbell consisting of a particle of mass m and a particle of mass M connected by a massless rod of length a receives an instantaneous impulse f in a direction perpendicular to the rod at a point on the rod a distance b from the mass M. Determine the subsequent motion of the dumbbell, assuming that it is initially at rest. 2. Two uniform rods PQ and QR, each of mass m and length 2a, are smoothly hinged at Q. Initially they lie in a straight line on a smooth table. The rod PQ is acted on by an impulse f at P. The magnitude of f is ma and its direction is perpendicular to PQ and horizontal. Use Lagrange's equations to determine the initial velocities of the respective centers of mass and the initial angular velocities of the rods. 3. A framework A BCD consisting of four equal uniform rods AB, BC, CD, and DA, freely jointed at the ends, lies in the form of a square on a smooth horizontal table. An impulse f, applied at the corner A, causes the framework to start moving as a rigid body. Prove that the impulse acts in a direction parallel to DB and that if M is the total mass of the framework, the kinetic energy generated by the impulse is 5/2/4M. 4. A rigid body that is initially at rest is subjected to an instantaneous impulse f at a point P. Determine the subsequent motion of an arbitrary point Q in the rigid body. For what conditions will the velocity of the point Q be the same as would be the velocity of P if the impulse were applied at Q instead of PI 5. Two equal uniform square plates A BCD and EFGH, each of mass m, lie at rest on a smooth horizontal table, and are connected by massless parallel rods DE and CF freely jointed to the plates so that AD EH and BCFG are straight lines. The system is set in motion by an impulse at A of magnitude / and in the direction AB. Prove that the kinetic energy generated is If2/Sm. 6. A uniform circular hoop of mass m and radius a can turn freely about a point on its circumference. When it is at rest, it is struck at right angles to its plane by a blow of impulse / at another point P on its circumference. Show that the angular velocity generated by the blow cannot exceed 3//(72 ma), and find the distance OP for which this greatest angular velocity occurs. 7. A plane nonrigid framework consisting of four equal uniform straight rods AB, BC, CD, and DA, each of unit mass, is in general motion in its own plane. At a certain instant when its form is square, the motion is such that the component velocities of AB, BC, CD, and DA along their lengths are a, b, c, and d, respectively. At this instant a constraint is suddenly applied that maintains the framework in the form of a square. Prove that the resulting loss of kinetic energy is (a — b + c — d)2/6.
PART APPLICATIONS OF LAGRANGIAN MECHANICS
i
SECTION 1 Central Force Motion
J
58 Central Force Motion INTRODUCTION In Chapter 23 we discussed the motion of a particle under the action of a conservative central force. We now reconsider this subject from the Lagrangian point of view, extending our discussion of central force motion to some topics not previously considered. We freely draw on results obtained in Chapter 23. THE EQUATIONS OF MOTION If we use spherical coordinates r, 0, and § to describe the motion of a particle of mass m under the action of a conservative central force for which the potential is V(r), the Lagrangian function is L = {m[r2 + r202 + r2sin20<j>2] - V(r) (1) The Lagrange equation associated with the coordinate 0 is dt\dO ) Substituting Eq. (1) in Eq. (2) we obtain g= - M_+sin0cos0<j>2 (3) From Eq. (3) it follows that if at t = 0, 0 = <ir /2 and 0 = 0, then at t = 0, 0 = 0. But if both 0 and 0 are zero at t = 0, then at time t = €, where e is an infinitesimal increment of time, we will still have 0 = <n/2 and 0 = 0. Proceeding in this fashion to times 2c, 3c, and so forth, we conclude that if at t = 0, 0 = 7r/2, and 0 = 0, then 0 will have the value 7t/2 for all time. Thus if we choose our polar axis to be perpendicular to the plane containing both the force center and the line of motion of the particle at some instant of time, at 597
598 Central Force Motion that instant 9 = tt/2 and 6 = 0; hence by the argument above we have for all times *-? (4) 2 It follows that the motion of a particle under the action of a conservative central force lies in a plane containing the force center. We have chosen that plane to be the plane 6 = m/2. Substituting Eq. (4) into Eq. (1) we obtain L = iw(r2 + r2<j>2)- V(r) (5) The system has now been reduced from a system of three degrees of freedom to one of two degrees of freedom. Since L is not a function of the time, it follows from Theorem 3 in Chapter 56 that the quantity (d~L/dr)f + (8L/3<j>)<j> — L, which in this case is just the total energy, is a constant of the motion. Thus {m(r2 + r2<j>2) + V(r) = E (6) where E is a constant. Since L is not a function of <j>, it follows from Theorem 4 in Chapter 56 that the quantity dL/d<j> is a constant of the motion. Thus mr2$ = h (7) where h is a constant. The quantity mr2<j> is the angular momentum of the particle. Noting that <j> = d<j>/'dt we can rewrite Eq. (7) as follows: dt=^-d^ (8) Solving Eq. (7) for <j>, substituting the result in Eq. (6) and noting that r = dr J'dt, we obtain * = — ui (9) {(2/m)[E- V - (h2/Imr2)])' where the plus sign is used when dr is positive and the minus sign when dr is negative. The set of equations (8) and (9) provide us with a nice starting point for determining whatever information we wish for a particle moving in a central force field. We pursued this subject in some detail in Chapter 23. We briefly review the general results we obtained there, and then consider some further topics in central force motion.
The Radial Motion 599 THE ORBIT The orbit of a particle in a conservative central force field can be determined by first equating Eqs. (8) and (9) to obtain d<t> ^^ (10) r2{2m[E- V- (h2/2mr2)]}1/2 and then integrating to obtain £(»• THE RADIAL MOTION The value of r as a function of the time can in principle be obtained by first integrating Eq. (9) to obtain t{f), and then solving for r{t). However, much information about the radial motion can be obtained without integrating Eq. (9) by noting that Eq. (9) is formally the same equation one would obtain for a particle of mass m that is moving in one dimension r, in the potential U{r)=V{r)+2^ (H) The function U(r) is called the effective potential. Since U depends on the value of h, there is no unique effective potential, but one potential for every value of h. To illustrate some of the information that can be extracted from the effective potential, suppose for a particular value of h, that U has the form shown in Fig. 1. Then, depending on the value of E and the initial conditions, we can immediately determine within what range r varies, and how. If for example the particle has an energy E = [/,, the particle can be only in the region r, < r < r\, or the region r > r'{. If initially the particle is in the region r, < r < r\, it will oscillate back and forth between r, and r\. On the other r Fig. 1
¢00 Central Force Motion hand if it is initially in the region r > r'[ and r < 0, the particle will move toward r'[, reverse the direction of its radial motion at r'[, and then move off to infinity. If the particle has an energy E = U2 the particle is free to move in the region r > r2. If initially r > r2 and r < 0, the particle will move toward r2, reverse the direction of its radial motion at r2, and then move off to infinity. THE ANGULAR MOTION If we know either the orbit r(<|>) or the radial motion r(t) we can find the angular motion <j>(t) by substituting either r(<j>) or r(t) in Eq. (8) to obtain either or rf»- hdt , m[r(t)] and then integrating. Integrating Eq. (12) will give us /(<|>). Integrating Eq. (13) will give us ¢(/). Note that d<j>/dt is inversely proportional to the square of the distance of the particle from the force center. APSIDES An apsis (pi. apsides) or an apse (pi. apses) is a point on an orbit for which r has p. maximum or a minimum value. At an apsis dr/d<j> = 0 and r = 0. The line joining the force center and an apsis is called an apsidal radius. The length of an apsidal radius is called an apsidal distance. The angle between two consecutive apsidal radii is called an apsidal angle. The orbit of a particle is symmetric with respect to a line drawn from the force center to an apsis. Hence if we know the behavior of a particle as r goes from its minimum value (which will be at an apsis) to its maximum value (which will be either at its next apsis if it exists or at infinity if it doesn't), we can use the symmetry property above to determine its behavior at all other times. POWER LAW POTENTIALS Of particular interest are potentials of the form V(r)=±^ /i=±l, ±2, ... (14) where k is some positive number. The force corresponding to such a potential (12) (13)
Power Law Potentials 601 /(,)=+/^-1 (15) The upper sign corresponds to an attractive force and the lower sign to a repulsive force. The equations describing the motion of a particle in a power law potential can be simplified (except for the case n = - 2) if we use the dimensionless variables r, E, and t defined by the relations r = E = t = mk h2 1/(2 + «) 1/(2 + «) k2h2n k2 mnh2- 1/(2 + «) t Substituting Eqs. (14) and (16)-(18) in Eq. (8) and (9) we obtain dt = r2d<j> dt = ±dr i*/2 (16) (17) (18) (19) (20) [2E T (2rn/n) - r~2} To determine the equation of the orbit, we equate Eqs. (19) and (20), divide by r2, choose the reference axes for <f> such that <J> = 0 when r has a minimum or maximum value r0, and integrate from the point r = r0, <j> = 0 to an arbitrary point r, <J>. Doing this we obtain *-r ^—-2 (21) Jr° r2\2ET(2?n/n)-r-2} The plus sign is used when r0 is a minimum, and the minus sign when r0 is a maximum. In general r0 will be zero, infinity, or one of the turning points obtained by solving the equation . = 2r" IE _ ^-'o -r0-2 = 0 (22) There are three cases in which either Eq. (10) or Eq. (21) can be easily integrated, namely, n = 2, n = - 1, and n = -2. For a number of other cases the integral can be obtained in terms of elliptic functions. The case n = — 1, corresponding to an inverse square force, was considered in great detail in Chapter 23. We will simply summarize the results. For a repulsive inverse square force V= kr~\ E is restricted to values greater than or equal to zero, and the orbit is given by r — i/2 r= (1 +2E) cos which is the equation of a hyperbola. *-l] (23)
602 Central Force Motion For an attractive inverse square force V = — kr~\ E is restricted to values greater than or equal to — 1 /2, and the orbit is given by r = |"(l +2£)1/2cos<J>+ ll (24) If E = — \, the orbit is a circle. If — \ < E < 0, the orbit is an ellipse. If E = 0, the orbit is a parabola. If E > 0 the orbit is a hyperbola. EXAMPLES Example 1. Discuss the motion of a particle of mass m that moves in the potential V = kr2/2, where k is positive. Solution. The analysis of this problem is simplified if we use the dimen- sionless variables ■( mk\i/4 ™ I r (25) h2 ) 1/2 ■(£) t (26) N SrTv <28> "^r» <29> The effective potential U is given by C7=±(r2 + r~2) (30) The effective potential £/ is infinite at r = 0, decreases to a minimum value of 1 at r = 1, and increases to infinity as r approaches infinity. It follows that all orbits are bounded and the minimum value that E can have is L_For_a given value of E the apsidal distances are obtained from the equation U = E, which yields the results r0 = [£-(£2-l)1/2]1/2 (31) ,'/2 ?,=[£ +(^-1)7] (32) from which it follows that V, = 1 (33)
Problems #» To obtain the orbit we set n = 2 in Eq. (21) and integrate. Thus Jrn =2 r O r _ -Z dr 1/2 * r2[2£-r2-r"2] £r2- 1 tan 1/2 77 2 (34) (-r4 + 2^r2- 1)" Rearranging Eq. (34), and squaring the result, we obtain \E2 - (E2 - l)cos22<f>]r4 - 2^2 + 1 = 0 (35) Solving for r2, and choosing the solution that makes r = r0 when <J> = 0, we obtain r2 = \E + (E2- l)1/2cos2<j>] = 1\e-(E2- l)1/2lsin2^ + |"^+(f2- l)1/2lcos2^| (36) Using Eqs. (31) and (32) in Eq. (36) we obtain (37) r2 = r0r\ rl sin2<J> + r\ cos2^> Equation (37) is the equation in polar coordinates of an ellipse with center at the origin. The period of motion can be obtained by substituting Eq. (37) in Eq. (19) and integrating from zero to 2it. Doing this we obtain f = 2tt (38) or J/2 T = -Mf) (39) Thus the period is independent of E or h. The results above can be obtained more simply using Cartesian coordinates. The purpose of this example is to illustrate the techniques developed in this chapter. PROBLEMS !• A particle is moving in a circular orbit under the action of an attractive inverse square force. The time of one revolution is T. Show that if the particle were suddenly stopped in its orbit, it would fall into the force center in a time r/25/2. 2. A particle moves in a bounded orbit under an attractive inverse square force. Prove that the time average of the kinetic energy is half the time average of the potential energy.
604 Central Force Motion 3. At the entry of a satellite into a circular orbit at a distance of 300 km from the earth, the direction of its velocity deviates from the intended direction by one degree toward the earth. a. How is the perigee changed? b. How does the perigee change if the actual velocity is 1 m/s less than intended? 4. The first cosmic velocity u, is the velocity of motion on a circular orbit of radius close to the radius of the earth. The second cosmic velocity v2 is the velocity necessary to escape from the surface of the earth to infinity. Find the magnitude of vx and show that v2=&vx. 5. A particle is attracted to a force center by a force that varies inversely as the cube of its distance from the center. Discuss the motion. 6. A particle is acted on by a central force of the form /(r) = - k/r2 + e/r3, where k and e are positive constants. Obtain the equation for the orbit. Discuss the motion for small €. 7. A particle moves in a central force field for which the potential is V = —ae~br/r, where a and b are positive constants. Obtain the effective potential U( r) for a given value of the angular momentum h, and discuss the nature of the motion for different values of E. Calculate the values of E and h that correspond to circular orbits, assuming that Z>r«l. 8. A particle of unit mass moves under the action of a central force of magnitude /(r). Show that when the orbit has two apsidal distances, a and b, the apsidal speeds are in the ratio b/a. Show also that the speed v at a distance r from the force center is given by the relation \(b2 - a2)v2 = a2(af(r)dr+ b2 P' f(r)dr 9. A particle that moves under the influence of a central force describes the curve r — a/(§ + l)2, where a is a constant. a. What is the force law? b. If when <|> = 0 the particle receives an impulse that reduces its radial velocity to zero and doubles its transverse velocity, show that its subsequent path is given by 3a/2r =1+^ cos(V3~<|>/2). 10. A particle of unit mass is subject to a central force for which the potential energy is — a/4r4. The angular momentum of the particle is h. a. Show that if the total energy is zero, the orbit consists of part of a pair of touching circles, each of radius (<z/8/*2)1/2. b. Show that if the total energy is h4/4a, the orbit is part of the curves r = Manh(<J>/V2 ) and r = &coth(<J>/V2 ), where b = (a/h2)l/2.
Problems <»05 11. One end of an elastic string of natural length a is tied to a fixed point on the top of a smooth table, and a particle attached to the other end of the string can move freely on the table. If the particle is slightly disturbed from a circular orbit of radius b, show that the apsidal angle of the new path is w[(b - a)/(4b - 3a)]1/2. 12. A particle of mass m moves in a potential V(r) = —ar~n with n > 2. Using the effective potential discuss the qualitative nature of the motion for different values of the energy. For the bound orbits show that the particle takes a finite length of time to spiral into the force center, passing through a finite number of revolutions. Can you say anything about the subsequent motion?
59 Bertrand's Theorem INTRODUCTION An orbit is said to be a bound orbit if the distance of the particle from the force center either is constant or ranges between two finite values. In this chapter we explore some properties of bound orbits. CIRCULAR MOTION From Fig. 1 in Chapter 58 it is apparent that there are two situations in which it is possible for the orbit to be a circle. If U has a minimum at some point r = r0 and if E = U(r0) and the particle is initially located at r = r0, the particle will move in a circle of radius r0 with a constant speed. If the particle is given a small impulse in the radial direction so that h remains constant but E increases, the motion will deviate only slightly from circular motion. The value of r will oscillate about r0. If on the other hand U has a maximum at some point r = r0 and if E = U(r0) and the particle is initially located at r = r0, then as before the particle will move in a circle of radius r0, but now if the particle is given a small impulse in the radial direction so that h remains constant but E increases, the motion of the particle will deviate radically from circular motion, and the value of r will leave the neighborhood of r0. In the first case the orbit is stable. In the second case it is unstable. For U to have a minimum at r = r0 the first derivative of U with respect to r must be zero at r = r0 and the second derivative must be greater than or equal to zero at r = r0. Thus if the particle is moving with constant speed in a stable circular orbit of radius r0, the following relations will hold: U(r0) = V(r0) + -^ = E (1) 2mrQ U'(r0)=V\rQ)-J^^0 (2) mri U"{r0) = V"(r0) + 1¾ > 0 (3) mr0 606
Bertrand's Theorem «07 where a prime indicates the first derivative of a function of r with respect to r, and a double prime indicates the second derivative of a function of r with respect to r. Equations (1)-(3) are equivalent to the following set of equations: U(r0) = V(r0) + —A^. = E (4) U'(r0) = 0 (5) ^0)=^0) + -—>0 (7) BERTRAND'S THEOREM An orbit is said to be bound if the radial position of the particle either is constant or ranges between two finite values. A bound orbit is said to be closed if after one or more revolutions the particle returns to a position it previously occupied and has the same velocity it previously had at that position. If this occurs after a single revolution the orbit is said to be a simple closed orbit. A bound orbit will be closed if the apsidal angle, the angle between two consecutive apsidal radii, divided by 2it, is a rational number. For certain force laws all bound orbits are closed. For example if a particle is moving under an attractive inverse square force, then as we have seen in the preceding chapter, the bound orbits will be either circles or ellipses. For other force laws some orbits will be closed and some will not. It is possible as we see in the following theorem, which apparently was first derived by J. Bertrand in 1873, to determine the force laws for which all bound orbits are closed. Bertrand's Theorem. If for a certain potential every bound orbit is closed, over that range of radii covered by the bound orbits, the potential is either of the form V = — kr~x or of the form V = kr2, where k is positive; and thus the force is either an inverse square attractive force or a Hooke's law force. proof: Solving Eq. (10) in Chapter 58 for dr/d<f>, squaring, rearranging, and making use of the definition of the effective potential given by Eq. (11) in Chapter 58 we obtain 2m [ji%)1=E-U^ (8) If we define r (9) n«)HU(r)UUu=u(l)=v(l)-b£ (10)
608 Bertrand's Theorem then Eq. (8) can be rewritten t{%)1=E-w^ (11) We use an asterisk to indicate a derivative with respect to u, and a prime to indicate a derivative with respect to r. Thus dW(u) dU(r) . W**(u) = d-^- = | [ -r2U'(r)] ± = 2PU\r) + r4U"(r) (13) If a particle is in a bound orbit, the value of u ranges between a lower limit ux and an upper limit u2. Between ux and u2 there is a point u0 at which W{u) has a minimum value. This point corresponds to the point r0 at which U(r) has a minimum, that is, Wn= — (14) Since W(u0) is the minimum value of W, it follows that W*(u0) = 0 W**(u0) > 0 If we define e and x by the relations E = W(u0) + € u = w0 + X 05) (16) (17) (18) Eq. (11) can be rewritten If a particle is in a bound orbit, then by gradually reducing its energy, hence e, while keeping its angular momentum, hence u0, fixed, we will eventually reduce the orbit to a stable circular orbit. The stable circular orbit corresponds to the case in which x = 0 and e = 0. If we consider orbits close to stable circular orbits, W(u0 + x) can be expanded in powers of x. Thus W(u0 + x) = W(u0) + W*(u0)x + W**(u0) £ + W***(u0) ^ + W****(u0)^+ • • • (20) Substituting Eq. (20) in Eq. (19) and making use of Eq. (15) we obtain 2^{%f= €" ^**("o)f - W~**("o)ff - W~***(K0)ff (21)
Bertrand's Theorem <W If we take the derivative of Eq. (21) with respect to <£ and divide by dx/d<j> we obtain ^ ( —2 ) = " W**(uo)x ~ ^***("o) t£ - W****(u0) |i (22) If we define where 0 = a<j> (23) m h2 1/2 then Eq. (22) can be rewritten d2x _ „ _ aJL d9 where _ j W***(u0) A =2\ W**(u0) 2 ^**("o) I (24) 2 = -x- Ax2- Bx5 - • • • (25) (26) _ j W****(w0) B = V. W**(u0) (21>> For orbits close to circular orbits, the value of x will remain close to zero and we can neglect all but the first term on the right hand side of Eq. (25); hence we have ^ = -x (28) d62 y } If we choose our reference axis for 0 such that 0 = 0 when the particle is at an apsis, the solution to Eq. (28) is x = acos0 (29) If the orbit is a closed orbit there must exist a set of positive integers nx and n2 such that when 0 has changed by the amount 2mnx then <|> has changed by the amount 2<nn2. It follows that for the orbit to be closed, 0/<f> must be a rational number; hence from Eqs. (23) and (24) m 1/2 ^ W**(u0)\ = rational number (30) Vh2 Making use of Eqs. (13) and (5)-(7) in Eq. (30) we obtain r, r"Co) H'o) +3 = a (31) % changing h we can change the value of r0 continuously. But as r0 changes, if the orbits are to remain closed, a must remain a rational number, and since it cannot change continuously, it must remain constant. It follows that Eq.
610 Bertrand's Theorem (31) must be true for arbitrary r, that is, rV'\r) 2 V\f) + 3 = a2 (32) Equation (32) is a differential equation that V(r) must satisfy. Solving for V(r) we obtain V(r) = kr"2-2 (33) The potential is thus restricted to a particular subset of the set of all power law potentials. We can put further restrictions on the potential V by considering orbits in which the departure from circularity is larger than we have considered up till now. If we increase E while keeping h fixed, the orbit will depart more from circularity. For the orbit to remain closed, x(0) must remain periodic over the interval 2tt; hence x(0) in the general case can be represented by a Fourier series of the form 00 x{9)= 2 ancosnO (34) n = 0 The dominant term in this series is the ax cos# term. Let us therefore rewrite Eq. (34) as follows: x(0) = xl(0) + x2(0) (35) where xx(0) = axcos0 (36) 00 x2(0) = a0 + ^ an cos n0 (37) Substituting Eq. (35) in Eq. (25) we obtain d2x, d2x7 . . —T + —f = -*i -x2-A(xl + x2)2- B(xx + x2)3- • • • (38) dO2 dO2 v } K } v If we keep only terms to second order, that is, terms in which the sum of the subscripts is either one or two, we obtain d X\ ci Xi *> //%m —t- + —t = ~xx-x2- Ax2 (39) d92 dO2 x 2 ] V Substituting Eqs. (36) and (37) in Eq. (39) we obtain oo oo - ax cos0 - 2 ann2cosn0= -axcos0 - a0- 2 an cos nO- Aa2 cos20 (40) n=2 n=2 Using the relation cos20 = (1 + cos 29)/2 and rearranging we obtain oo (a0 + ±Aa2) + (-3a2 + ±Aa2)cos20 + 2 (1 - n2)ancosn0= 0 (41)
Bertrand's Theorem 611 The coefficient of each cos nO must separately vanish. Hence a0= - \Aa\ (42) a2 = \Aa\ (43) an = 0 n > 3 (44) The solution to this order is thus x = -±Aa2 + a} cos0 + ±Aa2 cos20 (45) To obtain the next order solution we let x(0) = xx(0) + x2(0) + x3(0) (46) where xl(0) = a^cosO (47) jc2(0) = «0+ a2cos20 (48) 00 x3(0) = 2 ancosn0 (49) Substituting Eq. (46) in Eq. (25) we obtain d2x* d2Xj d2x 1 + —\ + —j = - *i - x2 - x3 - A (x} + x2 + x3)2 d»2 dO2 dO2 \3 -£(.*! + x2+ x3y- • • • (50) If we keep terms only to third order we obtain a Xa a X'y d A-? ** o —i- + —f + —^- = -x,- x7- x2- Ax\-2Axxx2- Bx\ (51) JD2 J02 jq2 12 3 1 12 1 V ) Substituting Eqs. (47)-(49) in Eq. (51) we obtain oo -tfjcosfl - 4a2cos20 - 2 ann2cosn0 oo = — tf i cos 0 — a0— a2 cos 20 — ^ A cos ^ « = 3 -^^0-2,4(^08 0)(00+ a2 cos 20)- ^cos30 (52) Using the relations cos20 = (1 + cos20)/2, cos0cos20 = (cos0 + cos30)/2 and cos30 == (3cos0 + cos 30)/4, and rearranging, we obtain (a0 + \Aa\) + (2,4¾¾ + ^4¾¾ + f Ai?)cos0 + (-3¾ + ^«2)cos20 00 + (-8^ + ^^^2 + ^^08 30+ 2 (1 - n2)ancosn0=O (53)
612 Bertrand's Theorem Equating the coefficient of each cos nO term separately to zero we obtain a0= - \Aa\ A (2a0ai + a^a2) + \Ba\ = 0 a2 = \Aa\ (54) (55) (56) (57) (58) a3 = \[Aaxa2 + \Ba\] an = 0 n>4 The solution to this order is thus x = - \Aa\ + axcos0 + \Aa\o,os2Q + ^ [2,42 + 35]aj>cos30 (59) But in this case we have an added condition, namely, Eq. (55), still to use. Substituting Eqs. (54) and (56) in Eq. (55) we obtain 10,42-95 = 0 Substituting Eqs. (26) and (27) in Eq. (60) we obtain W***(u0) W* >o) W**(u0) But from Eqs. (10) and (33) W(u) = ku2~a W**(u0) = 0 2,,2 + hzu 2m hence W*(u) = (2 - a2)kul-a2 + hJL W**(u) = (2- a2)(\ - a2)ku-al + m W***(u) = (2- a2)(\ - a2)(-a2)ku~l-a2 W****(u) = (2- a2)(\ - a2)(-a2)(-\ - a2)ku From the fact that W*(u0) = 0 we obtain £---(2-«>„-« m hence W**(u0) = (2- a2)( - a2)kuoa2 W***(u0) = (2- a2)(l - a2)(-a2)kuoX~al W****(u0) = (2 - a2)(l - a2)(-a2)(-l - a2)A:«0_2" Substituting Eqs. (68)-(70) in Eq. (61) we obtain (1 - a2)(4- <x2) = 0 (60) (61) (62) (63) (64) (65) (66) (67) (68) (69) (70) (71)
Bertrand's Theorem <>13 The only values of a that satisfy this condition are a = 1 and a = 2. The possible potentials have now been restricted to potentials of the form V- vr'x and V— kr2. If we investigate the orbits associated with these potentials, we find that for the repulsive cases, none of the orbits are bounded, but for the attractive cases, all bound orbits not only are closed orbits but are simple closed orbits. This completes the proof of the theorem.
60 The Differential Scattering Cross Section INTRODUCTION If a particle that is moving with some known velocity encounters a region of space within which there is a force field, the velocity of the particle will undergo a change. If the region within which the force acts is finite, and the particle enters and leaves the region, we say that the particle has been scattered by the force field. If the region within which the force acts is infinite, but the effect of the force on the particle is negligible beyond a certain range, we can still effectively treat the force field as if it were confined to a finite region of space. In what follows we therefore assume that the force field is confined to a finite region of space, even though we may apply the results to cases where it is not, and we refer to the region containing the force field as a scattering region. Frequently, when a particle with a given initial velocity is approaching a scattering region containing a known force field, we are interested only in the probability that the particle will collide with the scattering region and end up with its final velocity in some definite range of velocities, not in the exact details of the encounter. On the other hand it may happen that the force field is unknown and we cannot measure it directly, but we are able to measure the initial and final velocities of particles scattered by the field. In this case we can use the particles to probe the field, and quite often we can obtain in this way a great deal of information about the field. In the analysis of either of these situations, the concept of the scattering cross section is extremely useful. THE TOTAL SCATTERING CROSS SECTION Let us assume that the scattering region in which we are interested is centered at the origin of a Cartesian coordinate frame, and the incident particle is moving with some known speed v in the positive z direction. The force field 614
The Differential Scattering Cross Section 615 ♦y ■*- X Fig. 1 we are considering need not be spherically symmetric, although it generally is. (If it is not, the analysis that follows applies only to a particle approaching the scattering region in the given direction.) If we project the scattering region onto the x-y plane, as represented in Fig. 1, the area of the projection is defined as the total scattering cross section, and is designated as o(v). If the initial value of the x and y coordinates of the incident particle lies within the range of values encompassed by the projected area, the particle will be scattered; otherwise it will miss the scattering region. The total scattering cross section for a given speed v can be determined experimentally by directing a uniform beam of particles of speed v at the scattering region and measuring the fraction of the particles that are scattered. If the cross-sectional area A of the beam is large enough, the scattering region will be entirely encompassed by the beam, and the fraction F of the particles in the beam that are scattered will be equal to o(v)/A, and thus o(v) = FA. THE DIFFERENTIAL SCATTERING CROSS SECTION We now consider the situation introduced in the preceding section in greater detail. To facilitate the discussion we use polar coordinates (s, \p) to specify the position of the initial line of motion of the particle, and spherical coordinates (#,</>) to specify the direction of the final velocity, as shown in Fig. 2. The parameter s is called the impact parameter. For a given force field and a given value of the initial speed v, the final velocity of the particle can in principle be determined if we know the values of s and \p. Thus for each pair of values of s and \p there will be a unique pair of values for 6 and <j>. It follows that if we project the scattering region onto the x-y plane (Fig. 3a), then with each pair of values of s and \f/ in the projected area there is an associated pair of values of 0 and <J> (Fig. 3 b); hence a given region R in the neighborhood of the point (s, ty) within the projected area will correspond to scattering into a definite solid angle £2 in the neighborhood of foe direction (0,<f>). We now define o(v; 0,<j>)sinQd9d<f> to be the magnitude of the projected area that corresponds to scattering into the solid angle
616 The Differential Scattering Cross Section .%! Fig. 2 sin 9 dO d<f>, or equivalently to scattering in which the value of 9 lies between 9 and 9 + d0, and the value of § lies between § and § + d<f>. The quantity o(v; 9, <j>) is called the differential scattering cross section. If we know the differential scattering cross section o(v;9,<j>) we can determine the total scattering cross section o(v) by integrating o(v; 9,<p>)sm9d9d<p over the range of values of 9 and <f> covered by the scattered particles. It should be carefully noted that this range does not necessarily include all values of 9 and <J>. There may for example be an upper limit to the scattering angle 9. The differential scattering cross section o(v\9,<j>) of a given scattering region for a given speed v can be experimentally determined by directing a uniform beam of particles of speed v at the scattering region, and measuring the fraction of the particles that are scattered into each solid angle dQ. If the cross-sectional area A of the beam is large enough, the scattering region will be entirely encompassed by the beam, and the fraction dF of the particles in Fig. 3a Fig. 3b
Determination of the Differential Scattering Cross Section 617 the beam that are scattered into the solid angle dQ, centered about the direction (#,<£) will be equal to o(v;0,<j>)dQ/A, and thus o(v;0,<i>) = AdF/dQ. DETERMINATION OF THE DIFFERENTIAL SCATTERING CROSS SECTION FROM KNOWLEDGE OF THE IMPACT PARAMETER AS A FUNCTION OF THE SCATTERING ANGLE The area of the region R in Fig. 3a is given by I ( sdsdxp (i) and from the definition of the differential scattering cross section is also given by f (o(v;O,4>)sin0d0d<t> (2) where 2 is the solid angle in Fig. 3 b corresponding to the region R in Fig. 3 a. Equating (1) and (2) we obtain f (sdsdxp= f (o(v\0,<j>)sin0d0d<j> (3) If we change the variables of integration in the right hand integral from 0 and <t> to s and $, we obtain ffR,*#-ffRo(v;9,*ytoM%t) dsdxp (4) where J(0,<j>/s,\p) is the Jacobian for the transformation (see Appendix 19). Since Eq. (4) must be true for arbitrary regions R, the integrands on both sides must be equal; hence o(v;0,<i>) = sin# i&) (5) It follows that if we know s and \p as functions of 0 and <f>, for the given value of v, we can determine a(v\ 0,<f>) from Eq. (5). In most cases of interest § = \p and s is a function of 0 only. In such cases J(s,\p/0,<j>) = ds/d0, hence v J sin0 ds d0 (6) If the Jacobian J(0,^>/s,^) changes sign in the region R there will be more than one value of s and $ for a given value of 0 and <J>. In this case it will be necessary to break the region R down into subregions in each of which the Jacobian keeps a constant sign, and then to consider separately the contribution each region makes to the differential scattering cross section.
618 The Differential Scattering Cross Section DETERMINATION OF THE DIFFERENTIAL SCATTERING CROSS SECTION ASSOCIATED WITH A GIVEN SCATTERING POTENTIAL Consider a particle scattered by a potential V(r) that vanishes at infinity. Let a be the angle the final direction of the scattered particle makes with a line drawn from the scattering center to the point of closest approach. It then follows that the scattering angle 9 is given by 0 = |ir - 2a\ (7) When the force is repulsive it > 2a, hence 6 = it — la. When the force is attractive m < 2a, hence 6 = la — it. From Eq. (10) in Chapter 58 hdr ,g. r2{2m[E-V-(h2/2mr2)])]/2 where r0 is the distance of closest approach. The quantity h, the magnitude of the angular momentum, is given by h = mvs (9) where v is the incident speed; and since the potential energy vanishes at infinity, the energy E is given by 2 Using Eqs. (9) and (10) in Eq. (8) we obtain a=f°° Idr (U) Jr0 r2[\-{V/E)-(s2/r2)]X/2 Jr0 E=^- (10) If we define (12) then Eq. (11) can be rewritten «=fZ° T77 (13) Jo {l-[V(z)/E]-z2}]/2 where z0 = s/r0. The value of r0 is the minimum value of r, hence can be obtained by setting dr/d<j> = 0 in Eq. (10) in Chapter 58. Doing this and converting from the variable r to the variable z we find that z0 must satisfy the equation ' "<*> .,8-0 (.4) E Gathering results we obtain for the scattering angle 0 = W-2(Zo ^ — (15) J Jo {l-[F(z)/£]-z2}1/2|
r Examples 619 where z is defined by Eq. (12) and z0 obtained from Eq. (14). Equation (15) gives us 0 as a function of s. Knowing 0(s) we can find s(0) and then use Eq. (6) to find a(v;0,<j>). EXAMPLES Example 1. Determine the differential scattering cross section and the total scattering cross section for the scattering of a particle by a rigid elastic sphere of radius a. Solution. A typical scattering event is illustrated in Fig. 4, from which it is apparent that j=<isin(^=-£) (16) The same result could also have been obtained by substituting the hard sphere potential directly in Eq. (15). If we do this we obtain dz 0 = = ^-2(^ i Jo [l-z2]1/2 = it — 2 sin" '(;) (") which is the same result as Eq. (16). Using Eq. (16) in Eq. (6) we obtain for the differential scattering cross section s I ds ' -2 and for the total scattering cross section d0 a 4 o(v)= \ \ o(v\0,<j>)sin0d0d<j>= ira Jo Jo (18) (19) Example 2. Determine the differential scattering cross section for the scattering of a particle by an inverse square repulsive force of magnitude + k/r2. Fig. 4
620 The Differential Scattering Cross Section Solution. If we substitute V = k/r = kz/s in Eq. (15) we obtain 0 = dz J° [\-(kz/Es)-z2y f -2z-{k/Es) ^° \[(k/Es)2 + 4]1 Letting V(z0) = kz0/s in Eq. (14) we obtain — sin 1/2 Zn = -(k/Es) + \(k/Es)2 + 4] = _ 1/2 Substituting Eq. (21) in Eq. (20) we obtain 0=ir-2 = 2 sin" - sin"'(- 1) + sin- 'J k/Es {[(k/Esf + 4] if k/Es {[(k/Esf + 4]W2\ Solving Eq. (22) for 5 we obtain fccot(0/2) s = mv and thus v ' sin 8 d6 [ 2mv2sin2(0/2) (20) (21) (22) (23) (24) The total scattering cross section for the force above will be infinite because the force field extends to infinity. PROBLEMS 1. A beam of particles with initial velocities in the positive z direction is scattered by a smooth elastic surface of revolution p(z) = bsin(z/a), where 0 < z < ma. Show that the differential scattering cross section is given by a(v\ 0,<f>) = a2sec4(6/2)/4 for 0 < 9 < 2tan-1(Z>/tf) and by a(v\ 9,<f>) = 0 for 2 tan" \b/d) < 9. 2. A beam of particles with initial velocities in the positive z direction is scattered by a smooth elastic surface of revolution p = p(z). Find the surface for which the scattering cross section is the same as the cross section for scattering by a repulsive force of magnitude k/r2.
Problems 621 3# A beam of particles of energy E encounters the spherical potential V(r) = A r<a V(r) = 0 r > a where A is a positive constant. Show that the differential scattering cross section is given by •<••'•♦>-1455375) J{ [l + ^-2nm(S/2)f |+7 for 0 < 0 < 2cos-1H and is zero for 0 > 2 cos-^, where n = [1 - (A/E)]1'2. 4. Find the differential scattering cross section for the scattering of particles by the potential V(r), where V(r) = ar ~x - aR ~x for r < R and V(r) = 0 for r > R. 5. Calculate the cross section for a particle to hit a small sphere of radius a placed at the center of the potential V(r) = —k/rn, where n > 2. 6. A beam of particles with their velocities initially parallel to the z axis is scattered by the fixed ellipsoid (x2/a2) + (y2/b2) + (z2/c2) = 1. Find the differential scattering cross section for the following cases: a. The ellipsoid is smooth and the scattering is elastic. b. The ellipsoid is smooth and the scattering is inelastic. c. The ellipsoid is rough and the scattering is elastic. note: In an inelastic collision with a smooth surface the normal component of the velocity is reduced to zero but the tangential component is unaffected; in an elastic collision with a rough surface, the magnitude of the velocity is unchanged but the direction is randomized.
61 Two Particle Collisions INTRODUCTION If two or more particles are approaching one another, and if the interaction between the particles has a finite range, or at least is negligible beyond a certain range, the motion can be broken down into three periods: an initial period during which the particles are moving toward one another with constant velocities, an intermediate period during which the particles interact, and a final period in which one or more particles, not necessarily the same as the original particles, are moving away with constant velocities. If these conditions are satisfied, we refer to the interaction as a collision. In this chapter we are interested in collisions between two particles from which two particles emerge. The case of only one particle emerging can be treated as a limiting case of the case of two emerging particles. The cases of three or more colliding particles or three or more emerging particles are too tedious for consideration here. DESCRIPTION OF A COLLISION Consider a collision between two particles in which two particles are produced. Let m and M be the respective masses of the colliding particles, and m' and M' the respective masses of the product particles. For simplicity we designate the respective particles as particle m, particle M, particle m\ and particle M', or simply as m, M, m', and M'. To discuss the collision, we define the following precollision velocities: v = velocity of m (1) V = velocity of M (2) u = v — V = velocity of m with respect to M (3) w = mv + MV = velocity of the center of mass c (4) m + M x s = v — w = —T~T7 u — velocity of m with respect to c (5) m+ M 622
Description of a Collision 623 and the following postcollision velocities: v' = velocity of m! (6) V == velocity of M' (7) u' = v' — V = velocity of m' with respect to M' (8) w/ = m v + M V ^ velocity of the center of mass c (9) m + M v ' s' = v' — w' = —-: T77 u' = velocity of m' with respect to c (10) m + M v 7 The quantities above are not all independent. The total mass remains constant, that is, ra + M=ra' + M' (11) and the velocity of the center of mass remains constant, that is, w = w' (12) If we define Ae as the change in the total kinetic energy of the system as a result of the collision, then 2V ' 2 m + M 2V ' 2 m +M (13) Combining Eqs. (5) and (10)-(13) we obtain s' = s* (14) where (s*f=mM:t±Ais2 (15) and 2 m+ M v } The quantity e is just the precollision relative kinetic energy, and since the center of mass kinetic energy is unchanged in the collision, the quantity Ae is either the change in the total kinetic energy of the system or equivalently the change in the relative kinetic energy of the system. In an elastic collision W = m, M' = M, and Ae = 0, and thus for an elastic collision s' = s or equivalently u' = u. We consider a given collision to be completely described by the quantities m, Af, m\ M\ v, V, v', and V. If we assume that the values of m, M, m\ AT, v, V, and also Ae/e are given, these values together with Eqs. (12) and (14) leave Us with only two independent unknowns. It follows that if we want to completely describe a collision, we must in addition to the given parameters specify the values of two independent parameters associated with the postcollision velocities. There are a wide variety of choices for the two parameters. They might for instance be two components of the postcollision velocities, or
624 Two Particle Collisions two angles specifying the direction of one of the postcollision velocities. In the following section we introduce two pairs of parameters that turn out to be particularly useful for this purpose. LABORATORY AND CENTER OF MASS COORDINATES Let us define a set of orthogonal directions 1, 2, and 3 by the three unit vectors — =- u~~ s . - vxV 2~|vxV| (17) (18) e1=e2xe3 (19) If either v or V is zero, we are free to choose e2 in any direction perpendicular to e3. We now define 9 and <|> as the spherical angles that determine the orientation of s' or equivalently u' with respect to a frame whose axes are oriented in the 1, 2, and 3 directions; that is, 9 and <|> are the angles defined by the relation — = — = (ejsin 9 cos <|> + e2sin 9 sin <|> + e3cos 9) (20) Similarly we define 0 and $ as the spherical angles that determine the orientation of v with respect to a frame whose axes are oriented in the 1, 2, and 3 directions; that is, 0 and $ are the angles defined by the relation ^- = (ejsin 0 cos $ + e2sin 0 sin $ + e3cos 0) (21) The coordinates 9 and <|> are convenient coordinates to use when discussing the collision from a theoretical point of view, and are called center of mass or relative coordinates. The coordinates 0 and $ are convenient coordinates to use when the collision is being discussed from an experimental point of view, and are called laboratory coordinates. note: Rather than designating the orientation of v' with respect to the axes el5 e2, and e3, one frequently finds other axes chosen. One could for example argue that the 3 axis in this case should be chosen in the direction of v rather than s. However it turns out that the calculations are greatly simplified if we specify the orientation of both s' and v' with respect to the same axes, and the best set of axes to choose is one in which the 3 axis is in the direction of s. In any case, if two different sets of axes are used, the transformation between the spherical angles with respect to the different sets of axes is not difficult. Suppose for example that 0 and <$ are the spherical angles that determine the orientation of v' with respect to el5 e2, and e3, and 0' and <$' are the spherical angles that determine the orientation of v' with respect to a different set of axes defined by the unit vectors er, e2, and e3- The transformation equations between the two sets of coordinates can then be
Transformation between Laboratory and Center of Mass Coordinates 625 found from the equations sin 0 cos $ sin 0 sin <£ COS0 •e2, sin 0'cos $' sin 0'sin $' COS0' (22) TRANSFORMATION BETWEEN LABORATORY AND CENTER OF MASS COORDINATES The laboratory coordinates 0 and $ and the center of mass coordinates 0 and <p are not independent. In this section we derive the transformation equations between the two pairs of variables. We note first that the components of the vectors v', s', and w satisfy the relation v; = s; + wi (23) Combining (20), (21), and (23) and expressing the result in matrix form we obtain: t/sin 0 cos $ t/sin 0 sin <£ t/cos 0 /sin 0 cos $ + h>! s'sin 0 sin <|> + w2 s'cos 0 + W, (24) Since w,, w2, and w3 are constants, the matrix Eq. (24) together with Eq. (14) provides us with four equations in the six unknowns s\ t/, 0, <|>, 0, and ¢. We can thus solve for 0(0, <|>) and ¢(0, <|>). To do this we first substitute Eq. (14) in Eq. (24) and obtain sin 0 cos <£ = s* \n\ + Yil n2 + y2 n3 + Y3 j where and n =J u_ sin 0 sin <£ COS0 : ejsin 0 cos <|> + e2sin 0 sin <|> + e3cos 0 (25) (26) w _/ ra'M V2 mv + MY Note that the quantity y is a constant whose value depends on the given values of the masses m, M, m\ and AT, the initial velocities v and V, and the quantity Ae/e. Taking the sum of the squares of the three equations in Eq. (25) we obtain v' = s*(\ + y2 + 2n.y)1/2 Substituting Eq. (28) in Eq. (25) we obtain sin 0 cos <I> sin 0 sin <$ COS0 (1 + y2 + 2n-y) 1/2 *i + Yi n2 + y2 n3 + y3 (28) (29)
626 Two Particle Collisions The third equation in Eq. (29) gives us 0 = cos-1 n3 + y3 1/2 (1 + y2 + 2n- y)1 Dividing the second equation by the first equation in Eq. (29) gives us $ = tan -1 "2+ Y2 n\ + Yi (30) (31) Since y is a constant, and n is a known function of 9 and <|>, Eqs. (30) and (31) provide us with the desired transformation. To calculate the inverse transformation we start with the relation /sin 9 cos <i> s'sin 9 sin <|> /cos 9 t/sin0cos$ — h>! t/sin 0 sin $ — w2 v' COS 0 — W~> (32) We then substitute Eq. (14) in Eq. (32) and obtain sin 9 cos <j> sin 9 sin <|> COS0 = v'Nx — s*y} v'N2 — s*y2 v'N3 — s*y3 where Ni V —- = eisin 0 cos $ + e9sin 0 sin $ + e^cos 0 v (33) (34) Taking the sum of the squares of the three equations in Eq. (33) we obtain (s*)2= (©')2+ (j*)V - 2v's*y • N (35) Solving for t/ we obtain o/ = j*|y.N±[(yN)2-y2+ l] 1/2 (36) If y2 is less than unity, the plus sign must be used, since t/ must be positive. If y2 is greater than unity there will be two acceptable values of v'. This is because for y2 greater than unity, the angles 9 and <|> are not single valued functions of 0 and ¢. Substituting Eq. (36) in Eq. (33) we obtain sin 9 cos $ sin 9 sin <|> COS0 {y.N±[(y.N)2-y2+l] )Nl-yl {y.N±[(y.N)2-y2+l]1/2}7V2-y2 {y.N±[(yN)2-y2+l]1/2}7V3-y3 (37)
Laboratory and Center of Mass Differential Cross Sections 627 Solving for 9 and <|> we obtain 6 = cos-1 <j> = tan -l {Y.N±[(Y.N)2-y2+l],/}iV3-V3 {Y.N±[(yN)2-72+l]1/2}jV2-72 {Y-N±[(yN)2-Y2+l]1/2}iV1-Vl (38) (39) Since y is a constant, and N is a known function of 0 and $, Eqs. (38) and (39) provide us with the desired transformation. If y2 < 1, the plus sign is used. If y2 > 1, both signs are used. LABORATORY AND CENTER OF MASS DIFFERENTIAL CROSS SECTIONS We are frequently interested not in the exact values of 0 and <|> or 0 and $ but only in the probability that the particles will collide and end up in some definite range of values for these quantities. The concept of a cross section is useful when this is the case, just as it was in the case of a single particle that was scattered by a stationary force field. We describe the events leading up to a collision between particles m and M from the viewpoint of an observer who before the collision is at rest with respect to particle M. Such an observer sees particle m moving with a constant speed u in the 3 direction toward an interaction region centered on the stationary particle M. This situation is illustrated in Fig. 1, where we have for convenience chosen the particle M as the origin of our reference frame. Fig. 1
628 Two Particle Collisions We describe the events subsequent to the collision both from the point of view of an observer who is at rest with respect to our original inertial frame of reference and from the point of view of an observer who is at rest with respect to the particle M'. The observer who is at rest with respect to our original inertial frame of reference sees m' moving off with a constant velocity v' (Fig. 2a). For convenience we have chosen the intersection of the lines of motion of the two particles as the origin of our reference frame. The observer who is at rest with respect to particle M' sees m! moving away from M' with a constant velocity u' (Fig. 2b). For convenience we have chosen the particle M' as the origin of our reference frame. If we project the interaction region shown in Fig. 1 onto the 1-2 plane, the area of the projection is defined as the total cross section and is designated as q(u) or Q(u). I® m / / * Fig. la / * ♦ 2 Fig. lb
Transformation between Laboratory and Center of Mass Cross Sections 629 note: We consider only cases for which the interaction region is spherically symmetric; hence the cross section does not depend on the direction of approach, but only on the relative speed of approach. The center of mass differential cross section q(u;0,<}>) is defined by the condition that q(u;0,<f>)sin0d0d<j> is the magnitude of the projected area that corresponds to collisions in which the direction of u', the postcollision relative velocity, lies in the solid angle sin 0d0d<j>. The center of mass differential cross section can generally be found theoretically by first determining the differential cross section o(v; 0,<f>) for a collision of particle m, moving with a velocity v, with particle M held rigidly fixed, then replacing the speed v by the relative speed u and the mass m by the reduced mass /x, except that we do not replace m by /x if m occurs in the interparticle force. The laboratory differential cross section £>(w;0, ¢) is defined by the condition that Q(u; 0,$)sin0 d® d$ is the magnitude of the projected area that corresponds to collisions in which the direction of v', the post collision velocity of particle m\ lies in the solid angle sin©d®d$. The laboratory differential cross section is the cross section one would normally measure experimentally. Given the differential cross section q(u;0,<j>) or <2(w;0,$), the total cross section is obtained from the relation q(u)=Q(u)= f Cq(u;0,<f>)sm0d0d<f> = f f Q(u; 0,$)sin0 d@ d<& (40) The limits of integration are determined by the ranges of values of 0, <j> and 0, $ resulting from the collision. TRANSFORMATION BETWEEN LABORATORY AND CENTER OF MASS CROSS SECTIONS The center of mass differential cross section q(u;0,<j>) and the laboratory differential cross section are not independent. To obtain the relationship between q(u;0,<f>) and <2(w;0,$) we note first that a given region R on the projected area corresponds either to collisions in which the direction (0,<J>) lies in a certain solid angle co, or alternatively to collisions in which the direction (©,$) lies in a certain solid angle B. From the definitions of q(u;0,<f>) and 20; ©, ¢) it then follows that area of region R = f f q(u;0,<t>)sin0d0d<j>= f f Q(u;@,$)sin@d@d$ (41) " we transform the variables of integration in the last integral from 0 and $ to 0 and <j> we obtain j r?(w;e,^)sine^^=JJe(w;0,^)sin0l/(-|^j|^^ (42)
630 Two Particle Collisions where 7(0, ¢/0,4)) is the Jacobian for the transformation (Appendix 19). Since this equation must be true for arbitrary regions R, or equivalently arbitrary solid angles co, the integrands must be equal. Hence q(u;0,<t>)= Q(u;Q,$) sinQ sin0 \~0j) (43) If we substitute 0(0, <J>) and ¢(0, <j>) into Eq. (43) we obtain an equation that will enable us to transform from the cross section Q(u;&, ¢) to the cross section q(u; 0,<J>). Conversely if we substitute 0(0, ¢) and ¢(0, ¢) into Eq. (43) we obtain an equation that will enable us to transform from the cross section q(u;0,<j>) to the cross section £>(w; 0, ¢). However the evaluation of the Jacobians in the general case is exceedingly tedious. We therefore do not follow this direct route, but proceed to the desired relations by means of a number of short cuts. If one is not interested in the general case but in some particular case for which the transformation equations are not too messy, the direct route may be the easiest. We note first that (44) But r/o',e,$\ t( P',e,$ \r( °1 »°2.P3 \JS'i >*2.*3 \ = (t/2sin6) (l)(.y'2sin0) =ttf sinfl sin© and (45) (46) To obtain s'(v', 0, ¢), we take the sum of the squares of the three Eqs. (32), and find s' = f(t/)2+ w2-2i/N-w] i/2 where N is defined by Eq. (34). From Eq. (47) we obtain t/ — N • w V 9t>' /0,$ Combining Eqs. (27) and (44)-(48) we obtain s*[(v')2- v's*y< But from Eq. (25) v'N = s*(n + y) in0 rt 0,$\ = ;in0 \ 0,<J> ) sin N (47) (48) (49) (50)
Combining Eqs. (49) and (50) we obtain sin sin Summary <>31 (51) Finally substituting Eq. (28) in Eq. (51) we obtain (1+n-y) sin9 jl e,$\_ sin^ \ 0,<f> j /j + y2 + 2n • y) 3/2 (52) To obtain this result in terms of N rather than n we substitute Eq. (36) in Eq. (49) and find after simplifying the result sine jt 6,<l>\_ sin9 \ 9,<}> ) ±[(y.N)2-y2+l] {Y.N±[(Y.N)2-y2+l],/2} If we substitute Eq. (52) in Eq. (43) we obtain 1 + y *n (53) q(u;9,4>)=Q(u;e,&) (1 + y2 + 2yn) 3/2 (54) Since y is a constant, and we know 0(0, <f>), $(0, <f>), and n(0,<f>), Eq. (54) provides us with the means to transform from Q(u; 0,$) to q(u; 0,<j>). If we substitute Eq. (53) in Eq. (43) we obtain e(w;0,$) = ?(w;0,4>) {y-n±[i-y2 + (y-N)2],/2}' [1-Y2 + (Y-N)2] 2l'/2 (55) Since y is a constant, and we know 0(0,0), ¢(0, *), and N(0,$), Eq. (55) provides us with the means to transform from q(u;0,<j>) to Q(u; 0,0). Next we gather together the results obtained up to this point, to make their use as convenient as possible. SUMMARY Consider a collision between two particles of masses m and M and velocities v and V, respectively, from which two particles of masses m' and M' and velocities v' and V, respectively, emerge. Let u be the velocity of m with respect to M, u' the velocity of m' with respect to AT, and Ae/e the fractional change in the relative kinetic energy. Let the directions 1, 2, and 3 be defined by the unit vectors. _ u e2 _ vxV |vxV| C] = ¢2 X ¢3 (56)
632 Two Particle Collisions Let the center of mass angles 0 and <J> and the laboratory angles 0 and $ be defined by the relations u -^- =eisin0cos<i> + e9sin0sind> + e,cos# u z ~ = ejsin 0 cos $ + e2sin 0 sin O + e3cos 0 (57) (58) And finally let the center of mass differential cross section q(u;0,<j>) and the laboratory differential cross section <2(w;0,<I>) be defined by the condition that the projected area of the interaction region on the 1-2 plane associated with scattering into the solid angles sin 0d0d<j> and sin0d0d$, be respectively, q(u;0,<j>)sin0d0d<j> and Q(u; 0,$)sin0d0d$. It then follows that the transformation from laboratory quantities to center of mass quantities is governed by the equations q(u;0,<j>)=Q(u;®,<!>) 0 = cos 1 + yn (l + y2 + 2y.n)3/2 n3 + y3 $ = tan" (1 + Y2 + 2yn) n2 + y2 1/2 where and "i +Yi n = ejsin 0 cos <j> + e2sin 0 sin <f> + e3 cos 0 ,_/ m'M\x/1 r \ mM' J \ € + Ae ) V2 m\ + MV Mu (59) (60) (61) (62) (63) If the scattering particle is at rest, then y = e3(mm'/MM')1/2[c/(e + A«)]1/2. If in addition the collision is elastic, y = e3(m/M). The transformation from center of mass quantities to laboratory quantities is governed by the equations x211/2-|2 Q(u;e,$) = q(u;0,<t>) (yN±[l-y2 + (yN)2] [l-y2 + (y.N)2]1/2 0 = cos_1 <£ = tan"' y.N±[l-y2 + (y-N)2] }jV3- Y3 {yN±[l-y2 + (yN)2]/ }N2-y2 {yN±[l-y2 + (y-N)2]/ }jV,-y (64) (65) (66)
Problems 633 where N = e, sin 0 cos $ + e2sin 0 sin O + e3cos 0 (67) and y is given by Eq. (63). If y2 < 1 the plus sign is used. If y2 > 1 both signs are used. PROBLEMS 1. A neutron of mass m and speed v is incident on a nucleus of mass 1m, which is at rest. The total capture cross section is a0. If the neutron is captured, a compound nucleus of mass 8 m is formed, which in turn undergoes a fission process into two particles each of mass Am. Show that if the total kinetic energy is the same after the collision as before and there is no preferential direction in the center of mass frame for the emerging particles, the laboratory differential scattering cross section is given by [(6 + cos20)1/2 + cos0l G(«,e,*) = ^- ± r7r-i- l7T [7(6 + cos20)]1/2 2. In Problem 1 show that the probability that the kinetic energy E of an emerging particle falls between E and E + dE is a constant over a certain range of energies and zero otherwise. 3. Show that the laboratory differential scattering cross section for protons of energy E incident on target protons at rest is given by Q(u; 0,$) = (k/Ef[sin-4@ + cos~40]cos0 for 0 < 7r/2 and is zero for 0 > 77-/2, where the force between two protons is given by k/r2. 4. A particle with velocity V decays into two identical particles. Let a be the angle between the directions at which the two product particles fly off, and/(a) da the probability that a will fall between a and a + da. If the decay is isotropic in the center of mass system, and the speed of the product particles is V0 in the center of mass system, what is /(a)? 5. Find the range of possible values for the angle between the velocity directions after a particle of mass m has collided with a particle of mass M at rest. ^. Find the laboratory differential scattering cross section for the collision of a smooth perfectly inelastic sphere of mass m and radius a with a similar sphere at rest. '• An electron moving at infinity with a speed V collides with another electron at rest; the impact parameter is s. Determine the velocities of the two electrons after the collision.
62 The Restricted Three Body Problem INTRODUCTION In Chapter 25 we showed how the problem of two particles moving under the action of the mutual force of interaction between them can be reduced to a one particle problem. If the force of interaction is an inverse square force, exact solutions in the form of simple functions can be obtained. The problem of determining the motion of three particles moving under the action of the mutual force of interaction between them has never been solved in the same sense that the two particle problem has been solved. It is possible however to gain some insight into the three particle problem by considering the limiting case in which the mass of one of the particles is so small that it does not influence the motion of the other two particles. This case is called the restricted three body problem, and the problem becomes simply to determine the motion of one particle under the influence of two particles that have a preassigned motion. We designate the two more massive particles as particles 1 and 2, and their masses as mx and m2, respectively, and we assume m{ > m2 (1) We designate the particle of infinitesimal mass as particle/? and its mass as m. In the interest of simplicity we further restrict the problem, making the following assumptions: (1) the force between the particles is an inverse square attractive force, (2) particles 1 and 2 are moving in circles around their common center of mass, which is at rest, and (3) particle p is moving in the same plane as particles 1 and 2. THE MOTION OF PARTICLES 1 AND 2 By supposition each of the particles 1 and 2 is moving with a constant speed in a circle. The center O of both circles is their common center of mass. Since ^34
Coordinates <S35 the center of mass must always lie between the two particles, the two particles are located on opposite sides of the center and both are moving with the same angular speed co. If a particle is moving in a circle of radius r with constant angular speed co, its acceleration is directed toward the center of the circle and is of magnitude co2r. If we let rx and r2 be the radii of the orbits of particles 1 and 2 and if the magnitude of the force between the two particles is given by kmxm2/{r\ + r2)2> then from Newton's law kmxm2 ('i + r2): = mxco2rx = m2o)2r2 The following results, which we later use, can be obtained from Eqs. (2): k(mx + m2) co2 = ('1 + >*2)3 m~> rx + r2 m, + m2 my rx 4- r2 mx + m2 mlrl = m2r2 (2) (3) (4) (5) (6) COORDINATES We let X and Y be the coordinates of particle p with respect to a fixed Cartesian frame with origin O, and we let x and y be the coordinates of particle/? with respect to a frame with origin at O, but rotating with particles 1 and. 2 as shown in Fig. 1. Note that particle 1 falls on the negative x axis, and particle 2 on the positive x axis. The transformation equations between coordinates jc, y and coordinates X, Y are X = x cos cit — y sin ut Y = x sin cit + j cos cor (7) (8) Fig. 1
636 The Restricted Three Body Problem If we let pj and p2 be the respective distances of particle/? from particles 1 and 2 as shown in Fig. 1, then p2 = (^ + ^)2 + / (9) P22 = (* -r2f +y2 (10) from which it follows that *2+y2= 2 ' , -/v2 (ii) a*! + r2 v ' THE EQUATIONS OF MOTION The kinetic and potential energies of particle /? are given by T={m(X2+ Y2) = \m[x2 + f - luxy + luxy + u\x2 + y2)] (12) kmmx kmm>> V=- -- - (13) Pi Pi v ' Substituting Eqs. (12) and (13) in Lagrange's equations of motion, d_(ZT\_ZT = _W (U. dtXdx) dx dx V> A( 1Z\- 91= _ IZ n^\ dt\dy ) dy ty K ' we obtain mx — 2muy — mu2x = —r— (16) C\ xr my + 2mcox — mo^y = — -r— (17) If we define ^£-I«2(*2+/+'.'2) then Eqs. (16) and (17) can be written x - 2coy = - ^~- y ox ...o- 9£/ y + 2cox = — -TT- vy These are the equations of motion for particle /? that we use. (18) (19) (20) THE POTENTIAL U As we shall see in the following sections, a considerable amount of information about the motion of particle/? can be extracted from the potential U. We
The Potential U *37 therefore begin by gathering some properties of U to be used in later calculations. We note first that if we make use of Eqs. (13), (3), (11), (4), and (5) in Eq. (18), the definition of U can be written U = —km] + Pi Pi 2(r, + r2f — km- 1 + Pi Pi 2(r, + r2f The derivatives of U with respect to px and p2 are given by dU 3Pi = km. 9p2 1 p? 1 c Pi + r2f Pi p\ ('i + r2f (21) (22) (23) 3pi 3.x 3p2 3.x 3Pi 3y X + A*! Pi x — r2 Pi Pi 9P2 _ y dy p2 Making use of Eqs. (22)-(27) and Eqs. (3) and (6) we obtain Using Eqs. (9) and (10), we obtain for the derivatives of px and p2 with respect to x and y (24) (25) (26) (27) (28) (29) (30) (31) (32) 3^ kmx{x + rx) km2(x — r2) dx Pi 2 du km\y ^ kmiy 7 p\ d2u = km, dx2 ' ' 1 3(x+ ,-,)2 rt3 p? + km- 3(x - r2)2 Pi 4 d2u dy2 d2U dxdy = km, = —km, 111 P? + km- 3 5 Pi Pi — 0) 3(x + rx)y P\ — km- 3/ p! 3(* - ^)/ pI
638 The Restricted Three Body Problem CONSTANTS OF THE MOTION If we multiply Eq. (19) by x and Eq. (20) by y and add the resulting equations, we obtain xx which can be rewritten from which it follows «+»-(f )*-(*?)> <*> dt K+i->! f <34> {xz + \yz+ U=C (35) where C is a constant whose value is determined by the initial conditions. The quantity \ x2 + \y2 + U is thus a constant of the motion. The same result could have been obtained by noting that since L= T — V is not an explicit function of the time, the quantity (dL/dx)x + (dL/dy)y — L is a constant of the motion. Equation (35) is formally the same as the energy equation for a particle of unit mass moving in a potential U. It follows that if particle p is in a region R bounded by a closed curve defined by the condition U = E/0, where U0 is some constant, and if C < U0, particle/? will be trapped in the region R. The particle could be trapped for example in the neighborhood of particle 1 or particle 2, where the potential U has a deep well. If the equations of motion of the particle/? were x = —dU/dx and y = —dU/dy, its behavior within the region R, or more generally its behavior anywhere in the potential U, would be similar to that of an ordinary particle moving in a potential U. However, the equations of motion are not of this form but are given by x — 2o)y = — 9 U/dx andy + 2cox = — 9 U/dy; hence we cannot determine the motion of the particle by analogy with an ordinary particle in a potential. As the following sections demonstrate, it is even possible for a particle to be trapped not only near a minimum of the potential U but also near a maximum of the potential U. EQUILIBRIUM POINTS From Eqs. (19) and (20) it follows that it is possible for the particle p to remain at rest in the rotating coordinate system at points at which 9 £//9.x = 0 and 9 U/dy = 0. Such points are called equilibrium points or libration points. The task of finding the equilibrium points can be simplified if we write the equilibrium conditions in the form at/ = M^i + |t/^ = 0 (36) ox opx ox op2 ox v dy 9pj dy 9p2 dy {
Equilibrium Points 639 Equations (36) and (37) constitute a set of simultaneous equations for the quantities dU/dp} and dU/dp2. Hence if dPi 9P2 9.x 9.x dPi 9P2 9y 9j then the only solutions to Eqs. (36) and (37) are ones for which ^0 (38) = 0 (39) dU^dU 3Pi 9P2 Substituting Eqs. (22) and (23) in Eqs. (39) we obtain immediately Pi = P2 = ''i + ^2 (40) It follows that if we construct an equilateral triangle with particles 1 and 2 at two of the vertices, the third vertex will be an equilibrium point. There are two such points, one in the positive y region and one in the negative y region. The point with y greater than zero is called the fourth Lagrange point and is labeled L4, and the point with y less than zero is called the fifth Lagrange point and is labeled L5. It is still possible to have other equilibrium points, provided they are points at which the determinant in Eq. (38) is zero, that is, provided 3pi dx 3Pi dy dp2 dx dp2 dy x + rx P\ z Pi x — r2 Pi y_ Pi = 0 (41) Equation (41) can be true if and only if y = 0. Hence all other equilibrium points must lie on the x axis. If we set y = 0 in Eq. (28) and then the result equal to zero, noting when y = 0 that px = \x + r,| and p2 = |jc — r,|, we obtain '2\> du kmx(x + rx) km2(x-r2) —- = — — + — — - co x = 0 dx \x + rA3 x-rJ3 (42) one This equation has three solutions. One solution lies in the region x < - lies in the region — r, < x < r2, and one lies in the region x > r2. To verify this we note from Eq. (30) that the second derivative of U with respect to jc, with y = 0, is given by 2rai 2m9 ~ 1 2 2 (43) d2U dx2 \X + A*, — CO ii \x-r2\3 The second derivative is everywhere negative, except at the points x = — rx and x = r2, where it is undefined. It follows that dU/dx is decreasing in each °f the three regions described above. Thus as x goes from — oo to — r,, vU/dx decreases monotonically from + oo to — oo; hence there is one and 0nly one point in the region x< — r, at which dU/dx = 0. In a similar
640 The Restricted Three Body Problem fashion we can show there is one and only one point at which 3 U/dx = 0 in each of the other two regions. The three equilibrium points corresponding to these three solutions are called the third, first, and second Lagrange points respectively and are labeled L3, L,, and L2, respectively. If one investigates the potential U at the equilibrium points, it will be found that the points L,, L2 and L3 occur at saddle points and the points L4 and L5 occur at maxima. STABILITY OF MOTION AT THE LAGRANGE POINTS In the preceding section we found that it is possible for the particle p to remain at one of the five Lagrange points during its motion. Now we inquire whether the particle p will remain in the neighborhood of a Lagrange point if its equilibrium is slightly perturbed, that is, whether the equilibrium is stable or unstable. To answer this question we consider solutions to the equations of motion in the neighborhood of a Lagrange point. If we let jc0, y0 be the coordinates of a Lagrange point, define £ and ri by the equations x = x0 + £ (44) y=y0 + v (45) and use the notation Ux,Uy, Uxx, Uxy, Uyy, . . . to represent the various partial derivatives of U, then the equations of motion, that is, Eqs. (19) and (20), become t-2vii=-Ux(x0 + t,y0 + ri) (46) 7) + 2^=-^0 + ^0 + 7,) (47) If we expand the functions on the right hand sides of Eqs. (46) and (47) in powers of £ and ti and keep only terms linear in £ and ti we obtain I - 2cori = - Uxx(x0, y0)£ - Uxy(x0, y0)7i (48) ij + 2^ = - Uxy(x0, y0)£ - Uyy(x0, y0)V (49) If we consider motion in the neighborhood of one of the Lagrange points Lv L2, or L3 then y0 = 0, px = |jc0 + r,|, and p2 = \x0 — r2|; and thus from Eqs. (30)-(32) tUWo)--2Q2-«2 (50) £^(*c»/o) = 0 (51) ik(wo) = a2-«2 (52) where „-> km. km-, Q2= !—- + 2—- (53) |*0+,-,13 K+,2|3 V
Stability of Motion at the Lagrange Points ¢41 Substituting Eqs. (50)-(52) in Eqs. (48) and (49) we obtain £ - 2cotj - (2S22 + co2)£ = 0 (54) r, + 2co£ + (Q2 - co2)tj = 0 (55) The solution to Eqs. (54) and (55) is of the form € = Aex' (56) V = BeXt (57) where A, B, and A may be complex (see Appendix 9). Substituting Eqs. (56) and (57) into Eqs. (54) and (55) we obtain (58) If a nontrivial solution for A and B is to exist, it is a necessary and sufficient condition that the determinant of the square matrix in Eq. (58) vanish. Setting this determinant equal to zero we obtain A4 - (22 - 2co2)A2 - (222 + co2)(B2 - co2) = 0 Solving for A2 we obtain 1/2 " A2 - 2fi2 - co2 - 2coA 1 2coA A2 + fl2 - co2 J \a' [b. 0 .0 \2 = i(Q2 - 2w2) ± i(9fl4 - 8fiV) But from Eq. (42), the equilibrium point x0 satisfies the equation kmx(xQ + r\) km2(xQ - r2) (59) (60) — co x0 = 0 l*o+/"il \x0-r2\3 Combining Eqs. (53) and (61) and noting that mxrx = m2r2 we obtain 1 1 (fi2 - co2) = kmxrx I. 1*0 - r2? |x0+ r, | (61) (62) If x0< —/-,, or xQ > r2, the right hand side of Eq. (62) is positive, hence Q > co2. If — /-, < x0 < /-2, a direct comparison of the definitions of £22 and co2 will again show that S22 > co2. But if fi2 > co2 and we choose the positive sign in Eq. (60) then where and A2 = 1^(1-27) + (9-87)-/2] ^(A2) = ifi2[-2-4(9-8y)-1/2]<0 1 fl2 0< y< 1 (63) (64) (65) (66)
642 The Restricted Three Body Problem It follows that as y ranges from 0 to 1 then A2 is real and decreases continuously from 222 to 0. Thus there always exists a real positive value of \ that when substituted into Eqs. (56) and (57) will provide a solution to Eqs. (54) and (55). Such a solution increases exponentially with time, hence implies an unstable situation. The equilibria at L,, L2, and L3 are thus unstable. If we consider motion in the neighborhood of the Lagrange point L4 then Pi = Pi = r\ + r2 — P> xo + r\ = P/2, x — r2 = —p/2,y = 73 p/2. Hence Uxx=-3a2 uxy=-P2 uyy = -9a2 where CO 2 (67) (68) (69) (70) (71) (72) (73) 2^ Z$tmx -m2\2 4 \mx + m2 ) Substituting Eqs. (67)-(71) in Eqs. (48) and (49) we obtain £-4a7]-3a2£- (1^ = 0 Substituting Eqs. (56) and (57) into Eqs. (72) and (73) we obtain ' A2-3a2 -4ak-p2' 4a\- /32 X2 -9a2 Setting the determinant of the square matrix in Eq. (74) equal to zero, and using Eqs. (70) and (71) we obtain A4 + 4a2A2+a2i?2 = 0 (74) where Solving for X2 we obtain R2 = Ylmxm2 A2 = 2«2[-l±(l-i?2)1/2] (75) (76) (77) If R2 < 1 we obtain four purely imaginary values for A; hence the motion is stable. If R2 > 1 we obtain four complex values for A, two of which have positive real parts; hence the motion is unstable. The case R2 = 1 also leads to an unstable situation. It follows that the equilibrium at L4 or L5 will be stable if and only if R< 21mxm2 {mx + m2f < 1 (78)
Problem 643 0r equivalently if and only if mx 2 25 v J The stability of equilibrium at the points L4 or L5 is of some interest in two situations. At both the L4 point and also at the L5 point in the system in which particle 1 is the sun and particle 2 is the planet Jupiter, there is a cluster of asteroids called the Trojan asteroids. The L5 point for the system in which particle 1 is the earth and particle 2 is the moon was one of the first sites suggested for the colonization of space. The preceding investigation of the stability of equilibrium in the neighborhood of the Lagrange points is limited by the fact that it is valid only very close to the equilibrium points. To determine what happens for larger perturbations is a much more difficult problem. PROBLEM 1. Three point masses mx, m2, and m3, are located at the corners of an equilateral triangle with sides of length a. The only forces acting on the particles are those of their mutual gravitational attraction. a. Show that the net force acting on any one of the particles is directed toward the center of mass of the system. b. Show that a constant rotational motion of the system as a whole about the center of mass can be found such that the particles move without changing their positions relative to one another. What is the required angular velocity?
J
SECTION 2 Rigid Body Motion
J
63 Rigid Body Kinematics INTRODUCTION The subject of rigid body motion was considered in great detail in Section 2 of Part 2. In this section we reconsider the problem from the Lagrangian point of view and introduce some new ideas. This chapter briefly reviews concepts from Section 2 of Part 2 that are needed. The reader who has no previous experience with these concepts should go back and study the material in Section 2 of Part 2. THE INERTIA TENSOR The moment of inertia of a system of particles with respect to an axis that passes through a point o and is in the direction of the unit vector n is defined as follows: I(o,n) = ^m(p)[p(p)]2 (1) P where m(p) is the mass of the/?th particle and p(p) is the distance of the/?th particle from the axis. Noting that p(p) = |n x r(op)\, where r(op) is the vector position of the pth particle with respect to the point o, expressing n and r(op) in terms of their components with respect to a Cartesian frame, and substituting the result in Eq. (1), we obtain ^») = 22W^ (2) < j where ^j(o)^m(p)[r\op)8iJ- xl(ap)xJ(ap)] (3) Fhe set of quantities I^o) is called the inertia tensor for the point o. The following two theorems provide us with some useful properties of the inertia tensor. 647
648 Rigid Body Kinematics Theorem 1. Let S be a Cartesian frame with axes 1, 2, and 3 in the directions of the unit vectors el5 e2, and e3; and let S' be a second Cartesian frame with axes r, 2', and 3' in the directions of the unit vectors er, e2>, and ey. The components I{- in the frame S, and the components /,y in the frame S", of the inertia tensor with respect to a given point, are related as follows: 7/y = SS^/A (4) r s where erj = er-ej (5) Theorem 2. The inertia tensor of a dynamical system with respect to an arbitrary point a is given by I,j(a) = I0(c) + M[R\a)8ij - *,(«) W] (6) where I^c) is the inertia tensor with respect to the center of mass c, M is the total mass of the system, and R(a) = e}X}(a) + e2X2(a) + e3X3(a) is the vector position of the center of mass c of the system with respect to the point a. PRINCIPAL AXES It is always possible to choose a frame S, with axes I, 2, and 3, such that the inertia tensor with respect to a given point is diagonal, that is, IU=lT*U (7) We call the three axes I, 2, and 3 for which this is true the principal axes. The corresponding directions are called principal directions and are designated by the unit vectors eT, e5, and e5. The moments of inertia associated with the principal axes are called principal moments of inertia. From the equations above it is apparent that the quantities /T, /2, and /5 are the principal moments of inertia. The following theorem tells us how to find the principal axes and the principal moments of inertia from knowledge of the inertia tensor with respect to some frame. Theorem 3. Consider a dynamical system for which the inertia tensor 1^ for a given point is known. a. The principal moments of inertia /T, I2, and /5 are equal to the three roots of the cubic equation \Iy - 4-5,,1 = 0 (8) where the notation | • • • | stands for the determinant of the matrix
The Euler Angles 649 b. Given a principal moment of inertia /^, the direction e^ of the corresponding principal axis can be obtained by solving the set of simultaneous equations 2[^-4«//K-=0 '"=1,2,3 (9) J for the components e^j of the vector e^. THE EULER ANGLES Six coordinates are required to specify the configuration of a rigid body with no point fixed; three are needed to specify the position of one point in the body, and three to specify the orientation of the body. The techniques for specifying the position of a point are already quite familiar to us. We therefore concentrate on the specification of the orientation of a rigid body. To make the discussion as simple as possible, we assume that we are dealing with a rigid body in which one point O is fixed. There are many sets of coordinates that can be used to specify the orientation of a rigid body. One of the simplest sets, and the set most often used, is the set of Euler angles. Before defining the Euler angles it is convenient to define four frames S, S, S, and S, as follows: S An inertial frame with origin at O. S A frame fixed in the body with origin at O and axes in the principal directions for the point O. A A A »3 A frame with origin at O, axis 3 coincident with axis 3, and axis 1 in the direction of the vector e3 x e^. S A frame with origin at O, axis 3 coincident with axis 3, and axis 1 in the direction of the vector e3 x e5. The Euler angles <f>, 0, and \p are then defined as follows: 4> The angle that the 1 (or 1) axis makes with the 1 axis, or equivalently the angle that the projection of the 3 axis on the 1-2 plane makes with the negative 2 axis. 0 The angle that the 3 axis makes with the 3 axis. ^ The angle thaUhe T axis makes with the 1 (or 1) axis, or equivalently the angle that the 1 axis makes with the direction of the vector e3 x e3. Note that the frame S can be obtained by rotating the frame S about the 3 a*is through an angle <f>, the frame S can be obtained by rotating the frame S about the 1 axis through an angle 0, and the frame S can be obtained by r°tating the frame S about the 3 axis through an angle \p.
650 Rigid Body Kinematics THE ANGULAR VELOCITY OF A RIGID BODY The angular velocity 63 of a rigid body is identical to the angular velocity of frame S with respect to frame S (see Chapter 6), which can be written as the sum of the angular velocity of frame S with respect to S, the angular velocity of S with respect to S, and the angular velocity of S with respect to S. The angular velocity of S with respect to S is e3<j>; the angular velocity of S with respect to S is e\0\ the angular velocity of S with respect to S is e3^. Hence 63 = e3<j> + ef 0 + e3^ (10) The angular velocity 63 can be expressed in terms of its components in any frame by simply expressing the unit vectors e3, ef, and e3 in terms of the unit vectors in the desired frame. If we express e3, ef and e3 in terms of the unit vectors ej, e^, and e3 we obtain 63 = eT(<j>sin0sin\p + 0 cosxp) + e^sin0cos\p — 0 shu//) + e3(<j>cos# + ^) (ii) If we express e3, ef and e3 in terms of the unit vectors e,, e2, and e3 we obtain 63 = ej(^ cos<j> + \p sin<J>sin0) + e2(0 sin<J> — \p cos<J>sin0) + e3(<j> + \j; cos0) (12) THE KINETIC ENERGY OF A RIGID BODY The kinetic energy of a system of particles with respect to a point o is defined as T-\^m{p)i(op).i(op) (13) P where m(p) is the mass of particle/? and r(op) is the vector position of p with respect to o. If the system is a rigid body and the point o is a point fixed in the rigid body then i(op) = <exr(op) (14) Substituting Eq. (14) in Eq. (13), expressing the vectors in terms of their components with respect to an arbitrary frame S, and rearranging terms, we obtain '' J If we choose the frame S to be the frame S we obtain T{o)=\^Ij{o)ui (16)
Examples 651 Combining Eqs. (11) and (16) we obtain 2 • 2 T= \I-[(<j>sin9sin\p + 0 cos;//) + \Ifasin0cos\p — 0 simp) + ±Ii(j>cos0 + xp)2 (17) CONCLUSION Having obtained the kinetic energy of a rigid body with one point fixed as a function of the Euler angles and their time derivatives, we are now in a position to write down Lagrange's equations of motion. This is the subject of Chapter 64. EXAMPLES Example 1. A homogeneous cone of mass m, height h, and semivertical angle a, rolls without slipping on a horizontal plane. A point on its axis describes one complete circle in time r. Find the kinetic energy of the cone. Solution. Let the inertial frame S be chosen with origin at the tip of the cone and 3 axis vertically up, and the body frame S chosen with origin at the tip of the cone and 3 axis along the symmetry axis of the cone. With this choice /T=/2=^(tan2« + 4) (18) /5=^ftan2« (19) 0=§ -a (20) ¢=^ (21) *=--$- (22) sin a v ' Setting /5 = /T in Eq. (17) we obtain T=\ /T(<j>2sin20 + 6 2) + i /3(4, cos 9 + 4, f (23) Substituting Eqs. (20)-(22) in Eq. (23) we obtain T-^jjg-V+Sccta) (24)
652 Rigid Body Kinematics PROBLEMS 1. What is the ratio of the height of a right circular cylinder to its radius such that the principal moments of inertia for the center of the cylinder are all equal? 2. Show that the components of the angular velocity vector with respect to the frame S are (cof, co^, coj) = (0, — ^ sin0,<j> + \pcos0) and with respect to the frame S are (cof, u-;, u$) = (0, <j> sin 0, <j> + cos 0 + \p). 3. A uniform rectangular block of mass m has adjacent edges of length 2a, 2b, and 2c._ Let 5 be a frame fixed in the body with origin at the center of the block, I axis parallel to the side of length 2a, 2 axis parallel to the side of length 2b, and 3 axis parallel to the side of length 2c. The block is rotating with angular speed co about an axis that passes through the vertex at the point (X\,X2,x?) = (—a,—b,c) and is parallel to the principal diagonal passing through the point (X\,X2,x?) = {a,b,c). Show that the kinetic energy is (mu2/3)[(a2b2 + la2c2 + lb2c2)/(a2 + b2 + c2)]. 4. An axially symmetric disk of mass m, radius a, and moment of inertia mk2 about its axis, rolls along the ground at a constant inclination a, and its center C describes a horizontal circle of radius c, with speed v. Show that the kinetic energy is (mv2/2)[\ + (A;2sin2a/2c2) + (k2/a2)]. 5. Two wheels, each of mass m and radius a, are joined by an axle of mass M and length a. The moment of inertia of each wheel about its axis is A and about a diameter is B. The moment of inertia of the axle about a perpendicular line through its center is C. The system is axially symmetric. The combination rolls on a horizontal plane, and the angular speeds of the wheels are a and (3. Show that the kinetic energy is I (ma2 + M"L +Aya + £)2+ 1 ^ma2 + ,4+45 + 2C)(a -/3)2
64 Rigid Body Dynamics INTRODUCTION In the preceding chapter we obtained the kinetic energy of rotation of a rigid body. We are now in a position to use Lagrange's equations to obtain the equations of motion of a rigid body. LAGRANGE'S EQUATIONS OF MOTION FOR A RIGID BODY WITH ONE POINT FIXED The Euler angles 0, <f>, and \p constitute a complete set of generalized coordinates for a rigid body with one point fixed. If we choose the Euler angles as generalized coordinates, then Lagrange's equations of motion for a rigid body that has one point fixed, but is otherwise subjected to no constraints, are given by dt\u) 30 Qe () *{%)-%-* .« *(if)-SJ-* <3> where T = (0, <fc i//, 0, <fc ip, f) is the kinetic energy, and Qg, Q^ and Q^ are the generalized components of the force. If in addition to the given force the rigid body is subjected to a holonomic or an anholonomic constraint that restricts the displacements of the body to those satisfying an equation of the form fdB + gd$ + hdty+ kdt = 0 (4) 653
654 Rigid Body Dynamics where/, g, h, and k are functions of 0, <j>, ty, and t, then Lagrange's equations of motion are i(i)-if-a+v <5> /0 + g<j> + /z^ + A; = 0 (8) If the generalized components of the force are derivable from a generalized potential U, the equations of motion are d_ dt (f)-i=° m (f)-f=° <">> d_ dt \ 3<j> d(dL\_^=Q (11 where L=T-U (12) To use the equations above we need to know T(0, <f>, ^/, ¢, <j>, ^, /), and if the force is not derivable from a potential we need to be able to obtain Qg, Q^, and Qr In the preceding chapter we showed that T=\I-xo>\ + {I-2o>l + \I-^ (13) where coT = <j> sin 0 sin ip + 9 cos ip (14) C05 = <j> sin 0 cos i// — 0 sin ip (15) CO3 = <j>COS0+ ^ (16) and /T, 15, and /5 are the T, 2, and 3 principal moments of inertia. To obtain the generalized components of the force, we note that the work done in a virtual displacement of the rigid body is given by 8W= G\80+ G38<f>+ G3~&// (17) where Gf, G3, and G§ are the 1, 3, and 3 components, respectively, of the
Lagrange's Equations of Motion of a Rigid Body with No Point Fixed <S55 torque G with respect to the fixed point. It follows that Qe=Gi= Gf = G]COs4>+ G2sin<J> = Gjcosxp- Gjsim// (18) Q+ = G3 = Gj = G^sin 9 + G5COS 9 = GTsin 9 sin \p + G^sin 9 cos \p + G5cos 9 (19) £^ = G3 — G3 = GjSin^sin^ — G2cos4>sin0 + G3cos0 = ~G2sin^+ G3xos0 (20) We now have enough information to write down Lagrange's equations of motion for a rigid body with one point fixed. We will not substitute the expressions for T, Q9, Q^, and Q^ in the equations of motion, since this operation is straightforward, and the results in the general case where there is no symmetry and none of the components of the force vanish are not particularly informative. LAGRANGE'S EQUATIONS OF MOTION OF A RIGID BODY WITH NO POINT FIXED The extension of the preceding results to a rigid body with no point fixed is straightforward. In this case we break the motion down into the motion of the center of mass, and the motion with respect to the center of mass. If we let X, Y, and Z be the position coordinates of the center of mass with respect to an inertial frame, the configuration of the body is determined by the six coordinates X, y, Z, 9, <J>, and \p and the kinetic energy is given by T = {MX1 + {MY2 + {MZ2 + i/T6>? + {I-2o>\ + i/36>| (21) where M is the mass of the body, and /T, 1^, and /5 are the T, 2, and 3 Principal moments of inertia for the center of mass, rather than for the fixed Point, as was the case when one point was fixed.
656 Rigid Body Dynamics EULER'S EQUATIONS OF MOTION If we substitute Eqs. (13) and (20) in Eq. (3) we obtain A dt 9coT 9co2 9co5 9co-f 9co2 9co5 " 7^t "aX ~ ^ lu " ^ IT = G3 (22) 8^ 2 2 8^ 3 3 3)// From Eqs. (14)-(16) we obtain 3«t 3co^ 1 - 2 =0 (23) 3i// 3)// = 1 (24) 9C05 9co-f 9co2 9c03 (25) (26) = 0 (27) Substituting Eqs. (23)-(27) in Eq. (22) we obtain /5CO3 — COyCO^/j — 12) = G3 (28) Since the initial choice of the T, 2, and 3 axes required only that they be in the principal directions and form a right handed system, we can generate two more equations by replacing the respective subscripts T, 2, 3 in Eq. (28) by the subscripts 3, I, 2 or 2, 3, I. Doing this we obtain h^i ~ <*?P\(h - h) = Gi (29) /TcoT - 602603(/2 - /5) = GT (30) Equations (28)-(30) are called Euler's equations of motion. If we substitute Eqs. (14)-(16) in Euler's equations we obtain a set of equations sufficient in principle to obtain 9{t\ <J>(/), and \p(t). Euler's equations do not offer any particular advantages over Lagrange's equations obtained earlier if we simply want to solve for 9{t\ <f>(0> anc* *K0- However Euler's equations are extremely convenient in some situations. If for example we know the angular velocity of a body, Euler's equations can be used to determine the torque acting on the body.
Examples 657 EXAMPLES Example 1. Symmetric Top in a Gravitational Field. A rigid body, having an axis of symmetry, is constrained to rotate about a fixed point O on this axis. The body is in a uniform gravitational field. Describe the motion. Solution. If we choose the 3 axis to be vertically up, and the 3 axis to be along the symmetry axis of the top, and if we let ^ = /2 = ^,/3 = C, and h be the distance from the point O to the center of mass, the Lagrangian for the system is given by L = T - V = \I-Xu\ + \h<*\ + \h<*\ - mgh cos9 = \A($hm29 + 92) + \C(j>cos9 + x^f- mghcos9 (31) The Lagrangian L is not a function of <j> or \p, and therefore 3L/3<j> and dL/d\p are constants of the motion. Furthermore L =^ L(t), V= V{9\ and T is a homogeneous quadratic function of <j>, 9, and ip; hence r+ V is a constant of the motion. We thus have ^4 = A$sm29 + Ccos0(<j>cos0 + ^) = <fc4 (32) ^4 = C(<j>cos0 + yp) = j&4 (33) dxp T+ V=\A{$hm29 + 02) + ^C(<j>cos0 + ^)2 + mghcos9 = E (34) where a, /?, and £ are constants. Substituting Eq. (33) in Eq. (32) we obtain a — /? cos 9 sin20 Substituting Eqs. (33) and (35) in Eq. (34) we obtain I ao2 L I a(cl- P*os9a H2+Mf^F)+'"^-£' <*> where R2A2 *' = b-\c. (37) Equation (36) can in principle be used to obtain 0(/). The nature of the solution can be visualized by noting that Eq. (36) is analogous to the equation of motion of a particle of mass A and energy E' moving in one dimension 9 in a potential \A[(a - /3cos9)/sin9]2 + mghcos9. Given 9(t) we can substitute it in Eq. (35) and solve for <j>(t). Given 9{t) and <}>(t) we can use Eq. (33) to obtain ^/(/). Equations (33), (35), and (36) are essentially the same equations we obtained in Example 2 in Chapter 40. The reader can refer to that example for the remaining analysis of this problem.
658 Rigid Body Dynamics Example 2. A sphere of mass M and radius a moves on a perfectly rough horizontal plane under the action of a force F whose line of action passes through the center of mass. The distribution of mass in the sphere is spherically symmetric, and the moment of inertia about any diameter is I. Determine the equations of motion of the sphere. Solution. The sphere has five configurational degrees of freedom. If we choose an inertial frame OXYZ with X-Y plane on the surface, then the X and Y coordinates of the center of mass of the sphere together with the Euler angles <f>, 0, and \p constitute a suitable set of generalized coordinates. Letting Z = 0, /T = I2 = 12 = I, and using Eqs. (14)-(16) in Eq. (21), we obtain for the kinetic energy T = \I{$>2 + 02 + yp2 + 244 cos0) + \MX2 + {MY2 (38) Since the surface is perfectly rough, the point P on the sphere that is in contact with the surface must have zero velocity. If we let i, j, and k be unit vectors in the X, Y, and Z directions, respectively, the velocity of P is v = iZ + j7+<ox(-ktf) (39) The first two terms on the right hand side of Eq. (39) are the X and Y components of the velocity of the center C of the sphere, and the last term is the velocity of P with respect to C. Setting v = 0 we obtain X —acoy = X — a(0 sin<J> - \p cos<J>sin0) = 0 (40) Y + aax = Y + a(0 cos<f>+ ^sin<f>sin0) = 0 (41) or equivalently dX - asin^dO + a cos<J>sin0<A// = 0 (42) dY+ acos<j>d0 + asm<f>sin0d\p = 0 (43) The constraints represented by Eqs. (42) and (43) are anholonomic. Applying Lagrange's equations for anholonomic systems we obtain MX = Fx + Xx (44) MY= Fy + X2 (45) l4> + Icos04,-Ism00xj, = O (46) 10' + /sin#<j>^ = — AjtfshKj) + \2acos§ (47) I\p +/cos0<J> -Ism0<j>0 = H-Aj^cos^sin^ + \2asm<f>sm0 (48) These five equations together with the constraint equations (42) and (43) provide us with seven equations in the seven unknowns X, Y, <j>, 0, \p, Xl5 and \2. The reader should compare this result with the result obtained in Example 4 of Chapter 40. Example 3. An asymmetric rigid body is rotating about a fixed point O. No torques are acting. Show that if the body is rotating about the principal axis
Examples 1559 with which either the largest or smallest moment of inertia is associated, the motion is stable, but if it is rotating about the remaining axis the motion is unstable. Solution. To simplify the notation let Ij = A, 1^= B, /5 = C, coT = /?, <o5 = q, and co5 = r. Since there are no torques, Euler's equations can be written Ap-qr(B- C) = 0 (49) Bq-rp(C- A) = 0 (50) Cr-pq(A- B) = 0 (51) If one of the components of the angular velocity is constant and the other two vanish, Euler's equations will be satisfied. Hence steady rotation about any one of the principal axes is a possible motion. Let us assume that the body is initially rotating about the I axis with angular speed £2 and at time t = 0 receives a small impulsive torque. At an arbitrary but short time t later the values of p, q, and r will be given by p = 2 + a (52) q = P (53) r = y (54) where a, /?, and y are small quantities. Substituting Eqs. (52)-(54) into Euler's equations and dropping terms involving products of a, /3, and y with one another, we obtain to this order of approximation a = 0 (55) y = /mA^B (5?) Equations (56) and (57) can be combined to give /?=-*/? (58) Y=-Ay (59) where (A-B)(A-C) X = Q Jc (6°) If A is greater than both B and C or if A is less than both B and C, then X will be greater than zero. Hence according to Eqs. (55), (58), and (59) the values of ft and y will oscillate about their initial values while a remains constant; thus the motion is stable. But if the magnitude of A falls between the magnitudes °f B and C, then according to Eqs. (58) and (59) it is possible for /5 and y to ^crease exponentially with time; thus the motion is unstable.
660 Rigid Body Dynamics PROBLEMS 1. Two particles of mass m are joined by a rigid massless rod of length a. The rod is made to rotate with constant angular speed £2 about an axis passing through the center of the rod and making an angle a with the rod. Using Euler's equations, find the components, along the principal axes of the system, of the torque driving the system. 2. One end of a uniform rod of mass m and length a is constrained to slide along a smooth vertical axis. The other end slides on a smooth horizontal floor. Let 0 be the angle between the rod and the downward vertical, and <J> be the angular displacement of the vertical plane through the rod. If at / = 0, 0 = 77-/4, 9 = 0, and <j> = Q, show that at any time later 02 = (S2/4) (2 - csc20) + (3g/a)(2~l/2 — cos0), provided the rod remains in contact with the floor. 3. A pole AB of mass m, moment of inertia mk2 about a perpendicular axis through the end A, and having its center of mass located a distance c from the end A, is supported at end A, which is carried around in a horizontal circle of radius a with constant angular speed co. Prove that it can maintain a constant inclination a to the vertical provided gctana = co2(ac - khina). 4. A homogeneous circular disk is pivoted about its center. It is set spinning with angular speed co about a line making an angle a with the normal to the disk. Find the time as a function of co and a required by the axis of the disk to describe a cone in space. 5. The center of a thin uniform disk of radius a and mass m is attached to one end of a thin massless rod of length b so that the rod is normal to the plane of the disk. The other end of the rod is smoothly hinged to a vertical shaft that is driven at an angular speed £2. Thus the vertical plane containing the shaft and the rod rotates at an angular speed 2. Show that motion with the center of the disk in its lowest position, hence the rod vertical, is stable for all 2 if a > 2b, and is unstable for S2 > 4gb/(4b2 - a2) if a < 2b. 6. The center C of a thin uniform disk of mass m and radius a is attached to the end C of a thin massless rod AC of length b so that the rod is normal to the plane of the disk. The system is placed on a rough inclined plane that makes an angle a with the horizontal. The end A of the rod remains stationary at a point O on the plane, and the disk is free to roll along a circle of radius c = (a2 + b2)x/2. Choose the inertial frame S with origin at O, 3 axis perpendicular to the inclined plane, and 2 axis pointing up the plane. Choose the body frame S with origin at O and 3 axis pointing in the direction of the rod AC. Show that as long as the end A of the rod remains on the plane, the equation of motion for the Euler
Problems 661 angle <J> is that of a simple pendulum with angular frequency given by Q2 = (4gcsina)/(a2 + 6b2). 7. A uniform cube is spinning freely about a diagonal. Show that if one edge is suddenly fixed, the kinetic energy is reduced by a factor -^. 8. A uniform disk of radius a and mass m rolls down a flat inclined surface which makes an angle a with the horizontal. The disk is constrained so that its plane is always perpendicular to the inclined plane, but it may rotate about an axis perpendicular to the surface. Determine the motion of the center of mass of the disk. 9. Using Lagrange's equations for anholonomic systems, prove that if a homogeneous sphere rolls on a rough fixed plane under the action of a force whose line of action passes through the center of the sphere, the motion of the center will be the same as it would be if the plane were smooth and all the forces were reduced to five-sevenths of their former values. 10. A sphere of mass m, radius a, and moment of inertia I about any axis through the center, rolls without slipping on a horizontal plane, and the plane rotates about a fixed vertical axis with constant angular speed S. Show that the center of the sphere moves in a circle with constant angular speed S/[l + (ma2/1)]. 11. A gyroscope is mounted with its center of gravity at the center of the gimbals so that there is no gravitational torque. However, the figure axis is constrained to move in the horizontal plane only. If the gyroscope is set spinning on the earth's surface, there will be an additional rotational motion due to the earth's rotation. Show that if the angular frequency of the gyroscope is large compared with the earth's rotation, the figure axis will oscillate symmetrically about the meridian and can thus be used as a compass. Determine the period of these oscillations.
J
SECTION 3 Small Oscillations
J
65 Vibrating Systems INTRODUCTION In dynamics we frequently deal with systems that are vibrating about a configuration of stable equilibrium, that is, a configuration in which the system can remain permanently and stably at rest. We consider now certain types of systems like this. We are particularly interested in systems for which the energy is small enough to ensure that the system does not depart appreciably from the equilibrium configuration. NATURAL MODES OF MOTION A Lagrangian system is said to be harmonic with respect to the set of coordinates q if the Lagrangian in terms of the set of coordinates q is of the form L=2 2,¾¾¾¾ " iS/SA?^, where the matrices imij\ and ikij\ are real> constant, symmetric, and positive definite (see Appendix 12). In this section we are interested in the motion of harmonic systems. We summarize the results in the following theorem. Theorem 1. Consider a Lagrangian system of / degrees of freedom that is harmonic with respect to the set of variables q. Let the Lagrangian be given by L = 2 2 S^#r \ 2 SM* (1) '■ j ' j Such a system has the following properties. a- There is a set of possible modes of motion in which the point in the configuration space q representing the instantaneous configuration of the system oscillates in simple harmonic motion about the origin along a straight line passing through the origin. These modes of motion are called natural modes of motion. The directions in q-space of the straight 665
666 Vibrating Systems lines along which such motion is possible are called modal directions. A vector c whose direction is a modal direction is called a modal vector. In a natural mode of motion all the particles in the system are oscillating with the same frequency and are either in phase or 180° out of phase with one another. b. Associated with each modal direction there is a single angular frequency of oscillation to, called a natural or modal frequency. The value of a particular modal frequency must be one of the / positive roots of the equation \Kj -^1=0 (2) where \ktj — co2raJ is the determinant of the matrix [k^ — ufanX We label the/positive roots of Eq. (2) as co-,,¢03, . . . , w/> respectively. c. If we substitute a particular one of the roots of Eq. (2) into the equation 2(*ff-«Hte=° (3) J then any vector c whose components satisfy the resulting set of equations will be a modal vector with which the given frequency is associated. We label a modal vector associated with the frequency co- as c-. If the frequency co- is a unique frequency, Eqs. (3) will determine a unique one dimensional subspace within which c- must lie; that is, the vectors will all lie along a single direction. If the frequency co- is one of a set of s identical frequencies, Eqs. (3) will determine a unique s dimensional subspace within which c- must lie. d. A set of modal directions is said to be linearly independent if the corresponding modal vectors are linearly independent. From the set of all modal directions it is always possible to choose a set of / linearly independent modal directions. proof: The motion of the system is determined by Lagrange's equations A dt M*(q,q,Q1 8LM''>,q (4) Substituting Eq. (1) in Eq. (4) we obtain J J We assume a solution of the form qj = Cjcos(o)t + <j>) (6) These are the equations of a point in q-space that moves on a line that passes through the origin and is executing simple harmonic motion about the origin. The angular frequency is co, the amplitude is (2,c?/)1/2, tne direction of the line along which the oscillations take place is the direction of the vector c whose
Natural Modes of Motion 667 coinponents are Cj, and the phase of the oscillation is <J>. If we substitute Eq. (6) in Eq. (5) we obtain - co2 2 m^cos^t + <J>) + ]>) kyCjCos^t + ¢) = 0 (7) J J pividing Eq. (7) by cos(co/ + <f>) we obtain 2(*</-"Hb=0 (8) j The set of Eqs. (8) constitutes a set of simultaneous equations for the constants c.-. This set of equations will have a solution if and only if 1^-6)^-1=0 (9) It can be shown that if the matrices [k^] and [m^] are real, symmetric, and positive definite, there will be / real positive roots of Eq. (9). Given a particular root co? the components of the vector c-r can be obtained by letting o) = o)j. in Eq. (8) and solving for the c-. The proof of the rest of the theorem follows immediately from the theory of simultaneous equations. The computations involved in solving for the modal frequencies and modal vectors can usually be considerably condensed by exploiting a few simple shortcuts. For example it is frequently possible to reduce either the matrix [ky] or the matrix [m{j] to a simple numerical matrix, or the product of a numerical matrix and some other factor, by the introduction of new generalized coordinates q- = a^i, where the a{ have been chosen to effect the desired reduction. Along the same lines the following corollary is often helpful. Corollary 2. If we let and [Mg] = a[miJ] (10) M-/W (11) where a and (3 are arbitrary constants chosen to make [MfJ] and [KfJ] as simple as possible, and if AT, A2, . . . , A^ are the/positive roots of the equation l^.-AM^O (12) then the modal frequencies can be obtained from the equations K)2=Af(-f) (13) and the modal direction associated with u- can be determined from the equations ^(Ky-ArM^c^O (14) J
668 Vibrating Systems proof: The set of equations 2(fy-«S>;=0 (15) J is equivalent to the set of equations J L )am0 Cj-0 (16) If we let Mtj = am^., K(j = f}kij9 and A = oi2(f3/a) we obtain 2(^/ -A^H=° (17) These equations will have a solution if and only if Eq. (12) is satisfied, and the solution corresponding to a particular value of A will be the same as the solution to Eq. (2) for the corresponding value of to. The following corollary is useful in the determination of the natural frequency associated with a known modal direction. Corollary 3. If we are given a particular modal vector c, the corresponding modal frequency can be determined from the relation co2=|M (18) proof: The result above follows directly from Eq. (3). note: Frequently one can determine one or more modal directions by inspection of the system. When this is possible we can then use Corollary 3 to determine the corresponding modal frequencies. This in turn provides us with one or more roots of Eq. (2), hence appreciably simplifies the problem of determining the remaining modal frequencies and modal directions. There are not a great number of real systems for which the Lagrangian can be written in the form demanded by Theorem 1. However, if a system is moving under the action of a force that is derivable from a potential having a minimum at some point, and if the total energy of the system is such that the system is confined to a small region in the neighborhood of the minimum, then it is possible to choose generalized coordinates for which the Lagrangian is approximately of the desired form. This result is stated in the following theorem. Theorem 4. Consider a Lagrangian system with f degrees of freedom, for which the generalized potential is a function of the configuration only and has a minimum value for a particular configuration.
Natural Modes of Motion 669 jf we choose a set of generalized coordinates q = quq2, • • . , qj such that: a. The kinetic energy function is of the form ^(q>q) = :l22«//(q)<H (19) where at- = ajt. From Theorem 2 in Chapter 48 this will be true if the equations for the transformation between the set of generalized coordinates q and the set of Cartesian coordinates x do not contain the time, b. The minimum value of the potential energy which we assign the value zero occurs at q = 0, that is, |>(q)l . . = F(0) = 0 L V^/J minimum V / (20) Then for small displacements from the configuration q = 0, the Lagrangian function is given approximately by L(q> 4) = -i 2 2 mMr 2 2 2 *w < j < j where m = *»- 92^(q>q) 92^(q) Jq = 0 92r(0,q) V°) (21) (22) (23) q = 0 and the matrices [/n„] and [/c„] are real, symmetric, and positive definite. Hence the system is harmonic with respect to the coordinates q. proof: If we expand aJq) and F(q) in a Taylor series about the configuration q = 0 we obtain «i,(q) = °9(0) + 2 k H(q) 3¾ ^(q) = v(0) + 2 3F(q) 3* q=o ' y But F(0) = 0 and since V(q) is a minimum at q = 0, it follows that dV(q) H = o 9k+ , = 0 3^(q)" 39,3¾ q = 0 3 that (24) • (25) (26) (27) q = 0
670 Vibrating Systems Thus if we approximate ay(q) and V(q) by the first nonvanishing terms in the expansions we obtain ^(q)«^(Q) 92^(q) 3¾ 3¾ 4¾ (28) Jq = 0 It follows immediately that the Lagrangian is given approximately by Eq. (21). By virtue of their definitions the m^ and the k^ are constants, and m-. = ^ and ktj = kjt. Since the kinetic energy is always positive, it follows that [mJ is positive definite; and since the potential energy is zero at q = 0 and greater than zero for other points in the neighborhood of q = 0, it follows that [kr] is positive definite. This completes the proof of the theorem. note: The behavior of a system for which the kinetic energy is given by T = 2 SiSyaiy(fl)*?y + *i(<0 and the potential energy is given by V = <J>2(q) is identical with the behavior of a system for which the kinetic energy is given by T = 2 2/2^(¾)¾¾ and the potential energy is given by V = <j>2(q) - <h(q). It is sometimes possible to use this equivalence to handle problems that do not quite fit the conditions of Theorem 4. NATURAL COORDINATES If a system is harmonic with respect to a set of coordinates q, it is possible to choose a new set of coordinates that reduces the motion of the system to the motion of a set of independent harmonic oscillators. This result is contained in the following theorem. Theorem 5. Consider a Lagrangian system of / degrees of freedom that is harmonic with respect to the set of coordinates q. Let the Lagrangian of the system be given by < j < j Let cT,C2, . . . , cj be a set of / linearly independent modal vectors in the configuration space q and coT,co2, . . . , coj the corresponding modal frequencies. Let us define a quantity q-r, which we shall refer to as the natural coordinate associated with the modal vector c^, as follows: ?r=SS <limijCrj= 2 2 Cri^ij% (30) '' j * J
Natural Coordinates 671 0r equivalently in matrix form as follows: q? = [q\ •••?/] l[Cr\'"CTf\ m, m, m, yi m, m. mf lff m, m, V 9\ (31) Then the equation of motion for qf is qf + co^ = 0 (32) which is the equation of motion of a simple harmonic oscillator. It follows that if we choose as generalized coordinates the set of natural coordinates qx, fy...,qj then: a. The motion of each of the natural coordinates is independent of the motion of the other coordinates. b. A motion in which the natural coordinate qf is varying but the remaining natural coordinates are not corresponds to the natural mode motion associated with the vector c^. c. The modal directions in the q configuration space are along the coordinate axes. Finally, since the complete motion of the system can be described in terms of the set of coordinates q-x,q^ • • • , qj and each of these coordinates corresponds to a particular natural mode of motion, it follows that the complete motion of the system can be considered to be a superposition of natural modes of motion. proof: The modal frequency u-r and the components of the modal vector c-r satisfy the equation j K we multiply Eq. (33) by qi and sum over i we obtain From Eq. (5) we have i i since kjj and nty are symmetric, we can rewrite Eq. (35) (33) (34) (35) (36)
672 Vibrating Systems Substituting Eq. (36) in Eq. (34) we obtain 2 SWi+ "£2 2™//%?/ = o 1 i If we define < 7 then Eq. (37) can be rewritten q? + <$q? = ° The remainder of the theorem follows immediately from this result. (37) (38) (39) The transformation from the set of generalized coordinates q to the set of natural coordinates q is given in matrix form by the equation (40) |>- ••#]=[?! •• or equivalently ' T\' .i. ~C\\ 9> •<?/] ""»11 mjx ■ ■ ■ CT/1 9/J r«i„ Lm/i mi/l m/f\ \c-u [<nf w'/l mff\ \i\~ [^. c/i" cff. (41) The transformation from the coordinates q to the coordinates q can be obtained by inverting these equations. Doing this we obtain [9\ •••?/]=[?!•• -qj] or equivalently if YYly mf ' cTl • • c/i" _cy "' c/f. ■ »»i/ • mff_ - -i "«n • w 1/ _ "»/i ' " w//. cn ■ ■ ■ ci/ .91 •• 9/. -i "?t" .*. (42) (43) As long as the inverses of the matrices [m^] and [c-ri] are not too difficult to obtain, we can readily transform from the set of coordinates q to the set of coordinates q. However, if the system has a large number of degrees of freedom, determination of the inverses may be very tedious. We will find that if we are careful in our choice of modal vectors in such cases, the problem can
Normal Coordinates 673 be alleviated to a degree. In the next section we consider a particularly convenient choice of modal vectors. NORMAL COORDINATES By proper choice of the magnitudes of the modal vectors, and by proper choice of their directions in the case where not all of the modal frequencies are distinct, it is possible to obtain results a little neater than those obtained in the preceding section. Theorem 6. Consider a Lagrangian system of / degrees of freedom that is harmonic with respect to the set of coordinates q. Let the Lagrangian of the system be given by L = { 2 2 miMj ~ 2 2 2 kyqflj (44) < j < j Let Cj,C2, . . . , Cj be any set of / linearly independent modal vectors and (0f,6>2, • • • , w? the corresponding modal frequencies. a. If the u? are all distinct, the modal vectors satisfy the orthogonality condition SS«j^-0 '** (45) ' j If the co-r are not all distinct, it is always possible to choose the modal vectors such that Eq. (45) is true. b. If we are given a set of modal vectors Cj,C2, . • . , Cj satisfying the orthogonality condition (45), then the modal vectors c^Cf, . . . , cy defined as follows: c;=7 " TU2 (46) [2i2j^ijCrlCff\ satisfy the orthogonality conditions 2 2¾¾¾ = 8rs (47) '' j 2 2^//^^ = ^ (48) j r c- The set of natural coordinates corresponding to the set of orthonorma- lized modal vectors Cf,Cf, . . . , Cf are designated q\,q^ • . . , q/ and are called normal coordinates, that is, 9r= 2 2 qWfrj = 2 2 cHmyqj (49) /- J < J
674 Vibrating Systems or equivalently in matrix form <7; = [<7i •••?/] E[>i * * * Cff] m. m} m, yi mt lff m. m. m '/i mt lff ?i 9/ (50) Given the normal coordinates q the generalized coordinates q can be found from the relation ?/ = 2 cH<li r or equivalently in matrix form ?, = [ciV9] 9\ If d. In terms of the normal coordinates q, the Lagrangian is (51) (52) (53) proof: We proceed as follows: a. Let Cf and c^ be two modal vectors corresponding to two different natural modes of motion. Let u-r and co5 be the corresponding modal frequencies. From Eq. (3) it follows that < w j ' 2 Csi( 2 kijC?j - «£ 2 mijCrj) = 0 If we subtract Eq. (55) from Eq. (54) we obtain {">? - ^) 2 2 mijc?iCsj= o < y If r 7^ 5 and co^ 7^ 0¾ then from Eq. (56) it follows that 2 2¾^¾^0 r^s < y li r ^s and co-r = cos we cannot deduce a relation such as Eq. (57) from Eq. (56). However, in this case there is some arbitrariness in the possible directions of c^ and c3 and one can use the Gram-Schmidt orthogonal- ization process (see Appendix 21) to choose a set of modal vectors for (54) (55) (56) (57)
Normal Coordinates 675 which Eq. (57) is true. Let us assume that this has been done. This completes the proof of part a. b If we normalize the vectors Cy,C2, . . . , c? according to the normalization condition (46), then making use of the orthogonalization condition (45) we obtain 2j i 2jj ™ij Cri Csj i J (2/S;«!/Cn^),/2(2i2j»VC«c,y)1/2 2>»vWij- 7^ ,i/2 ' ' 7m = 8rs (58) which gives us Eq. (47). If we now multiply Eq. (58) by c$k and sum over s we obtain 2 Csk 2 2 mijCPiCsj = 2 Csk^rs (59) s i j s which can be rewritten 2 C?i2 2 miJCsjCsk = 2 CrAk (60) ' j s i This can be true for all values of r only if SSm/,^=^ (61) J s This completes the proof of part b. c. If we multiply Eq. (49) by cn, sum over r, and make use of the orthogonality condition (48), we obtain 2 c;/?;= 2 2 2 ^mijcrjcri= 2 ft 2 2 mijc?jcn= 2 fts//= ft (62) r < j r < j r < This completes the proof of part c. d. Making use of Eq. (51) and the orthogonality condition (47) we have T =i 2 2 miMr 2 2 2 ^2 <*,-# 2 ^¾ < y i i r s = i2 2ftft 2 2^//¾¾ r j ' y = !2 2<MA-=i2<72 (63) r s r Similarly r =i 2 2 *«- 2 2 2 ¢¢2 2 V«c*- (64) 1 j r s i j But from Eq. (3) 2 ktjcsj = *>i 2 ^//¾ (65) y y
676 Vibrating Systems hence v = 2 2 S "I** 2 2 ^/, ^/¾ = 2 2 2 wfoft *#* = 2 2 «;V (66) Combining Eqs. (63) and (66) we obtain Eq. (53). This completes the proof of the theorem. NORMAL COORDINATES AND LINEAR TRANSFORMATIONS From the results of the preceding section it is apparent that the normal coordinates are just the set of generalized coordinates that reduce the Lagrangian from the following form ^/2y™/y?/?y - iE/Ey *#?''%'to the form 2ErS/%9/9/~ I'ZiZftfiffiqj- Jt follows that the normal coordinates can be found if we can find the linear transformation q- = ^-a^qj that simultaneously reduces the quadratic form E/Syw//<7/<7y to tne f°rm SS/^J?/?/ an(i the quadratic form 2/Ey^//<7/<7y to tne f°rm SfSy\*%4747- The standard technique for doing this is a three stage process and is considered in Appendix 27. The technique we have presented is essentially a unification of the three stages. EXAMPLES Example 1. Two particles each of mass m are connected as shown in Fig. la. Each of the two outer springs has a spring constant k, and the middle spring has a spring constant ek. The system can undergo motion only in the direction of the line joining the masses. The system is initially held at rest with the left hand particle displaced a distance c to the right of its equilibrium position, and the right hand particle at its equilibrium position. The system is then released. Discuss the subsequent motion. Consider in particular the case in which c< 1. > i ,. k m €k m .. 9 i J ! I I I I A I I I I Con& (b) Fig. 1
Examples ¢77 Solution. Let x and y be the respective displacements of the particles from theif equilibrium positions. The kinetic and potential energies of the system are T=\mx2 + \my2 (67) V=\k(\ + i)x2 + \k(\ + e)y2- kexy (68) It follows that A:(l + e) -ke -ke *(! + €) m 0 0 m imij] = We use Corollary 2 and let M a = -i- and £ = | m k 1 0 0 1 Doing this we obtain The roots of the equation \Ky - AM^| = M = 1 + e -€ -€ 1 +€ (l + c)-A -€ -€ (l + e)-A = 0 are Since AT = 1 and A2 = 2e + 1 co2 = A4 =AA it follows that the modal frequencies are k \x/1 (26+1)-^ 1/2 (69) (70) (71) (72) (73) (74) (75) (76) COf — ( — ) u>2 V m J The components of the modal vectors c must satisfy the equations (i + e)-A -€ lr^i ro -€ (1 + e)- A c2 I 0 " in the equation above we set A = AT = 1, cx = cTl and c2 = cl2 we obtain [cTl c-l2]=[a a] (77) where a is an arbitrary constant. Similarly, if we set A = A2 = 2e + 1, c, = c2U and c2 = <- we obtain >i ^22] = [> -*] (78)
678 Vibrating Systems where b is an arbitrary constant. The natural coordinates q^ and q^ corresponding to the modal vectors cy and c^ are m 0 q-x = [a a] q-2 = [b -b] 0 m m 0 x y\ ma(x +y) = mb(x — y) 0 m Since a and b are arbitrary, for simplicity we choose them as follows, 1 a = b = 2m (79) (80) (81) Substituting Eq. (81) in Eqs. (79) and (80) we obtain ?T = ±(*+y) (82) <n = Hx-y) (83) If we solve Eqs. (82) and (83) for x and y we obtain * = ?T + ?2 (84) y = T\ ~ ?2 (85) The solutions to the equations of motion for q\ and q^ are <7y = ;4yCOS(c0y/ + 4>y) (86) <72 = A^cos^t + fa) (87) It follows that the solutions to the equations of motion for x and y are X = ^4yCOS(c0y/ + <J>y) + A^COS^Cljt + ¢5) (88) J = ^4yCOS(c0y/ + <J>y) — A^COSfjU^t + ¢5) (8^) At t = 0, x = c, 7 = 0, x = 0, and y = 0. It follows that at t = 0, #T = c/2, <72 = c/2, ^y = 0, and qi = 0. Using the boundary conditions above to determine A J, A^, <J>T, and fe we obtain A^ = c/2, A^ = c/2, <J>y = 0, and 4¾ — 0- Hence (90) (91) (92) (93) (94) (95) * = £< ■y cos(coT;) + -y COS(Wj*) j = I cos(uTr) - I cos(u^) If we use the trigonometric identities cos ,4 + cos B = 2cos B ~A cos B + A cos A — cos B = 2 sin —^— sin 2 2 then Eqs. (90) and (91) can be rewritten x = ccos[ ^(co2 — <o-f)j]cos[^(w5 + «t)'1 7 = c sin[ £ (w5 - wT)*]sin[ \ (w2 + uT)f ]
Problems <779 k§Mlfr<#fr MiNi^^i Fig. 2 If e < 1 then from Eq. (75) we have <o5-WT [(2e+l)(k/m)y/2-(k/my/2 [(2e+!)(*/»»)] ' +{k/m) \'/2 [/2 If we define -M 1/2 we can write Eqs. (94) and (95) : CCOS si — lcos(co0/) y^csmi —— sin(co0r) (96) (97) (98) (99) (100) It follows that the x and y motion consists of oscillations of frequency co0 whose amplitudes are sinusoidally modulated. The situation is illustrated in Pig. 2. There is a slow transfer of energy back and forth between the two basses. At time / = 0 the mass x is oscillating with frequency co0 and amplitude c and the mass y is at rest. Slowly the amplitude of the oscillations °f mass x decreases, and the mass y begins to oscillate. At time t = 7r/eco0 the mass x is at rest and the massj is oscillating with frequency co0 and amplitude c. PROBLEMS • Two uniform rods AB and EC are freely hinged at B and hang from a smooth pivot A. The mass and length of AB are 2m and a, and of BC
680 Vibrating Systems are m and 2a. The system performs small oscillations in a vertical plane about its position of equilibrium. a. Determine the modal frequencies. b. Show that if the system is oscillating in one natural mode the inclinations of the rods to the vertical are in the ratio —1:1, and in the second natural mode they are in the ratio 2:1. 2. A hemispherical bowl of radius 2b and mass Am rests on a smooth table with the plane of its rim horizontal. Within it lies a perfectly rough solid sphere of radius b and mass m. The system is set in motion in such a way that both the center of the sphere and the center of the bowl remain in the vertical plane in which they were when the system was in equilibrium. Prove that the modal frequencies for small oscillations are coT and C05 where cof and wf are the roots of the equation \56b2x2 — 260bgx + 75£2 = 0. 3. A smooth circular wire, of mass Sm and radius a, swings in a vertical plane, being suspended by an inextensible string of length a attached to one point of it; a particle P, of mass m, can slide on the wire. If the system is released from rest with the ring in its equilibrium position and P is displaced through a small angle /3 from its equilibrium position, show that at time t the angle that the string makes with the vertical is (j8/21)[cos(/if/2V2) - cos(/if)], where n2 = 3g/a. 4. A light elastic string is stretched between two fixed points A and B. The points A and B are a distance Aa apart and the tension in the string is t. A particle of mass 3 m is attached to the midpoint C of the string and particles of mass Am are attached to the midpoints of AC and CB. When the system is at rest small transverse velocities u are suddenly imparted in the same direction to both the particles of mass Am. Prove that the displacement of the particle of mass 3 m at time t is given by (4w/5co)[V6sin(co//V6) — sinco/], where to2 = r/'ma. 5. Three particles of masses m, M, and m, respectively, are free to move along a straight line. The middle particle M is joined to each of the outer particles mbya spring of spring constant k and unstretched length a. The particles are initially at rest. A fourth particle, of mass m, moving with speed v along the line formed by the three particles, strikes one of the end particles and rebounds elastically. Determine the subsequent motion of the particle of mass M. 6. Three heavy beads of masses m, m, and M, respectively, are free to slide on a smooth circular wire loop placed in a horizontal plane. The beads are connected by three straight identical springs of negligible mass and spring constant k. When the masses are distributed symmetrically the springs are unstressed. Determine the modal frequencies and modal directions, and find a set of natural coordinates.
Problems 681 7# A uniform equilateral triangular lamina ABC of mass m is supported in a horizontal position by three equal springs at its vertices. The springs are constrained to remain vertical, the center O of the lamina is constrained to remain on a vertical line, and the vertex A of the lamina is constrained to remain in a vertical plane containing the center of the lamina. Show that for small oscillations the modal frequencies of the system are the same as the frequencies of a set of simple pendula of lengths a, a/4, and a /4, where a is the amount each of the springs is compressed when the system is in equilibrium. If the system is set in oscillation by a small downward impulse I applied at A, show that the displacement of either of the other vertices is (I/mco)sincot (1 — 4cosco/), where co2 = g/a. 8. A thin uniform square plate of side la is suspended by two light inextensible strings, each of length 4a/3, which are attached to the ends A, B of an edge and to two points C, D at the same level at a distance la apart. Describe the natural modes of oscillation and prove that the lengths of the equivalent simple pendula are 4a/3, 4a/9, 1(1 ±^3)a/3. 9. Find the normal coordinates for a system for which the Lagrangian is L = \{x2+y2) - {{a\x2 + to2/) + axy. 10. Find the normal coordinates for a system for which the Lagrangian is L = \{mxx2 + m2y2) + /3xy - \(x2 + y1). 11. A uniform rod BC of length la is fastened at B to a light elastic string AB that is fastened at A to a fixed point. The unstretched length of the string is a and its modulus of elasticity is six times the weight of the rod. Show that the equivalent simple pendula for the natural modes of motion in a vertical plane are of lengths a/6, a/6, and la/3. Using as generalized coordinates x, the increase in the length of the string relative to its equilibrium length; 0, the angle the string AB makes with the downward vertical; and <f>, the angle the rod BC makes with the downward vertical, obtain a set of normal coordinates. 12. One end of a uniform rod of mass m and length la is driven with a constant angular speed Q in a horizontal circle of radius b. Prove that when the motion is steady the rod lies in the vertical plane through the center of the circle and makes angle a with the vertical, where a is determined by the equation tt2(k2 + ab cosec a) = ag sec a and mk2 is the moment of inertia of the rod about an axis perpendicular to the rod and through one end. Show that if the system is disturbed from its steady state motion, the modal frequencies are the roots of the equation (kWsina - Q2ab)(kWsma - ®2ab - £22A:2sin3a) = 4ft2A:Vsin2acos2a
66 Symmetrical Vibrating Systems INTRODUCTION The determination of the natural modes of motion and the corresponding modal frequencies of a harmonic system can be considerably simplified if the system has some kind of symmetry. In this chapter we show how one can use symmetry arguments to aid in the determination of the natural modes of motion and the modal frequencies. Much can be accomplished on an elementary level, but the full exploitation of the symmetry properties is most elegantly achieved by using group theory. (See Appendices 28-30) NATURAL MODES OF MOTION In this section we review briefly the results obtained in the preceding chapter and introduce some concepts and notational changes that are used in this chapter. Let us consider a system of / degrees of freedom for which the Lagrangian in terms of a set of generalized coordinates quq2, •••>?/ *s giyen by L = {2 2 mtMr {2 2 *w (1) * j * j where the quantities my and ktj are real constants and the matrices [wiy] and [ky] are symmetric and positive definite. Such a system is said to be harmonic with respect to the set of generalized coordinates quq2, • • • , #/• If we (1) designate a particular configuration by the notation (#p ?2> • • • > ?/)> (¾ define the sum of two configurations (q\,q2, . . . , q'f) and (q",q'i, ..., qf) as the configuration (q\ + q'{,q'2+ q'{, . . . , q'f + q'/\ and (3) define the product of the real number a and the configuration (qx,q2, . • • » ?/) as the configuration (aql,aq2, ..., aqfi, then the set of all configurations (<7i>#2> • • • » q/) together with the operations above constitute a real vector space, which we shall call the configuration space C. 682
Natural Modes of Motion 683 In the preceding chapters we have used the notation q to designate a particular vector (q],q2, • • • , ?/) in the configuration space C. In this chapter we use the notation [q) rather than q to represent a vector in the space C. The set of vectors [e)x,[e)2, ..., [e)f defined by the relation [q) = 2/?/[*)/ constitutes a complete set of basis vectors for the space C. We refer to this set of basis vectors as the basis S. As shown in Appendix 29 a vector in a vector space can be represented by a column matrix. The column matrix corresponding to the vector [q) with respect to the basis S is the column matrix [qt), that is, it is the column matrix whose /th element is the value of the /th generalized coordinate. From the results of the preceding chapter we know: 1. There are certain directions in configuration space called modal directions. If a harmonic system is released from rest at time t = 0 in a configuration [c), where [c) is a vector lying along a modal direction, the resulting motion is described by the equation [q) = cos cot [c). Such a mode of motion is called a natural mode of motion, and the frequency to, which is unique for a given modal direction, is called a modal frequency. A vector [c) lying in a modal direction is called a modal vector. 2. The modal frequency co associated with a particular modal direction will be one of the roots of the equation l*j,-«Hl = 0 (2) With each frequency co there is at least one modal direction. 3. If we are given a particular modal frequency w, the components Cj of the corresponding modal vectors [c) with respect to the basis S are obtained by solving the set of simultaneous equations 2(*(/-"Hb=0 (3) J 4. If we are given a particular modal vector [c), we can use the relation co2 = (4) to obtain the corresponding modal frequency. This equation is valid for any choice of i. Sometimes a particular choice of / leads to the indeterminate form 0/0. When this happens one simply chooses a different /. In the preceding chapter the first step in obtaining the natural modes of Motion was to obtain the modal frequencies; then the modal directions were found. In this chapter we show that it is possible to use symmetry arguments to determine a great deal about the modal directions and the modal frequencies without ever knowing the exact values of the modal frequencies.
684 Symmetrical Vibrating Systems GEOMETRICAL REPRESENTATION OF THE CONFIGURATION A harmonic system may be considered to consist of a set of point particles that oscillate about an equilibrium configuration. If we let r(7) be the vector displacement of particle i from its equilibrium position, the configuration of such a system is known if we know the equilibrium configuration of the system and the values of r(7). We can thus geometrically represent the configuration of the system by drawing a diagram of the system in its equilibrium configuration, then attaching to each particle site i the directed line segment that represents the displacement of particle i from its equilibrium configuration. A particular configuration of a system of three particles that in equilibrium form an equilateral triangle is represented in this fashion in Fig. 1. The above geometrical representation of the configuration is very useful when the system we are dealing with is symmetric. SYMMETRY OPERATIONS A dynamical system that is in static equilibrium is said to be symmetric if there exists a rotation of the system as a whole about some axis, or a reflection of the system as a whole in a plane, that leaves the system in a configuration in which the mass distribution and the force distribution are the same as in the original configuration. As an example consider a system consisting of three identical particles a, b, and c, which are confined to a plane and in equilibrium form an equilateral triangle. If this system is rotated in a counterclockwise fashion through an angle of 0, 27r/3, or 47t/3 about an axis perpendicular to the plane of the triangle and passing through its center, or is reflected in a plane that is perpendicular to the plane of the triangle and bisects angle a, angle b, or angle c, the resulting configurations have the same mass distribution and force distribution as the original distribution. If in Fig. 2 the configuration 2.1 is the original configuration, the operations above lead respectively to configurations 2.1-2.6. / \ / \ / \ / \ \ \ &-— Fig. 1
Symmetry Operations <S85 / \ / \ (gj-----fe (§5 te> (2.1) (2.2) / \ / \ / \ / \ / ^ / v (2.3) (24) / \ / \ / / \ (gj. &> <§} © (2.5) (2.6) Fig. 2 If we designate a particular configuration as the reference configuration, each symmetry operation can be characterized by the configuration it produces when operating on the reference configuration. We consider two operations that produce the same configuration to be equivalent. The set of all nonequivalent symmetry operations for a particular system forms a group. We use the notation G to represent an arbitrary symmetry group, and the notation g to represent an arbitrary element of the group G. It is sometimes useful to apply one of the symmetry operations g in a group G to the displacement vectors r(7) rather than to the system itself. For example suppose the system discussed earlier is in the configuration represented by Fig. 3<z. If we rotate the displacement vectors in a counterclockwise sense through an angle of 2ir/2> about the center of the triangle while leaving the equilibrium configuration unchanged, we obtain the configuration represented by Fig. 3b. / \ / \ / \ / \ / \ f (a) (b) Fig. 3 V
686 Symmetrical Vibrating Systems OPERATOR REPRESENTATIONS OF THE SYMMETRY GROUP Consider a system in an arbitrary configuration represented by the vector [q) in the configuration space C. If we subject the displacement vectors r(/) for this configuration to a symmetry operation g while leaving the equilibrium position of the system unchanged, the resulting set of displacement vectors, which have in general changed both their location and their direction, will represent a new configuration of the system. Thus the symmetry operator g can be used to associate with every vector [q) a new vector, hence can be used to define an operator on the space C We designate the operator on C corresponding to the symmetry operation g by the notation [g]. The set of operators so defined constitute an operator representation [G] of the group G. THE NATURAL MODE DECOMPOSITION OF CONFIGURATION SPACE Consider a symmetric harmonic system whose configuration is represented by a vector [q) in a configuration space C. Let g be one member of a group G of symmetry operations for the system and let [ g] be the corresponding operator in the operator representation [G] on the space C. If [#(0) represents a possible motion of the system, [g][q(t)) also represents a possible motion of the system. Thus if we operate on a modal vector [c) with the operator [g], the resulting vector [g][c) must also be a modal vector, and the frequency associated with the modal vector [g][c) must be the same as the frequency associated with the modal vector [c). It follows that the subspace of C that contains the set of all modal vectors associated with a particular modal frequency must be an invariant subspace with respect to the group [G] of operators [g]. With each of the distinct modal frequencies there is associated an invariant subspace of C. We designate the distinct_frequencies as co\co2, . . . and the corresponding invariant subspaces as C1, C2, . . . . The subspaces C" are disjoint, and since any vector in C can be written as a linear combination of modal vectors, the direct sum of all the subspaces Cl is equal to C. DEGENERACY If the subspace C associated with a particular frequency col is of dimension n, where n is greater than one, we say that the frequency co' is n-fold degenerate or is degenerate with a multiplicity n. If the subspace C" associated with a particular degenerate frequency oil not only is invariant with respect to [G] but is also irreducible with respect to [G], we say that the degeneracy is an essential degeneracy. The occurrence of an essential degeneracy is a necessary consequence of the symmetry of the system and can be broken only by breaking the symmetry.
The Basic Problem 687 If the subspace Cl associated with a particular degenerate frequency co' can be decomposed into m irreducible subspaces of dimensions w,,«2, . . . , respectively, co' is said to have an accidental degeneracy of multiplicity m and essential degeneracies of multiplicities nl9n2, . . . , respectively. The occurrence of an accidental degeneracy depends on the values of the elements in the matrices [ktj] and [m^] and can be broken without altering the symmetry by changing the values of the forces between the particles in the system. Group theory cannot predict the occurrence of accidental degeneracies, and when an accidental degeneracy does occur it does not give rise to any formal difficulties in the group theoretical treatment. In the following sections we therefore act as if there were no accidental degeneracies, and for the sake of simplicity we use the term "degeneracy" to mean an essential degeneracy. If an accidental degeneracy occurs it simply means that two frequencies we are treating as distinct turn out to be equal. If there are no accidental degeneracies, each of the subspaces Cl is irreducible; hence the set of subspaces C" represents a decomposition of the space C into irreducible subspaces. Since the decomposition of C into irreducible subspaces is not necessarily unique, there may be other decompositions of C, but barring accidental degeneracies the decomposition above is the only one in which all the vectors in a particular subspace are modal vectors with which the same frequency is associated. Thus the C" constitute a unique and complete set of irreducible subspaces in C, but they do not represent the only complete set of irreducible subspaces in C. note: It has been implicitly assumed in the discussion in this section and the preceding section that the group G of symmetry operations contained all the symmetry operations for the system. We could however just as well have used some subgroup G of G. If we had used G, then the subspaces Cl which are invariant with respect to [G] would have been subspaces of the subspaces C[ which are invariant with respect to [G]. We would then find that if Cl and Cj were two different subspaces contained within the same subspace Ck, the frequencies associated with Cl and CJ would be equal. The equality of these frequencies would however not be an accidental degeneracy but would be an essential degeneracy arising from the suppressed symmetry operations. THE BASIC PROBLEM From the point of view developed in this chapter, the problem of determining the natural modes of a harmonic system consists in determining the subspaces C- Once a particular subspace C is found, we know that every vector in C is a modal vector associated with the same frequency co7, and the value of the frequency co' can be determined by simply choosing an arbitrary vector in C and substituting it in Eq. (4).
688 Symmetrical Vibrating Systems In the following sections we see how one can use group theory to aid in the determination of the subspaces C. THE GROUP THEORETICAL DECOMPOSITION OF CONFIGURATION SPACE From the theory of group representations (Appendix 30) we know that it is possible to decompose the configuration space C into a set of irreducible subspaces with respect to [G], and with each of these subspaces there is associated one of the irreducible representations of the group. We use the notation Cij to designate theyth subspace that is associated with the irreducible representation /, and the notation Cl to designate the direct sum of all the irreducible subspaces with which the irreducible representation / is associated. For a given [G] the subspaces C" are unique, but the decomposition of C into irreducible subspaces is not unique. It follows that every irreducible subspace must be within one of the subspaces C". The natural mode decomposition of the space C into irreducible subspaces Cl represents a particular decomposition of C into irreducible subspaces. From the results of the preceding paragraph it follows that each of the subspaces C" is contained within one of the subspaces CJ. To bring out the relation between the C" and the CJ we use from now on the notation Cil,Ci2,C3, . . . to designate the particular subspaces CJ that are contained within C". It follows that d= c"®c*® • • • e c^ (5) where m is the numberj)f irreducible subspaces contained within C. If there is only one subspace C} = Cl contained within C", then Cl = Cl and all the vectors lying within C must be modal vectors with which the same modal frequency cil is associated. If there are m irreducible subspaces CiJ contained in C", where m > 1, any vector in C can be written as a linear combination of m modal vectors, one from each of the subspaces Cij contained in C. A subspace C" containing m irreducible subspaces will thus be associated with a particular set of m modal frequencies. From the analysis above it follows that if we can decompose the space C into the subspaces C", we have taken a step toward separating the space into the subspaces C7, hence to obtaining the modal vectors. THE NUMBER AND MULTIPLICITIES OF THE MODAL FREQUENCIES Although the decomposition of a given configuration space into subspaces that are irreducible with respect to a group of operators is not necessarily unique, the number and* dimensions of the irreducible subspaces contained in each decomposition will be the same for all decompositions. Furthermore from group theory we know that it is possible to determine the number and
Use of Character Tables in the Determination of Natural Modes 689 dimensions of the irreducible subspaces without actually carrying out the decomposition. It follows that without actually determining the modal vectors we can determine the number of irreducible subspaces 0 contained in C and also the dimensions of the subspaces 0; hence we can determine the number of modal frequencies and their degeneracies. Suppose for example that from the analysis of the group G of symmetry operations for a given system we have determined: (1) the number p of irreducible representations 1,2, . . . ; (2) the dimensions n\n2x . . a. of the irreducible representation 1,2,...; and (3) the numbers m\m2, . . . of subspaces in the decomposition of C that are associated respectively with the irreducible representations 1,2, . . . . It then follows that there will be m] + m2 + ' ': +.m^ different modal frequencies, ml of which will have a multiplicity n\ m2 of which will have a multiplicity n2, . . . , and m? of which will have a multiplicity of nK note: Some difficulties arise when a complex irreducible representation that is not equivalent to a real representation occurs. In this case the decomposition of the real vector space C spanned by the set of all modal vectors is not identical with the decomposition of the corresponding complex vector space C spanned by the set of all modal vectors. In resolving this difficulty we must keep in mind that the theorems concerning irreducible representations were derived on the assumption that the vector space was complex. If the irreducible representation [G]' is complex, then {[G]'}* the complex conjugate of [G]', will also be an irreducible representation. In the decomposition of C, for every subspace CiJ associated with [G]', there will be a complex conjugate subspace {CiJ}* associated with {[G]'}*. In the decomposition of C however there will be only one irreducible subspace Cij corresponding to the two subspaces Cij and {ClJ}* and this subspace will be the real^subspace contained within the direct sum of the subspaces Cij and {ClJ}*. Thus if theA dimension of each representation in a pair of complex representations [G]' and {[G]'}* is nl and if the number of subspaces in C associated with each of the representations is m', then the number of sub- spaces in C associated with the pair is m' and the dimension of each of these subspaces is 2nl. Consequently for every pair of subspaces of C associated with a pair of complex irreducible representations there will be associated one modal frequency and the degeneracy of this frequency will be twice the dimension of either of the representations. USE OF CHARACTER TABLES IN THE DETERMINATION OF NATURAL MODES *n the preceding section we have seen that we can determine a great deal about the decomposition of C from a limited amount of information concern-
690 Symmetrical Vibrating Systems ing the irreducible representations. If however we know not only the number and dimensions of the irreducible representations but also the characters x[g]7 of the operators in the irreducible representations, we can determine not only the structure of the decomposition of C but also the subspaces Ck that occur in the decomposition, where Ck is the direct sum of all the^ irreducible subspaces in C associated with the irreducible representation [G]k. To accomplish this we recall from Appendix 30 that when the projection operator [Pf = 2x*[gf[g] (6) g operates on any vector in C it will project out a vector lying in Ck. By operating with [p]k on each member of a set of vectors that span the space C we obtain a set of vectors lying in Ck. From these vectors we can choose a set of linearly independent vectors that span the space Ck, hence define the space c*. note: Since Ck is invariant with respect to [G], it follows that if we know one vector [q) in Ck we can generate a second vector in Ck by operating on [q) with any of the operators [g]. It is thus possible to obtain a complete set of basis vectors by using [p]k to obtain one or two vectors in Ck, then using the operators [g] to generate enough additional vectors. In any case we can always use one or the other of these methods to obtain a set of basis vectors for the space Ck. Knowledge of the subspaces C" considerably reduces the problem of determining the subspaces Cij\ hence of determining the modal vectors for the system. If the space C" contains only one subspace Cl then Cl = Cl and all the vectors in Cl are modal vectors with which the same frequency is associated. If Cl contains two or more of the subspaces Cij\ we can decompose Cl into the subspaces CiJ by assuming that our system is constrained to remain in C\ then using the ordinary techniques for determining modal frequencies and modal vectors. Since the dimension m of the space Cl will always be smaller than the dimension n of the space C, we have an m dimensional problem to solve to find the subspaces CiJ rather than the n dimensional problem we would have had starting with the whole configuration space C. USE OF THE IRREDUCIBLE REPRESENTATIONS IN THE DETERMINATION OF THE NATURAL MODES The preceding section showed that we can determine a great deal about the decomposition of the space C if we know the character table for the irreducible representations of the group G. If a particular subspace C" contains ml irreducible subspaces each of dimension nl, and if ml or n1 is equal to one, then the procedure used in the preceding section provides us with as much information about the decomposition of C" as we can obtain using projection
Use of the Irreducible Representations in the Determination of the Natural Modes <S01 operators. If however m' and n' are both greater than one, it is possible, if we know not only the character table for the irreducible representation associated with C" but also the irreducible representation itself, to push the decomposition even further. To see this let us suppose we are given the irreducible matrix representation [G ]k and assume that the invariant subspace Ck associated with [Gtj\k can be decomposed into m irreducible subspaces Ck\Ck2, ..., Ckm, where each subspace is of dimension n, and m and n are greater than one. From Theorem 27 in Appendix 30 we know that when the projection operator [pi^u-%[g] (7) operates on any vector in C it will project out a vector that lies in an m dimensional subspace 2)/ that contains a vector from each of the subspaces Ckl. By operating with \p\t- on each member of a set of vectors that span the space C, we thus obtain a set of vectors all lying in the m dimensional subspace 2)/ and can choose from these a set of m linearly independent vectors that span the space 2)/, and hence define the space 2)/. We know furthermore from the results of Appendix 30 that the m dimensional subspace 2)/ will contain not only a vector from each of a particular set of irreducible subspaces Ckl but also a vector from each of any set of irreducible subspaces Ckl. In particular 2)/ will contain one vector from each of the m subspaces Ckl that are contained in Ck. It follows that if we assume that our system is constrained to remain in the space 2)/ and solve for the normal modes of this constrained system, the resulting m modal directions and m modal frequencies will also be modal directions and modal frequencies for the unconstrained system. The m modal frequencies will be the m modal frequencies associated with the m subspaces Ckl contained in Ck, and each of the m modal directions will be in a different one of the m subspaces Ckl. Furthermore if we know one modal vector [c) in a subspace Ckl we can generate the subspace Ckl by operating on [c) with the symmetry operators [<?]• We can thus determine all the subspaces Ckl contained in Ck and the modal frequencies associated with each.' To obtain the same information using the method of the preceding section we would have had to solve an m X n dimensional problem rather than simply an m dimensional problem. It follows that knowledge of the irreducible representations takes us closer to the final solution than knowledge of the characters alone. note: The results of this section were true because the subspace 2)/ contained one modal vector from each of the subspaces Cki. This is not the same as saying that the vectors in Dk can be written as linear combinations of the vectors in Ckl. To see this, consider a nondegenerate two dimensional
692 Symmetrical Vibrating Systems system for which the Lagrangian is \m^q} + \k^q] + \m2q\ + \ k2q\. There are two modal frequencies for this system, co1 = (A:,/m,)1//2 and co2 = (k2/m2)l/2. The subspace C1 associated with co1 is the set of vectors pointing in the 1 direction, and the subspace C2 associated with co2 is the set of vectors pointing in the 2 direction. If however the system is constrained to remain in the subspace D defined by the line ^, = q2 the system will oscillate along the line <7, = q2 with a frequency [(/c, + k2)/{mx + m^]^2. Any vector in D can be written as a linear combination of a vector in C' and a vector in C2, but D does not contain a vector from either C1 or C2; hence we do not expect the frequency and modal direction for the constrained system to correspond to a frequency and modal direction for the unconstrained system. DETERMINATION OF THE MODAL FREQUENCIES WHEN THE KINETIC ENERGY MATRIX IS A SCALAR MATRIX Frequently the kinetic and potential energy functions are of the form T=\m^qf (8) i r-i22M'* (9) ' J or can be reduced to this form by some simple transformation of coordinates. If this is the case there is a simple method of determining the modal frequencies without first determining the modal vectors, and without making use of projection operators. To find the modal frequencies, note first that if we make an orthogonal transformation to a new set of variables q-, that is, if we make a transformation [?/) = [■*//][?/) where then the kinetic and potential energy functions become r«i/»2# (12) i r-*22*w (13) '' J where [%]=[^]r[^][^]=[^]_,[^][^] (14) The kinetic energy function is unchanged by this transformation, whereas the matrix of the coefficients in the potential energy function is subjected to a (10)
Determination of the Modal Frequencies €193 similarity transformation. The transformation is a similarity transformation, not simply a congruent transformation, because of the orthogonality of the matrix [stj\. Now let us suppose that there exists a group G of symmetry operations for the system and that we have, by the technique introduced earlier, defined an operator representation [G] of G on the configuration space C. Let S be the set of basis vectors [e)t with respect to which the components of a vector [q) are qi9 S the set of basis vectors [e)j with respect to which the components of a vector [q) are g^and [gy] and [gy] the matrix representations with respect to the bases S and S, respectively, of an arbitrary element [ g] of the operator representation [G]. The matrices [gy] and [gy] can be shown to be related by the same similarity transformation that connects the matrices [ky] and [kjj], that is, [8f]-M~l[8y]M (15) Since the product of the similarity transformations of two matrices is equal to the similarity transformation of the product of the two matrices, it follows that M%r=[*d^UM>d (16) where [kt]n and [kjAn are the nth powers of the respective matrices [ky] and [ky], and n is some arbitrary positive integer. If we take the trace of Eq. (16) and note from Theorem 6 of Appendix 24 that the trace is invariant under a similarity transformation we obtain •Mf«dM"}-*{[*][**]"} (17) For a particular system, the matrices on the right hand side of Eq. (17) can be found; hence the right hand side of Eq. (17) can be assumed to be a known quantity. If we choose the basis S to be a set of modal vectors for the system, the matrix [ gjj] either will be in the following form or can be put into it by a simple relabeling of basis vectors: 'r i* (18) [Sg] where [g(j]r is the matrix corresponding to the element g in an irreducible matrix representation [Gj-]r and the terms other than those indicated are zero. [or]-
694 Symmetrical Vibrating Systems With the same choice of basis S, the matrix [k-A assumes the form [%]=[^%] = A:11 A:11 •;,2l ,21 ,22 ,22 (19) where the off diagonal terms are zero. The element J<fs falls in the space occupied by the sth occurrence of the submatrix [gjj]r in Eq. (18), and the number of times it occurs is equal to the dimension of the submatrix [gijY- With respect to this basis T={m^qf (20) r=i2W (21) Hence the modal frequencies are given by -'■(S) (22) If we substitute Eqs. (18) and (19) in Eq. (17) and take the trace of the left hand side we obtain (23) Sx[g]W=Tr{[^][^,]"} v where x[gY is tne trace °f tne. matrix associated with the element g in the irreducible representation [G]'. This is essentially the result we need to calculate the modal frequencies. We will however go one step further and put it in a form that is more convenient for computations. Multiplying Eq. (23) by /3n and defining K'J = pk'J (24) (25)
Determination of the Modal Frequencies 4595 where /? is an arbitrary constant chosen to make [ktJ] as simple as possible, we obtain 2xU]WB=Tr{[^][^n (26) V With each of the ICj there is associated a modal frequency uP defined by the equation (^)2=g (27) The only unknown quantities in Eq. (26) are the quantities Kj. By choosing different values of n and different elements g we can use Eq. (26) to generate enough equations to determine all the unknowns Kj\ hence we can find all the modal frequencies. Let us look more closely at what is involved in the determination of the KK Let us suppose that the group has/? classes, hence/? irreducible representations 1,2, . . . , p. Then for a given value of n we can, by using one element g from each of the p classes, generate p independent equations. The p equations corresponding to a particular value of n provide us with p equations in the p unknowns 2y(J^,2y(J^, ..., 2,-( A^)". Now let us suppose there are m, of the p irreducible representations that occur once in the reduced representation, m2 that occur twice, m3 that occur three times, ..., and mr that occur r times, where r is the greatest number of occurrences of any of the/? irreducible representations. It then follows that the/? equations corresponding to n = 1 are sufficient to determine the Kj associated with irreducible representations that occur once, the /? equations corresponding to n = 1, and p — mx of the equations corresponding to n = 2 are sufficient to determine both the Kj associated with irreducible representations that occur once and the Kj associated with irreducible representations that occur twice, and so forth. Proceeding in this fashion we find that to determine all the Kj it is sufficient to have the /? equations corresponding to n = 1, /? — mx of the equations corresponding to n = 2, . . . , and p — mx — m2— • • • - mr_x of the equations corresponding to n = r. note: It is possible to formally solve Eqs. (26) for the quantities 27(^T by multiplying Eq. (26) by x*[gf> summing over g, and making use of the orthogonality condition given in Theorem 21 Appendix 30. If we do this we obtain 2(^)"= \Gr^X*[gYTr{[gij][KiJ]n} (28) J % where |G| is the order of the group. Equation (28) does not significantly reduce the computations involved in finding the A7, but it is interesting from a theoretical point of view.
696 Symmetrical Vibrating Systems EXAMPLES Example 1. Determine the modal frequencies and modal vectors for a system consisting of three identical particles of mass m that are constrained to remain in a plane and are bound together as shown in Fig. 4 by three identical springs of natural length I and spring constant k. Solution. For illustrative purposes we solve this problem using a variety of group theoretical techniques. Let us define directions 1, 2, 3, 4, 5, and 6 as shown in Fig. 4 and choose as generalized coordinates the quantities: in the 1 direction in the 2 direction in the 3 direction in the 4 direction qx = displacement of particle a q2 = displacement of particle a : q3 = displacement of particle b : q4 = displacement of particle b : q5 = displacement of particle c in the 5 direction q6 = displacement of particle c in the 6 direction In terms of these coordinates the kinetic energy T is given by T=\m{q\ + q\ + q\ + q\ + q\ + q\] and the potential energy V is given by 2 \ 2 <]5 + \q6 + V3 (29) (30) "V5 Fig. 4
Examples <897 It follows that [**H ki 6 0 3 -V3" 3 V3 = m 1 0 0 0 0 0 0 2 V3 - 1 -V3 — 1 0 0 0 0 0 10 0 0 0 0 10 0 0 0 0 10 0 0 0 0 10 0 0 0 0 1 3 -VI 3 73 -1 -V3 6 0 3 0 2 VI 3 V3 6 -V3 -1 0 VI -1 -VI -1 0 2 (31) (32) Let axis z be an axis that is perpendicular to the plane of the triangle and passes through the center of the triangle, and planes a, b, and c planes that are perpendicular to the plane of the triangle and bisect angles a, b, and c, respectively. Then the following set of nonequivalent operations constitutes the symmetry group G for the system: g{ = counterclockwise rotation about axis z through an angle of 0 radian g2 = counterclockwise rotation about axis z through an angle of 27r/3 radians g3 = counterclockwise rotation about axis z through an angle of 47r/3 radians g4 = reflection in the plane a g5 = reflection in the plane b g6 = reflection in the plane c From the results of Appendix 30 we know: 1. There are three irreducible representations of G; two of them, which we designate [G][ and [G]2, are one dimensional, and one of them, which we designate [G]3, is two dimensional. 2. The character table for the three irreducible representations is given by Table 1. 3. A particular set of irreducible matrix representations is given in Table 2. Table 1 Element Repi reservation - 1 - 1 1 1 0 1 -1 0 1 -1 0
698 Symmetrical Vibrating Systems Table 2 Element -*"" Representation 12 3 4 5 6 "~"^ 1 2 3 1 1 ri oi l l 2 2VJ 2U 2 J 1 1 1 -1 J"*!] 1 -1 2- 2-V3 1 _iV3 -ij 1 - 1 — iVJ -i »- 2 If we construct a representation of G on the space C by the method discussed in this chapter we obtain [i,] = [3,] = [5„] = 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 1 0- 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 -1 0 0 0 0 0 1 0 0 0 1 0 0 0 1 0- 0 0 0 0 0 0 0 0 1 o' 0 0 1 0 0 0 -1 0 0 0 0-1 0 0 0 0 M = K] = 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 1 0 0 0 0 0 0 1 0 0 0 0 M = 1 0 0 0 0 0 0-1 0 0 0 0 0 0 0 0 10 0 0 0 0 0-1 0 0 10 0 0 0 0 0-1 0 0 0 0 10 0 0 0 0 0-1 0 0 1 ooooo 0-1 0 0 0 0 0 0 0 0 10 0 0 0 0 0-1 (33) The character table for this representation is Element 1 6 2 0 3 0 4 0 5 0 6 0 (34) Using Theorem 26 in Appendix 30 to determine the number ml of sub- spaces in the decomposition of C that are associated with each of the
Examples G& irreducible representations [G]', we obtain mx = i [(6 X 1) + (0 X 1) + (0 X 1) + (0 X 1) + (0 X 1) + (0 X 1)] = 1 (35) ^-4[(6X 1) +(OX 1) +(OX 1) +(OX (-1)) +(OX (-1)) + (OX (-1))] = 1 (36) ^ = ^[(6X2) +(OX (-1))+(OX (-1))+ (0X0)+ (0X0) + (0X0)] = 2 (37) From the above it follows that there is one frequency associated with the irreducible representation 1, one frequency associated with the irreducible representation 2, and two different frequencies associated with the irreducible representation 3. We label these four frequencies coT, co2, co31, and co32, respectively. Since the representation 1 is one dimensional, co1 is nondegenerate. Since the representation 2 is one dimensional, co2 is also nondegenerate. Since the representation 3 is two dimensional, each of the frequencies co31 and co32 is doubly degenerate. Since there is only one irreducible subspace or equivalently only one frequency associated with each of the representations 1 and 2, we know CX=C~X (38) C2 = C2 (39) Since there are two irreducible subspaces or equivalently two frequencies associated with the representation 3, we know C^=CYx®C^ (40) To determine the subspaces Ck we can make use of the projection operators [p]k, which are given by [/>]f = [l]+[2]+,[3]+[4] + [5]+[6] (41) [p]*-[l]+[2]+[3]-[4]-[5]-[6] (42) [/»]J-2[l]-[2]-[3] (43) To use these projection operators it is helpful to know the result of operating °n each of the unit vectors [e)t with each of the operators [g]. This information is displayed in Table 3. The subspace C1 is one dimensional; and therefore to define the space we need to find only one nonzero vector lying in C . The projection operator [pf when operating on an arbitrary basis vector [e) i will yield either zero or a nonzero vector lying in Cl. We therefore need
700 Symmetrical Vibrating Systems Table 3 Operand Operator in [2] [3] [4] [5] [6] leh [eh [eh [eh [eh [eh [eh [eh [eh [e)4 [e)6 ~[eh ~[e\ ~[eh [eh [eh [eh [eh [eh [eh [eh [eh [eh [eh [eh ~[eh ~[eh ~[eh [eh [eh [eh [eh [eh [eh [eh [e)6 [eh [eh [eh ~[e)4 ~[eh ~[e)6 find only some value of i such that [/^[e), =^ 0. If we choose i = 1 we find [P]\e)r [!][*), + [2][e), + [3][e), + [4][g)l + [5][e)l + [6][e)l . = [eh + [e)3 + [e)5 + [e)} + [e)3 + [e)5 = 2[e), + 2[e)3 + 2[e)5 (44) It follows that C1 is the space spanned by the vector 2[e), + 2[e)3 + 2[e)s or equivalently is the set of all vectors of the form a[e)x + a[e)3 + a[e)5, where a is an arbitrary constant. We can express this analytically as follows: ^=(0^), + 0^)3+0^)5} (45) Since C' = C' we also have CT={o[e)1 + a[e)3+a[e)5} (46) A configuration represented by a vector lying in CT corresponds to a configuration in which qx = q3 = q5 = a and q2 = q4 = q6 = 0. Such a configuration is illustrated in Fig. 5. The motion that results when the system is released from such a configuration is called a breathing mode because the motion involves a rhythmic expansion and contraction of the triangle abc. The frequency co associated with Cx can be obtained by making use of Eq. (4). We can use the components c, of any modal vector lying in CT in Eq. (4) to determine co1. We use the modal vector [e)x + [e)3 + [e)5. For this modal vector cx = c3 = c5 = 1 / \ / \ / \ Fig. 5
Examples 701 and c2 = c4 = c6 = 0- Substituting these values in Eq. (4) we obtain (w ) = * * (47) Any choice of i that does not lead to 0/0 will suffice to complete the determination of co1. Choosing i = 1 we obtain a_ (fc/4)(6 + 3 + 3) _ 3fc (a> W(i + o+o) _^ (48) We have thus determined the subspace C] and the frequency coT associated with the modal vectors in this subspace. The subspace C2 is also one dimensional and can be obtained in a similar fashion. If we make use of the projection operator [p]2 we find [p]2[e)2 = 2[e)2 + 2[e)4 + 2[e)6 (49) hence C2 = C2 = {a[e)2 + a[e)4 + a[e)6) (50) where a is an arbitrary constant. The frequency co2 associated with V2 is given by the relation ^ ' mi2 + m/4 + /w/6 m22 + m24 + m26 ^ ' A particular configuration corresponding to one of the modal vectors in C2 is illustrated in Fig. 6. The motions corresponding to the modal vectors in C2 are rotations about the center of the triangle. Since co3 = 0, the system does not return to its equilibrium position but continues rotating in one direction. The space C3 is four dimensional, and therefore we need four linearly independent vectors to define C3. Using the projection operator [p]3 we obtain [p}he)x = 2[e)x-[e),-[e)5 = [a) (52) [p]\e)2 = 2[e)2-[e)4-[e)6 = [b) (53) [PV[e), = 2[e),-[e)5-[e)x = [c) (54) [p]\e)4 = 2[e)4 - [e)6 - [e)2 = [d) (55) Fig. 6
702 Symmetrical Vibrating Systems The four vectors [a), [b\ [c), and [d) can be shown to be linearly independent hence C'={a[a)+fl[b)+y[c)+8[d)} = {(2a - y)[e)x + (2/3 - 8)[e)2 + (2y - a)[e)3 + (28 - 0 )[e)4 - (a + y)[e)5 - (jB + 8 )[e)6} (56) where a, /?, y, and 8 are arbitrary constants. Since C3 = C31 ©C32 we must further decompose C3 to obtain C31 and C32. In the absence of the irreducible matrix representation [Gy]3, the most straightforward method for determining C31 and C32 would be to assume that our system was constrained to remain in C3, and then to solve the resulting dynamical problem. If we do this and choose as our generalized coordinates the parameters a, /?, y, and 8 we obtain ?! = 2a - y (57a) q2 = 2/3-8 (57b) q3 = 2y - a (57c) q4 = 28 - /3 (57d) q5 = - a - y (57e) q6=-/3-8 (57f) Substituting Eqs. (57) in Eqs. (29) and (30) we obtain for the kinetic and potential energies of the constrained system T= 2m(a2 + /32 + y2 + 82 - ay - /38 ) (58) F = X (**2 + £2 + 72 + S2 " <*Y " V3 aS + V3 Py- 08 ) (59) Solving for the modal frequencies and modal vectors by the methods developed in Chapter 65 we obtain ^=0 (60) + V3ate)4-(f«-^/8)te)5-(^a+|J8)[e)6J (62) -V3«[e)4-(|a + ^ ^)^ + (^^-| ^)^1 (63)
Examples 703 / \ / \ / \ / \ Fig. 7 Since C31 and C32 are each two dimensional, two modal vectors are required to span each of the two subspaces. To illustrate the modes, we choose in each space one vector for which qx = 1, q2 = 0 and one vector for which qx = 0, q2 = 1. The modal vector in V3X for which (q,, q2) = (1,0) is [e)i ~ \ [e)3 + 4" te>4 - I [*)s - X [«)« (64) The modal vector in V31 for which (ql,q2) = (0,1) is [eh ~ 4" Kb " £ [04 + 4 [e>* - I[e>« (65) The configurations corresponding to these two modal vectors are illustrated in Fig. 7. The motions corresponding to modal vectors in V31 are translational motions of the system as a whole. Since co31 = 0, the system will not vibrate in this mode but will simply translate in one direction. The modal vector in V32 for which (q{9q2) = (1,0) is [e)x ~ \ [e)3 ~ £- [e)4 - \ [e)s + ^ [e)6 (66) The modal vector in V32 for which (q],q2) = (0,1) is [e)2 + 4"[e)3 ~ 2 [e)4 ~ it Ms ~ \ &* (67) The configurations corresponding to these two modal vectors are illustrated in Fig. 8. .Since we know the irreducible representation [Gy3, we could have found C and C32 much more simply by using one of the projection operators [/»]is2(r')i[*] (68) g H we operate on an arbitrary vector with a particular one of the operators [p]3 we will obtain a vector lying in a two dimensional subspace Dfj of C3, and this subspace will contain one vector from C31 and one vector from C32. We shall choose i = j = 1. Using the results of Table 3 and noting that the reciprocals
704 Symmetrical Vibrating Systems a / \ / \ / \ 8r~ / \ / \ / v <^--> /- Fig. 8 of the group elements are given by Group element Reciprocal 1 1 2 3 3 2 4 4 5 5 6 6 we obtain [/»]?.-m-*[2]-i[3]-[4] + i[5]+i[«] (69) (70) Since the space Z)u is two dimensional, we need two linearly independent vectors to define D?x. The first two nonvanishing projections [/?]n[e); are |>]n[e)2 = 2[e)2-[e)4-[e)6 (71) It follows that [/>]n[03 = §K>3-![e)5 Dn = {2«[e)2 - a[e)4 - a[e)6 + 0[e)3 - j8[e)5} (72) (73) where a and /? are arbitrary constants. We know Dx\ contains one vector from C31 and onejxom C32. We therefore assume that our system is constrained to remain in Dx\ and determine the modal directions for the constrained system. If we choose a and /? as our generalized coordinates then <7. = ° q2 = la q4= -a qs = P lb « (74a) (74b) (74c) (74d) (74e) (749
Examples 705 Substituting Eqs. (74) in Eqs. (29) and (30) we obtain for the kinetic and potential energies of the constrained system T=rn(3a2+ p2) (75) y=ik(3a2 + P2 - 2V3 a/3) (76) Solving for the modal frequencies and for a pair of modal vectors we obtain co31 = 0 (77) »a-(£f <*> [c)P = 2[e)2-[e)4-te)6-V3[e)3 + V3[e)5 (79) [c)F = 2[e)2 - [e)4 - [e)6 + V3 [e)3 - ^3 [e)5 (80) We know that the vector [c)31 lies in C31. We can generate another modal vector [c)3,1 lying in C31 by operating on [c)31 with any one of the operators [g]. Choosing [g] = [2] we obtain [c)¥ = [2][c)V = 2[e)4 - [e)6 - [e)2 - fi[e)5 + $[e\ (81) hence C» = {y[c)T + 8[c)f) = (V35[e), + (2y - «)[e)2 ~V3 y[e)3 + (28 - y)[e)4 + fi(y-8)[e)5-(y + 8)[e)6) (82) where y and 8 are arbitrary constants. In a similar fashion we find [c)f = [2][c)F = 2[e)4 - [e)6 - [e)2 + V3 [e)5 - V§>), (83) hence C^ = { - VJ8[*), + (2y - S)[e)2 + fiy[e)3 + (28 - y)[e)4 -V3(y-8)[*)5-(y + 8)[*)6} (84) where y and 8 are arbitrary constants. The expressions above for C31 and C32 can be shown to be equivalent to the results we obtained in Eqs. (62) and (63). Since the kinetic energy function is of the form £/w2i?/*> we could have used Eq. (26) to determine the modal frequencies and thus avoided the use of
706 Symmetrical Vibrating Systems projection operators. Letting /3 = 4/k in Eq. (24) we M = 6 0 3 -V3 3 ^3 0 2 V3 - 1 -V3 - 1 - 3 V3 6 0 3 -V3 -V3 - 1 - 0 2 V3 - 1 obtain 3 -V3 3 V3 6 0 V3 - 1 -V3 - 1 0 2 (85) Letting n = 1, and choosing g successively as gu g2, and g4 in Eq. (26), we obtain K~l+KJ + 2(K31 +/:32) = Tr{[ly][^.]}=24 ^ + ^-(^ + ^)=^(^^^)=6 K'^ = Tr{[4,][*,]} = 12 It follows that ATT=12 AT2 = 0 KTx+KTl = 6 Now letting n = 2 and choosing g = g, we obtain (^)2+(^)2+2[(^)2 + (^)2]=Tr{[l,][^.]2} = 216 Substituting Eqs. (89) and (90) in Eq. (92) we obtain Combining Eqs. (91) and (93) we obtain K^-O and AT32 = 6 or (86) (87) (88) (89) (90) (91) (92) (93) (94) K3l = 6 and AT32 = 0 (95) Which of the latter two possibilities we choose is immaterial. We choose the first, to keep the notation consistent with our earlier results. Since
Examples 707 We have "T=(£f (97) co5 = 0 (98) co11 = 0 (99) V 2m ) These results agree with our earlier results. Summary. We have found by several different approaches that the system described at the beginning of this problem has two nondegenerate frequencies WT = (3k/m)l/2 and co2 = 0 and two doubly degenerate frequencies co31 = 0 and ^32 = (2k/2m)x/1\ and we have found the subspaces C\ C2, C31, and C32 that contain the modal vectors associated with each of these frequencies.
PART 6 QUASI-COORDINATES
J
67 Quasi-Coordinates INTRODUCTION In describing the motion of a dynamical system it is frequently convenient to use parameters that do not qualify as generalized coordinates or generalized velocities. For example the motion of a planet about the sun is frequently described in terms of the area that the line joining the sun and the planet sweeps out as the planet moves. The area swept out is a useful parameter but it cannot be used as a generalized coordinate, because it depends not only on the position of the planet, but also on the path of the planet. As a second example the rotational motion of a rigid body is frequently described in terms of the components of the angular velocity of a frame fixed in the rigid body with respect to an inertial frame. The components of the angular velocity are useful in describing the motion but are not generalized velocities, since there is no set of generalized coordinates whose time derivatives are equal to the components of the angular velocity. In this chapter we consider such parameters, and in the chapters that follow we see how the equations of motion can be modified to include such parameters. The material in Part 6 is not essential to an understanding of the later chapters and may be skipped. QUASI-COORDINATES Consider a dynamical system consisting of N particles, whose configuration is defined by a set of 37V generalized coordinates g = gl9 g2, . . . , g3N and whose Motion is defined by the corresponding set of 3N generalized velocities Let us assume that at time t0 the system has a configuration g0, and let rfy = 2*/(&0rf&+*(&0* (1) 711
712 Quasi-Coordinates be some arbitrary exact or inexact differential. It is then possible to define a quantity y by integrating the differential (1) along the trajectory of the system from the point (g0,f0) to an arbitrary point (g,t). If the differential (1) is an exact differential, the quantity y will be a function of g and t only and will not depend on the path. Such a quantity could be used as a coordinate for the system. However if the differential (1) is an inexact differential the quantity y will depend not only on g and t but also on the path. Such a quantity may be useful, but it cannot be used as a coordinate in the sense that we have used the term "coordinate" up to this point. We therefore call a quantity y that is obtained by integrating a differential along the trajectory of the system from a fixed point (gQ, /0) to an arbitrary point (g, t) a quasi-coordinate, and we call the quantity y, a quasi-coordinate velocity. A quasi-coordinate may or may not correspond to a real coordinate. Although the configuration g of a dynamical system cannot in general be defined by using quasi-coordinates, it is possible to define the displacement of a system using quasi-coordinates. GENERALIZED COMPONENTS OF THE FORCE FOR QUASI-COORDINATES Suppose g = gu g2, . . . , g3N is a set of generalized coordinates for a dynamical system, and g = gl9 g2, . . . , g3N is the corresponding set of generalized velocities. Let Y = Yi,y2> • • • > Y3w be a set °f quasi-coordinates that are related to the g and the g by the transformation equations dyi = '2*ij(&t)dgJ + ai(g,t)dt (2) J dgi = ^ibiJ(g,t)dyi + bi(g,t)dt (3) j Equations (2) and (3) could also be written j a-Sty&Ot + M&O (5) j It follows that the coefficients aijy bi}, a„ and Z>, can be defined as follows: 3"fc(&8.') , 3&(fcY»0 b* = ^^ — As mentioned in the preceding section, although quasi-coordinates cannot generally be used to define configurations, they can be used to define displacements. It follows that we can define a virtual displacement in terms of a,= b,= dt(g>g>0 dt dg,(g>"M) 3* (6) 0)
Holonomic Constraints 713 quasi-coordinates. The relationship between the virtual displacement Sy and the virtual displacement Sg is given from Eqs. (2) and (3) by *y/ = 2^(&0%/ (8) i «&-S6/y(ftO«Yy (9) J Theorem 1. The virtual work 8W of a force in an arbitrary virtual displacement Sy can be written in the form SW-2r,8y, (10) i where the quantities Ti depend only on the value of the force and are defined as the generalized components of the force corresponding to the quasi- coordinates y,-. proof: The proof of the theorem is the same as the proof of Theorem 1 in Chapter 47. Theorem 2. If Gl9G2, . . . , G3N are the generalized components of a force corresponding to the generalized coordinates gl9 g2, . . . , g3N and Tl9 r2, . . ., T3N are the generalized components of the force corresponding to the quasi-coordinates y!,y2, • • • , y3w> the two sets of generalized components are related by the transformation equations 1-,-2¾¾¾^ (..) proof: The virtual work done in a virtual displacement Sg is given by SW=^lGj8gJ=^lGJ^lbji(g,t)8yi J J i = 2 ¢2-5¾^ «r,-s2 J < i j H «Y/ (12) Comparing Eq. (12) with Eq. (10) we obtain Eq. (11). Holonomic constraints lf a dynamical system is acted on by a set of M holonomic constraints, then as We have seen earlier we can define a set q = ql9q2, . . . , q3N-M of generalized c°ordinates that together with the constraint conditions uniquely specify the
714 Quasi-Coordinates configuration of the system, and we can appropriately modify our definition of the generalized component of a force so that it applies to this situation. The extension of the notion of quasi-coordinates to this situation is straightforward and is left to the reader. We use the variables it = 77-,,77-2, . . . to designate quasi-coordinates for a holonomic system, and II = 11,, II2, . . . to designate the corresponding force components.
68 Lagrange's Equations for Quasi-Coordinates In this chapter we reformulate Lagrange's equations in terms of quasi- coordinates. For simplicity we restrict our considerations to situations in which the quantities atj and btj in Chapter 67 are not functions of the time, and the quantities at and bi are zero. LAGRANGE'S EQUATIONS FOR QUASI-COORDINATES Theorem 1. Consider a dynamical system. Let q = q{,q2, • • • > ?/ be a set of generalized coordinates that together with the holonomic constraint conditions uniquely specify the configuration of the system, q = ql9q2, •••>?/ the corresponding set of generalized velocities, and it = 7^,7^, . . . , -ny be a set of quasi-coordinates that are related to the q and q by the transformation equations */ = 2*/,(<lH- j (1) (2) Let T be the kinetic energy of the system and II, be the generalized component of the force associated with the quasi-coordinate ^. Then the motion of the system is governed by the equations dt 8r(q,*,Q 3 7T, + 2¾¾.¾ Uk -2¾. dq. -n, (3) where xikm = 2 2¾¾ 3% da ■ki 'ji ulm dq, dqj (4) 715
716 Lagrange's Equations for Quasi-Coordinates Equations (3) together with Eqs. (1) provide us with enough equations to determine q(t) and ir(t). proof: If we multiply Lagrange's equations A dt dT(q,q,t) 3r(q,q,Q Qj by 3£y(<l»*>9 dw: = b, and sum overy we obtain 2¼ dt dT(q,q,t) dTi^t) "2fy Ya 2½ j v J (5) (6) C) The first term in Eq. (7) can be rewritten 2¾ J1 dt ZT(q,q,t) -2¾. 3r(q,*,0 3^(q,q) k 9¾ H -2^4 7 ^ -"A = 22V*/^ dT(q,*,t) 37-(q,ir,/) dt Uk 3r(q,*,0 + 22Vv-a5- But and 2Vv2 3gy(q.*) 3^(q»q) 3w,- 3¾ = 8,-, 'ik dakJ dakj %• = 2 id- ?/= 2 2 *7' ^ 3¾ v ^4/ 3ft Substituting Eqs. (9) and (10) in Eq. (8) we obtain blm^m (8) (9) (10) 2*4 3r(q.q>0 ^. <# 3r(q,g,Q drt, + 2 2 2 2^Am-3^- 3¾ . 3r(q,*,0 ,.n irm — y11) k I m dwk
Examples 717 jhe second term in Eq. (7) can be rewritten J v J ar(q,*,0 f 3r(q,*,f) 3^(g,g) ty k 9^ 3¾ (12) But 9**(q>q) ^ 3%/ . ^^ 3% H "S^*"??"^*1"*- (13) Substituting Eq. (13) in Eq. (12) we obtain 3T(q^0 37^,/) Uu. 3r(q,*,/) (14) The third term in Eq. (7) can be written 2,^^== Z a^ 6/ (15) But from Theorem 2 in Chapter 67 the term on the right hand side of the equation above is just II,.. Hence 2fyGy-n, (16) J Substituting Eqs. (11), (14), and (16) in Eq. (7) we obtain Eq. (3). This completes the proof of the theorem. EXAMPLES Example 1. A particle P is moving in a plane under the action of a force directed toward a point O and of magnitude /(r), where r is the distance between O and P. Obtain the equations of motion of the particle in terms of the variables r, 0, r, and A, where 0 is the angle the line OP makes with a fixed direction in the plane, and A is the area swept out per unit time by the line OP. Solution. If we express r and A in terms of r, 0, r, and 0 we obtain r = r (17) A=\r20 (18) ™ in Theorem 1 we make the following identifications qx = r q2=0 (19) ttx = r 7T2 = A
718 Lagrange's Equations for Quasi-Coordinates then n,=/(r) n2 = o r1 au = \ a22 =— an=a2]=0 b\\ = * ^22=-7 *12 = *21 =0 ~ _ 2 .. _ 2 M22 — r u221 — f (20) a^ = 0 if //A: ^ 122 or 221 and the kinetic energy is given by "■* • 2 1 ~ 1 ^. ~ 1 ^ LYYVn-, T= ±rnr2 + ^mr¥ = ±mir? + —f- (21) Substituting these values in Eqs. (3) we obtain I [-.]--^ + -^=/(^) (22) r r Amir 2 ~7~ + f-^- - 0 = 0 (23) Equations (22) and (23) when simplified and expressed in terms of r, 0, r, and A become mf-^=f{r) (24) r A = 0 (25) These are the desired equations of motion. From Eq. (25) it follows that A is a constant of the motion. Example 2. A rigid body is free to turn about one of its points O, which is fixed. Obtain the equations of motion in terms of the coordinates 0, <f>, *//, and the velocities coT, co5, and co5, where 0, <f>, and i// are the Euler angles and coy, C05, and C05 are the components of the angular velocity with respect to a frame S fixed in the body with origin at O and axes in the principal directions for the point O. Solution. If we express coT, co5, and co5 in terms of 0, <f>, i//, 0, <j>, and ;// we obtain coy = cos i// 0 + sin 0 sin xp <j> (26) C05 = — sin $ 0 + sin 0 cos tp <j> (27) C03 = cos 0 <f> + \j/ (28)
jf in Theorem 1 we make the following identifications 7Ti = COt then U2=G-2 77" •> — CO? U3=G-3 where G- is the i component of the torque, cos \p sin 0 sin \p 0 [#//]= I — simp sin 6 cos i// 0 [*.]" 0 COS i// sin \p sin# sin ^/ cos 0 sin# cos# 1 — sin \p cos *// sin# — cos ^/ cos 9 sin# Examples 719 (29) (30) (31) (32) and aijk ~ eijk (33) where eijk is the Levi-Civita symbol (see Appendix 6), and the kinetic energy is given by T = f J™2 + \h*\ + {1*4 = i/t*? + i/j*! + {I?hl (34) where the I? are the principal moments of inertia. Using the values above in Eq. (3) we obtain which are Euler's equations. (35) (36) (37)
69 The Gibbs-Appell Equations of Motion INTRODUCTION In this chapter we introduce a new formulation of the laws of dynamics and show that there is a single function that by itself completely characterizes the dynamics of an anholonomic system in the same way that the kinetic energy function characterizes the dynamics of a holonomic system. A USEFUL LEMMA We will find the following lemma useful in the proof of the theorem in the next section. Lemma 1. Let g = g,, g2, . . . , g3N be a set of generalized coordinates and y = yl5y3, . . . , y3N be a set of quasi-coordinates. Then 3¾ at/ 0) proof: From Eq. (5) in Chapter 67 it follows that & = &(g>Y>0 (2) Taking the time derivative of Eq. (2) we obtain 6 = ? SQ- *+ ? -iy-*+ —37— (3) Substituting Eq. (2) in Eq. (3) we obtain & = 2 9^ gj(8> Y, 0 + 2 3^ Y, + gj (4) 720
The Gibbs-Appell Equations 721 Treating g, as a function of g, y, y, and t and taking the partial derivative with respect to yj9 we immediately obtain Eq. (1), which is what we wanted to prove. THE GIBBS-APPELL EQUATIONS Theorem 1. Consider a dynamical system consisting of N particles and moving under the action of a force. Let x = xx,x2, . . . , x3N be a set of Cartesian coordinates, mi the corresponding masses of the particles, g = g,, #2> • * * ' #3w a set °f generalized coordinates, and y = y,,y2, . . . , y3N a set of quasi-coordinates. Let the set of equations relating the gt and the y, be given by S, = S,(g,"M) (5) Then dS(g,y,y,t) = r, (6) where T{ is the generalized component of the force corresponding to the quasi-coordinate y/5 and S^^mtf (7) i The quantity S is called the energy of acceleration, or AppelPs function, or the Gibbs-Appell function. The set of Eqs. (6) together with the set of Eqs. (5) provide us with enough equations to solve for the set of unknowns y and the set of unknowns g. proof: From Newton's law "/*/=// (8) But dS(x) Substituting Eq. (9) in Eq. (8) we obtain dS(x) -3--=/1 (10) Multiplying Eq. (10) by dxi(g,y,t)/dyJ and summing over i we obtain f 3*, ty *}> dyj <U>
722 The Gibbs-Appell Equations of Motion According to Eq. (11) in Chapter 67 the right hand side of Eq. (11) above can be written 2//-3-- = r, (12) If we use Lemma 1 on the left hand side of Eq. (11) we obtain dS(x) 3x,(&v,Q _ dS(x) dxfawt) _ dS(g,y9y,t) f 3¾ dyj f 3¾ 3¾ 3¾ (13) Substituting Eqs. (12) and (13) in Eq. (11), we obtain Eq. (6), which is what we wanted to prove. The Eqs. (6) are called the Appell equations of motion, or the Gibbs-Appell equations of motion. They were independently discovered by Gibbs in 1879 and by Appell in 1899. ANHOLONOMIC SYSTEMS One of the most useful features of the Gibbs-Appell equations is that an- holonomic constraints can be handled in a fashion analogous to the handling of holonomic constraints in Lagrange's equations for holonomic systems. Theorem 2. Consider a dynamical system consisting of N particles, which is acted on by a given force, and a set of L constraint forces, M of which are holonomic and the rest anholonomic. Let q = qx, q2, . . . , ^-M be any set of generalized coordinates that together with the M holonomic constraint conditions uniquely specify the configuration of the system. Let w = 7rl5 tfr2> • • • 9^3N-L be any set of quasi-coordinates that together with the L constraint conditions can be used to specify a displacement of the system. Then 07Ti where ,S(q, *,#,*) is the Gibbs-Appell function of the constrained system, and IT, 877-,- is the work done in a virtual displacement in which only 777 is varied and only in such a way that none of the constraint conditions is violated. The set of 3N — L equations (14) together with the set of 3N - M equations *=*(**>') (15) provide us with the equations needed to solve for the set *n and the set q as functions of the time. proof: Let us suppose that we have chosen a set of generalized coordinates g = g,, g2, . . . , g3N and a set of quasi-coordinates y = *yl5y2, . . . , y3^-
Anholonomic Systems 723 Let the L constraint equations be 2^(&0&+ Ms>') = 2*/(&')7,-+ *0(&0 = o 2^'(& 0&+ A'<fe 0 = 2 */(& 0t+ «o(g, 0 = ° i i 2^(L)(g> 0&+ 4L)(& 0 = 2 */L)(& 0t+ ^L)(& 0 = o (16) i i If we define the quantities \p,\p', . . . , t//(L) as follows: *(&y>0=2*/(&0* + *o(&0 f(g>Y>0=25/(g.0Y/+fio(g>0 *(L)(g>Y,0 =2*/L)(&0y/ + ^H&O (17) then the constraint equations can be written *(«.*,<) = S 3V ;Y, +^0(g,0 = 0 *'(& Y, 0 = 2 3^ ^ Y, + ^o(g- 0 = 0 *(L)(g.T.O = 2 3 ^ Y, + *o(i)(g>0 = 0 (18) It should be noted that even though we can define the functions \p(g,y,t), it does not follow that there exists a function \p(g, i) of which \p is the time derivative. If the constraint is holonomic there will be such a function; if the constraint is anholonomic there will not be such a function. The generalized components of the constraint forces associated with the quasi-coordinate y, can be written \d\i>(g,y,t)/dyh A'3^gvY,0/37i'> . . . . The justification of this follows the arguments given in Chapter 49. The Gibbs-Appell equations for the system are thus ^ = 1 • + A ^-. h A 7T-. h • • • 9 7/ 3 7/ 9 7/ + x<«^|M (W) These equations together with the constraint equations * (g, Y, 0 = 0, i'(g, Y> 0 = 0, . . . , ^L)(g, y, t) = 0 (20)
724 The Gibbs-Appell Equations of Motion provide us with the necessary number of equations to solve for the y. as functions of g and t. Now suppose that we choose as our quasi-coordinates the set Y = Yi,Y2> • • • >y3N = *\>ir2> • • • >*3n-uW> • • • >t(L) = ^,^ (21) The corresponding set of generalized components of the force is r = r„r2,..., ^ = 11,,1¾...., n3N_L,%<y,..., *<L> =n,* (22) For this set the Gibbs-Appell equations become 3S(g. 3S(g, 3^(8. 3^,. 17,^,17 3^ *,&* 3i// •jt, ^, *ji -,&o ■>*,') ;*,*) (23a) = a = * + x = *' + A' = *<L)+X(L) (23b) and the constraint equations are j> = 0, f = 0, . . ., ^(L) = 0 (24) If we substitute Eqs. (24) into Eqs. (23a) we obtain aS(ftir,0, #,0,/) 3 77, = n, (25) Now suppose that the first M of the L constraints are holonomic; then there exists a set of functions ip(g,t),ip'(g,t), . . ., i//(A/)(g,/) such that Wis,') dt d^M\g,t) = ^<"> (26) and such that the first M constraint conditions can be written *(& 0 = 0, ^'(g, /) = 0, . . . , ^M\g, t) = 0 (27) Now suppose that we choose as our generalized coordinates the set S = gt,g2> •• -.^ = ^1.^2, • • -,q3N-M>ip>t', •• -,^M) = <lA (28)
Examples 725 where the set q is any set of 3 N — M coordinates that together with the M constraint conditions (27) uniquely define the configuration of the system. In terms of these coordinates Eq. (25) becomes 3S(q, ^,17,0,17,0,/) -^ --n' (29) Finally if we substitute Eqs. (27) into Eqs. (29) we obtain 3S(q, 0,17,0, £,0,/) 377-,- = n, (30) But S(q, 0,17,0,i7,0,/) is just the value of S for the constrained system and H/ &7> is the virtual work done in a virtual displacement in which only 7ti is varied and none of the constraints are violated. This is what we set out to prove. THE GIBBS-APPELL FUNCTION Although the Gibbs-Appell equations are simpler in appearance, and also more versatile in their handling of anholonomic constraints than Lagrange's equations, they are not as well known or as widely used. There are several reasons for this. In the first place the Gibbs-Appell function S(q, 17,17,/) is generally harder to obtain than the kinetic energy, !T(q,q,/). In the second place the number of anholonomic constraints is usually not so great that one is seriously handicapped by having to treat them by the methods used in Lagrange's equations for anholonomic systems, that is, by means of the Lagrange multipliers A, A', . . . . The usual method of obtaining the Gibbs-Appell function 5(^17,17,/) is to simply find jc,(q, 17,17,/) and then to substitute this result into the expression 5 = |2/m/^/2- There are several simplifications that are sometimes helpful in the process. In the first place any additive term in the final expression for 5(q, 17,17,/) that does not contain one of the 777 can be dropped, since it will have no effect on the equations of motion. In the second place if we tet S(c) be the Gibbs-Appell function with respect to the center of mass c, M the total mass of the system, and A the acceleration of the center of mass with respect to our origin of coordinates, then S = {MA2+ S(c) (31) The proof of this result is given in Chapter 70. EXAMPLES Example 1. A particle P is moving in a plane under the action of a force greeted toward a point 0, and of magnitude /(r), where r is the distance between O and P. Obtain the equations of motion of the particle in terms of
726 The Gibbs-Appell Equations of Motion the variables r,0,r,A, where 9 is the angle the line OP makes with a fixed direction in the plane, and A is the area swept out per unit time by the line OP. Solution. The Gibbs-Appell function S in terms of the polar coordinates r 6 is given by S=±m ^r(rcos9) dtlK ' 2 ^r(rsm9) dt2K ' = )r m(f + 4r292 + r292 + r294 - 2rr92 + 4rr99 ) (32) The area swept out by OP per unit time is given by A=\rH (33) hence 9 = M (34) fj=2l_4A±_ 35 r r Substituting Eqs. (34) and (35) in Eq. (32) we obtain S(r,0,M,M)= \m[r2+ *¥- - $Mi j + fm(r999r9A) (36) We can if we want drop the last term [i.e., fcn(r, 0, r9A)\ since it will disappear anyway when we take derivatives with respect to f and A, hence has no effect on the motion. The Gibbs-Appell equations for the system are dS(r,9,rJ,r,A) _tt M 11, V1) dS(r,9,r,A,r,A) '/ ) = n, (38) dA From the definitions of the generalized components of the force we have nr=/(r) and 11^=0 (39) Substituting Eqs. (36) and (39) in Eqs. (37) and (38) we obtain mr _ Am A r3 4mA r2 n = f{r) (4°) = 0 (41) These are the equations of motion. From Eq. (41) it follows that A is constant. note: Examples in which the Gibbs-Appell equation is applied to an- holonomic systems can be found in the following chapter.
70 The Gibbs-Appell Equations and Rigid Body Motion INTRODUCTION To apply the Gibbs-Appell equations to the motion of a rigid body it is necessary to be able to determine the Gibbs-Appell function for a rigid body. THE GIBBS-APPELL FUNCTION FOR A RIGID BODY Theorem 1. The Gibbs-Appell function of a dynamical system with respect to an arbitrary point o is given by S(o) = \M[A(o)]2+ S(c) (1) where M is the total mass of the system, A(o) is the acceleration of the center of mass c with respect to the point o, and S(c) is the Gibbs-Appell function with respect to c. proof: If we let a(//) represent the acceleration of a pointy with respect to a point i then S(o) = \^m(i)[a(oi)}2 i = i2m(0[a(oc) + a(c0] •[*(<*:) + a(c0] i 2m(OJ[a(00 -a(°c)] + i 2w(0[a(°0 * ■(*')] _ 1 2 + 2w(0a(°0 ■ a(oc) (2) 727
728 The Gibbs-Appell Equations and Rigid Body Motion But ]>>(/) = M (3) i a(oc) = A(o) (4) and from the definition of the center of mass 72 i dt L < Substituting Eqs. (3)-(5) in Eq. (2) we obtain Eq. (1). = 0 (5) Theorem 2. The Gibbs-Appell function of a rigid body with respect to a point o fixed in the rigid body is given by s(°) = 2 S Shf»Pj+ S S S S*ijklki<*i<*j<*i (6) i j i j k I where the o)i are the components of the angular velocity with respect to some frame R, the 1^ are the components of the inerti eijk is the Levi-Civita symbol (see Appendix 6). frame R, the Iy are the components of the inertia tensor with respect to R, and proof: We first find the Gibbs-Appell function with respect to the point o for a single particle fixed in the rigid body, and then sum over all the particles in the rigid body. For a single particle P fixed in the body S = ^mr-f (7) where r is the vector position of P with respect to o. If wejet R be a frame fixed in the body; co and f be the time rate of change in R of co and r; and note that r = 0 and <Jo = co we have (see Chapter 6) f = f + <oxr = <oXr (8) f = (<oxr) + <ox(<oXr) = <bXr + <oxf + <oX(<oxr) = <b X r + 63 X (« X r) (9) Substituting Eq. (9) in Eq. (7) we obtain S = \m[6z X r + 63 X (63 X r)] • [to X r + 63 X (63 X r)] = A + B+C (10) where A =^m(63Xr)-(63Xr) (11) B = m(63 x r) • [63 X (63 x r)] (12) C =^ra[63X («xr)] «[«x («xr)] (13)
The Gibbs-Appell Function for a Rigid Body 729 Using the summation convention, and the techniques described in Appendix 6 We can rewrite A: A = ^meiJko)Jxkeilm<blxm = 2meijkeilmXkXm^j^l = ±m(8jfikm - 8Jm8kl)xkxm6)J6)l = ^m(SJl8kmxkxm - Sj^x^ibj^ = im^Sjt - XjXf)^ (14) B = meijk^jXkeilm^lemrS^rXs = m(8jAm ~ $jm$lk ymrs^l^rXkXs = m(*krs"jUjUrXkXs ~ *jrs"j"kUrXkXs) (15) Similarly But Wrt = " ^,¾¾ = - (r X r)r= 0 (16) or equivalently ekrsXkXs ~ 2 (esrkXsXk + ekrsXkXs) = 2 ( ~ €krsXkXs + €krsXkXs) ~ 0 ( 17) Substituting Eq. (17) in Eq. (15) we obtain B= - rneJrsG)jOkorxkxs = "Kjrs"j°>kUr(rXs ~ XkXs) ~ "KjrsUj"k"rr\s (18) But ^jrs^k^rr\s = ejrk<*j<*k<*rr2 = ^j^jrk^r^k) = ° 09) Substituting Eq. (19) in Eq. (18) we obtain B = ejrS[m{r\k ~ *,**) ]tt/0r«* (20) The term C given by Eq. (13) does not contain any accelerations, hence may be dropped from the Gibbs-Appell function because it does not affect the motion of the system. Substituting Eqs. (.14) and (20) in Eq. (10) and dropping the term C we obtain S = ^(r2^ - Xjx^ufo + eJrs[m(r28sk - xsxk)]u/*rak (21) If we sum over all the particles in the system, drop the summation convention, relabel the indices, and recall that hj ~H'»(p)[r\p)8lj - xt(p)Xj(p)] (22) P we obtain Eq. (6).
730 The Gibbs-Appell Equations and Rigid Body Motion Corollary. If in Theorem 2 we choose as reference frame a frame R that is fixed in the body and whose axes are in the principal directions, the Gibbs- Appell function is given by s(o) = {2 itf + 2 2 2 ^4^ (23) where the /• are the respective principal moments of inertia. EULER'S EQUATIONS Euler's equations of motion for a rigid body follow immediately if we use the Gibbs-Appell function given by Eq. (23) in the Gibbs-Appell equation, with ir( = cof. EXAMPLES Example 1. A sphere of mass M, radius a, and moment of inertia I about a diameter is free to roll on a rough horizontal table, which is rotating with an angular velocity 2 about a vertical axis that passes through a fixed point O on the table. Determine the motion of the center of the sphere. Solution. Let R be an inertial frame with origin at O, and 3 axis vertically up. Let R' be a frame fixed in the table with origin at O and 3' axis vertically up. Let R be a frame fixed in the sphere with origin at the center C of the sphere. Let P be the point on the sphere that is in contact with the table, and P' the point on the table that is in contact with the sphere (Fig. 1). Since the table is rough, the velocities of P and P' must be equal. If we let x and y be the xx and x2 coordinates of C with respect to the frame R, the velocity of the point P' is (see Chapter 6) f (OP') = r {OP') + <*{RR') X r(OP') = <o(i?iT)xr(OP') = kQ x (ix + jy) = -Qyi + Qxi (24) 3 ^- I Fig. 1
Examples 731 If we let o)x, co^ and coz be the 1, 2, and 3 components of co(RR), the velocity of the point P is r(OP) = r(OC) + r(CP) = r(OC) + i(CP) + a(RR)xr(CP) = r(OC) + co(RR)xr(CP) = (Ix + jj>) + (icox + jco^ + kcoz) X (-kfl) = (x - aya)\ + (y + o>xa)] (25) Equating Eqs. (24) and (25) we obtain x - uya + Qy = 0 (26) j/ + coxa - Six = 0 (27) Equations (26) and (27) are the equations of constraint. The Gibbs-Appell function for the sphere is given by S = ±M(x2+f)+ S(C) (28) where s(C) = I s 2 W/+ 222 SV*M"' (29) < y i j k I The inertia tensor for the sphere is given by /(, = /¾ (30) substituting Eq. (30) in Eq. (29) we obtain S(C) = 1/(^ + (^ + ¾2) (31) Substituting Eq. (31) in (28) we obtain S = {M{x2 + y2) + ^/(<b2 + cb2 + <b2) (32) Of the five quantities x, j, cox, w , and coz, only three are independent, since there are two constraint equations. We choose x, y, and coz as the independent variables. From Eqs. (26) and (27) we obtain -.-5^ (33) fly + x »y-JLr- (34) Using Eqs. (33) and (34) to eliminate ux and ay from Eq. (32) we obtain S = I M(x2 +f) + ±-L [(q* _^)2 + (fly> + *)2 + a2^2j (35)
732 The Gibbs-Appell Equations and Rigid Body Motion Since there are no forces other than the constraint forces, the Gibbs-Appell equations are dS(x, y,x,y,uz) (36) (37) (38) Substituting Eq. (35) in dS(x dS(x dx ,y,x,y, dy 3wz Eqs. (36)-(38) we Mx + My- iw. 7<ai ^), ^1, 0 0 obtain + Jc) = -7) = = 0 = 0 (39) (40) /cbz = 0 (41) From Eq. (41) it follows that the z component of the angular momentum is conserved. If we define co0 = -& (42) 0 (Ma2/I)+\ K } then Eqs. (39) and (40) become x + co0 y = 0 (43) y - co0i = 0 (44) Solving these two equations we obtain x = A + B cos o)0t + C sin co0/ (45) y = D — Ccosco0/ + Bsinu0t (46) where A, B, C, and D are constants whose values are determined by the boundary conditions. If we eliminate the time from Eqs. (45) and (46) we obtain (x-Af+(y-D)2=B2+ C2 (47) This is the equation of a circle with center at x = A, y = D and of radius (B2 + C2)1/2. Thus the center of the sphere moves in a circular orbit with an angular speed co0.
PART HAMILTONIAN MECHANICS
SECTION Hamilton's Equations of Motion
71 Hamilton's Equations of Motion In this chapter we consider the Hamiltonian formulation of the equations of motion. We initially restrict our attention to Lagrangian systems whose behavior under the action of a given force can be defined by a Lagrangian function, that is, systems for which the constraints are holonomic and the generalized components of the given forces are derivable from a potential. THE HAMILTONIAN The behavior of a Lagrangian system is defined if we know the Lagrangian L as a function of the generalized coordinates q = quq2, • • • , qp the generalized velocities q = q\,q2, • • • , qp and the time t. One of the most useful sets of quantities encountered in studying the motion of a Lagrangian system is the set of generalized momenta, which are defined as _^(q,q,Q m h = ~W~ (1) In Appendix 22 we show that if we are given a function/(x, y) the Legendre transformation of /(x, y) with respect to the set of variables x when expressed as a function of the sets of variables 3/(x,y)/3x, and y, contains the same ^formation as the original function. It follows that the Legendre transformation of the function L(q,q, i) with respect to the set of variables q, when expressed as a function of the set of generalized momenta p, the set of generalized coordinates q, and the time /, contains the same information as the function L(q,q, t). The Legendre transformation of L(q,q, i) with respect to the variables q is called the Hamiltonian and is designated by the letter H. Thus H^Ptq-L (2) 737
738 Hamilton's Equations of Motion The function H(q, p, t) is called the Hamiltonian function. Just as the Lagrangian function L(q,q, t) defines the behavior of the Lagrangian system so also does the Hamiltonian function H(q, p, t). Although Eq. (2) is the fundamental definition of the Hamiltonian, there are many situations in which the Hamiltonian is equal to the sum of the kinetic energy T and the generalized potential U. The conditions for which this is true are given in the following theorem. Theorem 1. Consider a Lagrangian system whose configuration is defined by the set of generalized coordinates q = qx,q2, • • • , <7/ and whose behavior is defined by the Lagrangian function L(q, q, t). If the generalized potential U is not a function of q, that is, U = £/(q, t), and either the equations of transformation between the set of Cartesian coordinates x and the set of generalized coordinates q are not functions of the time, that is, x = x(q), or the kinetic energy T(q, q, t) is a homogeneous quadratic function of the set of generalized velocities q, then the Hamiltonian H is the sum of the kinetic energy T and the generalized potential U, that is, H=T+U (3) proof: The proof of this theorem is contained in the proof of Theorem 3 in Chapter 56. The proof is not altered by the possible dependence of U on the time in this theorem. HAMILTON'S EQUATIONS OF MOTION If we are given the Hamiltonian function #(q,p, 0 and wish to find q(/), we could proceed by first finding the Lagrangian function L(q,q, t) and then making use of Lagrange's equations of motion, that is, d \ ^M>0 dt dqt There is however a more direct way. V } = 0 (4) Theorem 2. Consider a Lagrangian system whose configuration is defined by the set of generalized coordinates q = q\,q2, • • • > ?/ and whose behavior is defined by the Hamiltonian function H(q,p,t). The behavior of the system can be determined from the equations 4.^|ElO (5a, . A = -4^ <5b> These equations are called Hamilton's equations of motion.
Constants of the Motion in the Hamiltonian Formulation 739 proof: From the results in Appendix 22 we can construct the following table *'= H (6a) q> W~ {) H = ^Piq-L (7a) L = ^qiPi-H (7b) i i 3L(q,q,/) 3#(q,p,/) 3L(q,q,/) 3#(q,p,0 (8a) (8b) 3/ 3/ Equation (6b) gives us Eq. (5a). If we substitute Eqs. (6a) and (8a) in Eq. (4), we obtain Eq. (5b). Equations (5a) and (5b) provide us with 2/ equations in the 2/unknowns quq2, • • • , qt>P\> Pi> • • • >/>/• Thus we can solve for p(/) and q(0- CONSTANTS OF THE MOTION IN THE HAMILTONIAN FORMULATION In Chapter 17 we saw that solving the equations of motion for a dynamical system is equivalent to finding the independent constants of the motion. In Chapter 56 we saw that the Lagrangian formulation of the laws of dynamics were useful in revealing constants of the motion. In a similar fashion each new formulation of the laws of dynamics tends to provide a different perspective that frequently is useful in finding constants of the motion. The following two theorems are useful in this respect. Theorem 3. If the Hamiltonian function H(q, p, t) is not an explicit function of the time, the Hamiltonian H is a constant of the motion. proof: The time rate of change of H is given by dH ^ 3#(q,p,Q , , ^ 3g(q,p,Q , , 3#(q,p,Q "S^— ft+S—^--/>,+ 3, (9) Substituting Eqs. (5a) and (5b) in Eq. (9) we obtain dH _ dH(W) dt 3/ ^ ^ #(<1>P>0 is not an explicit function of the time then 3//(q,p,/)/3/= 0; hence dH/ dt = 0. The theorem follows immediately. We could have also proved the theorem by noting from Eq. (8b) that if H is not an explicit function of the time, L is not an explicit function of the time. The proof of the theorem is then the same as the proof of Theorem 3 in Qiapter 56.
740 Hamilton's Equations of Motion Theorem 4. If the Hamiltonian function H(q, p, t) is not an explicit function of a particular generalized coordinate qi9 the conjugate generalized momentum pt is a constant of the motion. If the Hamiltonian function H(q, p, t) is not an explicit function of a particular generalized momentum pi9 the conjugate generalized coordinate qt is a constant of the motion. proof: The proof of this theorem follows immediately from Hamilton's equations of motion, Eqs. (5a) and (5b). EXAMPLES Example 1. A particle of mass m and charge q moves in an electromagnetic field defined by the scalar potential $ and the vector potential A. Choosing the Cartesian coordinates *,, x29 and x3 of the particle as generalized coordinates, obtain the Hamiltonian function and the equations of motion. Solution. From the example in Chapter 53, the force on the particle is derivable from the generalized potential U=q^-q^xiAi (11) i Hence the Lagrangian is given by L = ^mxj-q^ + q^xiAi (12) i i The generalized momentum corresponding to the generalized coordinate xt is p'=ur mk>+ qAi (13) Note that the generalized momentum corresponding to the coordinate x{ is not equal to mxh the xt component of the linear momentum. From Eq. (13) it follows that _ Pi ~ qAf i m The Hamiltonian is given by (14) i Z i i 2 i Substituting Eq. (14) in Eq. (15) we obtain the Hamiltonian function H(x>P) = J^2,(Pi-<lAi)2+q$ (16)
Problems 741 rjainilton's equations of motion for the particle are Xi = -W~ (17) Substituting Eq. (16) in Eqs. (17) and (18) we obtain i, = ^^ (19) A-*2(*-*4)^-,(fi) (20) Equation (19) is the same as Eq. (14). Using Eq. (19) to eliminate/?, from Eq. (20) we obtain -,+H=*s^-*(if) (21) Noting that we obtain 4-2(¾)^¾ m dt a$ Mi . v dAJ v a^< ,<m ""'"-^-'ir+^a^-^a^ (23) Comparing Eq. (23) with Eq. (16) in Chapter 53 we find that the right hand side of Eq. (23) is simply the jc,- component of the Lorentz force F = qE + q\ X B. Hence mxt = Ft (24) Thus Hamilton's equations of motion for the particle are equivalent to Newton's law. PROBLEMS 1. Find the Hamiltonian function for a system whose Lagrangian in cylindrical coordinates is given by L = {{pi1 + p2<j>2 + i2) + ap2<j>, where a is a constant. a. Write down Hamilton's equations of motion for the system. b. Obtain Lagrange's equations of motion from them by elimination of the variables pp, p^, and pz.
742 Hamilton's Equations of Motion 2. Apply Hamilton's equations to the plane motion of a particle of unit mass in a field of force for which the potential in polar coordinates is k cos <j>/ r2, and prove that if at time t = 0 it is given that r = a and r = o then t = [(r2 — a2)/2E]l/2, where E is the constant total energy. 3. A particle of unit mass is moving in a plane under the action of a conservative central force. The potential energy of the particle is K = 2Xr — [x2/2r2, where A and jit are constants. Find Hamilton's equations of motion, using polar coordinates r and <j> as generalized coordinates, and show that in the orbit for which the angular momentum is \i and the total energy is 2a\ there is an apse at r — a, and after a time (a/2X)]/2 the particle is at a distance a/2 from the origin. 4. A particle of mass m moves in three dimensions under the action of a conservative force with potential energy V(r). Using the spherical coordinates r, 0, and <j> as generalized coordinates, obtain the Hamiltonian function for the system. Show that the quantities p<p,(p}/2m) + {pl/2mr2) + (pl/2mr2sm20) + V(r\ and p] + (/?2/sin20) are constants of the motion. 5. Two particles of masses m and M, respectively, interact with each other. The potential energy of interaction is V(r), where r is the distance between the two particles. Choosing the Cartesian coordinates X, Y, and Z of the center of mass of the system, and the spherical coordinates r, 9, and <f>, which locate the particle m with respect to the particle M, find the Hamiltonian function, and obtain six constants of the motion. 6. A mass m is suspended by means of a string that passes through a small hole in a table. The particle is set in motion in a vertical plane and the string is drawn up through the hole at a constant rate a. a. Choosing the angle 0 that the string makes with the vertical as the generalized coordinate, find the Hamiltonian function for the system. b. Is the Hamiltonian equal to the total energy? c. Is the Hamiltonian function a constant of the motion? 7. A particle in a uniform gravitational field is constrained to the surface of a sphere. The radius r of the sphere varies in time, that is, r = r(t). Choosing the spherical coordinates 6 and <J> as generalized coordinates with the origin at the center of the sphere and the polar axis vertically up, obtain the Hamiltonian function and the equations of motion. Is the Hamiltonian the total energy? Discuss energy conservation. 8. A particle of mass m moves in one dimension under the influence of a force /= (/c/x2)exp[ — (t/r)], where k and r are positive constants. Obtain the Lagrangian and Hamiltonian functions. Compare the Hamiltonian and the total energy.
Problems 743 9# A particle of unit mass is moving in a plane under the influence of the generalized potential U= (1 + r2)/r, where r is the distance from the origin. Write down the Lagrangian in polar coordinates r, 0, and from it obtain pn p0, and H(r,0, pn pg). Find the equations of motion and show that angular momentum is conserved. Reduce the problem for r to a single first order differential equation for r. 10. A particle moves in a coordinate system S fixed to the earth, which is rotating with fixed angular velocity co = coe3 relative to an inertial frame. Find the Lagrangian in the earth's system. Find the momenta conjugate to the Cartesian coordinates *,, x2, and x3 of the earth's system, and calculate the Hamiltonian in this system. Show that H is not the total energy but that it is a constant of the motion.
72 Equations of Motion of the Hamiltonian Type INTRODUCTION In this chapter we consider a number of topics that are related to Hamilton-'s equations but are not essential to an understanding of any of the later chapters. Readers may skip this chapter if desired. HAMILTON'S EQUATIONS OF MOTION OF THE SECOND KIND If the components of the generalized force are not derivable from a potential then in the Lagrange approach we use Lagrange's equations in the form A. dt 37^0 -¾^ - a (i) From Eq. (1) it follows that the kinetic energy function r(q,q, t) defines the dynamical nature of the system. We can generate an alternate formulation of the foregoing equations of motion by taking the Legendre transformation of T(q, q, 0 with respect to the set of variables q. We summarize the results in the following theorem. Theorem 1. Consider a holonomic system whose configuration is defined by the set of generalized coordinates q= quq2, . . . , q^. If we define 8£(*M (2) K=2,M-T (3) 744
Hamilton's Equations of Motion of the Second Kind 745 where T is the kinetic energy of the system, the behavior of the system under the action of the force Q can be determined from the equations • ^(¾^) q, = —^- (4a) i,-—^+a (4b) If r(q,q,0 is a homogeneous quadratic function of the set of generalized velocities q, then K = T (5) If the generalized components of the given force are derivable from a generalized potential U, and if the generalized potential is not a function of q, that is, U = U(q, t), then si=Pi (6) We call Eqs. (4a) and (4b) Hamilton's equations of the second kind. proof: Since K is the Legendre transformation of T(q, q, t) with respect to the set of variables q, we can construct the following table, using the results of Appendix 22: ar(q,q,o K=lLW-T (8a) i 3T(q,q,0 = _ 3A"(q,M) i 3A"(q,M) (7b) (8b) m 3ft H v Equation (7b) gives us Eq. (4a). If we substitute Eqs. (7a) and (9) in Eq. (1) we obtain Eq. (4b). If r(q, q, t) is a homogeneous quadratic function of the generalized velocities q, then from Euler's theorem, Appendix 20, we have ?*"~a*- (10) Combining Eqs. (7a), (8a), and (10) we obtain Eq. (5). If the generalized components of the force are derivable from a velocity independent potential
746 Equations of Motion of the Hamillonian Type U(q, t), then ZL(q,q,t) 37-(q,q,0 ^ = ^^- = -3^- "*' (11) This completes the proof of the theorem. ROUTH'S EQUATIONS OF MOTION Given a Lagrangian function L(q, q, t) it is sometimes convenient to consider a Legendre transformation with respect to some subset of the set of generalized velocities q= q{,q2, • • • , ?/• The result of doing this is summarized in the following theorem. Theorem 2. Consider a system whose configuration is defined by the set of generalized coordinates q',q" = q\,q2, ..., ^,?i'>?2 > • • • > q/-r and whose behavior under the action of a given force is defined by the Lagrangian function L(q',q",q',q",/). If we define and Pi = ,q",q',qV) R^p"q'/-L i of motion for the system are d dt r 3*(q',q",q\pV) ] m 3*(q',q",q',pV) H .„ dR(q', q",q',pV) ?' = T~T> dR(i f>" = 1 il',q",q',p",0 w (12) (13) = 0 (14) (15a) (15b) The quantity R is called the Routhian, and the equations of motion are called Routh's equations of motion. note: The Routhian is more often defined as the negative of the definition given here. The present definition however is more consistent with the definition of the Hamiltonian H. proof: Since the Routhian is just the Legendre transformation of the Lagrangian function L(q',q",q', q",/) with respect to the variables q", we can
Examples onstruct the following table using the results of Appendix 22. 3L(q\q",q',qV) pr- w (16a) RSPW'-L (17a) i 3L(q',q",q',qV) 3L(q\q",q',qV) 3L(q',q",q',qV) .„ 3/?(q',q",q',p" *' = ^- L = ^q;'P;'-R i 3*(q'.q",q',pV) 3?; 3*(q',q",q',pV) 3i?(q',q",q',pV) A 747 (16b) (17b) (18) (19) C2(tt 3*' H Equation (16b) immediately gives us Eq. (15a). If we substitute Eqs. (18) and (20) into Lagrange's equations for the variables q[, that is, A dt 3L(q',q",q',qV) 3L(q',q",q',qV) = 0 (21) we obtain Eq. (14). If we substitute Eqs. (16a) and (19) into Lagrange's equations for the variables q", that is, A dt 3L(q',q",q',qV) m> 3L(q',q",q',qV) = 0 (22) we obtain Eq. (15b). Routh's equations are particularly useful when one or more of the generalized coordinates are cyclic, that is, the Lagrangian function does not explicitly contain these coordinates. Let us assume that these are the f—r coordinates q". From Eq. (15b) it immediately follows that the corresponding momenta are constants of the motion. It then follows that we can set the p" in Eq. (14) equal to constants, and since the Routhian function does not contain the coordinates q", we have thus eliminated all the cyclic coordinates in Eq. (14) and reduced the problem to r equations of the Lagrange type. EXAMPLES Example 1. A particle of mass m moves on the x axis under the action of a force — kx in a medium that resists the motion with a force ax, where a and k are positive constants. Use Hamilton's equations of motion of the second kind t(\obtain the equation of motion.
748 Equations of Motion of the Hamiltonian Type Solution. The kinetic energy of the system is Hence T=\mx2 s = —- = mx ox 2 K= sx-T=±mx2 = -£- z 2m (23) (24) (25) The x generalized component of the force is Qx = - kx - ax (26) Hamilton's equations of motion of the second kind are dK(x,s,t) x = ds (27) dK(x,s,t) i==--^+& (28) Substituting Eqs. (25) and (26) in Eqs. (27) and (28) we obtain jc = — (29) m v ' s = —kx — ax (30) Combining Eqs. (29) and (30) we obtain mx = — kx — ax (31) Example 2. Use the Routh procedure to eliminate the cyclic coordinates for the case of a symmetric top in a uniform gravitational field, and derive the equation of motion for the noncyclical coordinate. Solution. If we let m be the mass of the top; A, A, and C the principal moments of inertia for the bottom tip of the top; h the distance from the bottom tip of the top to the center of mass; 0, <j>, and \p the Euler angles that describe the orientation of the top with respect to an inertial frame with origin at the bottom tip, and 3 axis vertically up; then the Lagrangian for the top is given by L = \A02 + ^(<j>sin0)2+ ±C(j>cos0 + jf- mghcosO (32) Since the variables <J> and \f/ are cyclic or ignorable, we choose the following Routhian: R=p^+p^-L (33) where /^ = -^ = A$sin20+ Ccos0(<pcosO + ^) (34) P4 = ^ = C(4,cos9 + i) (35)
Problems 749 Substituting Eqs. (32), (34), and (35) into Eq. (33) we obtain R=_ Aii+(p.-p*"»9)2+4+mghcose (36) 2 2,4sin20 2C s ( ' Routh's equations of motion are dt\d9 ) f-° (3?) *=f ^ ^-"f (38b) *-|£ (39a) *"-!? (39b) Substituting Eq. (36) in Eq. (37) we obtain (P* - A/,cos 0 )2cos 6 pAp*- Az,cos 0 ) -A"+ 1J AM '+•+*-* W Substituting Eq. (36) in Eqs. (38) and (39) we obtain p+=c+ and /V = ^ (41) where c^ and c^ are constants. If we substitute Eqs. (41) in Eq. (40) and let a = c^/A and /3 = c^/A we obtain g = mghsing (a - ff cosfl)2cosfl _ fl(a - ftcosfl) ^ sin3^ sin^ ^ which is the same result we obtained in Example 2, Chapter 40. PROBLEMS 1. A particle of mass m moves on the x axis in a medium that resists the motion with a force kx2. Use Hamilton's equations of motion of the second kind to obtain the equation of motion. 2. A mass m slides under gravity on the inside of the surface z = x2 + y2. Gravity is acting in the negative z direction. Use Routh's procedure to reduce the problem in cylindrical' coordinates to one degree of freedom and find the equation of motion. 3- A dynamical system has a kinetic energy T = \{q2 + [qj/(a + bq2)]} and potential energy V= \{kxq\ + /t2), where a, b, kx, and k2 are constants. Obtain the equations of motion using Routh's procedure. 4. Solve the problem of the motion of a projectile moving in a nonresistive medium in the gravitational field of the earth using the Routhian procedure.
73 Point Transformations INTRODUCTION A transformation from one set of generalized coordinates q = qx, q2, . . . , qf to another set of generalized coordinates Q = g,, Q2, . . . , Qf is called a point transformation. A point transformation is represented by the set of equations fi=ft(q,0 The inverse transformation is given by the set of equations * = *(Q>0 (i) (2) The Lagrangian function L(Q, Q, t) is obtained by substituting the relations (2) into the Lagrangian function L(q, q, t). We express this fact by saying that the Lagrangian is invariant under a point transformation. The equations of motion in terms of the coordinates q are given by A dt 3L(q,q,/) dL(q,q,t) H = 0 The equations of motion in terms of the coordinates Q are given by A dt aL(Q,Q,Q 9fi 3L(Q,Q,Q 3ft = 0 (3) (4) Equations (3) and (4) are identical in form. We express this fact by saying that Lagrange's equations are invariant under a point transformation. EXTENDED POINT TRANSFORMATIONS The situation is not the, same in the Hamiltonian formulation. If we make a point transformation, the equations of motion are identical in form, but the Hamiltonians are not necessarily the same in the two sets of coordinates. 750
Extended Point Transformations 751 Theorem 1. Consider a system whose configuration is defined by the set of generalized coordinates q = quq2, • • • , <7/ and whose behavior under the action of a given force is defined by the Hamiltonian function H(p, q, t). If we introduce a new set of generalized coordinates Q = Qu Q2, . . . , £},, which are related to the set q by the point transformation & = G(q>0 (5) then the conjugate momenta are given by p v H(Q.Q and the Hamiltonian K, associated with the set of coordinates Q, is given by ^=^-S/»/—ay— (7) The transformation given by Eqs. (5) and (6) between the set of variables q,p and the set of variables Q, P is called an extended point transformation. proof: We note first that the Lagrangian is invariant under the transformation (5). It follows that SM-tf-S^a-* (8) i i or equivalently 2 p, dqt -Hdt = ^Pi dQt -Kdt (9) i i But <Hi = 2 3fl. ^+ 8f ^ (10) Substituting Eq. (10) in Eq. (9) we obtain SS^^^^+S^-^^^-^^ = S^^-^ (11) i j uty i 0l j Equating coefficients of dQj and dt we obtain Eqs. (6) and (7). From Theorem 1 it follows that if the transformation between the set of coordinates q and the set of coordinates Q does not depend on the time, then K = H. On the other hand, if the transformation between the set of coordinates q and the set of coordinates Q does depend on the time, then generally
752 Point Transformations EXAMPLES Example 1. The configuration of a certain dynamical system is defined by the generalized coordinate q and its dynamical behavior by the Hamiltonian function H(q,p,t) = ap2+bp(q+tf (12) where a and b are constants. A point transformation is made to a new generalized coordinate Q = q+t (13) Obtain the Hamiltonian function K( Q, P, t). Solution. From Theorem 1 we obtain p-pJw1-p (14) K-H-p^-L-H + p (15) Combining Eqs. (12)-(15) we obtain K(Q,P,t) = aP2+bPQ2 + P (16) PROBLEMS 1. Given the point transformation from Cartesian coordinates jc, y, and z to spherical coordinates r, 0, and <j>, determine pr, pg, and p^ as functions of x>y>z>Px>Pr and/?z. 2. The Hamiltonian function for a certain system whose configuration is described by the generalized coordinates qx and q2 is given H=(£^f + p2 + (qi + q2)2 Solve for qx and q2 as functions of the time by making use of the point transformation Qx = q2, Q2 = qx + q2> 3. A particle of mass m is moving in the x-y plane under the action of a force that is derivable from the potential V(r), where r2 = x2 + y2. A point transformation is made from the coordinates x, y to a new set of coordinates x', y' defined by the equations x' = x cos cot + y sin cot, y' =5 — x sin cot +y cos w/. Find the Hamiltonian functions H(x, y, px, pyJ) and HXx',y',px,,py,t).
SECTION Canonical Transformations
.^■JiB
74 Canonical Transformations INTRODUCTION In solving dynamical problems it is frequently convenient to transform from the set of generalized coordinates q = quq2, • • • , q/ and the set of generalized momenta p = /?,, p2, - . . , Pf to a new set of variables Q, P = Qx, Q2, . . . , Qp P]9P2, . . • , Pj that are related to the set of variables q,p by transformation equations of the form Q = Q(q,p,r) P = P(q,p,/) (1) Transformations of this form do away with any distinctions between coordinates and momenta, since it is possible within the framework of this form of transformation to interchange the role of coordinates and momenta. Nevertheless we shall for simplicity retain the term "coordinates" for the variables Q and the term "momenta" for the variables P. Out of all the possible transformations of the form given by Eq. (1) there is a subset of transformations called canonical transformations, whose properties are particularly useful in dynamics. CANONICAL TRANSFORMATIONS A time independent transformation Q = Q(q,p) is said to be a canonical transformation ^(q,p) such that ^(q>p) = 2/>/^/- i A time dependent transformation Q = Q(q,M) P = P(q,p) (2) if and only if there exists a function 2^(q.P)^fi/(q.p) (3) i P = P(q,M) (4)
756 Canonical Transformations is said to be a canonical transformation if and only if any one of the following three equivalent statements are true: 1. The transformation is canonical in the previously defined sense for every value of the time. 2. There exists a function F(q, p, t) such that for arbitrary fixed time t = tn where and dF{* P> h) = £ Pi da~ 2 pi{% P> *o) dQi{* P> M «^(4 P> 'o) = 2, jTI ^ft + Zj ^ <fo H dPi <%(q, P- M = 2 9^ ^-+ 2 9^ dPj * = 2 ^/(4 P> 0 + (5) (6) (7) 3. There exists a function F(q, p, t) such that ^(q.p.0 = SflA- 2^(4M)<%(4P»9 + *(<hm)<// (8) where a* a? (9) The equivalence of conditions 1 and 2 follows directly from the definition of a time independent canonical transformation. The equivalence of conditions 2 and 3 can be seen by noting that ^(q>Mo) = <%■(*&'(>) = dF(q,p,t)- 3/ A <%-(*P'0"—^—* (10) (11) If we substitute Eqs. (10) and (11) into Eq. (5) and note that it is true for arbitrary t0, we immediately obtain Eq. (8). If we are given a particular time independent transformation and we know that it is canonical, we can determine the function F(q, p) within an arbitrary additive constant by simply integrating Eq. (3). If the transformation is a time dependent canonical transformation we can similarly determine F(q,p, 0 within an arbitrary additive function of the time by integrating Eq. (5) and replacing t0 by t. To see the latter point suppose that we integrate Eq. (5) from
Definition of a Canonical Transformation by Lagrange Brackets 757 some known fixed point q0,p0 to the point q,p. We then obtain F(q, P> 'o) " ^(qo> Po> 'o) = fq'P'/0 [ 2 Pi d<li ~ 2 ^/(4 P> 'o) <%(4 p, *0) (12) Since the time t0 is arbitrary, we can replace t0 by t in Eq. (12) and we obtain an expression for F(q,p,/) in which all quantities are known except the quantity F{%,^0J). Thus we have determined F(q,p,t) to within an additive function of the time. If we wanted to use the definition above to determine whether a given transformation was canonical, we could set up the differential expression ^iPi^i:~ 2/^/(4P>'o)^2/(4P>'o) and use the results of Appendix 11 to determine whether the differential was exact. In the following section we develop an alternative definition of a canonical transformation that is based on this procedure and provides us with a practical norm for testing the canonical character of a transformation. DEFINITION OF A CANONICAL TRANSFORMATION BY LAGRANGE BRACKETS Let r- and rk be any two members of the set of variables q,p = q\,q2, • • • > ?/> P\,p2> • • • j/y- Then the Lagrange bracket [rj,rk] associated with the transformation is defined as Q = Q(q,p,/) P = P(q,p,/) (13) 30/(4 P>0 3^/(4*0 30-(4P.0 3^/(4*0 drj drk drk dry. (14) Theorem 1. The transformation Q = Q(q,p,0 P = P(q,p,0 (15) is a canonical transformation if and only if the Lagrange brackets [qi9qj], [?/> Pj] and [pi,Pj] satisfy the following conditions: [%<lj]=0 '[qi>Pj] = 8y [P,Pj]-0 (16) where 8.. is the Kronecker delta. That is, if i = j then 8U = 1, and if i ^ j then % = 0. J proof: Throughout this proof we will for simplicity suppress the arguments of the functions, which in all cases are the set of variables q, p, and t. Furthermore the variable t is to be constant throughout. The transformation (13) is a canonical transformation if and only if there exists a function F such
758 Canonical Transformations that dF=^Pidq-^PidQi -2^-2^12^ + 2^ If we define and ?(*-?^)*+?(-?'*f)* w Aj-Pj-^P*-^ (18) *>'-¥*-$ (19) then we can rewrite Eq. (17) dF=^lAJdqj+^Bjdpj (20) j J But from Theorem 1 in Appendix 11 a necessary and sufficient condition that there exists a function F such that Eq. (20) is true, is that dAj _ QAj dqt dqj dAj _ SB, dPi ~ dqj 95,- _ dBj If we substitute Eq. (18) in Eq. (21) we obtain (21) (22) (23) 4MMtK(*-?M£) <24) Carrying out the derivatives on both sides we obtain 3£,3&_ _^ = _ a^a&_v 3¾^ 25) f h d% YkHH it H H it kHH K Rearranging Eq. (25) we obtain it H H it 9¾ H {
Some Properties of Canonical Transformations 759 But the left hand side of Eq. (26) is just the Lagrange bracket [^,¾]. We thus have [<?/><?,] =0 (27) We similarly obtain from Eq. (22) Ui>Pj} = 8ij (28) and from Eq. (23) [Pi>P,]-<> (29) EXTENDED POINT TRANSFORMATIONS In the preceding chapter we introduced the concept of an extended point transformation defined by the equations ft.= fi(q,0 and P, = % Pj-1±> (30) If we apply Theorem 1 to this transformation it can be shown that the transformation is canonical. We could prove the same fact more directly by noting from Eq. (9) in the preceding chapter that for fixed time ^ iPidq( — Tii^idQi = 0 and therefore there exists an F(q,p,t), namely, any function of the time only, satisfying the requirements of the definition of a canonical transformation. POISSON BRACKETS It is customary at this stage in the discussion of canonical transformations to introduce the subject of Poisson brackets. Although they are useful in dealing with canonical transformations, we do not really need them at this point. The machinery we have already introduced is adequate for our immediate purposes, and the introduction of additional machinery would only overburden the subject. We therefore relegate the subject of Poisson brackets to a later chapter, where it is treated in complete detail and its usefulness is more obvious. SOME PROPERTIES OF CANONICAL TRANSFORMATIONS In this section we prove a number of useful properties of canonical transformations. Theorem 2. If the transformation Q = Q(q,p>0 P = P(q,M) (31)
760 Canonical Transformations is a canonical transformation, the Jacobian of the transformation is equal to ± 1, that is, v q>p I (32) (we later show that 7=+1, but the present weaker theorem is sufficient for our immediate purposes). proof: We note first that 9Si 3?i V <1>P I 9^ dqf 92, 92, 9/>/ ZQf 9P, 9^i 3^i 92^ 9P, dqf dqf HZf 9^ 3/7, 3/>, 9^ 3P, 9/>/ 9/y 9?, dqf dPi dPf (33) We now perform the following sequence of operations on the determinant in Eq. (33): (a) multiply the bottom set of/rows by - 1; (b) multiply the right hand set of/columns by - 1; (c) interchange the bottom set of/rows with the top set of/rows; (d) interchange the right hand set of/columns with the left hand set of/columns; .(e) interchange the rows and columns. The net result of all these operations leaves the value of the determinant unchanged, but changes its form as follows Q,p\_ q>p ) El dpx 9/>i 92, 9/>, 9/>i El 9/>/ 9Pf dPf 92, 9/>/ _ 9¾. 9/>/ _El 9?i _E 9^, 92i 9?, 92/ dq, _El dqf _E dqf 92i 3¾. 9^ dqf (34)
Some Properties of Canonical Transformations 761 If we now multiply Eq. (33) by Eq. (34) we obtain am [p\>p\] [pppi] 1 0 0 1 0 0 0 0 [9/> pf][qi>qf] [pvPf][quPx] [pj>Pf][qi>Pf] [is**] iippi] [fyPf] o o 0 0 1 0 0 1 = 1 (35) Equation (32) follows immediately from Eq. (35). Theorem 3. If the transformation Q = Q(q,P>0 P = P(q,p,0 is a canonical transformation, the inverse transformation q = q(Q,P,0 p = p(Q,P,/) exists and is also a canonical transformation. (36) (37) proof: According to the results of Appendix 18 the inverse of the transformation given by Eq. (36) exists if v q>p I ^o (38) From Theorem 2 this condition is satisfied, hence the inverse exists. Since the transformation given by Eq. (36) is canonical, it follows that there exists an F(q>p,0 such that ^(q.p.'o) = 2 Pi*i- 2 ^*p>'o)<%-(*p>'o) (39) If we substitute the inverse transformation, Eq. (37), into Eq. (39), we find that the function - F(Q, P, /0) satisfies the relation d[-F(Q,P,t0)] =^lPidQ- 2 p,(Q,P,t0)dq,(Q,P,t0) (40) It follows immediately that the inverse transformation is a canonical transformation: This completes the proof of the theorem. Theorem 4. If the transformations q"=q"(q',p',f) p" = P"(q',p'>0 (41)
762 Canonical Transformations and q' = q'(q>p>0 p' = p'(*P'0 (42) are canonical transformations, the product of the two transformations, that is, the transformation q" = q'(qp>0 p" = p"(q>p>0 (43) is a canonical transformation. proof: If the transformation (41) is a canonical transformation there exists a function F'(q', p', 0 such that for fixed time t = t0 dF'Spldqi-'Zprdqr (44) i i and if the transformation (42) is a canonical transformation there exists a function F(q, p, t) such that for fixed time t = t0 dF=%pidq-%p'idq'i (45) i i Adding Eqs. (44) and (45) we obtain d(F + F') = 2 P, dq~ S P" dq" (46) i i It follows immediately that the transformation (43) is canonical. In addition to Theorems 3 and 4 one can immediately prove the following two theorems, which we simply state without proof. Theorem 5. The identity transformation Qi = qi9 Pi = pt is a canonical transformation. Theorem 6. The set of canonical transformations obeys the associative law of multiplication. It follows from Theorems 3-6 that the set of canonical transformations forms an algebraic group. GENERALIZATION OF THE DEFINITION OF A CANONICAL TRANSFORMATION The definition of a canonical transformation given at the beginning of this chapter can be generalized in a way that puts the different variables on a more equal footing. We now consider this generalization. Consider a system whose dynamical state is defined by the set of variables q,p. We use the notation jt,, j, to represent either the pair of variables qi9pi or
Generalization of the Definition of a Canonical Transformation 763 the pair of variables pi9 — q(. Thus *t = ft OTPi y<=Pi if */ = ?/ (47) = "?/ ^ */=/>/ Similarly if Q, P is a second set of variables defining the dynamical state of the system, we use the notation Xi9 Yi to represent either the pair of variables Q.9 Pt or the pair of variables Pi9 — Qt. The notation above allows us to generalize our definition of a canonical transformation as follows. Theorem 7. Consider a system of/degrees of freedom whose dynamical state is defined by the set of variables q, p. Let Q, P be another set of 2/ variables defined by the transformation equations Q = Q(q,p,') P = P(q,p,0 (48) If we let xi9 yt = qi9 pt or pi9 - qt and Xi9 Y) = Qi9 P, or Pi9 - Qi9 the transformation is canonical if and only if for some arbitrary choice for the set of variables x, X there exists a function F(q, p, t) such that for arbitrary fixed time t0 dF(q, p, f0) = 2 yi dx~ 2 ^/(4 P> 'o) dXi(* P' ^o) (49) i i We refer to the function as the F function for the given choice. If there exists an F function for one choice of x, X, there exists an F function for all choices of x,X. The F function is different for each choice of x,X. If x',X' and x",X" are two choices for the set of variables x,X that differ only in that for a particular k9 x'k = qk but x'k' = pk9 and if F' and F" are the F functions for the two choices, then F" = F'-pkqk (50) If x',X' and x",X" are two choices for the set of variables x,X that differ only in that for a particular k9 X'k = Qk and X£ = Pk9 and if F' and F" are the F functions for the two choices, then F" = F' + PkQk (51) proof: Suppose that x',X' and x",X" are two choices for x,X that differ only in that xk = qk but xk = pk9 and there exists an F' such that dF^^yidxi-^YfdXf i i = 2 yi dx\ + Pk dqk - 2 y/ dX! (52)
764 Canonical Transformations It then follows that d{F' - pkqk) = 2 y[ dx[ - qk dpk - £ r/ <«/ i¥= k = 2//' dx»-%Yi"dXi" (53) i i Hence there exists a function F" = Ff — pkqk such that dF" = ^y" dx" - 2)/ Y" dX". Conversely if F" exists, F' exists. In a similar fashion we can show that if x',X' and x",X" differ only in that X'k = Qk but Xk" = Pk, then if there exists an F\ there exists an F", and if there exists an F", there exists an F\ and F" = F' + PkQk. Since any two choices of the set of variables x,X can be connected by a succession of exchanges of the forms <7,->#, #-» qi9 6/-»P., or P, -> Qi9 it follows that if there exists an F function for one choice of the set of variables x,X, there exists an F function for all choices of the set of variables x, X. The transformation (48) is by definition canonical if and only if there exists an F function for the choice q, Q of the set of variables x, X. It immediately follows from this definition and the analysis above that if the transformation is canonical, there exists an F function for every choice of the set of variables x,X; and if there exists an F function for some choice of the set of variables x, X, the transformation is canonical. This completes the proof of the theorem. EXAMPLES Example 1. Show that the transformation Q=p2+q P = p + t (54) is a canonical transformation and then find a function F(q,p,t) such that dF(q, p, t0) =pdq- P(q, p, t0) dQ(q, p, t0). Solution. The Lagrange brackets for the transformation are ["]-#£-$£" (57) From Theorem 1 it follows that the transformation is canonical. Since the transformation is canonical, there exists a function F(q, p, i) such that dF(q, p, t0) =pdq-P(q, p, t0) dQ(q, p, tQ) (58) Substituting Eqs. (54) in Eq. (58) we obtain dF{q, p, t0) =pdq-{p+ t0) d(p2 + q) = -tQdq-2p(p+t0)dp (59)
Problems 765 Integrating Eq. (59) at time t0 along any path from some reference point q0, p0 to the point q, p we find that F(<1> P> *o) =-^0- 2P3/3 ~ p\ + fcn(?0, p0910). (60) Setting t0 = t we obtain F(q, p,t)= ~qt-\ -p2t + fcn(f) (61) The function F(q, p, t) has the desired property for any choice of fcn(/). PROBLEMS 1. Prove that the following transformations are canonical transformations by using the Lagrange bracket test: a. Q = q2 + (p2/n2) i>= | tan"'[/>/(«<?)] b. <2i = q\ + p\ Q2 = 12(q2 + q\ + p2 + pi) '.->-(£)->-(£) '--It) 2. Prove that the transformation e = ln(^) P-qcotp is a canonical transformation by using the Lagrange bracket test, and find a function F(q, p) such that dF(q, p)= pdq — P(q, p)dQ(q, p). 3. Given the transformation Q = \np P= -(1 + q + In p)p a. Show that it is canonical by applying the Lagrange bracket test. b. Show that it is canonical by finding a function F(q, p) such that dF(q> p)=pdq- P(q, p)dQ(q, p). 4. Show that the transformation Q=C(2q)l/2cosp P=C-\2q)l/2smp where C is a constant, is a canonical transformation. 5. Show that the transformation Q = q + tep P=p is a canonical transformation, and find a function F(q,p,t) such that dF(q, P> 'o) =pdq- P(q, p, t0) dQ(q, p, t0).
766 Canonical Transformations 6. Show that the transformation Q = {p2 + q2)U1 P = {P2 + <12)X/2 tan m is a canonical transformation, and find a function F(q, p, t) such that dF(q, p, t0) =pdq- P(q, p, t0) dQ(q, p, t0). 7. Show that the transformation Q2 = pl + q2 Pi=Pi+t is a canonical transformation and find a function F(q,p,t) such that dF(q,p, g = 2tPidqt - 2^(q,p, /0)<%(q,P> * q). 8. For what values of a and /? do the equations Q = qacos fip and P = qasm pp represent a canonical transformation? 9. Show that the transformation Q\ = «11^1 + <*12?2 £>2 = alxqx + a22?2 Pl = h\\P\ + ^12^2 P2 = *2l/>l + ^22^2 is a canonical transformation provided Z^- = |^^|/|^| for all i, j, where 1^41 is the determinant of the matrix [a^] and \Ay\ is the cofactor of the element a^, and all the a^ and b^ are constants. 10. Prove that the transformation Q = ln(l + q]/2cosp) P = 2(1 + ^1/2cos/?)^1/2sin/7 is a canonical transformation by finding a function F\q, p) such that dF' = -qdp - P(q, p)dQ(q, p) and also by finding a function F"(q,p) such that dF"(q, p) = - qdp + Q(q, p)dP(q, p). 11. Let Q\ = q\ p\ = P\{q^q^PuPi) Qi = q\ + qi Pi = Pi(q\>qi> PuPi) be a canonical transformation. Complete the transformation by finding the most general expressions for Px and P2. Show that a particular choice
Problems 7tf7 will reduce the Hamiltonian function v2 H = (%f1)+P^(^ + ^2 to H= Pf + P2 12. Consider a particle subject to the force F--kq-± 9 Show that this system is described by the Hamiltonian 2m 2 q where A must be a properly chosen constant. Show that the transformation below is a canonical transformation. where R(q,p,t) is a function homogeneous of degree zero in the variables q and p, but otherwise arbitrary. Use the transformation above to solve for the motion of the particle. 13. If <2 = tan-^j P=P{q,p,t) is a canonical transformation, show that where R is a function homogeneous of degree zero in the pair of variables q, p but otherwise arbitrary. Use this transformation with R appropriately chosen to simplify the problem of the simple harmonic oscillator of mass m and angular frequency o) = X/m.
75 A Condensed Notation INTRODUCTION In the chapters that follow we make extensive use of canonical transformations. In this chapter we introduce a notation that shortens and simplifies many of the long and tedious proofs that we encounter in the theorems. THE SUMMATION CONVENTION In this chapter and whenever the notation introduced in this chapter is used we employ the summation convention; that is, we assume that whenever repeated indices occur in a term they are to be summed over. With this convention the expression 2/tfA, f°r example, is written as simply a^. Indices that are repeated are called dummy indices. Since dummy indices are to be summed over, it is immaterial what letter is used for the indices. Thus, for example, the quantity aibi can be written as a-b- or akbk. THE NOTATION Consider a system of/degrees of freedom whose dynamical state is defined by the sets of variables q = qx, q2, . . . , qf and p = /?,, p2, . . . ,/y. We define the set of variables r = r,,r2, . . . , r2f by the matrix equation r r\ . rv_ = ' 9\' pi _pf\ 768
Hamilton's Equations and Canonical Transformations and the 2/X 2/matrices [Xtj] and [^] as follows: [w/] = 7» (2) (3) where Or is the / X / null matrix and 1^ is the / X / unit matrix. Alternatively we can define ri9 Xy, and ^ as follows: *■/ = ft =/>,-/ \, = 1 = 0 My = 1 = -1 = 0 »=i,...,/ /-/+1,...,2/ /='+/ otherwise /=«'+/ /=«-/ otherwise The matrices [A,.] and [ p^] have the following properties: w- l>y "[VT= l>;] [%]T=-[M//] ][^]r=[^] 1^1 = 1 (4) (5) (6) (?) (8) (9) (10) where [ ]T represents the transpose of the matrix [ ]. Alternatively we can state the first three of the results above as follows: \j ~ ^// " ft/ ft/ = - fy ft/ft* = °jk (ii) (12) (13) HAMILTON'S EQUATIONS AND CANONICAL TRANSFORMATIONS In terms of the notation above the results of the preceding chapters can be expressed as given below. Hamilton's equations are n = ^ dH ■J dr, (14)
770 A Condensed Notation A transformation R = R (r, t) is defined to be canonical if and only if there exists an F(r, t) such that dF(r,t0) = \jrjdr, - X^y(r,/0)^/(r^o) (15) The Lagrange bracket [rk,rl] associated with the transformation R = R(r,r) is defined as lr»rA-*i-wk ^- (16) Theorem 1. The transformation R = R(r, t) is canonical if and only if the Lagrange brackets [rk, r{] satisfy the condition [rk>n] = /**/ (17) EXAMPLES Example 1. Prove Theorem 1 using the condensed notation introduced in this chapter. Solution. We note first that if the transformation is canonical there exists an F such that dR dF = \.rjdrt - \.R. dRt = \.rjdrt - XkjRj -^- drt = \kjrjdrk-\ijRjj^drk -(V/-M/^K (18) kj j U J $f J * A function F satisfying the relation above will exist if and only if _8_ dr, l;(v,-V$)-£(v,-V*$) (19) Carrying out the differentiations we obtain 9rj dRj dR, d% kJdr, y dr, drk V J dr,drk -^-^^-^ ** (20) u drk lJ drk 9r7 J J drk 9r7
Noting that d2R, ^ d2R, we obtain 9 'dr,drk » Jdrkdr, A MM=A MH 'Jdrk dr, J'drk dr, dR; dRj Making use of Eq. (11) we obtain dR, dRj _ Examples 771 (21) (22) (23) (24) (25) (26) which is the desired result. Example 2. Use the condensed notation to prove that the Jacobian for a canonical transformation is ± 1. Solution. Given a transformation R = R(r,0 we wish to show that if the transformation is canonical, then a*, 90 = ±1 (27) (28) where |3ft/3ry| is the determinant of the matrix [3ft/3ry]. If the transformation is canonical, the Lagrange bracket conditions give us ft 3ft 3ft _ y drk ~3^ " ftt/ In matrix form the equation above can be written 3*, 3r, [ty] 3*, dr. -[ty] (29) (30) where the summation convention is obviously not being employed. If we take the determinant of both sides, remembering that the determinant of the Product of matrices is equal to the product of the determinants of the matrices
772 A Condensed Notation and that the determinant of a matrix is equal to the determinant of the transpose of the matrix, we obtain 90 1 ft/I 90 = I /tyl (31) and thus 3i?, = 1 (32) from which Eq. (28) follows.
76 Hamilton's Canonical Equations of Motion INTRODUCTION In the preceding chapter we introduced the concept of a canonical transformation. Now we consider the effect of a canonical transformation on the equations of motion. CANONICAL VARIABLES Prior to the considerations of the preceding chapter we restricted ourselves to situations in which the configuration of a system was defined by a set of generalized coordinates q, and the dynamical state of the system was defined by the set of generalized coordinates q together with either the set of generalized velocities q or the set of generalized momenta p. If we make a canonical transformation from the set of variables q, p to the set of variables Q, P, the dynamical state of the system can be defined in terms of the new set of variables Q, P; but because the canonical transformation destroys the distinction between coordinates and momenta, the set of variables Q is no longer restricted to sets that define the .configuration of the system. The use of canonical transformations has thus opened up a whole new realm of possible formulations for a given dynamical problem. Any set of variables Q, P that can be associated, by means of a canonical transformation with a set of generalized coordinates q, and the corresponding set of generalized momenta p, is called a set of canonical variables. As we see in the next section, the equations of motion of a system in terms °f a set of canonical variables can always be expressed in a form that is !dentical to Hamilton's equations of motion. It is therefore customary to call the resulting equations simply Hamilton's equations of motion. However to 773
774 Hamilton's Canonical Equations of Motion emphasize the extended range of validity of these equations, we refer to them as Hamilton's canonical equations of motion. HAMILTON'S CANONICAL EQUATIONS OF MOTION One of the most useful properties of a canonical transformation is contained in the following theorem. Theorem 1. Consider a system that is acted on by a given force. Let the dynamical state of the system be defined by the set of variables q,p = qx, . . . , qpPi, . . . ,pp and let us assume that there exists a function H(q, p, t) such that the behavior of the variables q, p under the action of the given force is determined by the equations of motion 37/(q,p,0 *-—5S— (1) Pi=-—H~ (2) If we make a transformation to a new set of variables Q = Q(q,p,0 P = P(q,p,0 (3) and if the transformation is a canonical transformation, that is, if there exists a function F(q, p, t) such that for arbitrary fixed time t = t0 dF(q,p,t0) = ^lpidqi-^Pi(q,p,t0)dQi(q,p,t0) (4) i i then the equations of motion in terms of the variables Q, P can be written Q.JHS^l (5) 9ft where gsg+af(,,P,0+2pafl(,.P.,) (7) Furthermore if the determinant of the matrix [3(?/(q,p,0/3/>y] is not zero' Eq* (7) can be written K_H+»F(VQ,,) (8) ot
Hamilton's Canonical Equations of Motion 775 proof: Using the condensed notation introduced in Chapter 75 and adopting the summation convention, we are given the equation of motion 9H(r9t) ^ = ^-^1 <9> and we wish to prove that if we make a transformation to a new set of variables R = R(M) (10) and if there exists a function F(r, t) such that dF(r,t0) = \gfjdr, - XyRjQctddRtfrto) (11) then the new equations of motion are dK(R,t) where Throughout this proof for simplicity we suppress the arguments of all functions. The arguments in all cases are either the set of variables r, t or the set of variables R, t. The choice in a particular case will be obvious from the context. We note first a.M = aM^± (14) ^3R. n>drk dR, l ' j K j Using Eq. (13) we obtain _dH , d (dF\,x dRmdR, d2R, Rewriting Eq. (11) as follows: dF = (\kmrm - \lmRm -^- j drk (16) we obtain dF-\ r -\ P dRl ~ Akmrm Alm*m "aT- g^ Akmrm AlmKm "g^T (1 /)
776 Hamilton's Canonical Equations of Motion Substituting Eq. (17) in Eq. (15) we obtain M = M_x ^M + a dR» dR' drk drk lm dt drk lm drk dt -*H + a x \dR» dR> --^r + (A,m Am/) -^7--97- _ dH . Mm 3*/ " drk + fl*"~3^" dt (18) But dH _ « dH _ „ „ 37/ 3rft " w~3^ " ^^ "317 = ft**» (19) Since the transformation is canonical, (20) (21) Substituting Eq. (20) Substituting Eq. (21) Substituting Eq. (22) V>nk = [rK in Eq. (19) 37/ in Eq. (18) 3¾ ^/m = ^/m in Eq. (14) ^ dR,. »rk\ — V'lm we obtain dR/ we obtain • 3^m n m 3¾ we obtain = ftjPimRr dR, *Rm 9rk dR/ "37 9rk *Rm *rk K \dR, ) 9rk drk dRj (22) = ^=^ (23) This proves the body of the theorem. To complete the proof we note that if the determinant of the matrix [3<2/(q,p,0/9p/] *s not zero> tnen according to the results of Appendix 18 we can invert the relation (?(q,p, 0 to obtain p(q, Q>0> hence the function F(q, Q, t). We then note 3/ ^ 3ft 3/3/ l
Hamilton's Canonical Equations of Motion 777 But from Eq. (4) 3F(q,Q,Q 3ft = ~ Pi (25) If we substitute Eq. (25) in Eq. (24) we obtain 3F(q,p,Q_ ^p3a(q,p,Q , 3F(q,Q,Q 3/ ^ ' 3/ 3/ ( ' Combining Eqs. (7) and (26) we obtain Eq. (8). This completes the proof of the theorem. Corollary 1. Consider a system that is acted on by a given force. Let the dynamical state of the system be defined by the set of variables q,p = Q\y •••>?/> P\> - - - >Pf and let us assume that there exists a function #(q,p, t) such that the behavior of the variables q,p under the action of the given forces is determined by the equations of motion *—W~ (27) 3#(q,p,Q pt - —o^r~ ( } If we make a time independent transformation to a new set of variables Q = Q(q,p) P = P(q,p) (29) and if the transformation is canonical, the equations of motion in terms of the variables Q, P can be written Qi Jp-^ (30) dH(Q,P,t) ''-—IS"2 (31) proof: If the transformation above is canonical, there exists a time independent function F(q,p) satisfying Eq. (4). If we use this function in Eq. (7) we obtain K = H. The corollary then follows immediately from Theorem 1. The converse of Theorem 1 is not true. That is, if the equations of motion of the system in terms of the variables q, p can be written in the form given by Eqs. (1) and (2), and the equations of motion in terms of the variables Q,P can be written in the form given by Eqs. (5) and (6), it does not necessarily follow that the transformation (3) is a canonical transformation. Many authors define a canonical transformation as a transformation for which the abpve is true. If such a definition is used, the theorems of Chapter 74 are not
778 Hamilton's Canonical Equations of Motion strictly true and must be appropriately qualified before they can be used. In Theorem 2 we take up the conditions under which the converse is true. Nothing in the later chapters depends on this theorem, and the reader may omit it without loss of continuity. Theorem 2. Consider the transformation a = a(q,p,/) /* = /*(q,p,0 (32) If for every function #(q,p, 0 there exists a function y(a, )3,/) such that the equations 3#(q,p,0 3#(q,p,0 are equivalent to the equations *= 8ft A 3^- (34) then the Lagrange brackets for the transformation are given by [q„qj]=0 [%Pj] = b8iJ [Pi,Pj]=0 (35) where b is a constant. It follows that the transformation is a canonical transformation and the equations d _ a*(Q.P.Q ,_ a^(Q>P.Q r37. where *=| (38) are equivalent to the equations (33). proof: We use the condensed notation introduced in Chapter 75 and also the notation p for the set of variables a, )3; we suppress the arguments of all functions. Given P = P(M) (39) and let us suppose that there exists a function y(r, t) such that p< = * 8p aT (41)
Hamilton's Canonical Equations of Motion 779 It then follows that 3y = dy_ 9p* = 3y 9p* = dy dpk drj dpk drj lkd9l drj ^^3p/ fr. dpk ( 9Pm . , 3Pm If we define 9P/t . _ 9P/t / 9Pm . , = 0,1+^0^ = [^o] + [^o]^|f (42) ^['.'jl (43) l>jk=[rhrj]nlk (44) then we can rewrite Eq. (42) as 37 = ^/ + ^37: (45> From Eq. (45) we obtain 81-,8/). 3/-, f drt 8/-, f jk ^rk 3a. _ 36 7y* 37/^^^a , VH ar + 2-af |f + 22M*-ark (46) Interchanging the i and they we obtain 32Y 3a, , v 3Z>„ 37/ , v v, ^ 82ff ,47^ 3/-,3/-, 3/) f 3/)- 3/-, t « '* ^3^,3¾ Equating Eqs. (46) and (47) we obtain (48) Equation (48) must be true for any /f, and therefore 3a,- 3a, ^-^=° (49) 8/-, 8/) v ' dbik 3¾ ^--^f = 0 (5°) (bJk8,m ~ bikSjm) + (bjm8ik - bim8Jk ) = 0 (51) Equation (51) was obtained by setting the sum of the coefficients of the mk terni and the km term in the last term in Eq. (48) equal to zero. The left side of
780 Hamilton's Canonical Equations of Motion Eq. (49) can be rewritten daj 9a, 3 r, i 3 3r,- orj ort l JA 3^- L J 3 I dPk 9P/\ 3 / 9P* 9P/ " 3^fe dt drj) drj^" dt dr: d2Pk 9P/ _ d2Pk 9P/ " ^3^37 3^ fe 32p* 9P/ 3/)3/ 32P, 3/)3/ 3P/t < 3/-,. 3P* 3/-,. fyj_ ^3/-,3/ 3ry fe^,a, a, ^3/3/-,. 3/)^3/-,. 3/3/) _ 3 / 9P* 9ft \ _ 3 r -, " 37[^'J?: 3^j" a7L'V'OJ Substituting Eq. (52) in Eq. (49) we obtain from which it follows that Combining Eqs. (44) and (54) we obtain 1)7=° Let us now consider Eq. (51). If we let m = k = i we obtain If i ^= j then *yA "My,+ *,,«,,-My/= 0 fy + fy = o Thus if i 7^y then by = 0 and therefore we can write by = bu8y Substituting Eq. (58) in Eq. (51) we obtain (bMSJk8im - bit8ik8jm) + (bjj8jm8ik - bu8im8Jk) = 0 which can be rewritten (¾-*«)(*;**'«+ M»)"° Suppose j =^ i and we let k = j and m = i, then (¾-Z>,,)(1+0) = 0
A Generalization of Hamilton's Canonical Equations of Motion 781 and thus b^b^b (62) Substituting Eq. (62) in Eq. (58) we obtain by = My (63) Substituting Eq. (63) in Eq. (55) we obtain f = 0 (64) Substituting Eq. (63) in Eq. (50) we obtain 9%-f8* = ° (65) dr, "* a j Letting k=j and assuming that i =£j in Eq. (65) we obtain jjS-0 (66) From Eqs. (64) and (66) it follows that b = constant (67) Substituting Eq. (63) in Eq. (44) we obtain [rhrj}iilk = b8jk (68) Multiplying Eq. (68) by \imk and simplifying the result we obtain [^0-] = b^J (69) If we recast Eq. (69) in terms of the variables q and p we obtain [ *, qj] = 0 [ ft, Pj] = bSik [ Pj, Pk]=0 (70) We have thus proved the body of the theorem. The proof of the remaining part of the theorem is straightforward and is left to the reader. A GENERALIZATION OF HAMILTON'S CANONICAL EQUATIONS OF MOTION In terms of the generalized definition of a canonical transformation given in the preceding chapter, we can reformulate Theorem 1 as follows. Theorem la. Consider a system that is acted on by a given force. Let the dynamical state of the system be defined by the set of variables q,p= qx, ?2> . . . , qp px, p2, . . . ,pp and let us assume that there exists a function ^(fl> P> 0 such that the behavior of the variables q, p under the action of the
782 Hamilton's Canonical Equations of Motion given force is determined by the equations of motion 3#(q,M) q'-—9jT~ (71> dH(n,p,t) » h~ (72) If we make a transformation to a new set of variables Q = Q(q,p,/) P = P(q,p,/) (73) and if the transformation is a canonical transformation, that is, if there exists a function F(q, p, t) such that for arbitrary fixed time t = /0, dF{i\, p, /0) = 2 yM ~ 2 Y,dX, (74) /' i where x„ y{ = q„ /?, or pt, -q, and X„ Y, = 0,,^, or P.„ - Qt, then the equations of motion in terms of the variables Q, P can be written 9AYQ,P,0 a = a/> (75) 8g(Q,P,Q p'-" aa (76) where KSH+^+2Yi^ iV) Furthermore if the determinant of the matrix [dXj(q,p,t)/d)>j] is not zero, Eq. (77) can be written dF(\,X,t) K=H+ {d( } (78) proof: The proof of this theorem is perfectly analogous to the proof of Theorem 1 and is left to the reader. EXAMPLES Example 1. The Hamiltonian function for a simple harmonic oscillator can be written H(p,q) = HP2 + »V) (79) Show that if we make a canonical transformation to a new set of variables Q and P defined by the equations . f=esin(arf-^) (80) p-aQcos^t-JL) (81)
Examples 783 the Hamiltonian function for these variables is given by K(Q,P) = 0 (82) Use Hamilton's equations to find Q and P as functions of the time, and use your results to obtain q and p as functions of the time. Solution. If we solve Eqs. (80) and (81) for Q we obtain Q = (p2 + <o2<72)1/2/<*>- Since dQ(q, p,t)/dp is not zero, we can use the relation ,.*+^*m (83) to find the new Hamiltonian. From the definition of a canonical transformation the function F(q, 0, t) satisfies the relation dF(q, 0, t0) = p(q, 0, t0) dq - P(q, 0, t0) dQ (84) From Eqs. (80) and (81) we obtain p(q,Q,t0) = o(Q2-q2)]/2 (85) P(q> Q,*o) = "2Qt ~ "2 sin"1 -| (86) Substituting Eqs. (85) and (86) in Eq. (84) we obtain dF(q, 0, t0) = o( Q2 - q2f2dq - (co2Qt0 - oQ sin"' -| j dQ (87) Integrating Eq. (87), replacing t0 by t, and dropping the unknown additive function of the time, we obtain -\"2Q2t (88) F(q, Q,t) = I«[f( Q2 - q2f2 + g'siiT1^ Substituting Eq. (88) in Eq. (83) we obtain K= H-\u2Q2 (89) Combining Eqs. (79)-(81), and (89) we obtain K = 0 (90) Hamilton's canonical equations for the variables 0 and P are 3/qe,iV) 3g(g,/»,Q g= 9^— .and P ^- (91) Substituting Eq. (90) in Eqs. (91) we obtain g = 0 P = 0 (92) Hence 0 = A P= B (93) where ^-and B are constants. Substituting Eqs. (93) in Eqs. (80) and (81) we obtain q = Asin(ot- C) (94) p = o)Acos(o)t - C) (95)
784 Hamilton's Canonical Equations of Motion where C = A/coB. This is the same result we would have obtained if we had applied Hamilton's equations to the Hamiltonian function H(q, p). Example 2. The Hamiltonian of a certain system is H(q,p) = 12p2 Show that if the following transformation is made q = a p=m where /(/?) is an arbitrary function of (3 and if we define Y =[/(/*)# then a and (3 satisfy the following equations . 3Y da (96) (97) (98) (99) (100) (101) Show that the Langrange bracket [q, p] for the transformation is hence the transformation is not in general canonical, even though the resulting equations are in the canonical form. Show that the result above is not inconsistent with Theorem 2. Solution. Using Hamilton's equations of motion we obtain q=P (103) p = 0 (104) It follows that *-$f* + £>-'-A/i) <l05> But |I=0 (107) oa %-M) (108) Combining Eqs. (105)-(108) we obtain Eqs. (100) and (101). The Lagrange bracket for the transformation is i*']-!?$-$H (,09)
Problems 785 Substituting Eqs. (97) and (98) in Eq. (109) we obtain The quantity df/d/3 is not in general equal to unity as it would have to be if the transformation were canonical. Furthermore, even though there exists a function y(a, /?) such that the equations of motion are given by Eqs. (100) and (101), it does not follow that [q, p] is equal to a constant. This does not violate Theorem 2 because Theorem 2 assumes that for every Hamiltonian H(q,p,t) there exists a function y(a, /3,t) such that the equations of motion are given by Eqs. (100) and (101), not simply that for some Hamiltonian this is true. PROBLEMS 1. The Hamiltonian function for a simple harmonic oscillator is H(q, p) = (p2 /2m) + (mu>2q2/2). Show that there exists a time independent canonical transformation to a new set of variables Q = Q(q, p), P = (q, p) for which H{ £>, P) = (P2 + o>2Q2)/2. 2. The Hamiltonian function for a certain system is H(q, p,t) = q+ tep. A canonical transformation is made to a new set of variables Q = q + tep and P = p. What is the Hamiltonian function for the variables Q and PI 3. The Hamiltonian of a certain system is given by H = pxp2sm(et 4- co}q} — co2q2) where c, «,, and co2 are constants. We introduce a new set of variables defined by the canonical transformation What is the Hamiltonian function for the new variables? 4. The Hamiltonian of a certain system is given by H = co2p(q + t)2, where co is a constant. Determine q as a function of the time: a. By solving Hamilton's canonical equations for the variables q, p. b. By making use of the canonical transformation Q = q + t, P = p. 5. Use the canonical transformation Q\=P2\ Q2 = p22 + q2 P^-^+t P2=p2+t to determine the equation of motion of a system whose Hamiltonian is // = /?2+/?|+ q2. Compare this solution with the solution obtained by using Hamilton's equations for the variables qvq2, pv and/?2.
Hamilton's Canonical Equations of Motion Use the transformation P=^tan-im e=?2+4 2 \ ciq ) ~ co2 to solve the problem of the simple harmonic oscillator whose Hamiltonian is H = \(p2 + coV). The Hamiltonian function for a simple harmonic oscillator is H(q, p) = (p2/2m) + (mu2q2/2). Use the canonical transformation p = C(2P)l/2cosQ q= C~\2P)l/2smQ with an appropriate choice for C to simplify the Hamiltonian function. Solve for the motion in terms of the variables Q and P, and convert your results to the orginal variables q and p. Show that the transformation Qx = q2 + p\ Q2 is a canonical transformation. Apply this transformation to solve the equations of motion for the system whose Hamiltonian is H = \{p\ + p\ + q\ + q\) and compare this solution with the solution of the equations of motion in terms of the original variables. The Hamiltonian for a certain system having two degrees of freedom is given by H = p\ + p\ + ±(?, - q2f+ !(?, + q2f Determine the Hamiltonian function for the variables g,, Q2,P\> and ^2> which are defined by the following canonical transformation: ?2 = "( Gi)I/2cosP, + (2Q2)]/2cosP2 / O \1/2 ^-(fiO^sin^ + ^J sini>2 I O \1/2 ft--(ei)1/2sinP, + [-^J sini>2 Solve for gp <22>^i> anc* ^2 as functions of the time, and use these results to obtain q\,q2, P\, and/?2 as functions of the time. —-it)
SECTION 3 Hamilton-Jacobi Theory
77 Generating Functions INTRODUCTION The proper choice of variables can be an immense aid in solving a particular dynamical problem. Up till now we have developed no systematic way of making the proper choice. In Chapters 77 and 78 we develop some powerful techniques to aid us in doing this. GENERATING FUNCTIONS The definition of a canonical transformation not only enables us to test whether a transformation is canonical but also allows us to construct transformations that are canonical. This fact is brought out in the following theorem. Theorem 1. Suppose we are given an arbitrary function F(q, Q, t) that satisfies the condition 32F(q,Q,/) *0 (1) H 3¾ where the notation | | represents the determinant of the matrix [ ]. If we define _dF(q,Q,t) H = ^(4Q.O (2) dF(q,Q,t) p< = - \Qi =^/(q>Q.O (3) then we can solve Eqs. (2) and (3) for Q and P as functions of the variables % p, and t, and the resulting equations & = fi(q>p>0 ^/-^,(4p,o (4) 789
790 Generating Functions define a canonical transformation between the set of variables q, p, and the set of variables Q, P. The function F(q, Q, t) is called a generating function for the transformation. proof: Using Eqs. (2) and (3), the differential of the function F(q, Q, t) for arbitrary fixed time t = t0 can be written dF(q, Q, t0) = 2, ^ dq,+ 2j ^ dQt (5) According to the results of Appendix 18 we can solve the set of Eqs. (2) for Q as a function of q, p, and /, if and only if 3/>,(g,Q,Q dQj 32^(g,Q,Q 3¾½ ^o (6) Equation (6) is true by hypothesis. Therefore we can solve the set of Eqs. (2) to obtain fi = ft(q,P>0 Substituting Eq. (7) in Eq. (3) we obtain p,--^[q.Q(q. p.0>']-^(4*0 Substituting Eqs. (7) and (8) in Eq. (5) we obtain dF{q,p,*0) = 2 Pidq- 2^,(q,P, h)dQt(q,p,/0) (7) (8) (9) It follows immediately from the definition of a canonical transformation that the transformation defined by Eqs. (7) and (8) is a canonical transformation. Given a canonical transformation, can we find a generating function F(q, Q, t) for the transformation? The answer to this question is contained in the following theorem. Theorem 2. If we are given a canonical transformation Q = Q(q,p,/) P = P(q,p,/) and if 9fi(q,P>0 *Pj ^0 (10) (») where |9Q//9/?7-| is the determinant of the matrix [dQt/dpj], then there exists a generating function F(q, Q, t\ and the generating function can be determined
The Transformation of the Hamiltonian 791 within an arbitrary additive function of the time by integrating the equation dF+^Pi(q,Q,t0)dq- ^PfaQ^dQ, (12) along any path between some fixed point q0, Q0, t0 and the point q, Q, t0 and then setting t0 = t. proof: If the transformation (10) is a canonical transformation there exists an F(q, p, t) such that dF{% p, t0) = 2 Pid(l~ 2^/(4 P> 'o)dQi{% P> t0) (13) If condition (11) is true, we can invert the function Q(q, p, t) to obtain p(q,Q,0- Thus Eq. (13) can be rewritten dF(q,Q,t0) = ^Pi(q,QJo)dq- 2^/(4Q>'o)<% (14) i i From Eq. (14) it follows that ^(¾ Q>0 p af(g>Q>0 .... * ^~ Pi=~^Q- (15) From Eq. (15) it follows that F(q, Q,/) is a generating function for the transformation. Finally if we integrate Eq. (14) from the point q0, Q0, t0 to the point q, Q, tQ we obtain *■(*<*.'o)- ^(qo.QoM = fq'p,f0 f2M4Q.<o)^"SPfoQ'WQi] ^qo.po^oL i i 1 (16) If we replace t0 by t we have an expression for F(q, Q, t) in which all quantities are known except F(q0, Q0, /)• Since an additive function of the time does not affect the transformation a particular generating function generates, we can if we wish simply drop this term. THE TRANSFORMATION OF THE HAMILTONIAN If we use a function F(q, Q,/) to generate a canonical transformation, the Hamiltonian function for the new set of variables can be found as follows. Theorem 3. Consider a system whose dynamical state is defined by a set of canonical variables q, p and whose behavior under the action of a given force is defined by the Hamiltonian function H(q, p, t). If we make a transformation to a new set of canonical variables Q(q, p, t\ P(q, p, t) generated by the generating function F(q, Q, 0, the Hamiltonian K for the new set of variables
792 Generating Functions is given by K= H + dF(q,Q,t) dt (17) proof: If we compare Eq. (9) in the proof of Theorem 1 with Eq. (4) in Chapter 76 we see that the quantity F in the present theorem is the same as the F in Theorem 1 in Chapter 76. The proof of Eq. (17) then follows immediately from Theorem 1 Chapter 76. Suppose that instead of finding the Hamiltonian associated with a new set of canonical variables we wanted to find the canonical variables that would cause the Hamiltonian to assume a particular form. This problem is considered in Theorem 4. Theorem 4. Consider a system whose dynamical state is defined by the variables q, p and whose behavior under the action of a given force is defined by the Hamiltonian function H(q, p, t). Let AT(Q, P, t) be a known function of the variables Q,P, and also the time t. Then any function F(q, Q, t) that satisfies the partial differential equation 3F(q,Q,/) K\ V' 3Q = H 3F(q,Q,/) q>—^—->' and also satisfies the condition 32F(q,Q,/) 3ft 3¾ 3q ^0 dt (18) (19) will be the generating function for a canonical transformation from the variables q, p to the variables Q, P, and the Hamiltonian function in the Q, P representation corresponding to the Hamiltonian function H(q, p, t) will be the function AT(Q, P, t) proof: A function F(<\, Q, t) satisfying Eq. (19) will generate a canonical transformation defined by the equations 3F(q,Q,t) Pi P,= - H 9^(q,Q,Q 3 a (20) (21) If 7/(q,p,0 is the Hamiltonian function corresponding to the variables q,p and K'(Q, P, t) the Hamiltonian function corresponding to the variables Q, P> then K' = H + 3^(¾ Q,Q 3/ (22)
Generalized Generating Functions 793 If we use Eqs. (20) and (21) to express all the terms in Eq. (22) as functions of „ Q, and t, we obtain K' dF(q,Q,t) = H 8F(q,Q,/) 4 3^ >' 3q + —^r- (23) 3; But if F(q, Q, t) satisfies Eq. (18) then K'= K which completes the proof of the theorem. (24) Theorem 4 is an extremely powerful theorem. Its use will become evident in the following chapter. GENERALIZED GENERATING FUNCTIONS The results of the preceding section are generalized in the following theorems. The proofs are left to the reader. Theorem la. Let*,, y. = qi9 pi or pi9 - qt and Xi9 Yt = ft, Pi9 or Pi9 - Qt. Suppose we are given an arbitrary function F(x, X, t) that satisfies the condition 32F(x,X,/) dx.^Xj ^0 If we define yt = Y= - 3F(x,X,f) dxi 3F(x,X,/) 3X (25) (26) (27) then we can solve Eqs. (26) and (27) for Q and P as functions of the variables q, p, and t, and the resulting equations & = e,(q,p,0 i>, = i>,(q,p,0 (28) define a canonical transformation between the set of variables q, p, and the set of variables Q, P. The function F(\, X, t) is called a generating function for the transformation. Theorem 2a. If we are given a canonical transformation Q = Q(q>p,0 P = P(q,p,/) and if the determinant 3*,-(x,y,0 ty *0 (29) (30)
194 Generating Functions where xi9 y{ = qi9 pj or pi9 - qt and Xi9 Yt = Qi9 Pt or Pi9 - Q{, then there exists a generating function F(x, X, t) and the generating function can be determined within an arbitrary additive function of the time by integrating the equation ^=2 yfa x 'o) dx~ 2 Yfa x ^) dxi (31) along any path between some fixed point x0,X0,/0 and the point x,X,/0 and then setting tQ = t. Theorem 3a. Consider a system whose dynamical state is defined by a set of canonical variables q, p and whose behavior under the action of a given force is defined by the Hamiltonian function H(q, p, t). If we make a transformation to a new set of canonical variables <2(q,p, t)9P(q9p9t) generated by the generating function F(x9X9t)9 then the Hamiltonian K for the new set of variables is given by dF(x,\t) K=H+ 3, (32) Theorem 4a. Consider a system whose dynamical state is defined by the variables x,y where xi9 y( = qi9 p{ or pi9 — qi and whose behavior under the action of a given force is defined by the Hamiltonian function H(x, y, t). Let K(X,Y9t) be a known function of the variables X,Y and the time t where Xi9 Yi = Q., P. or Pi9 - Qt. Then any function F(x, X, t) that satisfies the partial differential equation dF(x,X,t) K X 3X J = H 3F(x,X,f) x» —. ^ and also satisfies the condition 32F(x,X,/) dX.dX; 3x 7^0 + 3/ (33) (34) will be the generating function for a canonical transformation from the variables x, y to the variables X, Y and the Hamiltonian function in the X, Y representation corresponding to the Hamiltonian function H(x, y, t) will be the function K(X,Y,t). THE JACOBIAN FOR A CANONICAL TRANSFORMATION In Theorem 2 in Chapter 74 we proved that the Jacobian associated with a canonical transformation was ± 1. We now show that it is + 1. Theorem 5. If the transformation Q = Q(q,PO P = P(q,p,0 (35)
Problems 795 is a canonical transformation, the Jacobian of the transformation is + 1; that is, J + i (36) proof: If the transformation is canonical, a generating function F(x,X, t) exists for at least one choice of the set of variables x,X. If F(x,X, t) is a generating function for the transformation, then 3^(*,X,Q , „ 3F(x,X,Q *~—55— and Y' txT (37) Making use of these equations and the properties of Jacobians, we obtain '(£) .(-i/ 3^,(x,X) dXj 1 3^(x.X) | dXJ ■-( iv' ( V d2F(x,X) ~ dxjdXi 1 a2F(x,x) 1 dXjdx, =(-l)2/ 32F(x,X) 32F(x,X) 3^.3*, = +1 (38) PROBLEMS 1. Determine the transformation generated by the generating function F = Ql/2q\ — § [ Q2 - ?2]3/2 ~~ (Q\ + Qi)1 anc* use tne transformation to determine the motion of the system whose Hamiltonian is H = p2 + p\ + q2. 2. Determine the transformation generated by the generating function F(q, Q,t) = (o/2)[q(Q2- q2)]/2+ Q2sm-\q/Q)]-u2Q2t/2 and use the transformation to obtain the solution to the equation of motion of a simple harmonic oscillator for which the Hamiltonian is H(q, p) = 3. Determine the transformations generated by the following generating functions: a. F(q,Q,0 = 2/ftft-
796 Generating Functions b. F(p,P,t) = ^iPiP, c. F(p,Q,t) = -(tanp){eQ - \). 4. Find the generating function that generates the identity transformation. 5. Find the generating function F(q, Q) that generates the transformation p = -K(2Q)x/2cos P, q= K-\2Q)l/2sin P. 6. Determine the generating function for an extended point transformation. 7. The Hamiltonian function for a certain system is H(q, p,t)= q + tep. A canonical transformation is made to a new set of variables Q = q + tep and P = p. a. Find two generating functions for the transformation. b. What is the Hamiltonian function for the variables Q and PI 8. Find a canonical transformation such that the Hamiltonian of a freely falling body in one degree of freedom becomes K(Q,P) = P. Solve the problem in terms of Q and P and transform back to obtain the usual solution.
78 The Hamilton-Jacobi Equations INTRODUCTION In the preceding chapter we showed that it is possible in principle to generate a canonical transformation that transforms the Hamiltonian into any desired form. This possibility provides us with a completely new approach to solving problems in dynamics. Instead of trying to solve a complicated set of equations of motion, we try to find the generating function that transforms the Hamiltonian into a form that leads to equations of motion that are easy to solve. THE TIME DEPENDENT HAMILTON-JACOBI EQUATION Given a Hamiltonian function H(q,p,t) we can, in principle, generate a transformation to a new set of variables Q,P for which the Hamiltonian vanishes. This fact is exploited in the following theorem. Theorem 1. Consider a system of/degrees of freedom whose dynamical state is defined by the set of variables q, p and whose behavior under the action of a given force is defined by the Hamiltonian function H(q, p, /)• If we set up the partial differential equation H\ 35(4 0 3q 3SYq,0 (1) and if we can find a solution to this equation of the form S=S(q,a,t) (2) where a = a,, a2, . . ., a.f is a set of constants, and if the solution satisfies the condition 32S(q, «M) 3¾ daj 7^0 (3) 797
798 The Hamilton-Jacobi Equations then q(/) can be obtained from the equations 9S(q,a,t) 3 a, = A (4) where )3 = /?,, /?2, • • • > fy is a set of constants. The set of equations (4) provides us with/ algebraic equations in the/ unknowns q],q2, • • • , qj. The values of the constants a and )3 are determined by the boundary conditions. Given q(t) we can find /?(/), if desired, from the equations *S(q,a,t) Pi=-H~ (5) The partial differential equation (1) is called the time dependent Hamilton- Jacobi equation, and the function S(q, a, t) is called Hamilton's principal function. proof: According to Theorem 4 in Chapter 77, any function F(q, Q,t) that satisfies the partial differential equation H 3F(q,Q,f) q>—^r—-,t 3q , aF(q,Q,0_0 3? and also satisfies the condition a2^(q,Q,0 ^o (6) (7) will be the generating function for a canonical transformation to a set of variables for which the Hamiltonian is AT(Q, P, t) = 0. The function F(q,Q,0-[S(q,a,/)]a_Q=S(q,Q,/) (8) is such a function. Therefore, S(q, Q, i) is the generating function for a canonical transformation leading to a new set of variables Q, P for which the Hamiltonian K is identically zero. The transformation equations associated with the generating function S(q, Q, i) are defined by the equations 3S(q,Q,0 Pi = P> = as(q,Q,f) 3g~ and since AT(Q, P, ?) = 0, the equations of motion are 3A"(Q,P,/) e,= /»,- 3P,. 8A-(Q,P,/) 3¾ = 0 = 0 (9) (10) (») (12)
The Time Independent Hamilton-Jacobi Equation 799 If we integrate Eqs. (11) and (12) we obtain ft = «, (13) Pi=-Pi (14) where a; and /3i are constants. The sign in Eq. (14) has been chosen negative to simplify later results. If we substitute Eqs. (13) and (14) in Eq. (10) we obtain ■A-- 3S(q,Q,Q 9S(q,a,t) JQ=« da- (15) Equation (15) is identical with Eq. (4). If we substitute Eq. (13) in Eq. (9) we obtain Eq. (5). This completes the proof. THE TIME INDEPENDENT HAMILTON-JACOBI EQUATION If the Hamiltonian function is not a function of the time, we can partially solve the time dependent Hamilton-Jacobi equation. This result is contained in the following theorem. Theorem 2. Consider a system of/degrees of freedom whose dynamical state is defined by the set of variables q, p and whose behavior under the action of a given force is defined by the time independent Hamiltonian function H(q, p). If we set up the partial differential equation dW(q) H\ 3q = E (16) where E is a constant whose value for a particular set of boundary conditions is equal to the value of the constant of the motion #(q,p) for the given boundary conditions, and if we can find a solution to this equation of the form W= W(q,a) (17) where a = al,a2, . . . , ot.f is a set of constants that explicitly or implicitly includes the constant E, that is, E = E(a\ and if the solution satisfies the condition d2W(q,a) then the equations of motion are given by 3S(q,«,0 9 a,- 7^0 = A where S(q,a,t)= W(q,a)-E(a)t (18) (19) (20)
800 The Hamilton-Jacobi Equations and j8 = /?,, /?2, . . . , j8y *s a set °f constants. The set of equations (19) provides us with / algebraic equations in the/unknowns qx,q2, • • • , <7/. The values of the constants a and )3 are determined by the boundary conditions. The partial differential equation (16) is called the time independent Hamilton- Jacobi equation, and the function W(q,a) is called Hamilton's characteristic function. proof: If the Hamiltonian function #(q,p) does not contain the time, the time dependent Hamilton-Jacobi equation becomes 3S(q,f) H q> 9q 3SYq,0 If we assume a solution of the form S(q,t)=W(q) + 4,(t) then Eq. (21) becomes dW(q) H\ * 3q (21) (22) (23) Since the first term is a function of q only, and the second a function of t only, and since Eq. (23) must be true for arbitrary values of q and t, it follows that dW(q) H 3q = E = -E (24) (25) where E is a constant. The solution to Eq. (25) is <j>(t)= -Et + constant (26) It follows that if we can find a solution W(q, a) to Eq. (24) and if the solution satisfies the condition \d2W(q,a)/dql?9a.| ^0 then the function S(q,a,t) = W(q,a) — E(a)t will satisfy the time dependent Hamilton-Jacobi equation and will also satisfy the condition |325(q, a,/)/3^3^1 =^0. It follows that the equations of motion are given by Eq. (19). This completes the proof of the theorem. SEPARATION OF VARIABLES It is frequently possible to solve the time independent Hamilton-Jacobi equation by assuming a solution of the form ^(q)-2 **;■(*) (27)
Generalization 801 If this is possible we will find that the time independent Hamilton-Jacobi equation can be replaced by / ordinary differential equations of the form dWt{q) = 0 (28) We can then solve for dWi{q^/dqi and integrate to obtain W^q^a). Having found the functions Wt we can proceed immediately to the solution of the original problem. In cases such as this the entire problem is reduced to the determination of a set of integrals. Such a situation is presented in Example 2. GENERALIZATION The Hamilton-Jacobi equation can be generalized. The result is stated without proof in the following theorem. Theorem la. Consider a system of / degrees of freedom whose dynamical state is defined by the set of variables x, y where xi9 yt = qt, pt or pi9 — qt, and whose behavior under the action of a given force is defined by the Hamilto- nian function H(x, y, i). If we set up the partial differential equation H dS(x,t) x,— ,t 3x dS(x,t) (29) and if we can find a solution to this equation of the form S= S(x,a,t) (30) where a = al,a2, . • • , ay is a set of constants, and if the solution satisfies the condition d2S(x,a,t) dX;dat j ^0 then the motion of the system can be obtained from the equations dS(x,a,t) dXj dS(x,a,t) = yi = A where 0=0,,02 3 a,- f$j is a set of constants. (31) (32) (33) The preceding theorem can sometimes be used to simplify a Hamilton- Jacobi problem. Suppose, for example, we are interested in the motion of a system whose Hamiltonian is H = p + sin p + q. If we let x, y = q, /?, we are faced with solving the partial differential equation dS/dq + sin(3S/dq) +
802 The Hamilton-Jacobi Equations q + 35/3/ = 0. If however we let jc, y = /?, — q we have the simpler partial differential equation/? + sin/? — dS/dp + dS/dt = 0 to solve. The same types of simplification that are achieved by use of the preceding theorem could also be achieved by noting that the transformation q'i = <li * ^ k = /?, i = k *-» <+" <34) is a canonical transformation. Using such transformations, we can make changes of the type qk,pk^pk, — qk in a given Hamiltonian until we have reduced it to the form that leads to the simplest Hamilton-Jacobi equation. EXAMPLES Example 1. Solve the one dimensional simple harmonic oscillator problem by Hamilton-Jacobi techniques. Solution. By proper choice of canonical variables the Hamiltonian for a one dimensional simple harmonic oscillator can be written in the form H-±(p2 + aY) (35) The time independent Hamilton-Jacobi equation for the Hamiltonian above is ,2 il(f)+"¥ Solving for dW/dq we obtain = E (36) dW _ „ v _ ..2„2^/2_ ,./„2 _ „2^/2 where a2 = 2E/u2. Integrating we obtain ^ = (2^-^2)17 2=«(«W),/2 (37) IF-f qi^-q^^+ahin-^^ + C (38) The solution above contains two constants a and C. We are looking for a solution containing one arbitrary constant that satisfies the condition (18). We therefore set C = 0. We thus have for Hamilton's characteristic function ^,«)=f[^-^),/2 + a2sin-«(i)] (39) and for Hamilton's principal function S(q,a,t)= %\q(a2 - q2)U2 + ahm-^l)] - *&-t (40)
Examples 803 The equation of motion is thus — , M) = oasm-l( 2-) - u2at = B (41) da \ a J v 7 Solving for q we obtain q = a sinl cot H J = a sin(co/ + y) (42) where y = /3/ua. Example 2. Determine the motion of a particle that is moving under the action of a force that is derivable from a potential of the form V(r\ where r is the distance of the particle from some fixed point. Solution. Choosing spherical coordinates r, 0, and <j>, the Hamiltonian function is H=pL + JL+ P} +V (43) 2m 2mr2 2mr2sin29 ' The time independent Hamilton-Jacobi equation is l (dw\2 l (Bw\2. l (w\2+ V_E (44, 2m { dr ) + 2mr2 V ™ ) 2mr2sm29 I 3* / ^ ' We now assume H^= Wx{r) + H^2(<9) H- W3(4>) (45) Substituting Eq. (45) in Eq. (44) we obtain 2m{ dr ) + 2mr2 I dB J 2mr2sm20 \ d<j> ) { ' Multiplying Eq. (46) by 2mr2sin20 we obtain dW\ \2 ■ 2n( dW2 \2 I dW3 rsin The third term on the right is a function of <J> only, and the remaining terms are functions of r and 6. This can be true for arbitrary r, 0, and <j> only if dW3 \2 , - > = a2 (48) r2sin20 I -^-1 J + sin20 I -^- I + 2mr2sin29 (V- E)= -a2 (49)
804 The Hamilton-Jacobi Equations where a3 is a constant. Dividing Eq. (49) by sin20 and rearranging the terms we obtain ot dWx \* , ( dW2 \2 a2 ,>(_2)+2mr;(F_£) + (_i)+_,„ (50) The first two terms are functions of r only and the last two terms are functions of 9 only. This can be true for arbitrary r and 0 only if / dW2 \2 of 2 M+^=a2 (5i) [-^) +2mr\V-E)=-a22 (52) where a2 is a constant. If we solve Eqs. (48), (51), and (52), noting that the constants of integration can be ignored, we obtain W3 = arf (53) W2 = sin2<J> [/2 w, =/ 2m(E- V)- 1/2 dr Hamilton's principal function is thus given by S = =/ of 2m(E- V)--j r dr + /« sin20 1/2 (54) (55) dO+a3<$>-Et (56) The equations of motion can now be obtained using the results of Theorem 1. Thus dS(r, 0,<j>, E, a2,a3) dE = mj 2m(E- V)- a} -1/2 dS(r,0,<f>,E,a2,a3) da-> C ^1 J r2 2m(E- V)- dr-t = ^ (57) -1/2 dr 'H^'^9 -1/2 d9 = p2 1/2 (58) ZS(r,0,*E,a2,a3) = _ ^ ,^ X" „ (59) °<*3 J sin20 \ sin20 J Equation (57) can be solved to obtain r(t), Eq. (58) can be solved to obtain r(9\ and Eq. (59) can be solved to obtain <j>(6). Combining these results we can determine whatever we want to know about the motion.
Problems 805 Example 3. The Hamiltonian of a certain system of one degree of freedom is H = p2-q (60) Solve for the motion using the generalized Hamilton-Jacobi equation, letting Solution. The generalized Hamilton-Jacobi equation is /+M + f = o (61) Note that it is — q and not + q that is replaced by dS/dp in obtaining the equation above. Solving for S we obtain p3 S= - -^- +ap- at (62) where a is a constant. The motion of the system is obtained from the equations dS(p, a, t) ^ } - (63) (64) (65) (66) (67) (68) PROBLEMS 1- The Hamiltonian function for a certain system is given by H(p, q) = p + aq2. Solve for the motion of the system using the Hamilton-Jacobi theory. 2. A projectile of mass m is fired with an initial speed v at an angle <J> with the horizontal. Use the Hamilton-Jacobi theory to determine the equation of the path of the projectile. 3. A particle of mass m is constrained to remain on a smooth moving wire. The equation of the wire is y = x + \ at2. At time / = 0 the particle is at rest at the point x = y = 0. Determine the path of the particle using Hamilton-Jacobi theory. which gives us from which we obtain -^F~ = -q dS(p,a,t) U P p2-a = q P~t = P p=t + p q = (t + j3)2-a
806 The Hamilton-Jacobi Equations 4. Use the Hamilton-Jacobi equation to obtain an equation for the orbit of a particle moving in a two dimensional potential U = k/r, describing the motion in terms of the coordinates u = r + x and v = r — x. 5. A particle of mass m moves in the x-y plane in a field created by force centers at (1,0) and (- 1,0). The potential energy is V'= Axrxx + A2r2~\ where rx and r2 are the distances from the respective force centers to the particle and Ax and A2 are constants. Use the Hamilton-Jacobi theory to obtain an equation for the orbit in terms of elliptic coordinates qx and q defined by x = cosh^ cos^2 and j = sinh^, sinq2. 6. Use the Hamilton-Jacobi equation to show that the orbit of a particle whose Hamiltonian is H = \{p] + p\){q\ + q\)~x + (?? + tf!)"1 is a conic in the qx-q2 plane. 7. A particle of mass m is moving in a force field whose potential is V=(A/r)— Bz, which is a combination of a uniform field in the z direction and an inverse square central force field with the center of force at the origin. Obtain Hamilton's principal function for parabolic coordinates £, tj, and <j> where £, tj, and § are defined in terms of cylindrical coordinates p, <j>, and z by the equations £ = r + z, -q = r — z, and § = <f>, where r1 = p2 + z2. 8. A particle of mass m moves in a potential that in spherical coordinates is given by V = f(r) + [g(0)/r2], where/(r) is some function of r, and g(9) is some function of 9. Obtain a general solution for the orbit. 9. A particle of mass m is moving in a potential that in cylindrical coordinates p, <f>, and z can be written in the form F = [f(r + z) + g(r - z)]/r, where r2 = p2 + z2 and f(r + z) and g(r — z) are arbitrary functions of r + z and r — z, respectively. Obtain Hamilton's principal function for the coordinates £, 77, and <J> where £ and ?] are parabolic coordinates defined by the equations £ = r + z and 77 = r — z or equivalent^ by the equations z = \{£ — 77) and p = (£t])1/2. 10. A particle of mass m is moving in a potential of the form V = [o2/(rxr2)] {f[(r2 + rx)/2o] + g[(r2 - rx)/2o]}, where rx and r2 are the respective distances of the particle from a fixed point Px and a fixed point P2, 0 is a constant, and /(x) and g(x) are arbitrary functions. Obtain Hamilton's principal function using the elliptic coordinates £, 77, and <f>, which are defined in terms of cylindrical coordinates p, <f>, and z by the equations p = a[(£2 - 1)(1 - T72)]1/2, z = a£r7, and <J> = ¢. Choose the value of 0 and the axes in such a way that Px and P2 lie on the z axis at the points z = —o and z = + a, respectively.
79 Action and Angle Variables INTRODUCTION The Hamilton-Jacobi approach provides us with a method for transforming from one set of canonical variables q, p to a second set of canonical variables Q, P, each of which is a constant of the motion. In this chapter we consider a method, valid for certain types of motion, of transforming from one set of canonical variables q, p to a second set Q, P, not all of which are constants of the motion but do provide a great deal of insight into and information about certain aspects of the motion. SEPARABLE SYSTEMS We restrict ourselves to systems in which the Hamiltonian function is not an explicit function of the time, that is, H=H(q,P) (1) and for which it is possible to find a solution to the time independent Hamilton-Jacobi equation of the form ^(q,«) = 2^-(<7/>«) (2) i We refer to such systems as separable systems. CYCLIC SYSTEMS Consider a system whose dynamical state is characterized by the set of generalized coordinates q = quq2, • • • , <7/ and the set of generalized momenta Ps/?i,p2, • • • ,/y- As the system moves, it traces out an orbit in the q,p sPace, the space whose coordinates are the variables q\,q2> - - • , q* Pi, 307
808 Action and Angle Variables Fig. 1 p2, . . . ,/y. It also traces out an orbit in each of the subspaces qi9 pr The orbit in the qi9 pi space is the projection of the orbit in the q,p space onto the qrp. plane, and can be represented by an equation of the form/?,- = #(¾) or by the pair of equationspt = pi(t),qi = qt(t). If for each value of /, the equation of the orbit pi= Pi(qj) in the qrpt plane is a closed curve, either because the qi oscillates between two definite limits qi = a and qi = b, as shown in Fig. 1, or because the qt moves from qi = a to qi = b and then because of the nature of the <7rspace starts over again at qi = a, as shown in Fig. 2, then we refer to the system as a cyclic system. note 1: The term "cyclic" has been introduced to simplify the statement of results in the following sections. The term is a little misleading because even though the system may go through a cycle in each of the qh pi subspaces, it does not necessarily follow that the system will ever return to its original state in the q, p space. note 2: If a cyclic system has only one degree of freedom, the time required for the system to execute a cycle in the q-p plane will be a constant; hence the motion in the q-p plane will be periodic in time. If a cyclic system has more than one degree of freedom, then in general the time required to execute a particular cycle in one of the qh pi spaces will not be a constant but will depend on the motion of the other coordinates; hence the motion in the given qi9 p{ space will not be periodic in time. This fact should be carefully Fig. 2
Action and Angle Variables 809 noted because many textbooks leave the reader with the erroneous impression that each of the projected motions is periodic in time. ACTION AND ANGLE VARIABLES Consider a separable cyclic system of / degrees of freedom, whose dynamical state is characterized by the canonical variables q,p. Let H(q,p) be the Hamiltonian function of the system, and let w(q.«)=2^(*.«) (3) i where a = al9a2, . . . , «y are constants, be a complete solution to the time independent Hamilton-Jacobi equation Let J = Ji, J2, ■ ■ • > J/ be the set of constants defined by the equations J>W-$ 3^ ^ (5) where the integral is over one complete cycle for the variable q,. If we use the function W(q,3)=W[q,a(J)} = 2^/(¾. J) (6) as a generating function for a canonical transformation from the variables q, p to a new set of coordinates w, and momenta J; that is, we define the variables w and J by the transformation equations a^(q,J) *w,{g„ J) *—*;, H~ ^ -, ^ (8) then the new coordinates wl9w2, . . . , ny are called angle variables, and the new momenta Jl9J2, • • • ,J/ are called action variables. From Eq. (7) we obtain ay,[g,,j(a)] 3^(^«) ^fl)" H = -^~ (9) Substituting Eq. (9) in Eq. (5) we obtain Ji(a)=(j)pl(qi,a)dql (10)
810 Action and Angle Variables The equation pi = p^q^a) is the equation of the projection of the orbit P = P(<D onto tne subspace #,¾. The integral on the right hand side of Eq (10) is thus the area enclosed within the orbit, or under the orbit, as illustrated by the shaded regions in Figs. 1 and 2. Thus the function Jt{a) can be geometrically interpreted as the area swept out in the pi9qt subspace during one complete cycle in this subspace. This area depends on the constants a or equivalently on the initial conditions and may in general assume any value Historically the first attempt to pass from classical mechanics to quantum mechanics consisted in assuming that the value of Jt was not completely arbitrary but had to be a multiple of h/2ir, where h is Planck's constant. THE MOTION OF A SYSTEM IN TERMS OF ACTION AND ANGLE VARIABLES Theorem 1. a. Consider a cyclic separable system of/degrees of freedom whose dynamical state is defined by the canonical variables q,p = q\,q2, • • • , %P\, p2, . . . ,Pf and whose dynamical behavior is governed by the Hamiltonian function H(q, p). If we transform to action and angle variables J, w, then the Hamiltonian H is a function of J only, that is, H = H(J) (11) and the motion of the system is given by Ji = 7i (12) w,. = v+ *,. (13) where the y, and ^ are constants that are determined by the initial conditions and the vt are constants, called the frequencies of the system, that are defined as follows: 3//(J) (14) If we consider a motion in which all the coordinates qj are held constant except the coordinate qi9 then in the passage of qt through a complete cycle the w. will all return to their original values except for the variable wi9 which changes by one unit. Conversely if all the Wj are held constant except wi9 which changes by one unit, the coordinates qt will all return to their initial values. It follows from the above that any function F of the set of variables q has the following properties: (1) F is a multiply periodic function (see Appendix 31) of the set ot angle variables (h>,,h>2, . . . , ny) with the fundamental period system (1,0,0, . . . , 0), (0,1,0, . . . , 0), (0,0,1, . . . , 0), . . . , (0,0,0, . . . , D-
The Motion of a System in Terms of Action and Angle Variables 811 (2) F can be expanded in a Fourier series of the form F= 2 2 • • • 2 {4,,¾ • ■nM^(n1w1 + n2w2 + • • • + nfwf)] "I "2 »f + ^n,n2...«/cos[2ff(«,w1 + n2w2 + • • • + nfwf)]} "I "2 + Dnini...ncos[2ir (»,*, + n2v2+ ■ ■ ■ + nfvf)t]} (15) where the A% B's, C's, and Z)'s are constants. (3) If F when expressed in terms of the action and angle variables is, in addition to being dependent on the action variables, an explicit function of some subset wl9w2, . . . ,wr of \he angle variables, where r < /, then F will be a periodic function of the time if and only if the reciprocals of the frequencies vu v2, . . . , vr have a common multiple; that is, if there exists a set of integers kuk2, . . . , kr such that /Ci Ky (16) proof: The transformation to action and angle variables is a time independent transformation; hence the new Hamiltonian is equal to the old Hamiltonian. The Hamiltonian function in terms of the action and angle variables is thus given by #(w,J) = tf[q(w,J),p(w,J)] 3W(q,J) = H = H q(w,J), q(w,J), But from Eq. (4) H dW(q,a) 3q 3q dW[n,a(J)] 3q = £(«) Combining Eqs. (17) and (18) we obtain H(v,J)-E[a(J)]-E(J) (17) (18) (19)
812 Action and Angle Variables Hamilton's canonical equations of motion for the action and angle variables are thus J,--^-0 dH(J) W: = u Solving Eq. (20) we obtain Ji = Y, Substituting Eq. (22) in Eq. (21) we obtain where Solving Eq. (23) we obtain v, = dH(J) J = Y w, = vtt + <.,. (20) (21) (22) (23) (24) (25) This completes the proof of part a. The relationship between the angle variables w and the coordinates q is given, from the definition of angle variables, by the set of equations *W(q,J) w. = J 3./,- (26) The Ji are constant during the motion of the system; hence the change in the variable w is given by 3W(q,J) dWj = 2 t d1k 3/.- dqk 3 ^ 8^(q>J) If we now hold all the qk fixed except qt, then dqk dW: = 1 W, dq, (27) (28) Integrating over one complete cycle of the qi9 while holding all the other q^
The Motion of a System in Terms of Action and Angle Variables 813 fixed, we thus obtain r ZW(q,J) (29) Thus if all the qk are held fixed except qi9 which goes through one cycle, hw. = 1, and Aw- = 0 fory ^ i. The converse is also true; that is, if all the wk are held fixed except wi9 which is changed by one unit, all the qt will return to their original values. To prove this we note that for W(q,J) to be an acceptable generating function it must satisfy the condition 32*F(q,J) dqJJj Substituting Eq. (26) in Eq. (30) we obtain dw,(q,J) H ^0 ^0 (30) (31) If Eq. (31) is true, then according to the results of Appendix 18, there is a unique value of q for each value of w. Thus if a given value of q generates a set of values of w, each of these set of values of w must correspond to this one value of q. Since w = (wl9w29 . . . , wi9 . . . , ny) and w = (wl9w2, . . . , w,- + 1, . . . , Wj) correspond to the same value of q, it follows that each of these values of w corresponds to the same q. This completes the proof of part b of the theorem. The proofs of parts cl and c2 follow from part b and the definitions of a multiply periodic function and a Fourier series and will be left to the reader. To prove part c3 we note that F= F(q) = F[q(w)] = F(wl9w2, ...,wr) = F(vxt + <t>l9v2t + <j>29 . . . , vrt + <t>r) = F{t) (32) The function F(wl9w29 . . . , wr) is a periodic function of the set of variables (^l'H^ • • • , wr) with the fundamental period system (1,0, . . . , 0),(0, \> • • • , 0), . . . , (0,0, . . . , 1). Hence F(t) will be a periodic function of the time t if and only if there exists an increment of time At for which each of the terms p.t + ^. changes by an integer. For this to be true we must have vx At = kx v2At = k2 vrAt = K (33)
814 Action and Angle Variables where kl9k2, . . . , kr are integers. These conditions are equivalent to the conditions K i Ky Kr V„ (34) This completes the proof of the theorem. EXAMPLES Example 1. The Hamiltonian of a simple harmonic oscillator is given by t + hi1 2m 2 Show that the frequency v is equal to (k/m)l/2/27r H^P) = in+\ (35) proof: Since H is a constant of the motion, the orbit in q, p space is given by where E is a constant. This is the equation of an ellipse with semiaxes (2m£)'/2 and (2E/k)l/2. The area enclosed by the ellipse is equal to the value of the action variable J. Hence J = H2mE)^(f)l/2=2,(f)l/2E (37) It follows that H(J) = E=K ' } J (38) and the frequency is dH(J) (k/m)x/1 v = dJ (39) Example 2. Use action and angle variables to analyze the motion of a particle that is moving in a plane under the action of a central force for which the potential is V(r)=-k_A (40) where r is the distance from the force center, k and ft are positive constants, and ft<^kr over the range of values of r spanned by the particle in its motion.
Examples 815 Solution. The Hamiltonian function in terms of the polar coordinates r and <}> is "(^/^♦)-s| + A-7-4 (41) The time independent Hamilton-Jacobi equation is 2^1 9r / 2mr2\ H) r r2~ ( } The system is separable. Separating variables and solving we obtain W= Wr{r9E9h)+W^9E9h) (43) where Wr(r,E,h) = (2m)"2 f(*+ * + 4 " A )^ (44) J \ r r 2mr J W<t>(<t>,E9h)=jhd<j>=h<j> (45) and is and h are constants. The orbits in the r-pr and the <$>-p^ planes can be determined from the equations dWr(r9E9h) .ni k B h2 \1/2 P* H h (47) If, as we shall assume, mk2 Amp - 2hA <E<0 (48) then the orbit pr = pr(r) is a closed orbit and is symmetric about the r axis; and the value of r ranges between the values rx and r2, where rx and r2 are the roots of the equation £ + - + 4 " -^r = 0 (49) r r1 2mr2 K } The orbit p^($) is also closed, with <f> moving from 0 to 2tt, and then back again to zero. The system is thus cyclic. Since the system is a cyclic separable system, we can define a set of action and angle variables. The action variables
816 Action and Angle Variables are given by •w*)-0 -V-2 Jr U2 \'/2 /2 L / 2m\,/2 = -2^-2,^+(-^) 7rA: T^{E,h)=j) d4> = <£hd<j>= f "hd$ = 2tt7i Solving for £ and h as functions of Jr and 7^ we obtain 2mw2k2 E= - [jr + (j2-^2m^/2] *-i Since H(Jr,J^) = E we have 2mm k ihi 7r + (7,2-87r2^)1/2] The frequencies of the system are given by p, = u [7r + (72-8^V)1/2] i/2i3 2m2wA: dH(J„JJ 7, '* 3^ /r2_o„2„.oNi/2"'' '♦ {Jl -$v2m/3y (50) (51) (52) (53) (54) (55) (56)
Examples From the definition of the angle variables dW(r,t,Jr,Jj where and wr = dJr _ dWr(r,J„JJ dW^>Jr,J^) dJr + dJr _ dWr{r,E,h) dE(J„Jj dWr(r,E,h) M dE djr + dh djr dw^,E,h) dE(jr,jj dw^,E,h) dh + dE djr + dh djr dWr(r,E,h) = -4^-^ + 0 + 0 + 0 = vrF{r) (57) dW(r,<t>,Jr,JJ w* 37;— _ 3 Wr(r,J„JJ dWJ^,Jr,J^) M* + 3^ _ dWr(r,E,h) dE(J„Jj dWr(r,E,h) M dE dJ^ + dh dJ^ dW„{<$>,E,h) dE{Jr,J^) dW^,E,h) u + dE dJ^ + dh dJj, = ^F(r)-G(r)+± (59) where F(r) is given by Eq. (58), and Equation (57) can be solved in principle to obtain r = r(wr)
818 Action and Angle Variables The quantity r is a periodic function of wr with a period one. Since wr changes by one unit in a time rr = \/vn it follows that r is a periodic function of the time with a period \/vr. If we substitute Eq. (61) in Eq. (59) and solve for <h we obtain 4> = «j>(wr,Hg (62) The angle <j> is a doubly periodic function of (w,.,^) with the fundamental period system (1,0), (0,1). Since wr changes by one unit in a time rr = \/v and h^ changes by one unit in a time r^ = 1/j^, the angle <j> will not be a periodic function of the time unless vr and v^ are commensurate, that is, unless there exists a pair of integers kr and k^ such that kr/vr = k^/v^. Once we know r(wr) and ¢(^,.,^) we can find r(t) and ¢(/) by using the equations wr = vrt + \r (63) w* = V + A* (64) If /? = 0 the problem reduces to the motion of a particle in an inverse square central force field. For this case (-2mE)3/2 v'=v*=p=^^r (65) and r(t) and <j>(t) are both periodic functions of the time with the common period I/p. The orbit in space is in this case a closed orbit, which can be shown to be an ellipse. If /3 is small but nonzero, the orbit, though not closed, is nearly an ellipse, and the pericentron (the point of closest approach to the force center) precesses slowly around the force center. The rate of precession can be determined, as we shall show, without solving explicitly for r(t) and <p(t). Combining Eqs. (59) and (64) we obtain <j> = Imv^t - 2mv+F{r) + 2irG(r) + 2ttX^ (66) Since r(t) is a periodic function of the time with the period rr = \/vr, the change in <j> during a time rr will be A*-2^Tr = 2*(^) (67) Note that the last three terms in Eq. (66) return to their original values at the end of a time rr, hence do not contribute to A<J>. If v^ = vr then A<J> = 2tt and there is no advance of the pericentron. If v^ ¥" vr then A<J> ^ 2m and the pericentron advances by an amount 27r(v^/vr) — 2m. It follows that the rate of advance of the perihelion is given by 2ir(vJvr) - 2tt q. R K^r) = M^ _ ^ (68)
Examples 819 jf R is small then to a first approximation 7* -^1 + ^¾ (69) '* (^-8»V^),/2 Jl The constant 7^ can be approximated by its value when j3 = 0. From Eq. (51) and the results of Chapter 23 we have J* " (2™) « (70) where a and Z> are the lengths of the semimajor and semiminor axes, respectively, of the ellipse. Combining Eqs. (68)-(70) we obtain R^—^r (71) kb2 v }
J
SECTION poisson Formulation of the Equations of Motion
80 poisson Brackets INTRODUCTION A Poisson bracket is a very useful analytical device for studying the behavior of a dynamical system. In this chapter we define "Poisson bracket" and consider some of its properties. Chapter 81 presents some applications of Poisson brackets to the study of dynamical systems. POISSON BRACKETS If u and v are any two quantities that depend on the dynamical state of a system and possibly the time, the Poisson bracket of u and v with respect to a set of canonical variables q, p is defined as («.«0=2 du(q,p,t) 3o(q,p,/) 9«(q,p,f) 3o(q,p,/) (1) 3* dPi tyi H In terms of the condensed notation introduced in Chapter 75 8«(r,/) 80(r,Q („,o)-^-^--^- (2) Poisson brackets have the following properties (where w, v, and w are arbitrary Unctions of q, p, and t; a is an arbitrary constant, and r is any one of the variables q., Pi9 or/): Property 1. (u,v)= ~(v,u) (3) Property 2. (u,u) = Q (4) 823
824 Poisson Brackets Property 3. Property 4. Property 5. Property 6. (w, v + w) = (w, v) + (w, w) ^ (w, VW) = V(U, W) + (w, V)W /£\ a(u, v) = (au, v) = (w, av) n\ d(u>v) /3n \./ 3o\ ^- = (^'U) + (W'97) (8) Property 7. Jacobi's identity (w, (o, w)) + (o, (w, w)) + (w, (w, o)) = 0 (9) The proof of the properties above can be accomplished by using the definition of a Poisson bracket to express each of the terms in these identities in terms of partial derivatives of w, o, and w, and simply noting the truth of the resultant equations. We demonstrate this procedure by proving Jacobi's identity. To simplify the task we use the condensed notation introduced in Chapter 75 and write the partial derivatives 3/3r, and 92/9r/ dr. as simply 9,- and 9^. We note first from the definition of a Poisson bracket that (K,t0 = fty3/w3/t> (10) and from Property 6 that 3,(ii,o) = (3,ii,o) + (ii,3,o) (11) It follows that (ii, (o, w)) = ny 3,11 S/o, w) = fty 3,w[(3,.0, w) + (o, djw)] = fty 3,w[ fiw 3^o 3/W + fe 3*o d/jw] = fyVki 3/w 3*yt> 37w + %f**/ 3/" 9^ 3/yw (12) The indices in Eq. (12) are all to be summed over, hence can be replaced by other indices without altering the meaning of the expressions. If we replace i by k,j by /, k byy, and I by / in the first term on the right hand side of Eq. (12) and make use of the fact that ^ = - ^, and 3iy = 3,., we obtain %f**/ 3/« d/cjV 3/W = tiklnj. dku 3y/o 3,w = - ^, 3*w 3/y o3,w 03) Substituting Eq. (13) in Eq. (12) we obtain (m,(o,w)) = tojUkii-dkUdfjvdtW + 3,tt3*o3/yw) (14)
The Relationship between Lagrange and Poisson Brackets 825 From symmetry (v, (w,u)) = iiijiikl(- dkvd0.w d,u + dtv dkw dljU) (15) (w, (w, v)) = fi^/(- 3^ w 3/yM 3.1? + 3.w 3* w 3/jf.t?) (16) If we add Eqs. (14)-(16), the right hand side of the resultant equation is zero, and we obtain Jacobi's identity. THE RELATIONSHIP BETWEEN LAGRANGE AND POISSON BRACKETS The Poisson bracket is closely related to the Lagrange bracket, introduced earlier. The relationship is given in the following theorem. Theorem 1. If we use the symbol ri to represent a member of the set of variables q, p and let [r"o] =? 1 "a^ "a^"" "a^ "aT" j (17) be the Lagrange bracket of the variables ri and r^ for the transformation Q = Q(q,p,0 P = P(q,p,0 (18) and if we let _,/ 3r, 3r. 3r 3r, \ be the Poisson bracket of the variables ri and r- with respect to the variables Q, P, then 2 [ri,rj](ri9rk) = 8Jk (20) proof: Using the condensed notation introduced in Chapter 75 we have drk dR„ k m cn 8¾ 3*„ ^^dRj, drj = 8 9rt dRm mPdRp dr, drk dRm -X /on jR-^r- 8"J (21) This is what we wished to prove.
826 Poisson Brackets POISSON BRACKETS AND CANONICAL TRANSFORMATIONS The Lagrange brackets can be used to test whether a transformation is canonical. The Poisson brackets can be used in a similar fashion. Theorem 2. The transformation q = q(Q,P,/) p = p(Q,P,/) (22) is a canonical transformation if and only if the Poisson brackets (qi9q.)9 (qi9 pj), and (/?,, pj) with respect to the variables Q and P satisfy the following conditions (<7/> qj) = 0 (% Pj) = «<,- (Pi> Pj) = 0 (23) note: In terms of the condensed notation introduced in Chapter 75, the conditions above are equivalent to the condition (^0) = % (24) proof: From Theorem 1 [Wfa.rJ-Sjt (25) If the transformation is canonical then [nsA-Hj (26) Substituting Eq. (26) in Eq. (25) we obtain ^j(n^k) = ^jk (27) Multiplying by ^ and simplifying we obtain (>7>>*) = f% (28) Conversely if Eq. (28) is true, Eq. (26) will also be true. Thus the relation (28) is a necessary and sufficient condition for the transformation to be canonical. CANONICAL INVARIANCE OF POISSON BRACKETS The notation (w, v) for the Poisson bracket of the quantities u and v does not indicate the set of canonical variables q, p with respect to which the Poisson bracket is determined. In this section we prove that the quantity (u,v) is invariant under a canonical transformation, hence does not depend on our choice of canonical variables. Theorem 3. If the transformation Q = Q(q,P>0 P = P(q,p,0 (29)
Canonical Invariance of Poisson Brackets 827 js a canonical transformation, the Poisson bracket of two quantities u and v with respect to the set of variables q, p is equal to the Poisson bracket of u and v with respect to the set of variables Q, P, that is, V ( 0^- — — -^ — ^ = V ( ^u du _ dv du \ ^/y. f { dq, dp, dqi 3/,,. j f \ 3ft dP, dQi 8/>, J ^ proof: In terms of the condensed notation introduced in Chapter 75, we wish to prove du_ dv_ = du dv (3U ^dRi dRj ^drt 3/) ^1) To show this we note du__dv_ = du drk dv dri ^ dR; dRj ^ drk dRt dn dRj JM5^|i«8£ (32) \^dR{ dRj)drk dr, ^ Since the transformation is canonical drk drt Substituting Eq. (33) in Eq. (32) we obtain Eq. (31).
81 The Poisson Formulation of the Equations of Motion In this chapter we consider the use of Poisson brackets in the analysis of the motion of a dynamical system. THE POISSON FORMULATION OF THE EQUATIONS OF MOTION Theorem 1. Consider a system whose dynamical state is defined by the canonical variables q, p and whose dynamical behavior is defined by the Hamiltonian function H(q, p, i). Let F be an arbitrary quantity that depends on the dynamical state of the system. The time rate of change of F is given by F-(F,lf) + —^ (1) where (F, H) is the Poisson bracket of F with H. proof: The time rate of change of F(q, p, t) is given by *-2Si*+2|£A*£ <2> But from Hamilton's equations qt = 3///9^ and/?, = -dH/dq^ hence -?(if-ff)+f=^)+f « which completes the proof. Theorem 2 (The Poisson Formulation of the Equations of Motion). Consider a system whose dynamical state is defined by the canonical variables q,p and whose dynamical behavior is defined by the Hamiltonian function H(q, p, 0- 828
Constants of the Motion in the Poisson Formulation 829 The motion of the system is governed by the equations 4,-(%H) (4) Pi = {P»H) (5) proof: The proof of this theorem follows immediately from Theorem 1. CONSTANTS OF THE MOTION IN THE POISSON FORMULATION Each new formulation of the laws of dynamics tends to provide a different perspective that can be useful in finding constants of the motion. The Poisson formulation is no exception. Theorem 3. If a given dynamical quantity F is not an explicit function of the time, and if the Poisson bracket of F with H vanishes, that is, if (F,H) = 0, then F is a constant of the motion. proof: The proof of this theorem is a direct consequence of Theorem 1. Corollary 3a. If the Hamiltonian is not an explicit function of the time, it is a constant of the motion. Theorem 4 (Poisson's Theorem). If F and G are two constants of the motion, then (F, G), the Poisson bracket of F with G, is a constant of the motion. proof: If F and G are constants of the motion then F={F,H) + ^=0 (6) G-(G,H)+*g-0 (7) and thus 17=-(^) (8) ff=-(G.tf) (9) Making use of the properties of Poisson brackets, Theorem 1, and Eqs. (8) and (9), we obtain ft(F,G) = ((F,G)H) + -t(F,G) = ((F,G),H) - ((F,H),G) - (F,(G,H)) = ((F,G),H) + ((H,F),G) + ((G,H),F) = 0 (10) Hence (F, G) is a constant of the motion.
830 The Poisson Formulation of the Equations of Motion note: Poisson's theorem provides us with a way of generating constants of the motion by taking Poisson brackets of pairs of known constants of the motion. There is, however, no quarantee that a constant so generated will be independent of the original two constants of the motion. To determine, in a particular case, whether the generated constant of the motion is independent of the constants of the motion from which it was generated, we can make use of the results of Appendix 14. EXAMPLES Example 1. Let hl9 h2, and h3 be the Cartesian components of the angular momentum of a particle. a. Determine the Poisson brackets (hi9h) and (hi9h). b. Show that it is impossible for two of the components h{ simultaneously to be canonical momenta. c. Show that if two of the components are constants of the motion, the third must also be a constant of the motion. Solution. a. From the definition of the angular momentum of a particle we have h\ = x2p2- x3p2 (11a) h2= x3px - xxp3 (lib) h3 = xxp2- x2px (lie) It follows that (hx,h2) = h3 (12a) (h3,hx) = h2 (12b) (h2,h3) = hx (12c) and (hl9h) = (h2,h) = (h39h) = 0 (13) b. The Poisson bracket of any pair of canonical momenta for a system must be zero, that is, (pi9 pj) = 0. But (hi9 hj) ¥> 0 for i =£y; hence two of the components of the angular momentum cannot simultaneously be canonical momenta. Since (hi9 h) = 0, it is possible for the magnitude of the angular momentum to be a canonical momentum at the same time as one of the components of the angular momentum. c. From the results of part a and Poisson's theorem, it follows that if two of the components of the angular momentum are constants of the motion, the remaining component must also be a constant of the motion.
Problems 831 PROBLEMS 1. Show that if the Hamiltonian H and a quantity F are constants of the motion, dF/dt must also be a constant of the motion. As an illustration of this result, consider the one dimensional motion of a free particle. The Hamiltonian is a constant of the motion, and the quantity F = x — (pt/m) is also a constant of the motion. Show by direct computation that the constant of the motion dF/dt is the same as the constant of the motion obtained by taking the Poisson bracket of F and H. 2. Set up the problem of the spherical pendulum in the Hamiltonian formulation, using the spherical coordinates 6 and <|> as generalized coordinates. Show directly in terms of these variables that (hx,hy) = hz, (hz,hx) = h , and (h ,hz) = hx, where hx, hy, and hz are the components of the angular momentum h = r X my. What is wrong with the following argument: pe and p^ are perpendicular components of the angular momentum h, but perpendicular components of the angular momentum cannot be used simultaneously as canonical momenta (see Example 1), and therefore pe and Pq cannot be used simultaneously as canonical momenta? 3. Show by direct evaluation that (hi9 h2) = 0, where hi is the i component of the angular momentum h = r x my and h1 is the square of the magnitude of the angular momentum. 4. Let <|>(r, p) be any function of the coordinates and momenta of a particle, that is spherically symmetric about the origin, and h = r x my be the angular momentum of the particle. Show that the Poisson bracket (<J>, h;) is zero.
PART VARIATIONAL PRINCIPLES IN CLASSICAL MECHANICS
82 D'Alembert's Principle INTRODUCTION D'Alembert's principle provides a useful starting point both for deriving Lagrange's equations of motion and also, as Chapter 83 reveals, for expressing the basic principles of dynamics in a variational form. D'ALEMBERT'S PRINCIPLE Theorem 1. The Newtonian equations of motion for a system of N particles acted on by a given force f, that is, the set of equations /■ = M/ /=1,2, ...,37V (1) is equivalent to the single equation 2(//-M/)**/-0 (2) / = 1 where Sx = &cl5&c2, .., Sx3N is an arbitrary infinitesimal virtual displacement of the system. We refer to Eq. (2) as the first form of d'Alembert's principle. proof: Let us first assume that Eq. (1) is true. If we multiply Eq. (1) by 8x{ and sum over i we obtain Eq. (2). Therefore if Eq. (1) is true, Eq. (2) is true. Conversely, Eq. (2) can be true for arbitrary values of 8xt only if the coefficient of each 8xt vanishes. Thus if Eq. (2) is true, Eq. (1) is true. This completes the proof of the theorem. Theorem 2. Consider a dynamical system consisting of AT particles acted on by a given force G. Let g = gu g2, . . . , g3N be any set of generalized coordinates that define the configuration of the system, 8g = 8gu8g2, 835
836 D'Alembert's Principle ..., 8g3N an arbitrary infinitesimal virtual displacement of the system, Q = Gl5 G2, . . . , G3/v the generalized components of the given force defined by the condition that 2/^7¾ is the virtual work done by the force G in the displacement Sg, and T the kinetic energy of the system. In terms of these quantities the first form of d'Alembert's principle becomes 37-(8.8.0 3/V 2 /=1 Gi~jt Hi + ^-1^=0 (3) proof: Let the transformation between the Cartesian coordinates x and the generalized coordinates g be given by x = x(g,/) It then follows that ^3x,.(g,Q (4) (5) Substituting Eq. (5) in Eq. (2) we obtain 22 < J S a*,(g,/) .. 3x,.(g,o Hj Hj sgj=o But from Eq. (9) in Chapter 47 Vf3^(g10 _ and from the proof of Theorem 1 in Chapter 48 (6) (?) v .. 3*,(g,Q v J 3T(x) 3x, 3*,(g.Q 9§ = 2 dt ■IJ rf/ ar(x) 3x,.(g,Q 3-*/ 3§. ar(x) 3x;(g,g,Q 3x 3r(*) j 3x- A 3*,(S.Q 8S ar(i) a*,(g,g,0 ^. a 3r(8.8.0 94/ 94/ 37-(8.8.0 3¾ dX; 3§ (8) Substituting Eqs. (7) and (8) in Eq. (6) we obtain Eq. (3). Theorem 3. Consider a dynamical system consisting of N particles that is acted on by a given force G and a set of 3 N — / holonomic constraint forces. Let q = <7i,<72> • • • > ?/ ^e anv set °f generalized components that together with the constraint conditions define the configuration of the system. Also let 8q = 8q],8q2, . . . , 8qf be an arbitrary infinitesimal virtual displacement that
D'Alembert's Principle 837 is consistent with the constraint conditions, Q = Ql9 Q2, . . . , Qj the generalized components of the given force defined by the condition that 2/S/&7/ *s the virtual work done by the given force G in the displacement Sq, and T the kinetic energy of the constrained system. It follows that the dynamical behavior of the system is completely determined by the equation / /=1 2 a- dt dT(q,q,t) (9) together with the constraint equations. We refer to Eq. (9) as the second form of d'Alembert's principle. proof: Let the transformation between the Cartesian coordinates x and the coordinates q be given by x = x(q,/) (10) It then follows, for virtual displacements consistent with the constraints, that Substituting Eq. (11) in Eq. (2) we obtain 22 3x,(q,0 .. 3x/(q,f) /■—^ m,x, H H 8qr0 (11) (12) The quantity 2/2///(^/(¾ 0/3¾)¾ ls tne virtual work done by the force f in a virtual displacement that is consistent with the constraint conditions. Since the constraint forces are holonomic they do no work in such a virtual displacement. It follows that only the nonconstraint forces must be considered. And thus 22/ a ' H=2 Q/H' < j "j j (13) The remainder of the proof is formally similar to the proof of Theorem 2. Thus we can show 2w/*, d*/(q>0 A. dt dT(q,q,t) (14) Then substituting Eqs. (13) and (14) in Eq. (12) we obtain Eq. (9).
83 Hamilton's Principle INTRODUCTION In this chapter we express the fundamental postulate of classical dynamics in the form of a variational principle. HAMILTON'S PRINCIPLE Although Theorem 1, below, is sometimes referred to as Hamilton's principle, we reserve this name for Theorem 2, which though somewhat less general, expresses the principle in a more useful form. Theorem 1. Consider a dynamical system having / degrees of freedom. If the system is in a configuration 1 at time tx and in configuration 2 at time t2, the path q(/) that the system follows between configuration 1 and configuration 2 will be the one that satisfies the condition (t2(8W+ST)dt=0 (1) where 8W is the work done by the forces acting on the system in an infinitesimal virtual displacement at time t, T is the kinetic energy, and the varied paths are restricted to those that begin in configuration 1 at time tx and end in configuration 2 at time t2. This condition is necessarily satisfied and is also sufficient to determine the actual motion of the system. proof: From d'Alembert's principle »-i(!f)+!fh=° <2) This equation is true at every instant. For every time t we are free to choose our 8qf however we wish. Let us choose 8qt{t) such that 8q({tx) = 8qi(t2) ~ ®m 838
Hamilton's Principle 839 Expressing the dependence of 8qt on time in Eq. (2) we obtain y' dt\*qt) H «*(') = o We now integrate Eq. (3) between tx and t2 and we obtain sM*-m,) dT H 8q,(t)dt=0 (?) (4) Integrating the second term by parts we obtain Substituting Eq. (5) in Eq. (4) we obtain H8q> But and ^Q^q^SW ?(g^+|f «*)"«• (5) (6) (?) (8) Substituting Eqs. (7) and (8) in Eq. (6) we obtain Eq. (1). Conversely Eq. (1) is equivalent to Eq. (4). But Eq. (4) can be true for arbitrary 8qi only if a A dt (S) + |I-o (9) for every value of i. Equations (9) are Lagrange's equations of motion, which are sufficient to determine the motion. This completes the proof of the theorem. Theorem 2 (Hamilton's Principle). Consider a dynamical system having f degrees of freedom for which the generalized components of the force are derivable from a potential U. If the system is in a configuration 1 at time tx and in a configuration 2 at time t2, the path q(/) that the system follows between configuration 1 and configuration 2 will be the one for which the functional S[q(t)]=f'2L[q(t),q(t),t]dt (10)
840 Hamilton's Principle where L= T — U, has a stationary value, or equivalently for which 8ft2L[q(t),q(t),t]dt=0 01) This condition is necessarily satisfied and is also sufficient to. determine the actual motion of the system. proof: If the behavior of the system is defined by a Lagrangian, there exists a U such that d(9U\ dU dt \ dqt ) dq, Sq, i i I And thus = 0 - ('28Udt Jtx And therefore in Theorem 1 C'2(8T+ 8W)dt= (h8(T- U)dt= (h8Ldt= 8 (hLdt Jt) Jt] Jtx Jtx (12) (13) (14) HAMILTON'S PRINCIPLE AND LAGRANGE'S EQUATIONS As shown in Appendix 32, the condition sf'2L(q,q, t)dt=0 is equivalent to the condition JL(*L\- dt \ 3¾ J dL H = 0 (I5) (16) Hence Hamilton's principle is equivalent to Lagrange's equations for a Lagrangian system. HAMILTON'S PRINCIPLE AND HAMILTON'S EQUATIONS OF MOTION Hamilton's equations of motion can be derived from Lagrange's equations of motion for a Lagrangian system. Since Lagrange's equations of motion for a Lagrangian system can be derived from Hamilton's principle, it follows that Hamilton's equations of motion can be derived from Hamilton's principle. In
Hamilton's Principle and Hamilton's Equations of Motion 841 this section we show how Hamilton's equations can be obtained directly from Hamilton's principle. The Lagrangian and Hamiltonian functions are related as follows: £(*q>0 = 2M*4>0ft- H[q,p(q9q,t),t] (17) i It follows that s('2Ldt = (h8Ldt -02/*** + 2**,-2|f«*-2|ftp)* (18) Integrating the first term by parts and noting that 8qf vanishes at tx and t2 we obtain 02/>/^)*=[2rt|2-02 A**)* = -02A*fc)* (19) Substituting Eq. (19) in Eq. (18) and the result in Hamilton's principle we obtain ?J,1-(*+ifh+(*-fK. dt=0 (20) One is tempted at this point to conclude from Eq. (20) that the coefficients of the 8qf and the 8pt must vanish. If this were true, we would immediately obtain Hamilton's equations. However the 8qt and the 8pt are not independent variations; hence this step is not justified. Instead of proceeding in this direction we note that H is the Legendre transformation of L(q,q,/) with respect to the set of variables q, hence from Theorem 1 in Appendix 22, ?/ = IK Substituting Eq. (21) in Eq. (20) we obtain ?£[(*♦£)* dt=0 Since the 8qt are arbitrary, Eq. (22) can be true only if p' H Hamilton's equations consist of Eq. (23) together with Eq. (21). (21) (22) (23)
842 Hamilton's Principle PROBLEMS 1. A particle of mass m is in a uniform gravitational field directed in the negative z direction. The particle starts from rest at the origin at t = 0 and falls to a lower point at t = T. Consider the path z = -(gt2/2) + at(t - T), and show explicitly that Hamilton's principle demands that a = 0 thus demonstrating the correctness of the principle in this simple case. 2. Determine which of the following trajectories for a falling particle best approximates the actual motion: a. z = At b. z = Bt3 3. Consider a system for which the Lagrangian is given by L = \x2 - ij2. a. Show directly that Sjl/SL dt has a value zero for the path x = A sin t. b. Show, for the path x = A (sin t + c sin 8/), which has the same value at t = 0 and t = 7r/8 as the path x = A sin/, that jl/sLdt has its least value when c = 0. 4. A uniform flexible string is stretched between two rigid supports A and B a distance / apart. The tension in the string is t and the mass per unit length is p. Small transverse vibrations are set up in the string. If y(x,t) is the transverse displacement at time t of the element of the string that is located a distance x from the support A, show that the kinetic and potential energies of the string are given by T = (p/2)JQ(dy/dt)2dx and V— (r/2)jl0(dy/dx)2dx, respectively. Use Hamilton's principle to derive the equation of motion of the string, namely 3^/3/2 = c232>?/3x2, where c2 = r/p.
84 The Modified Hamilton's Principle INTRODUCTION Hamilton's principle assumes a knowledge of the Lagrangian function L(q,q, t) and considers varied paths in the configuration space q. In this chapter we show that it is possible to set up an equivalent variational principle that assumes a knowledge of the Hamiltonian function #(q,p, 0 anc^ considers varied paths in the space q, p. THE MODIFIED HAMILTON'S PRINCIPLE Theorem 1. The functional %0.p(0] =^(2/-/(04(0 - %).pCV])* (1) where the functions q(0»p(0 are restricted to those having given values at both tx and t2, has a stationary value, or equivalently Sj^SAOftC) " H[q(t),p(t),t]}dt=0 (2) where the variations are limited to those that vanish at tx and t2, for a particular set of functions q(0>P(0> if and only if these functions satisfy the equations dp, a 3ft A--I? (4) 843
844 The Modified Hamilton's Principle proof 1: From Theorem 3 in Appendix 32, Eq. (2) is true if and only if q(/) and p(/) satisfy the following equations: d { 3 dt { 3¾ d [ 3 dt \ dpf J 3* )"¥ = o (5) = 0 (6) Carrying out the partial differentiations in Eqs. (5) and (6) we obtain !<»>+f=° o -*+f-° (») or equivalently ><--f m * = dH ' 3/,,. (10) proof 2: Carrying out the variation in Eq. (2) we obtain «{|Sm-%p.O dt -2 jfl*«*+**-ff«*-^fW dt=0 But jt2pi8qidt= -jpiSqtdt Substituting Eq. (12) in Eq. (11) and rearranging terms we obtain ?i:1 [-(*+f)] 8^,+ "• 3#" 8p,\dt=0 Since the &j>, and Spt are arbitrary, Eq. (13) will be true if and only if • dH _ ft" = M 3/>, (11) (12) (13) (14) (I5) Corollary (The Modified Hamilton's Principle). Consider a system of / degrees of freedom whose dynamical behavior is defined by a Hamiltonian #(p>q>0- If tne system is in the dynamical state 1 at time tx and in the
The Modified Hamilton's Principle 845 dynamical state 2 at time t2, the path q(0»p(0 that the system follows in q,p space between the states 1 and 2 will be the one for which the functional 4q(0.P(0] -J'2{ 2/»/(04(0 - H[q(t),p(t),t]}dt (16) has a stationary value, or equivalently for which «J'2{SM04(0 - H[q(/),p(0.']}*-0 (17) proof: From Theorem 1, Eq. (17) is equivalent to Hamilton's equations of motion. It follows that the particle will follow a path satisfying Hamilton's equations of motion if and only if it also satisfies the modified Hamilton's principle. note: The modified Hamilton's principle */(?M -^)^=0 (18) and Hamilton's principle 8fLdt=0 (19) are formally distinct principles. In Hamilton's principle only the function q(/) is varied. In the modified Hamilton's principle the functions q(/) and p(/) are varied. To demonstrate the caution that is necessary in distinguishing between these two principles, we use each of them to determine the motion of a free particle in one dimension, and then show some erroneous results that can be obtained by misuse of these equations. The Lagrangian function of a free particle in one dimension is and the Hamiltonian function is H(9>P>t) = ^ (21) Applying Hamilton's principle we obtain 8 fhL dt =8Jh^dt = fhmq 8qdt=- fhmq 8qdt=0 (22) from which it follows that <7 = 0 (23) which is the correct equation of motion. Applying the modified Hamilton's
846 The Modified Hamilton's Principle principle we obtain = ^((-/^ + (^-^)^=° (24) from which it follows that P = 0 (25) * = i (26) which again lead to the correct equation of motion. Now consider the following erroneous argument. From Hamilton's principle %2 „ „2 hence 8(hLdt=s('^dt=sfh^dt= ("P-Spdt-0 (27) p = 0 (28) This conclusion is not true and results from a misuse of Hamilton's principle. THE MODIFIED HAMILTON'S PRINCIPLE AND CANONICAL TRANSFORMATIONS The modified Hamilton's principle can be used to derive many of the results of canonical transformation theory. Theorem 2 (see Chapter 76). Consider a system that is acted on by a given force. Let the dynamical state of the system be defined by the set of variables q,p = qx, . . . ,qf,px, . . . ,Pf, and let us assume that there exists a function H(q, p, t) such that the behavior of the variables q, p under the action of the given force is determined by the equations of motion qi—W~ {) * H~ (30) If we make a transformation to a new set of variables Q = Q(q,p,0 p = P(q,p,o (31)
The Modified Hamilton's Principle and Canonical Transformations 847 and if the transformation is a canonical transformation, that is, if there exists a function F(q, p, t) such that where „-Vp, ^&(q,P>0 , 9F(q9f9t) * = Z Pfa P> 0 gj + 37 (33) then the equations of motion in terms of the variables Q, P can be written dK(Q9P9t) Qi = \Pi (34) 3AT(Q,P,/) p = ——- r35^ where K = H+R (36) proof: If Eqs. (29) and (30) are true then from the modified Hamilton's principle J^SM -#)*«<> (37) Now suppose that we make a canonical transformation to the variables Q, P. Since the transformation is canonical there exists an F(q, p, t) such that ^«2-M-2^a+* (38) i /' where i? is given by Eq. (33). It follows that F+^PiQ-H-R « 2M- # (39) i i and therefore «jr'2(^+s^a-^-*)*-*jr'2(SM-^)-o (40) But 6 f'2F<ft= 6 f2^= «[F(/2) - F(*,)] (41) And since the variations vanish at the endpoints 8[F(t2)-F(tl)]=0 (42) Combining Eqs. (40)-(42) we obtain Sf'2[ 2 PiQ>-{H +R )"U-0 (43)
848 The Modified Hamilton's Principle From Theorem 1 8JJT(Q,P,0 a = —jp—- (44) 3ff(Q,P,f) where K = H+ R (46) Theorem 3. Consider two sets of variables q, p and a, /3 related by the transformation equations a = a(q,p,t) 0 = 0(q,p,/) (47) The equations dt=0 (48) S£2 ?M"%IK) and "r_ l<fc=0 (49) 8jT'2[SM-7(«,ftO where the variations are restricted to those vanishing at tx and /2> are equivalent for all functions H(q, p, /) if and only if Sft«rr = ^M-^)-f (50) where c is a constant and F is a function of q,p, t. proof: To simplify the proof we use the condensed notation introduced in Chapter 75, including the summation convention, and use p for the set of variables a, )3. In terms of this notation Eqs. (48) and (49) become 8 ft\XiJr/i-H)dt = 0 (51) sf'XXiJpJpi-y)dt = 0 (52) and Eq. (50) becomes Wt-y-cOy/.-V-f (53) If Eq. (53) is true then 80V/. - v)<*- «X''['(V/'" ») ~ f ] <*
The Modified Hamilton's Principle and Canonical Transformations 849 Since the variations are assumed to vanish at the endpoints, 8I,'2f *««[^2)-^,)]=0 (55) Substituting Eq. (55) in Eq. (54) we obtain 8 (\\jPjP, -y)dt-c8 f''(A,//,- - H)dt (56) It follows that if Eq. (51) is true, Eq. (52) is true and conversely if Eq. (52) is true, Eq. (51) is true. Thus if Eq. (53) is true, Eq. (51) is equivalent to Eq. (52). To complete the theorem we need to show that if Eqs. (51) and (52) are equivalent, Eq. (53) is true. From Theorem 1, Eq. (51) is true if and only if ^W'f> (57) "» drj and Eq. (52) is true if and only if dy fty^-ft (58) Equation (53) can be rewritten dF = c\ikrkdri - \mkpkdpm + (y- cH)dt I c\krk - Xmkpk-7^- \dri + [y ~ cH ~ KkPk-jf- Jdt (59) There exists an F satisfying Eq. (59) if and only if and ij|r I c\krk - KkPk -£r ) = jfm I cXjkrk ~ K*Pk -97 I (60) Yt [CX*rk ~ XrnkPk -^ J = Jff (t - CH - XmkPk -^ J (61) Carrying out the differentiations in Eq. (60), rearranging terms, and making use of the fact that \y - Xy/ = ^-, we obtain Proceeding similarly with Eq. (61) we obtain ^"aT"^" a^~c~a^ (63) Thus instead of showing that Eqs. (51) and (52) lead to Eq. (53) we can show that Eqs. (57) and (58) lead to Eqs. (62) and (63). We note first 9y _ dy_ 9ft* « 9y 9p* _ 97¼ .,..
850 The Modified Hamilton's Principle Making use of Eq. (58) in Eq. (64) we obtain h dr, y _ dPk . 9Pfc / 9Pm . , 9Pm \ Substituting Eq. (57) in Eq. (65) we obtain aY _ „ , i. 3# ■j ~rk where 3r "^^ (66) aj = ,kmk~W%F. (67) °jk — fimn ~^~ "g^" /% (68) Taking the derivative of Eq. (66) with respect to rt we obtain 3r,3r, 3r, -f 3r, 3 ay 4^3/-,3 V u'i A: u'/ u'k k n _ daj dbJk dH diH = 37 + 2,-9791- + 2, 2jbJk8imj—^- (69) ari k 0ri 0Kk k rn orm0rk Interchanging the i and they we obtain 3^ =3o+?^Ta^ + ? 5M*" 3¾ (70) Equating Eqs. (69) and (70) we obtain (daj 3a/ \ , V ( UJk ^W+VV^A f, ft \ *H -0 (71) Equation (71) must be true for any H and therefore dr, 8/y ^-^=0 (73) (M*» - bikSJm) + (bJmSik - bim8jk) = 0 (74) Equation (74) was obtained by setting the sum of the coefficients of the mk term and the km term in the last term in Eq. (71) at zero. Substituting Eq. (67)
The Modified Hamilton's Principle and Canonical Transformations 851 in the left hand side of Eq. (72) and noting that ^. = — ^ we obtain da, 3 ^_^i = _i_/ *^L^!l\-JL[ df)"> dpk rt 3/) drt \ ^mk dt 3ry J 3ry [ ^mk dt dr( 92Pm dPk _ 92Pm dPk ^3/-,3/ 3/}. ^3/)3/ 3r, 92Pm 3Pfc 92Pm 9P^ ,-<- ^3^3/ 3ry Mfcm3ry3/ 3/¾ l j Interchanging the indices /c and m in the last term we obtain H _ 3¾ = 4,¼ d2P* ^Pm 3r, 3/) ^3/-,3/ 3ry ^3/)3/ 3/-,. 3 / 9Pm dPk\ n,. -Tty^^Wj) (76) Substituting Eq. (76) in Eq. (72) we obtain 87(^1^-^)-0 (77) Comparing Eqs. (68) and (77) we see that Eq. (77) implies that Let us now consider Eq. (74). If we let m = k = i we obtain (79) (80) (81) Substituting Eq. (81) in Eq. (74) we obtain (bM8jk8im - bAkSp,) + (bjj8Jm8ik - bu8im8jk ) = 0 (82) which can be rewritten (bM ~ *«)(*A + 8ik8jm) = 0 (83) Suppose thaty =£ i and we let k=j and m — i; then (fy-W+0)-0 (84) and thus btt = bjj = b (85) If i ¥=j then and therefore bjAi ~ h8jt + bjfia - bfi + bjt-Q bij = hsij " Jl
Substituting Eq. (86) in Substituting Eq. (86) in Eq. (78) Eq. (73) db 3/-/ h = bSy we obtain we obtain ^-^-o Letting k =j and assuming that i =£j we obtain |*-o dr. 852 The Modified Hamilton's Principle Substituting Eq. (85) in Eq. (81) we obtain (86) (87) (88) (89) From Eqs. (87) and (89) it follows that b - constant (90) and therefore b,j-c8y (91) where c is a constant. Substituting Eq. (91) into Eq. (68) we obtain . 9Pm 9Pm /(r)N cSjk^Hmn-faT -97- fe (92) Multiplying Eq. (92) by /x/Vfc and making use of the fact that [ilkiiik = S/7 we obtain ^// = ^-^--^7 (93) which is equivalent to Eq. (62). Finally if we substitute Eq. (91) in Eq. (66) we obtain ir-aJ + cir (94) drj ^ drj Combining Eqs. (94) and Eq. (67) we obtain Eq. (63). We have thus shown that Eqs. (57) and (58) lead to Eqs. (62) and (63). This completes the proof of the theorem.
PART RELATIVISTIC MECHANICS
1
SECTION Relativistic Kinematics
85 The Basic Postulates of Relativistic Kinematics INTRODUCTION The basic principles of Newtonian kinematics were presented in Part 1, Section 1. In this chapter we consider the modifications necessary when one is dealing with speeds comparable to the speed of light. INERTIAL FRAMES The discussion of space and time, and inertial frames of reference given in Chapters 1 and 2 requires no modification and should be reread at this point. It follows that Postulates I and II given below are the same Postulates I and II that were used in the Newtonian kinematics presented in Volume 1. Postulate I. There exists at least one frame of reference called an inertial frame having the following properties: a. The geometry is Euclidean. b. The clocks can be synchronized. c All positions, directions, and instants of time are equivalent with respect to the formulation of the laws of motion. Postulate II. Frames of reference have the following properties: a. Any frame all of whose points are moving with the same constant velocity with respect to an inertial frame is also an inertial frame. b. All inertial frames are equivalent with respect to the formulation of the laws of motion. The identification of inertial frames can in principle be accomplished by making use of the following theorem, which we have already proved. 857
858 The Basic Postulates of Relativistic Kinematics Theorem 1. If a particle is far removed from the influence of all other particles in the universe, the particle will move with a constant velocity with respect to an inertial frame. THE UPPER LIMIT ON THE SPEED OF A PARTICLE In Newtonian kinematics one assumes there is no upper limit to the speed with which a particle can move. Experiment reveals however that there is an upper limit, and in relativistic kinematics this fact is taken into account. It follows that Postulate III in Newtonian kinematics must be replaced by the following postulate. Postulate III. There is a finite upper limit c to the speed with which a particle can move with respect to an inertial frame. Experiment reveals that the value of c is given by c = 2.997925 X 108 m/s (1) The existence of an upper limit is the key difference between relativistic kinematics and Newtonian kinematics. The existence of this upper limit has as we shall see profound consequences, not only in kinematics but also in dynamics. SUMMARY Postulates I, II, and III stated above are the foundation on which relativistic kinematics is built. In the chapters that follow we consider some of the consequences of these postulates.
86 The Lorentz Transformation THE LORENTZ TRANSFORMATION In Chapter 3 we derived the following theorem relating the readings in one inertial frame to those in another. Theorem 1. Let R be an inertial reference frame, and R' a second inertial frame that is moving with respect to R. Let O be the origin and OX, OY, and OZ the axes of a Cartesian coordinate system S fixed in frame R. Let O' be the origin and 0'X\ 0'Y\ and O'Z' the axes of a similar Cartesian coordinate system S" fixed in frame R'. The coordinates of a point with respect to S are designated x, y, and z, respectively; and the coordinates of a point with respect to S' are designated x', y\ and z', respectively. The time reading on a clock in the R frame is designated by t, and the time reading on a clock in the R' frame is designated /'. If we choose the coordinate systems and the zero setting on the clocks in such a way that at zero time in both frames the origins coincide, the respective axes are collinear, and the relative motion of S' with respect to S is in the positive x direction, the coordinates x, y, z, t in frame R of a particle and the coordinates x',y\ z', t' in frame R' of the same particle are related as follows x' = T(x- Vt) (1) / = y (2) (3) <' = It<-iJ) (4) where V is the speed with which S' is moving with respect to S, c is the upper limit on the speed with which a particle may move with respect to an inertial frame, and -1/2 (5) -(-S) 859
860 The Lorentz Transformation note: The choice of coordinate axes and the setting on the clocks puts no physical restriction on the results. proof: See Chapter 3. In Newtonian kinematics the limit c was taken to be infinite. If c is set equal to infinity, the resulting transformation, the Galilean transformation, is relatively simple. In relativistic kinematics, c is finite and is in fact 2.997925 X 108 m/s. In this case the transformation remains as given and is called the Lorentz transformation. THE LORENTZ TRANSFORMATION IN VECTOR FORM In the preceding section we chose coordinate systems S and S' to make the results come out as simple as possible. This puts no physical restriction on our results. Nevertheless it is sometimes useful to write the transformation in a form that allows a wider choice of coordinate systems S and S'. Corollary. Let R be an inertial frame and R' a second inertial frame. Let V be the velocity in R of R' with respect to R and V the velocity in R' of R with respect to R'. Let r be the position in R of a particle with respect to a point O fixed in R, and r' the position in R' of the same particle with respect to a point O' fixed in R', Let S be any Cartesian coordinate system fixed in R and S' be a coordinate system fixed in R' that is chosen such that the orientation of S' with respect to — V is the same as the orientation of S with respect to V and such that the points O and O' coincide at times t = t' = 0. Then the relationship between the coordinates x, y, z, and t and x\ y\ z', and /' is given by the vector equations (r-l)(r-V)V r' = r + ^ !\ >_ - jyt (6) V 2 ''-i-'-f) (?) where c is the upper limit on the speed with which a particle may move with respect to an inertial frame, and r-(-2p proof: See Chapter 3. note 1: See Note following Corollary 1 in Chapter 3. note 2: Note that the coordinate systems S and S' do not necessarily coincide at t = t' = 0. Consider for example the x axis of coordinate system S,
Transformation of Velocities 861 which is defined by the equations y = z = 0. Setting y = z = 0 in Eqs. (6) and (7) we obtain x' = x + (T-l)^--TVxt (9) /-(T-l)^£-rK,r (10) ^ = (1--1)-^-1^ (11) xW V xV, r = r^-^j (12) Setting /' = 0 in Eqs. (9)-(12) and eliminating t and x we obtain y' = AVyx' (13) z' = AVzx' (14) where {T-\){Vx/V2) + T{Vx/c*) i + (r -1)( v?/v2) + T(v? A2) ^ = ■ T—^ T-V (15) It follows that at /' = 0, the x axis of the coordinate system S does not in general coincide with the x' axis of the coordinate system S'. However, in those cases in which V lies along either the jc, y, or z axes the axes do coincide at t = t' = 0. note 3: It is sometimes convenient to break the vectors in Eq. (6) into components parallel to V and components perpendicular to V. If we do this we obtain 1l = r('l|-V0 (16) *±=*± (17) TRANSFORMATION OF VELOCITIES From Eqs. (1)-(4) it follows that dx' = T(dx-Vdt) (18) dy' = dy (19) dz' = dz (20) dt' = T(dt-^dx\ (21)
862 The Lorentz Transformation Hence dx, T(dx - Vdt) [(dx/dt) - V] dt' T[dt-(V/c2)dx] [l-(V/c2)(dx/dt)] dy' dy (dy/dt) dt' T[dt-(V/c2)dx] T[l-(V/c2)(dx/dt)] dz' dz (dz/dt) dt' T[dt-(V/c2)dx] T[l-(V/c2)(dx/dt)] v =dx =dy _dz V*~dt Vy~H **=-% . -dx' , _ dy' _ dz' V*~dt' V>-dt' V*~dt' (22) (23) (24) (25) we can rewrite Eqs. (22)-(24) vY = \-(vxV/c>) v, v, = T[l-(vxV/c2)] °z T[l-(vxV/c2)] (26) 4=^, /„,^ (27) (28) which is the law for the transformation of velocities. We can obtain the vector form of the transformation from Eqs. (6) and (7), from which it follows that dx, dr + [(T- \)(dr • V)V/ V2] - TVdt di7 Tdt-[T(dr-V)/c2] (dr/dt) + {(T-l)[(dr/dt)-V]V/V2]}-TY = T-{T[(dr/dt)-\/c2]} (29) hence v + [(T - l)(v • V)V/ V2] - rv ~ r[l-(y.V/c2)] V = (30)
Transformation of Accelerations 863 It is sometimes convenient to break the vectors in Eq. (30) into components parallel to V and components perpendicular to V. If we do this we obtain v„-V v;, = —!—— (31) l-^F/c2) v, v, = x r[i-(0|lK/c»)] (32) The law of transformation for the speed can be obtained from Eqs. (26)-(28) or from Eq. (30) and is given by v = r2[(v. v/ v) - v] + v2 - (v • v/ V) r2[i-(v-v/c2)]2 2 , 2 2 ' r (V\\ ~ V) + V V Note that if v = c then t/ = c. (33) TRANSFORMATION OF ACCELERATIONS From Eqs. (26)-(28) and Eq. (4) it follows that [1 - (vxV/c2)]dvx + (vx - V)(V/c2)dvx dv' = ± — J [i-K^A2)] [\-{vxV/c2)]dvy + vy{V/c2)dvx dv'y = T[\-{vxV/c2)^ [l-(vxV/c2)]dv2 + v2(V/c2)dvx dv'. = dt T[\-{vxV/c2)]2 ' = r(dt--^dx\ If we define (34) (35) (36) (37)
864 The Lorentz Transformation then from Eqs. (34)-(37) we obtain aY = T>[l-(vxV/c2)]3 [l-(vxV/c2)]ay + (vyV/c2)ax ay = T2[l-(vxV/c2)]' [l-(vxV/c2)]a2 + (vzV/c2)ax a,= T2[l-(vxV/c2)]3 (39) (40) (41) We can obtain the vector form of the transformation from Eqs. (30) and (7), from which it follows that [1 -(v\/c2)]dv+ {[(1 -T)\/(TV2)} + (v/c2)}(V. df) d\' = (42) r[i-(v.v/c2)]2 d? = T\dt-^-^-} (43) hence [l-(v.V/c2)]a+{[(l-r)V/(rF2)] + (v/c2)}(V.a) a' = (44) r2[l-(v.V/c2)]3 It is sometimes convenient to break the vectors in Eq. (44) into components parallel to V and components perpendicular to V. If we do this we obtain a""r3[i-(^A2)]3 r2[i-(t^A2)]2 T2[i-(vl}v/c2)]2 _ a± + (V/c2) x [(v± x a„) + (v„ x ax)] r2[i -(v,v/c2)]3 THE QUANTITY y The quantity (45) (46a) (46b) occurs very frequently in relativistic mechanics.
The Modified Velocity u 865 To obtain the transformation law for y we note first from Eq. (30) that x2 . . i _/-„2/„2a t (pT_1 /v'.y'\_ 1-(»2A2) c2 \ c2 ) r2[l-(v-V/c2)]2 (48) hence v' = (l-^)rr (49) THE MODIFIED VELOCITY u The mathematics of relativity can frequently be simplified if instead of using v to describe the motion of a particle we use the quantity u = yv (50) which we refer to as the modified velocity. One of the advantages of using u rather than v is the absence of restrictions on the values of u, whereas v is restricted to values for which v is less than or equal to c. The law of transformation for u can be obtained by combining Eqs. (30) and (49), thus: 7 \ c2 )y\ r[l-(v.V/c2)] } ^, (r-i)(Yv-v)v = Yv-YrV+ — (51) If we let U = TV (52) Eq. (51) can be written fT-l)(u«U)U u' = u - yV + !±- — (53) U2 The quantity y, which was defined in the preceding section as ■1/2 I 2 \->/i ..-^ (54, can also be written in terms of u. To obtain y(u) we replace v by u/y in Eq. (54) and solve for y. Doing this we obtain ,2 J/2 (.♦£)"'
866 The Lorentz Transformation PROBLEM 1. At a given instant a particle is moving along the x axis of an inertial fratne R with a speed v. Show that the components of the acceleration of the particle with respect to the frame R, and the components of the acceleration of the particle with respect to an inertial frame R', which is momentarily at rest with respect to the particle, are related as follows: a'x = aj[l-(v2/c2)]3/2, a'y = ay/[\-{v2/c2)], a'z = aj[\ - (v2/c2)].
87 Some Consequences of the Lorentz Transformation INTRODUCTION In this chapter we consider some of the kinematic consequences of the Lorentz transformation. THE LAW OF THE ADDITION OF VELOCITIES If a particle is moving with a velocity v with respect to an inertial frame R, and the inertial frame R is moving with a velocity u with respect to an inertial frame R', the velocity w of the particle with respect to R' can be obtained by setting v = v, V = — u, and v' = w in Eq. (30) in Chapter 86. Doing this we obtain [l-(W2/c2)]1/2(u + v)+{l-[l-(W2/c2)]1/2}[l + (u-v/W2)]u l + (u-v/c2) 0) It follows that the law for the addition of two velocities is much more complicated than in classical mechanics, where w = u + v. If u and v are in the same direction, the law of addition of velocities becomes simply — u + v w = 1 + (uv/c2) (2) TIME DILATION Suppose that a clock that is at rest at a point r in a frame R gives out signals at times tx and t2. The coordinates of the first and second signals are (r, tx) and 867
868 Some Consequences of the Lorentz Transformation (A'>"r. ,.*' ^ (6) (r,t2\ respectively. In a frame R' that is moving with respect to R with a velocity V, the times of the two signals are hence t'i -1\ = r(/2 - tx) (5) or [l-(KVc2)]' Thus the interval between the two signals is longer in frame R' than in frame R. It follows that a clock that is moving with respect to a given observer will run slower than an identical clock that is at rest with respect to the observer. LENGTH CONTRACTION Consider a rod that is at rest in a frame R' with its ends located at the points r', and r'2, respectively. If R' is moving with a velocity V with respect to a frame R, then at a time t in frame R the positions r, and r2 satisfy the equations (r-lXr,.V)V V2 ri=ri + - „i ~rv/ (7) (r-l)(r2-V)V ^ = ^+- ;A,2 ' -IT/ (8) K hence or ^,-,)+(r-1)[Vv]v c) (r-l)[(Ar)-V]V /im (Ar)' = Ar + ^ ^^ *- (10) If we break (Ar) and (Ar)' into components parallel to V and perpendicular to V we obtain (Ar)± = (Ar)'± (H)
Examples 869 It follows that a rod that is moving with respect to a given observer will be shortened along the line of motion, but not perpendicular to it. EXAMPLES Example 1. Twins A and B are initially together in an inertial frame S. Twin B takes a rocket journey. The rocket initially moves away from A with a speed of V for a time r as recorded on the rocket's clock. The rocket then turns around and travels toward A with a speed V for a time r as recorded on the rocket's clock. Show that at the end of the trip, when the twins are reunited, twin A will have aged an amount 2Tr where T = [1 — (F2/c)2]_1/2, whereas twin B will have aged only an amount 2t. Hence twin B will be younger than twin A. Solution. Let us assume that the twins are initially together at the origin O of frame S, and that the rocket sets off at time t = 0 in the positive x direction. Let S' be an inertial frame moving in the positive x direction with a speed V with respect to S, and S" be an inertial frame moving in the negative x direction with speed V with respect to S, and let us assume that at times t = t/ = t// = 0 the origins of S, S", and S"' coincide. The transformation equations between S and S' are x' = T(x- Vt) (13) t' = r(t-?f\ (14) or x = T(x'+ Vt') (15) t = T(t+?f\ (16) The transformation equations between S and S" are x" = r(.x+ Vt) (17) t'' = T(t+¥f\ (18) or x = T(x" - Vt") (19) t = r(r-^f\ (20) On the outward trip the rocket is at rest with respect to frame S\ and on the return trip the rocket is at rest with respect to frame S". Let us consider three events.
870 Some Consequences of the Lorentz Transformation Event 1. The rocket takes off. Twin B passes from a state state of rest in S'. The coordinates of this event in S are x = 0 / = 0 and in S' are x' = 0 t' = 0 Event 2. The rocket turns around. Twin B passes from a state of rest in S' to a state of rest in S". The coordinates of this event in S' are x' = 0 t' = r (23) From Eqs. (15) and (16) it follows that the coordinates of the event in S are x = TVt t = Tr (24) and from Eqs. (17) and (18) it follows that the coordinates of this event in S" are x" = 2T2Vt t" = T2( 1 + -^ \r (25) Event 3. The rocket lands. Twin B passes from a state of rest in S" to a state of rest in S. The coordinates of this event in S" are x" = 2T2Fr (26) t" = T2(\ + ^ \T + r = 2T2r (27) From Eqs. (19) and (20) it follows that the coordinates of this event in S are x = 0 t = 2Tr (28) Clearly the twin B has aged an amount 2r because according to clocks at rest with respect to him he has traveled for a time 2t. But when he returns, the time that has elapsed on clocks at rest with respect to twin A is 2Tr; hence twin A has aged by an amount 2Tr. PROBLEMS 1. A particle at the origin of an inertial frame R executes simple harmonic motion along the y direction with angular frequency co and amplitude A. Determine the position and velocity as functions of the time in an inertial frame R' moving with velocity V in the x direction. Calculate the time and distance between successive maxima of y'. 2. A thin rod AB of proper length 4a is traveling along the x axis of a frame R with a speed V3 c/2 in the positive x direction. A hollow cylinder CD of of rest in S to a (21) (22)
Problems 871 proper length 2a is placed with its axis along the X axis, so that when the ends of the cylinder are open the rod will pass through the cylinder. The end C of the cylinder is located at x = —2a and the end D at x = 0, and both ends are equipped with hypothetical devices capable of closing them off with impenetrable and immovable walls. a. Show that in the frame R, in which the cylinder is at rest, both the rod and the cylinder have length 2a. b. Show that in the frame R', in which the rod is at rest, the length of the rod is 4a, whereas the length of the cylinder is only a. c. Suppose at time t = 0 in frame R the front end B of the rod is located at x = 0, and both ends of the cylinder are suddenly closed. Is the rod trapped in the cylinder as might be inferred from the answer to part a, or is the rod cut into two parts as might be inferred from the answer to part b? Reconcile this paradox.
SECTION Relativistic Dynamics of a Particle
88 Mass and Momentum INTRODUCTION In Part 1, Section 3, the basic principles of Newtonian dynamics were presented and should be reread at this point. In Chapters 88 and 89 we consider the modifications necessary when one is dealing with speeds comparable to the speed of light. Rather than have the reader skipping back and forth between this chapter and the corresponding chapter in Part 1, Section 3, we repeat the material that is common and simply make modifications where necessary. THE INTERACTION OF PARTICLES Consider two isolated particles, each of which because it is isolated is moving with a constant velocity with respect to an inertial frame. If these two particles come close together, the velocity of each will be found to change, and we say that the particles interact with each other. Our first objective is to describe this interaction. One of the simplest ways to imagine two particles to be interacting is to suppose that one particle has some thing or property that it exchanges or passes on unchanged to the second particle. As a result of this exchange each particle changes, and the amount of whatever is exchanged is a measure of the interaction. In the remainder of this chapter we consider whether such a property could exist, and if it could, does it? COLLISIONS If two particles are approaching each other, and if the interaction between the particles has a finite range, or at least is negligible beyond a certain range, the motion can be broken down into three periods: an initial period during which 875
876 Mass and Momentum the particles are moving toward each other with constant velocities, an intermediate period during which the particles interact, and a final period in which the original particles or some other set of particles are moving away with constant velocities. If these conditions are satisfied we refer to the interaction as a collision. In the following section we assume that we can treat the interaction between two particles as a collision in which the interaction takes place instantaneously at a point, and we speak of the particles being free before and after the collision. Because there are no preferred directions or points in an inertial frame, we can write down the following two theorems. Theorem 1. For every collision that occurs, a collision in which the velocities of the incident and the outgoing particles are reflected in a plane is a possible collision. Theorem 2. For every collision that occurs, a collision in which the velocities of the incident and the outgoing particles are rotated by the same arbitrary amount is a possible collision. QUANTITIES CONSERVED IN A COLLISION If two particles a and b, with velocities \(a) and \(b), respectively, collide and two or more particles c, d, . . . emerge from the collision with velocities v(c), \(d), ..., respectively, and if there exists a quantity g, associated with a particle, which depends only on the intrinsic nature of the particle and its velocity and satisfies the condition g[a,v(a)] + g[b,y(b)] = g[c,v(c)] + g[d,v(d)] +■■■ (1) then we say that the quantity g is conserved in the collision. We find it convenient in the following sections to use the modified velocity u = y v (2) where / 2 \ -V2 / 2 \*/2 -(-?) H'+!?) <3) rather than the velocity v. In terms of u a conserved quantity is one that satisfies the relation g[a,u(a)] + g[b,u(b)] = g[c,u(cj] + g[d9u(d)] + • • • (4) From the definition of conserved quantities we can write down the following two theorems:
Quantities Conserved in a Collision 877 Theorem 3. If g'(u) and g"(u) are conserved quantities, any linear combination of g'(u) and g"(u) is a conserved quantity, or equivalently g(u) = Ag'(u) + Bg"(u) (5) where A and B are arbitrary constants, is a conserved quantity. proof: The proof of this theorem follows immediately from the definition of a conserved quantity. Theorem 4. If there exists a conserved quantity of the form g(u) = <* + /?/* (u) (6) where a and /3 are particle parameters and h(u) is a universal function of the velocity other than a constant, and fi is not identically zero, the ratio of the value of /3 for one particle to the value of /3 for a second particle is unique for the given pair of particles. proof: See Proof of Theorem 4 in Chapter 8. From Theorems 1 and 2 we can derive the following theorems. Theorem 5. If g(u) is a conserved quantity, then g( — u) is a conserved quantity. Theorem 6. If g(w,, w , uk) is a conserved quantity, where / =£j =£ /c =£ /, then g{uj,uk,u^) is a conserved quantity. Theorem 7. If g(w/?«-, w^) is a conserved quantity, where i =£y =£ k =£ /, then g(— w,, «y, w^) is a conserved quantity. Since all inertial frames are equivalent, in the sense stated in Postulates I and II, we can write down the following theorem: Theorem 8. If a quantity is conserved in one inertial frame, it is conserved in all inertial frames. When one transforms from an inertial frame R to an inertial frame R' moving with a velocity -V with respect to R, the quantity u transforms as follows (see Eq. 53, Chapter 86). u'-u+7U + (r-l)(u.U)(-^) (7)
878 Mass and Momentum where UeeTV r.(.-sr.(.+sy It follows that Theorem 8 is equivalent to the following theorem. Theorem 9. If g(u) is a conserved quantity then g[u + yU + (T — 1)(U • u) (U/ U2)] is a conserved quantity for arbitrary constant U. POSSIBLE CONSERVED QUANTITIES Consider a collision between two identical particles initially moving along a line with the same speed but in opposite directions. If there exists a conserved quantity, the possible products of the collision and the possible values of the velocities of the products will in some way be restricted. Conversely if we know that certain products and velocities are either possible or impossible, this information tells us something about the possible conserved quantities. In this section we see that if certain fairly plausible products and velocities are assumed to be possible, we can put very severe restrictions on the possible conserved quantities. Whether our assumptions are correct is of course a matter of experiment. We assume first that a collision is possible in which the only effect is to change the directions of the two identical colliding particles. Secondly, if the two particles are approaching the point of collision along the same line, then from symmetry we expect that they will recede from the point of collision along a common line. We do not assume that this is necessarily so but merely that it is possible. Finally, and this is our least plausible assumption, we assume that the line of recession may have any direction. With these very plausible assumptions we show in the following two theorems that the possible conserved quantities are extremely limited. Theorem 10. If one assumes that when two identical particles moving along a line with the same speed but in opposite directions collide, it is possible for the same particles to emerge moving with their original speeds and in opposite directions but along a different line, and that the directions of the line of approach and the line of recession may assume any value, then any conserved quantity depending only on the nature and velocity of a particle must, in relativistic mechanics, be of the form g = \ + *2liiui+ryc (1°) i where A, ju/5 and v are particle parameters, that is, quantities that are constant for a given particle, although their values vary from particle to particle. (8) (9)
Possible Conserved Quantities 879 proof: Since the colliding particles and the products are all identical in the collision described, the functions g in Eq. (4) will all correspond to the same particle and we use the notation g(u) rather than g[a, u(a)]. Since g(u) must be conserved in the given collision, it follows that g(un) + g(-un) = g(un*) + g(-un*) (11) where w, the absolute value of u, is the same for each particle, n is a unit vector parallel to the precollision velocities, and n* is a unit vector parallel to the postcollision velocities. If there exists a quantity g(u) that is conserved in one inertial frame, it will be conserved in all inertial frames. If R' is a frame that is moving with a modified velocity — U with respect to R, sl particle with modified velocity u with respect to R will be moving with a modified velocity u' = u+YU + (r-l)(u-U)i (12) with respect to R'. Converting the precollision and postcollision modified velocities in Eq. (11) to the values they would have in frame R\ and assuming that g is still a conserved quantity, we obtain g(yU + wa) + g(yU - wa) = g(yU + wa*) + g(yU - wa*) (13) where the quantity a, which should not be confused with the acceleration, is defined (r- l)(U-n)U a = n+^ ^- }— (14) U2 V or equivalently by the relation _ (r-l)(U>a)U n = a - (15) TU2 V ' Since n is a unit vector, the components of a are not all independent but must satisfy the auxiliary condition (c2 + U2)(\ - S«,2) + 2 S U,Ujaiar 0 (16) which is obtained by squaring Eq. (15) and making use of the definition of T given by Eq. (9). It follows from the preceding development that if g(u) is a conserved quantity it must satisfy Eq. (13) for arbitrary choices of u and U and also for arbitrary choices of a and a* consistent with the constraint condition Eq. (16). This requirement puts severe restrictions on the form of g. We now determine what these restrictions are. If we take the second derivative of Eq. (13) with respect to w, remembering that y is a function of w, and set u = 0 we obtain 2 2 a,«/&/(U) = S 2 «/**/S,(U) (17)
880 Mass and Momentum where d2g(U) Z^^JuJUj (18) Since Eq. (17) must be true for any choices of a and a* consistent with the constraint condition, it follows that the function /(»)=SS«/% (19) 1 J must be independent of our choice of values of a consistent with Eq. (16). If we make use of the method of Lagrange multipliers, we can show that a necessary and sufficient condition for this to be true is that the function ' j (c2 + U2)(\ - 2¾2) + S 2 U.Uja.a] (20) be independent of our choice of the variables ax, a2, and a3. It follows that 32F(a) ddiddj = 0 (21) Substituting Eq. (20) in Eq. (21) we obtain gij(U) = h(V)[(c2+U2)8ij-UiU/] (22) To determine h(U) we note that &jk = h[{c2 + U2)StJ - U,Uj] + h[2Uk8tJ - Uj8ik - Ufijk] (23) The derivative gijk is equal to the derivative gkji, hence hk[(c2 + U2)8tJ - U,Uj] + h[2UkStJ - Uj8ik - U,8Jk] = h{(c2 + U2)8kJ - UkUj] + *[2^ - Uj8ki - Uk8jt] (24) Choosing i =£y = k we obtain \ {(c2+U2) -Uft + ^Wj --3U, (25) Interchanging i andy in Eq. (25) we obtain \[{c2+U2)- U*]+ ^1/,^- -3¾ (26) Solving Eqs. (25) and (26) for ^./A we obtain h c2+U2 (27)
Possible Conserved Quantities 881 Equation (27) provides us with the three partial derivatives of ln/*(U). If we solve for h we obtain h=r(c2+ U2)~V1 (28) where v is a constant. Substituting Eq. (28) in Eq. (22) we obtain gij = v(c2 + UY3/2[(c2 + U2)8V - UtUj] (29) Equation (29) provides us with all the second derivatives of g. If we solve for g we obtain g(U) = A + 2 A<U,+ v{c2 + U2)l/2 = X + 2 toU,+ vYc (30) i i where A, /xl5 jh2, and ju3 are constants. We have shown that if Eq. (13) is true for arbitrary values of u and U, and also arbitrary values of a and a* consistent with the constraint condition (16), then g must be of the form given by Eq. (30). Conversely it can be shown by direct substitution that if g is of the form given by Eq. (30) then-Eq. (13) is satisfied for all values of u and U, and also all values of a and a* consistent with the constraint condition (16). Theorem 11. If there exists a quantity g = A + 2 Wi+vyc (31) i that is conserved in a collision, then a. There exists a particle parameter ju and a set of universal constants 2?!,1?2,1?3, and C such that ft = BiVl (32) v=Cp (33) If at least one of the parameters ^, v is not zero, the ratio of the value of jit for one particle to the value of jit for a second particle is unique for the given pair of particles. b. The quantity G = AX + 2 Biiaii+ Cwc (34) i is a conserved quantity not only for the particular values of A, Bh and C for which G reduces to g, but for any other choice of the constants A, Bh and C. proof: Using Theorems 3, 5-7, and 9 we can show that if there exists a conserved quantity g = A + 2 ftw/+ vicl + "2)1/2 (35)
882 Mass and Momentum then the following quantities are also conserved: ft = k [ S(u) + g{ -u)] = X + v(c2 + «2)'/2 £2 = i[s(u)-s(-u)] = 2fw (r - i)(u • u)u gs = g3 u+yU + = HiyU, + [i, u2 (r-i)(u-u)£y,. t/2 &(») ft = 2¾ [ ft(«) + ft(-«)] = ftCY = M,(^2 + «2) (r - i)(u • u)u 1/2 ft = ft u+ yU + = p(r- i)(c2 + M2)' + v u2 ./2 (U-u) ft(«) 2(r-i) Z-n[^(u) + ft(-u)] = ,(c2+«2) 2n'/2 (36) (37) (38) (39) (40) (41) (42) (43) ft = § [ ft(u) - £7(-u)] = *U • u - *§ U,u, (44) £io = 2jj. [ ft(M'-'"/'M*) " ft( - w/> wy> uk) ] = ™/ (45) g\\ssg\-gs = * (46) Since nxui9 /x2M/, jtx3w/, and pw, are conserved quantities, by virtue of Theorem .4 there exists a set of constants BUB2,B3, and C and a particle parameter /a such that ft = £/f* v = C\i (47) (48) The same result follows from the fact that ^(c2 + w2)1/2 and v(c2 + w2)1/2 are conserved quantities. Finally from Theorem 3 any linear combination of the quantities above is a conserved quantity; hence G = AX + 2^/«+ QK*2 + w2)V2= A\ + 2^-/^/+ CWC (49) is a conserved quantity for any choice of the constants A, Bt, and C.
Linear Momentum 883 THE BASIC POSTULATE OF RELATIVISTIC DYNAMICS From the preceding results it follows that if any velocity dependent quantity at all is conserved in a collision, there exists a quantity jtiu that is conserved. Experiment reveals that there is such a quantity. Postulate IV. It is possible to associate with each particle i in the universe a parameter /x(/) having the following property: if two or more particles interact with each other, the quantity 2)/j^(0u(0> where u = [1 — (u2/c2)]~1/2v, has the same value at the end of the interaction as it had at the beginning of the interaction. MASS Given two particles i andy, the ratio jti(/)/jti(/) is unique but the values of /x(/) and ju(j) are not. We are free to assign any value to one particle in the universe. Once this value is assigned however the value of jti for all other particles is fixed. It is customary to pick some particle oasa reference, assign it unit value—that is, let \i{o) = 1—and refer to the quantity 11(1)/11(0) = m(i) as the mass of particle /. Then the mass of any particle i can in principle be determined by allowing the particle to interact with the reference particle o and noting that ju(/)Au(/) + f*(o)Au(0) = 0 (50) where Au(/) and Au(o) are the changes in the modified velocities of particles i and 0, respectively. This fact is summarized in the following definition. Definition 1. If we choose a particle o as a reference particle, the mass of any particle i is defined as the negative of the ratio of the change in the modified velocity of o to the change in the modified velocity of i when the two are allowed to interact and the modified velocities are measured with respect to an inertial frame, that is, Au(o) LINEAR MOMENTUM The quantity mu is of great importance in relativistic dynamics. Definition 2. If in an inertial reference frame a particle of mass m is moving with a modified velocity u = yv = [1 - (v2/c2)]~l/2y, its linear momentum p
884 Mass and Momentum is defined by the relation P = rnvi (52) The momentum obeys essentially the same transformation law as does the modified velocity u, that is, p' = p-myU + (r-l)(p.U)-^ (53) where -1/2 (54) r = U = TV (56) If we break the above vectors into components parallel to V and components perpendicular to V we obtain pf, = rP|| - myV = r(p„ - myV) (57) Pi = Px (58) CONCLUSION The introduction of the concept of momentum provides us with a means of quantitatively describing the interaction between two particles. In the following chapter we pursue this subject further. PROBLEMS 1. A particle of mass m, moving in the positive x direction with a speed v{ collides and coalesces with a particle of mass m2 moving in the negative x direction with a speed v2. Determine the mass and velocity of the resultant particle. 2. Show that if there are no external forces acting on a rocket, the differential equation governing its motion is M(dV/dM) + U[\ — (V2/c2)] = 0, where M is the instantaneous mass of the rocket, U is the constant speed of the exhaust gases relative to the rocket, and V is the instantaneous speed of the rocket. Verify that the solution can be put in the form V/c = [1 - (M/M0)2U/c]/[\ + (M/M0)2U/cl where M0 is the initial mass of the rocket.
89 Force INTRODUCTION In analyzing the interaction between two or more particles, we are frequently interested only in the behavior of one of the particles, not in the associated changes in the other bodies. The concept of force is useful in enabling us to perform such analyses. FORCE If an agent acts on a particle and causes the momentum of the particle to change, we say that the agent has exerted a force on the body. More precisely a force is defined as follows. Definition 3. If in an inertial frame the momentum p of a particle is changing with time, the particle is said to be acted on by a force and the force is defined quantitatively by the relation f=p (l) where p is the time rate of change of the momentum of the body. TRANSFORMATION OF THE FORCE If we are given the force in one inertial frame, how do we determine the force in a second inertial frame? In classical dynamics the force is unchanged. This is not so in relativistic dynamics, and in this section we determine the law of transformation. 885
886 Force From Eqs. (53) and (7) in Chapter 86 u'-u-TU + (r-1)(11-U)(i) (2) hence *, dP du' dt' dt' \ du - dy U + (r - 1)( du • U)(U/ U2) = m\ = m But T[dt-(dr-\/c2)] J ^(du/dt)-(dy/dt)U + (T- l)[(du/dt)-V][V/U2] ~m r[l-(v-V/c2)] (4) -(f)-i(-)-f-' (5) _ m d /v2„2x _ 1 d (n2x = * -^ (p • p) = —L_ P. ^- 2myc2 dt myc2 dt = -Lfflyv.i=vif (6) myc c u = rv (7) Substituting Eqs. (5)-(7) into Eq. (4) we obtain f, = f-(f-v/c2)rv + (r-i)(f.V)(v/r2) r[i-(vv/c2)] If we break the vectors above into components parallel to V and perpendicular to V we obtain ., _ f||-(/||t>n + fx-Vx)(VA-2) (9) l-(v,V/c2) r Ll (10) x r[i-(,„F/c2)]
Problems 887 PROBLEMS 1. Show that the relativistic motion of a particle that is attracted toward a point by a force that varies inversely as the square of the distance of the particle from the point is a precessing ellipse. Show that the precession of the perihelion of Mercury resulting from this effect is about 7" per century. The actual value is about 40" per century and can be accounted for correctly only by the general theory of relativity. 2. A particle of mass m, charge q, and initial velocity v0 enters a uniform electric field E perpendicular to v0. Find the subsequent trajectory of the particle and show that it reduces to a parabola as the limiting velocity c becomes infinite.
90 Work and Energy ENERGY From Theorem 11 in Chapter 88 it follows that if mu is a conserved quantity, myc2 is also a conserved quantity. The quantity myc2 is very important in relativistic dynamics. Definition 4. If in an inertial reference frame a particle of mass m is moving with a velocity v, its energy E is defined by the relation / 2\-*/2 E=myc2=mll-^\ c2 (1) Since the energy is a conserved quantity, we have the following theorem. Theorem 1. If two or more particles interact, the sum of their energies has the same value at the end of the interaction as it had at the beginning of the interaction. The energy obeys essentially the same transformation law as the quantity y, that is, = TE-p\ (2) MOMENTUM AND ENERGY The following relationship between energy and momentum is frequently useful in relativistic dynamics: E2=p2c2+m2c4 (3) It can easily be proved by simply substituting the definitions of E and p into Eq. (3) and reducing the equation to an identity. 888
WORK AND ENERGY Work and Energy 889 The work done on a particle by a force f in an infinitesimal displacement dr is defined by the relation dW = f • dr (4) Theorem 2. If a particle is acted on by a force f then dW=dE (5) where dW is the work done by the force and dE is the change in the energy E of the particle. proof: = _L d_//r2w, = _L Af»W+ mf-J\dt c2 dp . = ~E*dtdt myc = t*vdt = t*dr (6)
SECTION Four Dimensional Formulation of Relativistic Mechanics
91 Four Dimensional Formulation of Relativistic Mechanics INTRODUCTION Many of the results we have obtained so far can be simplified mathematically if we express them in tensor form, as shown in this chapter. The reader who is not familiar with tensors should read Appendix 23. SUMMATION CONVENTION Throughout the chapter we make use of the summation convention (see Appendix 23). THE LORENTZ TRANSFORMATION Let R be an inertial frame, S a rectangular Cartesian coordinate system fixed in R, and C a clock fixed in R. Let i?' be a second inertial frame, S' sl rectangular Cartesian coordinate system fixed in R\ and C" a clock fixed in R'. Let us further assume that the standard of length is the same in both coordinate systems and that the standard of time is the same for both clocks. The transformation connecting the coordinates x, y, z, and t and x\ y\ z' and t' is a Lorentz transformation. If we allow R and R\ S and S", C and C" to be arbitrary while maintaining the standards of length and time the same, the resulting set of transformations forms a group of transformations called the complete Lorentz group, the inhomogeneous Lorentz group, or the Poincare group. If out of the complete Lorentz group we consider only the transformations for which the origins of S and S" coincide at t = /' = 0, the resulting set of transformations forms a group called the homogeneous Lorentz group. If 893
894 Four Dimensional Formulation of Relativistic Mechanics out of the Lorentz group we consider only the transformations for which the axes of S and S' are collinear at time t = t' = 0 and the velocity of R' with respect to R lies along one of the axes of S, for example the x axis, the resulting set of transformations forms a group called the special Lorentz group. In this chapter we restrict ourselves to the special Lorentz group. The results we obtain are however more general. The transformations in the special Lorentz group are of the form x' = T(x - Vt) (1) y' = y (2) (3) where COORDINATE SYSTEMS When defining tensors the first step is to decide on the space within which one is operating and the coordinate systems that are to be considered. In this chapter we consider our space to be the space defined by the coordinates x, y, z, and t in some inertial frame R. We then define coordinates x1 = x (6) *2 = y (7) x3 = z (8) x4=ct (9) and consider the set of all coordinate systems related to the x' by the special Lorentz group, that is, all coordinate systems that can be obtained from the x' by a transformation of the form ^- = 1^-^4) (10) x* = x2 (11) xy = x3 (12) x4' = r^4_Zx.j (13)
Some Tensors 895 The transformation above can be written in the form or simply where \X\ \x2'\ = r r 0 0 -r- m= 0 0 1 0 0 1 0 0 xl = OjxJ 0 0 1 0 0 1 0 0 V 0 0 0 0 -r- "1 *' x2 x3 x4 _ 0 0 (14) (15) (16) TENSORS Since n/' _ dx' j dxJ (17) it follows from the definition of a tensor in Appendix 23 that any quantity Ajfj* ;; lj» obeying the transformation law Ar$"mfr = 0kl0k2' • 'OHM2' ' '0JpA^-jk- (18) is a tensor with respect to the set of coordinate systems described. SOME TENSORS From the definition of a tensor given in the preceding section we find immediately that xl and dx1 are contravariant vectors. In the following sections we generate some additional tensors.
896 Four Dimensional Formulation of Relativistic Mechanics THE METRIC TENSOR If for an arbitrary coordinate system we define then 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 -1 o o -r- o o -r- 1 o o 1 o o o o o o 0 0 1 0 0 0 0 -1 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 -1 0 0 (19) r 0 0 0 0 1 0 0 1 c 0 0 (20) It follows that gy is a symmetric tensor. If we define g{j as the metric tensor, the resulting Riemannian space is one for which the distance ds between neighboring points is defined by the relation ds2 = gijdxidxJ ={dxx) +(dx2) +(dx3) We assume this metric tensor throughout this chapter. (dx4) (21) FOUR-VELOCITY Since the components of the metric tensor are constant, the absolute derivative (see Appendix 23) of a tensor reduces to the ordinary derivative. It follows that if the trajectory of a particle in our four dimensional space is given by jc' = jc'(ii) (22) where u is a scalar, then dx1 /du is a vector. The trajectory of a particle can be described by giving x'(s), but rather than s it is customary to introduce a parameter r defined by the equation r2=-4 (23)
Four-Momentum 897 It follows that dx11 dT is a contravariant vector. We call dx'/dr the four- velocity and designate it ul. Noting that jT_[^-(l/c2)(^ + ^+^)]1/2_/t ^./2 dt dt we obtain -(<-£) -\ w M1 = M2 = " dr .dx2 ' dr Z-> ..yd*!. 7 dt ..yd*l 7 dt dx' dt = 7¾ = yvy (25) Hence (26) (27) Thus the components of the vector w' in a particular frame are yvx, yvy, yvz, and yc. FOUR-MOMENTUM If we take the product of the mass mofa particle and its four-velocity u' we obtain a contravariant vector called the four-momentum, which we designate pl. Thus px — mux = myvx (30) p2 = mu2 = rayt> (31) /?3 = mu3 = myuz (32) p4 = mi/4 = rayc == — (33) The first three components of the four-momentum are the ordinary relativistic components of the momentum. The fourth component is the energy divided by c. In the four dimensional formulation the usual laws of conservation of momentum and energy combine to form the single law of conservation of four-momentum.
896 Four Dimensional Formulation of Relativistic Mechanics FOUR-FORCE If we take the derivative of the four-momentum with respect to r we obtain a contravariant vector, which we define as the four-force/'. The components of /'" are /■-^-rj^)-* f-%-iiW4 f-g-iiw* r-^-ri^-If-J.-, (34) (35) (36) (37) The first three components of the four-force are related to the ordinary components of the force. The fourth component of the four-force is related to the work done per unit time by the force.
SECTION Relativistic Lagrangian and Hamiltonian Mechanics
92 Relativistic Lagrangian Mechanics LAGRANGE'S EQUATIONS OF MOTION Theorem 1. If a particle is moving under the action of a force f, the relativistic equations of motion of the particle in terms of the set of generalized coordinates #i,#2>93 are A. dt 3fl(q,q,Q d*(q,q,Q H = 0, (1) where R = -mcA -(7)1 1/2 (2) and the Qt are the generalized components of the force. If the generalized components of the force are derivable from a potential U, that is, if there exists a C/(q, q, 0 such that 0,= 9£/(q,q,Q 3£/(q,q,Q the equations of motion can be written (3) dt dL(q,q,t) 3L(q,q,/) = 0 (4) where L-R-U (5) 901
902 Relativistic Lagrangian Mechanics proof: Classically, the equations of motion of a particle can be written as follows: fx-i(m*)- d dt d dt ^(H+j^+H) dt \ dx ) f = <L(m Jl dt\dz ) Relativistically, the equations of motion can be written (6) (?) (8) A dt /* = ^("»^) = ^i^r mc2[ 1 x2+y2 + z2 1/2-1 dt\dx ) r = £( ML] ]y dt\dy ) f - d (dR\ Jz dt\dz ) (9) (10) (11) Starting with Eqs. (6)-(8) it is possible to derive the classical Lagrangian equations of motion (see Chapter 48). If instead we start with Eqs. (9) and (10), we can use the same arguments with T replaced by R and we end up with the results stated in the Theorem.
93 Relativistic Hamiltonian Mechanics THE RELATIVISTIC HAMILTONIAN Given the relativistic Lagrangian of a particle v211/2 L = — mc' 1 c -(!)-" <'> the generalized component of the momentum conjugate to a generalized coordinate qt is defined by P^~H~ (2) the Hamiltonian is defined by H = ^Piq-L (3) and Hamilton's equations are . 3*(q,P,Q ,,. ag(*p>0 »- H~ ^ Theorem 1. If U is not a function of q and the equations of transformation between the set of Cartesian components x and the generalized components q are not functions of the time then H = myc2 + U (6) where -my"2 <7> 903
904 Relativistic Hamiltonian Mechanics proof: ^-SM-i = S|rri if then dxj dx- Hi ~ H If furthermore U =£ U(q) then U =£ U(x) and -?(s||^)*- */ = */(<0 (9) *y-?aJ* (10) (11) Making use of Eqs. (8), (11), and (12) we obtain »-2S!^*-i-22>*%|f*-i = 2rnyXjXj- L = myv2 - L = myu2 - I - ^- - U J = myh2+ -^ J + U= my(v2 + c2- v2) + U = myc2 + U (13) From the theorem above we find that even though the relativistic Lagrangian L = —(mc2/y) — U does not have the simple interpretation of the classical Lagrangian, where L = T — U, nevertheless when the classical Hamiltonian is equal to the total energy, the relativistic Hamiltonian also is equal to the total energy.
SUPPLEMENTARY MATERIAL
Appendices
APPENDIX 18 Inverse Transformations Theorem 1. Let y = yx, y2, . . . ,yn be a set of n quantities that are uniquely defined over a region R of the space x = xux2, . . . , xn, by the set of equations y = y(x). If each of the functions yi(x) and all its first derivatives are continuous in the region R, and if the Jacobian 7(y/x) is not zero anywhere in the region R, the set of equations y = y(x) can be solved to obtain the variables jc,- as functions of the set of variables y, and the solutions will be unique. proof: We use mathematical induction to prove the theorem. Let us first consider the case of n = 1. For this case we are given a function y(x) and we wish to show that if y(x), and dy/dx are continuous in a region R and if dy/dx 7^0 anywhere in R, we can solve for xasa function of y and the solution is unique. We note first that if dy/dx is continuous in the region R, and if dy/dx =^ 0 anywhere in R, the sign of dy/dx is the same throughout R. From this fact and the continuity of y(x) it follows that y(x) is either a monotonic increasing or a monotonic decreasing function of x in the region R. It immediately follows that we can invert the function y(x) to obtain x(y). We now prove that if the theorem is true for n = k — 1 it will also be true for n — k. Let us therefore assume the theorem is true for n = k — 1. We then consider the case for which n = k. If we expand the Jacobian 7(y/x) for the case in which n = k in terms of the cofactors of the first row we obtain ,/^^) = 1¾^^) \ XX,X2, • • • , Xk J i={ OX; \ xx,x2,x3, . . . , xk ) If the Jacobian 7(y/x) is not zero, at least one of the cofactors must be nonzero. For simplicity let us assume that it is the cofactor of the term in the first row and first column. Thus \ -*l>-*2>"*3> ' ' ' ■> xk J 909
910 Inverse Transformations From Eq. (2) it follows that \ -^2,-^3, • • • , Xfc / Since the theorem is true for n = k — 1, it follows from Eq. (3) that we can invert the set of equations 7/=7/(x) / = 2,...,/: (4) to obtain xi = xi(xl9y29y39...9yk) (5) If we substitute the set of Eqs. (5) into the equation yx = yx(x) we obtain y\ =j>i(*i>*2> •••>**) = y\[xi,x2(xl9y29 . . .,^), . . .,xk(xl9y29 . . .,yk)] (6) It follows that 8^lj2 ^) _ 971(-^1^2,...,½) y 971(^1,^2,...,½) 9^1,72 ^) /=2 3¾ 3*i 97i(x) , £ 97i(x) r/ xi9y29...9yk / = 2 °*i V x\> S2> ' ' ' >Sk J dx Multiplying Eq. (7) by J(x]9 y2, . . . ,yk/x]9x2, . . . , xk) we obtain ^1(^1^2. • • • ,yk) ( xX9y29 ...,yk\j* tyi(*)J(xi9y29...9yk\ dxx \ xl9x29 . . . , xk ) /=1 dxt \ x]9x2, . . . , xk ) (8) But the right hand side of Eq. (8) is just the expansion of 7(y/x) in terms of the cofactors of the first row, as can be seen by comparing Eqs. (1) and (8). Thus ty\{xx,y2,...,yk) ( x]9y29...9yk \ = / yX9y2,...,yk \ (9) 3*i \ xX9x29 . . . , xk ) \ xX9xl9 . . . , xk J The two Jacobians in Eq. (9) have been assumed to be nonzero. It follows that dyl(xl9y2,...,yk)^Q (1Q) dxx From Eq. (10) it follows that we can invert the equation 7i=7i(*i>72>---7^) (11) to obtain xi = x](yl9y29...9yk) (12)
Inverse Transformations 911 If we combine Eqs. (5) and (12) we obtain xi = xi(yl9y29...9yk) /=1,...,/: (13) Thus if the theorem is true for n = k — 1 it is also true for n = k. This completes the proof of the theorem. It should be noted that if J(y/x) is zero somewhere in the region R it does not necessarily follow that we cannot invert the transformation y = y(x). This can be quickly seen by noting in the case n = 1 that if the function y = y(x) has an inflection point in the region R but not an extremum, then dy/dx will be equal to zero at the inflection point, but y = y(x) will still be either a monotonically increasing or a monotonically decreasing function of x.
APPENDIX Change of Variables in an Integral Theorem. If x = xl9x2, . . . , xn and y = yx,y2, • • • ,y„ are two sets of variables related by the set of transformation equations y = y(x), where the functions y^x) are continuous and have continuous first derivatives in some region R of the space x and if 7(y/x), the Jacobian of the transformation, is not zero anywhere in the region R, then jRf(x)dx=jRAx(y)]\J{^)\dy (1) where R' is the region in the space y corresponding to the region R in the space x, and the limits of integration are chosen in such a way that the integrals }Rdx and JR>dy, which give the "volumes" of the regions R and R\ respectively, are both positive. proof: Consider the integral I=J ■■■ Jfdx,---dxn (2) We first replace the integration over the set of variables xl9x29 . . . , xn by an integration over the set of variables yl9x29 . . . 9 xn. To do this we note that during the integration over the variable xx the variables x2, . . . , xn are held constant, hence = ^,(7,,^, . ■ . , x.) = / xux2, • • • , *, \^ (3) 1 3ji 7 \ yux2,...,xn) ' Substituting Eq. (3) in Eq. (2) we obtain '-//••■M^^^)*'*'-^ <4) We next replace the integration over the set of variablesy},x2, . . . , xn by an integration over the set of variables yl9 y2,x3, ..., xn; then we replace integration over the set of variables yl9 y2,x3, . . . , xn by integration over the set 912 19
Change of Variables in an Integral 913 of variablesy,, y2, y3,x4, . . . , xn. We proceed in this fashion until finally the variables of integration are ^, y2, . . • ,yn. Following this procedure we obtain T=f f fj{ Xx,Xl' ' ' ' > Xn\jl y\>X2>X3> ' ' ' > Xn\ J '" JJ \yi,x29...9xn) \yi,y2>x3,...,Xn) If the Jacobian J(y/x) changes sign in the region R, there will be two or more points x corresponding to a point y in the neighborhood of a point where the Jacobian vanishes (see Appendix 18). To eliminate this possibility we restrict ourselves to regions in which the Jacobian J(y/x) does not vanish anywhere, hence cannot change sign. If it is necessary to handle regions in which the Jacobian changes sign, the best procedure is usually to break up the region into subregions in each of which the Jacobian keeps a constant sign; then consider each region separately. So far nothing has been said about the limits of integration. In the integration over R the limits must be chosen in such a way that the magnitude of JR dx is equal to the "volume" of the region R, and in the integration over R' the limits must be chosen in such a way that the magnitude of fR,dy is equal to the "volume" of the region R\ If the Jacobian is positive the signs of the integrals fRdx and }R ,dy must be the same, and if the Jacobian is negative the signs of the integrals fRdx and fR ,dy must be opposite. This slight awkwardness can be avoided by simply replacing the Jacobian in Eq. (5) by its absolute value, and always choosing the limits in such a way that the integrals fRdx and fR>dy are positive. This would correspond for example in the one dimensional case to integrating from the smaller limit to the larger limit.
APPENDIX 20 Euler's Theorem A function f(xl,x2, . . ., xs) is said to be homogeneous of degree n in the variables xl9x2, . . . , xr, where r < s if and only if f(Xx},Xx2, • • • ,A*r,xr+1, • . .,^) = Ay(x1,x2, ...,¾) (1) for every positive value of A. Euler's Theorem. If f(x]9x2, . . . , xs) is homogeneous of degree n in the variables xl9x2, . . . , xr9 then nf(x},x2, ...,xs)=2j xt r- (2) proof: If we take the derivative of Eq. (1) with respect of A we obtain ' 9/(Ax!?Ax2, ..., \xr9xr+l9 . . . ,xs) d(Xxi) h 3(^/) ~^~ -n\n-lf(xl9x2,...9xs) (3) Noting that 3(Ax/)/3A = xi9 and then setting A = 1 in Eq. (3) we obtain Eq. (2), which is the desired result. 914
APPENDIX 21 The Gram-Schmidt Orthogonalization Process We are frequently given a set of linearly independent vectors that span a particular space but are not orthogonal. The theorems in this appendix provide us with a technique for constructing a complete set of orthogonal vectors from the given set. The technique is known as the Gram-Schmidt orthogonalization process. Theorem 1 is actually a special case of Theorem 2. Theorem 1 is treated separately because it is the case most needed, and because it makes the proof of Theorem 2 more apparent. Theorem 1. If we are given a set of linearly independent vectors a(l), a(2), ..., a(s) spanning an s dimensional space, then the set of vectors b(l), b(2), ..., b(s) defined as follows: b(l) = a(l) b(r) = a(r)-"] J- or equivalently in component form as MO = «/(0 a(r). HJ) HJ) _k hU) (la) (lb) (la') (lb') constitutes a set of linearly independent vectors that span the same space and satisfy the orthogonality condition b(i)-b(j) = b% (2) 915
916 The Gram-Schmidt Orthogonalization Process or equivalently ^bk(i)bk(j) = b% k (2') proof: The quantity b(j)/b(j) is a unit vector in the direction of the vector b(y); hence the quantity {a(r) • [b(J)/b(j)]}[b(j)/b(J)] is just the component of the vector a(r) in the direction of the vector b(y). It follows that the vector b(r) as defined by Eq. (lb) is just the vector a(r) minus its components in the b(l), b(2), ..., and b(r - 1) directions, hence is orthogonal to b(l), b(2), ..., and b(r - 1). Thus b(2) as constructed is orthogonal to b(l); b(3) is orthogonal to b(l) and b(2), . . . ; and finally b(s) is orthogonal to b(l), b(2), ..., and b(s — 1). The set of vectors b(l), b(2), ..., and b(s) are therefore mutually orthogonal. Theorem 2. If we are given a set of linearly independent vectors a(l), a(2), ..., a(^) that span an s dimensional space, and a symmetric matrix [T] of order s, the set of vectors b(l),b(2), ..., b(.s) defined as follows: i,(l) = *,(l) 7=1 2 2 MONACO k I k I Hi) (3a) (3b) constitutes a set of linearly independent vectors that span the same space and satisfy the orthogonality condition S 2 mow*) = 2 SMOW') k I I k I (4) proof: Let a be an arbitrary vector and, b(l),b(2), ..., b(.s) a complete set of vectors satisfying the orthogonality condition (4). The vector a can be decomposed into its components in the directions b(l),b(2), ..., and b(^); that is, there exists a set of scalars a(l),a(2), . . . , a(s) such that or equivalently a = 2«(0»(0 i ak = ^a(i)bk(i) (5a) (5b)
The Gram-Schmidt Orthogonalization Process 917 To solve for a(j) we multiply Eq. (5b) by Tk/b,(j) and sum over k and I. Doing this and making use of the orthogonality condition (4) we obtain SS^AC/') a (A : k—L (6) k I Comparing Eqs. (3b) and (6) it is apparent that b(r) as defined by Eq. (3b) is equal to a(r) minus the respective components of a(r) in the b(l), b(2), ..., and b(r - 1) directions; hence b(r) is orthogonal in the sense of Eq. (4) to b(l), b(2), ..., and b(r — 1). It follows that the set of Eqs. (3) defines a set of vectors that satisfy the orthogonality condition (4).
APPENDIX 22 Legendre Transformations A function f(xl9x2, . . . , xr) = /(x) contains a certain amount of information. Frequently it happens however that the set of variables 3/(x)/3x = 9/(x)/jc,, df(x)/dx2, ... is a more convenient set of variables than the set of variables x. Then we are led to ask whether there is some function of the variables 3/(x)/3x that contains exactly the same information as/(x), that is, a function of the variables 3/(x)/3x that can be obtained from/(x) and from which/(x) can be obtained. The answer to this question is contained in the following theorem. Theorem 1. Consider a quantity f that is a single valued, continuous, and differentiable function of a set of r + s variables x,y = xl9x2, . . . , xr,yx, y2, • • • ,ys and f°r which \d2f(x,y)/dxidxJ\ is not zero anywhere in the region R of the space x in which we are interested. If we define a new set of variables X = XX,X2, . . . , Xr and a quantity F as follows: X = JK } (1) F= S^/-/ (2) i then XJI^1 (3) f=^XiX-F (4) i Furthermore the information contained in the function F(X, y) is equivalent to the information contained in the function /(x, y). We call the quantity F the Legendre transformation of the function /(x,y) with respect to the set of variables x. 918
Legendre Transformations 919 proof: From Eq. (2) we obtain dF(X,y) 8x,.(X,y) 3/(x,y) 3x,(X,y) dXj dX, dX: ax,. (5) Substituting Eq. (1) in Eq. (5) we obtain Eq. (3). Equation (4) is simply a rearrangement of Eq. (2). This proves the first part of the theorem. We now note that if we are given/(x,y), we can use Eq. (1) to find X(x,y), hence x(X, y). If we substitute this result in Eq. (2) we obtain F(X, y). Hence if we know /(x, y) we can find F(X, y). Similarly if we know F(X, y) we can use Eq. (3) to find x(X,y), hence X(x,y). Substituting this result in Eq. (4) we obtain/(x,y). Hence if we know F(X,y) we can find/(x,y). Thus the functions /(x, y) and F(X, y) contain the same information. This completes the proof of the theorem. The proof above rests on the assumption that Eq. (1) can be inverted to obtain x(X, y) and Eq. (3) can be inverted to obtain X(x, y). From Theorem 1 in Appendix 18 it follows that Eq. (1) can be inverted if \dXi(x,y)/dxJ\ = \d2f(x,y)/dxidxJ\ =£0. In the theorem above this is true by hypothesis. To show that Eq. (3) can be inverted it is necessary to show that \dxi(X,y)/dX\ = \d2F(X,y)/dXidXJ\ ^0. If we express |3jcf.(X,y)/3^| and |3^(x,y)/3jcy| in Jacobian form we have and a*,-(X,y) 3*/(x,y) 3x,- _j( xx,x2, ..., xr,yx,y2, ..., ys \ \XX,X2, ...,Xr,yuy2, ...,ys) - j(Xl'X2' •■•'Xr'y\'y^ ■■■'yA { xux2, ...,xr,yx,y2, ...,ys ) It follows immediately from Theorem 2 in Appendix 3 that |3x/(X,y)/8^/| is just the reciprocal of \dXi(x,y)/dxf\. Thus if |ajc,.(X,y)/3*,| =^0 then |8^(x,y)/8xy|^=0. Theorem 2. If F(X, y) is the Legendre transformation of the function /(x, y) with respect to the set of variables x, then proof: From Eq. (2) dF(X,y) _ 3/(x,y) ty dyj (6) (7)
920 Legendre Transformations It follows that 3F(X,y) _ a*,(X,y) 3/(x,y) 8x,(X,y) 3/(x,y) 3jy r ty ' ~ 3¾ ty ty (8) Substituting Eq. (1) in Eq. (8) we immediately obtain Eq. (6). SUMMARY The relations between a function /(x, y) and F(X, y), the Legendre transformation of the function with respect to the set of variables x, are summarized below: X< = ^T <9a) x^^x~ (9b) F-TZxiK-f (10a) /= 2*/*/"J7 (10b) i /' 3/(x,y) 3F(X,y) dy, dy, The functions /(x, y) and F(X, y) contain equivalent information. (11)
APPENDIX General Tensors INTRODUCTION Physical quantities can often be described or represented by a set of numbers that are referred to a particular coordinate system. When a quantity is described in this manner, we refer to the set of numbers as a representation of the quantity in the given coordinate frame. Frequently we are given the representation of a quantity in a particular frame and we would like to know its representation in some other frame. To learn this, we need to know how the representation of the quantity transforms when we go from the one frame to the other. We considered this problem in Appendix 6 for a particular set of transformations. In this appendix, we investigate the same problem for a larger set of transformations, and we find as we did in appendix 6 that the representations of certain quantities have particularly convenient transformation properties. NOTATION We use the notation x\x2,x3, . . . , x? to represent the coordinates of a point in a particular / dimensional space. When we wish to distinguish coordinates of the same point with respect to different coordinate systems, we prime the superscripts. Thus (x\x2,x3, . . . , xf\ (xl\x2\x3\ . . . , xf\ (xv\x2\x3\ . . . , x? ), . . . , represent the coordinates of the same point with respect to different systems S, S", S", . . . . For example, in ordinary three dimensional space we could use Cartesian coordinates x, y,z or cylindrical coordinates p,<j>,z or spherical coordinates r,<j>,0 to specify a particular point, and we could make the identifications (x\x2,x3) = (x,y,z), (xl\x2\x3') = (p,<|>,z), and (xv\x2'\x3") = (r,<f>,9). Note that it is much more logical to prime the superscripts than the quantity x, since in each case we are referring to the same point x, but with respect to different coordinate systems. 921
922 General Tensors THE COORDINATE SYSTEMS In the following sections we assume that we are dealing with a set of coordinate systems S = x\x2, . . . , x-f, S' = xv,x2\ ..., x-F, S" = xl\ x2\ ..., x?\ . . . having the following properties: 1. There is a one to one correspondence between the values of the coordinates in one system and the values of the coordinates in any other system. 2. If S and S' are any two systems, the functions xl'(x\x2, . . . , x^) are continuous and have continuous first derivatives. 3. The set of all transformations between the coordinate systems S,S', S", . . . form a group. SUMMATION CONVENTION In this appendix we adopt the following convention. Summation Convention. When a subscript and a superscript are repeated in a given term, summation over these indices is understood. For example, atxl =^0^= axxx + a2x2 + • • • + a^ (1) i Al=^,AS=A\+"-+A{ (2) i NOTATION We make frequent use of the partial derivatives dxl/dxJ. To simplify the expressions involving these partials, we use the notation dxl(x\x2, . . . , xf) e;=—K—^——t (3) J dxJ Theorem 1. The quantities Oj satisfy the relation */#-«*' (4) where fi£ = 1 if V = k' and fi£ = 0 if /' ^ k'. SCALAR A scalar is a mathematical entity A having the following properties. 1. A can be represented in an arbitrary coordinate system S' by a number A'.
Tensors 923 2. If S " is any other coordinate system, the number A " representing A in S " and the number A' representing A in S" are equal, that is, A" = A' (5) VECTORS A contravariant vector is a mathematical entity A having the following properties: 1. A can be represented in an arbitrary coordinate system S by a set of numbers A1,A2, . . . , A? called the components of A in S. 2. If A1 are the components of A in S, and A' the components of A in a second system S\ then ^'" = OfA-i (6) Example. The set of numbers dxl are the components of a contravariant vector. A covariant vector is a mathematical entity ^4 having the following properties: 1. A can be represented in an arbitrary coordinate system S by a set of numbers AX,A2, • . . , A, called the components of A in S. 2. If At are the components of A in S, and ^4r the components of A in a second system S", then ^ = ejAj (7) Theorem. If / is a scalar function, then at any point, the set of partial derivatives df/dxl are the components of a covariant vector. TENSORS The preceding definitions of contravariant and covariant vectors suggest the following generalization of the concept. Definition. A tensor contravariant of rank m and covariant of rank n is a mathematical entity A having the following properties: a. A can be represented in an arbitrary coordinate system S by a set of numbers Aj*jl; jj called the components of A in S. b. If dyl'/Zj™ are the components of A in S and ^/^.'.'.Jf are the
924 General Tensors components of A in 5" then J\j2---Jn K\ K2 Km J\ Jl Jn l\l2---ln V / TERMINOLOGY Instead of saying "The tensor A contravariant of rank m and covariant of rank n with components Ajfj*; ;jr in system S," we generally abbreviate this statement thus: "the tensor A\x\2 "' k" J\j2 ■ ■ -Jn SOME PROPERTIES OF TENSORS Theorem 2 (Addition). If A-yyj™ and Bjy^ are two tensors of the same type, the sum of these two tensors defined as follows: C/' ••■/- = ^4/« •■■/-+ jj/' ■ • •/- (9) J\ ■ • -Jn J\ • • 'Jn J\ • ■ -Jn \ J is also a tensor of the same type. Theorem 3 (Exterior Product). If Aj*;;;j™ and B^\\\'^r are two tensors, the exterior product of these two tensors defined as follows c/,.. .ijcx. K = ^ /, - - ./„,£*,. -^ HO) y, . . .y„/, . . . ls J\ ■ • 'Jn '\ ' • • h V / is also a tensor. If ^4 is contravariant of rank m and covariant of rank n and 5 is contravariant of rank r and covariant of rank s, then C is contravariant of rank m + r and covariant of rank « + 5*. Theorem 4 (Contraction). If Afll2''' \k'' \-m is a tensor, the contraction of this V ' J\j2 ■ • -Jl ■ • -Jn t ' tensor with respect to the indices /^ andy, defined as follows: gixi2... (ik)... im = y ^ /,/2 .../*... '«= ^4 'i'2 •••'*.••'». n n jlJ2--(jl)---Jn • • J\J2---jl-.-jn J\h-'-k--Jn V ' is a tensor. If ^ is contravariant of rank m and covariant of rank n, then B is contravariant of rank m - 1 and covariant of rank n - 1. Example. The contraction of the tensor Ajk with respect to the indices i andy is the tensor Bk = A<k (12) Theorem 5 (Inner Product). If A!' •••*>•••'»■ and B^-;k' , are tensors, the inner product of these tensors with respect to the indices ip and / defined as
follows: Raising and Lowering Indices 925 S,- -^/..-. (/,)•■•/, y-..-.y„ ^/,.../...4 (13) is a tensor. Example. If ^ and £/" are tensors, then ^B™, AJkB{, AJkB^ AJkB{, and ^4^ are aH tensors. Theorem 6 (The Quotient Law). A set of quantities Aj^'//^ form the components of a tensor if an inner product of ^j\\\\lfn with an arbitrary tensor 2?/1,'2 • ■ • l[ is itself a tensor. 717 2 • • -Js METRIC TENSOR If in a given space, the distance ds between a point (x\x2, ..., x?) and a neighboring point (xl + dx\x2 + <ix2, . . . , x? + dx?) is defined by a relationship of the form ds2 = gijdx'dxJ (14) where g^ is a symmetric covariant tensor of rank 2, the space is called a Riemannian space, and the tensor is called the metric tensor of the space. Theorem 7. If g is the determinant of the matrix [ gtj\ and Gy is the cofactor of the ijth element of [ gy], then the quantity S*sy (15) is a contravariant tensor of rank 2, called the conjugate of the metric tensor. Theorem 8. The metric tensor gtj and its conjugate gij satisfy the following relation &*'* = */ (16) RAISING AND LOWERING INDICES Theorem 9. With every contravariant vector A' there is associated a covariant vector At defined by the relation ^/ = ¾^ (17) With every covariant vector Ai there is associated a contravariant vector A' defined by the relation Al = giJAj (18)
926 General Tensors If Bt is the vector associated with A' and C" the vector associated with Bh then A'= C". Theorem 10. With every tensor A)*1? ' Jw, there are associated tensors J J\J2' • -Jn ' griM--±-im (19) g*Aji\:::k...j. (20) These processes of generating tensors from tensors are called lowering an index and raising an index, respectively. If we start with a tensor A and lower (raise) an index to obtain a tensor B, the tensor obtained from B by raising (lowering) the given index is again the tensor A. NOTATION In the remainder of this appendix we use the following notation for partial derivatives: "mJL-»?—- <2i> CHRISTOFFEL SYMBOLS The Christoffel symbol of the first kind is defined as follows: r/y, k = \ tfigjk + ^jgik ~ ^kgij) (22) The Christoffel symbol of the second kind is defined as follows: IW%,* (23) note: Neither of the Christoffel symbols is a tensor. ABSOLUTE DERIVATIVES OF TENSORS If/is a scalar field, then 9,/ is a covariant vector. If a curve is defined by the set of equations xl(s), where s is a scalar, then the derivative dxl/ds is a contravariant vector. In general however, the derivative of a tensor is not necessarily a tensor. In this section and the next we generalize the concept of a derivative to obtain operations that generate tensors from tensors. Given a contravariant vector field A' defined along a curve x'(s), where s is a scalar, the absolute derivative 8Al /8s of the vector A' is defined as 8f = df + rAj<kl (24) 8s ds JK ds v
Covariant Differentiation 927 Given a covariant vector field Ai defined along a curve xl(s), where s is a scalar, the absolute derivative 8At/8s of the vector Ai is defined as 8A, dA, , JvJ V^-1^ (25) Given a tensor field Aj^'/'j™ defined along a curve x'(s), where s is a scalar, the absolute derivative &4/'.'2''' Jw/S^ of the tensor Af^2''' Y is defined as J\Jl ■ • -7« ' Jlj2 ■ • 'Jn (26) fo <fc ^ V 7.---7« <fc ^ P 7,-.-^-^ <fc Theorem 11. The absolute derivative of a tensor is a tensor. COVARIANT DIFFERENTIATION Given a contravariant vector field A1, the covariant derivative of A1 with respect to xJ is defined as follows: A'.^djA' + A"^ (27) Given a covariant vector field Ai9 the covariant derivative of Ai with respect to xj is defined as follows: A,.j = 3,.4,. - AkJ* (28) Given a tensor field A-^ ' K the covariant derivative of A,-1]2"' Y with respect to x is defined as follows: m n ^:::Jr;* = a^:::J:+ S^v.vi-^- S^'.v.t..^ (29) r=\ r=\ Theorem 12. The covariant derivative of a tensor is a tensor.
APPENDIX Matrix Transformations INTRODUCTION A rule that associates with each matrix [a] in a given set of matrices a unique matrix [b] is said to be a matrix transformation. note: The notation used in this appendix is the same as that used in Appendix 12. In particular a row matrix will be designated (a] and a column matrix [a). EQUIVALENCE TRANSFORMATIONS If [a] is any arbitrary matrix of order m X n and [q] and [p] are known nonsingular square matrices of orders m and n, respectively, the transformation defined by the equation [*]-[*][«][/»] (1) is said to be the equivalence transformation of [a] by [q] and [p]. note: If the sets of variables x and y in the bilinear form (j][«][^) are replaced by the sets of variables u and v defined by the transformation equations [x) = [a][u) and (y] = (v][fi], we obtain a new bilinear form (v][b] [w), where [b] is the equivalence transformation of [a] by [/?] and [a]. It follows that equivalence transformations are useful in the discussion of bilinear forms. Any matrix [b] that can be obtained from a given matrix [a] by an equivalence transformation is said to be equivalent to [a]; that is, [b] is equivalent to [a] if and only if there exist two nonsingular square matrices [q] and [/?] such that [b] = [#][**][/>]. 928 24
Similarity Transformations 929 CONGRUENT TRANSFORMATIONS If [a] is any arbitrary square matrix of order n and [p] is some known square matrix of order n and [p]T is its transpose, then the transformation defined by the equation [1>]-[P]T[°][P] (2) is said to be the congruent transformation of [a] by [p], note: If the set of variables x in the quadratic form (x][«][x) is replaced by the set of variables y defined by the transformation equations (x] = [a][y), we obtain a new quadratic form (7H&H7), where [b] is the congruent transformation of [a] by [a]. It follows that congruent transformations are useful in the discussion of quadratic forms. Any matrix [b] that can be obtained from a given square matrix [a] by a congruent transformation is said to be congruent to [a]; that is, [b] is congruent to [a] if and only if there exists a square matrix [p] such that [b] = [p]T[a] SIMILARITY TRANSFORMATIONS If [a] is any arbitrary square matrix of order n and [p] some known square matrix of order n and [p]~l its inverse, then the transformation defined by the equation [b] = [p]-\a][p] (3) is said to be the similarity transformation of [a] by [p]. Any matrix [b] that can be obtained from a given square matrix [a] by a similarity transformation is said to be similar to [a]; that is, [b] is similar to [a] if and only if there exists a nonsingular square matrix [p] such that [b] = [p]-l[a][p]. Theorem 1. If [b] is similar to [a] then [a] is similar to [b]. Theorem 2. If [a] and [b] are similar, and [b] and [c] are similar, then [a] and [c] are similar. Theorem 3. The similarity transformation of the sum [a] + [b] by [/?] is equal to the sum of the similarity transformations of [a] by [p] and [b] by [/?], that is, [p]-la°]+[i>])[p]-[prl[°][p]+[prl[i>][p] (4)
930 Matrix Transformations Theorem 4. The similarity transformation of the product [a][b] by [p] is equal to the product of the similarity transformations of [a] by [p] and [b] by [p], that is, [/»]"{[«][*]}[/'] = {[P]~l["][P]}{[Prl[b][p]} (5) note: From the preceding theorems it follows that a matrix equation involving sums and products of square matrices of the same order remains valid if each of the matrices is subjected to a similarity transformation by the same matrix [p]. Theorem 5. The determinant of a matrix is invariant under a similarity transformation. Theorem 6. The trace of a matrix is invariant under a similarity transformation. ORTHOGONAL TRANSFORMATIONS If [a] is any arbitrary square matrix of order n and [p] some known orthogonal matrix of order n, the transformation defined by the equation [b]=[P]T[°][P] (6) or equivalently by the equation [*]-[/']"'[«][/'] (7) is said to be the orthogonal transformation of [a] by [/?]. note: An orthogonal transformation is both a congruent transformation and a similarity transformation. Any matrix [b] that can be obtained from a square matrix [a] by an orthogonal transformation is said to be orthogonal to [a]; that is, [b] is orthogonal to [a] if and only if there exists an orthogonal matrix [p] such that [b] = [Prl[a][p] = [p]T[a][p].
APPENDIX Eigenvalues of Matrices Consider a square matrix [a]. If there exist a column matrix [x) and a scalar X such that [a][x) = X[x) (1) then the scalar is called an eigenvalue of the matrix and the column matrix is called an eigencolumn of the matrix. There are many different terms in use for the quantities we have named eigenvalues and eigencolumns. Instead of the prefix eigen- one encounters the words characteristic, proper, latent, and modal. Instead of the word value one sees the words root and number. Instead of the word column one finds the word vector. Theorem 1. If [x) is an eigencolumn of the matrix [a] associated with the eigenvalue X, then c[x], where c is an arbitrary scalar, is also an eigencolumn of the matrix [a] associated with the eigenvalue X. Theorem 2. The eigenvalues of a square matrix [a] of order n are the n roots XUX2, . . . , Xn of the equation K-Hi=° (2) This equation is called the characteristic equation of [a]. If Xk is a particular eigenvalue of the matrix [a], the eigencolumns associated with Xk can be obtained by making the substitution X = Xk in the equation [aiJ-X8u][xJ) = [o) (3) and solving for [xj) = [x). proof: Using the unit matrix [e], we can rewrite Eq. (1) {[a]-\[e])[x) = [o) (4) 931
932 Eigenvalues of Matrices The matrix equation (4) is equivalent to a set of n simultaneous linear equations in the variables xx,x2, • • • , xn and will have a nontrivial solution if and only if the determinant of the matrix [a] — X[e] vanishes, that is, if and only if \atj - XSy] = 0. The proof of the rest of the theorem follows immediately. Theorem 3. If the eigenvalues X,,X2, . . . , X„ of a matrix are distinct, the associated eigencolumns [x)!,[x)2, ..., [x)n are linearly independent Theorem 4. If the eigenvalues X,,X2, . . . , X„ of matrix [a] are not distinct but the matrix [a] is real and symmetric, there exists an associated set of eigencolumns [*)i>[*)2> ..., [x)n that are linearly independent. Theorem 5. If the eigenvalues of a matrix [a] are X1?X2, . . . , Xr, the eigenvalues of the matrix [a]n are X/,X2, . . . , X/1, and the corresponding eigencolumns are the same as the eigencolumns of [a]. Theorem 6. Similar matrices have the same eigenvalues. proof: Consider the equation [a][x) = \[x) (5) If we premultiply by [p]~l we obtain W'Mt^br'M*) (6) or equivalently {[/']",[«][/']}{[/']",[*)}-M[/']",[*)} (7) It follows that if X is an eigenvalue for [a], it is also an eigenvalue for [/»r "MM. Theorem 7. The eigenvalues of a diagonal matrix are the diagonal elements of the matrix. Theorem 8. The eigenvalues of a real symmetric matrix are real. Theorem 9. The eigenvalues of a real symmetric positive definite matrix are positive. Theorem 10. The eigenvalues of a real symmetric positive semidefinite matrix are greater than or equal to zero. The number of nonzero roots is equal to the rank of the matrix.
Eigenvalues of Matrices 933 Theorem 11. The eigenvalues A1? A2, . . . , Xr of a real symmetric matrix [a] of order r satisfy the equations Tr[«]"-2(M" (8) where Tr[af is the trace of the «th power of the matrix [a]. It follows that if [a] is a real symmetric matrix of order r, its eigenvalues can be determined from the r equations Tr[fl] = XX + A2 + • • • + Ar Tr[a]2 = A? + A22+...+\2 Tr[a]r = A[ + A2r + • • • +A; proof: If [a] is real and symmetric, there exists a complete set of eigen- columns that are real and independent (Theorem 4); hence there also exists a similarity transformation that reduces [a] to diagonal form (see Theorem 1 in Appendix 26). That is, there exists a square matrix [p] such that [/>]"'[«][/>]=['] (10) where [c]=[c,]=[A,.S,] (11) The diagonal elements A, of [c] are the eigenvalues of [c] (Theorem 7), and also the eigenvalues of [a] (Theorem 6). Since [a] and [c] are similar matrices, their traces are equal (Theorem 6, Appendix 24). We thus have Tr[a]-2\ (12) i where the A, are the eigenvalues of [a]. From Theorem 4 in Appendix 24 we can show that [a]n and [c]n are similar matrices and thus Tr[a]"=Tr[c]"=2(\)" (13) I This completes the proof of the theorem. The theorem above is particularly useful when we have already obtained by some other method partial knowledge of the eigenvalues. Suppose for example that [a] is a real symmetric matrix of order 6 for which the eigenvalues are A, B, C, C, D, D, and we have already determined A and B and know that the four remaining eigenvalues consist of two pairs. We can then determine the
934 Eigenvalues of Matrices values of C and D from two equations Tr[fl] = A + B + 2C + 2D (14) Tr[a]2 = ^ + B + 2C2 + 2D1 (15) In general it is tedious to calculate the trace of [a]n for n > 2 but simple when n = 1 or 2. The trace of [a] is the sum of the diagonal terms, and the trace of [a]2 can be shown to be the sum of the squares of the elements, provided [a] is symmetric, as it is here.
APPENDIX Diagonalization of Matrices INTRODUCTION In many problems in physics one seeks matrix transformations that transform a given matrix into a diagonal matrix. We investigate some important cases. DIAGONALIZATION OF A MATRIX BY A SIMILARITY TRANSFORMATION Theorem 1. Consider a square matrix [a] of order n. Let A1,A2, . . . , Xn be the set of n eigenvalues of [a]. If there exists a set of n linearly independent eigencolumns [*),, [x)2, ..., [x)n, as would be the case if the eigenvalues were distinct or the matrix [a] were real and symmetric, then the similarity transformation of [a] by the square matrix [p] whose columns are the matrices [*)i>[*)2> • • • > ix)n reduces [a] to diagonal form, and the diagonal elements are the eigenvalues A1?A2, . . . , A„. Thus if [/»]S[[*),[*V ••[*).] (i) then [PVS["][P]=[C] (2) where [c]=[c,]=[A,S,] (3) proof: From the definition of the eigenvalues and eigencolumns of [a] it follows that [a][x)=Xi[x)i (4) The set of equations (4) is equivalent to the single matrix equation [*][[*)i[*)2 • • • [*)„] -[t*),[*)2 • • • [*)j[vy (5) 935
936 Diagonalization of Matrices or [a][p]=[p][c] (6) If the eigencolumns are linearly independent, the matrix [p] is nonsingular and thus has an inverse. If we premultiply Eq. (6) by [p]~l we obtain Eq. (2). DIAGONALIZATION OF A MATRIX BY A CONGRUENT TRANSFORMATION The diagonalization of a real symmetric matrix by a congruent transformation is closely related to the problem of reducing a quadratic form to a sum of squares. We therefore refer the reader to Appendix 27 for material on this subject.
APPENDIX Quadratic Forms INTRODUCTION A function <j>, of the set of variables x = xx,x2, . . . , xn, which is of the form ¢00=2 2¾¾¾ (1) 1 j is called a quadratic form. A quadratic form can always be arranged in such a way that a0 = aJi (2) and we assume that this is always the case. In matrix form Eq. (1) can be written *(x) =(*][*][*) (3) The symmetric matrix [at] = [a] is called the matrix of the quadratic form. A quadratic form for which the atJ are real is called a real quadratic form. We assume throughout this appendix that we are dealing with real quadratic forms. The order of a quadratic form is the order of the matrix of the quadratic form. The rank of a quadratic form is the rank of the matrix of the quadratic form. A quadratic form <j>(x) is said to be positive definite if and only if 4> > 0 for every nonzero real value of x, nonnegative if and only if <f> > 0 for every nonzero real value of x, and positive semidefinite if and only if <j> > 0 for every nonzero real value of x and <f> = 0 for some nonzero real value of x. Similarly a quadratic form <j>(x) is said to be negative definite if and only if <j> < 0 for every nonzero real value x, nonpositive if and only if <J> < 0 for every nonzero real value of x, and negative semidefinite if and only if <f> < 0 for every nonzero real value of x and <J> = 0 for some nonzero real value of x. 937
938 Quadratic Forms TRANSFORMATION OF A QUADRATIC FORM Theorem 1. If the set of variables x in the quadratic form *WESSw (4) ' j is subjected to a regular linear transformation xi = 2 aifJk (5) k then the function <j>(y) = <J>[x(y)] is a quadratic form, that is, k I and the coefficients bkl are given by ^/ = 22¾0^ (7) ' y proof: Carrying out the transformation we have * = 2 2^/^== 222 2¾¾^°^ ' y A: / = 2 2 (2 2 «*«y/*0 W/= 2 2 huy*yi (8) k i \ i j i k i It is instructive to express Theorem 1 in matrix notation. If we do we obtain: Theorem 1. If the set of variables x in the quadratic form <f>(x)=(x][a][x) (9) is subjected to a regular linear transformation [*)=[«] b) (10) then the function <J>(y) = <J>[y(x)] is a quadratic form, that is, *(y)=O0[*]l>) (11) and the matrix [b] is given by [b]-[a]T[a][a] (12) proof: Taking the transpose of Eq. (10) we obtain (x]=(y][a]T (13)
Reduction of a Quadratic Form to a Sum of Squares by a Linear Transformation 939 Substituting Eqs. (10) and (13) in Eq. (9) we obtain 4>=(x][a][x)=(y][a]r[a][a][y)=(y][b][y) (14) note: It follows from the theorem above that a transformation of a quadratic form ^(x) = (*][#][ x) can be viewed either as a linear transformation of the set of variables x or as a congruent transformation of the matrix [a]. Theorem 2. If a quadratic form is subjected to a regular linear transformation, the resulting quadratic form has the same rank as the original quadratic form. REDUCTION OF A QUADRATIC FORM TO A SUM OF SQUARES BY A LINEAR TRANSFORMATION Theorem 3. Given a quadratic form n n <Kx) = 2 2<W/ (15) /=1 _/-! it is always possible to find a linear transformation n xt= 2 *tpj (16) that reduces § to diagonal form, that is, such that *[*(y)] 4S V#= S b„yf= 2 v,2 (17) 1 = 1./=1 / = 1 /-1 proof: Consider a quadratic form of order n: <t>=t 2 *//*/*/ (18) /=i/=i We consider separately the case in which at least one of the coefficients a\\>a22> • • • >ann ls nonzero, and the case in which all the coefficients au, a22, . . . , ann are zero. Let us take the nonzero case first and assume for simplicity that au =^ 0. If au =^ 0 we can rewrite Eq. (15) as follows: *-auxi + F*l + G-au(Xl + J^)2-(JL.)2+G (19) where F = 2anx2 + 2aux3 + • • • + 2alnx„ (20) n n G = 2 2 aijxixj (21) /=2y=2
then Eq. (15) where assumes 7i = *i" yi = xt the form <t> t = r2^ / = 2,3,.. = bxy\ + * n n 2 2 V* 940 Quadratic Forms If we make the transformation (22) (23) (24) / = 2y=2 We have thus shown that a quadratic form ^i2jjaijxixj °f order « for which at least one of the terms au,a22, . . . , ann is nonzero can be reduced by a linear transformation to the sum of a square term and a quadratic form of order /i-l. Let us now consider the case in which all terms an,a22, . . . , ann in the quadratic form ^i^jaijxixj vanish. In this case one of the cross terms containing x{ does not vanish. For simplicity we assume al2 i^ 0. If au = a22 = 0 and an^0, Eq. (1) can be written <j> = aX2xxx2 + Fxx + Gx2 + H -%(><+*>+*!?)'-•-?(*<-«- ^r-s- (*> where F = 2aux3 + 2a 14;t4 + • ■ G = 2a23x3 + 2a24x4 + • • n n /=3;=3 ' * + 2«lwx„ " * + 2tf2„x„ (26) (27) (28) If we make the transformation F+ G yl = xl + x2 + y2 = xx — x2 «12 F- G hi ax y, = X; i = 3,4, ..., n (29) then <j> takes the form *-*^? + M + ^ (30) where /-3./=3
Reduction of a Quadratic Form to a Sum of Squares by an Orthogonal Transformation 941 We have thus shown that a quadratic form S/Sy^*/-*/ °f order n for which all the terms au,a22, • • • , ann are zero can be reduced to the sum of two square terms and a quadratic form of order n — 2. Combining the results above we conclude that we are able to reduce any quadratic form to a sum of square terms and a quadratic form of lower order. The remaining quadratic form of lower order can in turn be reduced to a sum of square terms and a quadratic form of still lower order. Proceeding in this fashion we are able ultimately to reduce any quadratic form to a sum of squares. Theorem 4. If a real quadratic form n n /-1,-1 is reduced by a linear transformation n xi = 2 «(#■ to the form <Ky)=2V,2 (34) / = 1 then the number of nonzero coefficients bi is always equal to the rank of the coefficient matrix [a], and the number of positive coefficients bt and the number of negative coefficients bt always remain the same regardless of the transformation or the final diagonal form. REDUCTION OF A QUADRATIC FORM TO A SUM OF SQUARES BY AN ORTHOGONAL TRANSFORMATION Theorem 5. Given a quadratic form n n <Hx) = 2 2 aijxtxj (35) it is always possible to find an orthogonal transformation n Xt = 2 "tfj (36) that reduces <|> to a sum of squares, that is, such that *[*(y)] = 2 SVrS^,2 (37) /-17=1 /-1 proof 1. If we restrict ourselves to points lying on the n dimensional hypersphere S/*/2 = *> tnen tnere *s a point/*, at which <J>(x) has a maximum value. If we restrict ourselves to points lying on the intersection of the (32) (33)
942 Quadratic Forms hypersphere 2 tx? — 1 and tne hyperplane that passes through the origin O and is perpendicular to the line OPx, then there is a point P2 at which <|>(x) has a maximum value. If we restrict ourselves to points lying on the intersection of the hypersphere 2/¾2 = U tne hyperplane that passes through the origin O and is perpendicular to the line OPx, and the hyperplane that passes through the origin and is perpendicular to the line OP2, then there is a point P3 at which <j>(x) has a maximum value. Proceeding in this fashion we can locate n points PX,P2, . . . , Pn on the hypersphere 2/*/2 = *• Fr°m the definition of the points PUP2, . . . , Pn it follows that the line segments OPx,OP2, • • • , OPn are mutually orthogonal. We now show that if we choose a set of coordinates yx, y2, . . . ,yn such that the j, axis is in the direction OPx, thej2 axis in the direction OP2, ..., and the yn axis in the direction OPn, then in terms of the coordinates yx, y2, . . . ,yn the quantity <$> has the form <j> = 2/¾^ We note first that in terms of the set of coordinates j„ ¢ = 2/¾¾)^ anc^ ^ equation of the hypersphere is 2^/2 = 1- The point P, is the point that maximizes <|> subject to the constraint 2¾½2 = '> or equivalently the point for which the function *"(**)'= 2 2 V*- A( 2 yi -*) (38) /=1 y = 1 \/=l / has a maximum value (see Appendix 15). It follows that the coordinates yx, y2, . . . ,yn of the point Px must satisfy the equations 3F(y,A) JL -^1=^2biJyJ-2Xyi = 0 (39) But the coordinates of the point Px are /1 = 1 /2=/3 =y„ = o (40) Substituting Eqs. (40) into Eq. (39) we obtain bu=\ *I2 = *I3 *,„ = 0 (41) The point P2 is the point that maximizes <j> subject to the constraints 2j/2 = ^ and j, = 0, or equivalently the point for which the function G(y.M) = £ tb^j-Jtyf-l) (42) / = 2 j = 2 \ i = 2 / has a maximum value. It follows that the coordinates yx,y2, - - - ,yn °^ ^Q point P2 must satisfy the equations ^^--tUyyj-2^,-0 i = 2,..., n (43) But the coordinates of the point P2 are 7, = o j2=i y3 = y4= -' =yn = o (44) Substituting Eq. (44) in Eq. (43) we obtain b22 = n b23 = b24 = • • • = b2n = 0 (45)
Simultaneous Reduction of Two Quadratic Forms to Sums of Squares 943 Proceeding in this fashion we can show that all the values of by for which i ^j vanish. This proves what we set out to prove. Proof 2. Since the matrix [a] is real and symmetric, its eigenvalues X,, X2, . . . , X„ are real, and it is possible to choose an associated set of eigen- columns [x)u[x)2, ..., [x)n that are orthonormal; that is, they satisfy the relation (*],[*);- Sv (46) From the definition of eigenvalues and eigencolumns we have [a][x)=Xi[x)i (47) If we define [p] as the square matrix whose columns are [^)i,[^c)2, • . . , [x)n, that is, [/>]=[[*).[*V ••[*)„] (48) then the set of equations (47) can be written [a][p]=[p][\i8u] (49) Multiplying by [p]T and noting from the orthogonality of [p] that [p]T = [p]~] we obtain [p]T[a][P]=[\Aj] (50) Thus we have found a congruent transformation that diagonalizes the matrix [a]. If the congruent transformation of the matrix [a] by [p] diagonalizes the matrix [a], then the linear transformation [x) = [p][y) reduces the quadratic form (.x][tf][.x) to a sum of squares. But [p] is orthogonal; hence [x) = [p][y) is an orthogonal transformation. We have thus found an orthogonal transformation that reduces the quadratic form to a sum of squares. SIMULTANEOUS REDUCTION OF TWO QUADRATIC FORMS TO SUMS OF SQUARES Theorem 6. Given a positive definite quadratic form ¢ = 22¾¾¾ (51) * j and an arbitrary quadratic form ^ = 2 2V/*/ (52) ' j it is always possible to find a linear transformation */ = 2<W (53) J that simultaneously reduces <J> to a sum of squares and \p to a sum of squares.
944 Quadratic Forms proof: From Theorem 3 it is possible to find a linear transformation xi = 2/¾¾ tnat reduces <|> to a sum of squares. In the new variables the two quadratic forms become ¢ = 2^,2 (54) ^ = 22^ (55) ' j Since <|> is positive definite, it follows that the c, are positive, and we can introduce new real variables zi = (c,)1^, in terms of which the two quadratic forms become ¢ = 2¾2 (56) *-2 2M*y (57) ' J From Theorem 5 it is possible to find an orthogonal transformation zt = ^jfiijWj that reduces \p to a sum of squares. Since the transformation is orthogonal, the form <|> remains unchanged, so that in terms of the new variables the two quadratic forms become 4> = 2w,2 (58) * = 2 &»f (59)
APPENDIX Group Theory INTRODUCTION Often the analysis of systems that possess some kind of symmetry can be appreciably simplified by the use of group theory. In this appendix we consider some of the basic principles of group theory. Group. If we are given a set G of elements a,b, . . . , and an operation or rule that assigns to every ordered pair a,b, in G a unique element, which we designate element ab, then if the following conditions are satisfied, the set is said to form a group with respect to the given operation: Closure law. If a and b are elements of G then ab is an element of G. Associative law. If a, b, and c are elements of G then (ab)c = a(bc). Existence of an identity. There exists an element e of G called the identity element, which has the property that if a is an arbitrary element of G, then ea = ae = a. Existence of an inverse for each element. For each element a of G there exists an element a~x of G, called the inverse of a, such that a~xa = aa~l = e. note: It is customary to call the element ab the product of a and b, and to refer to the operation that assigns the element ab to the pair a, b, as multiplication. However one should not infer from this that the operation is necessarily multiplication in the ordinary sense. Order of a Group. The number of elements in a group is called the order of the group. The order of a group G is denoted by \G\. Examples. The two numbers 1 and - 1 form a group of order two with respect to the operation of multiplication. The rotations through the angles 0, 27r/3, and 4tt/3 about an axis form a group of order three with respect to the operation of successive rotations. The set of all integers forms a discrete group 945 28
946 Group Theory of infinite order with respect to the operation of addition. The set of positive real numbers forms a continuous group of infinite order with respect to the operation of multiplication. ELEMENTARY PROPERTIES OF A GROUP Theorem 1. There is one and only one identity element in a group. Theorem 2. There is one and only one inverse for a given element in a group. Theorem 3. If a set of elements G and an operation satisfy the closure law and the associative law and either there exists an element e such that ae = a for all a in G and for each a an element a~x such that aa~x = e or there exists an element e such that ea — a for a in G and for each a an element a ~x such that a~xa = e, then the set forms a group with respect to the given operation. Theorem 4. If a is any member of a group then (a-1)-1 = a. Theorem 5. Let a, b, and c be elements of a group. If ab = ac then b = c. Similarly if ba = ca then b = c. Theorem 6. If a and b are elements of a group G, there exists a unique element c in G satisfying the equation ca = b, and a unique element d in G satisfying the equation ad= b. Theorem 7. The product abc • • • of any number of elements in a group depends on the sequence of the elements in the product but not on the sequence in which the products are formed. Example. (ab)(cd) = a[b(cd)] = a[(bc)d] = [a(bc)]d = [(ab)c]d. Theorem 8. If a,b,c, . . . is any subset of elements in a group G then (abc • • - )~l = - - - c~]b~]a~l; that is, the inverse of a product is the product of the inverses taken in the reversed order. POWERS OF AN ELEMENT Power of an Element. If a is an element of a group and n an integer, we define the «th power of a, which we designate an, by the following relations: a°=e ax = a an = aan~x n>\ a~n=(a~x)n n>\
Isomorphisms, Automorphisms, and Homomorphisms 947 Theorem 9. If a is any element of a group G, and m and n are any two integers, then aman = am + n. Theorem 10. If a is any element of a group G, and m and n are any two integers, then (am)n = a™. Order of an Element. The least positive integer n such that an = e is called the order of the element a and is denoted by \a\. If such an integer does not exist the order of the element is said to be infinite. Theorem 11. The ratio of the order of a group to the order of an element in the group is a positive integer; that is, if a is an element of a group G, then |G|/|a| is a positive integer. MULTIPLICATION TABLE Multiplication Table. The properties of a group are completely determined if the product of every pair of elements is known. This information can be conveniently displayed in a multiplication table. The multiplication table for a group of order n is an n X n array with one row associated with each element in the group and one column associated with each element in the group and in which the element in row a and column b is the product ab of element a and element b. Such a table is indicated schematically in Fig. la. The multiplication table for a particular group of order four is shown in Fig. lb. 0 aa ba ca da b ab bb cb db c ac be cc dc d ad bd cd dd Fig. la b c e a be a b c |e b c e a c e a b Fig. lb Theorem 12. Each element of a group appears once and only once in each row and each column of the multiplication table. Theorem 13. Every row is different from every other row and every column is different from every other column in a group multiplication table. ISOMORPHISMS, AUTOMORPHISMS, AND HOMOMORPHISMS Isomorphic Groups. A group G' is said to be isomorphic to a group G if the following conditions are satisfied:
948 Group Theory a. There is a one to one correspondence between the elements a,b,c, . . . of G and the elements a',b',c\ . . . of G'. b. If a and b are arbitrary elements in G, the element in G' corresponding to the element ab in G is the product of the element in G' corresponding to a in G and the element in G' corresponding to b in G, that is, (ab)' = a'V. Theorem 14. If a group G' is isomorphic to a group G, the group G is isomorphic to the group G'. Automorphic Groups. A group G' that contains the same elements as a group G and is also isomorphic with respect to the same operation is said to be automorphic to G. Homomorphic Groups, A group G' is said to be homomorphic to a group G if the following conditions are satisfied: a. With every element a of G there is associated a unique element a' of G'. b. If a and b are arbitrary elements in G then (<zZ>)' = a'Z>'. note: If G' is isomorphic to G, the element a with which the element a' is associated is unique; but if G' is homomorphic to G, the element a with which the element a' is associated is not necessarily unique. Thus if G' is isomorphic to G the order of G' is equal to the order of G, but if G' is homomorphic to G the order of G' is less than or equal to the order of G. TYPES OF GROUPS Abelian Group. A group is called an Abelian or a commutative group if for every two elements a and b, the relation ab = ba holds. note: The multiplication table of an Abelian group is symmetric across the principal diagonal. Cyclic Group. A group G is said to be a cyclic group if every element in the group is a positive power of one given element g in the group, that is, G = {g, g2, g3, . . . }. The element g is said to generate the group. Theorem 15. The order of a cyclic group is equal to the order of the element that generates it. Periodic Group. A periodic group is a group in which every element has a finite order. Theorem 16. Every finite group is periodic.
SUBGROUPS Subgroups 949 Complex. A complex of a group G is any subset of elements of the group. Subgroups. Let G be a group and H a complex of G. If H forms a group with respect to the same operation used in defining the group G, then H is called a subgroup of G. Theorem 17. If H is a subgroup of the group G, the identity element of H is the same as that of G, and the inverse of each element of H is the same in G as in H. Coset. If H is a subgroup of the group G, and g is an element that is in the group G but not in the subgroup H, then the complex that is generated by taking the product of g with each element h of H is defined as the left coset of H with respect to G and is denoted gH. Thus gH = {ghu gh2, gh3, . . . }. Similarly the complex that is formed by taking the product of each element h of H with g is defined as the right coset of H with respect to g and is denoted Hg. Thus Hg={hlg,h2g,h3g, . . . }. Theorem 18. The cosets of a subgroup H have the following properties: a. Each of the cosets contains exactly |H | elements. b. No element in H is in any of the cosets. c. Any two of the left (right) cosets either contain the same elements or have no elements in common. d. Every element of the group is contained either in H or in one of the cosets of H. Theorem 19. The ratio of the order of a group to the order of a subgroup is a positive integer; that is, if H is a subgroup of G, then |G|/|//| is a positive integer. proof: The group G can be factored into the subgroup H and the collection of distinct left (right) cosets of H. If H has \H\ elements, each of the cosets contains exactly \H\ elements. It follows that the order of the group is an integer multiple of the order of the subgroup. Theorem 20. If the order of a group is a prime number, then: a. The only subgroups of the group are the group itself and the identity element. b. The group is cyclic. Invariant Subgroup. A subgroup if of a group G is said to be an invariant, normal, or self-conjugate subgroup, or a normal divisor if for every element g
950 Group Theory of G the left coset of H with respect to g is equal to the right coset of H with respect to g. Theorem 21. The cyclic group generated by an element g in a group G forms a subgroup of G. CLASSES Conjugate Elements. An element a in a group G is said to be conjugate with an element b in G if there exists an element c in G such that a = c~lbc. Theorem 22. Conjugate elements have the following properties: a. If a is conjugate with b, then b is conjugate with a. b. Every element is conjugate with itself. c. If a is conjugate with b and b is conjugate with c, then a is conjugate with c. Classes. The set of elements in a group that are conjugate with an element a in G is called the class of a. note: The class of an element a is generated by evaluating the product g~lag for all g in G. Theorem 23. If a belongs to the class of b, then b belongs to the class of a. Theorem 24. If b is not an element of the class of a, the elements forming the class of a are entirely different from those forming the class of b. Theorem 25. Every element belongs to its own class. Theorem 26. The inverses of the elements of a class form a class. Theorem 27. Every element in an Abelian group forms a class by itself. Theorem 28. The identity element forms a class by itself. Theorem 29. The ratio of the order of a group to the number of elements in a class is a positive integer.
APPENDIX Vector Spaces VECTOR SPACE Complex Vector Space. A complex vector space V is a collection of objects called vectors, which we designate [a), [b), [c), ..., satisfying the following conditions: a. For every pair of vectors [a) and [b) in V there exists a unique vector [c) = [a) 4- [b) called the sum of [a) and [b) that is also in V and has the properties [«) + [*)-[*)+[«) (i) {[a) + [b)}+[c) = [a)+{[b) + [c)} (2) b. For every vector [a) in V and every complex number a there exists a unique vector [6) = a[a) called the product of a and [a) that is in V. The product has the properties l[a) = [a) (3) «{/?[«)} = {«0}[a) (4) c. The sum and product defined above obey the distributive laws {a + /3)[a) = a[a) + /3[a) (5) a{[a) + [b)} = a[a) + a[b) (6) d. There is a vector [0) in F satisfying the condition [a) + [0) = [0)+[«) = [«) (7) for every [a) in F. e. For every vector [a) in V there exists a vector -[a) satisfying the condition [*)+{-[*)} =0 (8) 951 29
952 Vector Spaces Real Vector Space. If the numbers a, /?, . . . appearing in the definition above are restricted to real numbers, the space defined is called a real vector space. Subspace of a Vector Space. A subset of the vectors in a vector space is said to form a subspace if it forms a vector space under the same operations of addition and multiplication. Proper Subspace. Any subspace of a vector space V other than the space itself and the null space is called a proper subspace. Linearly Independent Vectors. A finite set of vectors [a), [b), ... in a vector space V is said to be linearly dependent if there exists a set of constants a, /?, . . . , not all of which are zero, such that a[a) + /3[b)+ ... =[0) (9) If no such set of constants exists the vectors are said to be linearly independent. Dimension of a Vector Space. A vector space V is said to be of dimension n if it contains n linearly independent vectors but every n 4- 1 vectors are linearly dependent. Vector Space Spanned by a Set of Vectors. If [a), [b), [c), . . . constitute a set of vectors, the space that is generated by taking all possible linear combinations of these vectors is said to be the space spanned by these vectors. Conversely a set of vectors is said to span a vector space V if every vector in V can be expressed as a linear combination of the given vectors. Basis Vectors. A set of vectors is said to form a basis for a vector space V if the set spans the space V and the vectors are linearly independent. In an n dimensional vector space V any n linearly independent vectors form a basis for the space. We use the notation [e)l9 [e)2, ..., [e)n to designate a particular set of basis vectors for a space of dimension n and we refer to the set of basis vectors as the basis S. To distinguish between two different bases we use the notation [e)x, [e)2, ..., [e)n for one basis set and [e)v, [e)T, ..., [e)n> for the other basis set, and speak of the basis S and the basis S", respectively. Components of a Vector. Let S = [e)l9 [e)2, ..., [e)„ be a complete set of basis vectors in a vector space V of dimension n. An arbitrary vector [a) can be expressed as a linear combination of these basis vectors; that is, we can write [a) = ax[e)x+a2[e)2+ ... + an[e)n (10) The set of constants ax,a2, • • • , an are said to be the components or coordinates of [a) with respect to the basis S.
Matrix Representation of Vectors and Operators 953 SUM OF TWO VECTOR SPACES If F is a vector space and Vx and F2 are two subspaces of F, the sum of Vx and F2, which we designate Vx 4- F2, is the vector space spanned by the set of vectors contained in Vx and F2. If F, and F2 are disjoint, the sum is said to be a direct sum and is designated Vx © F2. LINEAR OPERATORS Linear Operator on a Vector Space. A linear operator [p] on a vector space F is a rule that assigns to each vector [a) in F a unique vector [b) = [p][a) also in F, and also has the property that for every pair of vectors [a) and [b) in F and every pair of complex numbers a and /3 [p]{a[a) + fi[b)}-a{[p][a)}+fl{[p][b)} (11) The Product of Two Linear Operators. If [p] and [q] are two linear operators on F, the linear operator [r] = [p][q] defined by the condition ['][*)Hp]{[i][°)} (12) for arbitrary [a) is called the product of [p] and [q]. An Invariant Subspace. A subspace V of a vector space F is said to be invariant relative to an operator [p] if this operator acting on the vectors in the subspace V transforms them into vectors belonging to the same sub- space V. MATRIX REPRESENTATION OF VECTORS AND OPERATORS Matrix Representation of a Vector. Let [a) be an arbitrary vector in a vector space F. Let S = [e)i,[e)2> . . . ,[e)n be a complete set of basis vectors in F. The column matrix [at) whose ith element is the /th component of [a) with respect to the basis S is called the matrix representation of the vector [a) with respect to the basis S. If we know [a) and the basis S we can find [a,) from the foregoing definition. Conversely if we know [#,-) and the basis S we can find [a) from the relation [a) = 2/fli[*)/- Theorem 1. Let [a), [b% [c% . . . be arbitrary vectors in a vector space F and K)> [bi% [cd, . • • their matrix representations with respect to a basis S. If lc) = [a) + [b) then [c,) = [*,) + [£>,), and if [b)=a[a) then [£>,.) = a[*,.). It follows that the set of matrices [at), [6,.), [q), ..., together with the operations of matrix addition and multiplication by a number, constitute a vector space that is isomorphic to the space F.
954 Vector Spaces Matrix Representation of a Linear Operator. Let [p] be a linear operator on a vector space V of dimension n. Let S = [e)!,[e]2, . . . , [e)n be a complete set of basis vectors in V. The square matrix [/^.] whose //th element is equal to the /th component of the vector [/?][e), is called the matrix representation of the operator [p] with respect to the basis S. If we know [p] and the basis S, there is a unique [ptJ] determined from the relation above. Conversely if we know [py] and the basis S, there is a unique [p] determined from the relation [P][e)j = 2>y[e),- Theorem 2. Let [a), [b), [c), . . . be arbitrary vectors in a vector space V and K)> [*/)> to) • • • their matrix representations with respect to a basis S. Let [p\ [?L I/L • • • be linear operators on V and [#y], [fy], I//,], . . . their matrix representations with respect to a basis S. If [p][a) = [b) then [/^¾¾) = [i;), and if [/¾] = [r] then [/>..]fo..] = [rtj\. From the results above it follows that it is possible to set up a one to one correspondence between a set of vectors in a vector space V of dimension n and a set of column matrices of order n, and also to set up a one to one correspondence between a set of operators on V and a set of square matrices of order n. Furthermore these correspondences can be made in such a way that the laws of combination of vectors and operators are replaced by the corresponding laws of combinations of matrices.
APPENDIX Representations of a Group REPRESENTATIONS OF A GROUP Operator Representation of a Group. Let a,b,c, . . . be the elements of a group G. Let [a], [b], [c], ... be a set of linear operators on a vector space V. If the set of operators is homomorphic with respect to the operation of multiplication to the group G, the set of operators is said to be an operator representation of the group G and the space V is said to be the carrier space for the representation. An operator representation of a group G is denoted by [G]. Matrix Representation of a Group. Let a,b,c, . . . be the elements of a group G. Let [ay], [bt\ [ctj\ ... be a set of square matrices. If the set of matrices is homomorphic with respect to the operation of matrix multiplication to the group G, the set of matrices is said to be a matrix representation of the group. A matrix representation of a group G is denoted by [Gtj]. There are representations of a group other than operator or matrix representations. Our interest in this appendix however is restricted to representations of these two types. The unmodified word "representation" is to be understood to mean either a matrix representation or an operator representation. Dimension of a Representation. The dimension of the carrier space for an operator representation [G] is said to be the dimension of [G]. The order of the matrices in a matrix representation [G^] is said to be the dimension of [GJ. Faithful Representation. A representation that is isomorphic to a group G, rather than simply homomorphic, is said to be a faithful representation of the group. 955
956 Representations of a Group EQUIVALENT REPRESENTATIONS If [G] is an operator representation of a group G and S is a basis for the carrier space, we can generate an isomorphic matrix representation of the same dimension by replacing each operator in [G] by its matrix representation with respect to the basis S. Conversely if [G^] is a matrix representation of a group G and S is a basis in a vector space V of the same dimension as [G~], we can generate an isomorphic operator representation on the space V by replacing each matrix in [G^] by its operator representation with respect to the basis S. Equivalent Matrix Representations. Two matrix representations are said to be equivalent if they are of the same dimension and can be generated from the same operator representation. note: From the definition above it follows that two matrix representations that are equivalent but not identical are generated by the same operator representation but with respect to different bases. Theorem 1. Two matrix representations [G^]' and [G^]" of a group G are equivalent if and only if they are of the same dimension and there exists a similarity transformation that simultaneously transforms the elements in the one representation into the corresponding elements in the other representation, that is, there exists a square matrix [p-] such that [g^]" = [piJ]^l[gij]/ [py] for all values of g. Equivalent Operator Representations. Two operator representations are equivalent if they are of the same dimension and can be generated from the same matrix representation. note: From the definition above it follows that two operator representations that are equivalent but not identical are generated by the same matrix representation but with respect to different bases in either the same or different vector spaces. Theorem 2. Let G be a group, [G]' an operator representation of G on a space V, and [G]" an operator representation of G on a space V". The representations are equivalent if and only if they are of the same dimension and there exists a similarity transformation that transforms the elements in the one representation into the corresponding elements in the other representation, that is, if there exists an operator [ V" -> V'\ that converts a vector in V" into a vector in V such that [g]" = [K"-> VT\g\[V" ^ V'\ for all values of[£].
The Generation of Representations 957 note: Two operator representations will be equivalent if and only if they generate the same set of equivalent matrix representations, and two matrix representations will be equivalent if and only if they generate the same set of equivalent operator representations. Thus if two matrix representations are equivalent, operator representations generated from these representations will be equivalent irrespective of the set of basis vectors used; conversely if two operator representations are equivalent, matrix representations generated from these representations will be equivalent irrespective of the set of basis vectors used. Hence there is a one to one correspondence between the classes of equivalent operator representations and the classes of equivalent matrix representations. UNITARY REPRESENTATIONS Theorem 3 (Unitary Representation). If [GtJ] is a matrix representation of a group G, it is always possible to find an equivalent representation in which the matrices [g^] are unitary matrices. Such a representation is called a unitary representation. ONE DIMENSIONAL REPRESENTATIONS Ordinary numbers can be viewed as one dimensional matrices or as operators on a one dimensional vector space; hence it is possible for a set of numbers to form a one dimensional matrix or operator representation of a group. Theorem 4. The number 1 is a one dimensional representation of all groups. Theorem 5. The elements of a one dimensional representation of a finite group must be roots of the number 1. THE GENERATION OF REPRESENTATIONS There is an infinite number of representations of a group. In this section we discuss how some of these representations can be generated, and we introduce a few important representations. Theorem 6 (The Direct Sum of Two Representations). Let [Gtj\ and [Gtj\" be two different matrix representations of a group G. If each matrix [g^]' in [G/y]' is combined with the corresponding matrix [g^]" in [GtJY to form a new
958 Representations of a Group matrix [gy] as follows: [8y]' [°1 [*]- [o] [*]* (i) then the set of matrices [g] constitutes a new representation [G^], which we call the direct sum of the representations [G^]' and [G^]" and denote [G-]' © [G^]". If the dimension of [Gy]' is ra' and the dimension of [G^]" is m", the dimension of [GJ is ra' + ra". Theorem 7. If each of the matrices in a representation [G^] of a group G is replaced by the inverse of its transpose, the resulting set of matrices also constitutes a representation of G. Theorem 8. If each of the matrices in a representation [Gtj] of a group G is replaced by its complex conjugate, the resulting set of matrices also constitutes a representation of G. Theorem 9. If each of the matrices in a representation [Gtj] of a group G is replaced by its determinant, the resulting set of numbers constitutes a one dimensional representation of G. Theorem 10 (Regular or Natural Representation). Let us designate the elements of a group G by the numbers 1, 2, 3, . . . , n. If with each element g in the group we associate a column matrix [ g7) whose elements are all zero except for the element in the gth row, which has the value 1, then we can also associate with each element g a square matrix [ gtj] defined by the condition that [g;j][hj) = [(gA)/), where h is any element in G and [(gh);) is the column matrix associated with the element gh. The set of matrices so defined constitutes a representation of the group and is called the regular or natural representation of the group. proof: The elements of the matrix [&.) are given by &■ = ** (2) where «%-l ^ '• = * = 0 if i*g (3) If we let [ gy] be the square matrix whose elements are given by &J-*HgJ) (4) where 8KgJ) = l if i = gj = 0 if i*gj (5)
Irreducible Representations 959 then [fc]fe)-[2M) L j ; = *Z8i(gj)Sjh) =[si(gh)) L j ' "[(**),) (6) Thus the set of matrices so defined has the desired transformation properties. If we multiply the matrix [gtj] by the matrix [htj\ we obtain [gij][hj"]= 2&A* L j ^iSi(gj)Sj(hk) j -[(«*)*] (7) Hence the set of matrices forms a representation of the group. Furthermore the matrices are real, and are also orthogonal, since [gij\T[Sj*\ = 2j SjiSjl ijk I J S8y<g/)8y(**) ""[8gi(g*)] L j J -[«*] (8) It follows that the matrices are unitary, hence the representation is unitary. IRREDUCIBLE REPRESENTATIONS Irreducible Spaces. If the carrier space V for an operator representation [G] of a finite group G can be decomposed into the direct sum of two subspace V and V"> which are invariant with respect to the operations in [G], then the space V is said to be reducible with respect to the representation [G\ If the carrier space cannot be so decomposed, the space is said to be irreducible with respect to the representation [G]. Irreducible Operator Representation. If the carrier space V for an operator representation [G] is reducible with respect to [G], we say that the representation is reducible; and if the carrier space is irreducible we say the representation is irreducible. Irreducible Matrix Representations. A matrix representation that can be generated from a reducible operator representation on a space V is said to be
960 Representations of a Group reducible if the operator representation is reducible and irreducible if the operator representation is irreducible. Theorem 11. A matrix representation is reducible if and only if it is equivalent to the direct sum of two other nonzero matrix representations. note: The definitions above apply to the representations of finite groups only. A representation of an infinite group satisfying these conditions is said to be "completely reducible" rather than simply "reducible." For finite groups the two terms are essentially equivalent. There are a finite number of nonequivalent irreducible representations of a given finite group. We designate the nonequivalent sets of equivalent irreducible representations of a group G as 1, 2, . . . . An operator representation in the set k is designated [G]k. When we wish to distinguish between two equivalent representations in the same set k we use the notation [G]k and [G]k . A matrix representation in the set k is designated [G^. The operator in the irreducible representation [G]k corresponding to the element g in the group G is designated [g]k. The matrix in the irreducible representation [G ]k corresponding to the element g in the group G is designated [gu]k. The yth element in the matrix [g^f is designated g/f. The dimension of the matrices in the irreducible representation [Gy]k is designated nk. Theorem 12. The number of nonequivalent irreducible representations of a finite group is equal to the number of classes in the group. Theorem 13. The ratio of the order of a group to the dimension of an irreducible representation is a positive integer, that^ is, \G\/nk is a positive integer, where \G\ is the order of the group and nk is the dimension of the irreducible representation [G]k. Theorem 14. The sum of the squares of the dimensions of all the nonequivalent irreducible representations of a group is equal to the order of the group, that is, 2(«*V=|G| (9) k where nk is the dimension of the irreducible representation [G]* and \G\ is the order of the group. Theorem 15. All the irreducible representations of a finite group are contained in its regular representation.
The Characters of Representations 961 Theorem 16. The matrix elements gff in the irreducible matrix representations of a group satisfy the following orthogonality conditions: 2U-%U)l/=-^M,*s* (10) g nr where \G\ is the order of the group and nr is the dimension of the irreducible representation [G]r. If the representations are unitary then (*-%-(«?)• (ii) where (gp* is the complex conjugate of gjr Note that the order of the subscripts is different on the two sides of Eq. (11). note: In the theorem above if f — s then [G]r is assumed to be not only equivalent to [G]s but also identical. If r = s but [G]r and [G]s are not identical, the theorem must be appropriately modified. THE CHARACTERS OF REPRESENTATIONS Character. Let [G] be an operator representation of a group G and [G^] the corresponding matrix representation with respect to the basis S. Let [g] be an operator in [G] and [gfJ] the corresponding matrix with respect to the basis S. The trace of the matrix [gfJ] is called the character of [gfJ]. The trace of a matrix is invariant, under a similarity transformation. It follows that the trace of the matrix corresponding to the operator [ g] does not depend on the choice of basis S. Hence we can speak of the trace of [ gy] as being the character of the operator [ g] as well as the character of the matrix [ g^]. We designate the character of [giJ] or [g] as simply X{[g]} or {[g]}. Theorem 17. Let [G] be a representation of a group G and [g] one of the operators in [G]. The character of the operator [g~l] is equal to the complex conjugate of the character of the operator [g], that is, x{[g-l]}=X*{[g]} (12) proof: There always exists a unitary matrix representation [G-] corresponding to [G], that is, a representation in which (g- %■=(§•,)* (13) In this representation ?(rV?(&r (14)
962 Representations of a Group But the trace of a matrix [ gy] corresponding to the same operator [ g] must be independent of the choice of basis; hence Eq. (14) must be true for all matrix representations corresponding to [G]. Since Eq. (14) is equivalent to Eq. (12), we have proved the theorem. Theorem 18. Let [a] and [b] be two elements in a representation [G] of a group G. If [a] and [b] belong to the same class, then x{[#]} = x([^]}« Theorem 19. Two representations of a group are equivalent if and only if corresponding elements in the two representations have the same characters. THE CHARACTERS OF IRREDUCIBLE REPRESENTATIONS Theorem 20. A representation [G] of a group G is irreducible if and only if I,X*{[g]}x{[g]} = \G\ (15) g where x([g]} is the character of the operator associated with element g, X*([g]} is the complex conjugate of x([g]}> and \G\ is the order of the group. Theorem 21. Let [G]' and [G]7 be two irreducible representations of a group G. The characters of the operators in the representations obey the following orthogonality relations: Sx*{[^];}x{[ ^) = 1^ (16) g Theorem 22. Let [G]1, [G]2, . . . be the irreducible representations of a group G. Let a and b be two elements in G. If a and b belong to different classes then 2x*{[«]'>{[*]'}-° (17) i and if a and b belong to the same class then 2x*{[«]'MM')-71 (18) where p is the number of elements in the class and | G | is the order of the group. Theorem 23. Let G be a group. Let C,, C2, . . . be the classes of G, np the number of elements in the class Cp, CpCq the set consisting of all products of an element from Cp with an element from Cq, and apqr the number of times that each element in the class Cr appears in the set CpCq. Let [G] be a representation of G, n the dimension of [G], and x(Cp the character of an
Decomposition of a Space 963 element in class C . If [G] is an irreducible representation then npX(Cp)nqX(Cq) = n^apqrnrX(Cr) (19) r note: The preceding theorems are sufficient to determine the character table for each of the irreducible representations of a group without actually determining the irreducible representations. DECOMPOSITION OF A SPACE Theorem 24. Let G be a group and [G] an operator representation of G on the space V. The space V can be decomposed into a direct sum of a set of subspaces that are invariant and irreducible with respect to [G]. Each of these subspaces is a carrier space for an irreducible representation; hence with each subspace there is associated a particular one of the irreducible representations [G]k. If we let Vk\ Vk2, ... be the subspaces with which the irreducible representation [G]k is associated, and Vk the direct sum of all of these subspaces, then v= yu® vn® • • • e v2X® v22® • • • = vl®v2® - - - (20) The subspaces Vk are unique, but if a given Vk is the direct sum of more than one irreducible subspace Vkj the subspaces Vkj are not unique. Theorem 25. Let G be a group and [G] an operator representation of G on the space V. Let Vu, V12, . . . , V21, V22, be a complete set of irreducible invariant subspaces of V, where Vkj is theyth subspace associated with the irreducible representation [G]k. If from each subspace Vkj we choose a set of vectors [e)kJ, [e)%, . . . spanning the subspace, the set of all these basis vectors forms a basis S for the space V, and the matrix representation of [G] with respect to the basis S will be a direct sum of irreducible matrix representations, that is, [Gy]-[Ggf<B[Ggf2<B • • • 9[Gyf (B[Gyf(B ■ ■ ■ (21) where the representations [Gu]kI are irreducible and the representations [Gu]k\ [G^]^2, ... are equivalent. Theorem 26. In the decomposition of the space V into irreducible invariant subspaces, the number of subspaces associated with the irreducible representation [G]k is given by ^=T^2x*{[£]}x{Uf} (22)
964 Representations of a Group proof: g g j =s^sx*{[^];}x{[gr} J 8 = ^mJ\G\8^= mk\G\ (23) Theorem 27. Let [G] be an operator representation of a group G and V the carrier space. Let F11, Vn, . . . , F21, F22, . . . represent a particular decomposition of V into irreducible subspaces, where Vkl is the /th subspace associated with the irreducible representation [G]k. Let [G^f be any irreducible matrix representation generated from [Gf. Let [p]kj be an operator defined as follows: [P]h2(g-lti[g] (24) g where (g-1)| is the //th element in the matrix corresponding to the inverse of the element g in the representation [G^f and [g] is the operator corresponding to the element g in the representation [G]. If we operate with [p]kj on each member of a set of vectors spanning the space V, the resulting set of vectors spans a subspace that is contained in the subspace Vk = Vkx © F*2© • -^ and in addition contains one vector from each of the subspaces Vk\ Vk rkl \lm r „\/m proof: Let [e), , [¢)2 , ... be a complete set of basis vectors in the subspace F/m. Any vector [a) in F can be expressed as a linear combination of the basis vectors [e)1™, that is, /s /s [«) = S22^[^m (25) s and are by definition the now operate on [a) with the operator [p]k we obtain l m n where the a„/m are constants and are by definition the components of [a). If we [/^[fl)-2(r%[*]222^[«)r (26) £ l m n But from the definition of the matrix representation of a linear operator with respect to a particular basis we have r The basis vectors can in principle be chosen in such a way that the matrices [gr]l\ [gi]12, ... are equal and are furthermore equal to the matrix [gy]. We
Decomposition of a Space 965 assume that this has been done, and thus [*][«)?-2 «£[«)f (28) r Substituting Eq. (28) in Eq. (26) we obtain [/»]{[«) = 2 2 2 2 2(r liidna?[et (29) g / m n r From Theorem 16 SCr'fot^MA/ (30) Substituting Eq. (30) in Eq. (29) we obtain [/»]*[«) = ^ 2 2 2 2¾4¾¼¾ [«r. YiK i m n r = %A)f (31) n Since the vector [a) is arbitrary, the components akm are arbitrary. The subspace spanned by the set of all possible vectors [pfja) thus contains one vector in each of the subspaces Vk\ Vk2, . . . and is itself contained in the subspace Vk. The result above is completely independent of our choice of irreducible representation [G^f. To emphasize this point, let us let [Gty]k and [Gy]k be two different but equivalent irreducible representations. Since the matrices [gyj]k and [g;j]k are related by a similarity transformation that is the same for all values of g, it follows that there exist two matrices [c^] and [dr] such that Thus if we define then It follows that 4 = 22^4/ (32) r s [Pi -2(r'){[*] (33) g [rf/ = 2(r%[g] (34) g [/^- = 2 22^4/(^1)¾ s] g ' s -2 2<V,<V[/>]£ (35) [/»]£/[*)-2 2^4/bfr) (36)
966 Representations of a Group Substituting Eq. (31) in Eq. (36) we obtain >. km [/4[«) = 22W^2^K r s \ U m -^sfs^jfs^r) (37) YiK m \ r ) \ s ) The vector *Zsdsj{e)km lies in the subspace Vkm. Furthermore since the ,/cm coefficients akm *xt arbitrary, the coefficients ^rcrrakm are also arbitrary. It follows that [pfif projects out a subspace that contains one vector in each of rk\ T/k2 the subspaces V , F , . . . and is itself contained in the subspace F . Corollary. Let [G] be an operator representation of a group G and V the carrier space. Let Vl\ Vn, . . . , V2\ V22, . . . represent a particular decom- position of F into irreducible subspaces, where Vkl is the /th subspace associated with the irreducible representation [G]k. Let x[gf be the character of the element [g]k in [G]k. Let [/?]* be an operator defined as follows: [/»]*-2x*{[ *]*}[*] (38) If we operate with [p]k on each member of a set of vectors spanning the space V, the resulting set of vectors spans the subspace /s /s F*= F*!0 F*20 • • • (39) proof: The operators [pf and [/?]*• are related as follows: [/>]* = 2M (4°) It follows from Eq. (31) that [/»]'[«)-2 [/»]&«) ^2 2 «£[*)? (41) nK i m The subspace spanned by the set of all possible vectors [pf [a) thus contains all the basis vectors in Vk, hence spans the subspace Vk.
APPENDIX 31 Multiply Periodic Functions Definition. A function /(x,y) where x = x},x2, . . . , xr and y ~yl9y2, . . . , ys is said to be multiply periodic in the set of variables x with the period « = «i,«2» ' ' ' > °>r if /(x + <o,y)=/(x,y) (1) for all values of x. Theorem 1. If 63 is a period then nco is a period, where n is an arbitrary integer. Theorem 2. If 63(1) and 63(2) are periods then 63(1) + 63(2) is a period. Theorem 3. If a function has several periods 63(1),63(2), . . . , 63(g), every integral linear combination of these periods is a period, that is, 63= 2 *(*)«(*) (2) where «(/c) is an integer, is a period. Theorem 4. If/(x,y) is a periodic function of x = xl9x2, . . . , xr then there is a system of independent periods 63(1),63(2), . . . , 63(g) such that every period 63 can be expressed in the form g <°= 2 *(*)«(*) (3) *=i where n(k) is an integer. The highest possible value of g, the number of periods, is equal to r, the number of variables. A system of periods possessing the property above is called a. fundamental period system. 967
APPENDIX 32 The Calculus of Variations INTRODUCTION In this appendix we develop topics in the calculus of variations that are useful in classical dynamics. FUNCTIONALS If with every value of the quantity x we associated a value y, we then say that y is a function of x and we write y = y(x). If with every value of the set of quantities x = xux2, . . . , xn we associate a value y, we thaii say that y is a function of x and we write y = y(x). If with every function y(x) we associate a value z, we say that z is a junctional of the function y(x) and we write z = z[y(x)]. A function y (x) can be considered to be a point in an infinite dimensional space, since it assigns a particular value to each of the infinite number of points along the x axis. Hence a functional can be considered to be an extension of the idea of a function of a finite number of variables to a function of an infinite number of variables. Example. The distance between two points 1 and 2 on a curve y = y(x) can be written i dx (1) -r-(£) The value of S depends on the choice of the function y(x) and is thus a functional of y(x). Note that the set of functions y(x) are in this case restricted to those that pass through the points 1 and 2. 968
Stationary Values of a Function 969 THE FUNDAMENTAL PROBLEM IN THE CALCULUS OF VARIATIONS The fundamental problem in the calculus of variations is to find the functions for which a functional has a stationary value, that is, the functions for which the functional has a maximum value, a minimum value, or some kind of inflection. This problem is closely related to the problem of finding the points for which a function has a stationary value, the essential difference being that in the latter case we are dealing with a finite dimensional space, and in the former case an infinite dimensional space. It is very helpful in the understanding of the calculus of variations to keep this relationship in mind. Before tackling the calculus of variations we therefore review the process of finding the stationary points of a function, presenting it in a way that brings out the close relationship between the two problems. STATIONARY VALUES OF A FUNCTION Definition. If y(x) is a function of x, u an arbitrary vector in the space x, and e a scalar, then the function has a stationary value at a particular point x if and only if at the point lim = 0 (2) c-^o e v ' for all values of u. Theorem 1. The function y(x) has a stationary value at a given point if and only if 37(x) dx, = 0 /=1,2,...,« (3) at the point. proof: If we expand y(x + eu) in a Taylor series we obtain 3/(x) £2 d2y(x) (X-t- £U)=/(X}-t- £2, i and thus ,(, + ^)-^) + ^^+^22^^+--- (4) / j —,«*/ y(x+eu)-y(x) 9y(x) lim = X —5 ut- (5) Combining Eqs. (2) and (5) we obtain as a necessary and sufficient condition for y(x) to have a stationary value at a given point that Sir1*-0 (6)
970 The Calculus of Variations at the point. If dy(\)/dxl = 0 for all values of i then Eq. (6) is true. Conversely if Eq. (6) is true for all values of u, then dy(x)/dxi = 0. Thus condition (6) is equivalent to condition (3). This completes the proof of the theorem. THE d NOTATION Given a point x and a function y(\), we define X = X + €U (7) y=y(x)=y(x+eu) (8) 3x(e,u) dx': dy. 3e 3j(€,U) Je = 0 E = 0 de = ude (9) d^^dx, (10) In terms of this notation a stationary point of the function y(x) is a point at which dy = 0 (11) for all values of dx. AUXILIARY CONDITIONS Frequently the admissible values of x are limited to some subset of the points in the space x. When this is the case the stationary points for a function y(x) are different from those when there is no restriction on the values of x. Theorem 2. Consider the function y=y(x) (12) If the values of x are restricted to those that satisfy the equation z(x) = 0 (13) then the function y(x) will have a stationary value at a given point if and only if there exists a parameter X such that Eq. (13) together with the equations 3 3 X; [y(x) + \z(x)] =0 /=1,2,3,...,/1 (14) are satisfied at the given point. It follows that the determination of the stationary points of the functiony(x) when the values of x are subjected to the auxiliary condition z(x) = 0 is equivalent to finding the stationary points of the function f(x,\)=y(x) + \z(x) (15)
Auxiliary Conditions 971 proof 1: At a stationary point dx, dx;=0 (16) But not all the dx, are independent, since they must satisfy the condition 3z(x) dz(x)="2 dX: dx,= 0 (17) Equation (17) can be used to eliminate one of the dx, from Eq. (16). To eliminate dxx we multiply Eq. (16) by dz(\)/dxl and Eq. (17) by dy(x)/dxl and subtract the resulting equations to obtain 2 / = 2 9y(x) 9z(x) 9y(x) 9z(x) 9.x, 9xi 9.x, dx;=0 (18) Since Eq. (18) must be true for arbitrary values of dx2, dx3, ..., dxn, it follows that the coefficients of the dx{ must separately vanish, which leads to the result dy(x)/dx,. dy(x)/dx{ If we define 9z(x)/9.x,- dz(x)/dx{ dy(x)/dX] dz(\)/dxl and make use of Eq. (19) we obtain 9y(x)/9jc, i = 2,3, . . . , n dz(x)/dx, = -X /= 1,2,3,.. .,/i (19) (20) (21) which is equivalent to Eq. (14). Finally we note that the conditions for a stationary point of the function /(x, X) =y(x) + Az(x) are 3/(x,A) 3A 8/(x,A) 3y(x) = z(x) = 0 3z(x) 3^c, dxj ' " 3x, which are equivalent to the previous set of conditions. (22) (23) proof 2: The theorem could also be proved by multiplying Eq. (17) by an arbitrary parameter X and adding the result to Eq. (16) to obtain 97(x) +A8z(x) dX: dX: dx,= 0 (24)
972 The Calculus of Variations This equation must be true for arbitrary values of A and any n — 1 of the quantities dxr If we choose A such that then ox, ox, 2 /=2 Mx) +A8z(x) dx, dx,- dx;=0 (25) (26) Since dx2,dx3, ..., <ixw can be chosen arbitrarily, Eq. (26) will be true if and only if Mx) +A8z(x) = o dXj dxf Combining Eqs. (25) and (27) we obtain Eq. (14) i — 2,3, . . . , n (27) STATIONARY VALUES FOR A FUNCTIONAL Definition. If z[/(x)] is a functional of y(x), (f>(x) an arbitrary function in the given function space, and e a scalar, the functional z[/(x)] has a stationary value for a given function if and only if the function satisfies the equation lim Z[HX) + €HX)] ~Z[m] = 0 (28) for all values of <f>(x). Theorem 3. Consider a functional z[/(x)] of the form </(*)] = f/(*>j>/W* (29) •'a where y = y(x), y' = dy(x)/dx, f{x, y, y') is some known function of x, y, and y', a and b are constants, and the functions y(x) are restricted to those that have a given value at x = a and a given value at x = b. The functional above has a stationary value for the function y = y(x) if and only if the function satisfies the equation dx 9/(*> 7> /) 3/' 9/(*> 7> /) 3j = 0 (30) Equation (30) is known as the Euler equation. proof: If we substitute Eq. (29) in Eq. (28) we obtain as the condition for a stationary value iim± f*[/(*>y + €<k/ + **')-/(*'y>/)]^=° (31) €—»0 € Ja
Stationary Values for a Functional 973 and since we are restricting ourselves to functions that have a given value at x = a and a given value at x = b, we have <f>(a) = <f>(b) = 0 (32) If we expand the first term in the integrand in Eq. (31) in a Taylor series in the quantities ecf> and €</>' we obtain f(x, y + e<t>, / + €<#>') = 2Q J, (* ^ + V 3^7) /(*, y, /) ■ df(x, /, /) 3/( x, 7, /) 0 ^ +<t> dy 3/ + Substituting Eq. (33) in Eq. (31) and taking the limit we obtain r df(x, y, /) 9/( X, 7,7') (¢) — +<f> dx=0 dy dy' Integrating the last term by parts and making use of Eq. (32) we obtain (33) (34) i a, 9/(x> y> /) *■ ay dx = <t> *f(x> y>/) dy i K d 4> Jc ^ dx \ d 9/(*> j> y') dy dx dx 9/(*. /■ y') dy dx Substituting Eq. (35) in Eq. (34) we obtain b[ 3/(*. y, /) d dy dx df(x> y> y') dy cj>dx=0 (35) (36) Since (j> is completely arbitrary except for the conditions imposed by Eq. (32), Eq. (36) can be true only if Eq. (30) is satisfied. Conversely if Eq. (30) is satisfied Eq. (36) is true. This completes the proof of the theorem. The Euler equation (30) is a second order differential equation. There are two special cases in which the Euler equation can be immediately reduced to a first order equation. Case L If the function f(x, y, y') does not explicitly contain y, that is, if /(*>/>/)=/(*>/) (37) the Euler equation becomes simply _d_ dx df(x,y') dy = 0 which can be integrated to give 9/(*> /) dy = constant (38) (39)
974 The Calculus of Variations Case 2. If the function f(x, y, y') does not explicitly contain x, that is, if f(x, y,/) = f(y,/) the Euler equation becomes A dx 3/ 9/(/./) 9/ = 0 which reduces to »2. 92/(/,/) , , 92/(/,/) „ 9/(7./') n 87 dy' 9/'2 9/ (40) (41) (42) Multiplying Eq. (42) by y' and then adding and subtracting the term [9/(.y>y)/9y]y we obtain 92/(/,/) ,, 92/(j,/) , n 3/(,,/) w 3,3/ ■r + 9/2 7/ + 9/' 7 9/( J. J') „ 3/(7./) 9/' '7 dy / = 0 But }2r/„ „"> a2, 92/(7./) ,2 , 92/(j,j') , „ , 9/(7.7') „ d a a , 7 + : 7/ + 5-; 7 ="J- 97 9/ 3/2 -7 y 87 -7 <ix 9/(7.7') 97' 7 and 9/(7./) 3/(7./) , 7 + aT 7 dx dy' 7 dy Substituting Eqs. (44) and (45) in Eq. (43) we obtain A dx 9/(7.7') , , 7 -/ = 0 which can be integrated to give 9/(7.7') 97' 97' y' — j — constant (43) (44) (45) (46) (47) The preceding theorem can be readily generalized to include functionals of a set of functions y(x)=yl(x),y2(x), . . . ,yn(x). This is done in Theorem 4. Theorem 4. Consider a functional z[y(x)] of the form -!>(*)]= (?f(x>y>y')dx (48) where y = y(x), y' = dy(x)/dx, f(x,y,y') is some known function of x, y, and y', and the set of functions y(x) is restricted to those for which y(a) and y(b) have known fixed values. The functional above has a stationary value for the
The 8 Notation 975 set of functions y = y(x) if and only if the set of functions y(x) satisfies the equations A dx df(x,y,y') df(x>y>y') = 0 i= 1,2, . . . , n (49) proof: The proof is a simple generalization of the proof of Theorem 3. THE 8 NOTATION Given a function y(x), a function /(/), and a functional z[jy(x)] we define y=y(X) + e<t>(x) f=f(y) and 8y = «/ = Sz = 3j(€ de 9/(£. 3£ 3z(e ¢) *) ¢) de = <p(x)de Je = 0 de 3e de €=0 Using the relations above we obtain the following particular results: dy(x) dx _3_ 3e dy(x) dx de = de dy(x) dcf>(x) + € dx dx ?*=f[^] = |w (50) (51) (52) (53) (54) (55) de (56) and I'<'H =£.)>><*< 3e - fb\\ 9 de £{[-tf^iJe}d^£Sfdx (57) In terms of the 8 notation the functional z[y(x)] has a stationary value at a point if and only if 8z = 0 (58)
976 The Calculus of Variations for all values of 8y. Using the 8 notation, the proof to Theorem 3 becomes 8z = 8 I f(x, y, y')dx = J 8f(x, y, y')dx Ja Ja Ja -!' Ja -r Ja • by H -^-^ by dy dy' dx df(x,y,y') df(x, y, y') d 9^ » + 8/ &<W 9/(*> J. /) J / 9/(*> J> /) eke 37 J* 9/ 8y dx = 0 (59) The Euler condition then follows from the condition that Eq. (59) must be true for arbitrary 8y. AUXILIARY CONDITIONS Theorem 5. Consider the functional -!>(*)] = f f(x>y>y')dx Ja (60) If we restrict ourselves to functions y(x) that in addition to passing through a given point at x = a and a given point at x = b also satisfy the condition rb g(x> 7' 7 ) "■* = constant (61) then the functional z[j^(x)] will have a stationary value for the function y(x) if and only if there exists a parameter A such that Eq. (61) together with the equations A dx dh(x,y,/) dh(x,y9/) dy = 0 where are satisfied. h (X, y, /) = f(x, y, /) + \g(x, y, /) (62) (63) proof: If the functional z[/(x)] has a stationary value for/(x) then dy dx \ dy' s( f(x,y,y')dx= f Ja Ja But 8y is not arbitrary but must be such that Sydx=0 8ydx=0 (64) (65)
Auxiliary Conditions 977 If we multiply Eq. (65) by an arbitrary parameter X and add the result to Eq. (64) we obtain f[ 8y dx = 0 (66) "b\ m _ j_ ( Ml dy dx \ dy' where h=f+Xg (67) If we assume in Eq. (66) that dy is arbitrary, Eq. (66) will be true if and only if s(f)-tr° <«> If we solve Eq. (68) tor y{x) and choose the constants of integration such that y(x) passes through the given points at x = a and at x = b we obtain y=y(x,X) (69) We can now choose X such that Eq. (61) is satisfied. The resulting function will then be the desired function. Theorem 6. Consider the functional z[y(x)] = (bf(x>y>y')dx (70) If we restrict ourselves to sets of functions y(x) = yx{x\ y2(x), • ■ • >yn(x) that in addition to passing through a given point at x = a and a given point at x = b satisfy the condition g(x,y,y') = 0 (71) then the functional z[y(x)] will have a stationary value for the set of functions y(x) if and only if there exists a function X(x) such that Eq. (71) together with the equations A dx 3/z(x,y,y') M(x,y,y') = 0 /= 1,2, . . . , n where h(x,y,y')=f(x, y,y') + A(x)g(.x,y,y') (72) (73) are satisfied. proof: If the functional z[y(x)] has a stationary value for y(x) then ^■*n*-2jr/[£-£($)]**-o (74) But the 8y,- are not arbitrary but must be such that ^ = 2^,.+ 2^/,= 0 (75)
978 The Calculus of Variations Multiplying Eq. (75) by an arbitrary function X(x) and integrating, we obtain ?i &(*>*.+&<*>* dx=0 (76) But i \±<MW,\*-l -ftr &)£&,)*• -i b _8 b d | 3 Jx I dy- 7<*g) 8ytdx (77) Substituting Eq. (77) in Eq. (76) and adding the result to Eq. (74) we obtain 2/ i Ja dh _ d_( IK 97/ dx \ ax 8yidx=0 where If we choose A such that dh _ _d_( IK) =o Sy,- dx = 0 dy j dx \ 3y j then Eq. (78) becomes v fh\ AK-A( IK) Since 8y2,8y3, . . . , Syn are arbitrary, Eq. (81) is true if and only if dh d I dh \ n . ^^ ■^— r- I ^—r I = 0 / = 2,3,...,« dy. dx \ dy; (78) (79) (80) (81) (82) Equations (80) and (82) are equivalent to Eq. (72). This completes the proof of the theorem. EXAMPLES Example 1. Find the shortest path between two points (a,c) and (b,d) in the x-y plane. Solution. The distance between the two points is given by S = £[\ + (y')2]dx (83) The functional will have a stationary value for the function y(x) that satisfies the equation d< U1+iy)T}-l[i+w2]l/2-° dx \ dy' (84)
-1/2 = 0 Equation (84) reduces to Integrating with respect to x we obtain y>[\+(/fy-=A where A is a constant. Solving Eq. (86) for y' we obtain Examples 979 (85) -1/2 y = [A\i-A2)} Integrating with respect to x we obtain y = Bx + C 1/2 = B (86) (87) (88) This is the equation of a straight line. The constants B and C can be chosen in such a way that y(a) = c and 7(6) = d. It can be shown that some other neighboring curve between (a, c) and (b,d) covers a greater distance, hence the distance is a minimum if (a, c) and (b, d) are joined by a straight line. Example 2. A particle of mass m is free to slide on a smooth curve lying in the x-y plane and joining the points (0,0) and (a,b). Gravity is acting in the positive y direction. The particle is released from rest at the point (0,0). Find the shape of the wire that will result in the particle arriving at the point (a,b) in the shortest possible time. Solution. The energy of the particle is a constant of the motion. It follows that {mv2-mgy = 0 (89) Hence the speed of the particle at an arbitrary point on the wire is given by v = {2gy)'/2 (90) The time required to go from (0,0) to (a, b) is given by ^2^1/2 (91) J(0,0) v Jo n<™\x/2 dx Vgyy The integrand when expressed as a function of x, y, and y' is not an explicit function of x; hence T will have a stationary value for the function that satisfies the equation [see Case 2, Eq. (47)] y 3/" 1 + (/)2 y i1/21 i + (/)2 y J/2 = A (92)
980 The Calculus of Variations where A is a constant. Carrying out the differentiation and solving for y' we obtain A2y A^ The solution to this equation in parametric form is v= 0-sinfl 2,42 1 — COS0 1A which is the equation of a cycloid. y-\-izr\ (93) (94) y = ^T (95)
Answers
J
Answers CHAPTER 47 1. Gr = — T + mg cos 0 G9 = — mga sin 0 G^ = 0 2. Gx = GY=G9=G<t> = 0 Gz=-(M+m)g Gr =/(/•) CHAPTER 48 1. m(f> - p<j>2) = Fp m(pj> + 2p<j>) = F+ mz = Fz 2. a. r =(111/8)(^ + 1,)1½2^+ (^/1,)] G€ = (V2)[l+fo/0]I/2 Gv = (^/2)11+(^/1,)]1^ b. T = (m/2Z?2)[(a2 + Z?2)*2 - 2oxi] +172] Gx = ^11+(^/^/2 c. T=(m/2)(a2sin20 + A2){02 + [A2/(a2 + A2)]} G9 = (A2 + a2sin20y/2F9 Gx = [(A2 + a2sin20)/(A2 + a2)]]/2Fx 4. r -r02 + aAsin[0-(At2/2)- Bt]- a(At + 5)2cos[0 - (^/2/2) - 5/] — gcos0 + (X/ma)(r — a) = 0 r0 + 2r0 + a4 cos[0 - (^/2/2) - Bt] + a(>4/ + 5)2sin[0 - (^/2/2) - Bt] + gsin0 = 0 5. mi:' — mu*2x' — 2mciy' = Fx. my' — mu*2y' + 2mcox' = Fv. CHAPTER 49 1. 10 degrees of configurational freedom; 8 degrees of motional freedom 2. One degree of configurational freedom; one degree of motional freedom
984 Answers 3. 5 degrees of configurational freedom; 3 degrees of motional freedom 4. 6 degrees of configurational freedom; 4 degrees of motional freedom 6. Rr(t)=-aRx(t)/x(t) 7. RjRz=-a 8. Rx:Rv:Rz::x:y: -a CHAPTER 50 1. Q9 = Fa cos 0 Q^ = Fb cos § 3. z = [gb2/(a2 + b2)]t2/2 4. (1 + 4x2)x + 4xx2 + (2g - co2)x = 0 5. (b/a)]/2co 9. rax + ma cos 0 0 — ma sin 0 0 2 = F 3mcos9x + 4ma9 = 6Fcos9 — 3 rag sin 0 CHAPTER 51 1.0 = (v/a)[\ - (b/r)] F = - v2b2/r3 2« r = ( g/2co2)cos(cot + a) + A sinh cot + B cosh to/ F= — 2ragsin(co/t + a) + 2raco2(/l cosh to/ + B sinh cot) 3. ff = (l//x)ln[(/iV/«)+l] /V = -ma/[fit + (a/i>0)]2 ^ = -H^VI 4. 0 = (1/(^ + fcfl)]ln{[(j^ +A:a)v/a]+ U Fr = -ma/{[(ii + **)/ + (fl/t>0)]} ^ = -/x|/?r| 5. z = A sin(cot + 5), where A and 5 are constants co2 = (k/m)[a2/(a2 + Z>2)] The magnitude of the force is k(p2 + z2)1/2. Fp = kb- (mb/a2)z2 F^ = (mb/a)z Fz = kz + raz 6. Fr = 3mgcos9 — 2mg 0^=cos-1(§) 8. rap - rap<j>2 = 2pA (d/dt)(mp2<i>) = 0 raz = — rag + <zA p2 = a2 — <zz Fp = (2ma3gp)/(a2 + 4p2)2 /^ = 0 Fr = (ma4g)/(a2 + 4p2)2 The particle will not leave the paraboloid. 9. Fr = (Img cos 0/3) - (4mg/3) Fe = - mg sin 0/3 0,= cos"1 (4/7)
CHAPTER 52 Answers 985 1. mx + myz = Fx + zFy mz = Fz y - zx = 0 2. mtx + my = Fxt + Fy mz = Fz x - ty - a = 0 3. Let xx = x, x2 =y, x3 = z, mg = imgsina — kmgcosa <j> = angle the plane of the disk makes with the x axis \p = angle a fixed radius in the disk makes with a horizontal radius <$> = At + B ;//= {[2gsinacos(^/ + B)]/(3aA2)} + Ct + D x = [gsina/(3^2)]sin2(,4/ + B) - (aC/A)sin(At + B) + £ 7 = [gsina/(3^2)P/ - sin(^r + B)cos(At + 5)] + («C/^)cos(^/ + B) + F where A, B, C, D, E, and F are determined by the initial conditions 6. See Problem 9, Chapter 64. CHAPTER 53 2. mbO — my sin 0 + abO — ay sin 0 + mg sin 0 = 0 my — mb0 sin0 — mb02cos0 + ay — ab0sin0 + ky = 0 3. £/=- (mo)2/2)(x2 + j2) - mco(jy) -yx) CHAPTER 54 1. b0 —ao)2sin(o)t — 0) + gsin# = 0, where 0 is the angle the rod makes with the downward vertical. 2. (at2 + b)0 + 4a/0 - 2a sin 0 = 0, where 0 is the angle the line from the center to the bead makes with the line from P through the center. 5. L = (mx2/2) - [k(x - a)2/2] - pmgxx/\x\ 6. 27r[\la/24g]x/2 11. (2mg sin a cos a)/[3(M + m) — 2m cos2a] 12. (M+ m)x+m(b7 a)cos002+ m(b - a)sin00 = - kx sin0x + \(b - a)0 = —gsinO CHAPTER 56 1. The two equations that follow can be solved to obtain x(0) and 0(0): (M + m)x + 2Ma0 cos0 = P \(M + m)x2 + 2Ma202 + 2Macos0x0 - 2Mgacos0 = E where P and £ are constants determined by the initial conditions
986 Answers 2. r = {[mv20/(M + m)][\ - (r2/r2)] + [2Mgr0/(M + m)][\ - (r/r,)]}^2 <j> = 'o^oA2 3. L = ^ amr2 + ^ [amr2 + (4ma 2/3)](0 2 + sin2 ° <#>2) + mgco$0(ar + a) — \{nmg/a)(r — a)2 <j>} = \amf2 + \\amr2 + (4ma2/3)](92 + sin20<j>2) — mg cos 0(ar + a) + \{nmg/ a){r — df <j>2 = [amr2 + (4ma2/3)]sin20$ 4. 4>i = ^ (M + m)x2 + mi2 + mxs cos a — rags sin a <j>2 — (M + m)x + mi cos a <j>3 = s — st + {[(M + ra)g/2sina]/[4(M + m) — 2m cos2a]} <j>4 = x — xt — {(mgt2sina cosa)/[4(M + m) — 2m cos2a]} CHAPTER 57 A 1. Let ej be a unit vector in the direction of f and e2 a unit vector in the direction of the rod, the positive sense being from M to m. The center of mass will have a velocity [f/(M + m)]ex and the rod will be rotating with an angular velocity f [(Mb + mb — ma)/(mMa2)](e2 X ex). 2. PQ: v(C) = 5a/4 <o = 9a/4a QR: v(C) = - a/4 <o = - 3a/4a 6. 3a/2 CHAPTER 58 3. a. The height of the perigee is less by approximately 110 km. b. The height of the perigee is less by approximately 3.5 km. ^. tfj = (gR)]/2> where R is the radius of the earth. 5. If/(r)= -k/P then r = ^/cos{[l - (mk/h2)]l/2(<j> - B)} h2 > mk r = A/cosh{[(mk/h2) - l]1/2(<f> - B)} h2 < mk r = (A/<j>)+B h2=mk 6. r = p{ 1 + e cos[y(<J> — a)]} ~ \ where p = y2(h2/mk), e = [1 + y2(2Eh2/mk2)]1 /2, y2 = 1 + (em/h2), and a is an arbitrary constant. If c <^m/h2 and is < 0, the orbit is a precessing ellipse with the pericentron advancing by an amount (2m /y) — 2tt^ —(jrem/h2) each revolution. 7. u(r) = -(ae-br/r) + (h2/2mr2) If /*2 < (2 +V5 )exp[-(l + V5 )/2](ma/b\ then t/(r) has a minimum and bound orbits are possible. If the particle is moving in a circular orbit of
Answers 987 radius R, then E = (a/2R)(bR - \)e~bR &ab- (a/2R) h2 = maR(bR + \)e~bR « maR CHAPTER 60 2. p2 = kz/E 4. [(R2/4)(\ + A)]/[l + Asin2(0/2)]2, where A = 4ER(RE + a)/a2 5. If a < [k(n - 2)/2E]l^n then a = 7rn(n - 2)(2-n)/"(k/2E)2/". Ifa> [k(n - 2)/2E]x/n then a = 7ra2[l + (k/Ean)\ 6. a. a(0,4>) = tfW/{4sin\0/2)[tf2cos24> + Z>2sin2<f> + c2tan2(0/2)]2} b. o(0,<f>) = a2b2c2/{cos30[a2cos2<j> + Z>2sin2<f> + c2tan20]2} c. o(0,<f>) = a2b2c2cos0/{sin40[a2cos2<j> + b2sin2<j> + c2cot20]2} CHAPTER 61 4. f(a) = A2sin«/[cos3«(l - A2tan2«)1/2], where A = (V2 - V£)/(2VV0) 5. 7t/2 < a < ir when m < M a = ir/2 when m = M 0 < a < 7r/2 when m > M 6. 4a2(5 - 9sin20)/(l -9 8^¾)1^ where 0 < 0 < sin-l(l/9) 7. Let vx and v2 be the postcollision speeds, 0X and 02 the scattering angles, and k/r2 the magnitude of the force. Then vx = Vcos0x 0X = i?m-\2k/mV2s) v2 = Fcos02 02 = cot-1 (2k /mV2s) CHAPTER 62 1. b. [G(mx + m2 + m3)/a3]^2 CHAPTER 63 1. V3
988 Answers CHAPTER 64 1. G1=_-ma2tt2smacosa/2 Gj = 0 G5 = 0 The 3 axis is along the rod, the 2 axis is in the plane determined by the rod and the axis of rotation. 4. 2tt/[co(1 +3cos2«)1/2] 8. See answer to Problem 3, Chapter 52. 11. 2ir[I^/(I^LScosA)]1/2, where £2 is the angular velocity of the earth, S is the spin of the gyroscope, and A is the latitude. CHAPTER 65 1. <oT = (3^/2 <,-2 = (6g/\\a)^2 5. [mv/(M + 2m)][t — co-1sinco/], where <o2 = k(M + 2m)/(Mm) 6. <oT = 0 C02 = (3A:/4m)1/2 <o5 = [(M + 2m)k/(4Mm)]^2 cT = «(1,1,1) c-2 = >8(0,1, - 1) c5 = y(2m, - M, - M) q-x = (M0/m) + <f>+.\p qi = <i>- i> q^ = 20 -<j>- xp 9. <7f = x cos <J> + y sin <J> ^ = — x s^n <t> + y cos 4> cot(2<f>) = (cof-co2)/(2a) 10. q\ = xcos<j> + ysin<j> q2 = — .xsin<J> + ycos<j> cot(2<f>) = (m,- m2)/(2fi) 11. The correctness of a set of normal coordinates can be verified by expressing the Lagrangian in terms of the set and checking that it is in the appropriate form. CHAPTER 71 X- a- P=PP PP = {pl/p3)-a2P <#> = (/V/P2) ~ a P* = ° Z=Pz Pz=0 b. p - p$2 - 2ap$ = 0 (d/dt)(p2$ + ap2) = 0 £ = 0 5. 7/ = (/72 + />2+/>2)/[2(M+m)] + [/>,2 + (Pt/r2) + (/>2A2sin20)][(M + m)/(2mM)] + F />*> /V> Pz> H> P*> Pe + (/W sinfl) 6. a. // = [pi/2m(r0 — at)2] — ma2/2 — mg(r0 — at)cos0 b. No c. No
Answers 989 7. //(0, ¢, p9, Pp 0=- (mr2/2) + (p20/2mr2) + (/>2/2mr2sin20) + mgr cos 6 where r = r(t) and /• = dr{f)/dt. The term —(mr2/2) may be dropped because it does not affect the equations of motion. 8. H = {p2x/2m) + (fc/x)exp(- t/r) 9. r(r - 2)r2 + A2 + 2r + Br2 = 0, where ,4 = r20 = constant and 5 is a constant 10. H = (l/2m)(p2 + p\+ pl)+ V + o)X2p{ - uxxp2 CHAPTER 72 1. mx = — kx\x\ 2. p(l + 4p2) + 4pp2 + 2gp - A2p~3 = 0, where A = p2<j> = constant 3. p2 = 0 #2 = (a + Z>#2)/?2 qx + (/?|6 + fc,)^ = 0 CHAPTER 73 1. Pr = (*PX + J2P, + %)/(*2 + / + *2)1/2 a = [«a + Wy - (*2 + /)a1/(*2+/)I/2 p* = xPy-ypx 2. ?1 = {At + 5)1/2 q2=t+C-(At + B)x/2 where ^4, 5, and C are constants 3. ff = (p2x/2m) + (p2/2m) + V[(x2 + y2)^2] /T = (p2./2m) + (p2/2m) + V[(x'2 + /2)'/2] + u/px, - «x'/y CHAPTER 74 8. a = \ /3 = 2 11. Pl=(pi-p2)/(2qi) + [l/(2?i)]J{[3/(?i,fc)/3fil - [3/(^1^2)/3¾]}^ + g(<7.) ^2 = />2+/(?!>£>) where / and g are arbitrary functions of the variables indicated CHAPTER 76 1. P = p/m]/2 Q = m,/2q 2. A: = Q + ep 3. K = J^sinOo, S, - co2 fij) - (ePj/wj)
9!H) Answers 4. q = (l/w)tan(wf + C) - / 5. ^ = At + 5 #2 = -t2 + O + D where yl, 5, C, and Z> are constants 9. gi=>4 £?2=5 P\ 2/ + C P2=-/ + Z> where A, B, C, and Z> are constants CHAPTER 77 1. 9! = At + B q2= -t2+ Ct + D where A, B, C, and Z> are constants 3. a. Pi=-qi Qi=Pi b. Pi=-qi Qi=Pi c. P = (1 + ^cos^?)tan /? £> = ln(l + qcos2p) 4. F(p,Q)=-2^ft ^(q,P) -2/^/ft 5. F= - g sin"'[ty/Cg)1'2] - (Kq/2)(2Q - K2q2)x'2 6. F(q, P, 0 = 2 ;/-(q, 0^/ ^(P> Q> 0 = 2 MQ> OPi 7. a. F(p,Q,t)=-pQ+te? F(q,P,t)=Pq+tep F(q,Q,t) = (Q-q){\-ln[(Q -q)/t\) b. #=0+^ 8. F(q, Q,t)=- mgQq - (mg2Q 3/6) CHAPTER 78 1. q=t + A p= -at2-2aAt +B where A and B are constants 2. x = (vcos<j>)t y= -(g/2)t2 + (vsin<f>)t 3. jc = - a/2/4 y = at2/4 4. S(u,v,E,a,t) = j[(mE/2) + (a/2w) - (km/2u)]]/2du + j[(mE/2) - (a/2v) - (km/2v)]l/2dv - £/ The orbit is obtained from the equation dS/da = /?. 5. S(quq2,E,a,t) = j[a + 2mEcosh2q} -2m(Ax + ^2)cosh^1]1/2^1 + /[ — a — 2mEcos2q2 — 2m(Ax — A2)cosq2\l/2dq2 — Et The orbit is obtained from the equation dS/da — /3. 7. S(£,7),<j>,E,a2,a3,t) = J{(Bm£/4) + (mE/2) - [(Am - a2)/2fl-(a2/4|2)}1/2^ + f{-(Bmi1/4) + (mE/2) - [(Am + a2)/2i1] - (a3W)}1/2<fy + a3<t> — Et
Answers 991 8. S(r,0,<j>,E,a2,a3,t) = f[a2 - 2mg(0) - (a2/sin20)]^2d0 + }{2m[E - f(r)] - (a2/r2)} dr + a3<f> - Et The orbit is obtained from the equations dS/da2 — fi2 and dS/da3 = /?3. 9. S(^,-q,<j>,E,a2,a3,t) = f{(mE/2) + («2/20 - [mf®/$ - (<*32/4£2)}1/2^ + j{(mE/2) - (a2/2V) - [mgW/ift - (af/V)}1^^ + a^ - Et 10. S(^,7],4>,E,a2,a3,t) = f{2ma2E + [a2 - 2mo2j{£)/(i? - 1)] - [a2/^2 - 1)2]} rfg + f{2ma2E - [a2 + 2mo2g(<n)/(\ - v2)] - [a2/(\ - v2)2]} dV + a3<j> — Et CHAPTER 83 2. z = Bt3 CHAPTER 87 1. X'= - Vt' / = A sm(at'/T) v'x= -V v'y = (Aa/T)cQs(at'/T) The time between successive maxima is 2flT/co. The distance between successive maxima is 2aT V/co. 2. The event corresponding to the closing of end C of the cylinder and the event corresponding to the arrival of end A at the end C of the cylinder have the same coordinates. In frame S the coordinates of this event are x = —2a, t = 0. In frame S' the coordinates of this event are x = — 4a, t =fi(2a/c). Hence in both frames the rod will be trapped in the cylinder. Note that the end A of the rod travels a distance 3 a in frame S' between the time end D of the cylinder is closed and the time end C of the cylinder is closed. CHAPTER 88 1. v3 = [m]y(v])v] - w2y<>2)tf2]/[™iY<>i) + m2y(v2)] m3 = [mxy(vx) + m2y(v2)]/y(v3) where y(v) = [1 - (v2/c2)]~^2 CHAPTER 89 2. x = (v0/a)\n[at + (1 + a2t2)l/2] y = (c/a)[(\ + a2t2y/2 - 1] a = (qE)/(y0mc)
Index Abelian group, 950 Absolute derivative, 926-927 Acceleration: in cylindrical coordinates, 45 of gravity, 73,74 in moving frame, 38 relative, 27 in spherical coordinates, 45 Action and angle variables, 807-814 Action at distance, 56 Action and reaction, law of, 60, 66, 216 Amonton's law of friction, 75 Angular frequency, 160 of driven oscillator, 163 Angular impulse, 127,129-130 Angular momentum: and arbitrary point, 102, 226-227 axial component, 271-272 and center of mass, 227 change due to impulse, 127,129, 130, 338 conservation, 123,174, 248, 580, 585, 586,598 definition, 102, 226 of incident particle, 191 Poisson bracket of components, 830 of rigid body, 271,272, 312 time rate of change, 103, 229 and torque, 103, 229-230 Angular motion, 176-177, 600 Angular velocity, 33-36 definition and properties, 33-38 Euler angles, 310, 650 Anholonomic constraint, 532 Anholonomic system: definition, 554 and Gibbs-Appell equations, 722-725 Aphelion, 184 Apocentron, 183 AppelTs equations of motion, 722 AppelTs function, 721 Apse, 177,600 Apsidal angle, 177,600 Apsidal distance, 177, 600 Apsidal radius, 600 Apsis, 177,600 Associative law, 945 Automorphic group, 948 Axial direction, 269 Axis of rotation, instantaneous, 36 Basis vectors, 683 Bertrand's theorem, 606-613 Bilinear form, 928 Boundary conditions, 435 Bound orbit, 606, 607 Brachistochrone problem, 979 Calculus of variations, 968-980 Canonical transformation: definition, 755-756, 762-764 inverse, 761 Jacobian for, 760, 771-772, 794-795 Lagrange bracket test, 757 and modified Hamilton's principle, 846-852 Poisson bracket test, 826 product of two, 761-762 properties, 759-762 Canonical variables, 773-774 1-1
1-2 Index Carrier space, 955 Cartesian configuration space, 240-241, 349,514 Cartesian tensors, 417-425 Cartesian vector quantity, 421 Cartesian vectors, 420421, 424 Center of mass: acceleration, 67, 217 coordinates, 205, 624-627, 632-633 definition, 66 determination, 67 motion, 67-68,197, 216, 220, 254, 274, 320,655 position, 217 properties, 65-68 of rigid body, 269 table of values, 479482 of two particles, 196,622 uniqueness, 66 velocity, 217,622 Central force, 172-174 Central force motion: angular motion, 176-177, 600 circular, 606-607 fundamental equations, 175,598 Hamilton-Jacobi treatment, 803-804 Lagrangian treatment, 597-602 Newtonian treatment, 172-184 orbit in, 175,599 radial motion in, 176, 599-600 stability of circular orbit in, 184 symmetry of orbit, 600 for two-particle system, 198 Centrifugal force, 138 Character: of matrix group element, 961 of operator, 961 of representations, 961-963 Characteristic equation, 931 Character table, 689-690 Christoffel symbols, 926 Circular motion in central force field, 606-607 Class of element in group, 950,960, 962 Clocks, 9 Closed orbit, 607 Closed systems, 215 Closure law, 945 Coefficient of kinetic friction, 76 Coefficient of restitution, 131, 339 Coefficient of static friction, 75 Collision: change in energy in, 204, 623 definition, 49-50, 203, 622, 875-876 description, 203, 622-624 elastic, 131,204,339,623 inelastic, 131,339 of particle with surface, 131-132 possible, 50, 876 properties, 50, 876 quantities conserved in, 52-56, 876- 883 rigid body, 339 two-particle, 203-209, 622-633 Colonization of space, 643 Column matrix, 449 Column matrix representation of vector, 683,953 Complex numbers, 437 Complex of group, 949 Compression period, 131, 339' Configuration of harmonic system, 684 Configuration space: basis vectors in, 683 Cartesian, 240-241, 349,514 decomposition, by group theory, 688 definition, 241, 514 for holonomic system, 539,682 modal directions in, 683 vector in, 682-683 Configuration of system, 241,513 Conjugate elements in group, 950 Conservation: of angular momentum, 123,174, 248, 580,585,586,598 of energy, 123,174, 249,598, 888 of four-momentum, 897 of linear momentum, 57,122-123, 248, 578, 883-884 of mass, 57 see also Constants of motion Conservative force field, 114-115, 244 Conserved quantities, 50-58, 876-884 see also Conservation; Constants of motion Constants: , of motion, see Constants of motion in solution of differential equation, 433434 table, 477 Constants of motion: in central force motion, 174 definition, 120, 247, 575 generalized coordinate as, 740
Index 1-3 generalized momentum as, 577-578, 585,740 Hamiltonian as, 576-577, 585, 739, 829 in Hamiltonian formulation, 390,739- 740 importance, 122, 247-248 in Lagrangian formulation, 364-365, 575-584 number of independent, 120-121, 247, 575 in Poisson formulation, 829 in restricted three-body problem, 638 and symmetry, 578-584 see also Conservation; Conserved quantities Constraint, 349,528 Constraint forces: anholonomic: definition, 532 determination, 555-556 definition, 349, 528 generalized components, 532-534, 541 geometric, 529 holonomic: definition, 349-351, 531-532 determination, 549-550 ideal, 529-531 kinematic, 529 Contact: perfectly rough, 76 perfectly smooth, 76 Contraction of tensor, 422 Contravariant vector, 923 Conversion factors, table, 478-479 Coordinates: center of mass, 205, 624-625,632-633 generalized, 352-353,514, 539 laboratory, 205, 624-625, 632-633 orthogonal curvilinear, 43-46, 427431 relative, 205,624-625,632-633 Coordinate transformations, 470474 Coriolis force, 138 Coset, 949 Couple, 256-257 Coupled oscillators, 676 see also Vibrating systems Covariant derivative, 927 Co variant vector, 923 Cross product, 425 Cross section: differential scattering, see Differential scattering cross section hard sphere, 192-193,619 inverse square, 193, 619-620 total scattering, 187-188, 614-615, 628 Curl, 408409, 414,416, 425 Curvilinear coordinates, orthogonal: equations of motion in, 144-145 properties, 4346,427431 Cyclic coordinate, 577,747 Cyclic group, 948,950 Cyclic system, 808 Cylindrical coordinates: definition and properties, 4445, 429430 equations of motion in, 144 relation to rotating frame, 45 d'Alambert's principle, 56 Damped harmonic oscillator, 160-162 Decay modulus, 161 Decrement, 161 Degeneracy, 686-687 Degrees of freedom, 352, 534 Del operator, 409 Derivative, absolute, 926-927 Descartes, Rene, 12 Determinant, 452454, 930 Diagonalization of matrices, 935-936 Differential, exact, 442446 Differential equations: linear, 436440 ordinary, 432435 Differential scattering cross section: center of mass, 207-208, 627-629, 631-633 definition, 188-189,615-616 determination, 190-192,617-619 hard sphere, 192-193, 619 inverse square, 193, 619-620 transformation between lab and center of mass, 208-209,629-633 Differentiation: covariant, 927 of integral, 441 partial, 398400 of tensor, 424 Dimension: of representation, 955 irreducible, 960 Disk, rolling, 534-536, 556 Displacement, virtual, 349, 354, 515, 540,712-713
1-4 Index Dissipation function, 560-561 Distance, shortest, 978 Divergence, 408409,414,416,424 Dot product, 403, 424, 425 Driven harmonic oscillator, 162-168 Dynamical state, 120 Dynamical system, 65, 215 Dynamics, aim of, 122 Dynamics problems, method of solution, 88 Effective potential, 176,179,181, 326, 599 Eigencolumn, 931 Eigenvalues of matrix, 931-934 Einstein, Albert, 12 Elastic collision, 339 Electromagnetic field, generalized potential, 561 Elementary system, 521 Ellipsoid: of inertia, 291 Poinsot, 330 Elliptic functions, 330,467-469 Energy: of acceleration, 721 conservation, 123,174, 249, 598, 888 dissipation, 164 kinetic, see Kinetic energy and momentum, 888-889 potential, see Potential energy relativistic, 888 and work, 107-108,234-237, 888 Equations of motion: in configuration space, 241 Euler's, 322, 656, 719, 730 Gibbs-Appell, 720-725, 727-730 Hamiltonian type, 744-747 Hamilton's, see Hamilton's equations of motion Lagrange's, see Lagrange's equations of motion Newton's, see Newton's equation of motion in non-inertial frames, 136 Poisson's, 828-830 relativistic, 885, 901, 903 of rigid body, 274, 321-323 Routh's, 746-749 for two-particle system, 197 Equilibrium points, 638 Equimomental systems, 323 Equivalence of matrices, 928 Equivalent systems of forces, 255, 258 Escape speed, 182-183 Euclidean set of coordinate systems, 417 Euclidean transformation, 473 Euler angles, 303-307, 649 Euler equation, 972-974 Euler's equations of motion, 322, 656, 719,730 Euler's theorem, 914 Exact differentials, 442446 Extended point transformation, 750-751, 759 External forces, 78, 215 F function, 763 Fictitious forces, 137-138 Field, 113, 243,408,409,424 Fixed plane, definition, 269 Force: action at distance, 61, 78 active, 528 central, 172-174 classification, 61 components: generalized, see Generalized components of force ordinary, 518-519 physical, 517 conservative, 114, 244 of constraint, see Constraint forces contact, 61, 78 coplanar, 260 definition, 59, 65,885 derivable from generalized potential, 559 determination, 77 electromagnetic, 61 equivalent, 77, 255-258 external, 65,215,219 fictitious, 137-138 field, 113,243 generalized components, see Generalized components of force given, 528 gravitational, 60-61 instantaneous action at distance, 61 internal, 65, 215 interparticle, 216 inverse square, 178-184, 601-602 irrotational, 113, 243-244 line of action, 256 and linear momentum, 218-220
Index 1-5 Lorentz transformation, 885-886 net, 219 net external, 67,216 Newtonian, 59-62 parallel, 261 passive, 528 point of application, 77, 255 reduction of systems, 258 relativistic, 885-886 representation, 78 specification, 77 Foucault pendulum, 138-139 Four-force, 898 Four-momentum, 897 Four-velocity, 896 Frames of reference: inertial, 11-12,857-858 relative motion, 32 transformation between, 13-21, 859- 865 Frequencies of system, 810 Frequency, resonance, 164 Friction, 75-76, 80 Function, 968 Functional: definition, 968 in Hamilton's modified principle, 843 in Hamilton's principle, 839 stationary values, 972-975 Functional dependence, 460-463 Galilean transformation, 13, 20 Galileo, 12 Gauss's theorem, 412 Generalized components of force: Cartesian components, 355 as covariant components, 572 definition, 354,515 derived from potential function, 356 determination, 355, 357, 516 and Euler angles, 654-655 for holonomic constraint, 356 for holonomic system, 539-541 for quasi-coordinates, 712 transformation, 516-517 Generalized components of impulse, 588 Generalized coordinates, 352-353, 514, 539 Generalized force, see Generalized components of force Generalized force functions, 558-562 Generalized momentum: conservation, 577, 578, 585 definition, 387,577,737 Generalized potential, 558-560 Generalized velocities, 352, 514,539 General tensors, 921-927 Generating functions, 789-795 Geometry of space, 8 Gibbs-Appell equations of motion, 720-725, 727-730 Gibbs-Appell function, 721, 725, 727-730 Gradient, 408-409, 414416, 424 Gram-Schmidt orthogonalization, 674, 915-917 Gravitational constant, 73 Gravitational force: due to arbitrary mass distribution, 74 due to spherically symmetric mass distribution, 74 effective, 139 of homogeneous disk, 78 reduction to single force, 262 Gravity: acceleration due to, 73 law, 73 Green's theorems, 410412 Group: Abelian, 948 automorphic, 948 complete Lorentz, 893 conjugate elements in, 950 cyclic, 948 definition, 945 homomorphic, 948 inhomogeneous Lorentz, 893 isomorphic, 948 Lorentz, 893 matrix representation, 955 operator representation, 955 order, 945 periodic, 948 Poincare, 893 properties, 945-946 special Lorentz, 894 of symmetry operations, 685 Group theory, 945-950 Hamiltonian: canonical transformation, 774, 782 conservation, 390-391, 576, 585, 739 definition, 387-388,737 importance, 390 of particle in electromagnetic field, 740
1-6 Index relativistic, 903 and total energy, 388, 738 transformation, 774, 791 Hamiltonian function, 738 Hamilton-Jacobi equation: generalization, 801-802 for particle in central force field, 803 for simple harmonic oscillator, 802-803 solution by separation of variables, 800-801 time-dependent, 797-799 time-independent, 799-800 Hamilton's canonical equations of motion, 773-782 Hamilton's characteristic function: definition, 800 for simple harmonic oscillator, 802 Hamilton's equations of motion, 387-391, 737-740 in condensed notation, 769 and Hamilton's principle, 840-841 for particle in electromagnetic field, 740-741 relativistic, 903 of second kind, 744-745 Hamilton's modified principle, 843-852 Hamilton's principal function: definition, 798 for particle in central force field, 804 for simple harmonic oscillator, 802 Hamilton's principle, 838-841 relation to modified Hamilton's principle, 845 Harmonic motion, 158-168, 373-378, 665-707 Harmonic oscillator, simple: amplitude, 161 angular frequency, 161 critically damped, 162 damped, 160-162 decay modulus, 161 driven, 162 effect of constant force, 166 frequency, 161 overdamped, 161 period, 161 relaxation time, 161 solution by canonical transformation, 782-783 solution by Hamilton-Jacobi technique, 802-803 underdamped, 160 Harmonic system, 373, 665, 684 Holonomic constraints, 349-351, 356, 531-532 Holonomic systems, 351-352, 538 Homomorphic group, 948 Hooke's law, 74 Hooke's law potential, motion in, 602 Huygens, Christian, 12 Ideal constraint force, 529-531 Identity element in group, 945, 946, 950 Ignorable coordinate, 577 Impact parameter, 188, 615 Impulse: angular, 127,129-130 definition and properties, 126-132 generalized components, 588 on harmonic oscillator, 167 instantaneous, 127,128,130, 338, 588 Impulsive equations of motion: Lagrange's, 588-590 for particle, 126-132 for rigid body, 338-341 Indices, raising and lowering, 925-926 Inertia: law of, 12 moment of, see Moment of inertia Inertia ellipsoid, 291 Inertial forces, 137 Inertial frames, 11-12, 857 Inertia tensor, 286-298, 647-648 Initial conditions, 435 Integral: differentiation, 441 line, 409 surface, 409410 volume, 409410 Interaction between particles, 59, 875 Internal forces, 215 Interparticle forces, 216 Invariable line, 330 Invariable plane, 330 Invariant subgroup, 949 Invariant subspace, 686, 953, 963 Inverse of element in group, 945, 946, 950 Inverse square force field, motion in, 178-184,601-602 Inverse transformations, 909-911 Irreducible representations: definition and properties, 959-963
Index 1-7 and natural modes, 689-692 Irreducible sub space: of configuration space, 686-687 properties, 963 Irrotational force field, 113 Isomorphic groups, 947-948 Jacobian: for canonical transformation, 760, 771-772,794-795 in change of integration variables, 912-913 definition, 401 and inverse transformation, 909 properties, 401 of transformation from lab to center of mass, 630 Jacobian elliptic functions, 330,467469 Jacobi's identity, 824 Kinematics: Newtonian, 746 relativistic, 857-871 of rigid body, 303-319, 647-652 Kinetic energy: average, for driven oscillator, 163 change in collision, 623 definition, 107, 234 of dynamical system, 234 generalized coordinates, 524-525 of particle, 107, 234 relative, 623 of rigid body, 271-272, 312-314, 650-651 of two-particle system, 198 and work, 107-108, 234-237 Kronecker delta, 34,405 Laboratory coordinates, 205,624-625, 632-633 Lagrange bracket, 757, 770, 825 Lagrange multipliers, 53,464466,533, 880,970-972 Lagrange points, 639-643 Lagrange's equations of motion: for anholonomic systems, 554-556 for elementary systems, 521-526 and Hamilton's principle, 840 for holonomic systems, 361-365, 538-546 for impulsive forces, 588-590 for Lagrangran systems, 564-566 for quasi-coordinates, 715-719 relativistic form, 901 for rigid body, 653-655 and tensor analysis, 570-572 Lagrangian: definition, 363, 564 harmonic, 373, 665,682 approximately, 375-376, 668-670 indeterminate nature, 566 invariance, 578-581, 585-586 normal coordinates, 674 relativistic, 901 Lagrangian system, 564 Laplacian, 425 Latitude, definition, 140 Left-handed systems, 406 Legendre transformation: definition and properties, 918-920 of kinetic energy, 744 of Lagrangian, 737, 746 Length: contraction, 868 definition, 7 measurement, 7 Levi-Civita symbol, 35, 321,407,423424, 562 Linear differential equations, 436440 Linear equations, system, 458459 Linear momentum: at arbitrary point, 101, 218-219 at center of mass, 219 conservation, 57,122-123, 248, 578, 883-884 definition: for particle, 57,101 relativistic, 883-884 for system of particles, 218 effect of impulse, 126-130, 338 and force, 59, 218-220, 885 Lorentz transformation, 884 time rate of change, 59, 220, 885 Linear operator: matrix representation, 954 on vector space, 953 Linear operator representation: of group, 955-966 of symmetry group, 686 Linear transformations, 472, 676 Line integral: definition and properties, 409414 independence of path, 443446 Line of nodes, 303 Logarithmic decrement, 161 Lorentz force, 561, 741 Lorentz group, 893-894
1-8 Index Lorentz transformation: of accelerations, 863-864 consequences, 867-869 of force, 885-886 four-dimensional formulation, 893-895 of linear momentum, 884 statement and derivation, 13-21, 859-860 in vector form, 18-20, 860-861 of velocities, 861-863 Mass, 56-57, 883 Matrix: adjoint, 455 adjugate, 455 antisymmetric, 450 augmented, 459 coefficient, 459 column, 449 complex conjugate, 449 congruent, 929 definition, 447 diagonal, 450, 932 diagonalization, 935-936 eigenvalues, 931-934 equivalent, 928 inverse, 455-456 non-negative, 450 non-singular, 455 order, 447 orthogonal, 456, 930 positive definite, 450, 932 positive semidefinite, 450, 932 properties, 451 rank, 457, 459460 regular, 455 row, 449 scalar, 450 similar, 929, 932 singular, 455 skew-symmetric, 450 square, 450 submatrix,457 symmetric, 450, 932 trace, 961 transposition, 448,451 unit, 450451 unitary, 457 zero, 449,451 Matrix addition, 447,451 Matrix equality, 447 Matrix multiplication, 448, 451 Matrix representation: of group, 955 of linear operator, 954 of vector, 953 Matrix transformation: congruent, 929 definition, 928 equivalence, 928 orthogonal, 930 similarity, 929 Mechanics, definition, 1 Metric tensor, 896, 925 Modal direction, 374, 666, 671, 683 Modal frequency, 374, 666, 683, 688-689, 692-695 Modal vector, 374, 666, 673,683 Modified Hamilton's principle, 843-852 Modified velocity, 865, 876 Modulus of elasticity, 75 Moment of inertia: definition and properties, 269-271, 286-287, 647 principal, 292-295, 648-649 table of values, 482486 Moment of mass, 66 Momentum: angular, see Angular momentum and energy, relativistic relation between, 888-889 linear, see Linear momentum Motion: of center of mass, 67-68,197, 216, 220, 254,274,320,655 central force, see Central force motion constants of, see Constants of motion equations, see Equations of motion harmonic, 158-168, 373-378, 665-707 impulsive, 126-132, 338-341, 588-590 one-dimensional, 151-156 relative: of frames of reference, 32-39 of particles, 27-28 uniplanar, 269-275 Multiplication table for group, 947-948 Multiplicity: of degeneracy, 686-687 of modal frequency, 688-689 Multiply periodic function, 810,967 Natural coordinates, 376-378, 670-673 Natural modes of motion: definition and properties, 373-376, 665-670
Index 1-9 determination, using group theory, 682-695 superposition, 377, 671 Newton, Isaac, 12 Newton's equation of motion: application to elementary problems, 88-89 in configuration space, 240-242, 513-514 in cylindrical coordinates, 144 and d'Alembert's principle, 835 impulsive form, 126 and Lagrange's equations, 361-363, 522-524 in non-inertial frames, 136-139 in spherical coordinates, 145 statement, 59-60,102 for system of particles, 216, 240-242, 513-514 in tensor form, 570-571 and torque-angular momentum theorem, 103 and work-energy theorem, 108 Newton's law for collisions, 339 Newton's law of gravity, 73 Newton's laws: first law, 12 second law, 59-60 third law, 60-61 see also Newton's equation of motion Nodes, line of, 303 Noether's theorem, 581-586 Normal coordinates, 378, 673-676 see also Natural coordinates Normal divisor, 949 Nutation, 326 Operator: linear, see Linear operator projection, 690-691, 964-966 Orbit: bound,606 in central force motion, 175,599 in inverse square force field, 180, 182, 601-602 Order: of cyclic group, 948 of element in group, 947 of group, 945 Orthogonal curvilinear coordinates, see Curvilinear coordinates Orthogonality condition: for irreducible representations, 961-962 for modal vectors, 673 Orthogonal matrix, 456 Orthogonal transformation, 473 Oscillations, small, see Vibrating systems Oscillators: coupled, 676 harmonic, see Harmonic oscillator see also Vibrating systems Parallel-axis theorem, 270, 290 Partial differentiation, 398400 Particles: collisions between, 203-209, 622-623 definition, 1-2 equations of motion, see Equations of motion interaction, 49 relative motion, 27-28 statics, 83-85 system of two, 196-200 Pericentron, 183 Perihelion, 183 Period: fundamental, 813,916 of harmonic oscillator, 159 of inverse square orbit, 183 of multiply periodic function, 813, 916 in oscillatory one-dimensional motion, 153-156 Periodic group, 948 Permutation symbol, 35, 321,407, 423424, 562 Phase angle, 163 Poincare group, 893 Poinsot ellipsoid, 330 Point transformation, 750-751 Poisson brackets, 759, 823-827 Poisson's equations of motion, 828-830 Poisson's hypothesis, 339 Poisson's theorem, 829 Position, 27, 38,45,196 Postulates of mechanics: classical: 1,11 11,12 III, 20 IV, 56 V,216 relativistic: 1,857 11,857 III, 858 IV, 883
MO Index Potential: effective, 176,179,181, 326,599 generalized, 558-560 power law, 600-602 single-valued, 114, 244 Potential energy: average, for driven oscillator, 163 definition and properties, 113-115, 243-244 determination from orbit, 178 gravitational, 117 of harmonic oscillator, 158 and work, 115, 244 Potential function, 114, 243, 560 Power of group element, 946-947 Power law potential, 600-602 Principal axis, 292-298, 648-649 Principal directions, 292-298,648-649 Principal moments of inertia, 292-295, 648-649 Product: cross, 403,406 dot, 424 scalar, 403,424 vector, 403,424 Products of inertia, 287 Projection operator, 690-691, 964-966 Proper subspace, 952 Quadratic forms: definition, 937 properties, 937 reduction to sum of squares: by linear transformation, 939-941 by orthogonal transformation, 941-943 simultaneous, 943-944 transformation, 938-939 types, 937 Quasi-coordinates: definition and properties, 711-714 and Gibbs-Appell equations, 720-725 Lagrange's equation, 715-717 Quotient law, 925 Radial motion, 176, 599-600 Reduced mass, 198 Reducible representation, 959-960 Reduction of quadratic form, 939-943 Reference frames, 7, 9, 23,136 Regular transformation, 471 Relative coordinates, 205,624-625, 632- 633 Relative motion, 27-28, 32-39, 197-200 Relativistic mechanics: basic postulates, 857-858, 883 dynamics, 875-889 four-dimensional formulation, 893-898 Hamilton's equations in, 903-904 kinematics, 857-871 Lagrange's equations in, 901-902 Relativity: general theory, 1 principle, 11-12 special theory, 1 see also Relativistic mechanics Relaxation time, 161,166 Representation: of operator by matrix, 954 of quantity by tensor, 921 of vector by matrix, 953 Representations of group: characters, 961-963 dimensions, 955 direct sum, 957 equivalence, 962 equivalent, 956 generation, 957 irreducible, 956, 959-961 matrix, 955 natural, 958 one-dimensional, 957 operator, 686, 955 reducible, 959-960 regular, 958-960 unitary, 957 Resonance, 164-166 Restitution: coefficient, 131,339 period, 131,339 Restricted three-body problem, 634-643 Riemannian space, 896, 925 Right-handed systems, 406 Rigid body: angular momentum, 271-273, 312-315 collisions, 339-341 configuration, 253-254, 271, 303 definition, 253 kinetic energy, 271-273, 312-315, 650-651 statics, 264-266 Rigid-body motion: dynamics, 320-323,653-656 equations of motion: Gibbs-Appell, 727-730 Lagrangian, 653-655 Newtonian, 254, 261, 274-275, 320-323
Index Ml impulsive, 338-341 kinematics, 303-315,647-651 uniplanar, 269-275 Rocket, 220 Rotating frames, 33-39,136-139 Rotation: instantaneous axis, 36-37 as transformation, 474 Rotational motion, 33 Routhian, 746 Routh's equations of motion, 746-749 Scalar, 402,419,922-923 Scalar field, 408 Scalar potential, 470 Scalar product, 403,424 Scattering angle, 188-192, 205-206, 618-619,624-625 Scattering cross section, see Cross section Separable systems, 807 Separation of variables, 800-801 Simple closed orbit, 607 Simple harmonic motion, 158-168 Simple harmonic oscillator, see Harmonic oscillator Sine function, 397 Space: absolute, 22 configuration, see Configuration space definition, 7 Speed of particle, upper limit, 16, 20, 858 Sphere, rolling, 332-333, 658, 730 Spherical coordinates: definition and properties, 45-46, 430431 equations of motion, 145, 525-526 relation to rotating frame, 45-46 Spring constant, 74 Spring linear, 74 Statics: of particle, 83-85 of rigid body, 264 Stationary value: of function, 464-466,969-972 of functional, 972-978 Steady-state solution, 163 Stokes' theorem, 413 Strings, 76-77 Subgroups, 949-950 Summation convention, 425, 768, 922 Surface: perfectly rough, 76 perfectly smooth, 76 Symmetric top, see Top, symmetric Symmetry: and constants of motion, 578-584 and natural modes of motion, 682-707 Symmetry group, 685-686 Symmetry operations, 684-686 Tables, 477-486 Tensors: Cartesian, 288,417-426 general, 921-927 inertia, 286-298,647-648 and Lagrange's equations of motion, 570-572 metric, 896,925 in relativistic mechanics, 893-898 Three-body problem, restricted, 634-643 Thrust of rocket, 220 Time: absolute, 22 definition, 8 measurement, 8-9 Time dilation, 867-868 Time rate of change in different frames, 33-36 Top: asymmetric, 327-332 stability of motion, 658-659 spin, 325 symmetric: freely rotating, 323-325 in gravitational field, 325-327, 657, 748-749 Torque: and angular momentum, 103, 229-230 at arbitrary point, 103, 227-228 axial component, 273-274 at center of mass, 228 definition, 102-103, 227 produced by couple, 256-257 Total cross section, 187-188, 614-615, 628 Trace of matrix, 930, 933-934 Transformation coefficients, 308,418-419 Transformation matrices, 307, 308 Transformations: canonical, see Canonical transformation congruent, 929 coordinate, 470-474 equivalence, 928 extended point, 750-751, 759 Galilean, 13, 20 inverse, 909-911 lab to center of mass, 206-209, 625-633
1-12 Index Lorentz, see Lorentz transformation matrix, see Matrix transformation orthogonal, 930 point, 750-751 of quadratic forms, 938-939 similarity, 929 Transient solution, 163 Trojan asteroids, 643 Turning points, 152,177, 180 Twin paradox, 869-870 Two-particle collisions, 203-209,622-633 Two-particle systems, 196-200 Uniplanar motion of rigid body, 269-275 Unitary representation, 957 Units, abbreviations, 477 Vector: basis, 952 Cartesian, 420 components, 404-407, 952 contravariant, 923 covariant, 923 definition, 402, 420 derivative, 408 linearly independent, 952 negative, 402 unit, 403 zero, 403 Vector algebra, 402-407 Vector calculus, 408-416 Vector field, 409 Vector function, 408 Vector identities, 486 Vector potential, 740 Vector product, 403, 424 Vector space, 682, 951-954 Velocity: in cylindrical coordinates, 45 Lorentz transformation, 861-863 modified, 865 in moving frame, 38 relative, 27 relativistic addition, 867 in spherical coordinates, 45 Vibrating systems, 373-378, 665-676, 682-695 Virtual displacement, 349, 354, 515, 540, 712-713 Virtual work, 350, 354, 515, 540, 654-655,713 Weightlessness, 138 Width of resonance curve, 165 Work: in conservative force field, 114, 244 definition, 107, 235 and kinetic energy, 108, 236-237, 888-889 on particle, 107, 235 and potential energy, 115, 244 on system of particles, 235 virtual, 350, 354, 515, 540, 654-655, 713 Wrench, 261