Analytical Dynamics

Theory and Applications

Analytical Dynamics
Theory and Applications

Mark D. Ardema
Santa Clara University Santa Clara, California

Kluwer Academic / Plenum Publishers New York, Boston, Dordrecht, London, Moscow

Library of Congress Cataloging-in-Publication Data Ardema, Mark D. Analytical dynamics: theoiy and applications/Mark D. Ardema. p. cm. Includes bibliographical references and index. ISBN 0-306-48681-4 1. Dynamics. I. Title. TA352.A85 2005 620.1'04—dc22 2004054841

ISBN: 0-306-48681-4 ISBN E-book: 0-306-48682-2 ©2005 Kluwer Academic / Plenum Publishers, New York 233 Spring Street, New York, New York 10013 10 9 8 7 6 5 4 3 2 1

A C.I.P. record for this book is available from the Libraiy of Congress All rights reserved No part of this book may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microtilming, recording, or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Permissions for books published in Europe: Permissions for books published in the United States of America: Printed in the United States of America

may your forces be conservative, your constraints holonomic, your coordinates ignorable, and your principal function separable

Preface Chapter 1. Review of Newtonian Dynamics 1.1 Basic Concepts 1.2 Kinematics and Newtonian Particle Dynamics 1.3 Work and Energy 1.4 Eulerian Rigid Body Dynamics 1.5 Examples 1.6 Motivation for Analytical Dyanmics Notes Problems Chapter 2. Motion and Constraints 2.1 Newton's Second Law 2.2 Motion Representation 2.3 Holonomic Constraints 2.4 Nonholonomic Constraints 2.5 Catastatic Constraints 2.6 Determination of Holonomic Constraints 2.7 Accessibility of Configuration Space 2.8 Examples Notes Problems Chapter 3. Virtual Displacement and Virtual Work 3.1 D'Alembert's Principle 3.2 Lagrange Multiplier Rule 3.3 Virtual Velocity and Variations 3.4 Forms of the Fundamental Equation 3.5 Given Forces Notes Problems Chapter 4. Variational Principles 4.1 Energy Relations 4.2 Central Principle 4.3 Hamilton's Principle 4.4 Calculus of Variations 4.5 Principle of Least Action Notes vii xi 1 1 4 12 17 22 27 28 29 47 47 49 50 56 57 57 59 59 62 62 65 65 70 74 77 78 80 81 83 83 86 86 90 94 97

viii Problems Chapter 5. Generalized Coordinates 5.1 Theory of Generalized Coordinates 5.2 Examples Notes Problems Chapter 6. Lagrange's Equations 6.1 Fundamental Equation in Generalized Coordinates 6.2 Multiplier Rule 6.3 Lagrange's Equations 6.4 Special Forms 6.5 Remarks 6.6 Embedding Constraints Notes Problems Chapter 7. Formulation of Equations 7.1 Remarks on Formulating Problems 7.2 Unconstrained Particle 7.3 Constrained Particle 7.4 Example - Two Link Robot Arm 7.5 Example - Rolling Disk Problems Chapter 8. Integration of Equations 8.1 Integrals of Motion 8.2 Jacobi's Integral 8.3 Ignoration of Coordinates 8.4 Separation of Variables Notes Problems Chapter 9. Examples 9.1 Street Vendor's Cart 9.2 A Useful Identity 9.3 Indian Rope Trick Notes Problems Chapter 10. Central Force Motion 10.1 General Properties 10.2 Inverse Square Forces 10.3 The Time Equation

Contents 97 101 101 105 108 108 109 109 114 117 119 121 124 126 127 131 131 132 136 138 142 146 149 149 151 157 162 162 163 165 165 170 171 175 175 179 179 182 185

1 Definitions and Fundamental Equation 13.3 Indirect Methods 12. Hamilton's Equations 15.1 Introduction 12.2 Impulsive Constraints 13.1 Quasi-Coordinates 14.4 Lagrange's Equations for Impulsive Motion Notes Problems Chapter 14.5 Poisson Brackets 15.6 Reduction of System Order Notes Problems ix 187 187 189 189 192 199 201 201 205 205 208 211 217 219 222 223 225 225 228 233 240 241 241 245 245 248 249 251 259 259 261 261 264 265 268 270 274 280 280 .4 Stability of Hamiltonian Systems 15.3 Impulsive Motion Theorems 13.3 Gibbs' Theorem and the Gibbs-Appell Equations 14.Contents Notes Problems Chapter 11.1 Rigid Body Motion with One Point Fixed 11. Gyroscopic Motion 11. Stability of Motion 12.5 Liapunov's Direct Method Notes Problems Chapter 13.2 Heavy Symmetrical Top 11.2 Hamilton's Equations as a First Order System 15.1 Derivation of Hamilton's Equations 15.2 Fundamental Equation 14.2 Definitions of Stability 12.3 Examples 15. Impulsive Motion 13.4 Applications Notes Problems Chapter 15.3 Some Applications Notes Problems Chapter 12.4 Stability of Orbits in a Gravitational Field 12. Gibbs-Appell Equations 14.

3 Integration of the Hamilton-Jacobi Equation 17.3 Infinitesimal Contact Transformations Problems Bibliography Index Contents 281 281 283 286 288 290 294 295 297 297 303 308 310 313 316 316 319 319 325 329 332 335 337 .2 General Contact Transformations 16. Hamilton-Jacobi Equation 17. Approximation Methods 18.X Chapter 16.5 Separable Systems Notes Problems Chapter 18.1 Variation of Constants 18.2 Variation of the Elements 18.3 Homogenous Contact Transformations 16.1 Introduction 16.2 Hamilton-Jacobi Theorem 17.1 The Principal Function 17.4 Examples 17.4 Conditions for a Contact Transformation 16. Contact Transformations 16.5 Jacobi's Theorem Notes Problems Chapter 17.

" Indeed. The methods that I explain in it require neither constructions nor geometrical or mechanical arguments. Its basic principles are few in number and relatively easily understood. and its consequences are very rich. and it is this meaning of the term that is meant here. relativity theory. the number of figures per chapter tends to decrease. Decartes. Those who love Analysis will. Then Newton united both terrestrial and celestial dynamics into one comprehensive theory in the Principia. the "proof" of a new result in classical dynamics consists in showing that it is consistent with Newton's three "Laws of Motion". The term analytical dynamics has now come to mean the developments in dynamics from just after Newton to just before the advent of relativity theory and quantum mechanics. To this day. thermodynamics. Galileo. Mecanique Analytique (1788)-^Lagrange used the term "analytical" to mean "non-geometrical. and quantum XI . but only the algebraic operations inherent to a regular and uniform process. see mechanics become a new branch of it and will be grateful to me for thus having extended its field.Preface In his great work. It was not until Kepler." This was in marked contrast to Newton's Philosohiae Naturalis Principia Mathematica (1687) which is full of elaborate geometrical constructions. It has been remarked that the classical Greeks would have understood some of the Principia but none of the Mecanique Analytique. It was of great interest to the Greeks in the classical period. Among the branches of physics that have adapted the techniques of dynamics are the various theories of deformable continuous media. although not monotonically. Lagrange made the following boast: "No diagrams will be found in this work. dynamics has become the prototype of the branches of mathematical physics. electricity and magnetism. Dynamics is the oldest of the mathematical theories of physics. with joy. Although the Greeks had some of the concepts of statics correct. Huyghens. Being the oldest and best established. and others in the seventeenth century that the principles of dynamics were first understood. their knowledge of dynamics was seriously flawed. Frequent use will be made of diagrams to illustrate the theory and its applications. although it will be noted that as the book progresses and the material gets "more analytical".

the only exceptions are references to related material not covered in this book. . Although the division is somewhat artificial. sophisticated analysis. These are the books by Rosenberg {Analytical Dynamics of Discrete Systems. to name just a few. Essentially. This situation has now changed.Xll Preface mechanics. non-Euclidean geometry. Both subjects are covered in this book. It should be remarked that one of the primary sources for Rosenberg is the book by Pars. Over the years. differential equations. are usually quite complicated. No specific references will be made to material adapted from these two books because to do so would make the manuscript cumbersome. Many authors contributed to the theory. the excellent treatment of Rosenberg is the primary source. the dynamics of interest in engineering applications was extremely simple. theory of groups and fields. 1977) and by Pars {A Treatise on Analytical Dynamics. the Second Essay on a General Method in Dynamics (1835) by Hamilton. the first called Lagrangian and the second Hamiltonian dynamics. There are two principal sources for this book. The concept of a dynamic system has been abstracted and apphed to fields as diverse as economics and biology. the development of analytical dynamics begins with d' Alembert's Traite de dynamique of 1743 and ends with Appell's Traite de Mecanique Rationelle of 1896. 1965). Disadvantages of these books are that Rosenberg is out of print and Pars has no student exercises. especially. and. it is a useful one. functional analysis. the body of knowledge called analytical dynamics has coalesced into two parts. being confined to Lagrangian Dynamics. but two works stand out: First. the Mecanique Analytique of Lagrange. covering much material of limited interest to engineering analysts. The dynamics problems that arise in the fields of robotics. differential geometry. biomechanics. involving typically three-dimensional motion of collections of inter-connected bodies subject to constraints of various kinds. covering the subject in a single comprehensive and rigorous volume. being mostly concerned with two-dimensional motion of rigid bodies in simple machines. and space flight. and the second is much broader. Pars' book is a monumental work. The scope of the first of these is more limited than the present book. and this has sparked a renewed interest in the methods of analytical dynamics. Branches of mathematics that have benefited from concepts that first arose in dynamics are algebra. For the Lagrangian portion of the book. These problems require careful. A few decades ago. Thus the time span is the 150 years covering the last half of the eighteenth and the entire nineteenth century. and second.

as will be shown in this book. and central gravitational attraction. and nonholonomic constraints. the simple pendulum. These examples are often quite lengthy. Thus frequent use is made of. and these are deliberately kept as simple as possible. Features involved include three-dimensional motion. many specific dynamics problems are easier to solve by the Lagrangian than by the Newtonian method. although this may be of limited importance to engineering analysts. The first type is intended to illustrate key results of the theoretical development. there are hundreds of available books. multi-body systems. One of the aims of Hamiltonian dynamics is to obtain not just the equations of motion of a dynamic system. Throughout the book there are historical footnotes and longer historical remarks describing the origins of the key concepts and the people who first discovered them. it must be confessed that the usefulness of the techniques developed is often limited. a far-reaching impact on many fields of mathematical physics. Perhaps Gauss said it best: "It is always interesting and instructive to regard the laws of nature from a new and advantageous point of view. For readers less interested in the history of dynamics than I. The Hamiltonian approach to dynamics has had. real-life problems. Because the Hamiltonian formalism generally requires the same effort to solve dynamics problems as the Lagrangian.^ Because most dynamics problems may be solved by Newton's laws. but their solution. There are two types of examples used in this book. the one-dimensional harmonic oscillator (linear spring-mass system). Another advantage lies in the mathematical elegance of the presentation. One approach to dynamics is to develop it by the axiomatic method . In the case of Lagrangian dynamics this justification is easy. and continues to have. for example. alternative methods must have relative advantages to warrant interest. The other type of example is included to show the application of the theoretical results to complex.Preface xiii As would be expected of a subject as old and well-established as classical dynamics. since. I have listed in the Bibliography only those books familiar to me. so as to solve this or that problem more simply. this historical information may be skipped with no loss in continuity. or to obtain a more precise presentation". however. the study of Hamiltonian dynamics is more difficult to justify. and this is an important reason for its study. Also listed in the Bibliography are the principal original sources and two books relating to the history of the subject. comprising an entire chapter in some cases. rigid bodies.

I will write F = F{xs) to mean that F is a function of the variables xi.18.".XnThis book is intended both as an advanced undergraduate or graduate text. the material is covered in forty fifty-minute lectures. The background expected is an undergraduate understanding of Newtonian dynamics and of mathematics. It seems to me. are not necessary to subsequent developments. Finally. the distinction will be made. The symbolism F{xs) G C" will mean that F(-) is of class n. The symbol F{x) will be used to mean both "F is a function of a.are outside the main development.1 1 . at most. Chapters 2 . and.--. • •. when it is clear what is meant. virtual work. and as a reference for engineering analysts.^ Today we use the principles of classical dynamics because they give a sufficiently accurate model of physical phenomena. This includes the derivation of Lagrange's equations as well as numerous applications. it is with great pleasure that I acknowledge the faculty at the University of California at Berkeley who first imparted to me the knowl- . In my own graduate course. and the Gibbs-Appell equations .on stability. Some comments on function notation used in the book are required. when there is a chance of confusion. that this method is inappropriate for a subject that is experimentally based. concern the development of Hamiltonian dynamics and its applications. expecially differential equations.--.4 cover the foundations of analytical dynamics that will be used throughout the rest of the book . and this has been done by many authors. The next three Chapters . 15 .Xn.constraints. The remaining Chapters. if ^(a. virtual displacements. For example. but this effort will pay off in the long run.^) and Xg = f{qr) then F{qr) will mean F{fi{qi.--. A summary of the organization of the book is as follows. fnilir'TQn). This is not meant to be comprehensive but rather covers only concepts that are needed later.XIV Preface familiar from Euclidean geometry. and variational principles.qn).The same symbol will denote sometimes two different functions.--." and "the value of the function F for a specific value of a. and no others. Similarly. that is that it is continuous with continuous derivatives up to order n in all of its arguements xi.) Lagrangian dynamics is contained in Chapters 5 . which held that Newton's laws described how nature actually behaves. however. This is in marked contrast to the metaphysical view in Newton's time. Chapter 1 is a review of Newtonian dynamics. impulsive motion. (It may be somewhat frustrating to some students to spend so much time on preliminary material. although important topics. F ^ F{xs) will mean that F is not (allowed to be) a function of the variables xi.Xn.

This was first clearly recognized by Carnot. Notes 1 2 3 Specific works referenced here.Professors Rosenberg. devoting much space in Mechnique Analytique to the subject. I am especially indebted to the late Professor Rosenberg who granted me permission to use freely material from his book. are Hsted in the Bibliography. . Lagrange himself was deeply interested in the history of dynamics.Preface xv edge and appreciation of dynamics . and Goldsmith. and in the rest of the book. Leitmann.

We generally will take reference frames fixed relative to the surface of the earth to be inertial. all known materials deform under forces. Let X^F be the resultant (vector sum) of all the forces acting on a mass particle of mass m. vector addition by parallelograms. Classical mechanics rests on three basic assumptions: 1. 3.Chapter 1 Review of Newtonian Dynamics 1. Newton's Laws.1 Basic Concepts Assumptions. that is.^Assumption (4) is clearly an approximation. Assumptions (1) . Physical objects are particles or collections of particles constituting rigid bodies. 2. This implies that the Pythagorean theorem. Then Newton's Second Law 1 . There exist inertial (Galilean) reference frames in this space. but this deformation is frequently negligible. now they are regarded as engineering approximations. and all elementary geometry and trigonometry are valid. The physical world is a three dimensional Euclidean space. 4. An inertial frame is one in which Newton's three laws hold to a sufficient degree of accuracy. The quantities mass and time are invariant.(3) were regarded as laws of physics at one time. they are measured as the same by all observers.

Definitions. L. Newton stated these laws for a single particle.(i) = constant. 1-2 Newton's First Law states that ii ^F_ — 0 then t. The subject of mechanics is conveniently divided into branches as follows: £i2 —» ^ (1-2) F21 P2 • • e m2 . as we have just done. that is a collection of particles whose relative positions are fixed. This "Law" is therefore a consequence of the Second Law. Fig. say F_2i. 1-1 Newton's Third Law states that given any two particles pi and p2 with masses mi and m2. 1-2): Zi2 = |Zi2|e = .F = ma (1. F_i2.1) where a = dPr/dt'^ = f is the acceleration of the mass particle. Euler and others generalized them to a rigid body.Z 2 i where e is a unit vector in direction F_i2. and where r is the position vector of the mass particle in an inertial frame of reference (Fig. 1-1) and d{ )/dt = (") is the time derivative in that frame. p^ mi Fig. That is. force is proportional to acceleration with proportionality constant m. is equal and opposite to that exerted by p2 on pi.2 states that:^ Analytical Dynamics Y. and these forces act on a line adjoining the two particles (Fig. the force exerted by pi on p2.

It is clear that there are two basic problems: 1. for example. In practice. 3. velocity. 4. Dynamics is that branch of mechanics concerned with v(t) j^ constant. . Given the forces. 4. In the rest of this chapter. mixed problems frequently arise. It provides a benchmark with which to measure the worth of these new approachs.Review of Newtonian Dynamics 3 1. as defined by Newton's three laws. the position. 2. Kinematics is that branch of dynamics concerned with motion independent of the forces that produce the motion. 2. This is sometimes called the "backward" or "controls" problem. It gives insight that leads to other approachs to dynamics.2. Some of this material will be needed later. find the forces that produced it (actually. and acceleration as a function of time).= Q. Kinetics is that branch of dynamics concerned with the connection between forces and motion. find the motion (that is. given the motion and some of the forces. This impUes that a(t) = 0 and consequently J2E. Statics is that branch of mechanics concerned with the special case v_{t) = 0. Statics is a special case of this. we will review elementary Newtonian dynamics. It allows a chance in a familiar setting to get used to the approach and notation used throughout this book. for the following reasons: 1. Given the motion. one can usually only find the resultant force). Many of the following results will be presented without proof. Basic Problems in Kinetics. 3. what is the resultant of the remaining force (s)? Reasons for Reviewing Newtonian Dynamics. This is sometimes called the "forward" or "dynamics" problem.

1-3): „(. 1-3 Note that v_{t) is tangent to the curve C. Then the velocity of P is defined as (Fig. To obtain scalar equations of motion.3) Similarly the acceleration of the point is defined as: „/^^_^ _. (1. Denote the position vector of P at time t by r(i).2 Kinematics and Newtonian Particle Dynamics Motion of a Point.4) v. the position vector is given by: Z = xi + yj + zk From Eqns.4) At Fig.5) a = xi + yj + zk . is called the speed of the point. 1-4).^dr_ lim rit + At)-r{t) lim ^r (1. and a are: u = xi + yj + zk (1._.j. the vectors of interest are written in components. The magnitude of the velocity vector.-.3) and (1. In rectangular components (Fig.4 Analytical Dynamics 1.k}.7) (1._ lim 0 v{t + M)-v(^ (1.6) (1. Rectangular Components. Consider a point P moving along a curve C relative to a reference frame {i. v{t) = \ll{t)\.

Similarly. E ^ . 1-4 We call {x. normal-tangential components are frequently useful. v = \y_\ = \jx^ + p + z'^ (1.j. and frequently desirable. . = J2Fy = my. all basis vectors will be triads of mutually-orthogonal.z) the rectangular components of position (or the rectangular coordinates) of point P. z) and {x.i^ + E^^^ (1. It is possible. called basis vectors. For planar motion (take this to be in the {x. {x. Any vector in three dimensions can be written as a linear combination of any three linearly independent vectors.y) plane).F^ mx . y.8) Expressing the resultant force on the mass m at point P in rectangular components E£ = E^-*' + E^. z) are the rectangular components of velocity and acceleration.y.Tangential Components.10) This is a sixth order system of ordinary differential equations. Introduce unit vectors tangent and normal to v_ as shown in Fig.Review of Newtonian Dynamics Fig.1) and (1.k}.=mz (1. to express r. The distance of P from the origin and the speed of P are given by: r = \r\ = / dx^ ^-y^ + z^ . righthanded unit vectors and will be denoted by "hats".7) gives three scalar equations of motion: Y. v_. and a in components along directions other than {i.9) and combining this with Eqns. In this book. Normal . (1. y. 1-5. respectively.

1-5 separately.12) It is clear that in general it will vary with time and thus det/dt ^ 0. 1-6 Fig.Analytical Dynamics Fig. for 9>Q (Fig. 1-6): et = cos 9i + sin 9j e„ = — sin 9i + cos 9j det • •^ • ^ • —. We must consider the two cases shown on Fig.11) (1. First. 1-5 The velocity and acceleration vectors expressed in these components are: V = vet dv . 1-7 .„ det = vet + v—rdt dt (1.= —9sm9i + 9 cos 9j = 9en Fig.

i.e. J2E. from Eqn.14) We call {v. for rectilinear motion or at a point of inflection (Fig. a = vet + v\9\er. dt Thus for both cases: -dF = 1^1^^ so that. 1-7): 1 it = cos 9i + sin 6j e„ = sin 9i — cos 9j det = —0 sin 9i + 9 cos 9j = —9er.15) Note that e„ is undefined and p = oo for ^ = 0.16) Fig. Now let p = . v\9\) the tangential and normal components of acceleration. 1-9 . Hence we may write .Ft = mv . Y^Pn = m— (1. 1-8). 1-9). then S = R9 S = R9 = v R = 9 Thus we call in general p the radius of curvature. If the resultant force acting on a particle is expressed in normal-tangential components.15) give the scalar equations of motion: V Y.13) (1. V (1.1) and (1. (1. (1.2 a = vet-\ P e„ (1. = Yl Pt^t + Yl^n^ni then Eqns. 1-8 Fig.12).Review of Newtonian Dynamics Next for 6 < 0 (Fig.^ > 0. Suppose the motion is on a circle of radius R \9\(Fig.

(r.(j)) are shown on Fig. The cylindrical coordinates (r. Q) are just the familiar plane polar coordinates.19) . 1-10 Note that if z = constant.18) Fig.17) (1.zk (1. 4>. The spherical coordinates {r. From the geometry. In three-dimensional (3-D) motion it is often advantageous to resolve the velocity and acceleration into cylindrical or spherical components.Analytical Dynamics Cylindrical and Spherical Coordinates and Components. The relation to rectangular coordinates is given by X = r sin 9 cos (p y = rsinOsincj) z = r cos 9 (1.TQCQ -|. only the velocity components will be given here. 1-11.0. the cylindrical and rectangular coordinates are related by X = r cos ( y = r sin q z —z The velocity expressed in cylindrical components is v_ = rcr 4. z) are shown on Fig. 1-10.

Ak.J}by: a.r. 1-11 and the velocity in spherical components is V. It is sometimes necessary to relate the motion of a point as measured in one reference frame to the motion of the same point as measured in another frame moving with respect to (w.K Ki> .22) = ui = 9k = ek where we have used the fact that k = K is s.20) Spherical coordinates and components are particularly advantageous for central force motion (Chapter 10). 1-12 . First consider two reference frames moving with respect to each other such that one axis. 1-12).t. Fig. is always aligned (Fig.) the first one. frame {/.21) (1.t. say z.r. = a 9k = 9K (1. constant vector.j} w.Review of Newtonian Dynamics Fig. Define the angular velocity and angular acceleration of frame {i. Relative Velocity. = rer + r9ee + r4> sin 9e^ (1.

1-13 Figure 1-13 shows a point A moving in a plane w.t. DQ/Dt = dQ/dt. 1-13: (1.24) Differentiating and applying Eqn. {/.t.t. to two frames which are also moving w. The relation between the two is given by the basic kinematic equation DQ dQ -.t.r.10 Analytical Dynamics Fig. in general. {i. Let -—.r.r. We are now ready to derive the relative velocity equation. but for a vector Q.t.r.t. each other.r.r. J } 21B — Dr_BlDt — velocity of B w.= D t time derivative w.{i.j} dt For a scalar Q.t.j} . {/. = dr/dt — velocity of A w.f= — ^= +u X Q (1. (1.(J. From Fig. J } Vj.23) Dt dt which holds for any vector Q. DQ/Dt y^ dQ^/dt.25) where HA — DriA/Dt = velocity of A w. J] " ' • * d_ = time derivative w.23): DtA Dt Dr LB Dt Dr ^ ^ HA = VA dr w^-F — + w x r W^ + Vr + ^ X 21 = (1.r.

Review of Newtonian Dynamics 11 These results also apply to general 3-D motion provided that the angular velocity is suitably defined.23) and (1. For a full discussion of 3-D kinematics see Ardema. VBX\ f ^ (~^^) r = -xl ij = = VAi -VBJ f { R Vj. Eqns. This theorem states that any displacement of one reference frame relative to another may be replaced by a simple rotation about some line.t. 1-14). the other may then be thought of as a sequence of such rotations. Car B is rounding a curve of radius R with speed VB (Fig. Fig. Example. This is most conveniently done using Euler's Theorem. u is defined as the vector whose direction is the axis of rotation and whose magnitude is the rotation rate. With this definition of w. (1.{'~~§'^] . Newton-Euler Dynamics. The motion of the one frame w. We want the velocity of car A as seen by car B.\yB + —^ 1 J . .j} fixed in car B (the answer is required in this frame) Applying Eqn. Car A is traveling toward car B at speed VA and is distance X from car B at the instant shown. = VAI . (1.r. The cars are modelled as points.r = V-A ~ 'HB ~ ^'^ T. At any instant. 1-14 Introduce reference frames: {/. J } fixed in ground (the data is given in this frame) {i.25) are valid for 3-D motion.25): Wyl = WJ3 + ^ r + ^ ^ Zl VA == VAI VB == VBJ V.

= 0.27) where P = F_.27): m — •V = F • V dy_ dt _. . . 1-15 Now suppose F_ is the resultant of all forces and {i.v_is called the power. Eqn. (1.12 Analytical Dynamics 1. Fig. j'} is an inertial frame. (1.26) mdr_ . 1-15). bmce u = —-. holds: dv dt „ — Taking the scalar product of both sides with v_ and inserting the result in Eqn. Newton's Second Law. Define the work done by F_ during the displacement of m from rg to r^ along C by Uo F-dr (1. Suppose a force F acts on a particle of mass m as it moves along curve C (Fig.1). . this may be written as at U.1 F-vdt= P dt (1. then.3 Work and Energy Definitions.

dV.28) where the kinetic energy of the particle is defined as: T = ^mv'^ (1.^—dy . + ^ .31) and (1.k Then. In rectangular coordinates. dV^ grader) = ^ . Eqn.32) V{x. and writing V = dV so that dV F-dr = -^^dx ox dV dV ..i= fLi JjiQ E-dr= r^i Jxo F.34) dV dV (1.i (1. (1.-~-dz = -dV oy oz (1..30) Generally. (1. F = Fj + Fyj + F. this integration will depend on path C. + ^ ^ .31) Suppose that F_[r) is such that there exists a function V{r) such that Z(r) = -grad V{r) = FJ + Fyj + F.z). comparing Eqns. /•*' . Potential Energy. Recall that the gradient of a scalar function of a vector argument V{r) in rectangular coordinates is . and not just the end points.28) states that the change in kinetic energy from position 0 to position 1 is equal to the work done by the resultant force from 0 to 1. dV^..Review of Newtonian Dynamics /"*! dv .dv_= -m [vi — VQ\ Jto at Jto 2 V / = T i . dr_ = dxi + dyj + dzk .T o = ATo. . r = xi + yj + zk .k and if £ is a function only of position r.32). „ . (1. 1 / 2 2\ 13 / m •— • v_at = m y_.29) Uoi = In words. .y..dx+ rvi Jyo Fydy + fzi Jzo F^dz (1. Eqn. (1.26) gives Uo.

26).14 Therefore. F(r) is called a potential energy function and F_{L) with this property is called a conservative force.36) where K = 6. fVi Analytical Dynamics Uo. for example) 3> m (an earth satellite.sec^) is the universal gravitational constant.35) This shows that now the work done by F_ depends only on the endpoints and not on the path C. 1-16 Newton's law of gravitation gives the force acting on mass m as Z= -^T^er (1. If the two bodies are spherically symmetric they can be regarded as particles for the purpose of determing the gravitational force. Gravitation. If mg (the earth.: (1.2 + y 2 ^ ^2^1/2 we have KrupTn {xi + yj + zk) (1. from Eqn. Because r = rir r_ = xi + yj + zk r = (a. J Fig.673 x 10~-^^ m^/(Kg. for example). 1-16). = f \-dV) 'Vo JVn = . (1. we may take mg as fixed in an inertial frame.Vo) = -AFo.37) . Consider two masses with the only force acting on them being their mutual gravitation (Fig.( F i .

33) are satisfied for this function. 1.^ ^ ^ (1.40 V =0 Fig. In central force motion..(f)).0.38) Thus the gravitational force is conservative with potential energy function given by y. as mentioned earlier. the gradient of V in spherical components is ..39).my^-^-^^^-^^ Fz = —Knipm (^2 + y2 + ^2)3/2 (1. it is usually best to use spherical coordinates.el (a.2+y2 +^2)3/2 15 Fy = -Km. J^7f (1. 1-17 .39) This is verified by observing that Eqns. (1. 1-17). (1.Review of Newtonian Dynamics so that FT — —Krupm X = ±^ III.. Writing V = V{r. The force acting on the particle is F_ = Fxi = —mgi . dV ^ IdV ^ 1 dV ^ gvadV{r) =—-er + -^^eg + —T—-—-e^ or r oO r sm B ocp Thus the gravitational potential function is y =.41) r which could have been obtained directly from Eqn.. For motion over short distances on or near the surface of the earth it is usually sufficient to take the gravitational force as a constant in both magnitude and direction (Fig.

Suppose a number of forces act on m. energy is conserved. some conservative and some not. U IS defined over an interval of motion but T.16 Analytical Dynamics Therefore the gravitational potential energy function is V{x) = mgx because — — = —mg = Fx— ox Energy Equation.i = C/o^f (1-46) (1.43) (1.i = -AFo. Then m = -grad Vi for each conservative force.^ifeo)] + r E ^ r • ^C Using C/o. A£. F. and E are defined at an instant.\ ^-fci) . Remarks: 1.47) .45) and. (1.42) F = Y:F^ + ^F]'^ = J2 (-grad Vi) + ^ Z f 8=1 j=l i j The work done is Uo. The resultant force is (1.!.i + i7o"i (1-44) where V is the sum of all potential energies and C/"*^ is the work done by all nonconservative forces.dr = ^Y. (1.44) may be written A£?o. Let the total mechanical energy be defined by: E =T +V Then Eqn.o.i = ATo. in particular if all forces are conservative and accounted for in F . this becomes ATo.i = f"F.i = 0 that is.

p. the energy equation gives only one piece of information.48) where m = ^ J iTii is the total mass. Let {i. 5. 1-18. The energy equation is a once-integrated form of Newton's Second Law. 1. The center of mass is a position. The energy equation is most useful when a combination of the following factors is present: the problem is of low dimension. Fig. First consider a collection of particles (not necessarily rigid). and E are all scalars.j} be an inertial reference frame and: F_f = sum of all external forces on particle i. T. 4. not accelerations. forces are not needed to be determined. labeled G. U. V. 1-18 .Review of Newtonian Dynamics 17 2. Therefore. The energy equation involves only changes in T and V between two positions. (internal) force exerted by particle j on particle i. whose position vector is given by r = — > 1 m mijii (1.. Fig. thus adding a constant to either one does not change the equation. and energy is conserved. it is a relation among speeds.4 Eulerian Rigid Body Dynamics Kinetics of Particle System. 3.

Mil Sum these equations for all particles and recall from Newton's Third Law that £ j j = —F_jiforall i. Here we consider the 2-D motion of a rigid body.18 Newton's Second Law for particle i is: Analytical Dynamics FHT.r. the accelerations of their centers of mass will be the same (Fig. and y. The angular velocity of a rigid body relative to a given reference frame is just the angular velocity of any body fixed reference frame w. the sum of all external forces equals the total mass times acceleration of the center of mass. each having the same mass. which we take to be in the {i.t. for a system of constant mass.PiJ = rui dh df^ = rrija. A rigid body is a collection of particles such that there exists a reference frame in which all particles have fixed positions in this reference frame. Xi. Let {i'.^ Planar Rigid Body Kinetics. This means that if the same force F. 1-20).j: ST. the other frame.j} plane.^ The quantities d^. are all .j'} be a body fixed frame and let the position of a particle in the rigid body be given by d^ = Xii' + yij' (Fig. including a rigid body. 1-19). Rigid Body. i i j i EF-t m dt^ ma (L49) Therefore. is applied to two dissimilar rigid bodies.

Ki=rnx.^X= ^Aa (1.50) The third degree of freedom comes from relating the rate of change of angular momentum to the sum of the external force moments.51) Here. w.t. Also let the position of A. the origin of {i'.Pyi='^y (1. A special case is Y. It is clear that the location of all the particles of the rigid body are known if the values of x.Review of Newtonian Dynamics 19 Fig. The result is in general quite complicated. the inertial frame {i. The first two degrees of freedom are accounted for by Eqn. the sum of the moments of all external forces about the body fixed point A is (Fig. y and 6 are known.M%M = Uak ^Y.53) .j'}. 1-21) j:M%=Y. Y. (1. in rectangular components: Y. We say that the body has three degrees of freedom.52) (1. 1-20 constants.diXK and (1.j} be given by r^ = xi+yj.r.49).

Point A is either G. for 2-D motion T can be obtained from ^ 1 2 1^ 2 (1. however.54) For most of the approaches to dynamics developed later in this book.51) is only valid if ^(see Fig.55) Use of this formula is seldom convenient. its kinetic energy is.i + AFo. that is the body is regarded as a continuum. 1-21 is the mass moment of inertia of the body about the axis through A parallel to k. If a body is considered as a collection of mass particles. and a = ui = 9 is the angular acceleration of the body. Axis k is an axis of rotational mass symmetry or the body is "thin". and in what follows we present several alternative methods for obtaining T for rigid bodies.2 (1. or moves with constant velocity in the inertial frame. T=-^m. The first representation of IA in Eqn. Concepts of 3-D kinetics will be introduced as needed in future Chapters. As before. by definition. First.20 Analytical Dynamics mass symmetry Fig. Eqn.53) considers the rigid body to be a finite collection of mass particles and the second is the limit as the number of particles tends to infinity.44) applies. Equation (1. an essential step in deriving the equations of motion of a dynamic system is the determination of the system's kinetic energy. Work and Energy for Rigid Body.^. the center of mass. 1-21): 1. (1. repeated here: Aro.i = t/o"? (1. and 2. (1.56) .

k'} with origin at the body's center of mass that moves in such a way that it's axes always remain parallel to those of the inertial frame (thus this frame is not body fixed).yr. (1. like Eqn. A result that is sometimes useful is Koenig's theorem.t. (42 + f.ilr. Zf ^^ z ~T' i^T Koenig's theorem states that the kinetic energy of the body is given by T = im (^' + y' + i') + ^ 5] m. Consider a rigid body moving w.56).j'. j .57) where I A is the moment of inertia about an axis passing through a bodyfixed point that is also fixed in an inertial frame. 1-22.r. as shown on Fig. k}. Note that. respectively. As an alternative for 2-D problems T=-/V (1.Review of Newtonian Dynamics 21 where v is the speed of the center of mass and I is the moment of inertia about an axis passing through the center of mass and parallel to the axis of rotation. this equation divides T into a translational and a rotational part.c. Fig.'^r) in the inertial and the other frame.Zr) and {Cr. 1-22 . y^ ^ y ~r Tj^ .58) where m = 2J"^r. Introduce a frame {i'. an inertial frame {«.is the mass of the body. Then Xf — X -j. Let the coordinates of mass particle r (with mass m^) be {xr.y .l + ul) r r (1.

1-23).'n + v.v) (1. 1.5 Examples Simple Pendulum. Then for the system Eqn.. straight rod and suppose that at some instant the velocities of the ends of the rod are u and u (Fig. V is the total potential energy of the system. One final result is useful for a special case. Suppose a rigid body may be idealized as a homogeneous. Consider a system of constant mass. (1. and U'^'^ is the work done by all the nonconservative forces. A bob of mass m is suspended by an inextensible.54) applies where now T is the sum of all the kinetic energies of all the rigid bodies. weightless cord and moves in the {x. but consisting of a number of rigid bodies. 1-24).. ill rectangular components (Eqns.= 1^0. -x-g smt .59) System of Rigid Bodies.y) plane (Fig.1. 1-23 A method of obtaining T for a rigid body for the special case in which one point of the body remains fixed in an inertial frame is given in Section 11. Then T of the rod is T = -m{u • u + u. 1. not necessarily rigid. thin.22 Analytical Dynamics Fig. We obtain the equation of motion by three methods: (^) J2F.10): ^^Fx = mx 22 Py — "^y eliminate T to get: =^ =^ —Tsm6 = 'mx T cos 9 — mg = my = y cos 9.

— =^ =^ —mgsm9 = mi9 T —mg cos 9 = rriY The first equation provides the equation of motion: 9 + ^sm9 = 0 and the second gives the force T.Review of Newtonian Dynanaics 23 V =0 I = cos 6 I but X = i sin 0 =^ x = icos9-esin9e'^ y = —£ cos 6 = ^ Therefore y = isin 69+ 1008 00"^ 9+^sm9 =0 (1.v_dt = 0 and T does no work. (c) The work-energy relation (Eqn. U = I T_. 1.16): 22 Ft = mil 2_^Fn = m.60) (b) Z^£ = ?"« in normal-tangential components (Eqn.46): Since the chord tension T is perpendicular to v_. The . 1.

method (a) requires the most work and (c) the least. say B. mg.24 Analytical Dynamics only force doing work is weight. y) plane such that one of its points. 1-25 (a) Equation (1. M = external moment = Mk Fig. Fy = my . Note that method (c) does not give force T.F^i cos 9 = l9 .51) relative to point B: Y^MB = iBa =^ M = (I + mf)e [IB = / + ml by the parallel axis theorem] 1 9 -M = 0 (/ + mi^) (b) Equations (1. therefore energy is conserved and Eqn. which may be of interest.50) and (1. Robot Link.. M .47) applies: E = T + V = -m V —mgi cos 6 = constant E = 2-mfee + mgisinee =0 Clearly. and gives a once integrated form of the equation of motion.Fyi sin 9 .51) relative to point A: Fx = niT . and this force is conservative. A rigid body moves in the (a. 1-25). An external moment M = Mk is applied at B (Fig. remains fixed in an inertial frame. (1. We obtain the equation of motion by two methods.

mi^9 = 7e 9 .47). This may be treated as a special case of the robot Unk with the gravitational force providing the moment (Fig. Eqn.57): 1 2 E = T + V = -IB^ — "Tigt cos 6 = constant E = 2-(l + mi'^)ee + mgisindd = 0 mgi sin6' = 0 ' + (7 + mi^) .^ . 1-26). y — —icosO M . with T computed from Eqn. The equation of motion is obtained by three methods.61): A£ = r_x mg M = —mg'^sin^ mgi '+-= + {I mP) sin6» = 0 (b) Conservation of energy. = £ s i n ^ . (a) Equation (1.61) Physical P e n d u l u m .Review of Newtonian Dynamics a.. (1.. M = 0 (/ + m£2) 25 (1. (1.

even if there is slipping. Eqn. there is no velocity at the point of contact with the road. Friction is sufficient to prevent wheel slipping. Car Accelerating U p a Hill. A car of mass m has acceleration a and velocity w up an incline of angle 9 (Fig.60). We want to find the power. is the special case of / = 0. The kinetic and potential energies are found from Eqns. (1.26 Analytical Dynamics (c) Conservation of energy with T computed from Eqn. The wind resistance is D^ and each of the four wheels has a moment of inertia / and a radius r. Also. delivered by the engine to the wheels. Consequently. (1. Since the wheels roll without slipping.jDv dt . (1. 1-28): T V 1 _2 = -mv^ + = mg sin 9x 4-li- Ayfv-'^ The work done by force D is U^ = f D-Edt = . 1-27). N_i and N_2 do no work because they are normal to any possible velocity.27). Pg. from Eqn.56) and (1. F_i and F_2 do no work. (1.42) as (see Fig.56): E = T + V = -I u'^ +-m iP' -mgicos 9 = constant E = \{I + •mi^)6'^ — mgi cos 9 = constant The simple pendulum.

(2) to spin-up the wheels.6 Motivation for Analytical Dynamics Lessons from Newtonian Dynamics. This chapter has revealed the following: 1. (1. 1-28 Now differentiate the energy equation. They may be conservative or nonconservative (the former can be accounted for by potential energy functions).44).! + AFo. and substitute the above relations: t/o^5 = ATo. and (4) to overcome air resistance. 1. . And finally. There are differences in forces.Review of Newtonian Dynamics 27 v=rco Fig. Forces may be classified as external or internal (the latter cancel out in a system of particles). Eqn.! ['F!""- vdt = Ti-To + Vi-Vo Jto [ F^''-v_dt = T-To Jto F'^'^v = t + V Zi + V -VQ FgV — Dv = 2-mv a + 4-7—^—h mgvsmtf Zi I Pg = FeV = fnv a + 4—^—h mg sin 9v + Dv This shows that the power produced by the engine is used in four separate ways: (1) to accelerate the car. (3) to gain altitude. some forces do work while others do not.

not accelerations. Relativity theory was developed to account for these discrepancies. Vectors will be underlined here and throughout most of the book. 2 3 4 5 . then the equation contains products as well as moments of inertia. there are no inertial reference frames. free-body diagrams. In relativity theory. for example. Note that the origin of the body-fixed frame need not be "in" the body. any other frame. 3. it gives a relation of speeds. If these conditions are not met. Components along the coordinate axes are not always the best parameters to describe the motion. Obtain scalar equations of motion invariant to coordinate transformation in the minimum number of variables. 2.28 Analytical Dynamics 2. and thus possibly doesn't give the information desired. and the values of mass and time are different for different observers. These observations motivate the search for new approachs to dynamics. for a complete discussion of Eulerian dynamics of a rigid body.t. and is now the most accurate description of dynamics. (ii) It is cumbersome when forces do work and are nonconservative. Newton-Euler Dynamics) that all body-fixed frames have the same angular velocity w. Newton-Euler Dynamics. and coordinate systems are not needed. Purposes of Analytical Dynamics. See Ardema. It may be shown (Ardema. (ii) It gives an integral of the motion. The main goals of the rest of this book are to: 1. the physical world is a four dimensional. the exception is unit vectors. 3. The energy method is simple when appropriate. non-Eulclidean space.r. 9 is the best parameter for the simple pendulum. experiments showed that in certain situations Newton's Laws were significantly inaccurate. Eliminate constraint forces and treat conservative forces via potential energy functions. The disadvantages are: (i) It gives only one equation. that is. The advantages are: (i) Unit vectors. Notes 1 About 100 years ago. which will get "hats". Obtain solutions of the equations of motion.

Find the acceleration of the point when a. remains a constant. Problem 1/3 Fig. 1/4. = 2 ft. Grain is being discharged from a nozzle into a vertical shute with an initial horizontal velocity V_Q . 1/2 Problem 1/2 1/3. A point moves at a constant speed of 5 ft/s along a path given by y = 10e~^^. where ^4 is a constant. Find the normal and tangential components of acceleration. x. k a constant. where x and y are in ft. A point moves at constant speed v along a curve defined by r = A6. such that the horizontal component of velocity. 1/3 A point travels along a parabola y = kx"^. Determine the range of values of VQ for which the grain will enter the shute.65 ft 1 'I ft 1ft ft • t • "1 Problem 1/4 1/5. Determine the acceleration of the point as a function of position. Determine the range of values of h for which the water enters the trench in the ground. Fig.Review of Newtonian Dynamics 29 PROBLEMS 1/1. 1/2. . 0. A nozzle is located at a height h above the ground and discharges water at a speed VQ = 25 m/s at an angle of 55° with the vertical.

The water is discharged with a speed oi VQ = 8 ft/s at an angle of ^ = 40° to the vertical. A rotating water sprinkler is positioned at point ^ on a lawn inclined at an angle a = 10° relative to the horizontal. Determine the horizontal distances dc and ds where the water lands. Problem 1/6 Problem 1/7 .30 Analytical Dynamics Problem 1/5 1/6.

If the plane's speed is u = 800 km/h. Determine the maximum height h at which the ball can strike the wall and the corresponding angle 7. Determine the velocity and acceleration of collar C as a function of 0 for the following cases: (i) ^ = a. 31 A ball is thrown with velocity HQ against a vertical wall a distance d away. what must be the rate of rotation of the airplane 7 to obtain this condition at the top of its loop? The speed of a car is increasing at a constant rate from 60 mi/h to 75 mi/h over a distance of 600 ft along a curve of 800 ft radius. Express the velocity and acceleration of the point in tangential-normal components.87°.25 ft. (ii) 0 = 0 and 0 = a. 1/11. and the radius of curvature of its path is 6. It was observed that the radius of curvature of the stream of water as it left the nozzle was 35 ft. Determine the radius of curvature of the stream both as it leaves the nozzle and at its maximum height. The speed of the point is decreasing at the rate of 2 ft/s^. 1/12. The velocity of a point at a certain instant is v_ = 2i + Aj ft/s. and pin P slides in the slot attached to collar C. What is the magnitude of the total acceleration of the car after it has traveled 400 ft along the turn? Consider the situation of Problem 1/5. is the ball ascending or descending when it strikes the wall? What minimum speed VQ is needed to strike the wall at all? A condition of "weightlessness" may be obtained by an airplane flying a curved path in the vertical plane as shown. for 0 = 36. 1/14. and the radius of curvature of the stream when it reaches its maximum height. 1/13. Link OP rotates about O. and (9 = 0. Consider again the situation of Problem 1/5. 1/9. 1/8. In Problem 1/7. and g.Review of Newtonian Dynamics 1/7. 1/10. in terms of VQ. d. . Find the speed VQ with which the water left the nozzle.

. the magnitude of the total acceleration at P i . determine the radius of curvature of the path at A. the tangential component of acceleration is 3 in/s^. and the magnitude of the total acceleration at P2.5 in. The radius of pulley 1 is 1. 0. f. At Pi. and 9. Problem 1/16 1/17. the magnitude of the total acceleration of the airplane is 3g. At this instant. The shape of the stationary cam is that of a limacon. Tape is being transferred from drum A to drum B via two pulleys.32 Analytical Dynamics Problem 1/14 Problem 1/15 1/15. b > c. If the airspeed is 800 mph and is increasing at the rate of 20 mph per second. a point on pulley 1. the measurements of a rocket are 1/18. Determine the magnitude of the total acceleration as a function of 9 if the slotted arm rotates w^ith a constant angular rate co = 0 in the counter clockwise direction. compute the speed of the tape. The radar is in the vertical plane of the rocket's flight path. 1/16. defined by r = b — ccosO. a point on pulley 2. A radar used to track rocket launches is capable of measuring r.0 in and that of pulley 2 is 0. At a certain time. At the bottom A of a vertical inside loop. the normal component of acceleration is 4 in/s^ and at P2.

0072 rad/s^. using radial-transverse components. 9 = oj = constant.9i^ and 0 = ^7r(4i — 3t^). 1/21. and 6 = -0. determine the expression for the magnitude of the acceleration of a fluid particle when r = R. 1/20. The rod rotates according to ^ = 2i^. Also. r = 1600 m/s.20t^.02 rad/s. If the pump turns at a constant rate 9 = uj. 1/23. ^ = 0. measurements give r = 30. The path of fluid particles in a certain centrifugal pump is closely approximated by r = roe"^ where ro and n are constants. 1/24. and 9 = 0. determine the magnitude of the acceleration of P as a function of 9. Answer the same questions. but with r = 1. For the instant when 9 = 60°. determine r. 9 and 9. f.000 ft. show that the magnitudes of the velocity and acceleration of P are constant during this time interval. * What direction is the rocket heading relative to the radar at this time? What is the radius of curvature of its path? 1/19. . Slotted arm OA oscillates about O and drives crank P via the pin at P. Problem 1/21 1/22.25t^ — 0. r = 70 ft/s^. The pin A at the end of the piston of the hydraulic cylinder has a constant speed 3 m/s in the direction shown. Determine the magnitudes of the velocity and the acceleration of the rocket at this instant. where r = OA. Determine the velocity and total acceleration of the collar when t = Is. When 9 = 60°. Consider the same situation as in Problem 1/19. A vertically ascending rocket is tracked by radar as shown. with r in meters and t in seconds.000 m. with 9 in radius. For an interval of time. During this time. A collar A shdes on a thin rod OB such that r = 60i^ .Review of Newtonian Dynamics 33 r = 35.

= 120 in/s and at the same instant is at the center of the disk. find the acceleration of B relative to A. Find the velocity oiB relative to the occupants of ^ at the point of closest approach. The curves both have radius R = 100m and their point of closest approach is x = 30m. . the disk is rotating with an angular speed of w = 15 rad/s and the speed is increasing at a rate of 20 rad/s^. For the same conditions of Problem 1/25. 1/28. At a certain instant. 1/26. Two cars. 1/27. For the same conditions of Problem 1/25.1 /r/v " Problem 1/24 ^(i/ Problem 1/23 1/25. The slider moves in the slot in the disk at the constant rate i. are traveling on curves with constant equal speeds of 72 km/hr. Obtain the acceleration and velocity of the slider at this instant. labeled A and B.34 Analytical Dynamics Z/^/" //L/y^— . find the acceleration of B relative to A if ^ is speeding up at the rate of 3 m/s^ and B is slowing down at the rate of 6 m/s^.

1/31. Same as Problem 1/31.6 rad/s. At a certain instant.—' * 6 in Problem 1/29 1/29. Plate B rotates about point A.4 rad/s. When in the position shown. except that the rotation of the bracket is speeding up at a rate of 0. and the welding bracket with tips D and E moves in a cylinder C attached to B.Review of Newtonian Dynamics 35 D B [C A ^t 2 in 2 in . .5 mm B 10 mm D >=I 10 mm ' Problem 1/31 1/32. constant rate of 3 in/s and B is rotating counter clockwise about ^ at a constant rate of 1. For the same situation as in Problem 1/29. 1/30. except that the rod is slowing down at the rate of 2 mm/s^. bracket DE is moving to the right with respect to plate B at a.3 rad/s^. Shown is an automated welding device. Same as Problem 1/31. 1/33. Bracket ACD is rotating clockwise about A at the constant rate of 2. determine the velocity and acceleration of tip D. Find the velocity and acceleration of point B. 7. Determine the velocity and acceleration of tip E at that instant. rod BE is moving to the right relative to the bracket at the constant rate of 15 mm/s.

SOON \'-\* IXj 0. Prove Eqn. There is a constant force of 300 N in the cable and the spring attached to the block has stiffness 80 N/m. Prove Koenig's theorem. 1/38.36 1/34.57) for the case of 2-D motion.57).233 m.. (1. 1/39. A 50 kg block shdes without friction as shown.59). If the block is released from rest at a position A in which the spring is stretched by amount xi = 0. Problem 1/40 1/41. Prove Eqn. (1.58) reduces to Eqn.j Problem 1/41 .30.58). Analytical Dynamics Find the velocity and acceleration of point E for the situation in Problem 1/31. A 50 kg cart slides down an incline from ^ to B as shown. Eqn. (1. 1/40. (1. What is the speed of the cart at the bottom at B if it starts at the top at A with a speed of 4 m/s? The coefficient of kinetic friction is 0. Show that Eqn. (1.12 m . 1/36. what is the speed when the block reaches position B. 1/37. 1/35. (1. Prove Eqn.56).

The identical links are released simultaneously from rest at ^ = 30° and rotate in the vertical plane.6 m—*j Problem 1/44 1/45. Problem 1/43 The small 4 kg collar is released from rest at A and slides down the circular rod in the vertical plane. 30\ 400 ft .Review of Newtonian 1/42. Find the speed of the collar as it reaches the bottom at B and the maximum compression of the spring. What is the speed of the collar at Bl Neglect friction. at which time the engine is shut off. How far does the car roll up the hill before stopping? Neglect all friction and air resistance. Find the speed of each 2 lb mass when Q = 90°. ^ B — 0.4 m. The attached spring has stiffness 200 N/m and an unstretched length of 0. Ignore the mass of the links and model the masses as particles. !A BV- k = 20 kN/m Problem 1/42 1/43. The car starts from rest at A and the engine exerts a constant force in the direction of travel of 1000 lb until position B is reached. Dynamics 37 A 4000 lb car travels up a hill as shown. The spring is unstretched when Q = 90°. . The small 3 kg collar is released from rest at A and slides in the vertical plane to B. Neglect friction. 1/44. ^ .

If the frame is released from rest ai 9 = 90°. Neglect the masses of the links and all friction. determine the speed of the mass when 9 = 135° is passed. The device shown is released from rest with 0 = 180° and moves in the vertical plane. n 300 mm 200 mm Problem 1/46 1/47. 500 mm Problem 1/41 . The spring has stiffness 900 N/m and is just touching the underside of the collar when 9 = 180°. Shown is a frame of negligible weight and friction that rotates in the vertical plane and carries a 3 kg mass. Determine the angle 9 when the spring reaches maximum compression. The spring is unstretched when 9 = 90°.38 Analytical Dynamics Problem 1/45 1/46.

If the collar is 1/51. rolls down the track to B. Neglect friction and air resistance. A 600 g collar slides without friction on a horizontal semicircular rod ABC of radius 200 mm and is attached to a spring of spring constant 135 N/m and undeformed length 250 mm. The spring constant is 1. Determine the smallest deflection of the spring for which the pellet will remain in contact with the circular loop at all times.5 lb/in and is undeformed when the collar is at A. . determine (a) the force exerted by his or her seat on a 160 lb rider at both B and D. and (b) the minimum value of the radius of curvature of E if the car is not to leave the track at that point. Neglect all friction and air resistance. Problem 1/48 1/49. The 0. Problem 1/50 A 3 lb collar is attached to a spring and slides without friction on a circular hoop in a horizontal plane. and then moves over the hump at E. transits a circular loop of 40 ft diameter. Dynamics 39 A roller coaster car starts from rest at A. find the speeds of the collar as it passes through points B and A. li h = 60 ft. k = 3 lb/in Problem 1/49 1/50.5 lb pellet is pushed against the spring at A and released from rest. If the collar is released at C with speed 6 ft/s.Review of Newtonian 1/48.

dz Show that the force F_={xi + yj + zk){x^ + y^ + z^) is conservative by applying the results of Problem 1/52. The collar is initially held in position A against a spring of spring constant 2. As the rod is rotating at angular speed 12 rad/s the cord is cut. A 1/2 lb collar slides without friction on a horizontal rod which rotates about a vertical shaft. The spring is attached to the collar and the rod. releasing the collar to slide along the rod. Find the angular speed of the rod and the radial and transverse components of the velocity of the collar as the rod passes postion B. 1/54. Also find the potential energy function V{x^ y. dx dz dy dF. dx OF. z) associated with F_.40 Analytical Dynamics X = 240 mm y = 300 mm z= 180 mm r = 200 mm Problem 1/51 released from rest at A.5 lb/ft and unstretched length 9 in. dR. .. Prove that a force is conservative if and only if the following relations are satisfied: OF. what are the speeds of the collar at B and C? 1/52. dy 1/53. Also find the maximum distance from the vertical shaft that the collar will reach..

are mounted on the same shaft as shown.Review of Newtonian Dynamics 41 3 in 12 m A B 3=» s Problem 1/54 1/55. while the two disks roll on the xy-plane without slipping. Find the kinetic energy of the sphere for general motion. The center of the disk moves with constant velocity v. Problem 1/55 1/56. each of mass M and radius r. A homogeneous solid sphere of mass M and radius R is fixed at a point O on its surface by a ball joint. 1/58. The shaft turns about the ^r-axis. Prove that the ratio of the kinetic energies of the two disks is 1/57. Two uniform circular disks. Determine the kinetic energy of the uniform circular disk of mass M at the instant shown. 6{R + df + r^ 6iR-df+r^ . Find the kinetic energy of a homogeneous solid disk of mass m and radius r that rolls without slipping along a straight line.

8 in. 1/60. Find the kinetic energy of the disk when the line of contact turns around the cylinder at 10 cycles per second. with point D remaining at the origin. The moment of inertia of the disk and arm about axis OA is / and the total mass is M.y-plane without slipping. The disk rolls inside a cylinder whose radius is 4. Determine the kinetic energy of the disk if shaft CD rotates about the z-axis with constant angular speed n. A uniform circular disk of radius a and mass M is mounted on a weightless shaft CD of length h. . The disk rolls on the a. The shaft is normal to the disk at its center C.42 Analytical Dynamics Problem 1/57 Problem 1/58 1/59. with the center of gravity at G. A disk with arm OA is attached to a socket joint at O.

Problem 1/62 A particle of mass m slides along one radius of a circular platform of mass M. 1/64.-axis. At the instant shown.Review of Newtonian Dynamics 43 ooT 4 in. — • • Problem 1/59 1/61. What is the kinetic energy of the system at the lowest position? What is the velocity of the bob at the lowest position? A particle of mass m is attracted toward the origin by a force with magnitude {mK)/r'^ where K is a constant and r is the distance 1/63. Problem 1/60 A homogeneous solid right circular cone rolls on a plane without slipping.8 in. Determine the kinetic energy and the angular momentum of the system about point O. N v\ ^ l\\ 1 V ''1 1 L-—4. The line of contact turns at constant angular speed Q. The pendulum is released from rest at position A as shown. A pendulum consists of a uniform rod of mass m and a bob of mass 2m. Find the kinetic energy of the cone. the platform has an angular velocity w and the particle has a velocity v relative to the platform. about the 2. Problem 1/61 1/62. .

consisting of a massless rod and a bob of mass m. taking the horizontal plane passing through the support as the zero potential level. c. is initially held at rest in the horizontal plane. A spherical pendulum of mass m and length a oscillates between levels 6 and c. 6. Find the expression for the total energy of the system in terms of a. A spherical pendulum. located below the support. The particle is constrained to move in a frictionless tube which lies along the space curve given by z = h0 1 1 > in cyUndrical coordinates If the particle was at rest when z — 10.44 Analytical Vertical 60° Dynamics V 1 / / m / U 2m Problem 1/63 Problem 1/64 between the particle and the center of attraction. what is the velocity of the particle at ^ = 0? 1/65. A hori- . and g. Problem 1/65 1/66. m.

Review of Newtonian Dynamics 45 zontal velocity VQ is imparted to the bob normal to the rod. The tube is bent into a circular ring with the lowest point left open as shown. will the particle drop through the opening? Problem 1/67 . and simultaneously the particle is released from rest (relative to the tube) at ^ = -K/2. In the resulting motion. what is the angle between the rod and the horizontal plane when the bob is at its lowest position? 1/67. A particle of mass m is placed inside a frictionless tube of negligible mass. In the subsequent motion. The ring is given an initial angular velocity w about the vertical axis passing through the diameter containing the opening.

1) m„i"(i) = E £ " ( ^ . In the "strictly Newtonian" problem. Then Eqn. For most 47 . t) or mrx'it) = ^ £ ' ' ( x S ••. (LI) gives •mix}{t) = E z H ^ ^ ••. unbounded forces are allowed provided A force which is unbounded but measurable is called an impulsive force. ••. ••. x". Note that forces are not allowed to be functions of the particles' accelerations. Let the resultant forces on the particles be bounded functions of the particles' positions.) Let the position of particle r in an inertial reference frame be denoted x!'{t).^ In the "Newtonian" problem.2) If none of the forces depend explicitly on time.1 N e w t o n ' s Second Law Vector Form. s". fnn. x^. velocities. £"> ^ ^ •• ^"> t) (2. • • i " . as shown on Fig. 2-1. all forces are bounded.(Occasionally we will call a system of particles a dynamic system. r = 1. x}.2. we say the system is autonomous.Chapter 2 Motion and Constraints 2. n (2. x\ ••. and time. Consider a system of n mass particles of masses mi. •012-1 • • •. t).

63} are fixed in the inertial frame. For example. it may be of interest to determine the range of orbits accessible by an existing launch vehicle. Then. and hence the size of the vehicle. The second way is to determine the motion when the forces are given (problem of the first kind). 62. if {ei. Component Form. The exception is Chapter 13 where impulsive forces will be considered. One way (problem of the second kind) is to determine the forces acting on a system when the motion of the system is given. Thus it is characteristic that the equations of motion of a particle system give the accelerations of the particles in terms of their positions.63}. when designing a space launch system the motion is known (transition from earth surface to orbit location and speed) and Newton's Second Law can be used to predict the propulsive forces required. and time.48 Analytical Dynamics (inertial) Fig. This situation typically arises in the performance estimation of an existing system. For example. gf{t) = x[{t)ei + x^(i)e2 + xlit)e3 r ( i ) = x\{t)ei + xl{t)e2 + xl{t)e3 Label the components of Ui=x\ . velocities. although all the results obtained apply equally to either type of problem. This is typically the situation at the design stage. 2-1 of this book we restrict our attention to the strictly Newtonian problem. Recall that there are two ways Newton's Second Law can be used.3) follows . U4 = xl . 62. U^=x\. U2=x\ (2. In this book we approach dynamics as a problem of the first kind. Now introduce linearly independent unit vectors {ei.

and ma are all the mass of the first particle and Fi.U2. or "msUs = ^ F s ( u i . iiN. UN.Ul. F2.Motion and Constraints u^ = X2 . UN can be thought of as forming a vector in a subset of lE^. ••. UN. and F3 are the three components of the resultant force acting on the first particle. UN. ••. m2. U2 \ {n.6) Note the interpretations of rrig and Fg] for example. As the motion of the system proceeds. ••.ui. UN. 2. ". E v e n t S p a c e . t) ••. UN. t) miil2 = E-P2('^l5«2.5) ?Tl2«4 = E . ". ?7inWiV = E ^ 3 ' ( « l ) « 2 .U2. By the correspondence between n-tuples and vectors in Euclidean spaces.Fl(ui.2 Motion Representation C o n f i g u r a t i o n S p a c e . a p a t h is generated in this space called a C trajectory. UN. miilz = J2F3{ui.7) We call C the configuration space. •-.4) Then we can write the Second Law as miili ^Y. N (2. t) ••. ••.U2. the Ndimensional Euclidean space: U2 u = \ UN J eC cIE'' ?N (2.^ f («l:'"2.8) V i J .U2. ••. The combination of the configuration components and time is called an event. UN.Ul. UN. t) s = 1. ••. UN. -ui. an event is a vector in the event space E: ( u.ui.U2. UN = x^ 'i N = 3n 49 (2.U2. ••.U2. ill.Ui. the components of displacement. UN. (2. m i . ••.t) UN eEc lE^^'^ (2. t) ". UN. t) . ••.

The combination of states and time gives a point in state-time space ( ^1 \ (M. The combination of the configuration components and the components of velocity defines a point in state space S: / "1 U2 \ {U. z =0 . State-Time Space. suppose the motion of a single particle is constrained to be on a surface..50 Paths in this space are called E trajectories. t) = U\ U2 ETC IE^^+' (2. the third.9) U2 \ UN J Paths in this space are S trajectories. 2-2. y) plane.3 Holonomic Constraints Introduction. for which the constraint is: f{x. say z. A special case is motion in the {x.10) UN V t J 2. Analytical Dynamics State Space.u) UN Ml eS clE- •2N (2. The motion of a particle system is frequently subject to constraints. As an example. Note that now only two of the coordinates are independent.z) = z = 0 ==^ i = 0. M. as shov/n on Fig. is determined by the constraint. y.

Consider a system of n particles.y. •••. .x. 2-2 Equations (2.0. such a system has a N = 3n dimensional configuration space C. UN) = 0 (2.x.y) 51 Fig. A holonomic constraint on the motion of the particles is one that can be expressed in the form f{Ui. UN. for example: f{x.0.y.y.y.U2.12) then it is scleronomic.t) =0 (2.z) = 0.Motion and Constraints f(x. z = (t)(x. otherwise it is rheonomic. for this case: mx = Fx{x.z.0.0. •••. If all constraints are holonomic we say the system is holonomic. As a special case.y.0.U2. Of course constraints also may be prescribed functions of time. the C trajectories must lie on this surface. and if all holonomic constraints are scleronomic the system is scleronomic.y.0.t) =0 Definitions.t) Fy{x. The constraint equation is an iV — 1 dimensional surface in the configuration space C C lE^.x.11) Otherwise the constraint is nonholonomic.6) become.t) my = 0 = These are the equations of planar motion as expected. if a holonomic constraint can be expressed as f(ui.y.y.t) Fz{x. Of course there may be several such constraints.

2-5). z) = 0. if the constraints are scleronomic (Fig. for the case of two components. respectively. 2-6). x^ = x\ei + a. then there are two constraints in . If the surface is given by f{x. the E trajectories lie on a right cylindrical surface. y. U2 = X2 . 2-5 Fig. U^ = x'2 1 «3 = xl UQ = X^ Ui — x\. 2-6 Next consider two particles constrained to move on a single surface (Fig.2e2 + 2:363 Relabeling to put in component form: Ui= x\ . In E space. The position vectors of the particles resolved into components are: x^ = x\ei + x\e2 + x\e2.52 Analytical Dynamics f 1 = U3 = 0 C space C trajectory h=o Fig. f =0 E trajectory Fig. 2-4 Figures 2-3 and 2-4 show tlie cases of a single particle subject to one and two constraints.

Let the direction cosines of C. 2-8). fix a body-fixed reference frame with axes (. The number of degrees of freedom of the system is then DOF = N-L>0 (2. but is not necessarily an actual motion because actual motions also obey Newton's Laws.Motion and Constraints configuration space: / i = f{ui.U5. For a single particle. 1. First. If there are two holonomic constraints. 77. but the two constraints have reduced this number to four. Degrees of PVeedom.U2. Given a system of n = A^/3 particles suppose there are L independent constraints. with respect to a non-body-fixed frame with axes ar. v relative to x. y. /2 = f{u4. DOF = 0 and the particle is fixed. 2-7 The situation for a rigid body is more difficult. z (Fig. Every C trajectory satisfying all constraints is a possible motion. DOF = 1 and it takes one (motion on a line) and if there are three holonomic constraints. DOF = 2 and it takes two (motion on a surface). y. 2-7).13) li N = L the system is fixed in space if all constraints are scleronomic and moves with prescribed motion if at least one is rheonomic. If there is one holonomic constraint.U6) =0 53 The unconstrained particles had a total of six independent components. We establish the number of DOF of a rigid body in 3-D unconstrained motion in two different ways. y.U3) = 0 . v at the body's center of mass and let the coordinates of the origin of this frame be S. ??. z be £1 mi £2 rn2 ni ^2 k rnz na . if there are no constraints DOF = N — L = 3 — 0 = 3 and it takes three independent parameters to specify position in configuration space (Fig. L=o •r -^ • f ( DOF = 2 kA A X y"^ L=l ^ M( DOF = 1 ^ 7V^ 7V'^L = 3 DOF = 0 DOF = 3 -— Fig.

t] = 0.Thus the location of all the points of the rigid body are specified by {x. ii is the cosine of the angle between ^ and x). and that each additional particle takes 3 more (Fig.15) . the direction cosines are functions of three independent angles. We now derive local conditions. Thus if the rigid body has n particles.r. It is clear that the first three particles take 3 constraints. Alternatively. 'z. ^2. the total number of constraints is 3 + 3(n — 3). conditions on small displacements.11) or (2.12) is for arbitrarily large displacements. 02. UN{a). But since a particle without constraints has 3 DOF. 2-9). 03. The coordinates of the r^^ particle of the rigid body are then Xr =X + £i(r + hVr + Zr —'Z + ni(r h^r yr = y + mi(r + m^'qr + mzUr + "2'7r + n^Uj- Due to the orthogonality of the direction cosine matrix. 2. The constraint as specified by Eqn.[3 + 3(n .54 Analytical Dynamics Fig. r = 1. that is. y. L (2. 2-9 (For example. a: N El s= dfr du da dt da r = l. the DOF for the rigid body is DOF = 3n . Suppose we have L holonomic constraints and let Us = Us{a) and t = t{a) where a is a parameter: fr[ui{a). and therefore the body has 6 DOF. (2.14) Differentiating w. U2{a). say 9i. 2.t. we may view the rigid body as a system of constrained particles.^3) relative to the other frame. 2-8 Fig. (2.3)] = 6 Infinitesimal Displacements. 9i. • .

r = l. ^ In differential form N 55 r = l. 2.19) The first of these is a local condition on velocities and displacements and the latter is a local condition that small displacements must remain in the tangent plane of the constraint..17) For the special case of all constraints scleronomic. L (2.L (2. 2. L (2..18) dfr Y^ ^ dus = Q\ N r = l. 2-10) of the constraint at u*.2.. (2. 2. . L (2.16) Y^^dus s=l dus + ^dt dt = 0- r = l. 2-10 .Motion and Constraints An important special case is a = t: ^ dfr dfr E a ^ « ^ + -dt = 0 . then infinitesimal displacements dug satisfying E N duc = 0 are in the tangent plane (Fig. If u* is a position on the constraint. ••. constraint tangent plane f =0 Fig. Eqns.16) and (2. -.17) become N J2w^^s dus s=l = 0.

A nonholonomic constraint is one that is not holonomic. we cannot use this to get a relation between finite displacements. 2. Such a constraint may depend explicitly on t (rheonomic).56 Analytical Dynamics 2.20) it is called a configuration constraint. t) <0 (2. ••. U2i •-. This can happen in several ways and we discuss two.U2. 2-11 Equality Constraints.z) = z . UN. These are diflferential relations among the ui. t of the form N y ^ Arsdug + Ardt = 0 .h > 0 Fig. 2-11).4 Nonholonomic Constraints Configuration Constraints. UM. Recall that starting with a holonomic constraint fr{ui. An example of a configuration constraint is the requirement that an object must stay on or above a plane surface (Fig.21) that are not integrable. If a constraint can be reduced to an inequality in the configuration space. ••.t) = 0 . L (2. f(x. s=l r = 1. f{ui. UN. or not (scleronomic). •••. we diff'erentiated to get ^ dfr dfr .y. U2. that is.

Holonomic constraints usually come in integrated form. and that such constraints are restricted to those in which this dependence is linear. A constraint is catastatic if ^ r = 0 and acatastatic otherwise. is possible if and only if the system is catastatic. Equation (2.21) this cannot be done. If the Pfaffian form is an exact differential.22) implies that the condition of static equilibrium. ••. ••. We make an additional distinction. r = 1. s = 1. (2. we see that one way nonholonomic constraints can occur is as constraints on the velocity components. Thus the DOF of a system is the number of velocity components that can be given arbitrary values. but this is not necessary. then the constraint is integrable. we can go back to the finite form. Note that the Ars may be functions of t in a catastatic constraint. several example of this will be analyzed later.22) Each constraint may be either holonomic (integrable) or nonholonomic. if all constraints are catastatic. n. we say the system is catastatic.5 Catastatic Constraints Definitions. 2. However.16). Prom Eqn. Pfaffian Form. The differential form (whether integrable or not) of a constraint. Each constraint may be either scleronomic {Ars 7^ A^s {t) and A^ = 0) or rheonomic. Ug = 0. sometimes they come in Pfaffian form. L (2.21). . The most common situations in which such constraints arise involve bodies rolling on other bodies without slipping. 2.Motion and Constraints 57 Integrating. (2. We must then be able to distinguish between holonomic and nonholonomic. is called the Pfaffian form and it is the most general type of constraint we will consider in this book. (2. Eqn. Consider a system of constraints in Pfaffian form N ^ s=l Arsdus + Ardt = 0 . With a nonholonomic constraint of the type of Eqn. The following theorem (without proof) is a very general result.6 Determination of Holonomic Constraints Remarks.

VM) = 0 ''ydya dyp) ^ \dy^ dya J ^^^^^ +^"fl^-|^l=0.y2. of which (M — 1)(M — 2)/2 are independent. velocity components.23) is Ai{yi.25) and M(M-l)(M-2) _ 6 ~ ' (M-l)(M-2) _ 2 " Thus for the equation to be integrable the only requirement is: ^. y) plane such that the slope of it's path is proportional to time. dy/dx = Kt. (2. (1^ .y3)c?y3 = 0 (2. or time.58 Analytical Dynamics Theorem.\dyi .y2.y3)dyi +^2(yi. ••. this constraint is Kt dx .yM)dys = 0 (2. In Pfaffian form. For three variables.-. suppose a particle moves in the (x. Suppose an equation of independent variables yi.y2. that the equations -.7 = l. M = 3.26) In the application of this theorem to the constraints on a dynamical system.y2. the yr may be displacement components./3. Eqn. As an example.2.dy = 0 .y2.M be simultaneously and identically satisfied.1^) + A.23) Then it is necessary and sufficient for the existence of an integral of this equation of the form f{yi.-'i IjM is given in differential form as: M J2 Asiyi. There are M{M — 1)(M — 2)/6 such equations.y2. \^ dy^ dyry ) «.dyz J + A.y3)cJ?/2 + ^3(yi. C^-'^)= (1^ 1^) Koys dy2J \dy2 oyi J 0(2.

7 Accessibility of Configuration Space Definition. This means for example.8 Examples Example. Nonholonomic equality constraints do not reduce the DSAC. Ai = -z . If there are n particles and L holonomic constraints this number is N — L. consequently in general the DSAC is given by DSAC = N-L' (2.y2) This may be integrated under mild assumptions on the functions Ai{-) and A2{-). dy — z dx = G Relabeling: yi=x . thus we take yi = x. 2. if not analytically. and y3 = ^ so that Ai = Ky^. y2 = y . 2.27) dyi A2{yi. then numerically.28) where L' is the number of holonomic constraints (L' < L)? The fact that nonholonomic constraints do not reduce the DSAC will be shown by some of the following examples. in this case the constraint may be written p =-44^ (2. yz = z A3 = 0 . We say that there is an N—L fold 00 of motion or. that any time-independent constraint on the position of a particle in 2-D motion is holonomic. (2.Motion and Constraints 59 This is a differential constraint of three variables. 2/2 = y. alternatively. and ^3 = 0. so that the constraint is nonholonomic. A2 = l . One consequence of the theorem is that any differential relationship between only two variables is always integrable. A2 = —1. Recall that a holonomic constraint reduces the number of quantities required to define a point in configuration space. Indeed. that the dimensionality of the space of accessible configurations (DSAC) isN -L.26) equals K. The lefthand side of Eqn. Suppose a constraint on the motion of a particle is ^r = dy/dx. In Pfaffian form. where A^ = 3n.

Two particles pi and p2 moving in the {x. = a. dy — zdx = f'dx — f'dx = 0 at a.y2). Consider any path such that (Fig. y = yi and z = zi Example. Let the coordinates of the two particles be {xi. We now show this directly.yi.zi) that satisfies the constraint.26) is equal to one so that the constraint is nonholonomic and therefore the DSAC is not reduced.xif + {y2 .aadt = 0 (dx2 dxi\ (dy2 dyi\ da .i . a{t) G C^. f{xi) = yi z = dx f'ixi) = zi /(0) = / ' ( 0 ) .dxi) + {y2 .0 The following shows that the constraint is satisfied and that the endpoint is reached.Then the constraint is {X2 . (2. We need to show that there is at least one path from the origin to any arbitrary fixed point {xi.xi){dx2 . {x2 .60 Analytical Dynamics Fig.y2) and {x2. respectively.yif = a'^ In Pfaffian and velocity forms.yi){dy2 . y) plane are connected by a light rod of length a which changes as a prescribed function of time.dyi) . 2-12): y = fix) . 2-12 The left-hand-side of Eqn.

(i) the edge remains in contact with the plane.)— or. 2-13). and (ii) the no slipping condition. y) coordinates of the contact point and three angles usually taken as Euler's angles.Motion and Constraints 61 It is clear that the constraint is holonomic rheonomic and that DOF = DSAC = 3. =o {x2 . There are two constraints . A knife-edged disk rolls without slipping on a horizontal plane (Fig.Disk Rolling on Plane. Thus DOF = 2 and DSAC = 3.xi)^ + {y2 .dxx) + (y2 .yif I sdxi = a'^ dyi iy2-yi)^-i^2-x.xi)dyi = 0 The first of these is holonomic scleronomic and the second is nonholonomic.{x2 . The first is holonomic and reduces the DSAC from six (the general number for a rigid body) to five.yi){dy2 . say the {x. in Pfaffian form. Example . but there is additionally the constraint that the velocity of pi is always directly along the rod. Now suppose the rod has constant length a.dyi) = 0 (2/2 .yi)dxi . . The two constraints are {X2 .xi){dx2 . The second constraint is a relation between velocities (the contact point must have instantaneous zero velocity relative to the surface) and is in general nonholonomic.

Give the equations of constraint on the finite and infinitesimal displacements of the Cartesian coordinates. because the path can vary and still satisfy the constraint. Fig. 2/2) -^2) of the bobs of a double spherical pendulum of lengths li and I2. 2-15 One of the Euler angles is the angle of rotation of some fixed line in the plane of the disk.62 Analytical Dynamics (xi. It is clear that inequality constraints of the type of Eqn. however. yi. PROBLEMS 2/1. but we are not considering this type of constraint. namely (a.yi) rolling without slipping on a plane. zi) and {x2. 2-14 rolling without slipping on a line. {x. we will return to the problem of a disk rolling on a plane and obtain the equations of motion. Fig. It is clear that there is no finite relation between 0 and the coordinates of the contact point. (2.2 —a^i) = r{d2 — Gi). zi) and {x2. Two particles having Cartesian coordinates {xi. say 9. yi. If the rolling is confined to be on a line (Fig. and therefore the constraint is holonomic. y2i ^2). Fig.y). there is such a relationship. What are the equations of constraint on the finite and infinitesimal coordinates {xi. are attached to the extremities of a bar whose length l{t) changes with time in a prescribed fashion. respectively. . respectively? 2/2. In Chapter 7. 2-15). Notes 1 2 Pars shows that otherwise the law of vector addition of forces would be violated.20) do not decrease the DSAC but rather restrict configurations to regions. 2-14.

respectively. between the angular displacement Vi and i/'2 of the disks is in .Motion and Constraints 2/3. and / < 26. roll without slipping on the shaft. Two disks of radii ri and r2. y)? 2/4. Show that the relation. A mechanism permits the disks to rise and fall in such a way that the disk rims never lose contact with the shaft.y2)-i respectively. 63 A thin bar of length I < 2r can move in a plane in such a way that its endpoints are always in contact with a circle of radius r. Problem 2/6 2/7. Are these constraints holonomic? The motion of an otherwise unconstrained particle is subject to the conditions z = xy. If the Cartesian coordinates of its endpoints are {xi. as shown. The line PO makes the angle 6 with the x axis. free of cj. what constraints on finite and infinitesimal displacements must these coordinates satisfy? Answer the same questions as in Problem 2/3 if the circle is replaced by an ellipse having major axis 2a and minor axis 26. A circular shaft of variable radius r{x) rotates with angular velocity ijj{t) about its centerline. What are the constraints on the finite and infinitesimal displacements of the point at the free end of the string having the position {x. The shaft is translated along its centerline in a prescribed fashion f{t). 2/6. Discuss the constraint on the infinitesimal and finite displacements. as shown.yi) and {x2. A particle moving in the xy plane is connected by an inextensible string of length / to a point P on the rim of a fixed disk of radius r. 2/5.

State the general condition that must be satisfied in the exceptional case that the constraint is holonomic and give an example. Formulate this constraint mathematically and classify it.64 Analytical Dynamics general nonholonomic. sin ^)d^ = 0 Is this constraint holonomic? Prove your answer. A dynamic system is subject to the constraint (cos 9)dx + (sin 0)dy + [y cos ^ — a. 2/10. 2/11. Formulate and classify the equation(s) of constraint of the particle motion under the assumption that the positions of P and O never coincide. 2/9. A particle P moving in 3-space is steered in such a way that its velocity is directed for all time toward a point O which has a prescribed motion in space and time. CO A particle moving in the vertical plane is steered in such a way that the slope of its trajectory is proportional to its height. Write down and classify the equation of constraint of a particle moving in a plane if its slope is always proportional to the time. . € Problem 2/7 2Is.

Newton's Second Law tells us that the only way the motion of a particle system can be affected is by the application of forces.-. and we wish to exploit this fact to obtain a formulation of dynamics devoid of constraint forces. iV (3. A problem arises. and define constraint forces as those that do no virtual work.1 D'Alembert's Principle Introduction. Problem of Dynamics. ••. ••.L<N (3. ••. To circumvent this problem. iiN^t) . UN-. It is natural to call such forces constraint forces. •-. and a new type of work. UN. Thus it is the application of forces that restrict the motion so that all specified constraints are satisfied. 65 r=l. s = 1. however.1) and the constraint equations N ^Arsdus s=l +Ardt = Q.ill. we define a new type of displacements.Chapter 3 Virtual Displacement and Virtual Work 3.2) .t) and initial conditions •Ujj(O) and Us{^) find the functions Us{t) that satisfy msUs = ^Fs{ui. called virtual work. The problem to be addressed in most of the rest of this book is the strictly Newtonian problem of the first kind: Given bounded functions X]-^s(^i: "•> ''^•Ni ^ i . We have seen that in certain simple problems a hallmark of constraint forces is that they do no work. called virtual displacements. because some constraint forces do in fact do work.




Displacements. We now make the following definitions. (1) Actual displacements Us{t) are those that satisfy both Eqns. (3.1) and (3.2); (2) Possible displacements dus satisfy Eqn. (3.2); and (3) Virtual displacements 6us satisfy

J2 AsSus = 0 ;

r = l,-,L<N


Thus the actual displacements are to be found among the possible displacements. Comparison of Eqns. (3.2) and (3.3) shows that for catastatic systems, virtual and possible displacements are the same. Recall that static equilibrium is possible only in catastatic systems. Virtual Work. The virtual work done by a fovce F^= {Fi,F2, ••, -Fjv) in a virtual displacement 6u = {6ui,du2, ••, SUN) is defined to be the inner product

5W = F-5u = Y.^^^'^^


Constraint Forces. A force F_' = (F{, ••, Fj^) that does no virtual work, i.e. which is such that

5W = F! •5U = J2PS^'^S
8= 1

= 0


is a constraint force. All forces that are not constraint forces are called given forces. Therefore we may write
Y,FS = FS + F;,; S = 1,-,N


where Fg and F!. are the s components of the resultant given and constraint forces, respectively. A constraint force is a force that ensures a constraint is satisfied. In the case of a holonomic constraint, the constraint force is normal to the constraint surface and its magnitude is such that the particle stays on the surface. Since this force may be in either direction, technically speaking the constraint must be regarded as two-sided (Fig. 3-1), but, in the interests of simplicity, holonomic constraints will continue to be depicted as one-sided.

Virtual Displacement and Virtual Work


Fig. 3-1

Fig. 3-2

These ideas are easily extended to a rigid body. Consider a rigid body sliding on a smooth (frictionless) surface. Considering the rigid body to be a collection of particles, each of the particles in contact with the surface is subjected to a constraint force normal to the surface (Fig. 3-2). The sum of all these constraint forces is the total constraint force on the body. This force is clearly normal to the surface and will do no virtual work (and no actual work if the constraint is scleronomic). D'Alembert's Principle. Substituting Eqn. (3.6) in (3.1), taking the scalar product with Su, and summing over all components gives

F. = Fi

s = l,2,

N s = l,2, N

{msUs - Fs)5us = F^5us

'^(nisUs - Fs)Sus = 0


where Eqn. (3.5) was used. This is D'Alembert's Principle, also called the fundamental equation by Pars.-^ It is a formulation of dynamics independent of constraint forces. It is not too useful for writing the equations of motion of a specific system; rather, it leads to other equations that are. Example - Single Constraint On a Single Particle. First suppose the constraint is holonomic scleronomic (Fig. 3-3) and given by

fix,y,z) = 0





f(x,y,z) = 0

Fig. 3-3 The Pfaffian and velocity forms of this constraint are df = —-dx + -^dy + -^dz = 0 ox oy oz (3.9)






A vector normal to the constraint surface is given by'^: , , 5/. 5/. dfj n = grad / = -—t + ^ - j + --k ox oy oz In order that the constraint force do no work, we take it to be Z=An (3.12) (3.11)

where A is such that that particle stays on the constraint. Now consider an infinitesimal displacement consistent with the constraint given by djz = dxi + dyj + dzk The work done in this displacement is dW = F-dr=(?^dx \ox + %dy + ^dz)x oy oz J =0 (3.14) (3.13)

where Eqns. (3.9) and (3.11)-(3.13) were used. Thus the constraint force does zero work. It is easy to see that it also does zero virtual work. Next suppose that the constraint is holonomic rheonomic (Fig. 3-4): f{x,y,z,t) =0 (3.15)

Virtual Displacement and Virtual Work



f(x,y,z,t + At) = 0
Fig. 3-4 The possible displacements now satisfy: dj = —dx + —dy + —dz + -—dt


•A. A


Fig. 3-5


Ttie force is still normal to the constraint at any given time; thus
dW = F-dr =




But this is non-zero in general and therefore work is done. In order to maintain the idea that "constraint forces do no work", we must consider virtual displacements and virtual work; the virtual displacements satisfy: -^dx + —-Sy + —-dz = 0 ox ay az So that the virtual work done is
SW = F-5r = {)



where 5r_ = 5xi + 5yj + Szk is a virtual displacement. For example, consider a particle on an elevator floor (Fig. 3-5). From V = dz/dt, possible displacements satisfy dz — vdt = 0 and virtual displacements satisfy 6z = 0. Thus real work is done (the constraint force changes the potential energy of the particle) but no virtual work. Note that the set of possible and virtual displacements are not only different, they have no member in common; the former have a verticle component and the latter do not. Finally suppose that the constraint is nonholonomic acatastatic. Then possible and virtual displacements satisfy a dx + b dy + c dz + p dt = 0 a 6x + b Sy + c 6z = 0
(3.20) (3.21)




respectively. The virtual displacements define a tangent plan; the constraint force is normal to this tangent plane and does no virtual work: E = X{ai + bj + ck) 6W = Z • ^r = 0 (3.22) (3.23)

In summary, constraint forces sometimes do work and sometimes they do not, but, by definition, they never do virtual work.


Lagrange Multiplier Rule

Lagrange Multipliers. Recall that the Fundamental equation is Eqn. (3.7) where the 5us are not arbitrary but are subject to Eqns. (3.3). A common technique (due to Lagrange) in problems with constraints is to adjoin the constraints with multipliers:

Y,{msUs - Fs)Sus + Y.^'-Y1

^rsSus = 0



Note that there is one multiplier, A^, for each constraint. Factoring out the Sug, Y, I rrisiis -Fs + J2 '^^'^rs ]Sus = 0 s=l V r=l / (3.25)

The advantage of this approach is that now the Sug are completely arbitrary.^ Thus the only way Eqn. (3.25) can be valid is if each of the coefficients of the Sus are zero:

msUs -Fs + Y. ^'rArs = 0 ;

s = l, -, N


Since the Fg are the given force components, the constraint force components must be:

^.' = - E ^'rArs


Thus the multipliers are directly related to the magnitudes of the constraint forces.

Virtual Displacement and Virtual Work


HZ w

Fig. 3-6

Example. A cart is constrained to roll on a horizontal plane (Fig. 3-6). It starts at rest on the y-axis and is subject to a contant force F in the negative y-direction. The constraint is z = 0 so that dz = 0 and 5z = 0 which is holonomic scleronomic. In this case L = 1, N = 3, and ^11 = Ai2 = 0. The given forces are Fx = 0, Fy — —F, and F^ = —w. Equation (3.25) gives mxSx + {my + F)5y + {mz + w + \)5z Since 5x, Sy, and Sz are arbitrary, this implies X= 0 , with solution IF z=0 -t' 2m' Note that the constraint force is equal to w as expected.
y = yo


my = —F ,

A = —w

Example. The point of suspension of a simple pendulum moves in a prescribed manner as shown on Fig. 3-7. The equation of constraint is





Fig. 3-7

y-g-\y =0 These two equations plus the constraint equation give x{t). 3-7.25) with L = 1 and A^ = 2: [mx + A'(a. X = f + IsinO y = £ cos 9 x = f + eecosd-ee^sm9 y = -iesme-£e^cos9 These relations satisfy the constraint equation.y{t). The virtual displacements satisfy {x . Let A' = —mX. since 5x and 5y are arbitrary: X-X{x-f) = 0. Prom Fig.f)dx + ydy-{x-f)fdt =0 =0 Therefore the constraint is acatastatic.f)Sx + ySy = 0 The given forces are F^ = 0. — f)]6x + [my — mg + X'y]Sy = 0 where A' is the Lagrange multiplier. The PfafRan form is 2[x-fmdx-fdt)+2ydy {x . These results may be expressed compactly in terms of the angle 0. and X{t). substituting them into the equations of motion and eliminating A gives X [x.72 Analytical Dynamics This constraint is holonomic rheonomic (integrable and time dependent). so that virtual displacements are not the same as possible displacements. (3.f) icos9{f y-g y + i9cos9-i9'^sm9) = {f + ism9{-i9sme-e9^cos9-g) = -yCos6' f) 9+jsm9 . Fy=mg Now apply Eqn.

2 All dui + Ai2 du2 + ^13 dus + Ai dt = 0 A21 dui + A22 du2 + A23 du3 + A2dt Comparing: All — -— .Fi+ XiAii + X2A21) 5ui + (171211.t) = 0 ax + by + cz + d = 0 Changing to component form: f{ui.y. OU3 ^1 = ^ 7 ot A2i = a .u2. A2 = d The virtual displacements satisfy -— dui + -— du2 + -^— 0U3 = 0 OUi OU2 OU3 a Sui + b 5u2 + c 5u3 = 0 Equation (3. r = 1.25) gives (miiii .2 ~ F2 + XiA^ + X2A22) Su2 + (m^ils .— du2 + T.z. OUi = 0 A12 = ^ OU2 . ^13 — -^— .t) = 0 aiii + bu2 + CU3 + d = 0 The PfafRan forms of these constraints are -— dui + . Suppose a particle moves in 3-D space subject to two constraints. A22 = b .m. A23 = c.Virtual Displacement and Virtual Work 73 E x a m p l e . one holonomic rheonomic and one nonholonomic acatastatic: f{x.— du3 + -— dt = 0 OUi OU2 OUs Ot a dui + b du2 + c dus + d dt = 0 The general form of the constraints is 3 y ^ Ars dus + Ar dt = 0 .-F3 + A i ^ i 3 + X2A23) Suz = 0 .

If these displacements and velocities satisfy TV Y^ArsUs + Ar=(i\ s=l r = l.Fy + Ai .29) . ••.3 Virtual Velocity and Variations Virtual Velocity. ••.Fx + Xi TT + -^2« = 0 ox my . y. If the constraints are sufficiently smooth. These three equations plus the two constraint equations give five equations for the five unknowns a. we must have mx . UN) A virtual charge of state is {u + 6u. and (J^r are now completely arbitrary. UN + 5UN. we have ^ :{5u) = 5 (^^^ dV = 5u (3.ui. |^7 + A26 ) 6y -—h A.u} = {ui. Ai. ••. ••.a)^.UI + 5ui. + Xi-^ + A2C j dz = 0 Since Sx.i . Uf^ + 5UM) where 5u is a virtual displacement and 5u is a virtual velocity. UN. L (3. z. Recall that the state of a system is the set of components of displacement and velocity {u.28) This defines virtual velocity. u + 6u) = (ui + Sui. Fy and Fz are the rectangular components of the total given force. and A2.74 or.+ A26 = 0 dy mz-F^ + Xi ~ + X2C = 0 oz df df df where Fx. changing back to the original variables. dy. ••. + ( my — + A . 3.+ (my-F. Analytical Dynamics mx — Fx + Xi | + A2a ) ^a. Fy + Xi + (mz -F..

G C'{x. this can be put into the form /db \dz dc\ dyj /9c \dx da\ dzj f ^ _ ^ \ _ n \dy dxj But from Eqn. (One might have thought the condition would be catastatic because in that case possible displacements are same as virtual displacements. Consider a function f{xi. {aSx — xSa) + {bSy — ySb) + {cSz — z5c) = 0 Using the chain rule for d. 6a. On the varied and actual paths 5{ax + by + cz) = 0 and on the actual path — {a5x + bSy + c5z) = 0 Substracting these equations. it is a possible motion).Virtual Displacement and Virtual Work they are possible states.26) this is just the necessary and sufficient condition for the constraint to be integrable. and 6c. (2. z) are time independent.) This is proved as follows.t) where all 2n + 1 arguments are regarded as independent of each other. This proves the theorem. c. . Virtual changes from a possible state lead to another possible state if and only if the system is holonomic. 75 Theorem. a contradiction. y. Variation of a Function. 6. •-. Xn.xi. c. b. but this is not the case. ••. The derivative form is ax + hy + cz = Q and the virtual displacements satisfy a5x + bdy + cSz = 0 Assume that the varied path satisfies the constraint (that is. Suppose there is a nonintegrable constraint on the motion of a particle: adx + bdy + cdz = 0 where a. Xn. 6b.

E = E.X2.X2. a. .x^.76 The differential of / is defined as Analytical Dynamics •'f=t&-' + t§-/-r^%dt We now define the variation of / as (3. in the 5 operation the Xr and Xr are varied but t is not. Consequently the variation of W is SW = y SXr + > 7^1 ^^r ^ But virtual work is defined as -WT. Let F b e a force acting on a particle.{xi. When the symbol 6W is used subsequently. it will denote virtual work.OXr OXr 3 5W = F_-^x = Y^ FrSxr r=l which contains no terms in 8x. then.2.X2. This definition causes ambiguity when appUed t o virtual work. X3.xi. R e l a t i o n t o V i r t u a l W o r k . a. in general. not t h e variation oiW. Therefore the virtual work 5W is generally not the same as the variation of the work 5W.31) that is.t) general. Thus W = W{xi. i i .30) ^/ = E r=l §^^^r '' +t r=l §fsxr ' " (3.t) T h e work done by F_ during a displacement dx is by definition dW = F-dx W -L = fp-dx in where c is a possible curve.X3.3.

(3.29). ••.33) gives N Y^AsAiis^O.-.34) we see that possible velocity changes satisfy the same constraints as virtual displacements. L (3. ••. Substituting this into Eqn. r = l. (3. N (3.37) Thus possible acceleration changes also satisfy the same constraints as virtual displacements. (3. Recall that possible displacements dug satisfy Eqn. (3. .. (3.34) Comparing Eqns.t.r.3) and (3.3).L (3.29): N J2 Arsiiis + Ails) + Ar = 0 . (3. -.36) Substituting this in Eqn. (3.. (3.35) gives N J2ArsAus = 0.32) where the Aiis are not necessarily small. r = l.33) Subtracting Eqns.35) and subtracting the result from Eqn. time to get J: (Arsus + ^ u s ) s=l + ^ = 0. (3. Recall also that possible velocities iis satisfy Eqn.29) w.35) is a possible acceleration. (3. Differentiate Eqn. L (3.L (3.. •-.Virtual Displacement and Virtual Work 77 3. s=l r = l. s=l r = l. (3. Possible Accelerations.35) If iig is a possible velocity.2) and that virtual displacements 5us satisfy Eqn. Now consider another possible acceleration: ils + Ails. N (3.4 F o r m s of t h e F u n d a m e n t a l E q u a t i o n Possible Velocities. then any ilg satisfying Eqn. Now consider another possible velocity Us + Ails . s = l.29) from Eqns. s = 1.

Equation (3. (3.5 Given Forces Conservative Forces.3) are virtual displacements by definition. (3. Note that conservative forces cannot be functions of velocity components or time. Every force for which this is not true is a nonconservative force. Virtual Work Done by Conservative Forces. Now suppose the force components of a given force depend only on the displacement components.40) the components of a conservative force are dV .ill. this will be the topic of Chapter 13. Eqn. their components are of the form Fs{ui.7). (3. ••. called conservative.Fs)Aus = 0 (3.3). ••. UM) such that a force F_ satisfies F = -grad V (3. If all the given forces acting on a system are conservative. Recall the fundamental equation. we say that the system is conservative.78 Analytical Dynamics Fundamental Equation. where the Sug must satisfy Eqn. UN. UN.UN)If there exists a scalar function V{ui.t) Here we wish to single out a sub-set of these forces.40) then F is a conservative force and F is a potential energy function.38) J2{msUs-Fs)Atis s=l =0 (3. generally. Fg = FS{UI. Equation (3. After the constraint forces are accounted for.39) will be used in Chapter 14. consequently either the Ails or the Ailg serve as virtual displacements and we obtain two other forms of the fundamental equation: N J2imsils .39) Recall that the Aiig and Aiis are not necessarily small. because in such cases there may be large finite instantaneous velocity changes. the given forces remain. From Eqn. (3.--. But any quantities satisfying Eqn. • • •.38) is useful in cases where there are impulsive forces. 3.

however. W_ does not vanish in this case but rather becomes a given force. 3-8). this force is technically outside the theory we are developing (forces must depend only on displacements. AN Ff- u W = mg Fig. velocity. Example. it is not zero because Fg is a given force. " The term on the left-hand side is the virtual work done by Fg-. If £ / depends on \N_\ (Coulomb friction). The other term is the variation of V because V is not a function of velocity. Therefore we have^ SW -SV (3. Let's classify and discuss the forces acting on the block: 1. From a practical standpoint. and time). 3.42) Thus the total virtual work done by a conservative force in going from ci G C to C2 G C through a sequence of virtual displacements depends only on the endpoints and not on the path. The forces N. Consider a block sliding on a rough surface (Fig. and W_ are different in that N_ vanishes if the constraint is removed. The normal force iV and the gravitational force W_ are both constraint forces because they do no virtual work (and no real work either). It is clear that they can be ignored in determining the motion. 3-8 . 4. 2. such forces usually may be included.Virtual Displacement and Virtual Work Forming the inner product of F_ and 5u: N N s=l 79 J2FSSUS=-Y: s=l dV 5uc du. The friction force F j does work in a virtual displacement and is therefore a given force.

3-9 Recall the following definitions: 1. A dynamic system is holonomic if all constraints on it's motion are holonomic. 2. 4. We have seen that there are differences among the forces acting on a system of particles. A system is conservative if all given forces are conservative. It will be convenient to make two other definitions. which we assume. for example electro-mechanical ones. A system is scleronomic if all holonomic constraints are scleronomic. This equation may not be true for more general systems than strictly mechanical. Provided the constraint function is sufficiently smooth. forces given constraint ("eliminated" via virtual work& energy) nonpotential potential (handled via potential energy) Fig. 3-9. Thus a natural system is always closed. and that these differences can be exploited to simplify the equations of dynamics. The classification of forces is summarized in Fig. The Lagrange multiplier rule will be precisely stated and proved in Section 6. 2. but not conversely.2. A system is catastatic if all constraints are catastatic. A dynamic system is natural if it is holonomic. The simple and concise form of the principle stated here is due to Lagrange. and conservative.80 Analytical Dynamics Classification of Forces and Dynamic Systems. 1. scleronomic. 3. Notes 1 D'Alembert's statement of his principle is somewhat difficult to comprehend for the modern reader. A dynamic system is closed if it is catastatic and conservative. 2 3 4 .

A particle of mass m is subjected to a force whose Cartesian components are x'^ + y'^ — z'^ — a"^ " (x2+y2)3/2 ^ ' 3/3. . A heavy.yi) and those of its upper {x2. and virtual displacements of the particle position in Cartesian coordinates. when the virtual displacements satisfy the knife edge constraint sin 6 5x = cos 9 6y. write down the fundamental equation and the equations of motion and discuss the C trajectories. Show that. The particle and string are placed on a smooth horizontal table so that the string is straight. and virtual displacements for the cases y2 > h and h > y2 > 0? Problem 3/1 3/2. uniform. virtual displacements from a possible state do lead to another possible state. smooth ladder of length I stands on a horizontal floor and leans against a wall of height h < I as shown. Show that. the free end of the string is set in motion with uniform velocity VQ in the plane of the table and normal to the string. possible. possible. Give the equations of constraint on finite.y2) with y vertical. 3/5. At the time io. virtual displacements from a possible state do not lead to another possible state. Let the coordinates of its lower extremity be {xi. A particle of mass m is attached to one end of a massless inextensible string of length /. 3/4. this velocity is maintained constant for all t > IQ.Virtual Displacement and Virtual Work 81 PROBLEMS 3/1. positive in the up direction. What are the equations of constraint on finite. when the virtual displacements satisfy sin ^o Sx = cos 9o Sy where OQ is a nonzero constant.

Dynamics Z 2z (a. y2)3/29 — y . A particle of mass m is constrained to move on the curve defined by X = 2 sin^ 0 . + y)2 ' x +y ' {x + z)"^ ' Calculate the work done by this force when the particle moves on the arc corresponding to 0 < ^ < n/2.2 9 + y2 + Z'^ ' y x'^ + yl z + 9 ' x^ +y'^ + z'^ Find the equilibrium positions.2 +. 3/6. y = cos 26* .2+y2)i/2 • Suppose the particle is constrained to move on a smooth sphere centered at the origin of the Cartesian coordinate system.^ — a^ (p/. An unconstrained particle is acted upon by a force whose Cartesian components are X = Y = Z = a. 3/7.9x. It is subjected to a force whose Cartesian components are (a._9 . Write the fundamental equation and the equations of motion. ./ (a.82 ^^ Y = Analytical rc^ + y^ — 2. z = 2 cos^ 9 . Find the equilibrium positions.

=x 1 ^ N = 3n 83 (4.Chapter 4 Variational Principles 4. Suppose a particle p has mass m. U2= xl .1) ^ ' For a system of n particles. U3= xl . (i.1 E n e r g y Relations K i n e t i c E n e r g y . 2. ••. U4=x'l . r = 1.3) .'-)2 = ( i r ) 2 + (^. n r=l 1 " r=l 1 " r=l where i ' ' = ij^ei + ^262 + 3:363 . (4. position x_. with of and if the position and velocity of particle r with mass m^.••• .m l i p = -mjp' 2 '-' 2 (4.)2 + (^r)2 Now change to component form: Ui=x\ . r = 1. Then the kinetic energy of the particle is defined as T — -mx 2 • X = . and velocity x_ relative to an inertial frame of reference. n Thus Eqn. ••.2) gives^ T=-Yl^suV. n.

Fs)dus = 0 s=l (4.5) or as N N ^ s=l rrisUsUa = X I ^sUs s=l (4.6) where the iis satisfy N X^„M. — = 2 ^ nisUsUs dt ~ s=l N . (4.8) Comparing Eqns. =0. Eqn. r = l.4) The virtual displacements always satisfy Eqn. „..3): N '^ArsSus s=l = 0.2)): N J2Arsdus s=l = 0. (3.7): JV s=l In a catastatic system.L Since the possible and the virtual displacements now satisfy the same equations. .6) and (4. (4.L (4.-.L (4. (3. the constraints are (see Eqn.7) Now differentiate T dT ^ .-. Recall the fundamental equation. (3. the fundamental equation may be written N Y.imsUs .84 Analytical Dynamics Kinetic Energy in Catastatic System.-..9) .8) gives: N dT ^ = J2FSUS dt (4. r = l. s=l r = l.

10) s = l where his a constant of integration.dus + JY^F^dus + h s—l s=l r ^ BV r ^ N T + V= r^F^^dus+h s=l (4. Suppose that some of the given forces are conservative and some are not and let F^ = s component of resultant of all conservative forces = —dV/dug F^^ = s component of resultant of all nonconservative forces. the total mechanical energy of the system is constant over time for actual motions: T + V = h = constant (4.11) If all forces are conservative and included in V. N •' N T= r^FsUsdt + h= r^Fsdus + h s = l •' (4. (4. Then Eqn.9) to get . in a closed system (catastatic and conservative).Variational Principles 85 This states that in a catastatic system. they hold for holonomic or nonholonomic systems.F^. If the iig are all continuous (as we are assuming because all forces are bounded) and if the number of particles is constant (as we are also assuming). and only in a closed system. Note that these relations are true for all catastatic systems. the time rate of change of the kinetic energy equals the power of the given forces under possible velocities. . we can integrate Eqn. (4.10) gives N ^ N T = JY. Energy Relations in Catastatic Systems. the total mechanical energy is a constant (is conserved).12) That is.

in vector form is n Y^i'^rx'r=l F') • Sx'= 0 (4.3 Hamilton's Principle First Form.2 Central Principle Central Principle. (4.7).17) This is called the central principle by Hamel. using the fundamental equation.14) Prom Eqn.^T (4. Now consider a general system.rp n nrp n r=l — 0 r=l — r—l Combining Eqns. Eqn.18) .16) Finally.17) between times ^o and ti. (4.15) gives n A ( ^ \ Y. 4.14) and (4. each more specialized. *o + 6W)dt (4.13) where FJ' are the resultants of the given forces.2) the variation of T is n p. Eqn. Consider I / n \ n n j -— I ^ nirif • SxJ' = ^ THrx!' • SxJ" + ^ nirif • —{Sx_^) \r=\ n / r=l \ / r—l " A ( ^ y ^ TTiri'^ • ix^ = —. (4.13).86 Analytical Dynamics 4. The fundamental equation.^ nir^ • dxf r=l \r=l — ^ rurif • Sif r=l (4.F! • Sx' = SW r=l (4. *i "*: 22 m-rX^ • ^x^ = f\dT Jtn r=l M. (4. We will derive several forms of Hamilton's principle. Integrating Eqn. • / n \ / n ^ [Y^rif \r-=l •5x:-\-5T = Y. (3. mrx^ •^^'' = Y \ ^ ^^^^ • %'• .

(3.18) therefore reduces to: '{6T + Jto 5W)dt^0 (4. 5T + SW = ST-5V = 5{T .42) which is the variation in T — F .V) (4. in this case. Second Form. Therefore.Variational Principles 87 varied path E C E N+1 actual path tl Fig. 4-1 Now consider virtual displacements from the actual motion satisfying Figure 4-1 shows the situation in the event space. not the variation of W. Equation (4. Define the Lagrangian function L^T-V Then Hamilton's principle for a conservative system is: SLdt==0 Jto (4. but that 6V is the variation of V. which states that "The time integral of the sum of the virtual work and the variation of the kinetic energy vanishes when virtual displacements are made from the actual motion with endpoints held fixed". known as the extended or unrestricted form.20) Eqn. If all given forces are conservative.21) (4. Note that the variations take place with time fixed.19) This is the first form of Hamilton's principle.^ applies: 5W = -5V Recall that SW is the virtual work. thus they are called contemporaneous.22) .

in Hamilton's principle constraints are generally violated and this is not a problem of the calculus of variations.mgf Since the only given force is conservative and the only constraint is holonomic. we will not do this for this problem.88 Analytical Dynamics or "The time integral of the variation of the Lagrangian function vanishes for the actual motion". A particle moves on a smooth surface with gravity the only given force (Fig. ^ J We could carry out these variations to get the equation of motion. Since. stated earlier. Then the variations satisfy the constraints and Hamilton's principle is s / 'to Jtn Ldt = 0 (4. The derivation of the various forms of Hamilton's principle given here are completely reversible.23) or "The time integral of the Lagrangian is stationary along the actual path relative to other possible paths having the same endpoints and differing by virtual displacements". . Now suppose the system is conservative and holonomic. in general. We have: T-i •'2 ^ / -2 I -2 I •2\ 1 = -mv = -m{x +y +z ) V = mgz Z = fxX + fyV L = T-V = \m x'^ + f + ifxx + fyvf] Zi . starting from them we may derive the corresponding fundamental equations. This equation is usually referred to as simply Hamilton's Principle. but will do it for the following one.^ Third Form. Thus Hamilton's principle is necessary and sufficient for a motion to be an actual motion. 4-2). The exception. It is precisely an integrated form of the fundamental equation. Example. that is. Remark. virtual changes from a possible state do not lead to another possible state. the third form of Hamilton's principle applies: 5 r Ldt = 5 f' ^ {^2 ^ ^2 ^ (^^^ ^ j^^2 _ 2 ^ ^ | ^^ ^ Q Jto Jto 2. is a holonomic system.

mgi(l .. therefore by the Fundamental Lemma of the calculus of variations (see next section) we must have 6'+^sin6' = 0 ../ 59 9dt JQ Jto / ' (-19 59-g r 'to sm9 59^ dt = 0 =0 (i6 + gsin9)5ddt But 59 is an arbitrary virtual displacement. ti i-tx 659 =0 . Consider again the simple pendulum (Fig.cos Q) The third form of Hamilton's principle applies: S Jtn 'to Ldt = S Jtn J to -mre^ .c o s 61) dt = 0 We will now carry out the variation to get the equation of motion: f ' [mfe 59 . We have: V = mgl{l -cos6') L = T-V = -mte^ . esedt Jtc Consequently / e—{5e)dt= Jto dt pti . .mgesine SO] dt = 0 Integration by parts gives^ I ti .d f*i (j .mgi{l . 4-3).Variational Principles 89 Fig. 4-3 Example.

to be derived shortly. in the calculus of variations necessary conditions are obtained by considering the first and second variations of J.ti] and ti M{t)rj{t)dt = 0 to .4 Calculus of Variations Statement of the Problem. to zero. Attention will be restricted to the "simplest problem" of the calculus of variations. We seek the function x = x{t). Analogously. stated as follows.26) Two key lemmas are needed to establish this result. necessary conditions for the minimum of a function are obtained by considering the first and second derivatives. Carrying out the differentiation gives the long form of the Euler Lagrange equation: fx .90 Analytical Dynamics We see that getting equations of motion from Hamilton's principle is somewhat cumbersome. it is a candidate for the minimizing function.24) a minimum subject to fixed endpoints x{to) = XQ and x{ti) = xi.x.Lagrange Equation. Euler . Because of the close connection between the variational principles of dynamics and the calculus of variations. obtained from setting the first variation. essentially carry out this variation in general for all problems and are much easier to use.t)dt (4. The Fundamental Lemma states that if M{t) is a continuous function on [to. In ordinary calculus. is that iix{t) minimizes J then it must satisfy the Euler-Lagrange equation: fx -jJ- =^ (4-25) where subscripts indicate partial derivatives.fxxX ~ fxxX = 0 (4. A function satisfying this equation is called an extremal. t G [ioj^i]) that renders the integral -h J= / I to f{x. 4. . the latter will be briefly reviewed. The most important result. Lagrange's Equations.

27) Applying Eqn.a.31) (4.t) are the extremals of rti i-ti (4.29) twice.ti].x.ii] and if rti / riN dt = 0 Jto for all r}{t) G C" with rj{to) = r}{ti) = 0 then N{t) = constant for all t G [to. The other necessary conditions.23): fti 6 I to Jtn L{x. Hamilton's principle for such a system is Eqn. we are given a two parameter family of curves x = g{t.x. Application to Dynamics.a. Consider a holononomic.) Inverse Problem. will not be discussed here. (4. x.l3) x = gtt{t.a.Variational Principles 91 for all r]{t) G C^ with ri{to) = r]{ti) = 0 then M{t) = 0 for all t € [to.t)dt =0 (4. d fdL\ dL This is in fact Lagrange's equation for the system. ti].x. x = gt{t.30) Differentiating Eqn. arising from consideration of the second variation. In the inverse problem of the calculus of variations. (4.P) and we want to find a function f{x.29) such that the family members J= / 'to Jtn f{x./3) (4. (Lagrange's equations for general systems will derived in Chapter 6.t)dt (4. The Du Bois . (4.25).32) . conservative system with a single coordinate.Reymond Lemma states that if N{t) is continuous on [io.

a. il){x.t). must be chosen so that / satisfies Eqn.29) and (4. ^ x =F = dV{x) ~ dx .t) (4.26). From Newton's Second Law we know that the equation of motion is .dM BM ^ . (4.x./3).33) This equation must be the Euler-Lagrange equation. Eqns.x. (4. (4. this becomes dM .36) twice gives an expression for / : f = f f M dxdx + x\{x. (p{x. (4. (4. a = p{x. t) (4.34) may be shown to be of the form 9{t.x.26).. it must be identical to Eqn.p))dt The solution of Eqn.37) where A and JJ. Since $ is arbitrary there are an infinity of such functions / and thus the solution to the inverse problem is not unique. Substitute Eqn.g{{t. ^ Now define the function 9{t^ a. t) + /i(a. /3) = exp lGi{t. Integrating Eqn. (4.92 Analytical Dynamics Under general conditions.31) may in principle be solved for a and /3.x^t)) (4. that is.33) into Eqn.x. Consider again a system with a single generalized coordinate subject to a conservative force.a.35) where $ is an arbitrary but nonzero function of (p and ij).. (4.t) /3 = and then substituted into Eqn. Example.32) to obtain an equation of the form x = G{x..26) and differentiate the result with respect to x to get Jxtx I XJxxx "T f-'jxxx ' ^xjxx ^^ U Letting M = fa.t) ip{x. (4.

37).t) To get the "simplest case".x.).t)dxdx + xX{x.x.35) gives d{t.26). (4. t) _ dX{x. from Eqns. and / becomes f = -x'^ -V = T-V = L ^ 2 Hence the "simplest" variational problem that leads to the correct equation of motion for this case is 6 1 Ldt = 0 /to 7' l-tl which is. M / = = $ = /ix / ^{x. select $ = 1 to obtain / = -cc +xX + ii Substituting into Eqn. (4. We want to find a function f{x.t)dt Jto That is. Choosing other functions $ gives other variational principles.t) + fi{x.a.t) _ dV{x) dx dt dx Thus we must have A = 0 and /i = — F(a. Jx Jxt XJxx ~ XJxx ^ U 1 2 diijx. . Hamilton's principle.36) and (4. (4. rti 5 f{x./3) = exp / Grdt = 1 Therefore.x.t)dt =0 Jto From Eqn.33) we see that ax so that Gx = 0 and Eqn. of course.t) such that this differential equation is the EulerLagrange equation for fh J = f{x. (4.Variational Principles 93 where V{x) is the potential energy function of F.x.

the principle becomes a problem of finding geodesies in a Riemann space. lent his reputation to the PLA and the glory of his invention to Maupertuis. The only requirement is that they be "valid". a truly great man. The previous Section shows that it is possible to generate an infinite number of variational principles. Mach remarked that "Euler. thus putting the principle an a sound basis. satirizing the PLA in some of his books. Maurepertius' viewpoint was that "nature in the production of her works always acts in the most simple ways". On the one hand. It is surprising that the principle that is perhaps the most straight-forward and useful. that is. (Euler was also the first to consider the inverse problem. that of Hamilton. Immediately after statement of the principle. He stated the principle in metaphysical terms and never proved the PLA in the sense of showing that it was equivalent to the established laws of dynamics.) We now have the PLA in two forms. who knew little of mathematics and science. In this view. a controversy started. a matter of great concern to eighteenth century scientists. Variational principles arose initially to meet this perceived need. practical and useful". Many years later. most notably Koenig. did not emerge . in other words. The idea was that of all the possible motions of a dynamic system. associated with the names of Lagrange and Jacobi. The latter's version has path length as the independent variable and may be viewed as a geometrical statement. Although Newton's Second Law gives highly accurate results in most situations. the one that is actually followed is the one that minimizes some fundamental quantity. The first successful variational principle was Fermat's principle of minimum time in optics. that the principle was not valid or that Leibniz had discovered it previously. that they lead to the same equations of motion as does Newton's Second Law. or both! Even the great Voltaire. but he made a new thing of the principle. nature acts in the way that is most efficient. Fermat was able to derive the laws of refraction. some claimed. got into the act. Maurpertius stated the Principle of Least Action (PLA) in dynamics from analogy to Fermat's principle. With this principle. it doesn't seem to emanate from any deeper philosophic or scientific principle.5 Principle of Least Action Historical Remarks.94 Analytical Dynamics 4. He starts "from the principle that Nature always acts in the shortest ways". Euler sided with Maupertuis and managed to prove (in the sense just stated) the PLA for a particle.

39) at and is illustrated on Fig.41) . 4-4 Lagrange's Principle of Least Action. varied path actual path Fig.Variational Principles until much later than the PLA. We consider closed systems only. (4. t) is given by dF StF = SF + —St t (4.12). the variations from the actual path take place with time fixed and the variations are zero at the endpoints. so that energy is conserved. In the principle of least action we consider variations from the actual path with energy held fixed.40) The Lagrange form of the principle of least action is 5tA = Q (4. and denote the noncontemporaneous variation operator by 5t. Thus from Eqn. In Hamilton's principle. as shown on Fig. 95 Noncontemporaneous Variations. Now we relax this restriction and consider noncontemporaneous variations. 4-5. 5tT + StV = 0 (4.38) The relation between the operators 6 and 5t for a function F{x.^ The action is defined by A r 2T dt (4. 4-4. the details of the derivation will be omitted and the results will be summarized. Because the principle of least action is largely of historical interest only.

96 Analytical Dynamics In words.V) dt (4.42) However. the Jacobi form of the principle is StA" = 0 This is a problem in the calculus of variations. (4.42) in terms of s. Clearly.42) is homogeneous of degree one in the iis. Eqn.40) may be omitted. the path length in configuration space. from Eqn. (4. Jacobi's Principle of Least Action. (4.^ Writing Eqn. Since energy is conserved. The principle is both necessary and sufficient and thus it may be used to derive equations of motion.44) . (4.40) may be written A' ^ f' 2^T{h . (4.V) ds Consequently.3) T is a quadratic function of the its and thus the integral in Eqn. We note that the varied motion does not in general take the same time as the actual motion and in fact the varied motion is not in general a possible motion. the factor 2 in Eqn.This means that A depends only on the path in the configuration space and not in the event space. (4.43) (4. gives A' = f ^J2{h . "the action is stationary for the actual path in comparison with neighboring paths having the same energy".

for both. inextensible string which passes over a frictionless.. 1963. V du : here we take u = 9 and dv = —(69)dt. Hamilton's principle for a class of non-conservative systems may be found in "Some Remarks on Hamilton's Principle". homogeneous pulley of mass m. (c) Write down Hamilton's principle. This was proved in Section 3. as shown on Fig. an alternate proof will be given in Section 6. (b) Give the energy integral.ndT'{ui. --jis) a. The other end of this string is attached to the center of a frictionless. Mech. Dec. T. --TUN) are two different functions but we will use the same symbol. T = T{xl.xl.Variational Principles 97 Notes Strictly speaking. Gravity is the only force acting on this system. massless pulley.xl. dt See Rosenberg or Pars for the details. A weight of mass 4m is attached to a massless. G.5. 4/1. US Problem 4/1 Problem 4/2 . A second massless inextensible string having masses m and 2m attached to its extremities passes over the pulley of mass m. if one exists. Recall that u dv = uv — See Pars PROBLEMS 4/1.3. Appl. J.X2. Leitmann. (a) Give the kinetic energy for this system.

98 4/2. (b) Give the energy integral. (c) Write down Hamilton's principle. as shown. respectively. if one exists. linear spring of rate k is attached to the center of the disk and to a fixed point. it hangs vertically down over the table edge. rolls without sliding on a horizontal line as shown. (c) Write down Hamilton's principle. A heavy. . horizontal table. are constrained to move so that they lie for all time on a straight line passing through a fixed point. Problem 4/3 4/4. (a) Give the kinetic energy for this system. Three particles of mass mi. constrained to remain in a vertical plane. Analytical Dynamics A homogeneous disk of mass M. homogeneous inextensible string of given length remains for all time in a vertical plane. For the force-free problem in Cartesian coordinates: (a) Give the kinetic energy. The particle velocity is directed for all time toward a point P which moves along the x axis so that its distance from the origin is given by the prescribed function ^(t). It lies in part on a smooth. Give the same answers as in Problem 4/2 when the configuration is changed so that the line is inclined by the angle a to the horizontal. If the free length of the spring is I. (b) Give the energy integral. 4/3. A massless horizontal. and the disk radius is R. m2. What is Hamilton's principle? A particle of mass m moves in the x. 4/6. and in part. if one exists. and ma. y plane under a force which is derivable from a potential energy function. 4/5.

zo.Variational Principles (a) How many degrees of freedom does the particle have? (b) What is Hamilton's principle? (c) Give the energy integral.t) = 0. explain why. Does an energy integral exist? If so. If none exists. if one exists. 99 One point of a rigid body is constrained to move on a smooth space curve defined by f{xo. If the forces and moments acting on the body are conservative. give Hamilton's principle. write it down. yo. . 4/7.

the first of these doesn't appear at all and the second appears via a potential energy function (see Fig. only one coordinate is needed and 0 is the obvious choice. We now take up this second point and introduce "generalized coordinates". A second important shortcoming is that coordinates x and y are awkward and. for example. In the energy method. one is redundant. Generalized Coordinates. use of Newton's Second Law in rectangular coordinates has the shortcoming that both the constraint force T and the gravitational force W appear explicitly. of which an example is 9 for the pendulum. Suppose a system of N/3 particles is 101 .1 Theory of Generalized Coordinates Remarks. As revealed by the simple pendulum (Fig. 3-9). 5-1). more fundamentally. however.Chapter 5 Generalized Coordinates 5.

that is n = N — L' (note that. Ajs dus + Ajdt = {).102 Analytical Dynamics constrained by L independent constraints.3 and 2.L' (5.t) = as .t) . Introduce transformation functions qs = ps{ui. Prom now on we will use the symbol n to denote the DSAC.'-. Now make the following definitions: 1. and the remaining Ps{-). n will not denote the number of particles as before).7 that DOF = iV . ••. L Suppose further that L'{< L) of these are holonomic and L — L' = i are nonholonomic. 2.5) where the a^ are constants. ••.t) = ai] i = l. Any finite set of numbers {qi. Ps{ui. i = l. Eqns. Thus n is the least possible number of coordinates and excedes the DOF by the number of nonholonomic constraints.L' (5. We wish to transform from rectangular coordinates to generalized coordinates.--. i. > n. s=l j = L' + 1. L (5. s=l r = 1.4) such that the first L' satisfy the holonomic constraints. if the integrated form of the holonomic constraints is fi{ui.2): N y~^ Arsdus + Ardt = 0 .-. s = l.-.e.-'jQn} is called a set of generalized coordinates where n is defined as above. Any set of numbers {qi.--. s = l.1) then the PfafRan form of the constraints is j^f^dus 8= 1 ^ + ^dt^Q.N are .q2. (3.--.e. ••.-.UN.L and DSAC = N — L'.2) N Y.L' (5. therefore. that completely defines the configuration of a system at a given instant is a set of coordinates.^ Transformation of Coordinates. s = L' + 1.UN. i.Qn}j n.3) Recall from Sections 2.N (5.UN.-.

6) du N Under these restrictions. Single-valued. qn. qN = <ln These then are the generalized coordinates. 3. Q!L'. ••. . s = 1. ••• . s = l. ••. UN. From the last of these. the differential displacements are related by dus = Y] -^dqk + -KT-dt. qN. dpi dui J dpN dui dpN dpi dupf ^0 (5. t) .-. ••. the transformation is one-to-one and onto (Fig. Continuous with continuous derivatives.8) Us = Us{qi. t) . we have now qs = psiui.-.9) . 2. n s = l. therefore by the implicit function theorem the transformation can be inverted to give Us = = Usiqi.t) . ••.. QN. 5-2 not one-to-one Now re-label: QL' + 1 =11. dt s = 1.Generalized Coordinates 103 1.N (5. t) Usiai. 5-2). Such that the Jacobian is not zero. that is. iV (5. -. QL'+i.N (5. ".7) one-to-one and onto not onto F i g .

-.15) . these are n J2 BrkQk + Br = 0 . fc=i where ouc ^rk — / .-. The nonholonomic constraints are N ^Arsdus s=l +Ardt = ()\ r = l. the discarded L' coordinates have accomplished this.10) ti ^^i' k P o s s i b l e a n d V i r t u a l D i s p l a c e m e n t s in G e n e r a l i z e d C o o r d i n a t e s .i (5.14) As before.11) where i = L — L'.104 Analytical Dynamics and therefore the virtual displacements are related by n n S^s==^^5qk. r = l.N (5.i (5. Substituting Eqn. these equations define possible displacements and velocities.-.-. s = l. k=l r = l.11): ^ Brkdqk + Brdt = 0 . Virtual displacements satisfy n J2BrkSqk = 0.i (5.9) into (5. (5.-.13) Br — 2_^ '^'Ts "^ + -^r s=\ In terms of velocity components.12) (5.i (5. dqk s=l N r = l. The generalized coordinates are not subject to the holonomic constraints. -^rs p.

16) (5. (3. and have any dimensions.Generalized Coordinates 105 It is important to realize that generalized coordinates are general. Let the rectangular components of the bob be {x. angles.28): d{5u) = 5{du) It may be shown that also d{5q) = S{dq) (5. they can be distances. 5-1) yx"^ + y'^ = i which is of the form f{x. the d and 5 operators are communicative in generalized coordinates.y.t) x .y) and catastatic.17) In words.y. Recall Eqn. We have: = a and which is holonomic. Variation Operator. etc. we transform to new variables such that first is equal to a and second is convenient: q^ = a = i = 92 = tan pi{x.2 Examples Simple Pendulum.t) .= p2{x. 5. scleronomic. N = number of rectangular components = 2 L' = number of holonomic constraints = 1 i = number of nonholonomic constraints = 0 L = number of constraints = L' + i = 1 DOF = degrees of freedom = N — L = 1 n = number of generalized coordinates = N — L' = 1 According to our approach. then the constraint is (see Fig.y).

106 The inverse transformation is X = y = Analytical Dynamics qi cos §2 = « i cos q2 = ^ cos §2 qi sin q'2 = a i sin g2 = ^ sin §2 The Jacobian of this transformation is dx dqi J = dy_ dqi dy_ dq2 dx dq2 COSQ2 sin 92 -qismq2 -qi cos 92 qi cos^ 92 + 91 sin^ 92 = 9i = -^ 7^ 0 Therefore the transformation satisfies all three requirements.3 it was shown that a rigid body in unconstrained motion has D O F = 6. thus it has 6 generalized coordinates.9s) .92. 5-3): {9i. 5-1) and the choice of 9 could have been made by inspection.^3) determine the location of p .9 The 9i and 92 in this problem are of course just the polar coordinates (Fig. 5-2).^2. This. but onto {x. Then either {x. A common choice of angles are the Euler angles.92.92. and three angles defining the location of body fixed axes. is not true because the transformations have the properties (see Fig. These are usually taken to be the coordinates of some body-fixed point. however. In Section 2. This seems to imply t h a t there is a relationship f{9i.y) or (^1.y) —> i9i. these are defined and used in Chapter 11. E x a m p l e . Now relabel to get the generalized coordinate: q2 = qi and the transformations are now 9 = tan"^^ X =e x = i cos 9 y — ism.y) is one-to-one. but not onto is not one-to-one.9^) —> {x. say the center of mass.93) = constant because it only takes two parameters to give the location of p. Three bars are hinged and lie in a plane such that one end is attached at 0 and the other end carries particle p (Fig.

generalized coordinates usually suggest themselves from the geometry of the problem. . In some cases there are isolated points in configuration space for which one or more of the three conditions on the coordinate transformation are not satisfied. 5-4 A basic assumption is violated and the transformation theory does not apply. the transformation is restricted to regions of the configuration space not containing these points. The general theory of transformation is needed. In this case.Generalized Coordinates 107 Fig. For specific problems. Remarks 1. because we desire to put the key equations of dynamics in terms of generalized coordinates. the choice of 9 for the simple pendulum is an example of this. however.

108 Analytical Dynamics Notes 1 In some texts.7) for this problem. (5. How many constraints must the Cartesian coordinates satisfy? What are they? Choose suitable generalized coordinates to describe the position of the flyballs. Problem 5/3 .7) for this case? (c) Examine the Jacobian. PROBLEMS 5/1. 5/2. A particle moves on the surface of a three-dimensional sphere. Construct Eqns. they are sometimes called Lagrangian coordinates. A particle moves on the surface of a right circular cylinder whose radius expands according to the law r — f{t) while its axis remains stationary. 5/3. If unconstrained. (a) Choose suitable generalized coordinates for the motion. "generalized coordinates" are not necessarily minimal. Also. Answer the same questions as in Problem 5/1. (b) What are the Eqns. (5. A centrifugal governor has the configuration shown. six coordinates would be required to define the configurations of the flyballs.

--. Previously we've expressed the kinetic energy of a system of particles in component form: T = -1 m . i . We now wish to put our previous results in terms of generalized coordinates. ^ = 2 X!X!""^^"^^+ X1^"^"+ ^ 9a ' 109 (^•^) .iV (6.1 Fundamental Equation in Generalized Coordinates Kinetic Energy.3) From Eqns. ^ dur (6. dUr 1 Ar (a ^\ '^r = 2^ jr-Qk + ^T'^ Therefore^ 1 ^ rrir r=l r = l.1) lk=l E dur . (5.2) Expanding.Chapter 6 Lagrange's Equations 6. 2 ^ ^ (4. instead of rectangular coordinates.9) the iir in terms of the qr are given by: ^ dUr . We begin with kinetic energy.

Example. (6. = / ( i ) + £sin6'.6) a^^ must be a symmetric positive definite matrix.m ( i 2 + y2) k.3). T = .5) As a special case. the transformation equations are a. Therefore T = ]-m \t^d^ + 2£^ cos ^ / + /2] y = icosB .f(t) - Fig. T > 0 and therefore from Eqn. for dugjdt — 0: ^ = o E E '^a/?^" ^? = (6. ( where daB = 2^ r Y^ Analytical Dynamics dUr dUr dqa oqp dUr dUr dqa Ot r (6. Consider a simple pendulum with moving support point (Fig.6) Prom Eqn.4) = i E rur dt >0 and where the following notation has been adopted T V i i=l E =E (6. 6-1). In rectangular components. 6-1 Choose Q as generalized coordinate.

qnTwo Equalities. ^^s = 0 (6. d / dUr dt\^dqsJ '^ dqs '^ dt \dqs .3) becomes '-anO'^ + bi9 + c Comparing.Ur -^— dqs dt \ dqs J dqs .9) Now consider .1+[ f)r.--.Lagrange's Equations For one generalized coordinate. dqs Substituting the second into the first. ^1 = mi cos Of .. (6.10) y^ r {rrirUr — Fj. Eqn. dUr \ .„ ^ _.)Sur — 0 Sur^y]~6qs. dUr d (. dUr \ . dUr Ur -^— = T : «r -^^ . Note that the coefRcients aa/3) ^a-.-. 1 ("^rUr -Fr)P^] r = l.N ^ I J2 dq. (3.7) and (5.10) . Recall Eqns. I ^^'^^ Fundamental Equation. ail = "^-^^ . c= -mf 111 This extra complexity in T due to choosing 0 as a coordinate is worth it in the long run. .1) we have diir dUr dqs and diir dq^ ~ dqs (6. (6. From Eqn. ^"^ ' a+ Fir..7) 1^ P. (6. and c are in general functions of the generalized coordinates qi. dt \dq. . dqadqs'^'^ '^ dt dqs ~ d (du. dUr .„ Fir.

(6. ' d q s dn- dUr 5g. Fr^}\5qs = 0 (6.14) Now define the generalized force Q as that force with components Qs — 2-j Fr -^— r dqs Then we get finally (6. (6.112 where Eqns.2-j rnr2Ur -^— 1 v^ ^.7) and (6. provided only they are sufficient t o specify the configuration of the system. dur .12) dT —oqs = 1 v^ ^. —- dT dqs = . then Qs is a force.15) " 'l(^)-^^Q.11): E d^ /dT dT =^ ^ dt \'d^.11) But from Eqn. (i) The coordinates are any convenient set of parameters. s=l n _dt \dqsJ dqs Sqs = 0 (6.10) into Eqn. There are several advantages to this form of the fundamental equation. li Sqs is a distance. (4. QsSqs is the work done by generaHzed force component Qg in a virtual displacement Sqs.13) Substituting these in Eqn. if 6qs is an angle then Qs is a moment. 1. 2.2^ mr2Ur -^T2 r dqs (6.8) were used. Substitute Eqn.16) This is the fundamental equation in generalized coordinates. (6. = 0 ' ' dqs (6. . dur oqs 2 r (6. further they are the minimum number to do so.9): d dUr rrir dt \ ^ dqs Analytical Dynamics s \ r dq. Remarks.3). (6.

Lagrange's Equations


(ii) The Sqs must satisfy only the nonholonomic constraints; the holonomic constraints do not appear at all; they have been eliminated by our choice of coordinates. Conservative Forces and Potential Functions in Generalized Coordinates. Recall from Section 3.5 that if F*^ is a conservative force, there exists a function V{ui,U2, ••,uj^) such that E" = -grad V or, in component form, (6.17)

Now make the change of variables to generalized coordinates Us{qi,- • •, qn,t) in V: dV ^ dqr ^ s dV^ ^ = _ V i^c ^ ^ _Qc dug dqr s ^ dqr ^ ir

Therefore the relations Q^ = - 7 r - ; s = l,-,n (6.19) oqs hold for generalized coordinates. Now suppose some given forces are conservative and some are not. Each of the conservative forces has a potential function and the sum in terms of components in generalized forces is





Recall the definition of the Lagrangian function, Eqn. (4.21), and the fact that V depends on the qg but not the qg] therefore dL oqs dT oqs , ,

Consequently, Eqn. (6.16) can be written


dt \dqsJ


5qs = 0





where the Q^^ are the components of the nonconservative given forces. R e m a r k . It is also possible to generalize the concepts of conservative forces and potential functions to include velocity dependence of a certain restricted kind, namely linear in the velocities. This will be done in Section 6.4.


Multiplier Rule

T h e D y n a m i c a l P r o b l e m . Recall the fundamental equation in generalized coordinates, Eqn. (6.16). The qs are any set of generalized coordinates, that is, any minimal number of independent parameters completely specifying the configuration of the system at any time. The minimal number of coordinates is: n = DSAC = N-L'

where, as before, N is three times the number of particles and L' the number of holonomic constraints. T h e functions T and Qg are in general functions of (gi, ••, g„, gi, ••, ^ „ , i ) . T h e virtual displacements are not arbitrary but must satisfy Eqns. (5.15): J2BrsSqs s=l = 0r = l,-,i (5.15)

where the constraints in velocity form are given by Eqns. (5.14): n Y,Brsqs s=l + Br = 0; r = l,-,i (5.14)

As before, any {qs,qs) satisfying Eqns. (5.14) are a set of possible displacements and velocities; i.e. they are a possible state. Among the possible states is the actual state that also satisfies Eqn. (6.16). N o t a t i o n C h a n g e . Let

d fdT\
dt \dqsJ R = (Ri, ••) Rn)


Rr — {Brl,Br2,

••, -^rn) ;

r = 1, •• )' )

Lagrange's Equations Then Eqns. (6.16) and (5.15) may be written
R-Sq = 0



By assumption, the constraint equations are Hnearly independent; this means that there do not exist constants c^; r = l,--,i not all zero such that

J2 ^B_^ = Q


Multiplier Rule for Single Particle and Single Constraint. For a particle in 3-space with a single nonholonomic constraint, N = 3 and i = 1; thus there is one vector B_i and this vector and R have three components.

Fig. 6-2 It is clear that there is no ci for which ci^^ — 0 except ci = 0. B^i • Sq = 0 means that 6q is such that it must be normal to B^^ but is otherwise arbitrary (Fig. 6-2). R- 6q = 0 means R must be such that it is normal to Sq for all such possible Sq. The only possibility is that R is in the same direction as 5^; thus we may write R = -Ai^i Rearranging and dotting with a completely arbitrary Sq, {R + XiB_^)-6q = 0 In essence, this combines the fundamental and the constraint equations into a single equation; in this new equation, the 5q are not required to satisfy the constraint.




General Multiplier Rule. We now return to the general problem. Consider vectors^ a = X = k such that

{ai,--,an) E. IE"' {xi, ••, Xn) G -ZE" (&i„, ••, biJelE'';

given arbitrary i= l,--i<n,


all 6j independent, given

< a,x> <ki,x>

= =

^ aiXi = aixi + a2X2 + • • +anXn = 0
2= 1


Without the last of these, the first implies a = 0 ; (ai, ••, a„) = (0, • • 0). Since the 6j are independent, they form a basis in lE^ C IE"'. Write IE" = lE^ © iE^, where iE^ is the orthogonal compliment of lE^ in iE". Then x J^ -g^ y_e lE^ <=^ x £ iE^ and < 6^, ^ > = 0 = ^ ^ G i E | . Also, < a,x >= 0 = > a ^ iEj^ = > a e lE^ = ^ a can be expressed in terms of a basis in lE^, say 6j. Therefore, i e a = Y^ Ciki , or, aj = ^ Cibij ; j = 1, -, n (6.26)
i=l i=l

Application to Dynamics. Identify a = R; Then Eqn. (6.26) is x = 5q; k = B,. ; Ci = -Xi

e E = - E ^rRr


Dotting with any completely arbitrary vector 6q:

lR + J2^rB_r) •h = ^
Finally, using Eqns. (6.23) to revert to the original notation


d (dT\





SQS = 0


which is another form of the fundamental equation. This equation is a statement about the system as a whole.

Lagrange's Equations



Lagrange's Equations

Derivation from Fundamental Equation. Since the 5qs are now arbitrary, from Eqn. (6.28) it is necessary and sufficient that TA7^)-7r--Q^ dt \dqsJ dqs + T.^rBrs ;^ = 0; s = l,-,n (6.29)

These are Lagrange's equations of motion. They are n equations in the n + i unknowns qi,--,qm M,'-,^e I the additional i equations to be satisfied are

J2Brsqs + Br = 0;

r = l,-,i


Equations (6.29) are statements about each individual component. Derivation from Hamilton's Principle. The first form of Hamilton's Principle is Eqn. (4.19): f \ST + 5W)dt = 0
I to Jtn

where dus{to) = 5us{ti) = 0 ; s = l,--,N T{qi,",qnjQii

Transforming to generalized coordinates, we now have qn,t) and

SW = J2 Qshs
The virtual displacements must satisfy Eqns. (5.15),


and the possible displacements and velocities must satisfy the nonholonomic constraints given by Eqns. (5.14). By the multiplier rule for integrals, the equivalent problem is /*' ('^^ + E Q^^'is - E E >'rBrsSqs] dt = 0
•^*o \ s=l s=lr=l 1





where the 5qs are now arbitrary. The variation of T is

Consider the SQS term; integrating by parts:

* ~i o

* i A ST d 9qs dt_


.=1 ^1^

i:mw>'- ^ . o L J.

Consequently, Eqn. (6.31) becomes /•ii"!" d fdT\ dT ^
= 0 (6.32)

Jto ^ 1 [ dt \dqsJ




Since the 5qs are arbitrary, we must have Eqns. (6.29):

d fdT\





Historical Remarks. Lagrange was born in 1736 in Turin, France. His abilities were recognized early and he was appointed professor of mathematics at the local technical college at age 16, after having mastered calculus on his own. At this time he began development of the calculus of variations. This subject was originated in correspondence between Euler and Lagrange. Euler and other leading scientists arranged to have him appointed to the Berlin Academy of Sciences. While in Berlin, Lagrange developed a friendship with Fredrick the Great. After Fredrick's death, he returned to Prance in 1787 where he was appointed to the French Academy of Sciences (before the French Revolution) and then as the first professor of mathematics at the famous Ecole Polytechnique (after). While at the French Academy, the queen, Marie Antoinette, befriended him. It is somewhat remarkable that Lagrange kept his head during the Revolution, a time when many scientists and other intellectuals, and many friends of the royal family, were losing theirs. The last of Lagrange's famous acquaintances was Napoleon, who cultivated socially the leading French intellegencia of the time. Napoleon once said that "Lagrange is the lofty pyramid of the mathematical sciences."

Suppose some of the forces are derivable from a potential energy function V{qi. s = i. see Section 12. and orbits in gravitational fields.qn) and some are not: Qs = Q1 + QT = . (5) Studied small motions about equilibrium positions. (6. in this book he: (1) Applied his equations to solve a wide variety of problems. expressed in Eqns. probability. Lagrange collected all the known principles of dynamics and presented them in a new. One final contribution deserves to be mentioned. In addition. and his last effort before dying in 1813 was to revise his great book. (6) Recognized that the equations of motion of any dynamical system could be written in first order form (see Section 12.4 Special Forms Conservative Forces.-. (2) Gave a method of approximation for the solution of dynamics problems. He also recognized that there might be constraints which are not holonomic. 6. He was instrumental in the decision to adopt base 10 for the system. and unified form.5). (4) Asserted that an equilibrium position is stable when the system potential energy is minimum (this was later proved by Dirichlet.1).29). (7) Gave the most complete statement of the principle of conservation of energy up to that time. The result of this committee's deliberations is today's metric system. although he did not pursue the matter. sound.Lagrange's Equations 119 The seminal work of Lagrange's career was the Mechanique Analytique. algebraic equations. (8) Introduced the use of generalized coordinates (sometimes called Lagrangian coordinates). He first conceived of this book when he was 19. vibrations. including the theories of limits.33) .n (6. functions. Lagrange also contributed to many other branches of mathematics (some of which he invented). and (9) Introduced the use of multipliers to account for constraints. His interest revived in his 70's. but didn't finish it until he was 52 years old. He was somewhat of a perfectionist! For a long period in mid-life. (3) Analyzed the stability of equilibrium positions. In the Mechanique Analytique (1788). powerful.--. he lost all interest in mathematics and science (we would say he was burned-out). Lagrange was president of the committee to reform the system of weights and measures during the Revolution. numbers and arithmetic.^ + QT.

this reduces to This shows that knowledge of the function L is sufficient to derive the equations of motion for a specific system. A common type of force is the dissipative damping force given by^ Qi = ^2 ^saQa 5 a=l dsa negative definite (6. Such a function is called a descriptive function. Rayleigh's Dissipation Function. (4. ••.120 Analytical Dynamics Also note that.e.36) Such a force may be written as Q^ = I ? dqs where 1 n Q=l (6. Other descriptive functions will be introduced in later chapters. Eqn. dg^ = das)The proof is as follows: = ^ 91 + X ] da2 qaq2-\ h^ das qa <is + •• . Lagrange's equations may be written as When the system is natural (recall that this means that all given forces are conservative and all constraints are holonomic scleronomic).qn).37) n D = ^Y1 ^da^QaQis P=l (6. because V ^ V{qi. _ 5T dqs dqs Consequently.38) provided d is a symmetric matrix (i.21) gives at.

The essential idea is to find a function 4>{qi. Lagrange's equations have been stated in the minimal number of coordinates. Another example is Lur'e's dissipation function. 6. qn).'* A restriction is that cj) must depend linearly on the velocity components g^. With dissipation forces present.qi. d-sa qa + / ^ " a s 9Q — rj / J yisa ' "'as) qa — / . any set of coordinates (gi. however. ••.1). ••. to do this. General Potential Functions. n> n that. otherwise the force would be a function of acceleration components. completely define the system .-'iqn. Lagrange's equations are also derivable from the Central Principle and from the Principle of Least Action. It is possible to extend the notion of potential and dissipation functions to more general situations.t) such that a given force may be represented as either d(f)/dqs = Qf or dcp/dqs = Qf. There is no requirement.Lagrange's Equations 121 "'an Qa Qn a dsl qi+ds2q2-\ \-^ das Qa +dssqs-\ Vdsn qn dqs / . ^as qa The force form Q^ = dD jdqs is particularly useful in Lagrange's equations just as the conservative force forms are useful. Non-Minimal Coordinates. n = N — L . together with the holonomic constraints. which is not allowed (Section 2. Lagrange's equations become ( rfr \ BT BD ^ where the Q" are the components of the sum of the non-dissipative forces.5 Remarks o t h e r Derivations. One example is Rayleigh's dissipation function.q'n. Other derivations are possible as vi^ell.

--.--. Eqn. Eqn.3): ^ = 9 XI Yl «Q/3 ^a g/3 + XI ^c^Qa + C. / nil fti (T .. (6. On occasion. Usually.22). We remarked earlier that Hamilton's Principle for a conservative system. S{T . dynamic coupling makes the equations difficult to solve and should be avoided.n Thus. One holonomic constraint must be added for every excess coordinate.e. (4. " a=l . d /9T\ dt \dqsJ d 1 /A dt \a=l d dt / ^ 0. however. every term qi. it is nonsingular and therefore there exists a linear transformation of coordinates that diagonalizes O^Q. Hamilton's Principle as a Variational Qa la=l .122 Analytical Dynamics configuration (i. all components of all the particles) will suffice. it's useful to use nonminimal coordinates. Dynamic Coupling. . Recall that the kinetic energy in generalized coordinates is of the form given by Eqn. (4. this is called dynamic coupling. but this is always possible to achieve by a linear transformation of variables. There is no dynamic coupling if the matrix aga is diagonal.V)dt = 0 /" Jto can be written as a problem in the calculus of variations.qn appears in every equation. in general. Because Uga is positive definite. a=l 1 apa a=l / f^T' "" Now consider the — I -— 1 term of Lagrange's equations: at \dqs.23).V)dt 0 Jto if and only if the system is holonomic. ' + bs terms with the qs and the qs but not the qs l.

14): T-V-^^Xr{Brsqs+Br) s=l r=l dt = 0 (6.".23) with constraints given by Eqns.n Y^ dL dqs ^ dqs dq^ -s-^ dL d /dqs ^ dqs dt ydq^ y^ s dlJ_ dq'r dL dqs ^^s ^^r dL dqs ^is dq^ Y^ dL dqs . qn.t) = L'{q[.41) is only valid for the special case i = 0.e. unless all constraints are holonomic. t) Form the derivatives dL dq'j. g^. ••. dqs dq'r . such that L{qi. (6.". We now show this directly for a natural system. Invariance of Lagrange's Equations.41) J to Carrying out this variation gives a result different than Eqn. 4 . For all forces conservative. _ _ Y^ 5-^ 9qs ^ dqs dq^ \-^ dL dqs ^ dqs dq^ y^ s r=l. (6. qn. q[. Eqn.34) with Q"'^ = 0 applies and Hamilton's Principle in generalized coordinates becomes /•*! A r d fdL\ ^ h 2_] XrBr Sqadt = 0 dqs r =^ l (6.40) unless Brs = Br = 0 for all r.q'n:ty. ••. It is often stated that Lagrange's equations are invariant to coordinate transformation. suppose we attempt to formulate this problem by applying the multiplier rule to Eqn. Thus Eqn. i.40) On the other hand. s . (5. ••.Lagrange's Equations 123 We now show this directly. qi. ••. Consider a transformation of generalized coordinates qr = qr{q'i. (6. (4.

We have solved the dynamic problem by adjoining the constraints to the dynamic equation with multipliers A^. Under our restrictions for transformations. dt \dqsJ dqs so that Lagrange's equations are the same in any set of generalized coordinates. 6.L (6. and only in a holonomic system. the constraint equations may be solved for L of the velocities qs in terms of the remaining n — L.42) where qi.124 Now form d (dL'\ dt ydq'j. The class of constraints we are considering (holonomic or nonholonomic) is: n Y.q2. ir d (dqs dt \dq^ f dqs \dq'j.6 Embedding Constraints Definitions. r = 1. ^ dqs dt d fdL\ dL dqs _dt \dqsJ dqs] dq. In a holonomic system.qn is a set of suitable (not necessarily minimal) coordinates. is a nonsingular matrix (Jacobian of the transformation is nonzero).-.. L to get Lagrange's equations. dqs/dq. consequently dt ydq'j. J dq'j. Alternatively. s=l r = l. . ••. Embedding of Holonomic Constraints.--.Brsqs + Br = 0. the result . J OL' Analytical Dynamics _ dq'r E •sr^ d / dL\ dqs ^^ dL ^ dt \dqsJ dq'^ ^ dqs •^ dL dqs -^ dL d ^ s dqs dq'j. this is called embedding. we may eliminate the constraints directly.

[Bi^i^i dqe+i H \-Bin 6qn = 0 h-Bj„ 5qn] All the equations are thus of the form Bii B.^+1 •• Bin ^%+i 6q„ Sqe Be. The virtual displacements satisfy Eqn.44) .Lagrange's Equations 125 substituted into the expression for T.e+i •• Bin BSq = -B 5q Since B^ is non singular (the constraints are independent) 6q = -B~^ BSq = ASq Sqi = X! " y "^^i '•> ^= 1:")^ (6. (5.14): J2BrsSqs s=l = 0.^ Hence in nonholonomic systems we must embed constraints in terms of virtual displacements and not in terms of possible velocities. Now assume there are i nonholonomic constraints.-. E m b e d d i n g of N o n h o l o n o m i c C o n s t r a i n t s . and will be done many times in what follows.l< n Writing out the i^^ equation.n •• Bu Bf 6qi -^1. and Lagrange's equations applied. This is what was done in the first example of Section 4. Bii 6qi-{ Bii 6qi-\ \-Bii Sqi + Bi^t+iSqi+i H \-Bii Sqi = . This is only valid for holonomic systems because in nonholonomic systems virtual displacements from a possible configuration do not lead to another possible displacement. r = l.3.

the fundamental equation may be written as n J2RsSqs = 0 s=l I n Y.126 Analytical Dynamics Using the notation of Eqns. In t h e interest of simplicity. In Section 8.23). Mathematically. Notes 1 2 Again. j = i + l.Rsasj s=l + Rj=0.n (6.2 we will show t h a t .45) These n — i equations along with the £ equations of constraint give the necessary n equations for the n unknowns Qg. (6. however. > will denote t h e inner p r o d u c t of two vectors. See Rosenberg. some n o t a t i o n from set t h e o r y will b e used in this p a r a g r a p h . as a consequence of dsa being negative definite.44). Rs hs + Y. £ n n J2RS s=l Yl 3=1+1 ««i ^"ij + J2 ^i ^1j = 0 j=£+l E ( E j=e+i \s=i Rsasj+Rj]sqj=0 J But now there are no side conditions to be satisfied and the 6qj are arbitrary: i Y.-. (6. also < . we use t h e same symbol. 3 4 5 . T. t h e r e is no reason not t o allow dsa t o b e non-negative definite. these dissipative forces cause a loss of s y s t e m energy. for two different functions. RsSqs = o s=l s=i+l Substituting Eqn. Rosenberg gives an e x a m p l e t h a t shows explicitly t h a t t h i s p r o c e d u r e for a nonholonomic constraint gives an incorrect result. ..

utilize Eqn. The free length of the spring is a — r. therefore. 6/2. masslesss spring of rate fc. Z = 2az{x + y). The Cartesian components of a force are X = 'lax{y-^z). y system. 6/4. spherical pendulum. and of free length I. It is attached to a linear spring of rate k. (6. as shown. calculate the generalized forces acting on the particle. The other extremity of the spring is free to rotate about a fixed point. Y = 2ay{x + z). Using the angle 6 as generalized coordinate. Does the answer change if a < r and.15) to calculate the generalized forces arising from the gravitational and the spring force. . if so.Lagrange's Equations 127 PROBLEMS 6/1. which is anchored at a point on the x axis a distance a from the origin of the x. how? Problem 6/2 6/3. A particle of mass m moves on a smooth surface of revolution about the ^ axis (a) What is the kinetic energy in cyUndrical coordinates? (b) Specialize this result for the cases in which the surface is a cone and a sphere. A heavy particle of mass m is constrained to move on a circle of radius r which lies in the vertical plane. Using spherical coordinates. an elastic. This system is. A heavy particle of mass m is attached to one extremity of a linear. and a > r.

The three weights of mass mi. (b) Calculate the Xi in terms of the qj and show that the qj satisfy the equation(s) of constraint identically. One extremity of a heavy. rj. pulleys. m2. 6/5. ////////////////// ///// ll _ -mi Xo /^ \ qo • ^ ' X3 v_> I 5 Problem 6/6 (a) Write down the equation (s) of constraint satisfied by the coordinates xi. X2. A particle is constrained to move in a plane. and X3 shown. (c) Construct Lagrange's equations of motion. and inextensible strings shown. (L59). uniform straight rod of length 21 and mass M can slide without friction along a vertical line. (d) Dynamically uncouple Lagrange's equations when mi = 6.128 Analytical Dynamics Calculate the generalized forces for cylindrical and spherical coordinates. 6/7. and ms. The other . the QJ are generalized coordinates. What is its kinetic energy in parabolic coordinates ^. where Is the transformation one-to-one? 6/6. To solve Problems 6/7 and 6/8 refer to Eqn. 1713 = 5 . are the only massive elements of the system of weights. respectively. 1712 = I. Hence.

M3. (a) How many and which are the constraints on the Cartesian coordinates of the endpoints of the rod? How many degrees of freedom does the rod have? (b) Construct Lagrange's equations in terms of the variables 6 and if and their time derivatives. h and of masses Ml. I2. uniform rods of lengths li. M3 h Problem 6/8 (a) Which are the constraints on the Cartesian coordinates of the end points of the rods? . as shown. Three heavy. Let 9 be the angle subtended by the vertical line and the string. M2. are linked together and can move in a vertical plane as shown. ••z Problem 6/7 6/8. respectively. inextensible string of length 21 whose other end is tied to a fixed point O on the vertical line. and let (f be the angle which the yz plane makes with the plane formed by the rod and string.Lagrange's Equations 129 extremity is connected to one end of a massless.

of mass m and of radius a. but so that it remains with all its points on the table. A smooth thin ring is mounted in the vertical plane on a smooth horizontal table so that it can rotate freely about its vertical diameter. as shown. . 6/10. is free to move in a vertical plane. are generaUzed coordinates? (c) Construct Lagrange's equations. What are Lagrange's equations of motion of the rod so long as it does not slip out of the ring? A heavy homogeneous hoop of negligible thickness. A ring having the same mass as the hoop slides without friction along the hoop. A straight uniform rod of length / and mass m passes through the ring. The rod is set into motion in any way.130 Analytical Dynamics (b) Is it true that the angles 9 and (p. Determine the motion of this system for arbitrary initial conditions. 6/9.

Identify and classify all constraints and given forces. 131 . Substitute the results into Lagrange's equations. the path through the configuration space.any set that completely specifies the configuration of the system. 3. general guides may be helpful: 1. especially in simple problems.1 Remarks on Formulating Problems Procedure for Formulating Problems via Lagrange's Equations. The goal is to choose the coordinates that make formulating the problem easiest. the potential energy. Choice of Coordinates. Write the kinetic energy. and the constraints in terms of the choosen coordinates. These coordinates eliminate the holonomic constraints directly and give the fewest number of differential equations. Although there is a wide variety of dynamics problems.e. It is almost always desirable to choose the minimal number of coordinates. that is. 4. 2. generalized coordinates. the following procedure is generally applicable: 1. the nonconservative given forces.Chapter 7 Formulation of Equations 7. i. Choose suitable coordinates . The result is a set of ordinary differential equations. although this is frequently obvious. the solution of which gives the motion.

oy dT dy dT -—. y.= mz oz dT dz ^ . 3. The holonomic constraints frequently indicate the most desirable coordinates.132 Analytical Dynamics 2. z) serve as generalized coordinates: In these coordinates. -— =my . from Eqn.1) the terms in these equations are: dT . (7.= mx . if the motion is confined to be on the surface of a sphere. (4.29). they should always be considered.2 Unconstrained Particle Rectangular (Cartesian) Coordinates. -—. (7. ox dT dx ^ dT . Eqns.2). 7. (6. for example. Since n = T = 3 for a V particle moving without constraints in 3-D. spherical coordinates are indicated. the rectangular components {x. Because of the simplicity in T for rectangular coordinates.1) The three equations are d fdT\ dt \dx J dT dx F =0 dt \dz J dz Prom Eqn. T=]^m{x'' + y^ + z') Apply Lagrange's equations.

X = rcos(f) — rcpsincj) .4) Cylindrical Coordinates. ^ (dT_ Jt\di = mz Qs = Fz Qi = Fx Thus the equations of motion are mx — Fx = 0 .1): T=-m Also Qi = Fr . my — Fy — 0 .6) . Q2 = F(h . (7. z = z (7.(t). 7-1 We need to express the kinetic energy in cylindrical coordinates. Substituting into Eqn.7) y =^ r sin (j) +rep cos (f). (1-17) (Fig.5) Fig. Differentiating. the transformation equations giving the rectangular coordinates in terms of the generalized coordinates are Eqns. mz-Fy =0 (7.z).Formulation of Equations d (dT\ -—.q2. Qs = Fz (x^ + y^ + i^) = -m (r^ + r"^^^ + i^ j (7. 7-1): X = r cos ( y = r sin q z—z (7. d dt Xdij) ~""^' Q2 = Fy .Q3) = (j'..= mx 133 dt \dx J _ (dT\ __ . Choosing iqi.

3 „ dz dr jp dz „ dz ^'^^^Yz^^^Yz^^'d-z Fr = Fx cos 4> + Fy sin 4> F<j.8) If the force is given in rectangular components. d(t> dT dT .15) may be used to get the generalized force components: Ps = EFi^-^ „ dx or „ dx „ dx J-.= mz dz dT Substituting into Eqns. = -F^r cos 4) + Fyr cos 0 F. dy dr „ dy j^ dy ^ = 1^2.F^ = 0 mz — Fz = 0 (7.= mr (b . the terms are dT .134 Analytical Dynamics Lagrange's equations then give the equations of motion. -—.= mr . dr dT -2 dT o.9) . —:. (7. -—.29): _d / a r \ _dT^_„ dt\dr J dr dt\d4>) dt \dz J d<P dz ^ _ „ mr — mrcf^ — F^ = 0 2mrr(t> + mr'^ij) . Eqns. (6. (6. = F.

89 ST .4>).93) = {r. (1. 7-2): X — r sin 9 cos (j) y = rsinOsincj) z = r cos 9 We now need to express the kinetic energy in spherical coordinates.^2. Differentiating.2 al2 = mr sin'' ^^^ -|.mr9 dr dT_ = mr'^ sin ^ cos 94)^ '89 dT = 0 8T 2 2—r = mr sin 9(j) d(f) .11) The generalized force components are Qi = Fr . the transformation equations are Eqns.10) (7. Choosing (^1.= mr 9 . The partials are Q2 = Fe .Formulation of Equations 135 Fig. Qs = FA.19) (see Fig. T = l-m{x^ + f + z^) = \m{r'^ + r^ sin^ 9^"^ + r^^^) Zi Zi (7.0. 7-2 Spherical Coordinates. 9r or df —:.

{mr"^ sin^ 0^) .12) will be our starting point for a future topic .2. 7-3 .central force motion (Chapter 10).Fr = 0 (7. 7-3: y = fix.t) Fig. as shown on Fig.3 i=l Fr = Fx sin ^ cos ^ + Fy sin Osmcj) + Fz cos d Fe = FxT cos 9 cos 0 + Fj/r cos ^ sin 0 — F^r sin ^ Fs = —Fxr sin ^ sin 6 + FyV sin 0 cos d) (7.12) di' d_ : [mr^O) — mr^ sin 9 cos 6*0^ .FQ = Q d_ . Equations (7. 7.3 Constrained Particle One Holonomic Constraint in Planar Motion. Assume the constraint is rheonomic (scleronomic will be a special case). s = 1.F^ = 0 dV If the force is given in rectangular components.136 Thus Lagrange's equations give d ^{mr) — mr sin^ 0^^ — mrr Analytical Dynamics .13) Note that F^ is a force whereas Fg and F^ are moments.

DOF = N-L. Here. L = L + t. The constraint does not appear (either geometrically or as a force) because it has been explicitly removed. (7. the terms are rlT m -^ = -(2x + 2flx + 2Ut) dt total derivative \dx partial derivative holding X &z t fixed m X + 2fx{fxxX + ifxxX + fxt)x + fxX + ftt) + fxt)ft + fxiftxX dT ^ox = 1 -m{2fxfxxX^ + ^fxxxft + 2fxxftx + 2ftftx) 2 partial derivative holding X &i t fixed Substitution into the following equation then gives the equation of motion: where Fx is the a. it can be eliminated directly by embedding as follows y = fxx + ft T = ^m{x^ + f) = ^m{x^ + f^x^ + 2f.Formulation of Equations 137 Since the constraint is holonomic. y = f{x).-component of the given force.14) reduces to: m {I + fi)x + fx fxx x^ -Fx =0 (7.xft + f^) Lagrange's equation gives the differential equation of motion. This is because a nonholonomic constraint does . suppose the constraint is scleronomic.15) Nonholonomic Constraints. n = N — L > DOF. ^ > 0. As a special case. then Eqn.

This may be considered a typical robot arm. Bi2 = b .138 Analytical Dynamics not affect the accessibility of the configuration space.16) 7. and A. in rectangular coordinates. Consider two rigid bodies connected together and moving in a plane as shown (Fig. (5. (7. B12. suppose a particle is subject to a single catastatic nonholonomic constraint a5x + bSy + cSz = 0 Comparison with Eqn. mx — Fx + Xa = 0 my-Fy + Xb = 0 mz . Most nonholonomic constraints arise as constraints on velocities. . As an example. one particle in 3space subject to a nonholonomic constraint has 3 generalized coordinates. along with ax + by + cz = 0 provide four equations in the four unknowns x. z. * q2 = O2 We could choose (j) instead of 62 but 62 is somewhat easier since we need the velocities and angular velocities relative to the inertial frame {x. y.y) for use in determining the kinetic energy.29) then give.4 Example — Two Link Robot Arm Problem Definition. 7-4). M^ and Mo are motor torques. All the constraints are holonomic and there are two generalized coordinates.15) with n = 3 and i = 1 gives Bn=a . for example in problems involving the rolling of one body on another.Fz + Xc = 0 These three equations. We choose gi = 6 1 . = c Equations (6.

56) gives ^ = E {I'^i ^f+\ii ^i) First consider body # 1 (Fig. Eqn.Formulation of Equations 139 V =0 D m2. Fig.57).18) is the moment of inertia about A by the parallel axis theorem. (1. 7-5 . For two rigid bodies. 7-5). (7. (1.17) Ti = ^mi4 + \hu^l = \m£lel + \hej = \h9l where / i = Ji + mill (7. 7-4 Kinetic Energy.l2 Fig. This agrees with the alternative expression given by Eqn.

^1)23 + {hk + ^^1 cos(02 .72}.^3} be reference frames fixed in the ground. respectively. t^ =: ^2^3 . {22. (1.^1)] "h (7. and in link CD. VD = = ^Q\h + Q + ^2^3 X £223 te^ sin(e2 . j i } : VD=ilC+ where 2lD = ^rel + W X r dt ' ^c = dr.9i)h + cos(6l2 Substituting. £ = ^2^ The required unit vector transformation is J2 = sm{92 .ji} and let w be the angular velocity of {h. Use Eqn. 7-6). and {23.140 Analytical Dynamics Now consider body # 2 (Fig.J3} relative to { n .25) to relate the velocity oiD relative to {23. Let {hji}. ^rel = Q . LC dt T h e terms are •He = ^^li'2 .19) 9i)h . in link AC.73} to its velocity relative to {ii.

) T2 is then T2 = -m2vl.^i) d9i dt \d9i) = m2^^2^i^2 sin(02 — ^1) — migii sm9i — m2g£sm9i . Consequently. /2 = T2 + m2t\ = \l'Ol + \l2el + m2ii2Q\k cos(^2 . T = Ti+T2 where J' = 7i + m2t^ . Since the only given force is gravity. Equations (6.^1) (7. the constant and linear terms are missing.^l) + ^hOJ (7.22) L = T-v = li'9f + h2e^ + m2U29ie2cos{e2-ei) +'migii cos 9i + m2g{i cos 9i + £2 cos ^2) Lagrange's E q u a t i o n s .20) 2 Note that is is quadratic in the velocity components with displacement dependent coefficients. + -I2UI2 -mo n2Q2 I 1)2 tef + Ip^a2 + 2^£2^102 COS(^2 . y = -migiicosei-m2g{icos9i+£2Cose2) (7.23) ^tKwJ'm-^''^ Computing some of the terms r\ J- di[W2)~d9~2-^' ^^-'^^ -^ = l'9i + m2ii2e2 cos(^2 .34) are d (dL\ dL d (dL\ dL (7.Formulation of Equations 141 (This also could be obtained geometrically from the law of cosines.21) P o t e n t i a l E n e r g y a n d Lagrangian.

B = migli + m2g(. 7.^1)6/1 + Asm{e2 . say (x. there is one holonomic constraint (y = 0 in our case) and n = N — L = 6 — 1 = 5. Linearizing for small angles and small angular rates.9i)9l + Csin^2 = M2 where Qi = Ml.z. Recall that a rigid body in motion in 3-space without contraints has six degrees of freedom and every point of the body is specified by 6 independent parameters. ip.25) are dynamically coupled and highly nonlinear.^sin(6'2 . If the body rolls on a plane.. A thin homogeneous disk of radius r rolls without slipping on a horizontal plane (Fig. 9).9) as generialized coordinates as shown. Modal analysis of these equations for typical cases reveals two modes. .Rolling Disk Problem Definition.142 Finally. 7-7).5 Example . For the disk we choose {x. the location of some body-fixed point with respect to an inertial frame. three angles of body-fixed lines.25) l2^2 + Acos{92 . and setting Ml = M2 = 0 gives the special case of the double physical pendulum. z). C = m2gi2 Q2 = M2 Equations (7. making their solution difficult.di)02 . l'9i + Ah + B9i^0 (7.^1)^2 + ^ s i n ^ i = Mi (7.4'. . and {(f). one rapid and one relatively slow. and A = 1X12^2 . They may be easily dynamically uncoupled by a change of variables.ip. y.26) l2^2 + A9i + C92 = 0 These equations are linear but still dynamically coupled. substitution gives the equations of motion Analytical Dynamics l'9i + ^008(^2 .

Formulation of Equations 143 Fig.j.k"} disk. {i".j". body fixed with k" along the axis of symmetry of the By the definition of pure rolling.k} ground fixed (inertial). the velocity of the disk at the contact point is zero and the contact point itself moves with velocity in the {x. 7-7 Introduce the following reference frames as shown on Fig.k'} neither ground nor body fixed with k' along the axis of symmetry of the disk and / along a diameter of the disk passing through the contact point and the disk center. y) plane in the i' direction.28) (7. 3. 2.j'. therefore the rolling constraint is v_ = —ripi' = —rip[cos (j)k + sincjii) Since v_ = zk + xi. {i'. 7-7: 1. we have in component form X = —rtpsincj) dx + r sin (f> d tp = 0 5x + rsin(l)Sil) = 0 z = —ril) cos (j) dz + r cos 4> dij) = 0 6z + r cos (f) 6 xp = 0 (7. {i.27) .

7-8 Differentiating X = X — r9 sm9 cos (f> — rcf) cos 9 sin ( y = r9 cos 9 'z = z + r9 sin9 sin (b — rd> cos 9 cose .56) First consider the translational term: V From Fig.) Kinetic Energy. (A special case is rolling along a straight line. For a rigid body.144 Analytical Dynamics These are two nonholonomic constraints. now. 7-8: X = X + r cos 9 cos ( y = r sin 9 'z = z — r cos 9 sin ci = X +y +z (7. 6 = 7r/2. cj) = constant and these constraints are integrable.29) t r cos 9 • ^ — \ r cos 0 1 X- z • "^— > - » • Fig. This shows that the issue of nonholonomic constraints doesn't usually come up in two-dimensional problems. T = -mu^ + -JLO'^ 2 2 (1.

let I = Ix'x' = ly'y' . J — h'z' (7. ^. B12 = B13 = Bu = 0 B21 = B23 = B24 = 0 .32) /.'4') ^^^ comparing with the constraint equations.9. the {x .4>.) The angles {9. The nonholonomic constraints are of the form 5 (7. Figure 7-7 shows the directions of the angular velocity components: CO ~ 9i + ipk + 4>j = 9i +ipk + 0(cos 9k + sin 9j ) = 9l'+4>sm9j'+ {ip + 4> cos 9)k' f I (7.z ) axes are principle axes of inertia.34) ^ B . r = l.3O) Next consider the rotational term.y . (The products of inertia are of course all zero. = 0 . (7. ^22 = 1.33) Also F ^ m ^ r sin 6 * and L = T .36) .^ + i : ^ ' ^ " = »^ » = l. . ' The appropriate form of Lagrange's equation is: | ( f ) . <5g.29) gives if = x^ + z^+ r^(P+ r^(t>^cos^9+ 2r{-x9sm9cos(j) 145 —x4> cos ^ sin ^ + z6 sin 0 sin 0 — zcj) cos 6 cos 0)(7. Bu = l.")?5) = {x. (7. 5i5 = rsin(/).27) and (7.5 (7. ip.-. The rotational term is then = l l (9^ + ^"^ sin^ 9)+ IJU + <J)cos 9^ (7. B25 = rcos<l). 4>) are Euler's angles.2 Letting (9i.F . Lagrange's Equations.Formulation of Equations Substitution of these in Eqn.28).31) Because of the rotational mass symmetry.z. Eqns.

37) plus (7.^^^. (7. and A2. — [J('i/' + 4> cos ^)] + Air sin (/) + A2r cos cf) = Q . The two nonholonimc constraint equations in velocity form are: i + rV'sin < = 0 ^ z + rip cos 0 = 0 .56).(7.33) .6^ sin^cos9 +J(V' + </>cos0)^sin^ + mgr cos 9 = 0 . (1. — [mr^(^cos^^ + mr(—i cos 0 sin 0 — icos^cos0) +I4> sin-^ 9 + J{ip + 4> cos ^) cos 0] + mr {—x9 sin ^ sin < > / +x4> cos ^ cos 4> + z9 sin ^ cos 4> — zcj) cos ^ sin 0) = 0 . ZQ are constants. Yo. (7.z.9.146 Analytical Dynamics Combining Eqns.r 9 + mr(—irsin^cos0 + i s i n ^ s i n ^ ) + I9\ +mr^(j)^ cos 9 sin 9 + mr {x9 cos 9 cos 4> ~ x(j) sin 9 sin 0 —z9cos9s\ncf) — zcf)sin0cos (f)) — 7(.35).tp.Xi.38) are seven equations in the seven unknowns x. where Xo. PROBLEMS 7/1.37) Equations (7.30).(p. — [mz + mr{0sinOsincj) — ^cos9cosc^)] + A2 = 0 . and (7. An unconstrained particle of mass m moves in 3-space under a force F = Xoi + Yo] + Zok . --r[m. the resulting equations of motion are: -j-[mx + mr{—9sin0cos 4> — <j>cos9sin(f)] + Ai = 0 . Write the Lagrangian equations of motion in the generalized coordinates C) ^ Ci which are connected to the Cartesian coordinates by . and thus difficult to solve. They are highly coupled and highly nonlinear.

and •0 are connected to w. and z by w ~ R cos 9 cos (f . cyclindrical coordinates. respectively. A heavy bead of mass m slides on a smooth rod that rotates with constant angular velocity J7 about a fixed point lying on the rod centerline. If gravity is the only force acting on the system. If 9. ip. Let the Cartesian coordinates of a 4-space be 10. and (. A heavy eccentric disk can rotate about a fixed. Let its mass moment of inertia about the axis of rotation be I. and P lie on a straight line. A particle of unit mass moves on the surface of a four-dimensional sphere of radius R under a potential force. which is constrained to move on a smooth horizontal surface as shown. H. a. and let its mass center G be a distance s from the axis of rotation. .Formulation of Equations 147 and state why ^. X = R cos 9 sin tp . z. A massless connecting rod of length I is smoothly hinged to the disk at a point P a distance R from the axis of rotation. show that 9. G. y. y.x^ = constant. define suitable coordinates and construct Lagrange's equations of motion for this system. as shown. and construct Lagrange's equations in 9. ip. horizontal axis at O. and ( are called parabolic. rj. Problem 7/3 7/4. Denote the generalized force components by E. and tp are suitable generalized coordinates. 7/3. smooth. (p. z = R sin 9 sin ij. -q. What are Lagrange's equations of motion of the bead in suitably chosen generalized coordinates? . 7/2. and connected to a particle of mass m. and the potential energy is constant on the cylindrical surface vp' -{. and Z. O. y = R sin 9 cos ip . and ij). x. Calculate an arc length ds in terms of ^.

3. . Use Lagrange's equations to find the equations of motion of the system described in the first example of Section 4.148 Analytical Dynamics Problem 7/4 7/5.

we are faced with solving.Vn =qn.qn.-.t) = 0 fniqi.Qn.-.. Having formulated Lagrange's equations of motion for a system with n generalized coordinates.V2n. these equations are in general of the form fiiqi. r=l . Let Vi = qi.2) 149 ..1 Integrals of Motion Equations in First Order Form.qi.<in.--.(in.-.qn.1) Then the kinetic energy becomes T = T{qi.« " + ?:VB„ = 0. = l.n (8. ( 8 ^ ) .-.qi.S .qn.--.-.<ii.. We now do this for Lagrange's equations.-.qi.V2n = Cln (8.qn.qi. Vn+1 = qi.qn.t) Lagrange's equations in the qs are given by In the Vs they become dt \dVn+s.Chapter 8 Integration of Equations 8. in general.t) = T(vi.-.t) =0 It is always possible to write a system of ordinary differential equations in first order form. a set of n second order differential equations in the generalized coordinates.".".

8) If we can find all 2n of the integrals F^ and G-y.-'-.2) and (8.BrsVn+s+Br = 0.<in..t) + V{qi. n (8.--.QmQi->'.-.qn) = h = constant . (6. 7=l. Also to be satisfied are the nonholonomic constraints n '£Brsqs s=l + Br = 0. There are a total of n independent such first integrals Fp{-) = Cp.-.t.-. for systems in which energy is conserved. r = l. s = l.-.0 = l . s=l r = l.3) Equations (8.t) such that whenever {qi.4) Definition of an Integral of Motion.Cn) = C'^.-.qn.{.-. then Ff.3) give 2n first order equations that are entirely equivalent to the n second order equations of Eqns. the problem is said to be completely solved.29).7) The integrals of these n first order equations are of the form Gry{qi.-.Qn) satisfy Lagrange's equations and the nonholonomic constraints. the energy is an integral of the motion: T{qi.qnAi.n (8.14) In terms of the Vg these are n Y.-An.'. As an example.150 Analytical Dynamics These are first order equations in the Vg.-. also to be satisfied are the first order equations -~=:Vn+s.Qn. . If one can find a function Fl3{Qi.i (5.n (8.Ci.) = Cp (8.".i (8.6) is a first integral of the motion.qi.

part of the solution will be obtained in closed form (the most desirable). We note that numerical evaluation of integrals is much more efficient and robust than is numerical solution of differential equations. 8. all integrals are known but some or all cannot be expressed in terms of elementary functions. in this case. then the solution can be expressed explicitly in terms of elementary functions of time. much of analytical dynamics is devoted to this goal.Uno).robustness (numerical procedures may become unstable) . however. In the typical case. . finding closed form or quadrature solutions (in whole or in part) is desirable for reasons of: .computational efficiency (speed and storage) .--. in fact. therefore. and the remaining part must be obtained by numerical solution of differential equations (the least desirable). wliere the Ugo are the values of the Ug at a specific time to. particularly Chapter 11. A solution of a dynamical system consists of a set of functions ui = ui{t. We now find a more general form of the energy integral. part in quadratures. obtaining integrals of the motion is important. or trajectory. then we say the solution has been reduced to quadratures. such that when these functions are substituted into the equations of motion identities result. ••. the equations of motion always can be solved numerically on a digital computer. Of course. on the other hand. Geometrically. we shall give examples of qualitative analysis. Obtaining a partial closed form solution is sometimes sufficient to obtain a great deal of information about system validation . in configuration space. It is one of the major goals of the theory of nonlinear oscillations. If.Integration of Equations 151 Solution of the Equations of Motion. In later chapters. a solution defines a path.insight into solution behavior For these reasons.uio. Uno).2 Jacobi's Integral General Form. If all 2n integrals of the motion are known and can be expressed in terms of elementary functions. we say the solution has been found in closed form.-':Un = Unit. uio. The science and art of deducing system behavior without obtaining complete quantitative solutions is sometimes called qualitative integration.

••.9) where V{qi. may be written as E dt \dqsJ dqs. dqs ±-f '^ '^^ ••. All given forces can be expressed as^ _ d dV dt dqs dV dqs Analytical Dynamics (8.qn. = 0 (8. dL y-^ dL . Eqn. dV_ y ..n d f dL\ dL J-^ . g. first consider dt pUt[Ws)'-p'dqs V.12) Substituting this into Eqn. -dTt^^Ws'-^^Ws'' (8. All constraints catastatic. ••. Then virtual displacements are actual displacements and the Fundamental Equation.^ ^ _ Q .152 Assume: 1. (8.Then Lagrange's equations.13) di Now consider the total time derivative of L: dL -ir-^ dL .11) To find Jacobi's integral. (6..22). are ±(dT\_dI^_d_/dV\ dt \dqsJ dqs dt \dqsJ -Qs s = 1. (6. y-^ dL .14) .29). (8. L^L{t) 3.qn) is a generalized potential linear in the qs.11). dL ^^. . qi.. Eqns. 2. £ (^\ sr •• ^ (8.

Recall T expressed in generalized coordinates. 153 dt[^'^'Ws This may be integrated to give y ^ (Is TTThis is Jacobi's integral. (8.15) Explicit Form of Jacobi's Integral. Let T = T2 + Ti+To where T2 = . that all forces are conservative. a 13 (8.15) and note that V ^ V{qi): ^ fdT2 \dqs dn dqs dTo\ dq.18) > —^ gg = Ti s oqs _ . (6. more strongly.17) Tl = ^ba a qa To = c Thus a (8. (8.14). +V =h (8. V = V{qi).18) were used. (8.Integration of Equations Combining Eqns. Eqn.19) T^V-h 2T2 + T1+O-T2-T1-TQ T2-To + V^h where Eqns.16) aap positive definite. Now assume. (8. .3).16) into (8.= 6^ ^ dqs Substitute Eqn. dt L = h = constant (8. i.e.13) and (8.^ ^ tto^ qa qp .

When Jacobi's integral is stated in generalized coordinates (minimal implies that all holonomic constraints are eliminated in the coordinate transformation).20) being for rectangular coordinates and Eqn. (8. the transformation from the Ug to the qg is time independent: Qs = qs{ui. This is true because for a scleronomic system. The difference has to do with Eqn. Thus the energy result is more restrictive than Jacobi's result.n r = l. Let's recall the classification of constraints. all constraints must be considered and "catastatic system" means all constraints (holonomic and nonholonomic) are catastatic. we would expect that for the case of all holonomic constraints scleronomic and all nonholonomic constraints catastatic.e. the energy integral and Jacobi's integral would be the same. Ur =Ur{qi. 8-1 Conservation of Energy. in the present terms T + V = T2 + Ti+To + V = h ^.--.-.1 thiat in a closed system energy is conserved.19). ••. only the nonholonomic constraints apply and "catastatic system" means all nonholonomic constraints are catastatic.UN) . 81. i. (8. s = 1. Since scleronomic implies catastatic for a holonomic constraint.-.19) being for generalized (minimal) coordinates.iV (8.21) . When the energy result is stated in rectangular coordinates. (8.20) But this seems to be in conflict with Eqn. summarized in Fig.qn) '. We proved in Section 4.154 Analytical Dynamics constraint holonomic nonholonomic scleronomic rheonomic catastatic acatastatic catastatic acatastatic Fig.

Integration of Equations Therefore. 1. the energy integral in general does not exist. rheonomic. The transformation . a given function Fig.23) Example — Rotating Pendulum. scleronomic =^ catastatic) 2. from Eqns.4). 8-2 Since all constraints are not catastatic. in spherical coordinates. all nonholonomic constraints catastatic. Does Jacobi's integral exist? One requirement. is met. (6. r = a (holonomic.Y ^ rrir ( ^V^ 1 = 0 2Z^r "" \ dt J and thus Jacobi's integral becomes T2 + F = r + F = /t = constant (8. ^^ dur dur >Ti = 0 155 (8. The constraints are. A bead slides without friction on a rotating hoop (Fig. ^ = LJ d(j) — ujdt = 0 (holonomic. acatastatic) (p = J u!dt+ constant CD co(t). 8-2). How about the others? Choose 6 as generalized coordinate (n = N—L = 3—2 = 1).22) 1 ^=^ / dUr c = .

it can be used to establish the stability of equilibrium positions. In this problem.^ for example. = 0. Dissipative Forces. as the hoop rotates at constant speed and the bead slides on the hoop. however. this will be true only if ui{t) = constant. Study of this equation provides much useful information about the motion of the system. It will be assumed that all forces except the dissipative forces are conservative. The exception is cj = 0 which implies that TQ = 0. From . and we have found only one.e. Jacobi's integral. is not generally conserved. T. Under this assumption. This integral gives an equation of the form 9^ = f{9). A further requirement for Jacobi's integral is dL/dt = 0. i. there is a constant interchange of energy in and out of the system. energy is conserved only if a. there are four integrals of the motion required to completely solve the problem. We are now in a position to give an interesting physical interpretation of Rayleigh's dissipation function. Jacobi's integral exists and is T2 + V — To = -zma 9 — mg a cos 9 — -ma sin 9 uj = constant Energy. for which T2 + V = T + V = h = constant That is. in the same manner as the energy integral is used in Chapter 11.156 Analytical Dynamics from generalized to rectangular coordinates is (taking (f){Q) = 0): X = a sin 0 cos / wdt y = a sin ^ sin / ojdt z = —acosO Note the time dependence due to the rheonomic constraint. V. and L are V L = = ~mg a cos 6 T-V Note that there is no Ti term in T. there is a continually varying torque required to keep w = constant.

We choose cylindrical coordinates .37). Historical Remarks. the conservation of energy and Newton's Laws were thought to be independent principles. 8-3). can be traced back to the fourteenth century. (6.Integration of Equations Eqn. D'Alembert. The principle was first expressed in its modern form by Green and by Helmholtz. The concepts of kinetic and potential energy and the principle of conservation of energy pre-date Newton's Laws. then this reduces to d {T + V) = 2D dt Thus D represents one-half the rate of loss of energy by the dissipation forces (recall that das is negative definite and therefore D < 0). and Carnot recognized that the energy conservation principle was a consequence of Newton's Laws for certain special cases. dD 2D 157 Using Eqn. dqs dL_ dL = 2D Invoking Eqns.18).3 Ignoration of Coordinates Example. 8. the fundamental equation is now E dt \dqsJ dqs_ qs = J2QUs = ^D Proceeding as before.38). this may be written as If the transformation from the Us to the QS does not involve time explicitly. (6. according to Dugas. A massless spring slides without friction on a vertical wire with a particle at it's end (Fig. kinetic energy was called "living force". and in fact. For a while after Newton. we arrive at dt E^. Lagrange. The first to state these concepts relatively clearly was Huyghens. At this time. (8.

this is equivalent to dL Ps = -K^. di 21 {mr 4>) = 0 =^ mr (f> = constant Thus we have found one integral of the motion. the kinetic energy is in this case . 8-3 Using Eqn. (7.24) If F = V{qi.qn).". Definitions.7).158 Analytical Dynamics Fig.n (8. the Lagrangian is L = T-V + r -^ • + . oqs s= l. The generalized momentum components are Ps = w^\ oqs dT s = l. This motivates the following definitions.--. A coordinate qs is ignorable if L 7^ L{qs).z ) ] — 2-k{r — £) — mgz • • • — o"^Vy 2 - We see that L / i^(</>) so that Lagrange's equation for < is > / = 0 dt \d^ Therefore.n (8.-.25) This latter definition is motivated by considering the motion of a particle in rectangular coordinates.

qn) and Lagrange's equations are (8.26) dt s O ° '"'"'• " = ''•••' *'•'" Equations (8. Define the Routhian function as: B=l ^^^ . ••. (2) conservative. /5 = 1.Integration of Equations and thus the x component of the generalized momentum is dT Px = ^T = mx ox 159 which is the a.= P3 = CB = constant .6 < n are ignorable and that the other qg are not. 6 (8. /S = 1. Continue with the same assumptions.-. Suppose g^ . qu--.29) Routhian Function. and that (3) dL/dt = 0. Then L = L{qb-^.-component of the linear momentum. Lagrange's Equations with Ignorable Coordinates. Assume that the dynamic system is (1) holonomic.qn.27) give h momentum integrals wr. the Lagrange equation for qg is dt \dqsJ so that Ps = constant This integral of the motion is termed a momentum integral.". If a system is conservative and holonomic and QS is ignorable.i.

--.n oqa oqa dqa dqa Putting Eqns.35) into Lagrange's equations gives d f dR\ dR .".26): Analytical Dynamics ^E^/^'^C'^ + /3=1 ^ ^ ^ . (8.qn.--. Equations (8. 8-4).-.qn. (8. 6 and substitute into Eqn. the result is R = R{qb^i. (8. dL dR dL dR .r = 1. Example. -^^ = ^^ . 6 (8.33). •-.\k{y . a = 6 + l.E #^^" (8-31) If we now solve Eqns. Choose generalized coordinates (x. (8. qb+i.37) The Routhian R is in effect the Lagrangian for the nonignorable coordinates.34) .y). Ci.32) ^R= E ¥^^1'^+ E dR dCp ' ^^<ic + i:^^Cp (8. .b (8. the Lagrangian is L = T-V = ^mii.31) and (8.2 + i ^ 2 ( i + yf .160 Take the variation of R.29) for the q^ as functions of the Cr] P. (8.Cb) Taking the variation of this function.^ ^E+ i -Q^^'i^ E ^^/5 /3=1 j Cga -Ef^%. -^— = ^—. . (8. (8.35) .. . /3 = l .if . /3^1.--.33) Equating the coefficients of Eqns. Two masses connected by a spring slide on a smooth surface (Fig.„^^. recalling Eqn.30).34) lead to h integrals of the motion: q0 = l ^ d t + Kp.

i+m. ._ dy y + ky = ki The Routhian approach removes all traces of the ignorable coordinates from the equations of motion.36): C X = -m2y + 7712 mi R = C-m2y^ 1 (C-m2yV G — -mi m.2 mi +7772 _dB. •^ 6 = — = mix + m2{x + y) is one integral of the motion.i.2 2 \m. (8. the one remaining equation of motion is found by substituting x for C in i? and applying the equation of motion. (8.30) the Routhian is R=y2^^^^~^ = ^^~ o"^i^^ ~ o"^2(i + yf + 7. neither would be ignorable. Since y is the only non-ignorable coordinate.) From Eqn.29) ^ ^ ^ • / .i + m. x is ignorable. . Eqn. 8-4 Since L ^ L{x).^f where from Eqn.2) are selected as coordinates.Integration of Equations 161 X2- X- ^ •y- mi k —AAA/^— m2 H Fig. (Note that if (a.2j 1 (C-m2y . 7712 Hy 2 ^1^1+^2 +\Hy-i) d_ /dR\ dt \dy ) mi77T.a.Ky . (8.

.o dwg Vs{qs)qs + 2 dqs Qs +-r~ 7:-r~ dqs Since each coordinate appears in one and only one equation. Liouville Systems. It can be shown that for systems with energies of the form T = 1 .39) . (6. \dvs . if all the forces are either conservative or dissipative damping (see Section 6..40) a=l that Lagrange's equations are also separated. and no other coordinate appears in this equation.".^ Notes 1 2 3 For example.38) (Note that only the quadratic term is present in T in a scleronomic system.) Substitute Eqns. holonomic.35): dt \dqsJ dqs ' = ^'^ s = l.n (8. scleronomic) system with kinetic and potential energies: -in s=l n (8. equations of motion can be uncoupled by a transformation of variables.4 Separation of Variables Uncoupled Systems.4). Each equation can be integrated independently in this case. but not all. Some. .cx=l (8.. See Rosenberg or Pars.162 Analytical Dynamics 8. See Pars. (8. the system is uncoupled and the variables are said to be separated. Eqns. Assume a natural (conservative.38) into Lagrange's equations.

The system of Problem 7/2. Examine the same question as in Problem 8/1 when the point of support is fixed. Discuss the existence of integrals. uniform rod is constrained to move in the vertical plane. use the constants of the motion to reduce the solution to quadratures to the extent possible. A heavy. but the length of the pendulum changes in the prescribed fashion f{t). Examine the same question when the point of support is moved in a prescribed fashion f{t) along the horizontal a. The system of Problem 6/10. and one of its extremities is constrained to move on a horizontal line so that its distance from a fixed point on that line is a prescribed function f{t). . 8/4. 8/8. 8/5. The systen of Problem 6/7. 8/3.-axis. Examine the same question as in Problem 8/1 when the point of support is moved in a prescribed fashion f{t) along the vertical z-axis. 8/6. 8/2.Integration of Equations 163 PROBLEMS 8/1. The simple spherical pendulum with fixed support and constant length possess an energy and a momentum integral. In the problems that follow. Are there conditions such that the system can rotate about the z-axis with 9 = constant 7^ 0? The system of Problem 6/9. 8/7.

This is an example of a multi-body problem with nonholonomic constraints. A two-wheeled cart of dimensions shown on Fig. 9-1 165 . — u~~. 9-1 moves in a horizontal plane.Chapter 9 Examples 9.1 Street Vendor's Cart Formulation. path Fig.

166 Let: TTic = niw = Ic =



C A Ic

= = =

mass of the cart without wheels mass of each wheel mass moment of inertia of the cart without wheels about a vertical axis through point C on the axis of rotation of the wheels mass moment of inertia of each wheel about the axis of rotation mass moment of inertia of each wheel about a diameter I + rricS^ (parallel axis theorem)

Constraints. Three rigid bodies in 3-space have 3 x 6 = 18 coordinates A'^. But by inspection, there are 5 generalized coordinates, say {x,y,9,(f)^,(f>r)- Therefore, there must be 18 — 5 = 13 = L holonomic constraints, but we need not consider them if we choose generalized coordinates. There are three other (nonholonomic) constraints: (1) Velocity tangent to path. From the figure, v_ = xi + yj = V cos 6 i + V sin 0 j X = vcosO; y

y = vsin9 cos 0 y — sin 9 x = 0

V sin 6
V cos U

Thus the constraints on the virtual displacements are cos0 5y-sin9Sx =O (9.1)

Recall from Section 2.1 that in the dynamics problem of the second kind, the motion is specified and the forces are to be determined. Thus if the path of the cart is specified, say by y = f{x), Eqn. (9.1) contains only two independent variables and the constraint is holonomic. In the problem of the first kind, the forces on the cart are specified and the path is to be determined. In this latter case, which is the case of interest, the constraint Eqn. (9.1) is nonholonomic. (2) Left wheel rolls without slipping. Using the relative velocity equation, Eqn. (1.25),
VD=VC + Vvel + W X r

Examples The terms are "^D = (pl f i (no slip condition)


vc = xi + fj ; ^Zrei = Q w = ^fc ; r = bj i k Therefore (j)(^r i — i;(cos ^ i — sin 0 j ) + y(sin ^ i + cos 0 j ) + Q + ^fc xhj Equating components gives 4>ir = X cos 0 + y sin 9 — Oh 0 = —i; sin ^ + y cos 0 The second of these leads to the same constraint as Eqn. (9.1) but the first is an additional nonholonomic constraint: - cos e5x~ sin 6 5y + h5d + r Scpt = 0 * (3) Right wheel rolls without slipping. Proceeding as before, (pr r = X cos 0 + y sin 9 + 6b so that -cos9Sx-sm9Sy-bS9 + r5(t)r=0 (9.3) (9.2) = cos ^« — sin 9j = k' j = sin di + cos ^j

The three nonholonomic constraints are of the form


+ Br = 0;

r = 1,2,3


Therefore, taking iqi,q2, Qa, QA, Qb) = {x, y, 9, (pi, cpr), comparison of Eqns. (9.1) - (9.4) gives: Bii = — sin^ , Bi2 = cos9 , B13 = Bu = B15 = 0 B21 = -cos9 , B22 = - sin0 , B23 = b , B24 = r , B25 = 0 (9.5) B31 = -cosO , S32 = - sin6' , B33 = -b , B34 = 0 , B35 = r


Analytical Kinetic and Potential Energies. From Fig. (9-1), X =^ X + s COS 9 ; y — y + s sin 0


Forming T of the cart from Eqn. (1.56), Tc = = ^mew2 + l 7 a ; 2 ^ i m e ( ; ^ ' + y ' ) + ^T^2 7:rnc \x^ + y^ + s^O'^ - 2s0 {x sin ^ - y cos 61)1 + ^(/e-m,s2)^2 T of the left wheel is: X(^ = X — bsinO , Tt = = \m^{xj yi = y + bcos9 (9_g)

+ yj) + \c4>j + ^A9^ 2b9 {xcos9 + y sin61) + 6^^^] (9.7)

i^rriyj [i;^ + f + IA9'+^-C^J

For the right wheel, replace b by —b in this equation. The total system kinetic energy is then T T = = Tc + Te + Tr
1 1




+4<^(<^r+<^l) 2where



= 1^ + 2m^fe2 + 2A

Consider the gravitational forces; are they constraint or given? They are constraint because they do no work (either actual or virtual). Consequently, y = 0 (9.9)



F=Xi+YJ F=Xi+Yj

Fig. 9-2 Lagrange's Equations. The applicable form of Eqns. (6.29) is

d fdT\
dt \dqs)


-Qs + J2^rBrs ^^^

3 ^

= 0;

s = l,-,5


Now assume that the cart is pulled by applying a force to the end of the tongue. The equivalent force system consists of the force acting at the center of mass plus a moment (Fig. 9-2). Thus let X, Y, M be the (x,y) components of the given force and the moment acting on the cart. Then from Eqns. (9.5), (9.8), and (9.10) the equations of motion are mx — rric s f^sin^-l- ^^cos^j — Ai sin^ — (A2 + A3) cos^ = X my + 'mc s \0sm9 — ^^sin^j + Ai cos0 — (A2 + A3) sin^ = Y -mcs{xsm9-ycose) C(i)r + A2r = 0 C(j)i + A3r = 0 + ie + b{X2-X3) =M (^-^l)

These five equations, along with the constraint equations in velocity form, give eight equations in the eight unknowns {x, y, 6, 4>i, (pr, Xi, A2, A3). By a great deal of manipulation, including transformation to pathwise coordinates and eliminating <;6r,(/'^, Ai, A2, and A3, we get^ Du — rric s6^ = F(u, 9, t) = force component tangent to path. (9.12) J9 + rric s9u = M{u, 9, t) where u = path length (see figure) D = m + 2C/r2 J = 1 + 2bC/r^ These equations are dynamically uncoupled but still nonlinear. To integrate Eqns. (9.12), we need initial conditions u(0), u(0), 9{0), 9{0), and functions F{t), M{t).





A Useful Identity

Lemma. Any equation of motion of the form fi{q)q + f2iq)h{q) =0 (9.13)

can always be reduced to quadratures by the identity

q = qf
dq Proof of Identity. dq dq dq dt dt dq .. 1 q


Proof of Lemma. Substitute Eqn. (9.14) into (9.13) and integrate: fi{q)q^ + f2{q)Mq) =0

f q ,.

h{q) f h{q) , ^

Fi{q) = F2{q) + ci

^ = F3b,c,) = |
dq + C2 F3{q, ci) t = F^{q, ci)+C2 q = F^it, ci, C2)
/ " * = /

Initial conditions q{0), q{0) give ci and C2:

0 = i^4(g(0),Ci)+C2 t = Fe {q, g(0), m) q = Fr{t,q{0),q{0)) Thus the solution has been reduced to quadratures. This technique is the same as applying the integrating factor q to Eqn. (9.13), and the resulting first integral is often equivalent to the energy integral.

Examples Example. Suppose /3(g) = 1; that is, Eqn. (9.13) becomes fi{q)q + f2{q) = 0 Applying Eqn. (9.14) and integrating j q dq = F2{q) + ci = -q^
+ | = VSb) Cl



dq ^2[F2{q)+ci where

+ C2

9.3 Indian Rope Trick

f2{q) dq {q)

Problem Formulation. A uniform inextensible rope of length t is thrown into the air bent double vertically. The initial speed of one end is «o and that of the other is UQ (Fig. 9-3). A particle of mass m (a boy or a monkey) climbs along one side of the rope with constant speed v relative to the rope, starting at the end of the rope at i = 0. Let p be the mass per unit length of the rope. In terms of x, y, and z, T and V are 2 P • x)x^ + {z- y)y^J + -m{x + v)"^ 1 1 x)2(^ + x) + iz-y)~{z 1 + y) + mg{x + vt) (9.16)

V = pg


TT xx+vt
Fig. 9-3

T T y=u+X (9.u). Since all given forces are conservative (gravity) and all constraints are holonomic. (6. T=j \2ii?. dx~A dx \Ux + 2u[i .21) where qi = x.20) V = pg Lagrange's Equations. u where u = y —X (9.35): dt \dqs dL_ dqs 0.22) .+ 2ux{t . and q2 = u. 1.19) In terms of {x.« ) + u{i -u)-u)-u'^ -ii^ = 0 (9.2 (9.u)\ + m{x + v) l-mg ^[2xii-u)+2uii-u)] 4 [—2ux — ii — pg aL _ _ pg ^ dL du = so that the equations of motion are {x + g){pi + m) + ^ \il{i {x + £ ? ) ( £ .172 The constraint is Analytical Dynamics {z-x) + {z-y) =i (9. + mg{x + vt) + ~pg^ xi + --iu —-u^ (9. the suitable form of Lagrange's equations is Eqn.17) We pick generalized coordinates x.u)] + -m{x + vf 1 1 .u) + v?{i .18) so that i u -^ = 7 + T + ^ . Computing the partial derivatives.

9-4). initially (Fig.14): /i = dji djj. Let the initial conditions be: .n) [/i. and the second is a rescaling of the initial velocity for convenience. (9.(l . (9. Noting that Eqn.«) .23) is in the form of Eqn.24) ^Jl + 2M The first of these is the condition that the rope is bent in the middle. «• m F ig . (9. For convenience.M) .A'] = (1 + M) [2/i(l . we apply Eqn.(0) = ^ = ^ ( ° ^ =0 (9./i'] This is a nonlinear.13).22) only in the combination {x + g). 9-4 . introduce: )[/ = M = — = dimensionless distance — = dimensionless mass pi 173 Noting that variable x appears in Eqns. (9. second order.Examples Analysis. we eliminate {x + g) to get {1 . ordinary differential equation in variable //.

which keeps the velocity finite.M ) V / ( 1 . It can also be shown that the tension in the rope —^ oo as / — 1.(1 + Mf 2 „ .i ( ^ ± ^ ) MVI + 2M + (1 + Mf sin"^ M M + 1 (9.)d^i /i)(l + 2M + /i) A*o M v / ( l ./ i ) ( l + 2M + /i) (9.25) Variable /i ranges from 0 at t = 0 (rope bent double) to 1 at some time ti (rope becomes straight).) + (l + M ) 2 s i n . Analytical Dynamics 1 2V1-/" 1 + 2M + II M M + fi dfj./ i ) ( l + 2M + /i) _ r^' ff' dfi 7/i(0) A io (1 (M + ij./ / ) ( ! + 2M + /.l sin M M +1 (9. this rising tension brings the > U > elasticity of the rope into play.26) The time ti when the rope straightens out is found by putting /j. We see that lim • lim /^^l M O V ( l .MVl + 2M .x{0)] (9. In an actual rope. This explains the crack of a whip. djj.28) . Integrating Eqn. „ . The crack occurs when the tip speed exceeds the speed of sound. = 1: fioh = 1 (1 + Mfn . y " ^{l-|J){l + 2M + |Ji)d^l = iioj dt fiot = ^ | ( M ./ i ) ( l + 2M + Ai) oo Thus the speed of the rope tip becomes infinite as it becomes straight.174 Substituting and integrating once: /" djj. M = 0 and 2/io 2/i(0) 2l[y{0) . (9. ( l .25).27) If there is no child or monkey.

Let the initial velocity of one extremity be UQ. as appropriate. e. as shown. and UQ ^ VQ. use Lagrange's equations to derive the equations of motion. in particular. Problem 9/3 The given forces acting on a particle of unit mass are derivable from the potential energy F r. 9/3. and use the integrals and the identity (9. explain what happens when the rope straightens out.+ . are interconnected by a massless. In the problems that follow. Problem 9/1 9/2. as shown. PROBLEMS 9/1. respectively. if) = f{r) + —3. and (p are spherical coordinates. 0. Pi is constrained to move on the surface of an inverted smooth rigid circular cone. both are vertically up.) Examine this system for the exsitence of an energy integral and of momentum integrals and reduce the solution to quadratures to the extent possible. A uniform heavy rope is thrown in the air bent double. (The string passes through a smooth hole in the apex of the cone. Two heavy particles Pi and F2 of mass mi and m2. to reduce the solution to quadratures to the extent possible. Show that Lagrange's equations can be reduced to quadratures. Discuss the motion.Examples 175 Notes 1 See Rosenberg for the details.2 r^ r^ sm 0 where r. that of the other VQ. The other particle is constrained to move on a vertical line.14). . as shown. investigate the existence of integrals. inextensible string of length I.

radius r. and the unrestrained length of the spring is (. 9/7. The mass of the string per unit length is p and two particles of mass mi and m2 are attached to the two ends of the string as shown. A homogeneous disk of mass m. and uniform thickness is attached to the end of the rod and can rotate freely with respect to it. Obtain a formula for the velocity of the string as mi leaves the table. Initially. Analytical Dynamics The flyball governor shown has four rods that are pivoted at the fixed point O and at each mass in such a way that..176 9/4. Obtain the equations of motion of the system by Lagrange's equations. The moment of inertia of the wheel about its axis of rotation is / . the mass M moves smoothly in the vertical direction along rod OB. (/rod = tk"^-^^ ' -^disk = \mr'^). Problem 9/4 9/5. Use coordinates 9 (angle of rotation of the wheel) and x (elongation of the spring). An inextensible string of length L lies on a horizontal frictionless table. A flywheel rotating in a verticle plane about its center of mass O carries a particle of mass m that can slide freely along one spoke and is attached to the center of the wheel by a spring with spring constant k. a portion of the string of length I hangs over the side of the table and the string is at rest. The moment of inertia of mass M about the OB axis is I and the masses m may be modeled as particles. 9/6. . Problem 9/5 A uniform rod of length L and mass m is pivoted at one end and can swing in a vertical plane. as the masses 771 move outward.

Particles mi and m2.2). Problem 9/9 A massless circular hoop of radius a rotates about a vertical axis with given constant angular speed uj.Examples 177 Problem 9/6 9/8. The particles move on a frictionless horizontal plane such that the motion of mi. A particle of mass m slides on the hoop without friction (see Section 8. . are connected by a massless rod of length i. Same as Problem 9/9 except that the hoop is free to rotate. is confined to a fixed frictionless track of radius R. each of mass m. 9/10. Qco Problem 9/8 9/9.

Chapter 10

Central Force Motion
10.1 General Properties

Lagrange's Equations. In Section 7.2 we derived Lagrange's equations for an unconstrained particle in spherical coordinates. In central force motion, the only force on the particle is directed towards the origin of an inertial frame, depends only on r, and is conservative (Fig. 10-1). Thus: Pr = Fr{r) = - ~ = dV_ dr (10.1)

Fig. 10-1 Substituting into Eqns. (7.12): mf — mr sin Ocf) — mrO"^ + ^^ = 0 dr


d_ „2h {mr 9) — mr sin 9 cos 9(f) = 0 {mr ,2„- 2 aj. sin' 179





From these we see that (p is an ignorable coordinate with corresponding momentum integral: mr"^ sin^ Ocj) = k = constant = p^ = (f) component of generalized momentum (10.3) Solving this for (f) and substituting into Eqns. (10.2): mr r ^— — mr0 H—-- = 0 mr'^sirrO dr d . 2/,N k'^cosd — {mr^e) 0-: = 0 dt mr-^ sin 9

f i n 4^1

Since these are independent of <^, the motion takes place in the (r, 0) plane (this agrees with intuition). Hence, from Eqn. (10.3), (j) = constant = ^ ^ = 0 ^ A ; = 0 and Eqns. (10.4) become mr — mrd -I—-- = 0 dr j{mr^)=0 This gives another momentum integral: mr 6 = K = constant = pe = 0 component of generalized momenta (10.7)
A2 dV



This may be used to prove Kepler's second law; consider the area swept out by the position vector in time dt (Fig. 10-2): dA dA -— dt = = -r^de 2 I odd IK -r^—- = = constant 2 dt 2m -dA •de Fig. 10-2

,_„, (10.8) ^ '

Central Force Motion


where Eqn. (10.7) was used. Thus "the area is swept out by the position vector at a constant rate."'^ Next eliminate 6 between Eqns. (10.6): mr - ^ + ^ = 0 (10.9) mr'^ dr We see that this is in the form of Eqn. (9.15); using the identity r = r dr/dr: .dr mrdr 1 .2 -mr K^ dV ^ ^ + —— = 0 mr'^ dr K^ r dr -T +


dV = h = const.

The constant h is in fact the total energy, as can be seen by forming T + V. Reduction to Quadratures and Orbit Equation. Solving Eqn. (10.10) for r, 2 f, „ K^ r = W — 1/ m ( ' ' - ' ' - 2 S J ) = ^ M Thus f depends only on r and not on 6. Integrating,


This gives r{t)\ 9{t) is then obtained from Eqn. (10.7): rt K r K ho —7:dt / mr


The orbit equation comes from eliminating t between Eqns. (10.7) and (10.11) to get an equation in (r,5). We have dr dr dd ^^'dt^de'dt^ K dr mr^ d9 " •'^^'




It is convenient to make a change of variable from r to u, where, u = - ; du = ^dr (10.15)

Now restrict F{r) to be a power of r F{r)=ar'' so that Vir) = Then Eqn. (10.14) becomes


F{r)dr = -^^—^





'"A/^ + ^ ^ I I ^ -

It has been determined that this will have a solution in terms of trigonometric functions when, and only when, n = +l,-2,-3


Inverse Square Forces

Gravitation. Equation (1.36) expresses Newton's law of gravitation. If me 3> m \ then m^ may be regarded as fixed in an inertial frame and m has central force motion about m^. Equations (10.16) and (10.17) become F{T) = - ^ ; V{T) = -^ (10.19)

where JJL = KrUe is the gravitational constant of the attracting mass and F{r) is the force per unit mass of the orbiting body (Fig. 10-3).^ In terms of u, these equations are F{u) = -iJLu'^ ; V{u) = -nu (10.20)

The orbit equation, Eqn. (10.18), becomes «-e„ = ^ r

^h +

f l U - ^


Central Force Motion


Integrating and substituting u (10.22)

^ ^ ( - \ / - ^ - < - ^ '

where 0 is a constant of integration. The equation of a conic section in polar coordinates with one focus at the origin is

- = c{i + ecosie-e')')


Comparing Eqns. (10.22) and (10.23) we see that the orbit is a conic section, and that C = mn e= \l + 2hK^ (10.24)

where e is called the eccentricity. Prom the theory of conic sections: h > 0 => e > 1 , h = 0 =^ € = 1 ,
m^ <h<0 2i^2

hyperbola parabola 0 < e < 1 , ellipse (10.25)

h =


e= 0


Since the only one of these paths that closes on itself is the ellipse (with the circle as a special case), we have Kepler's First Law: "the planets travel around the sun in ellipses" (Fig. 10-4). Period of Elliptical Orbits. From Kepler's Second Law, Eqn. (10.8), the area of an ellipse is A IK = 1 dA=r'4dt=r 2 m JA Jo at Jo ,






Fig. 10-4

Fig. 10-5

where r is the orbital period, that is, the time for one complete revolution. But for an ellipse.

A = nab 1 1 a= C 1 (10.27)

where a and b are the semi-major and semi-minor axes, respectively (Fig. 10-5). Combining these results, the period of the orbit is

2mA K

2m,7rab 2m7ra^\/l — e^ K " H (10.28)

2m7r ,3/2 7=^

Thus "the square of the period is proportional to the cube of the semimajor axis"; this is Kepler's Third Law. Remarks: 1. Kepler's First and Third Laws are valid for inverse square force only while the Second is true for any central force motion. 2. Newton started with Kepler's Laws and arrived at the inverse square law of attraction, while we have done the reverse.

3 The Time Equation Remarks. fromEqn.29) K KC^il + ecosO*)^ where 6* = 9 — 0 is the angle from periapsis and is called the true anomaly.8): dt = "^r^de = ^ 2 d A (10.30) K K At this point we could use substitutions and integral tables to evaluate this integral to get the relation between 9 and t. (10. this has the advantage that it gives a geometric interpretation of the quantities involved. Time Equation for Ellipse. The derivation for hyperbolic orbits is similar. (10.23): K r 9 = — = constant m Thus dt m m c{i + ecos{e-e' dO (10. We will do this for an ellipse by a graphical method similar to that first used by Kepler.7) and (10. Also. labeled 0. elliptical orbit 0 circumscribing circle (apoapsis) A Fig. Instead we use a geometrical construction. for example. Recall Eqns. for rendenzvous problems. 10-6 Figure 10-6 shows an elliptical orbit about an attracting center at one of the foci. This is important. Determining the time between two points on an orbit takes one more integration. and the circumscribing circle tangent to the ellipse at apoapsis (maximum distance from 0) and at periapsis (minimum .Central Force Motion 185 10. and 0 < e < 1 for an ellipse.

32) for E and then Eqn.34) for e*. Q is the position of the orbiting object. Given t. Given 6*. using Eqn. solve Eqn.34) for E and then Eqn. 10-7: oe-I-r cos 0* = a cos £^ It can also be shown that (10.33) Use of E q u a t i o n s . . The circled dimensions are properties of an ellipse not proved here. 2. Therefore.32) for t. (10.23) the radius at periapsis is r. (10. These equations are of practical use in two ways: 1. (10.186 Analytical Dynamics distance from 0).-esin£. From Eqn. = re. iteratively solve Eqn. From Fig.t r i a n g l e O C Q ' ) K a 2 n 2p 1 ^ .triangle OCQ) ^ (sector P C Q ' .32) = eccentric anomaly (see Fig. (10. (10.) = j^ (10. Kepler used the fact that an ellipse is a "squashed" circle in the ratio b/a. m \b ^ where E M ty (£.31) Thus 9* is the angle from periapsis. 10-6) = E — e sin E — mean anomaly = time at periapsis Some Identities. t-tp 2m = — = (sector PCQ.=o = ^ ^ ^ (10.30). (10.

Find the equations of motion if the particle has some nonzero initial angular velocity. Problem 10/1 10/2. Same as Problem 10/1 except that the string has mass per unit length p. This will be approximately true for the motion of the planets around the sun and for satellites around the earth. The string passes through a hole in a horizontal table. 10-7 Notes 1 2 Kepler deduced this from observations of the orbit of Mars. The particle of mass M lies on the table. . PROBLEMS 10/1. A particle of mass M is connected by an inextensible string of length £ and negligible mass to a particle of mass m.Central Force Motion 187 orbit ]r*—a c o s E Fig. and the other particle moves on a vertical line beneath the hole.

. The spring resists a length change u in the amount f{u). these forces act along a line connecting the particles. is connected to a spring under the table by means of a massless inextensible string which passes through a hole in the table. There are forces equal in magnitude and opposite in direction acting on the particles. A particle of unit mass is attached to one extremity of a linear spring. Use Lagrange's equations to get the equations of motion and use constants of the motion to reduce the solution to quadratures. Use Lagrange's equations to obtain the equations of motion. 10/8. verify the formulas (a) a(l — ecosE) = r (b) c o s ^ = (c) smE = (d) tan e + cos 9* 1 + e cos ( 10/4. 10/7.188 10/3. find the equations of motion and reduce the solution to quadratures. A particle is attracted to a force center by a force which varies inversely as the cube of its distance from the center. Analytical Dynamics Two particles of mass mi and TO2 move on a smooth horizontal plane in such a way that the line connecting their positions passes for all time through a fixed point in the plane. Find the equations of motion using Lagrange's equations. 10/5. If the particle is set into motion in any way that stretches the spring initially. ^ TT^sinr 1 + e cos 9* ^ 2 = "1-e 1+e tan — 2 U^ . where f{u) is of class C^.. lying on a smooth. When the spring is unstretched and the string is taut. For elliptic orbits. the other extremity being hinged at a fixed point in an inertial frame. A particle of mass m. Same as Problem 10/3 except that the mutual forces on the particles are due to a linear spring of stiffness k that connects the particles. Show that finite orbits are closed. the distance between the particle and the hole is /. . and find all momentum and energy integrals of the motion. horizontal table. 10/6.

say B. 11-1). The velocity of a typical mass particle of the rigid body is given by Eqn. Consider the motion of a rigid body such that one of its points.1 Rigid Body Motion with One Point Fixed Kinetic Energy. 11-1 189 . (1.25) as (inertia!) Fig. Let {i.k} be body-fixed principal axes of inertia with origin at B.Chapter 11 Gyroscopic Motion 11. is at rest in an inertial frame (Fig.j.

4) . Wj-ei — Qj and u di Thus = = ooxi + ojyj + oJzk Xii + yij + Zik Analytical Dynamics = {u)yZi . Then these holonomic constraints are: x = {).Wj^ajj) fc Consequently. y = 0. the kinetic energy of the rigid body is y t + [yf + xlj ul . all the products of inertia are zero: ^xy ^^ / J 'fT^iXiyi =^ U i hz = ^ rriiyiZi = 0 i •l-zx ^^ / ^ frgZiXi = U E u l e r ' s A n g l e s . choose the origin of the inertial frame as point B. = \[lx<^l + IyU^l + h^t) (11. z = 0.2xi Zi w^ u^ .j.k} are principal axes.2) where the moments of inertia are i Iy = J2rniiZi+x1) i i (113) Note t h a t because {i.ojzyi) i + {uJzXi . The motion of a rigid body without constraints is described by 6 generalized coordinates.2zi yi ujy co^ .ujxZi) j + (w^yj . Motion with one point fixed imphes three holonomic constraints.190 where v ^ = 0. (11. Without loss of generality.2yj Xi oj^ W.

the components of w. in terms of 0. Then rotate about i = i by ^ to get 3.j.ip. 11-2): 1.j. Finally rotate about k = k by ip to get {i". we need to write Wj. 6. 2. k} to line up with any arbitrary body-fixed axes./. 9. 11-2 The three generaHzed coordinates left are orientations or angular measurements. cuz. in body-fixed principal axes. ip. we can get {i. By a suitable choice of 4>. the angular velocity of {i. {/.k"}. K} is u 'k' +9t" +tpk (11.7) . k} w. cos ipi — sin ipj 3 cos ipj + sin ipi (11. Thus {(p. cOy.k' xp Fig.j".6) k = cos 6k + sin 6j Thus + (J)cos9 + 7p^k (11.k}.K} (Fig.5) where. 11-2. For the kinetic energy.Gyroscopic Motion 191 K. from Fig. J. {i. ip) serve as generalized coordinates. First rotate about K = k by 0 to get {«'. j . 11-3.. A common choice is the Euler angles} Now consider three reference frames. From Fig. obtained by rotations relative to the inertial frame {/. J.r. k'}.t. 0.

Now suppose the rigid body is axially symmetric and is spinning about its axis of symmetry.k} be body-fixed.8) 11. I^ = Iy\ let h — y — ' 2 — J (11. we will consider the following special case.J. (11-8) into Eqn.8). (11.2).2 Heavy Symmetrical Top Formulation. and {i. Instead of doing this in the general case. (11. (11.2) then gives the kinetic energy. Fig. The angular rates in this case have been given names: (p — precession 9 = nutation "ip = spin Because of the symmetry. 11-4.9) Then.K} be inertial. Let frame {I.192 Analytical Dynamics k"v A k' Fig.10) . and (11.j.9): T = \[l[ujl + u>l)+Jut\ 1 I (9^ + ^'^ sin^ 9^ + J (^iP + <j) cos 9y 2 (11. 11-3 so that Ux = 4> sin 9 sin ip + 9 cos tjj iOy = (j) sin 9 cos ip — 9 sin '0 Uz = (p cos 9 + ip Substituting Eqns. from Eqns. (11.

For a natural system with three generalized coordinates.35): dL_ dt dqs Computing partials: dL d9 dtp dL.11) Lagrange's Equations.12) oe d(f> dL. (6. V = mgi cos 6 (11. line of nodes Fig.Gyroscopic Motion 193 plane perpendicular to axis of symmetry.2. = J (ii) +(pcosO) = L(p sin0 cos 0 — J yp + 4>cos 9\ (f)sin9 + rngisin9 dL = 0 dip L4)sm^B + J (ip + </)cos6') cos^ dL dqs = 0. the proper form of the equations is Eqn. s = l. = 0 .3 (11. 11-4 Also. .

•_ ^~ Pi-P2COs9 Ism^9 (Pi c o s ^ .2a + I .16) E n e r g y Integral.P2) .14).13) = Q$ = moment due to gravity d dt d dt /</)sin^0 + j(4> + 4>cose) cos0 J Up+ 4)cos0\ = 0 We see t h a t 0 and tp are ignorable with momentum integrals J {ip + 4>cos Oj = p^ = P2 = constant (11. we now have 3 integrals of the motion.15) P2 '^ = — .17) This is a first integral of Eqn.^^ ~ ^ y ° ^ I sm 9 with.15).13) gives a second order differential equation in 9: 19 .194 Lagrange's equations are then — ( l ^ ) . Since the system is closed (catastatic and potential). I Pl~P2 COS u X . Eqn. T + V = h = constant 52 . (11. (11. (11.\ J fPi-P2COs9\ . (11.14) I(j) sin^ 9 + P2 cos 0 = P(f. may be written in quadrature as t j f{9)d9 + constant . The energy integral. • 2/1—)<=os^ Ism^9 (11.I^"^ sinecos9 + J U + ^cosO) Analytical Dynamics <j)sme — mgi sinvO ^ V =0 (11. from Eqns. • o„ I sm 9 Ism'^9 +mg£ cos 9 = h \ ' + Pf (11.mglsine = 0 (11.17). = Pi = constant Solving these for (p and -ip and substituting into the first of Eqns.

14) and (11. f{u) looks as shown on Fig. (11. Substitution into Eqns. where ui and «2 are the two zeros oi f{u). Qualitative Analysis. if any exist. Eqn.18) gives •Q • a (j) cos ^0 . (11. —1 < u < 1. Let u = cos 0 . (e) If f{u) has three zeros. Now. 7=y Then the energy integral. Fig. First consider the special case Ui = U2 = UQ.u^){a .18) "^f.Gyroscopic Motion 195 which can be inverted in principle to give 9 = fe{t).19) First we note that for real u we must have /(«) > 0 and that f{u) = 0 gives « = 0 =^ u = constant ==^ 9 = constant. The solution has been reduced to quadratures. that is without numerically evaluating the integrals. (11. 2mgi a = 2h p2 IJ (11. (11.79^ + ^ = 0 .17). Consequently. becomes {(2 = j(^u) = (1 . combining the first of Eqns. We now analyze the behavior of the top via "qualitative integration".{t).13) with Eqns. The function f{u) is a cubic equation with properties: (a) f{u) -> +00 as u —)• +oo {—u^) {—au) — {-^uf = +00 because o > 0 (b) f{u) (c) — • —00 > as « —)• —00 /(±1) = -(/3T7)'<0 (d) f(u) has a zero between +1 and +00. ^ = 0. the other two must lie in — 1 < M < 1.16) then allows integration to get (f) = /^(i) and ^ = fii. 11-5 (—1 < u < +1 because u = cos 9). -{/3^uf (11.

21) there will be two real roots.This is called steady. precession. 11-6 with solution ^1. To visualize the motion.20) r Provided * 7 ^ > 2a cos 60 that is (tp + ij) cos Ooj 2 4Im5'^ (11. 11-5 Fig. 1/2' (11.196 Analytical Dynamics Fig. It takes special . consider the path traced out by body axis 2 on a sphere (Fig. 11-7). or regular. This is called the sleeping top. Another special case is UQ = 1{6 = 0) for which the z axis stays vertical. It traces out a circle at cone angle OQ at rate (pi or (p2.2 Fig. This can be satisfied in either of two ways as shown on Fig. The inequahty will hold when the spin tp is sufficiently high. 11-8 (only one is stable). 11-7 7 2 cos ^0 1± 1 2a cos 9. corresponding to a fast and a slow precession.

11-8 initial conditions 61(0).Gyroscopic Motion 197 1 •>J"f^^'°^''H. nutates stable.u)"^ {u2-u)Ul-u'^)a-j'^{u2-u)] (11. the other two are given by {l-u'^)a-j'^{u2-u) which implies that 7 ^^'=^ = 2 ^ ^ 4^ =0 U2 + I [11. where U2 = cos 62.t^ unstable. continues to spin about vertical Fig. •?/'(0)5 0(0).19) becomes 0 = f l — M2)(a — 0^2) so that a = au2.23) Clearly. (11. That is.u^)a{u2 . u = U2 is one zero as expected. ^'(0).14) and using Eqns.24) . (11. the top is released with no precession and no nutation at some angle 62.u) . f{u) = = {1 .22) P2U2 = Pi =151 Thus at i = 0 Eqn. (11.j'^{u2 . Substituing these initial conditions into Eqns. Next consider the special case of initial conditions: 0{O) = 92 e{o) = 0 . ^(0). cf){o) = 0. </'(0) to produce these special cases.Then for any time.18): (11.

6(0) = {j)(0) = 0 Fig. we get either of the two cases shown on Fig. the initial state is duplicated. such a point is cusp. the top falls under gravity but then begins to precess and nutate (Fig. 11-9 (t)(0) > 0 Fig. 11-9). only the lower one is physically possible (see Fig.26) U^ Hence both (j) and ii.25) Therefore the z axis falls from 62 to 9i and then oscillates between the two values. 11-10.14).198 Analytical Dynamics As mentioned previously.u) 1 (11. 11-10 (t)(0) < 0 . 11-5): ui = cos 9i = 1^ 2a 7^ (11. if the initial conditions are 4'{0) ^ 0. are zero simultaneously at u = U2. Prom the second of Eqns. the decrease in potential energy is accounted for by an increase in kinetic energy of the same amount. cj) is J{U2 . The motion is thus as follows: After release. When u = U2 is again reached. Essentially. geometrically. More generally. and thus 9 as well. (11.

and using Eqns.Q 0 = 0 dt where the Qi's are the components of the torque exerted by the bearings required to keep the motion as specified.2 ' 9=0 ip = constant (11. 11-11 ecliptic (sun-earth) plane Gyroscope. Consider now a spinning.Q^ = 0 dt [ j ( ^ + 0cos0)] -Q^ = 0 (11. it acts like a top (Fig.27) Lagrange's equations are — [/^ sin^ e + J ('ip + <i> cos e) cos 6'1 . and would be zero if the earth was spherical. Q^ = 0.000 years. precession perpendicular to ecliptic Fig."" ^ .3 Some Applications Precession of the Equinoxes.Gyroscopic Motion 199 11.28) (l^) -/(^2sin6'cos6' + J < / ) ( V ' . Because the earth had an initial spin on its polar axis when formed and because it is an oblate spheroid (slightly flattened at the poles). Qe = JH (11-29) . in 80 years the spin axis precesses r. This torque is extremely weak and gives a processional period of 26. heavy body with no gravity torque {£ = 0) and constrained such that ft. Q^=0. Carrying out the difi^erentiation. (11. primarily by the sun and moon. 11-11).27). The torque is due to gravitational attraction.^ c o s i 9 ) s i n 6 ' .

This is one application of the gyroscope. Therefore. axis attached to vehicle Fig. the gyroscope remains fixed in orientation relative to an inertial reference frame. perpendicular to both the <f> and tp axes. The gyroscope is mounted in gimbals such that it nominally experiences no torque (Fig. This can be used to detect motion. 11-12). Then motion in a perpendicular direction can be detected by measuring moments in the bearings in the third orthogonal direction.200 Analytical Precession Dynamics ^ moves in (x. The body is set spinning about its axis of symmetry. measuring the orientation of the vehicle relative to that of gyroscope gives the orientation of the vehicle relative to an inertial reference frame. 11-12 Therefore there is a torque required in the line of nodes (Fig. Three such devices in perpendicular directions will detect any angular motion. As the vehicle moves. 11-13 .y) plane (precesses) axis Fig. Another use of the gyroscope is as an angular reference in an inertial navigation system. 11-13). to keep the motion as specified.

This is a gyroscope fixed to the earth in such a way that the rotation of the earth causes the gyroscope to precess with a period of one day. See Ardema. 11-14) and why a football travels with a constant orientation along its path when spun and thrown^.these produce moments that make the gyro precess and nutate. 11-14 It is very important to minimize friction in the bearings and drag on the spinning disk . Newton-Euler Dynamics. . or the northernly direction. Notes 1 2 Other choices are the Rodriques parameters or quaternions. This causes bearing torques to act in such a way that the gyroscope axis always lines up with the direction of precession. Gyrocompass. Other Applications.Gyroscopic Motion 201 precession caused by turning handle bars moment ^ spm Fig. A disk of mass 2 kg and diameter 150 mm is attached to a rod AB of negligible mass to a ball-and-socket joint at A. The disk precesses at a steady rate about the vertical axis of ^ = 36 rpm and the rod makes an angle of/3 = 60° with the vertical. Gyroscopic motion also partly explains why one can stay up on a bicycle when it's moving (Fig. The latest technology is a "ring-laser" gyro which has very low friction and in which angles are measured by lasers. PROBLEMS 11/1. known as drift. Determine the spin rate of the disk about rod AB.

respectively. The radii of gyration of the top are 0. The fan spins at a rate of 1725 rpm with the fan turned on. Problem 11/3 Same as Problem 11/1 except that /3 = 30°.2 rad/s. The figure shows a top weighting 3 oz. and 6 = 30°. about the axis of symmetry and about a perpendicular axis passing through the support point 0. and 1.2 kg with a radius of gyration of 60 mm about the spin axis. Find the value of h that will produce a steady precession about the vertical of 0.202 Analytical Dynamics 600 mm Problem 11/1 11/2. The length C = 1. The parts of the fan that spin when it is turned on have a combined mass of 2. Determine the two possible rates of precession.. Problem 11/^ . 11/4. the unit is balanced when 6 = 180 mm.5 in. 11/3. With the fan turned off. A fan is made to rotate about the vertical axis by using block A to create a moment about a horizontal axis. The block A may be adjusted.84 in.80 in. the steady spin rate of the top about it's axis is 1800 rpm.

Determine the forces in the bearings A and B. and the turntable revolves about a vertical axis at a constant rate of 48 rpm in the direction shown. 203 The motor shown has a total mass of 10 Kg and is attached to a rotating disk. rolls on a horizontal plane on a circle of radius 2 ft. Determine the normal force between the wheel and the horizontal surface.5 Kg and a radius of gyration of 35 mm. Neglect the weight of all components except the disk. A rigid homogeneous disk of weight 96. 180 mm 120 mm Problem 11/5 11/6.6 lb. The motor rotates with a constant angular speed of 1725 rpm in a counter clockwise direction when viewed from A to B. Problem 11/6 .Gyroscopic Motion 11/5. The steady rate of rotation about the vertical axis is 48 rpm. The rotating components of the motor have a combined mass of 2.

Neglect the masses of all components except that of the disk. If the assembly is released from rest at the vertical position (^ = 0. compute the moment that the assembly exerts on the shaft at C. A homogeneous thin disk of mass m and radius r spins on its shaft at a steady rate P. The 64. Problem 11/8 . This shaft is rigidly connected to a horizontal shaft that rotates in bearings at A and B. If the cylinder spins at steady rate p = 50 rad/s and the bracket at 30 rad/s. ^ = 0).4 lb homogeneous cylinder is mounted in bearings at A and S to a bracket which rotates about a vertical axis. Neglect the mass of everything except the cylinder.204 Analytical Dynamics 11/7. Problem 11/7 11/8. determine the forces in the bearing at A and B as the horizontal position {6 = 7r/2) is passed.

--. ••.1) '^Qnsqs + 4>n{qi. generally coupled.-.29).1 Introduction First Order Form of Equations of Motion.qn. As discussed in Section 6. z\.5.qn'.Chapter 12 Stability Of Motion 12. application of Lagrange's equations.2) 205 .t) s =0 n where qi. Because positive definite. and (3) The matrix Urs is positive definite. In Section 8.1. (6. there exists a transformation to new generalized coordinates. (2) They are in general dynamically coupled. form. to a dynamic system results in a system of differential equations of the form s : (12.qi.qn. Eqns.--. We now do this by a different method that results in uncoupled equations.-.1) are called the mathematical model of the system. Note that: (1) These equations are linear in the acceleration components qi. Equas=l tions (12. say.qn are suitable generalized coordinates and ^ s = ^ . the equations of motion were put into first order.Zjn qr = Y^trsZs s (12.

in.t) = 0 : Zn + (l)'n{zi.--. (12.Zn.206 Analytical Dynamics such that in the new variables the equations are dynamically uncoupled: Zi +(f)[{zi.--.--.5) are said to be in state variable form.i) =0 (12. Frequently these are taken as current conditions in control applications. . we mean the solution of Eqns.3) Now let yi = zi (12.--.y2n) is called the state vector} The initial conditions are y(to) = (2/1(^0).Zn.Zi.4) Vn+l = k y2n = Z2n Then the equations of motion may be written as a system of 2n first order equations of the form yi = yiiyi.y2n. Intuitive Notion of Stability. There are generally three types of perturbations of interest: 1.-.^n.--. These perturbations may be due to sensor error or to disturbances.y2mt) : (12. ••)2/2n(^o)).Equations (12. This is a convenient form for further analysis and computation.5). In initial conditions.t) where y = {yi-.Zir-.5) hn = y2n{yi. Stability has to do with the following question: Does the motion of a system stay close to the motion of some nominal (reference) motion if the conditions are somewhat perturbed? By motion.

o and i. The unperturbed motion is given by UQ . Since any dynamic model is only an approximation of a physical system. . (uo + Vu) . and there may be significant unmodeled dynamics (for example.(0) = UQ. 2.(0) = a. x' — X = rjx cos nt ^ . the controller dynamics are neglected in many control problems).(0) = XQ + rjx and x{0) = UQ + rju so that the perturbed motion is ' t . 1. x = [XQ + rix) cos nt-\ smnr n The difference between the two is sin nt n Since this will stay small if r/x and rju are small. We investigate the stability of the motion of the harmonic oscillator. The reference motion is frequently an equilibrium condition (no motion). Only perturbations in initial conditions are usually considered in dynamics. Remarks.Stability of Motion 207 2. This is sometimes called structural uncertainty. In the dynamic model. X= XQ cos nt~\ sm nt n where a. The question of stability is usually of vital importance for a dynamic system because an unstable system is generally not usable. . the motion is stable. whose equation of motion is x + n?x — 0. Note that the perturbation in the motion does not tend to zero over time. but rather persists at a constant amplitude. 3. \ . mass. or aerodynamic coefficients). Example. there is always unmodeled dynamics. 3. Let the perturbed initial conditions be a. when the initial conditions are perturbed. all three are of importance in control system design. 4. Significant terms may have been neglected in formulating the equations. stiffness. In parameters (for example.

12-1 in T space for the case of two states for the general case and in Fig.6) ysit)-fsit) (12. Consider a motion (i.-. in the present terms.2 Definitions of Stability Geometrical Representations of the Motion.y2n{t) with initial disturbances Vsito) .5)) /i(t). t) €T C iE2"+^. (12./2n(^).This motion is L-stable if for each e > 0 there exists a r]{e) > 0 such that for all disturbed motions yi{t). Recall the representations of motion introduced in Section 2. t) e E C iE'*+\ y e S C lE'^"'.e. 12-2 for perturbed reference . where C is the configuration space where E is the event space where S is the state space {z. z_eC C lE^.208 Analytical Dynamics 12. where T is the state-time space Liapunov Stability (L-Stability). (y.7) for all t and s = l.2n.--.--.2. a solution of Eqns.fs{to) we have <vie) < e (12. The situation is depicted in Fig.

This motivates another definition of stability. consider the motion of the harmonic oscillator expressed in the form: X = Asin{cot + B) (12. further. lim t —>• o o ys{t) . but is L-Unstable with respect to parameter oj (see Fig. For example. f2n{l) where 7 is an arc length parameter. Poincare Stability (P-Stability). however.9) It is clear that this motion is L-Stable with respect to perturbations in both A and B. we see that the perturbed motion remains close to the reference motion and it may be that this is all that's desired (Fig.--.Stability of Motion reference — 209 perturbed Fig. In some cases. If the motion is plotted in the state space. In words.fsit) = 0 (12i for all s then the motion is asymptotically stable. a type of stability other than L-Stability is of interest. then they remain close thereafter. Consider a motion fi{j). 12-3). 12-4). 12-2 the special case of the reference motion being equihbrium. the motion is L-stable if when the perturbed motions are sufficiently close to the reference motion at some time. Then the motion is P-Stable if for each e > 0 there exists a 77(e) > 0 such . If.

Analytical Dynamics perturbation in A perturbation in B ^vX^^^ . 12-3 that for all disturbed motions i/sij) with initial disturbances IVsilo) ~ fsho)\ we have |y. A motion that is L-Unstable and P-Stable is illustrated on Fig. (12. then it is always P-Stable. 12-4 for all 7 and s. 12-5 in T space. but not necessarily conversely.(7)-/s(7)l <e perturbed motion at time t v reference motion at time t ^J^*^ ~'^'^y^^y^^^ 'f/"^ / / Fig. for obvious reasons.11) <ri{e) (12.10) \ . Note that if a motion is L-Stable. Poincare stability is sometimes called orbital stability. perturbation in co Fig.

2n (12. by expanding the disturbed motion in a Taylor series in the perturbations and retaining only the first (linear) terms. Variational Equations. no matter how small the disturbances. In some exceptional cases. s= l. that is. The stability of the linear system is then investigated. is not usually known. What constitutes "large". unfortunately.Stability of Motion 211 reference perturbed 12. The stability properties of linear systems are well-known as compared with the stability properties of iionlinear systems. There are two potential dangers in this approach: 1.-.t) . If the disturbances become "large".3 Indirect Methods Introductory Remarks. (12. the first order terms no longer dominate and the approximation is not valid. Let the disturbed motion be equal to the reference motion plus a perturbation: ys{t) = fs{t) + r]s{t) . stability of the linear system does not guarantee stability of the nonlinear system.12) Substitute this into Eqns. The nonlinear system is approximated by linearizing about a reference motion. 2.5) and expand in a Taylor's series in the perturbations: fs+rjs ^ s ( / l +mr-j2n + V2n. . This observation suggests the following procedure.

(12.15) where Igr is the identity matrix.— evaluated at y^ = fr.212 2n Analytical Dynamics Ysih.15) is a polynomial equation of order 2n.But fg — Ys(/i.14) In general.14) as follows: (1) If all roots As . Eqns. f2mt) because oyr fs{t) is a motion. The 2n roots of Eqn.14) are a time-invarient linear system. (3) If all roots are distinct and some roots have zero and some negative real parts. (12. (12. therefore. (12. the equations are stable but not asymptotically stable. If. the reference motion is an equilibrium position (/g = 0 for all s. Stability of Time-Invariant Linear Systems. . 2n have negative real parts. ••.15) are called the eigenvalues of Ugr and their signs determine the stability of Eqns. have been developed to determine the stability of a system directly from the coefficients of the system's characteristic equation.AlsrI = 0 (12. the equations are unstable.2n (12. The characteristic equation.14) are asymptotically stable. (12. s = 1.14) can be obtained by taking Laplace transforms or by substituting rji = Aie^^.f2n) + J2°''^rr): r=l +nonlinear terms in the 77^.13) gives 2n r]s ='^ r=l asrr]r ] s = l. This will not be pursued here.--. Eqn.13) QY where Usr = -7. neglecting the nonhnear terms in Eqns. then the Usr do not depend on time and Eqns. (2) If one or more root has a positive real part. The characteristic equation associated with Eqns. (12.-. the stability properties of which are well-known and easily stated. Criteria. which implies that all velocities and accelerations are zero) and if the functions Yj do not depend explicitly on time. ••. s = 1. called the Routh-Hurwitz Criteria. The key results will be stated without proof. 2n (12. ••. the agr will be explicit functions of time. • •. however. 'r]2n = ^2ne^^\ the result is Iflsr . (12.

In this case. 180°. (7. 180° We investigate the stability of these positions by using the rules in the previous section.Q\ sm(ei . First consider equilibrium position (a). Substitution into Eqns.16) Q2 + cos(0i . The masses of the links are negligible. 91 = 62 = 0.sin6*1 = 0 (12. 12-6). 0^ = 0°.6'2)6'2 + Ql sin(6'i .^2)^1 . 12-7): ^5^ = 0°.16) and retaining only the first order terms gives: f]i = m 2772 + m = ~ 2 T ? ? I m=m . (12.^2) + f sine2 = 0 We investigate the stability of the equilibrium positions. (12.Stability of Motion 213 Example . Let where the rji are small perturbations from equilibrium. Substitution into Eqns. these are defined by ei=Ql = 9* = 0* = O. Consider a double pendulum with both links of length i and both bobs of mass m (Fig. Eqns.6*2) -h 2.Double Plane Pendulum.16) gives sin(9^ = 0 and sin^l = 0 so that the four equilibrium positions are given by the combinations of (Fig.25) reduce to 2^1 + cos(6'i .

214 Analytical Dynamics a) 01 = 0 02 = 0 b) Oi = 180° 02 = 0 c) 6*1 = 0 02 = 180° d) 0* = 180° 02 = 180° Fig. all four of these eigenvalues have only imaginary parts (Fig. proceeding as before. 0| = 0.4 = ± A / 7 ( 2 ± \ / 2 ) . 12-7 (Note that these are not in state variable form.2.) The eigenvalues of the coefficient matrix of this system are AI.3.2.3. * Vl =112 2??2 + m = 2T??I m=m m + m = jm AI. Thus this is case (3) above and the equilibrium is stable but not asymptotically stable. 12-8).4 = ±A/7(-2±\^) Since g/i > 0. Next consider position (b). & = TT.

the solution with initial conditions a.(0) = XQ and y(0) = yo is X = Xo+yot1 -gt 2 . Equilibrium positions (c) and (d) are also unstable. 12-8 Fig. Some Simple Examples. Here and in the next section we consider motions in which the reference motion is not an equilibrium position.-gt y =yo + r]y-gt The difference between the two motions is: X -X = rij:+rjyt . 12-9). Letting y = x. y = yo.y2 ay y=x-y+ . y . consider the motion defined by the system ax X = —X — y +\Jx'^ -I. with x measured upwards from the earth surface. consider the motion of a particle moving vertically near the surface of the Suppose the initial conditions are now perturbed so that x{0) = XQ + rix and y{0) = yo + rjy\ then the perturbed motion is given by x 1 XQ + rix + {yo + riy)t . The equation of motion is i = —g. This is case (2) above and the equilibrium position is unstable.Stability of Motion 215 ^m *m -^ R. Rp Fig. As a second example. First.y = r]y Thus the difference between the reference and the perturbed motion grows with time and the motion is not L-stable. 12-9 The eigenvalues thus have only real parts and two of them are positive (Fig.

Q' . As a third example. 12-4. y(0) = a s i n a . y) plane with radius a and period 27r/a. r(0) = ro + ?7r and ^(0) = Oo + r]g.(0) = a c o s a . Then the motion is X = acos{at + a) y = a sin(ai + a) The motion therefore describes a circle in the {x. The situation is similar to that shown on Fig. consider the motion defined by = -y\/x^ + y2 y = x\/x^ + y^ Take the initial conditions.0 = r]e Therefore the perturbation in the motion stays small if -qr and rjg are small and the system is L-stable. 9 = 6o + t Now suppose the initial conditions are perturbed. without loss of generality. the system is not L-stable but is F-stable. 0=1 With initial conditions r(0) = rg and ^(0) = ^O) the solution is r = (ro . to be a. Note that r' — r ^ O a s f — > .o o but that 9' — 6 remains constant. then the perturbed motion is x' = {a + ria) cos[(a + r]a)t + a + r}a] y' = {a + rja) sin[(a + ria)t + a + ria] Because the period has been changed.a)e"* + a . Now let the initial conditions be perturbed to a + r/a and a + Tja'.216 Analytical Dynamics Transforming to polar coordinates. Then the perturbed motion is r' = (ro + r]r — a)e~* + a e' = eo + r]e + t The difference between the two is r' -r = r]re~* . the system equations are r = a — r. Thus the stability is not asymptotic.

Second. the Euler-Hill. First.(2—-1-3x0 ) c o s n t + — s i n n t \n J \ n J n yit) = yo — 2 n 3(yo + 2nxo)t + 2— cos nt n (12. are X — 2ny — Sn^x = 0 y + 2nx^0 z + n^z = 0 where n = \J[ijo? is called the mean motion and [i and a are the earth's gravitational parameter and the radius of the nominal orbit. the case of a satellite or space craft in a nominally circular earth orbit will be discussed. and orbital station-keeping. (12. or the Clohessy-Wiltshire equations.4 Stability of Orbits in a Gravitational Field Here we consider the important problem of the stabihty of closed orbits (i. z{t).17) is x(t) = 2 I — + 2x0 ) .17) are called in various places the Hill. for example they are used in the analysis of rendezvous between two spacecraft in neighboring circular orbits. docking maneuvers between two spacecraft.17) + 2 ( 2 — + 3x0 Isinnf n sin nt n In these equations. Eqns. and this component of the motion is L-stable but not asymptotically so.18) (12. y{t). For specificity. yo. yo. is uncoupled from the in-plane motion. The radial component. elliptical and circular orbits) of a body moving in a central gravitational field. (12. ZQ. however. XQ. zo are the perturbations in position and velocity components at time t = 0 from a nominal circular orbit. relative to a frame that travels with the orbit (Fig. Inspection of Eqns. The solution of Eqns. The perturbation equations. z{t) are the perturbations at some later time t. XQ. the in-plane perturbations are not generally bounded and the motion is not L-stable. (12. 12-10).e. derived in most books on orbital dynamics. is bounded and thus the ZQ z{t) = cos nt -\ .Stability of Motion 217 12. respectively.18) shows the following. the motion perpendicular to the orbital plane. They have found wide application in orbital dynamics. because of the term linear in t in the y{t) equation. and x{t).

12-11) and arrives back at the spacecraft at the time at which the spacecraft has completed one revolution of the earth.18) become xltj = — smnt n 2i. In this case the initial perturbations are xo = yo = ZQ = yo = ZQ = 0 . In an inertial frame. 12-11 perturbed (elliptical) orbit remains close to the nominal circular one. with an object in the perturbed orbit either pulling ahead or falling back relative to an object in the nominal one. Thus.o y[t) = z{t) = 0 n cos nt n These are the equations of an ellipse with semi-major and semi-minor axes of 2xQ/n and xo/n. Again the situation is similar to that shown on Fig. respectively. 12-10 Fig. the particle travels around the earth in an ellipse neighboring the circular orbit with the same period. . relative to the spacecraft the particle travels in a elliptical orbit (Fig. 12-4 and the motion is P-stable.218 Analytical Dynamics Fig. rendezvousing with the circular orbit after each revolution. io > 0 and Eqns. (12. This motion is clearly L-stable. As a simple example. suppose a spacecraft in a circular earth orbit ejects a particle of small mass in the outward radial direction.o ^ 2i.

If V{-) is positive everywhere except at the origin. if V{0. 12-12).19). C -ZE-^" containing the origin. These definitions are generalizations of the idea of positive definite and negative definite quadratic forms. but it can be zero other than at the origin. then V{-) is called definite in Q. 12-13). then 271 dt ~t^^dvs'^' (12. ". (12. 0) = 0. it is positive definite (Fig.2n be a perturbation from equilibrium of a dynamical system.5 Liapunov's Direct Method Autonomous Case. Now let the rjs in these definitions be solutions of Eqns. 3.V2n) . if negative everywhere except at the origin. these functions do not depend explicitly on t.". 2. and if V{-) has the same sign everywhere in fl except at the origin. r]2n) is of class C^ (i. 12-13 .19) In the autonomous case. Then these functions satisfy equations of the form ris = gs{rii.20) Fig. ••.2n . s = l. If F(r/x.e. continuous with continuous derivatives) in an open region Q.-. We define the following classes of functions: 1. it is called semidefinite. it is negative definite (Fig.--. s = l. Let rjsit).i2. 12-12 Fig.Stability of Motion 219 12. Figure 12-14 shows a positive semidefinite function. If V(-) has the same sign everywhere where it is not zero.

••. A geometric proof of the theorem follows from Fig. Liapunov's Theorem. 12-15 We now define a Liapunov function as a function V{-) in ft definite in sign for wliich dV/dt is semidefinite and opposite in sign to V{-). (12.--. The properties of V{-) ensure that the motions due to small perturbations from equilibrium either tend to zero or remain small. The key result is tlien tfie following.220 Analytical V Dynamics Fig. 12-14 Fig. the transformation equations to generalized coordinates are time independent. energy is conserved: E==T + V = h (12.2) the equilibrium condition is T = 0. (6.n (12. If a Liapunov function can be constructed for Eqns.n (12. from Eqn. and therefore. Also. scleronomic. 12-15.23) . then the equilibrium position is stable. Application to Dynamics. since the system is closed. The equations defining the equilibrium condition are thus: dV_ = 0 dqs s = 1.22) This equation is a necessary condition for an unbounded extremal point of V\ that is. conservative) system. or dV/dt = 0.19). Lagrange's equations for such a system are dt \dqsJ dqs dqs 0 l. V has a stationary value at an equilibrium point. Consider a natural (holonomic.21) Because the system is scleronomic.

t) .r]2n. Consequently.1.2n (12. Then. and let r?i. there is no systematic procedure for finding Liapunov functions and they are generally very difficult to find.--. ••.--. that is qr = q* + %.cos Oi) + mge{l . Thus V is positive definite. . Choose the datum for V such that y(0. r^2n) e ^ . 5 = l.-'' iQn be the equilibrium values. since V has a minimum at rji = 0.r]n).r]2n) e !^. 6*2) > 0 for all sufficiently small 9i and 6*2. ••. Example. In this case. then V{-) is positive definite. for V = 2mgi{l . ••.25) (2) If F(-) < Wi-) for all im. (12. An equilibrium solution of the equations of motion for the class of systems defined above is stable if the stationary value of the potential energy is a minimum relative to neighboring points. Since the system is then V{-) is negative definite. 0) = 0 and (ii) V(6'i. E is a Liapunov function and by Liapunov's Theorem the equilibrium is stable. it suffices to show that E is a Liapunov function. £^ is a Liapunov function and the equilibrium is stable. one or more of the perturbation equations contains time explicitly: r]s = 9s{m. (ii) they are singlevalued and of class C^ in fl. 9^ = 62 = 0.r]n = 0. We see that: (i) F(0. r = 1. Then V = V{rii. Prom Section 6.-^O). Remark.-.cos 82) Consider the equilibrium position (a). which Consider again the double pendulum (Fig. Also.0) = 0. 12-6). Consequently. ••. and (iii) W{-) is positive definite. implies that dE/dt = 0.--. K is positive at neighboring points and F is a positive definite function.-. Let q*.t) and W{rii. ••. To prove this theorem. In all but the simplest problems. n. E = h = const. ••. N o n a n t o n o m o u s Case. ••.?]2n) = (0.24) We now need to introduce two functions V{rii. energy is conserved and dE/dt = 0. Then: (1) If V{-) > W{-) for all (r?i.Stability of Motion 221 Dirichlet's Stability Theorem.r]2n) such that (i) they vanish at (??i. T is always positive definite so that E = T + V is positive definite. rjn be small perturbations from these values.

ensuring that motion will tend to zero or remain small. Liapunov's Theorem (nonantonomous case). for two different functions.25) and dV/dt as given in Eqn. . Fig. If the conditions of the theorem are satisfied. and that Liapunov functions will be even more difficult to find.222 Analytical The change in V{-) along a trajectory is now dV dt 2n ^ dV Dynamics dV dt (12. y2n)^ where T denotes transpose. the function V{-) will stay "completely inside" the function W{-) (Fig. 12-16 Notes 1 2 More precisely.26) is semidefinite with opposite sign of V{-). The proof of this theorem is similar to that for the autonomous case. y = {yi.24). we call a function V{-) a Liapunov function if it is definite in sign in accordance with definitions (12.26) For this case. Caution: We are using the same symbol. the reference motion is stable. If there exists a Liapunov function for Eqns. ••. It is clear that this is a much stronger requirement than for the autonomous case. Liapunov and potential energy functions. (12. (12. 12-16). V.

as shown. The rigid rod has negligible mass. Investigate the stability of equilibrium position (d) of the double plane pendulum by using linearized equations. vertical position by means of the linearized variational equations. In particular. and examine the stability of all three steady motions by means of the linearized variational equations. as shown. Iy)uJxOJy = 0 . 12/2. Find a Liapunov function for the system of Problem 12/3 thus verifying the results of the linear analysis. inverted pendulum of mass m is restrained by identical linear springs. A heavy. show that the motion is unstable if Ix is intermediate in magnitude between ly and Iz12/5. 12/3. Show that there exist three steady-state motions. Investigate the stability of equilibrium position (c) of the double plane pendulum by using linearized equations. 12/6. (4- Use the linearized variational equations to examine the stability of the steady-state rotation w^ = O = const. for one of which the pendulum angle with the vertical is a non zero constant. Problem 12/3 12/4. . ujy = uj^ = 0. A heavy pendulum of mass m rotates with constant angular velocity about the vertical. The rigid rod has negligible mass.Stability of Motion 223 PROBLEMS 12/1. Examine the stability of small motions about the inverted. Problem 12/5 The torque-free motion of a rigid body about a point is given by lyUJy Ix)^z^x = 0 .

can you find one? 12/9. by both the indirect and direct methods. Show. Investigate the functions Analytical Dynamics Ix^l + Iy<^l + h'A and IXUJI + lyCol + IzUj'i . 12/8.Ix^'^ as Liapunov functions for the motion of Problem 12/4. that the equilibrium position of Problem 4/2 is stable. .224 12/7. of Problem 7/3 is stable 12/10. Investigate the function V = 9'^ + oj^ [cos^ e + 2a(l .cos 9)] as a Liapunov function for the steady motion with 6 not zero of Problem 12/5. If this is not an L-function. Show that the equilibrium point 6 — by finding a Liapunov function.

that is. say tj. and F'^ is the r**^ component of the impulsive force on particle r. Mathematically. Now we relax this restriction and consider impulsive forces. In this equation. 13-1 225 time . it is a limit of functions that is not itself a function (Fig. xf = {x\.X2. we have considered only forces that are everywhere bounded.Chapter 13 Impulsive Motion 13. An impulsive force is a force that tends to infinity at an isolated instant. Distrif(t) If(t)dt const. the strictly Newtonian problem. 13-1). ast-^tj t Fig. an impulse is an example of what is called a distribution.x^^) denotes the position vector of particle s in an inertial frame. Until now.1) We call P'^ the impulse of force F^ at time tj.1 Definitions and Fundamental Equation Impulsive Force. the Newtonian problem. such that its time integral remains bounded: lim f F'{3f. that is. that is.x\T)dT = P' (13.

If a function / has a discontinuity at i = tj.226 Analytical Dynamics m 'i+ fj- time Fig. but not all.2) lim 0 < T2 .> 0 f{tj+T2)=f. Physically. 13-2): lim 0 < Ti . (13. Let an impulsive force act on the particle at time tj with impulse components P'itj) = I PUtj) PSitj) . an impulsive force is a very large force that acts over a very short period of time. this may be integrated once to give ft. 13-2 butions have some.- Impulse-Momentum Relationships.^ 0 fitj-ri) = fj (13. Newton's Second Law for particle r is (13. of the properties of functions.3) Since rur is a constant. we write (see Fig.4) Jtn 'to which states that the change in linear momentum of the particle equals the impulse of the force acting on the particle.

1 rt ^ rur {if {t)-of (to)) .E' = E-' + F'' (13. (13.4) for all n particles. this is the fundamental characteristic of an impulse.£ 1 -^x' = 0 where rt (13. which is again governed by Newton's Law.+ . _ ) = ^ J (13.^ ^ .7) where FJ" is the resultant given force and FJ' is the resultant constraint force on particle r. The main goal of the analysis of impulsive forces is to determine the values of the velocity components after an impulse.. Adding up Eqns. but that the position components are continuous there.5). This equation thus provides the initial conditions for the subsequent motion. (13.9) .6) with a virtual displacement Sx^. Til (i5. Now consider the scalar product of the last term of Eqn. m. where there is continuity of displacement and a discontinuity of velocity given by Eqn. (13.Impulsive Motion Then from Eqn.^'(^o)) . m. so that '* F^'dr) 8x' = (fF^'dr) •-Sx' =^r f {E!' • 5if) dT = 0 I to J J to Thus if we take the scalar product of Eqn. by definition. K (2^''(^) . we obtain Y. (13.4)./ F'dT 'to' =J2 ^''^'^ (13.6) with Sif. Fundamental Equation.8) F!dT to (13. (13.6) where we have written Y.5) We see that the velocity components are generally discontinuous at an impulse. The motion proceeds according to Newton's Second Law until time tj. a constraint force does no virtual work.

The assumption is made that the bodies and constraint surfaces do not deform during this time.-.4 it was shown that possible velocity changes satisfy the same conditions as virtual displacements.L (13.P'] • A r = 0 r=l (13. (2.11) is N Y.10) This is the fundamental equation of impulsive motion.14) .11) The displacement components of the particles were introduced in Eqns.13) The constraint forces associated with these constraints are N F.10) is n 5 3 [mr {x\_ . (13. In Section 3. in terms of displacement components Eqn.2 Impulsive Constraints Definitions. We see that the velocity components after the impulse are governed by linear equations and thus this is a relatively simple problem. or they may be arbitrary. Asdus + Ardt :^ 0 .£ l -Saf = 0 r=l (13. 13.£_) . In an impulse. The Saf may be subject to impulsive constraints.siL) .38). The Phaffian form of the constraints on a dynamic system are TV Y. (13. then. (13.=Xr{t) r = l. Eqn. s=l r = l.12) where N = 3n. at t = to. thus another form of Eqn. This is the impulsive form of Eqn.L<N (13. the constraints must exert very large forces (infinite in the mathematical approximation) over very short times. In this Section. Both the given and the constraint forces may be impulsive.--. Now suppose at least some of the forces are impulsive at time to. we consider the latter case.228 Analytical Dynamics is the impulse of the given force acting on particle r.8) gives n Yl h r (i^+ .[ms{us+-Us^)-Ps]Aus s=l =0 (13. (3.4).

. .3-) . If / = 0.17) gives ^= t^^to £^M-«(r)rfr (13.17) (13. if F ' tends to infinity at ^o but / remains bounded.Impulsive Motion where the \r{t) are Lagrange multiphers.19) lim /•* where A(i) = /'A(t)di (13.20) 5(i) = I gSfjdt We see that impulsive constraints may come about in essentially two different ways. Assume that \{t) is discontinuous at to but that a{t) e C^ there. Let A = (Ai(i). Therefore. that is. as follows.Ai(i)) / N N s=l 229 \ a = 5Z ^i*^s + ^ 1 ' ••' Yl ^LsUs + AL J \s=l / (13.Fi) Then Eqns. (13.15) F: = {Fi.. the force is not impulsive.---.16) and (13. Combining Eqns. we limit our consideration to the case 0 < I < oo.^ l7^^ £ A(r) • a!{r)dr (13. (13.A.18) Now integrate this equation by parts in two different ways: I = a{to) • (4o+ . A(^) Discontinuous. fE:{r)dr (13.14) may be written compactly as Z ' = A-a The impulse of this force at time to is L= /!!r.16) t ^ to JtQ By definition a constraint is impulsive if £ ' is impulsive.

21) and the first of Eqns.+ -kt.z) = 0 Fig. Just before the encounter. the constraint force is £ ' = A grad / and A 7^ 0. At t = to the constraint is encountered.22) where Lt. the particle is unconstrained. t^to ^™ f^B{T)-X!{T)dT = 0 Jtn (13.^= ^ lim to A(r)dr to (13.23) is nonzero by assumption. Also note that the DOF changes instantaneously at to- F' = X grad f f(x. there is no constraint force and A = 0. just after. 13-3 a{t) Discontinuous.y. For t < to. consider a single particle encountering a constraint surface (Fig.230 Analytical Then lim Dynamics I A(T) • d{T)dT = 0 (13.19) becomes I = a{tQ) • (A•fo + A. Then. and for t > to the particle is constrained to move on the surface. We see that X{t) is discontinuous at ^oAs an example of this type of constraint. Note that the position of the particle is continuous at to.24) . (13. t < to.0- (13. but that A(i) G C^. Now assume a{t) is discontinuous at to. 13-3) with velocity not tangent to the surface. but that the velocity is discontinuous.

25) so that a{t) is discontinuous at toRecall from Eqns. (13. of course.--. If the r— impulsive constraint is holonomic.) ^ « + (^'-*o+-^-*o-) s=l (1^-28) We see that this type of constraint force arises if either some of the E^-s or some of the Er are discontinuous. may be also present.29) where L' is the number of impulsive constraints. --^UNit) = 0 and \:^dfr . (13. then it may be expressed as friui.Ers{to))Us + (E.L' (13. (13.^ I^a{r)d7 t ^ to jtg (13.25) becomes.(i) - Er{to)) E ( ^ " e o + .27) Then Eqn. Nonimpulsive constraints. for the r— constraint. Thus the impulsive constraints are N Y^ ErsUs + Er = 0 ] s=l r=l.^_= lim '•* ^--.15) that N Ij.^ . s=l N dfr dt du.19) becomes 231 where B. lim I ^ N ^Ars{T)Us{T)+Ar{T) dT lim t-^to N = to Jti ls=l N ^{Ers{t) Ls=l . Er = f Ar{t)dt (13.* o .26) and let Ers = I Ars{t)dt .Impulsive Motion and the second of Eqns. .^^-B. — y ^ A^gUs + Af s=l (13.

= 0) and that some of the A^s are unbounded (indeed. Then the function f{x. t) is again discontinuous. if this were not true the constraint would not be impulsive) at some time t^._. y. rs Analytical so that Dynamics 9fr . continuous displacement). = 0 5= 1 (13.y. (13. A live impulsive constraint also occurs when there is a discontinuity in the velocity of a constraint surface. the constraint is i . First suppose that the constraint is scleronomic {Aj.y.12).z. 13-4). 13-5 Next suppose that some of the Ar are unbounded but that the Ars are all bounded at some time to. This is called an inert impulsive constraint. 13-4 Fig.]Au.. 9fr lit''' ^' =f lt'' f'^-™' As an example. Note that the constraint is not allowed to change in such a way that the position of the particle is discontinuous.t+) f(x.y = 0. Fundamental Equation.z.z) = 0 Fig.232 In this case Ars = dfr/dug and A^ = dfr/dr E. 13-5). After the impulse. z) is discontinuous there (Fig. the constraint on a particle on the elevator floor is i = 0. z. consider a single particle and a single holonomic constraint. f(x. resulting in discontinuous velocity for the particle (but of course. For example. as usual.31) . Before the impulse. This is called a live impulsive constraint.[( f(x. T. When impulsive given forces and impulsive constraints are present.y. suppose an elevator initially at rest is given an instantaneous speed V. z being in the direction of the elevator's travel.y.Then / ( x . repeated here: N ^m. the fundamental equation is Eqn. the surface changing its location instantly (Fig._)-P.+ -u.

The theorems which follow in this section relate impulsive forces to energy changes. for convenience. Since velocities change during impulses. if all constraint forces are live.Impulsive Motion From Eqn. (2) If there are no given impulsive forces. Let the difference in z between that for any possible motion and for the actual motion be Az = z — z .!j_ = i''(to+) = velocity of particle r just after impulse at time to In this section. then all the Pg = 0.3 Impulsive Motion Theorems Remarks. the constraints. we drop the underbars for vectors. (13. If both (1) and (2) are true.e. then Eqn.L' (13.ErsAus s=l = 0. In this section we consider impulsive given forces only (no impulsive constraints).^ This theorem states that the quantity z = -'^rrir ix\ — xL ] (13.-. As before. then all the Ers = 0 and the Aug are unconstrained.32) There are two important special cases: (1) If there are no inert constraint forces. the following notation is adopted: xL = x'^{tQ-) = velocity of particle r just before impulse at time to i. The proof is as follows. (13. the system kinetic energy generally does also. i. r = l.!J_ relative to all other possible velocities x_^. Because the impulse takes place so quickly. the velocity components Aiig are subject to N 233 Y.33) is a minimum for the actual values of i. even the rheonomic ones. r r=l Gauss' Principle of Least Constraint. We n also let y~] = V^. 13.31) is merely a statement that all velocity components are continuous at to. are considered fixed.29).

234 Analytical Dynamics and let the possible motion of particle r after the impulse differ from the actual motion by Ax'^: x. ••.11).34) Superposition Theorem. JJ -. I fir- -. so that Az= 1 " -Y^mr{Ax'f Thus Az > 0 for all possible velocities and ^ is a minimum for the actual velocities.=x^. Expanding Eqn. r = 1.33). (13.'^rix r \2 A necessary condition for z{x'^_^_) to be a minimum is dz 9T+ -r r.'' .r n i —(ii.!j_ = dx^j^— mrX_ — r = (J . Then after k impulses whose sum is i=l . A useful expression may be obtained by applying the first order necessary condition for a minimum of z.+ Then Ax^ A^ = -oE TJir U l + Ai. n (13.[ be the velocity of particle r after an impulse P/". Let i. „ r=l ^ r=l r where T+ = -J2mr{xlf T_ = -Y.A i ! p \2'^ ~ rrir Pr rrir = ( Pr\ 1 " ^ vrir [x^^ — x^_ ^ J Ax^ + . (13.^ r\2 mr{Ax But the first term on the right-hand side is zero by Eqn.

As an example. -xDix^. givesxf_) = ^P''{x\. (13. (13. (13. Fig. let T+ be the kinetic energy after the impulse of the unconstrained system Ti be the kinetic energy after the impulse of the constrained system i. > Tc.36) Bertrand's Theorem. if the system is initially at rest. Tc = 0. Eqn. The theorem says that Ta > T.35) reduces to ^T=^J2Pri+ (13. If an unconstrained system is subjected to an impulse.!J_ be the velocity of the unconstrained system after the impulse Xi be the velocity of the constrained system after the impulse . the kinetic energy of the subsequent motion is greater than if the system had been constrained and subjected to the same impulse.11).11) ^mr{x\ + r r +x r X_ V Ylmr{x\f -Y^mrixLf = E^'(^+ + ^ X_ But the term on the left-hand side is 2AT so that Eqn. 13-6 shows an unconstrained and two constrained rods subjected to the same impulse P.Impulsive Motion the velocity is k X — / ^ X-i 235 The proof follows from repeated use of Eqn. (13. Energy Theorem.35) follows. Of course. To begin the proof of the theorem. we may choose Substituting in Eqn. This states that the increase in kinetic energy due to an impulse P^ is equal to ^^=^E^'"(^. In particular..+^-) r (13-35) Since the Ax'^ are unrestricted.

= \Y. 13-6 (c) We must show that r + > Ti. subtraction of the two equations gives y~^ mrx\{x^j^ ~ x\) = 0 But il{i\ Also -^\) = \ \^^\f . Thus we may take Ai'' = x\ in both^ > 0 .(^i)' .( i . whereas in the second Ai'' = x\ is{xlf so that Ri=T+-Ti which proves the theorem. = -J2mr{xl . any A f is admissible. With this choice.236 Analytical Dynamics (a) (b) Fig. - x\f\ ^ r T. Equation (13.11) applied to the two cases results in r r r r In the first of these.

m < n have prescribed velocities and let ^ — /^l .Impulsive Motion 237 Kelvin's Theorem. ••. The proof proceeds as follows. Let particles l. i!p . the kinetic energy is a minimum relative to other possible motions in which these particles have the same velocity. . If some particles of a connected system are suddenly set into motion with prescribed velocity. Choose Ax^ = x\ ~ x'2 in the fundamental equation: r r n m n r=l r=l r=m+l But x\~xl = 0. ^rn x.i : ^ ) 2 and T+ = -Y^mr{xlf r 22 = ^ E " ^ ^ ( ^ 2 ) ' 2 . respectively.m+l _ _ •n\ X2 — V'*'+) j .^ + j -''2 ' T''2) be the actual velocities after the impact and other possible velocities. i.. P'^ = 0 so that r Also.)'-(i2)' + ( i . -x'2)>0 .-i^2) = ^[(i. ••.(i. Therefore.i ^ = 0 and P'^+^ = 0.--. R2 = T2-T+ = l^mrii\ 2 which proves the theorem.

with this choice. Ax^ = X2 is also a possible velocity change in Bertrand's theorem. 2 j m r ( i + — x\){x2 — i i ) = 0 r But {xi .) m .) . A« t^ • B q ^ ^t' D • I I Fig.i?l = ^ E mrixl . To prove this.^i)] Combining results i?12 = i?2 .i. We will analyze the subsequent motion using the previous theorems.238 Analytical Dynamics Taylor's Theorem. r — x\) = 0 However. Three uniform thin rods of equal length and mass.i. we obtain y^mrX2(^+ ~ ^1) = 0 r Subtracting these two equations. m.i^i . begin by recalling from Bertrand's theorem that with Ax"^ = Xi we obtained 22''^rx\{x\. 13-7 . Example.x\w2 . The gain i?2 in Kelvin's theorem is greater than the loss i?i in Bertrand's theorem. 13-7).X[f > 0 ^ r which was to be shown. are hinged together and are initially at rest in a straight line on a horizontal friction-less surface (Fig. An impulse of magnitude P is applied to one end of the assembly.i\) = {i^i .

Applying Eqn.. Eqn.^ = fYp 1 Solving these equations simultaneously gives _52P^ ^~15M' _IAP_ ^^15M' _ ±f_ ''~15M' _Ai!. q.34). (1. Using Eqn.^ = ] -M{iq-p-r) -M{Ar-q-s) BT 1 . s 0 = 0 = BT .Impulsive Motion 239 Let the speeds of the ends of rods after the impact be denoted p. Then s = 0 and in this case Gauss' principle gives 45 P 12 P ^ ~ 13M ' ^ ~ 13M ' and the energy after the impact is 1„ r = 3 P 13 M 45 P2 . x^^p. (13. r and s as shown. *"15M Next. which is in this case Pr This gives + •r = ^^7. we apply the Energy theorem to compute the kinetic energy.59). r. Now suppose point D is held fixed. q. the kinetic energy of the system is r+ = TAB+TBC+TcD + -M{r'^ -rs = = -^M{p^-pq + s'^) rs) + q^) + -M{q^-qr + r^) -Mip^ + 2g2 + 2r^ + s^ -pq-qr6 First we apply Gauss' principle.36). (13.

Eqns. without proof.= 2ir2"" ' ''^ = 24-. So far in this chapter we have formulated problems in terms of position vectors and rectangular components."!" /"* / Qsdt (13. 52 P ^ 115.. 113.37) and take the limit i -> to to obtain lim t^to to d_ / 5 r \ dt \ OQS J dT_ dt= dqs J . Finally.29) are ^.n (13.. Carnot's Theorem.34). Carnot's theorem states that the change in energy due to an impulsive inert constraint is always negative. Finally. (13.".dt A(^—j=i?.^ 2 "'mJi^.. Then. 13. 2 rr. .n (13. respectively.240 Analytical Dynamics Thus T+ > Ti.38) t -^ to J to Now integrate Eqns.39) This is in fact just another form of Eqn.. R12 = R2-Ri = (T2 . If there are no constraints. s = l.(T+ .T+) . (13. Q.37) is dt KaqJ dQs where n is now the number of generalized coordinates. lim I'QS t ^ toJto . verifying Kelvin's theorem. Thus r^_ > T2.Ti) = T2 + Ti . Now we consider systems defined in terms of generalized coordinates. Let Qs be an impulsive force component at to\ then the impulse component is Rs= lim .4 Lagrange's Equations for Impulsive Motion Lagrange's Equations. . All of the theorems of the previous section also apply when the system is defined in terms of generalized coordinates. (6.2r+ > 0 verifying Taylor's theorem. verifying Bertrand's theorem. Next suppose that P is changed such that end A has the same speed p after the impact in the two cases above."^ where T+ and T2 are the energies with end A free and fixed. s = l.(^)-^-Qs = 0. we give one result for impulsive constraint forces.

5). A homogeneous rod of length / and mass m lies on a smooth horizontal table. PROBLEMS 13/1. The framework moves such that at time ti the string becomes taunt. 13/5. They are smoothly hinged at A and lie on a smooth horizontal table. Find the velocities of the ends of the rod immediately after the blow is struck. What is the smallest blow that the rod can be struck at the bottom at right angles to the bar which will just make the bar reach the inverted position? Repeat the problem of the four rods of the example of Section 3. Find the motion immediately after the impulse at ti. This is an adaptation of a similar principle for the strictly Newtonian problem. 13/2. 13/3. not the number of generalized coordinates. A heavy uniform rod of mass m and length / hangs vertically from a smooth pin a distance nl from one end (0 < n < 0. v. Two homogeneous rods OA and AB have equal mass m.Impulsive Motion 241 Notes 1 2 In this chapter. x as shown. The rod OA is smoothly hinged to a vertical pin at O. each of mass M and length /. the rods form a straight line. It is struck a blow in the plane of the table a distance nl from one end (0 < n < 1). but OA has length li and AB has length I2. and in a direction normal to the rod. are freely hinged together at their corners as shown. Label the components of velocity of the ends of the rods u. Two opposite corners are connected by a light inelastic string of length l/\/2 so that the string is taunt when the framework is square. The framework moves on a smooth horizontal surface. An impulse P is applied normal to OA and in the plane of the table to a point C lying between O and A. The blow is such that it would give a velocity i) to a particle at rest of mass m. This principle can be applied also to inert constraints. Initially. Four uniform thin rods. See Pars. unless stated otherwise. .3 except that the impulse P is applied at point B. n refers to the number of particles. w. 13/4. Calculate the reactions at the hinge pins at O and A and the angular velocities of the two rods immediately after the impulse.

. ' Problem 13/7 JVi M - Problem 13/6 13/7. homogeneous rod OA and AB of mass m and length / are smoothly hinged at A.V A Problem 13/4 13/6. Three heavy equal homogeneous rods AB. oVJ m Am B f t . Find the initial motion.242 V Analytical Dynamics -» C 'X . Problem 13/5 Two equal heavy. An impulse is applied at B in such a way as to give B a velocity v toward D. CD are smoothly hinged at B and C. and OB is smoothly hinged at O to a fixed point. BC. This mechanism hangs at rest when a bullet of mass ji is shot with horizontal velocity v into the block as shown. Find the angular velocities of the two rods immediately after the impact. w B •/i H^ h D ' u . and so that all rods are in a vertical plane. The rod AB has fixed to it at 5 a block of mass M and of negligible dimension. This assembly is suspended at A and D from smooth pins so that the distance between A and D is twice the rod length..

Calculate the kinetic energy imparted to the box by the bullet when the edge DD' is fixed. and compare it to the kinetic energy when DD' is not fixed. It is filled with sand. Find the angular velocity of the rods immediately after the impact. Four equal. and the mass of box and sand is M. The density of the material of the box is the same as that of the sand. homogeneous rods are smoothly hinged together to form a rhombus which is initially at rest on a smooth horizontal table.Impulsive Motion 13/8. 243 A box of dimensions AB = 2a. as shown. A bullet of mass m is fired with velocity v into the center of the face AA'BB'. AD = b. Problem 13/8 13/9. AA' = 2c rests on a smooth horizontal table. Let an impulsive force acting on a hinge point along one of the diagonals give the hinge point a velocity v as shown. Problem 13/9 .

Thus the requirement that the displacement components of the particles are explicit functions of the generalized coordinates is relaxed (see Eqns. linear. In this and the next four chapters. Nonholonomic constraints are accounted for in Lagrange's equations by tlie use of Lagrange multipliers. nonintegrable functions of the time derivatives of the generalized coordinates. that is Xl — Ul = x\ X2=U2 X3=U3 = x\ = xl X^ = U4 = xf XN = UN = JV/3 X^' 245 . we will loosely follow Pars. Therefore.Chapter 14 Gibbs-Appell Equations 14. we make a notation change to bring our notation into line with that of Pars. and we consider coordinates such that the velocity components are explicit.1 Quasi-Coordinates Introduction. We now develop an approach to systems with nonholonomic constraints that does not depend on multipliers .the use of quasi-coordinates and the Gibbs-Appell equations. (5. Quasi-coordinates are analogous to nonholonomic constraints in that they are defined by differential relations that are not integrable. The displacement components will now be denoted by Xr.7)).

Let a set of generalized coordinates be gr. such that dOr = ^ CrsdQs + Crdt .xl) .X2. (14.--.4) . u is the number of particles and N = 3u is the number of displacement components.p (14.. Consider a dynamic system with u particles. ^ S will denote ^ . (.p (14.3) We now require that the matrix have maximum rank.2) may be rewritten as dqk+i+r = Yl ^rsdqs + Crdt . and k — N — L will be the degrees n of freedom of the dynamic system. As usual. r = l. and no holonomic ones.£ (14.i +p (14.--. + £) x p.-.u are the position vectors of the particles. s r = 1. ••^qk^^j^p so that Eqn. the matrix is {k + tj x (^ + p).p Relabel this new set of coordinates gi. L will be the total number of constraints.t). s=l Quasi-Coordinates.-)-£-|-p where qk+i+r = &r\ r=l. nonholonomic constraints. Now. r = l.246 Analytical Dynamics where af = {xi. ••. as before. Call the remaining ones pg. n will be the number of generalized coordinates. r = l. The total number of coordinates is now A.k + i = n — 'iu = N The I constraints are ^Brsdqs s + Brdt = Q. (Since Bra is {k + () X I and Cra is (A.-. Br 'rs Ore r = 1.1) Introduce p new coordinates 9r. Cr G C^{q.) Under this condition the implicit function theorem guarantees that we may solve for l + p oi the dqr as functions of the remaining k dqr.Then k dqr = YDrsdpa+Drdt.2) where Cr5. ••. called quasi coordinates. i will be the number of nonholonomic constraints.-. r = l.

-.N (14-6) Equation (14.6).9) Example. Thus the system now appears to be holonomic because it takes k coordinates to specify the system. virtual displacements satisfy k Sxr = ^arsSqs.n =k+i (14. From Eqns.4) becomes k dqr = Y^ Prsdqs + Prdt . the coefficients Drs and Dr will be functions of all k + i generalized coordinates QrThe displacement components in terms of the generalized coordinates are.8) ^Ir = Y>l^^^^'i«'' s=l r = l.6) and (14. r = l.-.4) to eliminate the superfluous coordinates in Eqn.5) and (14. and polar. (14. {x. s=l r = l.7).y). r = l. We have q = xy rt -yx q= / {xyJto yx)dt . 14-1). the degrees of freedom. Two possible choices of generalized coordinates are rectangular. (14. (14.--.N (14.5): k dxr = ''^arsdqs s=l + ardt .1) and (14.6). First consider a particle moving in a plane (Fig.Gibbs-Appell Equations 247 Equation (14.4) is equivalent to Eqns. (14.d).9): dxr = j:P^dqs s=l '^^ + ^dt. r=l.5) Next we relabel pg = QS .k and use Eqn. (5. as usual.n (14. In general. defined by dq = xdy — ydx It is easy to show that this is nonintegrable (see Section 2.--. s=l k r = l. A possible quasi-coordinate is q. given by Eqns.N (14. and no multipliers will be needed. (14.--.3).7) Comparing Eqns.-. we see that we have reduced the number of dqr upon which the dxr depend to k. (r.

9.12) We want to obtain a fundamental equation similar to Eqn. respectively. repeated here in the new notation: N 'y^^{mrXr ~ Fr)5Xr = 0 N (14. namely Eqns. (11. (3.39).8) the total rotation about the axis of a spinning top is q. 14-1 so that q is twice ttie area swept out in time {t — to) by the position vector. and 4> are the spin. the virtual work is obtained using Eqn. (14. (14.8): N ^ Fr5Xr r=l N k k Qs^qs = ' ^ F r ^ OirgSqs = ^ r=l s=l s=l .12) in our new coordinates. from Eqn. where dq = dip + cos where ip.38) and (3.^r ~ Fr)AXr = 0 r=l (14. (3. First.248 Analytical Dynamics Fig.10) y^{mrXr N — Fr)AXr — 0 (14. 14. For example. the total rotation about a given line of a rigid body is often a convenient quasi coordinate.2 Fundamental Equation Fundamental Equation with Quasi-Coordinates. Recall the three forms of the fundamental equation established in Chapter 3. and precession angles. nutation.11) y^(mj. As a second example. We will come back to this problem later.7).

--. Define the acceleration.N (14. function by (14. r = l. (14. r = l.15) into (14. or Gibbs. r = l.12) and use (14.Gibbs-Appell Equations so that iV Qs = '£Frars r=l 249 (14. then fc Xr + Axr = 22 '^rsiQs + '^^s) + terms without the (js .6).N Thus k Axr = J2(^rsAqs.13): N y ^ nirXrAXr r=l W r=l N k — ^ Fy. =0 y ^ rrirXrAxr — / ^ QgAqg = 0 s=l (14.--.-. Note that this involves a mixture of rectangular and generalized coordinates.3 Gibbs' Theorem and the Gibbs-Appell Equations 1 ^ G=7:T.rnrxl ^ r=l Gibbs' Function.13) From Eqn.17) . 14. s=l r = 1. (14.16) which is what we wanted to derive.--.15) Substitute Eqn.14) Consider another possible acceleration Xr + Axr'.N k Xr = y ^ OirsQs + terms without the QS . k Xr = ^arsqs + ar .N (14.2 ^ arsAijs r = l s=l A .

18) which proves the theorem.-.-.250 Analytical Dynamics Note that this is similar to the definition of kinetic energy. Given the displacements and velocities at some time t. Substituting Eqn. (14. Gibb's Theorem.k (14.16): / fc \ 1 T V k ^[G -^Qsiisj \ s=l J = -^mrixr+Axrf-^Qsiqs r=l -. N -^'ll'^r^r r=l -. Gibbs-Appell Equations. s=l r = k + l. Let ijs be the actual accelerations and let Qs + Aijs be possible ones.7): k Qr = J2 ^rsq. namely. (14. dG Qs = ^ ] dqs s = l.14) into (14. These equations are the first order necessary conditions associated with Gibbs' Theorem. (14. the accelerations at that time are such that k s=l is a minimum with respect to the QrThe proof is as follows. obtained from Eqns.s + Pr . and Go does not contain the Qr.19) Also to be satisfied are the constraint equations. N = -^mriAXrf r=l 1 ^ + s=l k Y^Qsqs s=l / N + \r=l + Aqs) k \ i'YmrXrAXr-'YQs'!^qs] s=l / = ::J2mr{Axr)^ >0 ^ r=l (14.n (14.We now state the following. except that accelerations are used instead of velocities. Gi is linear in the Qr. Form the change in the function just above and use Eqn.20) .17) gives a function of the form G = G2 + Gi + Go where G2 is quadratic in the ijr.

Gibbs-Appell Equations 251 Equations (14. Equations (14. k 3.4 Applications Particle in a Plane. Consider the work done in a virtual displacement to get ^ QsSqs s=l and hence the Qs4. 2.20). (14. either [x. respectively. 14. They were later discovered independently by Appell who first realized their full importance.19) and (14. (14. as shown on Fig.20) serve to determine the equations of motion of a dynamic system.19) were first discovered by Gibbs but attracted little attention. the following steps are required. but only k of the g • The k preferred Qr may be either generalized V or quasi-coordinates. 2. 3.19). The Qr are in general a mixture of generalized coordinates and quasi-coordinates. As mentioned previously. 1. We pick coordinates (r. The Gibbs-Appell equations are equivalent to Kane's equations (see Baruh and Kane and Levinson) Solution Procedure.19) and (14. Note that generally all the g^ and (jr will appear in G. 6) serve as generaUzed coordinates. Remarks. q) defined by r'^ = x'^ + y^ dq = xdy — ydx . Obtain G by expressing the x"^ in terms of k of the (jr (see Eqn. 14-2. To solve problems using Eqns. Form the equations of motion from Eqns. (14.17)). y) or (r. 1. Let a particle in a plane be subjected to a force with radial and transverse components R and S. Determine k = N — i. the degrees of freedom of the system.

q. We have q = — {rsm9){rcos9 Thus dq = r d9 and 5q = r^59 xy-yx + rdcos9) = {rcos6){rsm9 — r0sm6) = r^9 .r. 14-2 Coordinates r and q are generalized and quasi-coordinates. q. To form the Gibbs function G from Eqn. The generalized forces are obtained by considering the virtual work done by R and S.17). r. respectively.q) y{r. x and y are needed: rr = XX + yy r^ + rr = x"^ + XX + y"^ + yy q = xy -yx q = xy + xy — yx — yx = xij — yx These expressions are to be solved for x = x{f.r.19) they do not enter into the equations of motion.q. (14.q. (14. q) and substituted into and y = G = ~m{x^ + f) The result is G = -m Ir 1 1 2' ^^ r H—KQ where all terms not having the factors r or q have been omitted because in view of Eqn.252 Analytical Dynamics Fig. r.

G = -M(i' +f+f)+Y: mriC? + € + i^r) (14. Analogue of Koenig's Theorem. dV Using the identity of Eqn. with accelerations replacing velocities.58). -Y = oq 253 Consider the special case of central force motion with conservative force.21) . and q — a = constant. the second of these integrates to r^ + 2V + — = 2h = constant which is the energy integral. (14. in this case y = 0. (1. or to obtain m[r^ \ =R. Recall that Koenig's theorem states that the kinetic energy of the body is given by Eqn. (9. Eqns. 14-3. Let G be the center of mass of a rigid body and fix an axis system at G that does not rotate relative to an inertial frame (but the body may rotate) as shown on Fig. r Now we apply the Gibbs-Appell equations. for the same situation.14). we have immediately.19). m \r \ s. Since the Gibbs function is analogous to the kinetic energy.Gibbs-Appell Equations and therefore 5W = R6r + SrSe = R6r + Sr-^ Consequently.= —m—— "f" I dr R = -rr.

'nr. Let a rigid body move in a plane (Fig.yr. but 6 varies with time.254 Analytical Dynamics (xr. We note that r is a constant for each particle. 14-3 where M = Y^ rrir is tlie mass of tlie rigid body. C.Vr) A 0 A j Fig.Zr). 14-4). Consequently. Two-Dimensional Problems. = —rsinOO C = —r cos OO"^ — r sin 99 T] = r sin^ rj = r cos 99 f = —r sin 99 + r cos ( y so that /•'2 I "2 2n2 I 2M C + ?? = r e / +r 9 Fig. = r cos 9 C. (£r. 14-4 .

Gibbs-Appell Equations 255 Substituting into Eqn.22) where p = x -\.21) 1 9 1--0 2 •' 2 [14. G = -M{c'9^ + c2A4) + ^r -Ma2 \ 12 . 1 1 /I 2fl2 .y is the acceleration of G squared. We first get the rolling without slipping condition (Fig.22). Letting c = h — a^ AB = A'B be = a{9 + 4>) acj) = c9 a(j) = c9 ad) = c9 From Eqn. (14. Cylinder Rolling in a Cylinder. and the r'^9'^ term has been omitted because it does not contain any acceleration factors. / = YI'IT^T'^'^ is the moment of inertia relative to an axis passing through G and perpendicular to the plane of the motion. 14-5) by noting that A' is at A when 9 = Q. (14.

In the preceeding two examples. Let the angular velocity of the body be U = UJxi + UJyj + LOzk.256 Analytical Dynamics Because k = 1. k'} frame is parallel to the fixed frame with origin at G.{t) € C^ about the z-axis.j'.j. Equation (14.19) then gives the equation of motion: -MgcsinO = -Mc^O 3 0- ^+--sin^ = 0 3c which is a form of the equation of a simple pendulum. Sphere Rolling on a Rotating Plane. The plane rotates with variable rate Q. the situation in which the use of quasi-coordinates and the Gibbs-Appell equations is particularly advantageous. Now we consider a nonholonomic system. 14-6). a). the rolling constraint is used to do this: 2 G = ^M{c'e^ + c^e') + ^Ma^(^9 3 G = -Mc^e"^ + terms without 6 4 Since the contact force does no virtual work. we must write this in terms of only one acceleration component. Consider a spherical rigid body with radius a and radial mass symmetry (i. If the plane were at rest. y. the rolling-without-slipping conditions would he X = acOy and y = —acOx. the systems were holonomic and the equations of motion could have been obtained by more elementary means.e. the mass density depends only on the distance from the center) rolling without slipping on a rotating plane (Fig. the only given force doing virtual work is gravity: SW = Mg5{c cos 9) = -MgcsindSB so that Qe = —MgcsinO. the center of mass of the sphere.k} frame is fixed (inertial) with origin at the center of rotation and the {i'.If the sphere were at rest on the rotating . The {i. The rectangular coordinates of the center of mass relative to the fixed frame are {x.

we choose x. We choose the five coordinates x. aijy = X + ily + Cly (14.24) Using the analoque of Koenig's theorem.23).Gibbs-Appell Equations 257 Fig. qx. i = 2 and k = 3. Combining the rotating and rolling gives the nonholonomic constraints: X — acOy = —r^y (14. y. x = —fiy and y = Qx. Differentiating Eqns.y.25) where all non-essential terms have been omitted. the Gibbs function is^ G=\M (f + f) + \l [iil + g2 + q^-j (14.23) y + acjx = fix Thus there are one holonomic {z = a) and two nonholonomic constraints on the motion so that L' = 1. (14. and q^.<ly. 14-6 plane.26) aijx = —y + ^x + tlx . % =^y : iz= ^z (14.) The Gibbs function must now be expressed in terms of the acceleration components of three of the coordinates. and where / is the moment of inertia of the body about any axis passing through G. (For a body with radial mass symmetry any axis passing through G is a principal axis of inertia and the moment of inertia about all such axes is the same. Qz where x and y are generalized coordinates and the three quasi-coordinates are defined by qx=^x .

(14. (ii) 2 the body is a homogeneous sphere.24) dx — a dqy = ~^y dt dy + a dqx = ^Ix dt so that virtual displacements satisfy 5x — a 6qy = 0 5y + a Sqx = 0 Consequently.25) gives Analytical Dynamics +^ {-y + nx + nxy + \iql (14. From Eqns.258 Substituting these relations into Eqn.27) Now suppose that the external force system acting on the body has been resolved into a force F_ = Fxi + Fyj + F^k acting at the center of the sphere and a moment M = Mxi + Myj + Mzk about the center. Eqns.23) and (14. = (FX + ^)^^+ (Py ~~-)^y + ^^^'i(14. 5z = 0. of course. so that / = -Ma'^.28) where. We are now in a position to apply the Gibbs-Appell equations. the work done in a virtual displacement is FJx + FySy + F^Sz + MJq^ + MySqy + M.6q. the result is Mx + -^{x + ny + Cly) = F^ + FyM. (14.^x ..nx) = Consider the following special case: (i) the rotation i7 = const.19).29) My + — (y . (14. (14. and (iii) there is 5 .

J. My. %.z. y. and I^ and let the mass be m. Use the Gibbs-Appell Eqns. z) are the coordinates of the center of mass relative to {/. and q^ are the components along {i'.Gibbs-Appell Equations 259 no external moment acting on the sphere. k'} be a body-fixed frame aligned with the principal axes of inertia. j ' . K} be non-moving (inertial) axes. ly.q3. Then the equations of motion of the mass center reduce to: 7x + 2VLy = .2(4 Iy)u)xU)yUz\ where [x.Ix)uJz^x^y + hf^l . Let {i'. and where qx. to generate the equations of motion. three of which are called in this case Euler's equations. 14/3. 216): G = -M (. and qi.[4cj2 .j'. Then the Gibbs function is (Pars. 14/4. Fill in the details of the particle in a plane problem. Fill in the details of the sphere rolling on a turntable problem. Consider a rigid body moving in space under the action of any given system of forces. and w^ = q^. which are generalized coordinates. K}.r2 + f + z^^ + .q2. Notes 1 See Section 11. The system is holonomic with 6 DOF. pp. . Suppose the resultant force has components F^.h)ujyOJz(^x + Iy<^l -2{Iz . Choose as coordinates x. k'} of the angular velocity of the body. Fy. ^y = Qy. and My about the center of mass along the body-fixed axes.2{Iy . and Fz along the inertial axes and the resultant moment has components Mx. Ux = Qx.p 7y -2nx = ^ ^ (14-30) These hnear equations are easily solved in terms of convolution integrals.1 PROBLEMS 14/1. which are quasi-coordinates. Let the moments of inertia be Ix.y. 14/2. Fill in the details of the cylinder rolling in a cylinder problem. J. and {/.

Consider a holonomic system with all forces embodied in a function V = V{qr. the variation of L is^ « = E(f** + f*) r n (15-2) with.n 261 15.t).28). the system is not conservative because in that case V = V{qr) is required.35).1) implies dL Pr---=0. Recall that the generalized momenta are der=l fined by Eqn.Chapter 15 Hamilton's Equations 15. (Strictly speaking. Because the system is holonomic. (6.4) r = l. the Sqr are independent. (6. repeated here: dL PJ. (8.1 Derivation of Hamilton's Equations Another Fundamental Equation. dqr r = l. Eqns. Since L = L{qr. dqr Thus Eqn. leading directly to Lagrange's equations.3) .1) where n (the number of generalized coordinates) =A. the fundamental equation in this case is E dt \dqrJ dqr Sqr = 0 (15. as usual. (15.) Prom Eqn. =—^ .-.qr.25). Y^ = V ] .n (15.t).-. (the degrees of freedom).

n (15. (15.4) into (15.Pr.If the result is substituted into Eqn.5) to get 5H = 2Z Qr^Pr + \]Pr^Qr r r ~ ^L 5H = ^{qrSpr r .7) Next recall that the kinetic energy in generalized coordinates is given by Eqn. Hamilton's Equations.9) gives E dH qr = ^~.^ .6) Form the variation of this function and use Eqn.\ Oqr = 0 (15.262 Analytical Dynamics Substituting Eqns.".3) and (15. (6.--. Note that Eqns.6).8) Since ars is nonsingular. Now take the variation of this function: Combining Eqns.5) Pars calls this the sixth form of the fundamental equation. H will be of the form H = H{qr. we may solve these equations for the qg in terms of the qs and ps. . this gives r = l.3) and therefore dL dT Pr = —^ = —^ ^y^arsqs dqr dqr V + br\ r = l. ^ (15.n . Define the Hamiltonian function by H = J2prqr-L r (15.t). (15. (15.PrSqr) (15.7) and (15.4) are just Lagrange's equations.11) OPr . dH Pr = . (15. (15.10) But because the 5qr and 5pr are independent.2) gives SL = 'Y2iPrSQr +PrSQr) r (15.

Thus for a natural system. Thus.15) T + V = E = h = constant. (15. An explicit form of H for a natural system^ may be obtained as follows. T = T{qr. using Eqns.n (15. they are to be treated mathematically as of equal stature. (15.6) and (6.5) a natural system is one that is holonomic. They give the motion in the phase space P.16) . the H function becomes H = Y. Their special form is called canonical.qr) and V = F(g^).4) give 6^ = 0 so that Eqns. defined by: / qi\ {q. In such a system.% * + 4-5^^^~^ (15. (15. (6.11) this is dt . If Eqns. Although these two sets of equations have different meanings. the Hamiltonian is equal to the system mechanical energy and is an integral of the motion. and conservative.^ + 4 . (15.14) Consequently. H ^ H{t) and from Eqn. The second set expresses the dynamics. Next we compute the total time derivative oi H.Hamilton's Equations 263 These are Hamilton's equations.13) H = constant. scleronomic.Y. Recall that (Section 3. there results — / ^ CgrPr r (15.11) follows directly from the definition oi H and is equivalent to the set Eqns.13) Natural Systems.3). (15.14) are inverted.6). (15.p) = Pi eP ciE 2n (15. (15.^rsqsqr-{T-V) = 2T-T +V (15. using this and Eqns. Also. for such a system Eqns.12) \ Pn J Note that the first set of Eqns.8) become Pr E' . which is the definition of the Pr.

18) which are no longer of canonical form.17) 2^T. (15. (15. t).25). This will be the primary motivation for the developments in the remaining chapters. Pr^-^^ + Qr^'-i^^sBsr. Eqns. . to "solve" the dynamics problem in whole or in part).6) lead to gr = T ^ . (8.t).6) to obtain Analytical Dynamics r = s \ +y r s / (15. In what follows. Eqns.11).Pr. The important difference between L and H is that we regard L = L{qr. (15.6. It is clear that Hamilton's equations. 2. This will now be made explicit. r = 1.qr. that is as a system of uncoupled first order differential equations. (6. come naturally in this form.t).PrCrsPs r s General Systems.11). If there are nonholonomic constraints and forces not derivable from a function V{qr. Recall from Section 12. The functions L and H are sometimes termed descriptive functions. 3. one of the main aims of analytical dynamics is to find integrals of the motion (i. As remarked in Section 1. and (15.e. 1.-. In subsequent chapters we will identify other such functions.t) and H = H{qr.2 Hamilton's Equations as a First Order System Remarks. that is the canonical Eqns.264 Substitute this into Eqn.1 that Lagrange's equations always may be written in state variable form. because once they are known for a given system the equations of motion of the system can be produced. we will consider only holonomic systems with forces derivable from T^(g^. n (15.34). Remarks. 15.

(15. We can also write these equations compactly in matrix form. Set Xn+1 = Pi x\ = <?i (15.19) ^n — Qn X2n = Pn Xi dH dpi X:n+1 dH dqi (15.20) ^n — dH dpn X2n = dH dqn Then Hamilton's equations are in state variable form.3 Examples Example.Hamilton's Equations 265 Hamilton's Equations in First Order Form. Use of Hamilton's equations is often a convenient method for solving specific problems. and we first illustrate this use by obtaining the . respectively. Let f qi \ ( dH/dqi dH/dqn dH/dpi \ X = Pi Hz \Pn J and define the 2n x 2n matrix Z by \ dH/dpn J where / and 0 are the n x n identity and zero matrices.21) 15. Then Eqns.11) are x = ZHz (15.

is a known function . Only the motion of the mass centers of the bodies is of interest. It is desired to obtain the equations of motion of the system. massless tether.^ = -mglsm9 + |sin6' = 0 Example. p.cos 6) .cos 6») H = p0e~L = ^mfe^ + mgi{l . T L = = \mi^Q^ .^ We next consider a more substantial example. The length 2. 6) as generalized coordinates.266 Analytical Dynamics Fig. For this system.cos 6) Applying Eqns. 15-1). n e = dH opQ P0 mt'^ ^ .11). (15. 2n = mi e -— = -—: =^p0 Thus. we choose (r. Pe = .mgi{l . ^mfe'^ de V = mgi(l . The system is natural with three degrees of freedom. Figure 15-2 shows two bodies of equal mass connected by a rigid.. 15-1 equation of motion of the simple pendulum (Fig. The system is traveling in a planar earth orbit.

15-2 of time. The potential and kinetic energies are (using Koenig's theorem for the latter).Hamilton's Equations 267 Fig.^ — 2zr cos 0 The L function and the generalized momenta are „2A2 L = — \r"' + r'^'p'' + z" + z"{p-0?) +^ { - 2 Vn + r2 dL Pr = ^^r = mr dr de = —mz {p — 0) z^{p-9)) dL = — = m(r'^p + . V = m n m n T=^{r'+r^p') where + '^{z + z\p-ef) rf = r'^ + z^ + 2zr cos 6 r | = r^ + 2.

defined by r=:ro.= + ^ ( r — 2. (15. Here we write Eqns.4 Stability of Hamiltonian Systems Variational Equations. mr'^ I . . mr'^ 1 siPp+Pe) .11) in vector form as q = Hp. p= -Hg (15. p = ojt. the Hamiltonian function is 3 Analytical Dynamics r=l Hamilton's equations then give the equations of motion: dH ?= 1 —Pr ^— = opr dH opp • dH opff m 1 .268 Consequently.cos^) Pp = —— = 0 — op (p is ignorable) ( 1 ^^ " dH miizrsinO ~"5^" 2" Of particular interest in applications is the "spoke equilibrium". CJ and ZQ are constants. o 1 raz^ mu. 9 = 0. It may be shown (see Problems) that this equilibrium is possible only for a certain specific value of cj.22) .-TT irf [r + z cos 0) Pr = dH -^. z = ZQ where TQ. 15.

expanding. p{t) = p*{t) + I3{t) dH V dqn ) where a{t) and I5{t) are small perturbations.p Hp = dqi .Hamilton's where Equations 269 ( ^JL\ / ^ i \ f Pi \ . Recognizing that H* = H* .p*)+Hpg{q*. Eqns.24) P qq -^pq 13 It may be shown that the eigenvalues of such a matrix occur in positive and negative pairs.p/3 These equations are of Hamiltonian form with Hamiltonian H' = -a^H*ga + /3^ij.p*+P)=Hp{q*.^ q*+a = Hp{q* + +Hppiq*.« + ^^S^i/^^^ ^-a^H^^a + p'^Hlga = -Hg{q\p*)-H. . Substituting these into Eqns. since the coefficients of the matrix in Eqn.23) may be written a a T (15. and retaining only first order terms.p{q\p*)/3 But since q*.giq*. Further. (15. p* satisfy Eqns.22)..p*)P p*+(3 = -Hg{q*+a.p*)a Stability of Motion. and that a perturbed motion is q{t)=q*it)+ait). (15.p*+(3) -H. Now suppose (f{t\ p*{t) is a reference motion satisfying Hamilton's equations.23) ^ = -H*ga-H.p*)a a.22).Hq = dpi \qn J \Pn J dH V dpn Only the case of H not an explicit function of It will be considered. a = H*ga + H*pl3 (15. where T de notes transpose. (15.

15-3 (15. where we now take the state a. Recall from Section 8. (15.1 that an integral of the motion is a function that remains constant along a solution in state-time space {x.t). Al Analytical Dynamics -^Re Xoi ?^3 Fig.24) are real. Thus if Ai = a + ib is one eigenvalue then so are A2 = —a — ib. (15. as a. for example if 6 = 0 then Ai = A3 and A2 = A4). Taking the total derivative of this: (15. 15. these eigenvalues exhibit a "box form" (Fig. Since eigenvalues with positive real parts and those with negative real parts denote unstable and stable modes. some eigenvalues have zero real parts.11) were used. X3 = a ~ ib. speaking loosely we may say that the system is "half stable and half unstable". unless.Pr)F{x. = (qr.5 Poisson Brackets Definitions.26) d£ dt ^I'dFdH . 15-3). the eigenvalues occur in complex conjugate pairs.270 ^mn X4.t) = constant. . Plotted in the complex plane. ^ \dqr dpr dF dH dpr dqr = 0 (15.27) where Eqns. and A4 = —a + ib (some of these may not be distinct. This stability property may cause numerical problems when integrating Hamilton's equations. of course. respectively.25) dF^_dF_ ^ d£_.

-v) = -{u. r = l.27) may be written as ^ + {F. .t) and let c be a constant. v)) = 0 .H}^ {xr.2n where Eqns.H). u) = 0 (ii) (u.. v) = {u..Pr.-. Then the following properties of the Poisson brackets follow directly from Eqn.--. u)) + {w. fdu \ f dv (m) 9^(-'-) = U''J + r s ^ (iv) (u. ••. 2n. and the qr and pr are independent. Let u.--. by definition: iXr.^ = jo^ for r = n + 1.11) were used. (15. u) = (u. l. Consequently Hamilton's equations in terms of Poisson's brackets are Xr = {xr.s d .= qr for r = 1. {v. and w be class C^ functions of {qr.H).30) Properties..H)=J2( dxr OH dqs dps dxr OH dps dqs Since Xj. (15. {u. (15. v) .28): (i) {u. this reduces to {xr. ••.H)=0 (15.H) Thus Eqn. v.2n (15.Hamilton's Equations Define the Poisson bracket of F and H by 271 ^FH) = yi^P^H dFdH\_j^d{F.29) This equation is satisfied by any function F that is an integral of the motion. w)) + (v. u) = {-u. Consider {xr. n and a.n r = n + l.H) = oqi—n -—^qr] OPr F)TT r= =Pr~n. c) = (c. (u!.

If (f) and ip are functions of class C^ and are integrals of Hamilton's equations then {(f).^).^)+ ((</>. We show this by using Poisson's brackets. ^(0.i?) = -i{cP.^l. (15. (V.{i^.272 Analytical Dynamics Poisson's Theorem.29) gives We need to show that ^(</<. It is obvious that independent integrals cannot be constructed indefinitely by this method.)-{<j>. Furthermore. 0)) = 0 An important special case is when the system is natural. which of course we already knew. (15. The . (0. Then H ^ H(t) and. in view of property (i). ( 15. H = Y. ^)) + {H. 15-4). Since 0 and ^ are integrals. Eqn.Piii -L = -{PI +PI+PI) =1 ' 1 + V{i It is known that in this system angular momentum is conserved but linear momentum is not. We now prove the theorem. i l ) = 0 Using the properties of the Poisson bracket stated above and Eqn. F = H satisfies Eqn. because only 2n such integrals exist.29). may or may not be independent of the two used to generate it. v ) + ( < / ' . sometimes the new integral produced is identically zero.Central Force Motion. ^ ) + ((</>. tp) is also an integral. Consider a particle of unit mass subject to a conservative central force (Fig. however. This result provides a means of constructing a new integral of the motion if at least two are already known. ^).H)) + {i<|>. H)) + {cf>.H). ^ ) .H) = (^. Example .i.if) = ( ^ . For this system.29) and consequently i7 is a constant of the motion. {H.V') + ((0. This new integral.).

k = xy .Hamilton's Equations 273 m=l F = -grad V(r) Fig. similarly.yPz = (-x In fact.yx = xpy . 15-4 angular momentum and its z component are i = r X V = {xi + yj + zk) x {xi + yj + zk) iz = I. Next Poisson's theorem is used to get a third constant of the motion: dlz diy dqr dpr diz diy dpr dqr • PyZ .--T. the y component off. . taking the Poisson bracket of any two components of | gives the third.= 0 r ar dprdqrJ dV dV dV where the properties of the Poisson brackets were used and where dV_ _ dV_dr__ xdV_ dx dr dx r dr dV _ 9 F 5 r _ ydy_ dy dr dy r dr We can show that.ypx This is an integral of the motion because 54 dt ~^\dqrdpr dV = xdV y~r dr ydV ^ a. is an integral of the motion. ly.

. using Eqns.274 Analytical Dynamics Finally we show that the components of the linear momentum are not integrals of the motion. The linear momentum is h =^ v_ and its z component is hz = v_.34 Similarly.. For a natural system.11).h) (15.^ m^yf^^^ "TTJ \d<lr dpr dhzdH\ dh^dH\ dpr dqr J z dV ^ _^dV ^ ^ r dr "at and similarly for the other two components. dqr qr &^ d(j) l^roA\ r = 2. (15.Pn) = h = constant. qn. apr oqi opr jr Therefore.31) and take the derivatives with respect to the last n — 1 of the Pr'. ••.-:Pn. 15.Pi.n 15.Pi. (15. _ J^fdhzdH ^^f.32) (15.6 Reduction of System Order Use of the Energy Integral.-. n 15.-.35) .31) Now substitute this into Eqn. ^ + ^ ^ = 0 dqr dqi dqr (15. say qi: qi =H(l2. -.= 2 . dH dH d(t> ^ 7 ^ + 7 ^ ^ = 0.qn..31) and differentiating with respect to the last n — 1 of the qr'.33 -j—= — =-iw = ^^'^ dpi Pi 1^ dpr r.k = z = Pz Thus dhz . substituting Eqn. (15. (15. the Hamiltonian does not depend on time explicitly and iJ = T + F is an integral of the motion: H{q\.32) into Eqn. Suppose that we solve this equation for one of the Qr.

+ Acos -px + k g Bsm-px g 1 f (Q k k \ y = .| * .sm -px .36) Equations (15.kyf +PI h 9 Using (j) as the new Hamiltonian.n (15. (15. Consider a system for which the Hamiltonian is H = -^{PI +PI) .Hamilton's Eq ua tions which gives f)ff 2 75 ^ = 5: = | | = . dpi Pi §^ dqr r = 2. k B k y = — + 7.Px.34) and (15..36) are a new Hamiltonian system with pi the independent variable (taking the role formerly played by t) and with (f) the Hamiltonian function (taking the role formerly played by H). Note that the new Hamiltonian system is of order 2(n — 1) and that the new Hamiltonian system is nonautonomous because ^ is a function of independent variable pi.Py. Eqns.kypx + i^k'^y'^ . Solving H = h iov x we have X = 'f>{y../ 7 + A cos -px + B sin -px \dpx + C 9 J \k 9 9 J Px A . • k'^ Py g'^ k Py = . px as the independent variable.34) and (15. Example.Qx) where x and y are generalized coordinates and k and g are positive constants. and y and Py as the remaining dependent variables..h) = — [{px .— cos -px + C k k g k g .36) give dy _ d(p_ _ py^ dpx dpy g dpx oy g The solution of these equations is obtained as follows: (Ppy dpi 9 k 9 A k"^ dy 9 dpx k k 9 T-.

11) as dH Px = . m—1 m which is equivalent to X'^dxm-i . Consider a system of first order differential equations: —r^ = Xr{xr).39) Then to complete the solution for the trajectories we need only to integrate (J'Xrn.rTi —2 (15. r = l. (15.C.X'^_idxm = 0 (15.--.C. the corresponding momentum integral may be used in the same way as the energy integral to reduce the order of a Hamiltonian system.37) Suppose we have found (m — 2) independent integrals of the motion given by fr{xr.38) r = l. this theorem tells us how to find the last two.t) = Or .= g ox Analytical Dynamics Px = gt + D We now have all the equations necessary to express the solution of the problem as x = x{t.A.^ Theorem of the Last Multiplier.A.B.m (15. — ^ O'Xm. y= y{t.B.D).^ . or equivalently dxi Xi dx2 X2 dxm Xr.276 The variable Px is obtained from Eqns. If we have found 2n—2 integrals. When there is an ignorable coordinate.40) .D) Use of a Momentum Integral. (15.--.

q2. ••. the energy integral plus one other.Jacobi's theorem of the last multiplier (TLM)^ then states that an additional integral of motion is given by im-l fm-l where r^ _ djfl.Although we have considered an autonomous system. " . The final integral. to obtain the time.Xm).Cm~i.-.37).Xm) satisfying this equation are called the multipliers^ for the system of Eqns.i ( c i .P2) =h (15.41) and M is any solution of the partial differential equation ^ — ( M X i ) + —-(MX2) + -OXi 0X2 + j^{MXm)=0 (15. (15.m) and X!^{xjn-i. (15. Application of the TLM to Hamiltonian Systems.Assume that two integrals of the motion are known.44) where X^{xm) = X^ici.46) F{qi.X'^_idXra) = Cm-1 (15.43) The TLM then provides an integrating factor for dxm . Xm) = Xm{ci.Pi.P2).Pi. The final system equation is (JrJbrj-i (J/b Xm 1 (15.Xm) = ^ m .Q2.P2) = a .a. c^-2.X'^dt = 0 (15. the TLM works for nonautonomous systems as well.38) are dqi dH/dpi dq2 dH/dp2 dpi -dH/dqi dp2 ~dH/dq2 (15 45) where H = H{qi.--. ••. ••.q2.a^m-i.Hamilton's Equations 277 where X'^_^{xm-i. thus H{qi.Xm-2) = / = / -jTiiX'^dXni-i f M' M .42) The functions M{xi.pi.1171-2) d{xi. Then m = 2ri = 4 and Eqns. Cjji-2iXm-iiXm). Now consider an autonomous Hamiltonian system with two degrees of freedom. is then obtained by one more application of the theorem.

F) _dHdF dHdF dipi.a) Now differentiate Eqns.50). For this system.46) with respect to a: dpi da dp2 da (15.49) dpi da dp2 da Using Eqns.46) are solved for pi and p2: / Pi = fi{qi. Then Eqn.h.a) .h. h.q2.Q2. (15.51) / da = constant Equating the expressions in Eqn.50) \da da J Thus / i dqi + f2dq2 is a perfect differential and there is a function K{qi-.a) such that dK = fidqi + f2dq2 and the third integral.41) gives a third integral as — I -—dqi — -—dq2 ) = constant (15.49) in (15. (15. we arrive at I -—-dqi + ——dq2 1 = constant (15. The integral may be put into a different form.48) J \dp2 dpi J where the coefficients are expressed in terms of qi. p2 = f2{qi. (15. dK = t + constant (15.q2. may be written as dK (15. (15.52) / I In summary.h. Suppose Eqns.47) and (15. (15.48).q2.278 Analytical Dynamics It is assumed that the Jacobian j _ d{H. Eqn.45) to dt and using the same procedure. the four integrals of the motion are H =h F = a I I da dK -dh='^'' . M = 1 is a multipher and we use it. and a.P2) dpidp2 dp^dpi is not zero. we arrive at the fourth integral. (15.

11) and (8. /2 = a and fi{qi.55) J^dq. pr = r and pg = r^9 so that Eqn. a) = h for pi.Hamilton's Equations 279 In many specific problems. the second integral will be a momentum integral corresponding to an ignorable coordinate. From Eqns. (7. Consider a particle of unit mass in a central force field with potential energy function V(r) then two known integrals are H{qi.h.53) give the third and fourth integrals as df J 'da^^ +q2 = -p ( = h pe = a Solving the first of these for pr gives Pr = frir.a) = J2h ~ 2V -^ Equations (15. (15.24). .pi.54) In this case.15) gives H=\{pi+^2pi)+y Since energy is conserved and 9 is ignorable. Now dK = fidqi + adq2 and the last two of Eqns. (15. Suppose 52 is ignorable.=t-to Example.P2) =h\ P2 = a (15. two integrals of the motion are H{r. a) is obtained by solving H{qi.55) then give the other two integrals as XoiV(C)~'"'° which are the same equations as were obtained in Chapter 10.

Problem 6/7. Hpq is the matrix \\d^H/dprdqr\\. See Pars or Whittaker for the proof. P. 15/5. 15/6. and Dynamics. 15/3. Not to be confused with Lagrange multipliers. Subscripts here will denote partial derivatives. Problem 10/1. Prove properties (i) . Jan-Feb 1994. 5 6 7 See Pars for the details. K. "Control of Orbital Drift of Geostationary Tethered Satellites".(iii) of Poisson's brackets. Problem 6/10. Vol. Anderson.280 Analytical Dynamics Notes 1 2 3 Recall from Section 3.S. Pars gives explicit forms of H for types of systems other than natural.. and Hagedorn. . Journal of Guidance. 15/8. Control. 15/2. Problem 7/1. Problem 7/3. Problem 4/2. 17. 1.3 that in the 5 operation t is not varied. Prove property (iv) of Poisson's brackets. No. for example. 15/4. 4 PROBLEMS Obtain the equations of motion using Hamilton's equations for the systems described in the following six problems: 15/1. with Xo = Yo = Zo = 0. 15/7..

qi and en are regarded as the independent variables. Pr = (l)n+r {QT . Pl. Pr ^ t) . however.t) . ". The motion is now envisioned as the motion of a point in phase space (see Eqn.12)). say qr =(f>r{qr. the motion of a dynamical system defines a continuous group of transformations in phase space that carry the point at i = 0 to the point (91.Pn) at time t.9^. one that will be exploited in this and the following chapters. 281 r=l.Chapter 16 Contact Transformations 16. use of Hamilton's equations oifers no particular advantage over Lagrange's equations. There is an obvious connection between the two sets qi and qi. the latter being the derivatives of the former.n (16. In the Hamiltonian formulation.1) .--. (15. the procedures and the amount of work required are essentially the same. In the Lagrangean formulation.". Rather.Pr. In the last chapter we have seen that for solving specific problems. the variables qi and pi must be regarded as truly independent.1 Introduction The Nature of Hamiltonian Dynamics. The equations defining this transformation are the solutions of Hamilton's equations. In modern mathematical terms. the Hamiltonian formulation offers a new point of view.

P1. x. y. In either case. For example. Equations (15. 8-4 neither xi. Consider the case of a natural system. As another example. for example a transformation of coordinates.n (16. Then L ^ L{qi. for the problem of Fig. Qr = OJrt + q^ Pr = 0. (15. y one is. for which H = constant.11) then give dH dpr These have solutions Pr — Pr = constant . say the initial conditions q^. a. for the problem of Fig. This case is not quite as restricted as it first appears.n (16. and for some choices more of the coordinates may be ignorable than for others. . . Other types of transformations are also of interest. Idea of Contact Transformations.2) r = l. is.282 Analytical Dynamics solution curve phase space (qr'Pr) Fig. 16-1). (p. A dynamic system may be described by any set of suitable generalized coordinates. Thus in this special case the dynamics problem is easily completely solved.-. it would be of great value if we could find transformations such that either the new variables were constants.3) where cOr = cvr (pg) are constants.) and from Eqn.2 are ignorable but for the choice x. z. 8-3 none of the rectangular coordinates. 16-1 Thus the dynamics problem becomes the study of transformations (Fig.6) H = H{pk). and suppose all of the coordinates are ignorable. Based on the preceeding observations. are ignorable but one of the cylindrical coordinates. r = l. or the new coordinates were ignorable. Ignorable Coordinates. namely y.

t) . A.Qn. one not necessarily giving the motion of a dynamic system: {QU-.n (16. Case W = Wi(qr.Pn) In this case.Qr.2 General Contact Transformations Definition. we can solve the first set of Eqns. In the following chapter. Pl. r = l. The functions R .t). W = W2{qr.Pk combined with any set of the Qk.-.-.-Pn) such that the differential relation is true: Y^ PrdQr r = ^Prdqr r + Rdt .Ps. by the Implicit Function Theorem.qn.t) . (16. satisfy Hamilton's equations) then the new ones must be as well. Qr. It must be remembered that not only the generalized coordinates but also their corresponding generalized momenta must be transformed.Contact Transformations 283 the problem would be completely solved.".5) for the pr in terms of the Qr. Because any set of the qk. there are four possibilities: W = Wi{qr.4) This equation defines a contact transformation (CT). the most important application of contact transformations will be covered. and then prove that such transformations have the desired property of preserving the Hamiltonian structure.Pn) >(Ql. Such transformations must have the property that if the original variables are Hamiltonian (i.e.dW (16.".t). The CT generates the transformation Qr = (f>riqs.t). and t.Pr.Ps. Such transformations are called contact transformations} We will first define and study contact transformations in general. and W = W4{pr. 16.-.6) d{Pl.Qr.Qrji)First suppose that the Jacobian 7^0 (16. W = Ws{pr. The function W is called the generating function of the CT. Pr = (f>r+n{qs.t).Pr.Pk may be regarded as the 2n independent variables of the transformation.5) These are 2n equations in An variables. Now consider a general transformation of 2n variables.

Next consider the case W = W2{qr. dWi Pr = ^ UQr .10).284 Analytical Dynamics IS and W can then be expressed in terms of the qr.Qr.7) Since the qr. The Other Cases. (16. we proceed as follows. that the rank is n — 1.--. t are independent. Using a Lagrange multiplier A to account for the constraint. Qr.t are not independent but are constrained by Eqn. Suppose.Qr.4) and (16. Qr.t)=Q Take the differential of this: (16. the qr. ^^^ _^ (16. (16.n There will be a Lagrange multiplier for each such relation 9.8) These are the explicit equations of the CT. If the Jacobian defined in Eqn.9) Now. (16.12) . then there is one relation e{qr. for definiteness. comparison of Eqns. take as the generating function W2 = Wi+J2PrQr Now (16. de ^^^—'^ Oqr r = l. and t and there i no relation connecting these variables.7) gives _ dWi _ ^^^ dWr _ _ ^ .Pr.t). Taking the differential of Wi: dWi = E -Q^d^r + E ^dQr + -g^dt (16.6) is zero.

8) gives: 285 dW2 = dWi+J2PrdQr + J2QrdPr r r = Y.PrQr-J2Pr'ir r r which gives ^^^ ^^^ (16.t).14) The next case^ is W = Wz{pr^Qr-.t).^rPr r (16. oPr (16.18) Of course.Pr.Contact Transformations Forming dW2 and using Eqns.15) and this gives __dW^ ^^^ __dW^ ^^'^ _ (16.P^^'ir + JL Q^^^ + ^^^ r r Since in this case the qr.7) and (16. this inipHes _dW2_ Oqr (16.13) _dW2.t are independent. (16.16) The final case is W = W4{pr. The generating function is (16-17) W^ = Wi + Y. in all these cases if there are relations among the variables one Lagrange multipHer will have to be introduced for each such relation.Pr. Example.The appropriate generating ^ function is W3=Wi-Y. Consider the transformation described by the generating function of the first kind: Wl=Y^ QrQr .

r = l.Pr) 7^0 Transformation of Coordinates. (16.20) . QrLet the inverse transformation be qr = Fr{Qs). One application of HCT's is the transformation of one set of generalized coordinates. _dW2 Pr — r^ _^ — -fr oqr n _ ^ _ ^' ~ dPr . that is. 16. the Jacobian of the transformation is not zero: d{Qr.286 Prom Eqns.14).-. From Eqn.Pr £ C*^.19) It is assumed that the Qr.3 Homogeneous Contact Transformations Definition. Next consider an example of a CT with a generating function of the second type given by W2 = J29rPr r Applying Eqns. Pr are independent functions of the qr-.n (16.4) this is defined by Y^PrdQr r = ^PrC^^r r (16. The special case of a CT with Rdt — dW = 0 is called a homogeneous contact transformation (HCT). Qr."^^ Thus the old and new variables are the same. (16. Pr — Analytical Dynamics n — ^r ) '^ — J-) " t'lT' oqr oQr Thus this transformation interchanges the qr and the Pr (except for a change of sign).8). W2 generates the identity transformation. Example. to another. (16.Pr d{qr.

20).Q2) . this is qi = Qicos(52 = Fi{Qi. T=lY. using Eqn.3). y = r sin 9 Letting qi = x.-. q2 = QismQ2 = i^2(Qi. and Q2 = 9.22) define the CT. For such a system (see Sections 6. Qi = r..n (16.1). The transformation is given by X = r cos 9 . The corresponding generahzed momenta are denoted Pr and Pr.22) Equations (16.PrQr r (16.P^7^' r = l. O We now obtain the explicit relations for the transformation of the corresponding generalized momenta. r s r s so that. (16. (6.Contact Transformations 287 where F^ (E C"^.We assume a natural system so that the transformation does not explicitly contain time. Such a transformation is sometimes referred to as a continuous point transformation.Q2) . y) plane. Note that the Pr are homogeneous linear functions of the Pr • Example. for a natural system.1 and 15.21) Thus Y^PrdQr r = r Y^Prdqr S that the transformation is a HCT.Mr r = lT. q2 = y. From Eqn.20) and (16. Pr = T. Consider the transformation from rectangular coordinates to polar coordinates for a particle moving in the (a.

(16. Pi. Consider Case 1 of Section 16.Pn be C^ functions of variables u and V. If we can find a function W such that Eqn.Pr) djPr) ^[g^) The same result holds for the other cases.2 for which Eqns.. In this section. but this is often difficult to do.Pr) ~^ {Qr.Pr)/d{qr. C.4) is satisfied.Pr) n«o/l^ T h e o r e m . Then the Lagrange bracket of u and v is r 1 sc^fdqrdpr dpr dqA ^^d{qr. This is sometimes called Liouville's theorem.Pr) is a C T if . Pr) is a C T with no relations between the Qr and qr.Pr) to {Qr. (16.20). Liouville's T h e o r e m . P2 = Py. L a g r a n g e B r a c k e t s . ••.288 Analytical Dynamics These correspond to Eqns.Pr).22) give expressions for the new generahzed momenta in terms of the old: dFi dF2 n ^ • n ^1 = = P i ^ 7 r + ^ 2 ^ 7 ^ =piCOS(y2 +P2SmQ2 oQi oQi dFi dF2 ^2 =Pi^7r +^2-H7r = Qi{-pismQ2 +P2COSQ2) where p i = p^. then the transformation is a C T .23) expresses the fact that the transformation is measure preserving.. d{Qr. Suppose the transformation {qr.8) define the transformation. we derive tests that are usually easier to apply. T h e transformation from {qr. Equation (16. Forming the Jacobian d{Qr. It may be required to determine whether or not a given transformation is a C T .4 Conditions for a Contact Transformation R e m a r k s .^ p . and P2 = pe- 16. Pi = Pr. --jqu. (16. Let gi. Equations (16.

26) For such a generating function to exist.Contact Transformations 289 and only if.P).Ps]=0 (3) [Qr. (16. The necessary and sufficient conditions for this to be a perfect differential is that Eqn.25) 0 if s 7^ r 1 if s = r (2) [Pr. Consider the rectilinear motion of a particle in a uniform gravitational field. s = 1. for all r. {q. But Eqns. one obtains dW = y^Prdqr r — r }^PrdQr dqr = E E^ dQs PA dQs + E \yiPr^)j dPs (16. The familiar solution of the equation of motion is: X = xo + xot + -gt .24) be satisfied. given by Q = q + pt+-gt\ P=p + gt We check to see if this is a CT by the previous two methods. X = xo + gt In our current viewpoint.P) diq. in Eqn. The proof is as follows. d{Q. (16. this is regarded as the transformation of the generalized coordinates and momenta at time zero to those at time t. and the theorem is proved. ••. with t fixed.Qs]^0 (16.p) -^ iQ. Example. (1) [Qr.Ps]=Sr.p) 1 t 0 1 . this must be a perfect differential. (2.25) are just the conditions for this to be true. Treating the Qr ^'iid P^ as independent variables. n and fixed t.4). First.

28) for all r. (15. q = Q-Pt+ ^gt^ . Inverting the transformation. Second. •••.Pr) = 5ij s=l (16. Oqr . s = 1. we check Eqns. (16.25). Up to now.29) .11).U2n{<ir-. n and fixed t.. we have been considering general contact transformations. r = l.t) and Hamilton's equations are Eqns.5 Jacobi's Theorem R e m a r k s . (16. =-—.Pr)Then it can be proved that^ 2n 'Y^[us.Pr -^ QrPr to be a CT in terms of Poisson brackets are {Pr.23) is satisfied and the transformation is a CT.28) and (16.Ps)=0 (16.27) The necessary and sufficient conditions for the transformation qr. One of the reasons CT's are important is because if the original variables satisfy Hamilton's equations. Consider independent functions ui{qr. the Qr and the pr. ••. then the transformed ones do also. (15.24) seems to indicate that there is some sort of inverse relationship between Poisson and Lagrange brackets. repeated here: dH Qj. Opr dH p^ = .290 Analytical Dynamics so that Eqn.n . H — H{qr. In the original variables. Comparison of Eqns. Relation Between Poisson and Lagrange Brackets.Ui]{us. (16. 16. Now we will consider the transformation of the generalized coordinates and momenta of a dynamics problem. as will be proved shortly..— . This is indeed the case. p= P-gt Thus conditions (1) and (2) are satisfied trivially and condition (3) is \0 p] = ^ ^ ^^' ^ dQdP _ dp_dq_ ^ dQdP again verifying that the transformation is a CT.--.

= [i.t) ..l2n. (16.Pr. Pr = Pr+nhl.n (16.1211.-. (16. .". (16. (16.29)) then they are also Hamiltonian for Pr. This is the converse of Lemma 1.-y2n.n are the general solutions of Eqns. i = l.24): dG dji _ Y^ dH dqr r ^'ir 9ji • ^ dH dpr r ^Pr ^li Lemma 2. To prove this. ••.". t = l. -^.72n. (16. Qr as given by Eqn.2n (16. Pr are Hamiltonian (that is. if they satisfy Eqns. Pr are 2n independent functions of 71. (16.31) l.t) then G G C^ and dG .31) in H to get H = 0(71. To prove this.31) as the functions just described and solve them for the 7i. ••.32) To prove this.30).-.t) t) . If Qr = Prill. .li\.t) = G{'yr. (16.Substitute the result into G to get a function H such that H{qr. QrJ2PrdQr = J2Prd<lr + Rdt-dW 291 (16.= [t.Contact Transformations Consider a CT defining new variables Pr.29).-.29) and if we substitute Eqns. r = r = l. If the equations of motion for qr. liqr.t and if there exists a function G(7i.--. 72n> ^) such that dG .29) and (16.30) Jacobi's Theorem. we use Eqns. Lemma 1. ••.7. two lemmas are needed.2n then qr.]. Pr satisfy Eqn. think of the functions of Eqns.72n.--.

33) (16.Qr. t) is of class C^ so that d'^F _ djidt d'^F dtdji i = 1.8).t) Using Eqns. (16.37) dji i = I. ••. this equation implies Eqns. dt dji _ Analytical Dynamics = E i = 1.36) From Eqns.29). (16.t)=F{ji. 2n (16.36): d x-^ f dt dqr d-fi ^ dQ d-fi J d dj.2n where Eqns.35) The function F{'yi. ••. Y:(P^ dt dqr pdQr\ '' dt dW dt ^ /dQr dPr dPr dQr\ _ ^ ^ f dqr dpr dpr dqr\ d'^W ^ Kdfd^i " ~dt~d^) " ^ Vdt^i r 1 dG ~ ~dt~d^i) + 'd^t' (16.--. (16. ^/dWidqr^ ^ V dqr d-ji S^ I dqr dWi dQr dQr d-yi „ dQr i = l. P r o o f of J a c o b i ' s T h e o r e m . ••.2n (16. 2n . Choose generating function and W^Wi{qr. (16.292 Now differentiate 5G dji _ ^ / a f f % ^ V dqr d^i 'dqr dpr dt d'ji dHdpr\ dpr d-fi J dpr dq.34) ap dt ^[dqr s-^ f dqr dt dt ^ dQr „ dQr \ ^' dt dt J ^ . dWi dt (16.34) ~ (16. Since the dpr/d-yi dqr/dji are all independent.24) were used.32) and (16.

p to new variables Q.29). that is.Q) = -ujq^ cot Q Equations (16.Contact Transformations where. (16. 7. P as defined by the generating function W = Wi{q. r 1 L = -mq .Qr satisfy Hamilton's equations with Hamiltonian function G. Then T = -mq^ . Now consider a CT from q. Pr. dH dli _ _ ^/dl^dqr^ dHdpr r \ dqr dji dpr d-fi sr^ f dq^dpr^ _ dpr^dqr^ \ dt dji dt djiJ 293 . __ „ was used.8) give dWi p = —-— = muq cot Q aq dWi dQ muq^ 2sin^Q . PrBy Lemma 2.] is the Lagrange bracket of t and 7i in terms of the Qr-. dQr OH* dPr dH* where the new Hamiltonian is dW H* = H + ^ (16.39) Example."^ Consider a linear harmonic oscillator with linear restoring force constant k. using Eqns. and where G = H + dW/dt and [t.2 2 ^ kq^ 2 kq^ V = 2 ' OL p = dq = mq "-^-l^"' where w^ = k/m.

Q is an ignorable coordinate with momentum integral H E P = — = — = constant The equation of motion for Q is now easy to solve: .^ 2 sm(a!i -I.294 Solving for q.p in terms of Q. Goldstein. thus H = LoP cos^ Q + uP sin^ Q = LoP Therefore. Eqn. P) = H{p. P. Notes 1 The term originated in optics where it has to do with preserving the contact point between wave fronts. The details of the analysis of the last two cases may be found in Goldstein. q). (16. See Pars or Goldstein.39) gives H*{Q. the well-known solution to this problem is obtained as 2E . 2P q= \ V mu) Analytical Dynamics smQ p = v 2 m w P cos Q Since the transformation does not contain t explicitly. they are called canonical transformations in some texts. 2 3 4 . dH Q = Lot + a In terms of the original generalized coordinate.a) This is an example of how a CT can be used to obtain a Hamiltonian for which all coordinates are ignorable.

and (iii) Using the Lagrange brackets or Poisson brackets tests. aCT? P = q" sin^p 16/4.k linear transformation of coordinates. Q^e^^VgT^.--. Consider the transformation described by the generating function W2 = '^fi{qi. Q = lnl-smp] . Qi = y^Oikqk and that the k generalized momenta also transform linearly. . show that the indicated transformation is a contact transformation by three methods: (i) Directly using the definition of a CT.t)Pi where the fi are any smooth functions. 16/3. Show that the generating function W2 = y~^ ajkqkPi generates a i. 16/1. Prove Liouville's theorem for the case n = 1. P = -(^)e-''P^/^T^ P = gcotp For what values of a and /3 is Q = q"' cos f5p . 16/6. (ii) Using Liouville's theorem. i Show that the new coordinates depend only on the old coordinates and time and thus W2 generates a continuous point transformation.Contact Transformations 295 PROBLEMS In the two following problems. 16/2. 16/5.qn.

Chapter 17 Hamilton-Jacobi Equation 17. 17-1 297 . (Sq) = Sq dt and Eqn.5) may be written: 5L ^PrSqr dt (17. (15. Figure 17-1 shows the situation for two generalized coordinates. Consider the motion of a holonomic conservative system in configuration space and consider a varied path such that the 5qr occur at a fixed time. In this case.1) qi qj Fig.1 The Principal Function Hamilton's Principle Again.

Prom Eqn. Define Hamilton's principal function by ftl S = / I to Ldt (17.4) We want.u)!^.1) from to to ti with the variations Sqr zero at the endpoints gives / 5Ldt = /^Pr^Qr SI Ldt = 0 /'to ' Jtn which is the third form of Hamilton's principle (see Section 4.ujlto.(g. 17-2).3)..2) the principal function will be of the form S = S{q°.^ to express S in terms of boundary conditions S = Siq'„ql. this replaces a point-slope specification of the solution curves by a point-point specification (see Fig. s = l. Principal the parameters being the q^ and the wj?. (17. instead. this relation may be inverted to give '^'^ = ^s{qr^ql..".ti) (17. the solutions may be parameterized by the q^ and the g^. In effect.--.^r^to. s = l. and we now proceed to replace the dependence on the wj? by dependence on the q^. ql = Ps{qr.(i)=p.3) where gj! = qr{to) and wj? = qr{to). Integrating Eqn.5) where Q^ — qr{ti). (17.n (17.n If the Jacobian of this transformation is non-zero.6) .t) and from Eqn.ti) (17. they will be functions of the form g. as follows.-. Then the Lagrangian will be of the form L = L{q^.to.2) Now suppose that all n integrals of Lagrange's equations are known. We see that the solution may be thought of as a 2n parameter family of functions.ti) .298 Analytical Dynamics This leads directly to Hamilton's principle.ti) • S = l. Alternatively.n (17. (17.t).3).^a.

First. (17. (17.7) Further. by Eqn. Variation of the Principal Function.4) gives Eqns. Eqn.5) with t^ and ti fixed. 17-3 for one q. which also varies the q'^. Since to and ii are fixed.8) Fig. (17. (17.1) applies SS = 5 Jto Ldt= Jto 6Ldt 0 (17. as shown on Fig. 17-3 . fix to and ti and vary the ql.Hamilton-Jacobi Equation 299 Substitution of these relations in Eqns. (17.5).

(17. their coefficients must be equal..2).9) provides the ql in terms of the g ? { and p^.9). . Indeed. From Eqn.J2P'M '^ = -Hi (17.11). = . that is.t) space. (17. fix the q^ and oj^ and vary ti (and hence also the ql).8) and recalling that the 5q^ and the Sql are being regarded as independent. (17. from Eqns. using the second set of Eqns. all the integrals of the motion are known. S exists. . they provide the solution (i.7) and (17. (17. r = l. if this can be done.. all the integrals) of Hamilton's equations. Remarks 1. we see that if S is known. and to and ti. (17. dS P.12) the total variation in S due to variations in all the 2n + 2 arguments of 5* is dS = Y^pldql .p[!) —)• [qltP]-) is a contact transformation (CT) with generating function S. = ^ .13) Thus the transformation (g|?. Next. We note that if L 7^ L{t) then the Qr are functions of time only of the form {t — to) and hence S* is a function of time only of the form (ti — to). Thus.7). the dynamics problem is completely solved. (17.^^. and (17.Y^pldq^r " Hidti + Hodto r r (17.n . then. (17. 55 Cti Li .9) From these equations. The goal has been to construct a unique trajectory through any two points in the event {qr-.300 Analytical Dynamics Comparing Eqns.^ . the first set of Eqns.. 0 dS P. . Together.e.11) Similarly. it may be shown that Finally.

vo. (17. Prom this solution.tof Fig. yo = y{to). (17. 17-4 where XQ = x{to). Eqns.yo. Consider a particle of unit mass moving in a plane under constant gravity (Fig.4). Next we do the inversion. Example. Thus none of what we have done indicates how to find S. are known. We will turn to the problem of finding S in Section 17.2) as follows L = T-V = ^{x' + f)-gy S = / Jto Ldt = S{x(i. but one with an important difference from L and H. We have shown only that if S can be found. Eqn.ti) This corresponds to Eqn.6): Xl -XQ Uo h — to vo h — to .Hamilton-Jacobi Equation 301 2. (17. We have started by assuming that the integrals of motion. the dynamics problem is --g{t . UQ = x{tQ) and VQ = y{to).to. S may be computed directly from Eqn. Thus S is another descriptive function. 17-4). (17.uo.2.3). The solution is known to be X = y = VQ Xo+Uo{t-to) + vo{t .

this is valid for all cases except the trivial one.302 Analytical Dynamics Substitution of these into the expression for S gives (the details are left as an exercise) c-/ ^ ^N b{xo. (17.xi. This is the solution we started out with.5) is S = -z'n{xi + Xo) cot n{ti — to) 2 ' ' "' ' ' ' sinn(ti 1 ^ 2 2\ _„j.9) to get Uo dS dxo dS oyo xi — xo h . yi-yo h-to 1 . the result of putting S in the form of Eqn. t_L ± \ nXiXo -to) .tof Now apply the first set of Eqns.n^x^)dt We can solve for uo = uo{xo. The equation of motion and its solution are X + r?x = 0 x = xo C S n{t — to) -{ O n Computing 5* as before T '{yi . ti) = 7: ^ Jtn {x'^ . Under this restriction. then the solution (all integrals of the motion) is readily obtained. Uo. ^ . this demonstrates that if 5 is known.yo. (17.xi. 17-5). Example .yo) .yi. r an integer (see Fig.h) = {xi XQ)'^ + {yi _ -yof -^gih .-2 sinn(t — to) •'-2 2 L — -X — -n X S{xo.Harmonic Oscillator.^g^h . ti — to = uniquely provided n{ti — to) ^ r-K. Clearly.

It is often convenient.ti) (17. 17.19) results in dS = dS = 2_jPr'^(lr ~ "^Pr^^Qr ~ Hidti + Hodto "^pldql so that dS = dar -pr — "^(^rdar — Hidti + Hodto (17.13) and (16.3).to.qj.9) UQ dS dxo -nxo cot n{ti — to) + — nxi sinn(ti — to) which is the solution we started out with. 17-5 so that from Eqns. to use other parameters Ur and (3r related to the q^ and p^ by a HCT (see Section 16.qj.9) gives the ql and p^ as functions of the 2n parameters q^ and p'^. (17. The solution of Eqns. (17.14) Using Eqns...ti) = S{ar. We seek such a transformation that leaves S invariant: S{q^. however.2 Hamilton-Jacobi Theorem The Hamilton-Jacobi Equation. (17.16) .to.15) (17.Hamilton-Jacobi Equation 303 Fig.

18) In this as r = l. (17.19) With this new notation.304 Analytical Dynamics These equations are the solutions to Lagrange's equations. Pr.23) r = l. We now turn to the task of finding an equation for S.21) (17.Hdt r r (17.22) (17. at Substituting Eqn. t) =0 dS „f dS \ ^ qi.XI l^rdar . = ^pOdgO r r (17.t).Pr. (16.g-^. then the solution to the dynamics problem is easily (17. fir define a HCT. oqr da ^ Pr .t) (17. d a . (17.20) so that -—-=Pr.n (17.23).15) becomes dS = J^Prdqr . we shall take to = 0.19). Also.24) . we have shown that if we could find the principal function.Q-. thus S = S{ql.21) into (17. that is.17) Previously. § = r f (17. we arrive dS — + H{qr. and suppress the superscript write S = S.--.-.ti) = S{qr.-. S. ^ / 3 .to.n dS _ where H = H{qr. from Eqn. Eqn.qn. Let ar.

If S* = S{qr.21) was used.21) and (17. (17.21) and (17.27) dtdai +E d'^S dqr }qrdai dt . (17. raising the question of which of the solutions solve the dynamics problem. Regard the qr and a^ as independent parameters and differentiate Eqn.24) then the integral's of Hamilton's equations are given by Eqns.24) also satisfies Eqns.19).--.r.24). Also. (17.21) and (17. ( such that dqidai dqidar. dS dai -A ' dt dtdai -E dqr dh dqr dt (17. T h e solutions of Eqn. 0 (17. (17. (17. however.24).Hamilton-Jacohi Equation 305 This first-order. The proof proceeds as follows.J a c o b i T h e o r e m .^ or sometimes as Hamilton's equation. (17. (17.22). is a solution (complete integral) of Eqn. (17. a complete integral of Eqn.24) is a function of class C^ containing n arbitrary constants ai.22). We know that the principal function. non-linear. Eqn.22). /O ( Hence we have replaced the problem of solving a 2n-order system of ordinary differential equations (Hamilton's equations) by the problem of solving one first order partial differential equation (the Hamilton-Jacobi equation).t) is a complete integral of Eqn. By definition. from Eqns. are not unique. (17.t.24) w. H a m i l t o n .25) dqndai dqndar.26) daidt dpr daidqr where Eqn. and thus solves the problem. ar. (17. partial differential equation is known as the Hamilton-Jacobi equation. T h e following theorem establishes that any solution of Eqn.

25).21) and (17. (17.306 Since S eC^. (17. Eqn. (17. dpx _ d'^S dt dtdqi ^ ^ d'^S dqr dqrdqi dt (17.--.r. (17.n (17. we arrive at dpr dt and the theorem is proved. d'^S dtdai d^S daidti Analytical Dynamics d'^s ' daidqr d^S dqrdai so that Eqns. (17. qi and use Eqns. and not the time rate of change along a trajectory.31) Also. repeating this for q2.26) and (17.28) If this procedure is repeated for a2.--iqn. /5. the following matrix equation results dqda dq dt dH = 0 dp (17.29) implies dqr ^ dH dt dpr l.33) . and f. ••.24) w.--.27) combine to give E dqrdai dqr dt OH dpr 0 (17. To get the other n of Hamilton's equations.21) to obtain d'^S dqidt dH dqi ^ ^ d'^S dH _ dq\dqr dpr (17. dH dqr r = l.32) Combining these two equations.30) which are the first n of Hamilton's equations. Diff'erentiate Eqn. a„. In view of Eqn. and forming a matrix equation as before.22). we proceed much as before.t.29) The first of these factors in an n x n matrix and the second is n x 1. from Eqns.n (17. (17. Note that we have written dqr/dt here instead of dqr/dt because we are considering the family of trajectories generated by independently varying a.

Hamilton-Jacobi Equation 307 Historical Remarks. Hamilton." . Eqn. in 1805. he was appointed professor of astronomy at Trinity. Lagrange's function states. teaching himself the known mathematics of the time. French.24) and Hamilton's canonical equations. Eqns. S = L dt. and was learning a half a dozen others. and used it to formulate the dynamics problem. In his First Essay on a General Method in Dynamics (1834) he introduced the "characteristic function". not having even applied for the position. while still a student. to the search and the study of which he has reduced mathematical dynamics. by the age of 17. Mr. Hamilton's function would solve the problem. comprehensive theory. by the age of 10 he was proficient in writing Latin. As Hamilton stated in his paper.11). and had started his revolutionary research in optics. He was by then a student at Trinity College. indeed.unifying the particle and wave concepts of light into one elegant. At this time. and Sanskrit. (17. Italian. namely. Dublin. in the impersonal style then in vogue: "Professor Hamilton's solution of this long celebrated problem contains. Hamilton's goal was to bring the theory of optics to the same "state of perfection" that Lagrange had brought dynamics. applying the methods he had developed in optics. he derived both what we now call the Hamilton-Jacobi equation. Attracted to foreign languages as a child. He then became interested in mathematics. it succeeded in resolving the most outstanding problem of mathematical physics of his time . By the age of 22 his work on optics was complete. Already he was called "the first mathematician of the age" and it was said that "a second Newton has arrived". Arabic. Greek. t /0 2Tdt. one unknown function. In the Second Essay on a GenJo eral Method in Dynamics (1835). He fully realized the importance of finding the function S as well as the difiiculty in doing so (he gave an approximation method). (15. Hebrew. born in Ireland. was the ultimate child prodigy. Hamilton then turned his attention back to dynamics. This function must not be confounded with that so beautifully conceived by Lagrange for the more simple and elegant expression of the known difi'erential equations. clearly motivated by the Principle of Least Action. In this work he also introduced the principal function. the principal function S.

and Poisson brackets) proved to be ideal as the basis for the modern theory of quantum mechanics. His main contributions to dynamics were the proof of the Hamilton-Jacobi theorem and putting HamiltonJacobi theory into its modern form. he was not productive during these years. Hamilton was awarded every honor possible to a scientist. Qr) = T{qr. Jacobi was a first-rate mathematician and he is perhaps best known for his contributions outside of dynamics. contact transformations. "it is the aim of many workers in particular branches of theoretical physics to sum up the whole of a theory in a Hamiltonian principle. (17. interspersed with plates of partially finished meals. including being knighted and being named the first foreign member of the U. As a consequence. Qr) + V{qr) = h = constant (17." Most remarkable. For such a system Hiqr. These contributions were given in a series of lectures in 1842 and 1843. He became reclusive. canonical equations. During his life. his study was found piled high with mathematical papers. specifically to the fields of elliptic functions. is that when. Hamiltonian dynamics has had a far-reaching impact on all of mathematics and physical science. When he died at age 61. Academy of Sciences. which were not published until 1866.S.3 Integration of the Hamilton-Jacobi Equation Natural Systems. and irregular in his eating and sleeping habits. which he regarded as his greatest achievement. Consider a natural system.24) . He was born in Prussia in 1804 and spent most of his professional life as a professor (he was by all accounts an excellent teacher). solution of algebraic equations.34) We see by direct substitution that in this case a solution of Eqn. As Bell states. Later in life he spent all of his mathematical energies on the development of quaternians. and differential equations. 17. alcoholic. number theory. the tools of Hamiltonian dynamics (HamiltonJacobi equation.308 Analytical Dynamics Beginning in his late 20's. about 100 years ago. Hamilton suffered severe psychological problems. experiments began to reveal the nature of the motion of atomic particles. Jacobi's life was apparently relatively settled.

qn-i.qn-i) . The function K{-) is called Hamilton's characteristic function.-.Hamilton-Jacobi is of the form Equation 309 S = -ht + K where K = K{qi. Note that Eqns. (17.h. suppose Qn is ignorable with corresponding momentum integral Pn = 1 = constant.22).. Then H{qi..o (17.36) By the Hamilton-Jacobi Theorem.37) where to is written in place of /3i.24) gives «{'.35) into (17. the first of Eqns.40) = T{qi.Qn) In this case we write K = jqn + K' (17. (17.n (17.21) and (17.-. r = 2.'^ Substitution of Eqn.<in) + V{qi.n (17. ~)-h (17.37) then gives time elapsed along the path. (17.35) and where we have taken a i = h.38) r = l. Now. the integrals of motion are given by Eqns.-.--.-. and Eqn.Qn-i.qi. ( (17. (17.39) The remaining problem is to find the function K. N a t u r a l S y s t e m w i t h I g n o r a b l e C o o r d i n a t e .-.qn.38) determine the p a t h in configuration space. in addition.-. The rest of the equations are dK dar -—=p^oqr .-.-.22) yields: dS _ dai dS dh f = t-.

We have L = ^{x'^ + y^) dL .22) the first n integrals are given by dK' dh dK' dUr dK' ^7 t-to ~/3r .h.a2.310 where Analytical Dynamics K' = K'{qi.qn-i. dK' (17.'y) where we have taken a„ = (17.-. (17. ••.21) give the other n integrals: dK' dqr = Pr . Consider again a particle of unit mass in 2-D motion in a uniform gravitational field (Fig.. The last of Eqns. We now return to the two examples of Section 17.42) 17 A Examples Example.n=Pn=l oqn The remaining problem is now to find the function K'. -gy OL .-.41) ions (17. 17-4).= -Pn = Qn where we have taken /3„ = —q'^.n — 1 = — 'V o .".1. r = 1.22) gives oa-n 9K' ^ 0 Qn + -Q. . Thus from Eqns.ra — ] (17. n (In -(In r = 2.

/ dS dS\ ^ Since H = h = constant and x is ignorable.(^-j^ + gkj t + jx + ^/2^ J^ Now apply Eqns. (17. (17. Hamilton's equation is dS . Eqns. J2g{k-rj)dn rk-y K' = v ^ / Thus v ^ dz/ S = .40) apply: S=-ht + 'yx + K'{y) Substitution into Hamilton's equation gives 1 2 Letting gk = h — --y . the solution of this equation is K'= f" .22) OS dS_ = -\/^9{k-y) dy ^ V ^ dv =Py = y as dS I — = -gt + ^2g{k-y) = -fi^ .21) and (17.35) and (17.Hamilton-Jacobi Equation 2 I 311 r=l ^ = ^{x^ + f) + gy = T + V = -{pi +PI) + gy We see that x is ignorable.

2 ' Zi 1 2 2 V = -n X .312 The last two of these may be written as Analytical Dynamics 2g{k .yo) = / 3 | which give J = Uo .^ E x a m p l e . Pi = -Xo p2 = vo so t h a t the solution may be written as X — XQ = UQt Mvo -y)= g^t^ .2/32gt If a projectile is launched at position {xo.2 T = -X . 1.= X ^ dx H = -x^ + -r?x^ Thus Hamilton's equation is = -{p^ + v?x^) = h dS 1 ((dS\^ 2 2\ n . these equations give at time to 7 = UQ -^/2g{k-yo) = VQ 29{k .'^vogt which is the well-known solution to this problem. k = yo + ^ . We note that h is the energy integral and k is the maximum height of the projectile. 2 ' Zi Zi dL . The problem has been completely solved.yo) at time to = 0 with velocity components (UQ^VO). For this problem. p = -—.H a r m o n i c Oscillator.y) = Pi + gH^ .

This shows that the solution of Hamilton's partial differential equation is not unique.ri^)~^''^dr] =-p where Liebnitz' rule for differentiating under the integral sign has been used. In Section 17.35) S was written as the sum of a function .5 Separable Systems Separability. however. we gave partial solutions of the Hamilton-Jacobi equation by writing the principal function as the sum of two or more parts.1. In this section we continue to consider natural systems. we found (by starting with a known solution) a different function S that satisfies Hamilton's equation for this problem than the one found here. Then S = —-r?o?t + n / \jo? — •rf' dr\ Equation (17. For example.Hamilton-Jacobi Equation We try a solution of the form 313 S = --n'^aH + p{x) Substitution gives ¥^^\m--'^\ p= n 10 Ja^ — rp drj where r/ is a dummy integration variable. Both functions.3.22) gives the solution of the problem as: da rx -n" at + na Jo {a^ . In Section 17. provide the complete solution to the problem. The constants a and to may be expressed in terms of initial conditions if desired. this may be written in the more familiar form^ X = Q!sinn(i — to) where /3 = n^ato. (17. in Eqn. 17.

h.a2.--.40). for example the famous three-body problem. are completely separable. say g„. if all coordinates but qi are . gives some results for systems for which the kinetic energy contains only squared terms: T=\Y. or {17AA) where the a^ are constants. are not separable. In this case we say the problem is completely separable. see for example Pars.-. (17. the most common case of separability occurs when there are ignorable coordinates. Some problems. not stated here. the central force problem is not separable in rectangular coordinates but is in polar (because in the latter case one coordinate is ignorable).314 Analytical Dynamics depending only on t and a function depending only on the Qr. (17. then a partial separation occurs. (8. The most general separable system of the type Eqn. if coordinates q^.4 were completely separable. is ignorable. Thus Liouville systems.4). They will not be reviewed here.43) Not surprisingly.4) are separable in the Hamiltonian sense. separability depends on the choice of coordinates. Solution of Separable Systems.43) is given by Stackel's theorem. the Hamilton-Jacobi equation is only useful when there is some degree of separability. For other problems. We have already seen that if a coordinate.''rer (17. Conditions for Separability. developed in the previous chapter. More generally. defined by Eqns.qn are ignorable then the characteristic function may be written as K = amqm H l-an?n + K'{qi. Both of the examples of Section 17. In practice. General methods have been developed for solving completely separable systems. the same systems for which Lagrange's equations are separable (see Section 8. For example.Such a separation is always possible when H ^ H{t). As a practical matter. but Pars (who devotes two chapters to the subject of separability). In particular. as expressed by Eqn. Goldstein. In general. each containing just one of the qi or just t. These methods depend on the theory of contact transformations. it may be possible to write 5* as a sum of functions. A classic and important special case of complete separability is that of linear systems written in terms of modal coordinates. ••. General conditions for complete separability are not known.

02.(?^'\ 2J\dip) dip 1 (dK i\ 2 / V d9 so that 2 + . because then K = a2q2 -\ l-anQn + K'{qi. {a2 — 03 cos 9) + -—rOo -f. we see that H ^ H(t) so that Eqns.24) and (11.36) apply. consider again the heavy symmetrical top analyzed in Section 11. we see that 0 and if) are ignorable so that K has the form K = a2(j) + a^ip + K'{9) The problem is thus completely separable. (8. Second.2 and shown on Fig. we arrive at H = + Isin^9 J + mg£ cos 9 First.Hamilton-Jacohi Equation 315 ignorable. the problem is completely separable. is 21 \ 89 J 2Ism'^9\ +mg£ cos 9 = h Substituting for K.h.15). The Hamilton-Jacobi equation. PfP = I4> siv? 9 + J{ip + (p cos 9) cos 9 p^ = J{ip + 4>cos9) Using Eqns. From^ cos 9 = h p 21 sm 9 2J dK' ~dF = \fm .35) and (17. Example. (17.10). "n) (17. Eqn.36).^ As an example of a non-trivial problem. we obtain L(?E\\^J^J?I^_»JL^e\\l-. the problem has been completely solved. 11-4. (11. which is always reducible to quadratures. (17.11) and (15.45) In this case the Hamilton-Jacobi equation reduces to an equation in gi. ".

McCuskey. (17.37): f ^^^ and the other two are I I {a2 — a^ cos 9) d9 sin2 9^F{e) f d9 ^^ ^ f {02. the first of these is given by Eqn.316 and Analytical Dynamics K' = f ^/Fi9)d0 where F{e) = 2IhThus K = a24> + «3V' + / jF{d)de -al J . PROBLEMS Solve the following three problems by the Hamilton-Jacobi method. 17/1.0:3 cos 9) cos 9d9 "^-T'l^m^j—sin^^yFM—^~^' The constants /i. Eqn. (17. It may be shown that K is equivalent to the action integral.CK3 may be determined by initial conditions. .21 mglcose ^(0:2 .40). where it is called the Hamilton-Jacobi-Bellman equation. Now apply Eqns. see cose)^ sin 0 and S is given by Eqn. (4. Problem 4/2. The details of the solution to this problem are left as an exercise.22). Notes 1 2 3 4 The same equation plays a central role in the subject of dynamic programming. a2.35). (17.

. Fill in the details of the solution to the second example of Section 17.5. Problem 6/7. 17/6.4. Fill in the details of the solution to the first example of Section 17. 317 Show that linear systems written in terms of modal coordinates are completely separable. 17/5. 17/3.4.Hamilton-Jacobi Equation 17/2. 17/4. Problem 10/1. 17/7. Fill in the details of the solution to the example of Section 17.

n (18. In this chapter.1 Variation of Constants Remarks.-.-.In) 319 /O (18.t) and assume that d{pi. approximation methods are introduced which are based on the Hamiltonian formulation of dynamics.t) = Pi{ci.4) .t) = Pii'yi.Cn.Xn. i = i. i = l.2) .Pn) dill.Chapter 18 Approximation Methods 18. General Case.n (18.^n.-.n (18.1) To account for the fact that the solution will change.-.t) +gi{xs.t) with solution Xi = Pi{Cs. Consider a general system of first order ordinary differential equations: Xi = fi{xs.t) Now suppose the system is altered or perturbed so that Xi = Mxs.t) • i = l.3) (18.-.5) .t) = fi{xi.-.--. beginning with the method called the variation of constants. It is clear that obtaining solutions (either closed form or in quadratures) of dynamics problems is difficult for all but the simplest problems. introduce new variables by replacing the constants Cj by functions of time 7i(i): Xi = Pi{7s.-.

n (18.t) where li{i) = Aibs. the matrix in Eqn. bn. (18. Xr = Pr{'yi{t). the unperturbed motion of a small body orbiting around a massive body is that of a Kepler ellipse for which the eccentricity. initial conditions. where the 7i are called oscillating elements. crosses these surfaces. Xr = Pr{ci. t) .1 gi-. T" 7. The main feature is that we let the constants of integration of the unperturbed system vary to get the solution of the perturbed system. that is Xi = Pii7sit). (18. (18.8) Comparing Eqns.t) +gi{ T^s-i'^j v ^ dpi — -^i — o.5). ^ = l. The solution of the perturbed system. (18.= E dpi -\ .3). ? = 1.t).6) The bi are constants to be determined by.2).1) and (18.7) and (18.-. Figure 18-1 provides a way of looking at this. t) = ipiih. fi{xs.••. The surfaces of constant Cj are solutions of the unperturbed system. One application of the method has been to orbital mechanics.n (18. giving (18. For example.10) This is the equation for the variation of the constants that was sought..--. for example. using Eqn. (18. Prom Eqns. i = l.7) From Eqn.n (18. Interpretation. If the motion . changing the value of Cj as time evolves.8): i = l.9) is invertible. By Eqn. Jiy^si ''I ~^'~ dxi dpi dpi ~dt ~ ~dt ~ dt ' i = l. ".320 Analytical Dynamics We seek differential equations for the 7s such that the solution of Eqn. n . (18.2).-. (18.4). is a constant. i = h-. The method just described is called the variation of constants or the variation of parameters.t).3) is in the same form as Eqn. (18.n (18. e. •-.9) 7.

-. Now suppose. (18.t) . 18-1 is perturbed (due to the presence of a tiiird body or to air resistance.11). Pr = Gr{ci. in fact. then the eccentricity will no longer be constant. that Eqns. for example). n (18. 18-2). say 7 = e{t).C2n. 18-2 Variation of Constants in Hamilton's Equations.--.t) . (15.-.1) are Hamilton's equations. dH Pr with solution qr=Fr{ci. The orbiting body is now instantaneously on a certain Kepler ellipse but an instant later it will be on another one (Fig.Approximation Methods solutions of unperturbed system 321 solution of perturbed system Fig. Eqns. C2n.12) • dH % ' " = ^ ' - ' " . Perturbed motion family of Kepler ellipses Fig. r = 1.

15) .Pr. \-Rr. .t) ] r = 1.Oqr .n (18.To account for the perturbation we let the constants of the unperturbed solution vary: qr = Fr{'yi.n ^ (18.18) r = l.13) where the Rr are C^ functions of the qj. r = l. the first set are direct consequences of the definition of the Pr.16) dH _ .-. =i?^. Prit) = Grhs{t).t) .'y2n. r = l.t) . dqr ~ dt '^j^^d^s^' _ ^'•' _ ^--^'••'" Comparing Eqns. For the unperturbed motion. ^.16) and (18. OPr Analytical Dynamics dH p^ = .". dqr dFr dGr (18. = 0.17) gives Z . Pr = Gr{-ji. n (18.t) We seek solutions of Eqns.^2n. ••.14) We proceed as previously. ^ 7 . ^—7.Ps.t) = R*{'yi.13) in the form qrit) = Frijs{t).". Note that only the second set of equations is perturbed because only these equations contain the dynamics of the system.n E dGr . (18. r = l. _ dpr and for the perturbed system dH dFr ^ dF (18.-. (18.-.t) Also let Rr{qs. ••. dH .t.17) _dIl_dG^^dG^.n (18.322 Suppose that the equations are perturbed dH Qr =-—..

(18.18) by the second set of Eqns. arbitrary virtual displacements satisfy 7= 1 '^ (18. r = l. sum each from 1 to n. (18. each coefficient of this equation must vanish independently.21) become 2n Qjj E[7i. (18.-. Equations (18. Special Case. the result is 2n 2n " ^ fdFrdGr 2n dGrdPA . giving 2n i=l n r=l r.2n (18. (16. Consequently.19).n (18.p Ebj.20) = 1 r=l "^^ Since the Sjj are arbitrary. using Eqns. n op (18.24) was used. An alternative method of obtaining the 7i(i) is available.18) by the first of Eqns.18).7^]ii = EK^•^ ^'J j = h-. (18.19) Now multiply the first set of Eqns.-.22) then.21) are equivalent to Eqns.14). i = l.t) .24) .21) where Eqn.2n (18.7i]7i = . If there exists a function U{qr) such that dU Rriqr) =-JT = Kil. Eqns. Prom Eqn. and subtract one from the other. (18. (18.19) and the second of Eqns. (18. (18.15).Approximation Metliods 323 Solution of these linear ordinary differential equations gives the functions Alternative Form.T ^ .

324 Analytical Dynamics Example. Eqns. (18. Pr dH + Rr dqr 1. Forming the Hamiltonian of the unperturbed system: T L Px = = \{x^ + f).2 . V = gy T-V=~{x^+f)-gy dL dx 2 = X dy r=y H Y^PrClr -L=-ix^ r=l + y^)+gy= -{pi + p^ + gy m=l Fig.13) are Qr dH dpr . 18-3). A particle of unit mass moves in a plane under constant gravity and some perturbing force R = RJ + Ry] (Fig. 18-3 Hamilton's equations are dH X — dpx y Px dH Px = TTdx = 0 OH = Py dPy Py = dH dy '9 The solution of these equations is Px = Cl\ Py = -gt + C2 X = Cit + C3 y = -yf^ +C2t + Ci Now consider the perturbed motion.

oar r = l. C2. 012 all others zero Equations (18. Also. for example. (17. the latter being functions of t. thus X = y Px Py Computing partials. i ( 0 ) . Consider a transformation from {qr. OPr Pr = .26) .. 74. these equations may be solved for the 7J in terms of four constants of integration.18) give the differential equations for the jf.-.t) is a known complete integral of Hamilton's equation.rt (18.-. Eqn. Note that this approach is not restricted to the case in which the perturbing force Rr is conservative. suppose '5' = S{qi.--. 73. y(0).^ + QT\ Oqr r = l. *7i + 73 = 0 . C3.qn. y(0).-) such that Pr = — .t) .". oqr p^ = --—.24). Let H be the Hamiltonian of an unperturbed system for which we have the solution.2 Variation of the Elements T h e M e t h o d .t) l2t + 74: = F2{-ys. R e m a r k .t) Gi{'ys.25) 18. 571 ' 973 ' 972 ' 974 ' 971 = = = 71^ + 73 = -^gt^+ 11 = Fi{ja. 72. these constants may be determined.Approximation Methods 325 Next replace the constants ci. specifically. ai. it is applicable to the more general case in which the unperturbed problem is not conservative: Qr^-T^.n (18. i^72 + 74 = 0 ./?. 71 = Rx 72 = Ry liRx{t) and Ry{t) are known functions. C4 in the unperturbed solution by 7 i .an. <) —— = 1 .Pr) to (a^.3 ^ + 72 = ^2(7s. from initial conditions x{0).

39).--. OPr Pr = -^—.dS (18. For the perturbed problem with Hamiltonian H + K the solution is still of the form of Eqns.23)./??-).30) R e m a r k s .By Jacobi's theorem. (16. Pr.30) imply that the a^. O-r as the independent variables (any combination of Qr. provided that the Jacobian is nonzero).t) ] r = l.i) . Pr that completely solves the unperturbed problem.Pr are allowed to vary. oar r = l.Pr) to (a^.(17.Hdt . /?r are constants.21) . thus a transformation satisfying Eqns. Pr to a set of constants a^.26) and (17.326 Analytical Dynamics We have chosen qr. Eqns. therefore Using Eqns. with Hamiltonian H + K.38). For this problem make the same transformation from {qr. For this problem. as required. For the unperturbed system.--.Oir. (18.29) where K is expressed in the variables a^. K = H* = 0 and Eqns.n (18.28) with (16. Pr. at But H = -dS/dt so that H* = K (18. Pr are functions of time given by Eqns. = ^Prdqr r . /3r are constants.26) is a CT with generating function S. (18. Pr could be chosen.31) except that now the a^. (16. (18.28) Comparison of Eqn.Then by Eqn. . the new equations of motion are ar==—z-. Now consider a perturbed problem. For the unperturbed system.n (18. (18.30).31) where the a^. in which the ar. (18.4) shows that this is a CT.Pr."pr-. the solution will be of the form Qr = Pr{ar. Pr = Pr+n{oir. S is the generating function for a transformation from the qr. (18.

from Eqn.3. the solution of this equation is S = —nat + / y2na — vP'p^dp .l + . The perturbed motion at any instant is one of the old motions whose elements are a^. . 24 First consider the unperturbed system. The method of variation of elements also has been widely applied to orbital dynamics. the exact Hamiltonian is H = -p'^ + Expanding about ^ = 0: m .\ ^ Using the methods of Section 17. K continually modifies the old motion. 15-1).but these elements are not now constants. approximately.Approximation Methods 327 This method differs from that of Section 18. n'^{l~cos9) (l_e„s9) = l . dS 1 ffdS\^ . Letting p = 6 and n^ = g/£. Consider a simple pendulum with mass 1 and length i (Fig. Pars has a detailed discussion of the variation of the elliptic elements in the two-body problem. + . For example. -p^ + -n 6^ is the first approximation and -n 9 is the function K of the previous section. = . the variation of constants.24).1... in that in the present method the constants that are allowed to vary satisfy themselves Hamilton's equations. /3j. "' "* 9' «' + .^ - 1 1 We call this the second approximation. in effect.^. (17.. Example — Simple Pendulum.^ Thus.

(18. we ignore second and higher order terms in a to get the approximation: a =0 which have solutions a = constant ^' + -at 13 =-a /3 = where /3' is a constant./?)] ~~da' ~ 24 Since 9 is small and 9 is proportional to -^/a./S)] dH* _ 1 a [3 . P with solution dS_ d9 P = . (17. We have H* = K = -^n^9^ 48 so that Eqns.21) and (17. Eqns.4 cos 2{nt . n -a\t~0 ^sin n 1 16 t-0 .30) give a 13 dH* 5/3 1 a^ [2 sin 2{nt -/3) .sin 4(ni . Substituting the solutions for a and /S into the expression for 9 gives 9 = A'sm./?) + cos 4{nt .328 Analytical Dynamics By the Hamilton-Jacobi theorem./3) + cos 4(nt . a and /? become functions of time.22).(3)] 12 = -^a'^sin^{nt-p) a^ [3 .4 cos 2{nt . Let A = \j2ajn denote the amplitude of the motion.da dS_ 6 = p = Vn \~smint-l3) V2na cos{nt — (3) For the perturbed system (second approximation).

t).Pr) to {Qr.t) .SW (18. the new period r is given approximately by: 27r 27r / A'^\ 1+ The constants a and /3' may be determined by initial conditions ^(0) and 18.34).t] = p(qr. (18.QrSPr+YlPrSQr r + r Y.3 Infinitesimal Contact Transformations The Transformation.Pr.Pr (18. For the transformation from {qr. Eqn.Pr.Approximation Methods 329 Thus the ampHtude is not altered but the period is.Pr.34) Note that these resemble Hamilton's equations.33) results in r + r SM = J2iPr-Pr)Sqr r r Y.{Qr~qr)SPr r so that dM --Q— ^ = Pr-Pr\ dM -Q^ ^ = Qr-qr .Pr) to be a CT. (18. .33) Substituting Eqn..Pr. (16.4) must hold. Consider a generating function r r where M G C^{qr. (18. write 9{qr. we -e-—. Now let M = ee where e -C 1 and 9 G C^{qr. the variational version is J2 PrSQr r = J^PrSqr r .t).35) Using the first of Eqns.t) = d (qr.32) into (18.

q^ = e dp dpr ' ' •Pr = dp "'' -e: ^dq. ". n r=l. dA^_ dPr de__ dPr dp_dp^_ dps dPr dp_ dpr ' ' (18. (18. (18.37): Qj.. 18-4). (18.36) and (18.34).34) and (18. 18-4 . take /9 = i J .38) become Aqr At dH dpr Apr dH dqr ' r = 1. from Eqns. r = l. to first order.Pr) in time increment At and the motion may be viewed as a finite sequence of ICT's (Fig.Pr) gets transformed to {Qr.36) Similarly. ••.330 Analytical Dynamics Hence. Then Eqns.39) At Thus {qr. dM _ dqr de _ dqr dp dq^ r = l.35).38) which defines an infinitesimal contact transformation I n t e r p r e t a t i o n . Combining Eqns.-. To see this.n (ICT). (18. . e = Ai^ r = 1.37) to first order. n (18.-.--.n Aqr = Qr ~ qr 'i ^Pr = Pr-Pr. (QpPr) Phase space t+At time F i g . Equations (18.38) may be viewed as first order approximations to Hamilton's equations. (18.n (18.

Pr. holding t fixed and retaining only the first order terms.= 89 dt e- dp dt (18. (16.42) resuhs in H{Qr.H) dp (18.Pr. then by Poisson's theorem 'di + ip.41) with dW dt dM dt e-TT. Now identify with H{qr.t)fiqr.t) = Y^{Qr .t) H*{Qr.qr)^ df + Y^{Pr .t) = H{Qr.p) + e dp H*{Qr.PT.t) is an integral of the motion of the original system. df f{Qr.Pr. using Eqn.t) = H{qr.t) = H{Qr.t)=H{Qr. By Jacobi's theorem.44) (18.H) = 0 and thus in our case. p) is the Poisson bracket of / and p.t) = e{H. p) .Pr.t) .Pr.t) dp (18.t)=H{qr.Pr. Eqn. Recall from Section 15. H*iQr.Pr. if p is an integral.40) where (/. H*{Qr.e{H.Pr.Pr) dpr r _ ^ _ ^ _ df dp dqr dpr dpr dqr (18.t) be any function and form the Taylor series expansion.Pr.39). (18. Let / G C'^{qr.(18.t) + f{qriPr.40) .t)+e 'di + ip. The converse of what we have just proved provides a means .Pr.45) Hence an ICT with generating function an integral of the motion leaves H invariant.Approximation Metliods 331 The Hamiltonian.H{ Combining Eqns.t).5 that if p{qr.38). (18.43) which is the Hamiltonian of the transformed system.t) dW dt (18.Pr.Pr.Pr.

n oqr Equations (18. r = l.18) for this problem. the generating function will be then an integral of the motion. (18. linearized solution as a starting point. Shown is a simple pendulum with perturbing force £ . Eqns. . n r = l.n Since q^ is the only qr to change and H does not depend on q^.45) would imply Eqn. Find Eqns. Fill in the details of the solution of the example of Section 18.44).= 0 f o r r 7 t 2 .43) and (18.38) yield in this case dp^ Qr . Pr = Pr . (18. ••. 18/2. r = 1.--.= eSn . dpr dp Pr-Pr =-e^—= 0. Completely solve the problem in terms of initial conditions for the case in which R^ and Ry are both constants. the transformation leaves H invariant and thus p is an integral of the motion. n r = 1. PROBLEMS 18/1. ••. The procedure is to seek ICT's that leave H invariant.1.332 Analytical Dynamics of getting candidate integrals of the motion. H is independent of qi. p = Pi = constant which of course we already knew. (18. suppose Qi is ignorable. Then choose p such that p = Pi so that —. Ignorable Coordinate.-. As an example of the above procedure. which is Qr = Qr + e^Ti . ^— = 1 opr opi dp . In other words. Use the small angle. -— = 0 . that is.Qr = ETTT.

Approximation Methods 333 Problem 18/2 18/3. Fill in the details of the solution of the example of Section 18. .2.

McCuskey. "Theoretische Mechanik". J. T.. published in Paris 1788. 4... Wiley. Newton. D.T. 1977.B. "Mechanica. 2. Petersburg. 2001. 1944.. Kane. E.. I. Sive Motus Scientia Analytice Exposita".A. First edition published by Cambridge University Press. H. "An Introduction to Advanced Dynamics". 4. "Mechanique Analytique". Addison-Wesley. J. d'Alembert. "Traite de Dynamic".. Rosenberg. published in St. Berlin. Reprinted by Ox Bow Press. 3. 1959. 7.. 1743.Bibliography Primary Sources 1. le R.. 5. Prentice-Hall. 1904. 3. Bhat. "A Treatise on the Analytical Dynamics of Particles and Rigid Bodies".V. S. and Dukkipati. Addison-Wesley. 1949. L.T. L. "Advanced Dynamics"... R. Lagrange. 1985. F.. Secondary Sources 1.A. Springer-Verlag. "Classical Dynamics". 1736. R. Euler. "A Treatise on Analytical Dynamics". 1979. published in Paris. "Dynamics Theory and Applications". First edition published in London.. "Philosophial Naturalis Principia Mathematica". and Levinson. Whittaker. Hamel.R. 2. Original Sources 1. 2. "Analytical Dynamics of Discrete Systems". Narosa. "Classical Mechanics". R. 1965.W. 335 . Goldstein. McGraw-Hill. many later editions..L. Fourth edition published by Dover. 1687. Pars. 6. many later reprints and translations. Plenum. 1977. Greenwood. 1950. D.

A. 11. 2. 10. J. C.W. 1937. 1988. published in Berlin.R. 1986. "Men of Mathematics". Dugas.. published in London.. "Vorlesungen uber Dynamik". J. Hamilton. republished by Touchstone. Soc. First and Second "Essays on a General Method in Dynamics". Liapunov. translation of paper published in Kharkow.J. Bell. Trans. Editions de Griffon. 7. "Advanced Rigid Dynamics".. Vol. Phil.336 Analytical Dynamics 5. II. 6. published in Paris.T.. 1892. 8. 1829. 9. "Traite de Mecanique Rationelle". 1955. Routh. 1896.. 1928. 1884. P. Historical Sources 1. .. 1834 and 1835. R. Appell. respectively.G. E.F.. Simon and Schuster. "Histoire de la Mechanique". "Uber ein Neues Grundgesetz der Mechanik".. Jacobi. "Collected Works". Princeton. "Probleme General de la Stabitite du Mouvement". Gauss. published in New York. Gibbs. Roy. E. Journal de Crelle.. W. 1947. C.. English translation published by Dover. 1860.

287 coordinates. 57 catastastic. 65 problem of the second kind. 56. 8. 66. 102 cylindrical.Index acceleration of a point. 283 continuous group of transformations. 272 central principle. 67 degrees of freedom. 86 characteristic equation. 66. 4 acceleration. 212 embedding constraints. 10 basic problems in kinetics. 19. 301 differential. 154 classification of forces. 77 accessibility of configuration space. 57 337 nonholonomic. 78. 3. 80 configuration space. 48. 264. 51 constraint forces. 281 continuous point transformations. acatastatic. 95 actual motion. 131 classification of constraints. 156 distribution. 104 disspative forces. 57 configuration. 3. 9 angular velocity. 240 center of mass. basic. 102 rectangular. 51. 132 spherical. 8. 120. 185 assumptions. 137 rheonomic. 9. 183 conservative force. 122 dynamics. 76 Dirichlet's stability theorem. 53 descriptive function. possible. actual. 14. 133 generalized. 3 Bertrand's theorem. 221 displacements. 49 conic sections. 124 . 59 action. 17 central force motion. 212 choice of coordinates. 3 problem of the first kind. 135 D'Alembert's principle. 113. 119 constrained particle. 263 Carnot's theorem. 56 equality. 90 canonical equations. 56 holonomic. 104 virtual. 91 dynamic coupling. 48 eccentric annomaly. 51. 18 apoapsis. 225 DuBois . 66 infinitesimal. 68 contact transformations.Reymond lemma. 66. 186 eccentricity. 51 scleronomic. 47 basic kinematic equation. 235 calculus of variations. 136 constraints. 1 autonomous systems. 53 angular acceleration. 183 eigenvalues. 54 possible.

250 given forces. 308 Jacobi's integral. 1 infinitesmal contact transformations. 291 Jacobi's theorem of the last multiplier. 13 gravitation. 232. 87. 112. 90 Gauss's principle of least constraint. 67. 298 Hamilton-Jacobi equation. 78. 14 gravitational constant. 286 ignorable coordinates. 233 general potential functions. 237 Kepler's laws. 158 generating function. 49 extremal. 225 inertial reference frame. 83. 228.180 third.338 energy theorem. 112 generalized goordinates. first. 13. 217 homogeneous contact transformations. 235 Euclidean space. 123 inverse square forces. 14 gyrocompass. 263 Hamilton's principle. definite. 78 gradient. 199 Hamilton. 90 formulating problems. 3 kinetic energy. 90 Euler's theorem. 232 live. 283 Gibbs-Appell equations. 192. 229 inert. 109 . 219 fundamental equation. semidefinite. 183 second. 277 Kelvin's theorem. 131 functions. 102 generalized momentum. 228. 201 gyroscope. 121 generalized force. 309 Hamilton's equations. 330 integrals of motion. Analytical Dynamics 122. 232 impulsive force. 153 explicit form. 249. 184 kinematics. 87 second form.Lagrange equation. 66. 11 event space. 250 Gibbs' function. 304 Hamilton-Jacobi theorem. 88 Hamilton's principal function. 307 Hamiltonian function. 315 Hill equations. 297 first form. 117. negative definite. 182 Jacobi. 262 fundamental lemma. 150 invariance of Lagrange's equations. 305 heavy symmetrical top. 1 Euler angles. 220 positive definite. 158. 153 Jacobi's theorem. Liapunov. 249 Gibbs' theorem. 87 third form. 262 Hamilton's characteristic function. 282 impulsive constraints. 191 Euler .

113 339 potential energy function. 7 Rayleigh's dissipation function. 225 Newton's laws of motion. 263 Newtonian problem. 182 orbits. 222 Liouville systems. elliptical. 12 precession. 70. 50 state variable form. 5 nutation. 53 potential energy. 16 momentum integral. 117 last multiplier. 314 separation of variables. 246 radius of curvature. 17 Routh-Hurwitz criteria. 47 nonconservative forces. 151 stability. 13. 159 separability. 3 strictly Newtonian problem. 183 particle dynamics. 276 Liapunov's theorem. 120 rectangular components. 151 solution. 95 non-minimal coordinates. 206 statics. 162 Liouville's theorem. 212 Routhian function. 186 mechanical energy. 220. 192 precession of the equinoxes. 208 orbital. 70 Lagrange's equations. hyperbolic. 96 qualitative integration. 78 noncontemporaneous variations. 115. 53. 4 relative velocity. 67 rigid body dynamics. 21 analogue of. 9 relativity theory. 288 Lagrange multipliers. 268 Liapunov. 151 solution of dynamical system. in quadratures. 47 speed. 272 possible motion. 162 solution. 185 Pfaffian form. 210 Poincare. 1. 28 rigid body. 57. 4 spin. 209 state space.Index kinetics. 253 Lagrange. 2 Newton's second law. circular. 159 multiplier rule. 68 Poisson brackets. 3 Koenig's theorem. 94 Lagrange's. 95 Jacobi's. 116 natural systems. 192 . closed form. 199 principle of least action. 4 periapsis. 50 state-time space. 118 Lagrange brackets. 271 Poisson's theorem. 192 orbit equation. 288 mean anomaly. 78 power. 206 Hamiltonian systems. 121 normal-tangential components. 18. 206 state vector. 151 quasi-coordinates. 47. parabolic.

132 uncoupled systems. 102 one-to-one. 162 variation of a function. 76 work. 185 unconstrained particle. 76 variation of constants. 4 virtual change of state. 234 Taylor's theorem. possible. 319 variation of the elements. 238 time equation. 77 velocity of a point. 103 onto. 12 . 66. 325 velocities. 103 true anomaly. 74 virtual velocity. 186 transformation of coordinates.340 Analytical Dynamics superposition theorem. 74 virtual work.

Sign up to vote on this title
UsefulNot useful