Lecture Notes Classical Mechanics
Uploaded by
suudfiinLecture Notes Classical Mechanics
Uploaded by
suudfiinLecture Notes on Classical Mechanics
(A Work in Progress)
Daniel Arovas
Department of Physics
University of California, San Diego
May 8, 2013
Contents
0.1
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
0 Reference Materials
xiii
1
0.1
Lagrangian Mechanics (mostly) . . . . . . . . . . . . . . . . . . . . . . . . .
0.2
Hamiltonian Mechanics (mostly) . . . . . . . . . . . . . . . . . . . . . . . .
0.3
Mathematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1 Introduction to Dynamics
1.1
1.2
1.3
Introduction and Review . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1.1
Newtons laws of motion . . . . . . . . . . . . . . . . . . . . . . . .
1.1.2
Aside : inertial vs. gravitational mass . . . . . . . . . . . . . . . . .
Examples of Motion in One Dimension
. . . . . . . . . . . . . . . . . . . .
1.2.1
Uniform force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2.2
Uniform force with linear frictional damping . . . . . . . . . . . . .
1.2.3
Uniform force with quadratic frictional damping . . . . . . . . . . .
1.2.4
Crossed electric and magnetic fields . . . . . . . . . . . . . . . . . .
Pause for Reflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
10
2 Systems of Particles
11
2.1
Work-Energy Theorem
. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
11
2.2
Conservative and Nonconservative Forces . . . . . . . . . . . . . . . . . . .
12
Example : integrating F = U . . . . . . . . . . . . . . . . . . .
14
Conservative Forces in Many Particle Systems . . . . . . . . . . . . . . . .
15
2.2.1
2.3
ii
CONTENTS
2.4
Linear and Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . .
16
2.5
Scaling of Solutions for Homogeneous Potentials . . . . . . . . . . . . . . .
18
2.5.1
Eulers theorem for homogeneous functions . . . . . . . . . . . . . .
18
2.5.2
Scaled equations of motion . . . . . . . . . . . . . . . . . . . . . . .
18
2.6
2.7
Appendix I : Curvilinear Orthogonal Coordinates
. . . . . . . . . . . . . .
20
2.6.1
Example : spherical coordinates . . . . . . . . . . . . . . . . . . . .
21
2.6.2
Vector calculus : grad, div, curl . . . . . . . . . . . . . . . . . . . .
21
Common curvilinear orthogonal systems . . . . . . . . . . . . . . . . . . . .
23
2.7.1
Rectangular coordinates . . . . . . . . . . . . . . . . . . . . . . . .
23
2.7.2
Cylindrical coordinates . . . . . . . . . . . . . . . . . . . . . . . . .
23
2.7.3
Spherical coordinates . . . . . . . . . . . . . . . . . . . . . . . . . .
24
2.7.4
Kinetic energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
3 One-Dimensional Conservative Systems
3.1
3.2
3.3
3.4
Description as a Dynamical System . . . . . . . . . . . . . . . . . . . . . .
27
3.1.1
Example : harmonic oscillator . . . . . . . . . . . . . . . . . . . . .
28
One-Dimensional Mechanics as a Dynamical System . . . . . . . . . . . . .
29
3.2.1
Sketching phase curves . . . . . . . . . . . . . . . . . . . . . . . . .
29
Fixed Points and their Vicinity . . . . . . . . . . . . . . . . . . . . . . . . .
31
3.3.1
Linearized dynamics in the vicinity of a fixed point . . . . . . . . .
31
Examples of Conservative One-Dimensional Systems . . . . . . . . . . . . .
33
3.4.1
Harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . .
33
3.4.2
Pendulum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
34
3.4.3
Other potentials . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
36
4 Linear Oscillations
4.1
27
41
Damped Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . .
41
4.1.1
Classes of damped harmonic motion . . . . . . . . . . . . . . . . . .
42
4.1.2
Remarks on the case of critical damping . . . . . . . . . . . . . . .
44
CONTENTS
4.1.3
iii
Phase portraits for the damped harmonic oscillator . . . . . . . . .
45
Damped Harmonic Oscillator with Forcing . . . . . . . . . . . . . . . . . .
46
4.2.1
Resonant forcing . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
49
4.2.2
R-L-C circuits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
50
4.2.3
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
51
4.3
General solution by Greens function method . . . . . . . . . . . . . . . . .
54
4.4
General Linear Autonomous Inhomogeneous ODEs . . . . . . . . . . . . . .
55
4.5
Kramers-Kr
onig Relations (advanced material) . . . . . . . . . . . . . . . .
59
4.2
5 Calculus of Variations
61
5.1
Snells Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
61
5.2
Functions and Functionals . . . . . . . . . . . . . . . . . . . . . . . . . . . .
62
5.2.1
Functional Taylor series . . . . . . . . . . . . . . . . . . . . . . . . .
66
Examples from the Calculus of Variations . . . . . . . . . . . . . . . . . . .
66
5.3.1
Example 1 : minimal surface of revolution . . . . . . . . . . . . . .
66
5.3.2
Example 2 : geodesic on a surface of revolution . . . . . . . . . . .
68
5.3.3
Example 3 : brachistochrone . . . . . . . . . . . . . . . . . . . . . .
69
5.3.4
Ocean waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
70
Appendix : More on Functionals . . . . . . . . . . . . . . . . . . . . . . . .
72
5.3
5.4
6 Lagrangian Mechanics
79
6.1
Generalized Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
79
6.2
Hamiltons Principle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
80
6.2.1
Invariance of the equations of motion . . . . . . . . . . . . . . . . .
80
6.2.2
Remarks on the order of the equations of motion
. . . . . . . . . .
80
6.2.3
Lagrangian for a free particle . . . . . . . . . . . . . . . . . . . . . .
81
Conserved Quantities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
82
6.3.1
Momentum conservation . . . . . . . . . . . . . . . . . . . . . . . .
82
6.3.2
Energy conservation . . . . . . . . . . . . . . . . . . . . . . . . . . .
83
6.3
iv
CONTENTS
6.4
Choosing Generalized Coordinates . . . . . . . . . . . . . . . . . . . . . . .
84
6.5
How to Solve Mechanics Problems . . . . . . . . . . . . . . . . . . . . . . .
85
6.6
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
85
6.6.1
One-dimensional motion . . . . . . . . . . . . . . . . . . . . . . . .
85
6.6.2
Central force in two dimensions . . . . . . . . . . . . . . . . . . . .
86
6.6.3
A sliding point mass on a sliding wedge . . . . . . . . . . . . . . . .
86
6.6.4
A pendulum attached to a mass on a spring . . . . . . . . . . . . .
88
6.6.5
The double pendulum . . . . . . . . . . . . . . . . . . . . . . . . . .
90
6.6.6
The thingy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
92
Appendix : Virial Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . .
94
6.7
7 Noethers Theorem
7.1
97
Continuous Symmetry Implies Conserved Charges . . . . . . . . . . . . . .
97
7.1.1
Examples of one-parameter families of transformations . . . . . . .
98
7.2
Conservation of Linear and Angular Momentum . . . . . . . . . . . . . . .
99
7.3
Advanced Discussion : Invariance of L vs. Invariance of S . . . . . . . . . .
100
7.3.1
The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . .
101
7.3.2
Is H = T + U ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
103
7.3.3
Example: A bead on a rotating hoop . . . . . . . . . . . . . . . . .
104
7.4
Charged Particle in a Magnetic Field . . . . . . . . . . . . . . . . . . . . .
106
7.5
Fast Perturbations : Rapidly Oscillating Fields . . . . . . . . . . . . . . . .
108
7.5.1
Example : pendulum with oscillating support . . . . . . . . . . . .
110
Field Theory: Systems with Several Independent Variables . . . . . . . . .
111
7.6.1
114
7.6
Gross-Pitaevskii model . . . . . . . . . . . . . . . . . . . . . . . . .
8 Constraints
117
8.1
Constraints and Variational Calculus . . . . . . . . . . . . . . . . . . . . . .
117
8.2
Constrained Extremization of Functions . . . . . . . . . . . . . . . . . . . .
119
8.3
Extremization of Functionals : Integral Constraints . . . . . . . . . . . . .
119
CONTENTS
8.4
8.5
8.6
Extremization of Functionals : Holonomic Constraints . . . . . . . . . . . .
120
8.4.1
Examples of extremization with constraints . . . . . . . . . . . . . .
121
Application to Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . .
123
8.5.1
Constraints and conservation laws . . . . . . . . . . . . . . . . . . .
124
Worked Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
125
8.6.1
One cylinder rolling off another . . . . . . . . . . . . . . . . . . . .
125
8.6.2
Frictionless motion along a curve . . . . . . . . . . . . . . . . . . .
127
8.6.3
Disk rolling down an inclined plane . . . . . . . . . . . . . . . . . .
130
8.6.4
Pendulum with nonrigid support . . . . . . . . . . . . . . . . . . . .
131
8.6.5
Falling ladder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
133
8.6.6
Point mass inside rolling hoop . . . . . . . . . . . . . . . . . . . . .
137
9 Central Forces and Orbital Mechanics
9.1
143
Reduction to a one-body problem . . . . . . . . . . . . . . . . . . . . . . .
143
9.1.1
Center-of-mass (CM) and relative coordinates . . . . . . . . . . . .
143
9.1.2
Solution to the CM problem . . . . . . . . . . . . . . . . . . . . . .
144
9.1.3
Solution to the relative coordinate problem . . . . . . . . . . . . . .
144
9.2
Almost Circular Orbits . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
147
9.3
Precession in a Soluble Model . . . . . . . . . . . . . . . . . . . . . . . . . .
148
9.4
The Kepler Problem: U (r) = k r 1 . . . . . . . . . . . . . . . . . . . . . .
149
9.4.1
Geometric shape of orbits . . . . . . . . . . . . . . . . . . . . . . .
149
9.4.2
Laplace-Runge-Lenz vector . . . . . . . . . . . . . . . . . . . . . . .
150
9.4.3
Kepler orbits are conic sections . . . . . . . . . . . . . . . . . . . .
152
9.4.4
Period of bound Kepler orbits . . . . . . . . . . . . . . . . . . . . .
154
9.4.5
Escape velocity . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
155
9.4.6
Satellites and spacecraft . . . . . . . . . . . . . . . . . . . . . . . .
156
9.4.7
Two examples of orbital mechanics . . . . . . . . . . . . . . . . . .
156
Appendix I : Mission to Neptune . . . . . . . . . . . . . . . . . . . . . . . .
159
9.5.1
162
9.5
I. Earth to Jupiter
. . . . . . . . . . . . . . . . . . . . . . . . . . .
vi
CONTENTS
9.6
9.5.2
II. Encounter with Jupiter . . . . . . . . . . . . . . . . . . . . . . .
163
9.5.3
III. Jupiter to Neptune . . . . . . . . . . . . . . . . . . . . . . . . .
165
Appendix II : Restricted Three-Body Problem . . . . . . . . . . . . . . . .
166
10 Small Oscillations
173
10.1 Coupled Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
173
10.2 Expansion about Static Equilibrium . . . . . . . . . . . . . . . . . . . . . .
174
10.3 Method of Small Oscillations . . . . . . . . . . . . . . . . . . . . . . . . . .
174
10.3.1
Can you really just choose an A so that both these wonderful things
happen in 10.13 and 10.14? . . . . . . . . . . . . . . . . . . . . . . .
175
10.3.2
Er...care to elaborate?
. . . . . . . . . . . . . . . . . . . . . . . . .
175
10.3.3
Finding the modal matrix . . . . . . . . . . . . . . . . . . . . . . .
176
10.4 Example: Masses and Springs . . . . . . . . . . . . . . . . . . . . . . . . . .
178
10.5 Example: Double Pendulum . . . . . . . . . . . . . . . . . . . . . . . . . .
180
10.6 Zero Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
181
10.6.1
Example of zero mode oscillations . . . . . . . . . . . . . . . . . . .
182
10.7 Chain of Mass Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
184
10.7.1
Continuum limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
187
10.8 Appendix I : General Formulation . . . . . . . . . . . . . . . . . . . . . . .
188
10.9 Appendix II : Additional Examples
. . . . . . . . . . . . . . . . . . . . . .
190
10.9.1
Right Triatomic Molecule . . . . . . . . . . . . . . . . . . . . . . . .
190
10.9.2
Triple Pendulum . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
192
10.9.3
Equilateral Linear Triatomic Molecule
. . . . . . . . . . . . . . . .
195
10.10 Aside : Christoffel Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . .
199
11 Elastic Collisions
201
11.1 Center of Mass Frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
201
11.2 Central Force Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
205
11.2.1
Hard sphere scattering . . . . . . . . . . . . . . . . . . . . . . . . .
207
11.2.2
Rutherford scattering . . . . . . . . . . . . . . . . . . . . . . . . . .
208
CONTENTS
11.2.3
vii
Transformation to laboratory coordinates . . . . . . . . . . . . . . .
12 Noninertial Reference Frames
208
211
12.1 Accelerated Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . .
211
12.1.1
Translations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
213
12.1.2
Motion on the surface of the earth . . . . . . . . . . . . . . . . . . .
213
12.2 Spherical Polar Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . .
214
12.3 Centrifugal Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
216
12.3.1
Rotating tube of fluid . . . . . . . . . . . . . . . . . . . . . . . . . .
216
12.4 The Coriolis Force . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
217
12.4.1
Foucaults pendulum . . . . . . . . . . . . . . . . . . . . . . . . . .
13 Rigid Body Motion and Rotational Dynamics
13.1 Rigid Bodies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
13.1.1
220
223
223
Examples of rigid bodies . . . . . . . . . . . . . . . . . . . . . . . .
223
13.2 The Inertia Tensor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
224
13.2.1
Coordinate transformations
. . . . . . . . . . . . . . . . . . . . . .
226
13.2.2
The case of no fixed point . . . . . . . . . . . . . . . . . . . . . . .
226
13.3 Parallel Axis Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
227
13.3.1
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
228
13.3.2
General planar mass distribution . . . . . . . . . . . . . . . . . . .
229
13.4 Principal Axes of Inertia
. . . . . . . . . . . . . . . . . . . . . . . . . . . .
230
13.5 Eulers Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
231
13.5.1
Example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
234
13.6 Eulers Angles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
235
13.6.1
Torque-free symmetric top . . . . . . . . . . . . . . . . . . . . . . .
237
13.6.2
Symmetric top with one point fixed . . . . . . . . . . . . . . . . . .
238
13.7 Rolling and Skidding Motion of Real Tops . . . . . . . . . . . . . . . . . . .
241
13.7.1
Rolling tops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
241
viii
CONTENTS
13.7.2
Skidding tops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
242
13.7.3
Tippie-top . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
244
14 Continuum Mechanics
247
14.1 Strings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
247
14.2 dAlemberts Solution to the Wave Equation . . . . . . . . . . . . . . . . .
249
14.2.1
Energy density and energy current
. . . . . . . . . . . . . . . . . .
250
14.2.2
Reflection at an interface . . . . . . . . . . . . . . . . . . . . . . . .
251
14.2.3
Mass point on a string . . . . . . . . . . . . . . . . . . . . . . . . .
252
14.2.4
Interface between strings of different mass density . . . . . . . . . .
255
14.3 Finite Strings : Bernoullis Solution . . . . . . . . . . . . . . . . . . . . . .
257
14.4 Sturm-Liouville Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
259
14.4.1
Variational method . . . . . . . . . . . . . . . . . . . . . . . . . . .
261
14.5 Continua in Higher Dimensions . . . . . . . . . . . . . . . . . . . . . . . . .
264
14.5.1
Membranes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
265
14.5.2
Helmholtz equation . . . . . . . . . . . . . . . . . . . . . . . . . . .
266
14.5.3
Rectangles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
266
14.5.4
Circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
267
14.5.5
Sound in fluids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
268
14.6 Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
270
14.7 Appendix I : Three Strings . . . . . . . . . . . . . . . . . . . . . . . . . . .
272
14.8 Appendix II : General Field Theoretic Formulation . . . . . . . . . . . . . .
275
14.8.1
Euler-Lagrange equations for classical field theories . . . . . . . . .
275
14.8.2
Conserved currents in field theory . . . . . . . . . . . . . . . . . . .
276
14.8.3
Gross-Pitaevskii model . . . . . . . . . . . . . . . . . . . . . . . . .
277
14.9 Appendix III : Greens Functions . . . . . . . . . . . . . . . . . . . . . . . .
279
14.9.1
Perturbation theory . . . . . . . . . . . . . . . . . . . . . . . . . . .
281
14.9.2
Perturbation theory for eigenvalues and eigenfunctions . . . . . . .
284
CONTENTS
ix
15 Special Relativity
287
15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
287
15.1.1
Michelson-Morley experiment . . . . . . . . . . . . . . . . . . . . .
287
15.1.2
Einsteinian and Galilean relativity . . . . . . . . . . . . . . . . . . .
290
15.2 Intervals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
292
15.2.1
Proper time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
294
15.2.2
Irreverent problem from Spring 2002 final exam . . . . . . . . . . .
294
15.3 Four-Vectors and Lorentz Transformations . . . . . . . . . . . . . . . . . .
296
15.3.1
Covariance and contravariance . . . . . . . . . . . . . . . . . . . . .
299
15.3.2
What to do if you hate raised and lowered indices . . . . . . . . . .
301
15.3.3
Comparing frames . . . . . . . . . . . . . . . . . . . . . . . . . . . .
301
15.3.4
Example I . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
302
15.3.5
Example II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
302
15.3.6
Deformation of a rectangular plate . . . . . . . . . . . . . . . . . .
303
15.3.7
Transformation of velocities . . . . . . . . . . . . . . . . . . . . . .
305
15.3.8
Four-velocity and four-acceleration . . . . . . . . . . . . . . . . . .
306
15.4 Three Kinds of Relativistic Rockets . . . . . . . . . . . . . . . . . . . . . .
306
15.4.1
Constant acceleration model . . . . . . . . . . . . . . . . . . . . . .
306
15.4.2
Constant force with decreasing mass . . . . . . . . . . . . . . . . .
307
15.4.3
Constant ejecta velocity . . . . . . . . . . . . . . . . . . . . . . . .
308
15.5 Relativistic Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
309
15.5.1
Relativistic harmonic oscillator
. . . . . . . . . . . . . . . . . . . .
311
15.5.2
Energy-momentum 4-vector . . . . . . . . . . . . . . . . . . . . . .
312
15.5.3
4-momentum for massless particles . . . . . . . . . . . . . . . . . .
313
15.6 Relativistic Doppler Effect . . . . . . . . . . . . . . . . . . . . . . . . . . .
313
15.6.1
Romantic example . . . . . . . . . . . . . . . . . . . . . . . . . . . .
314
15.7 Relativistic Kinematics of Particle Collisions . . . . . . . . . . . . . . . . .
316
15.7.1
Spontaneous particle decay into two products . . . . . . . . . . . .
317
CONTENTS
15.7.2
Miscellaneous examples of particle decays . . . . . . . . . . . . . . .
318
15.7.3
Threshold particle production with a stationary target . . . . . . .
319
15.7.4
Transformation between frames . . . . . . . . . . . . . . . . . . . .
320
15.7.5
Compton scattering . . . . . . . . . . . . . . . . . . . . . . . . . . .
321
15.8 Covariant Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . .
322
15.8.1
Lorentz force law . . . . . . . . . . . . . . . . . . . . . . . . . . . .
324
15.8.2
Gauge invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . .
326
15.8.3
Transformations of fields . . . . . . . . . . . . . . . . . . . . . . . .
327
15.8.4
Invariance versus covariance . . . . . . . . . . . . . . . . . . . . . .
328
15.9 Appendix I : The Pole, the Barn, and Rashoman . . . . . . . . . . . . . . .
330
15.10 Appendix II : Photographing a Moving Pole
332
. . . . . . . . . . . . . . . . .
16 Hamiltonian Mechanics
335
16.1 The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
335
16.2 Modified Hamiltons Principle . . . . . . . . . . . . . . . . . . . . . . . . .
337
16.3 Phase Flow is Incompressible . . . . . . . . . . . . . . . . . . . . . . . . . .
337
16.4 Poincare Recurrence Theorem . . . . . . . . . . . . . . . . . . . . . . . . .
338
16.5 Poisson Brackets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
339
16.6 Canonical Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . .
340
16.6.1
Point transformations in Lagrangian mechanics . . . . . . . . . . .
340
16.6.2
Canonical transformations in Hamiltonian mechanics . . . . . . . .
342
16.6.3
Hamiltonian evolution
. . . . . . . . . . . . . . . . . . . . . . . . .
342
16.6.4
Symplectic structure . . . . . . . . . . . . . . . . . . . . . . . . . .
343
16.6.5
Generating functions for canonical transformations . . . . . . . . .
344
16.7 Hamilton-Jacobi Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
347
16.7.1
The action as a function of coordinates and time . . . . . . . . . . .
347
16.7.2
The Hamilton-Jacobi equation . . . . . . . . . . . . . . . . . . . . .
349
16.7.3
Time-independent Hamiltonians . . . . . . . . . . . . . . . . . . . .
350
16.7.4
Example: one-dimensional motion . . . . . . . . . . . . . . . . . . .
351
CONTENTS
xi
16.7.5
Separation of variables . . . . . . . . . . . . . . . . . . . . . . . . .
351
16.7.6
Example #2 : point charge plus electric field . . . . . . . . . . . . .
353
16.7.7
Example #3 : Charged Particle in a Magnetic Field . . . . . . . . .
355
16.8 Action-Angle Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
357
16.8.1
Circular Phase Orbits: Librations and Rotations . . . . . . . . . . .
357
16.8.2
Action-Angle Variables . . . . . . . . . . . . . . . . . . . . . . . . .
358
16.8.3
Canonical Transformation to Action-Angle Variables . . . . . . . .
359
16.8.4
Example : Harmonic Oscillator . . . . . . . . . . . . . . . . . . . .
360
16.8.5
Example : Particle in a Box . . . . . . . . . . . . . . . . . . . . . .
361
16.8.6
Kepler Problem in Action-Angle Variables . . . . . . . . . . . . . .
364
16.8.7
Charged Particle in a Magnetic Field . . . . . . . . . . . . . . . . .
365
16.8.8
Motion on Invariant Tori . . . . . . . . . . . . . . . . . . . . . . . .
366
16.9 Canonical Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . .
367
16.9.1
Canonical Transformations and Perturbation Theory . . . . . . . .
367
16.9.2
Canonical Perturbation Theory for n = 1 Systems . . . . . . . . . .
369
16.9.3
Example : Nonlinear Oscillator . . . . . . . . . . . . . . . . . . . .
372
16.9.4
n > 1 Systems : Degeneracies and Resonances . . . . . . . . . . . .
373
16.9.5
Particle-Wave Interaction . . . . . . . . . . . . . . . . . . . . . . . .
375
16.10 Adiabatic Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
378
16.10.1 Example: mechanical mirror . . . . . . . . . . . . . . . . . . . . . .
379
16.10.2 Example: magnetic mirror . . . . . . . . . . . . . . . . . . . . . . .
380
16.10.3 Resonances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
382
16.11 Appendix : Canonical Perturbation Theory . . . . . . . . . . . . . . . . . .
382
17 Physics 110A-B Exams
385
17.1 F05 Physics 110A Midterm #1 . . . . . . . . . . . . . . . . . . . . . . . . .
386
17.2 F05 Physics 110A Midterm #2 . . . . . . . . . . . . . . . . . . . . . . . . .
390
17.3 F05 Physics 110A Final Exam . . . . . . . . . . . . . . . . . . . . . . . . .
397
17.4 F07 Physics 110A Midterm #1 . . . . . . . . . . . . . . . . . . . . . . . . .
405
xii
CONTENTS
17.5 F07 Physics 110A Midterm #2 . . . . . . . . . . . . . . . . . . . . . . . . .
411
17.6 F07 Physics 110A Final Exam . . . . . . . . . . . . . . . . . . . . . . . . .
415
17.7 W08 Physics 110B Midterm Exam . . . . . . . . . . . . . . . . . . . . . . .
425
17.8 W08 Physics 110B Final Exam . . . . . . . . . . . . . . . . . . . . . . . . .
430
0.1. PREFACE
0.1
xiii
Preface
These lecture notes are based on material presented in both graduate and undergraduate
mechanics classes which I have taught on several occasions during the past 20 years at
UCSD (Physics 110A-B and Physics 200A-B).
The level of these notes is appropriate for an advanced undergraduate or a first year graduate
course in classical mechanics. In some instances, Ive tried to collect the discussion of more
advanced material into separate sections, but in many cases this proves inconvenient, and
so the level of the presentation fluctuates.
I have included many worked examples within the notes, as well as in the final chapter,
which contains solutions from Physics 110A and 110B midterm and final exams. In my view,
problem solving is essential toward learning basic physics. The geniuses among us might
apprehend the fundamentals through deep contemplation after reading texts and attending
lectures. The vast majority of us, however, acquire physical intuition much more slowly,
and it is through problem solving that one gains experience in patches which eventually
percolate so as to afford a more global understanding of the subject. A good analogy would
be putting together a jigsaw puzzle: initially only local regions seem to make sense but
eventually one forms the necessary connections so that one recognizes the entire picture.
My presentation and choice of topics has been influenced by many books as well as by my
own professors. Ive reiterated extended some discussions from other texts, such as Barger
and Olssons treatment of the gravitational swing-by effect, and their discussion of rolling
and skidding tops. The figures were, with very few exceptions, painstakingly made using
Keynote and/or SM.
Originally these notes also included material on dynamical systems and on Hamiltonian
mechanics. These sections have now been removed and placed within a separate set of
notes on nonlinear dynamics (Physics 221A).
My only request, to those who would use these notes: please contact me if you find errors or typos, or if you have suggestions for additional material. My email address is
darovas@ucsd.edu. I plan to update and extend these notes as my time and inclination
permit.
xiv
CONTENTS
Chapter 0
Reference Materials
Here I list several resources, arranged by topic. My personal favorites are marked with a
diamond ().
0.1
Lagrangian Mechanics (mostly)
L. D. Landau and E. M. Lifshitz, Mechanics, 3rd ed. (Butterworth-Heinemann, 1976)
A. L. Fetter and J. D. Walecka, Nonlinear Mechanics (Dover, 2006)
O. D. Johns, Analytical Mechanics for Relativity and Quantum Mechanics (Oxford,
2005)
D. T. Greenwood, Classical Mechanics (Dover, 1997)
H. Goldstein, C. P. Poole, and J. L. Safko, Classical Mechanics, 3rd ed. (AddisonWesley, 2001)
V. Barger and M. Olsson, Classical Mechanics : A Modern Perspective (McGraw-Hill,
1994)
0.2
Hamiltonian Mechanics (mostly)
J. V. Jose and E. J. Saletan, Mathematical Methods of Classical Mechanics (Springer,
1997)
1
CHAPTER 0. REFERENCE MATERIALS
W. Dittrich and M. Reuter, Classical and Quantum Dynamics (Springer, 2001)
V. I. Arnold Introduction to Dynamics (Cambridge, 1982)
V. I. Arnold, V. V. Kozlov, and A. I. Neishtadt, Mathematical Aspects of Classical
and Celestial Mechanics (Springer, 2006)
I. Percival and D. Richards, Introduction to Dynamics (Cambridge, 1982)
0.3
Mathematics
I. M. Gelfand and S. V. Fomin, Calculus of Variations (Dover, 1991)
V. I. Arnold, Ordinary Differential Equations (MIT Press, 1973)
V. I. Arnold, Geometrical Methods in the Theory of Ordinary Differential Equations
(Springer, 1988)
R. Weinstock, Calculus of Variations (Dover, 1974)
Chapter 1
Introduction to Dynamics
1.1
Introduction and Review
Dynamics is the science of how things move. A complete solution to the motion of a system
means that we know the coordinates of all its constituent particles as functions of time.
For a single point particle moving in three-dimensional space, this means we want to know
its position vector r(t) as a function of time. If there are many particles, the motion is
described by a set of functions ri (t), where i labels which particle we are talking about. So
generally speaking, solving for the motion means being able to predict where a particle will
be at any given instant of time. Of course, knowing the function ri (t) means we can take
its derivative and obtain the velocity vi (t) = dri /dt at any time as well.
The complete motion for a system is not given to us outright, but rather is encoded in a
set of differential equations, called the equations of motion. An example of an equation of
motion is
m
d2x
= mg
dt2
(1.1)
with the solution
x(t) = x0 + v0 t 12 gt2
(1.2)
where x0 and v0 are constants corresponding to the initial boundary conditions on the
position and velocity: x(0) = x0 , v(0) = v0 . This particular solution describes the vertical
motion of a particle of mass m moving near the earths surface.
In this class, we shall discuss a general framework by which the equations of motion may
be obtained, and methods for solving them. That general framework is Lagrangian Dynamics, which itself is really nothing more than an elegant restatement of Isaac Newtons
Laws of Motion.
3
1.1.1
CHAPTER 1. INTRODUCTION TO DYNAMICS
Newtons laws of motion
Aristotle held that objects move because they are somehow impelled to seek out their
natural state. Thus, a rock falls because rocks belong on the earth, and flames rise because
fire belongs in the heavens. To paraphrase Wolfgang Pauli, such notions are so vague as to
be not even wrong. It was only with the publication of Newtons Principia in 1687 that
a theory of motion which had detailed predictive power was developed.
Newtons three Laws of Motion may be stated as follows:
I. A body remains in uniform motion unless acted on by a force.
II. Force equals rate of change of momentum: F = dp/dt.
III. Any two bodies exert equal and opposite forces on each other.
Newtons First Law states that a particle will move in a straight line at constant (possibly
zero) velocity if it is subjected to no forces. Now this cannot be true in general, for suppose
we encounter such a free particle and that indeed it is in uniform motion, so that r(t) =
r0 + v0 t. Now r(t) is measured in some coordinate system, and if instead we choose
to measure r(t) in a different coordinate system whose origin R moves according to the
function R(t), then in this new frame of reference the position of our particle will be
r (t) = r(t) R(t)
= r0 + v0 t R(t) .
(1.3)
If the acceleration d2R/dt2 is nonzero, then merely by shifting our frame of reference we have
apparently falsified Newtons First Law a free particle does not move in uniform rectilinear
motion when viewed from an accelerating frame of reference. Thus, together with Newtons
Laws comes an assumption about the existence of frames of reference called inertial frames
in which Newtons Laws hold. A transformation from one frame K to another frame K
which moves at constant velocity V relative to K is called a Galilean transformation. The
equations of motion of classical mechanics are invariant (do not change) under Galilean
transformations.
At first, the issue of inertial and noninertial frames is confusing. Rather than grapple with
this, we will try to build some intuition by solving mechanics problems assuming we are
in an inertial frame. The earths surface, where most physics experiments are done, is not
an inertial frame, due to the centripetal accelerations associated with the earths rotation
about its own axis and its orbit around the sun. In this case, not only is our coordinate
systems origin somewhere in a laboratory on the surface of the earth accelerating, but
the coordinate axes themselves are rotating with respect to an inertial frame. The rotation
of the earth leads to fictitious forces such as the Coriolis force, which have large-scale
consequences. For example, hurricanes, when viewed from above, rotate counterclockwise
in the northern hemisphere and clockwise in the southern hemisphere. Later on in the course
we will devote ourselves to a detailed study of motion in accelerated coordinate systems.
Newtons quantity of motion is the momentum p, defined as the product p = mv of a
particles mass m (how much stuff there is) and its velocity (how fast it is moving). In
1.1. INTRODUCTION AND REVIEW
order to convert the Second Law into a meaningful equation, we must know how the force
F depends on the coordinates (or possibly velocities) themselves. This is known as a force
law. Examples of force laws include:
Constant force:
Hookes Law:
Gravitation:
Lorentz force:
Fluid friction (v small):
F = mg
F = kx
F = GM m r/r 2
F = qE +q
v
B
c
F = b v .
Note that for an object whose mass does not change we can write the Second Law in the
familiar form F = ma, where a = dv/dt = d2r/dt2 is the acceleration. Most of our initial
efforts will lie in using Newtons Second Law to solve for the motion of a variety of systems.
The Third Law is valid for the extremely important case of central forces which we will
discuss in great detail later on. Newtonian gravity the force which makes the planets orbit
the sun is a central force. One consequence of the Third Law is that in free space two
isolated particles will accelerate in such a way that F1 = F2 and hence the accelerations
are parallel to each other, with
m2
a1
=
,
(1.4)
a2
m1
where the minus sign is used here to emphasize that the accelerations are in opposite
directions. We can also conclude that the total momentum P = p1 + p2 is a constant, a
result known as the conservation of momentum.
1.1.2
Aside : inertial vs. gravitational mass
In addition to postulating the Laws of Motion, Newton also deduced the gravitational force
law, which says that the force Fij exerted by a particle i by another particle j is
Fij = Gmi mj
ri rj
,
|ri rj |3
(1.5)
where G, the Cavendish constant (first measured by Henry Cavendish in 1798), takes the
value
G = (6.6726 0.0008) 1011 N m2 /kg2 .
(1.6)
Notice Newtons Third Law in action: Fij + Fji = 0. Now a very important and special
feature of this inverse square law force is that a spherically symmetric mass distribution
has the same force on an external body as it would if all its mass were concentrated at its
CHAPTER 1. INTRODUCTION TO DYNAMICS
center. Thus, for a particle of mass m near the surface of the earth, we can take mi = m
and mj = Me , with ri rj Re r and obtain
F = mgr mg
(1.7)
where r is a radial unit vector pointing from the earths center and g = GMe /Re2 9.8 m/s2
is the acceleration due to gravity at the earths surface. Newtons Second Law now says
that a = g, i.e. objects accelerate as they fall to earth. However, it is not a priori clear
why the inertial mass which enters into the definition of momentum should be the same
as the gravitational mass which enters into the force law. Suppose, for instance, that the
gravitational mass took a different value, m . In this case, Newtons Second Law would
predict
a=
m
g
m
(1.8)
and unless the ratio m /m were the same number for all objects, then bodies would fall
with different accelerations. The experimental fact that bodies in a vacuum fall to earth at
the same rate demonstrates the equivalence of inertial and gravitational mass, i.e. m = m.
1.2
Examples of Motion in One Dimension
To gain some experience with solving equations of motion in a physical setting, we consider
some physically relevant examples of one-dimensional motion.
1.2.1
Uniform force
With F = mg, appropriate for a particle falling under the influence of a uniform gravitational field, we have m d2x/dt2 = mg, or x
= g. Notation:
x
dx
,
dt
7
= d x ,
x
dt7
d2x
,
dt2
etc.
(1.9)
With v = x,
we solve dv/dt = g:
Zt
Zv(t)
dv = ds (g)
v(0)
(1.10)
v(t) v(0) = gt .
Note that there is a constant of integration, v(0), which enters our solution.
(1.11)
1.2. EXAMPLES OF MOTION IN ONE DIMENSION
We are now in position to solve dx/dt = v:
Zt
Zx(t)
dx = ds v(s)
(1.12)
x(0)
Zt
x(t) = x(0) + ds v(0) gs
(1.13)
= x(0) + v(0)t 12 gt2 .
(1.14)
Note that a second constant of integration, x(0), has appeared.
1.2.2
Uniform force with linear frictional damping
In this case,
m
dv
= mg v
dt
(1.15)
which may be rewritten
dv
= dt
v + mg/
m
d ln(v + mg/) = (/m)dt .
Integrating then gives
v(t) + mg/
= t/m
ln
v(0) + mg/
mg
mg
v(t) =
+ v(0) +
et/m .
(1.16)
(1.17)
(1.18)
(1.19)
Note that the solution to the first order ODE mv = mg v entails one constant of
integration, v(0).
One can further integrate to obtain the motion
mg
mg
m
t.
v(0) +
(1 et/m )
x(t) = x(0) +
(1.20)
The solution to the second order ODE m
x = mg x thus entails two constants of
integration: v(0) and x(0). Notice that as t goes to infinity the velocity tends towards
the asymptotic value v = v , where v = mg/. This is known as the terminal velocity. Indeed, solving the equation v = 0 gives v = v . The initial velocity is effectively
forgotten on a time scale m/.
Electrons moving in solids under the influence of an electric field also achieve a terminal
velocity. In this case the force is not F = mg but rather F = eE, where e is the
CHAPTER 1. INTRODUCTION TO DYNAMICS
electron charge (e > 0) and E is the electric field. The terminal velocity is then obtained
from
v = eE/ = e E/m .
(1.21)
The current density is a product:
current density = (number density) (charge) (velocity)
j = n (e) (v )
=
ne2
E.
m
(1.22)
The ratio j/E is called the conductivity of the metal, . According to our theory, =
ne2 /m. This is one of the most famous equations of solid state physics! The dissipation
is caused by electrons scattering off impurities and lattice vibrations (phonons). In high
purity copper at low temperatures (T <
4 K), the scattering time is about a nanosecond
9
( 10 s).
1.2.3
Uniform force with quadratic frictional damping
At higher velocities, the frictional damping is proportional to the square of the velocity.
The frictional force is then Ff = cv 2 sgn (v), where sgn (v) is the sign of v: sgn (v) = +1
if v > 0 and sgn (v) = 1 if v < 0. (Note one can also write sgn (v) = v/|v| where |v| is
the absolute value.) Why all this trouble with sgn (v)? Because it is important that the
frictional force dissipate energy, and therefore that Ff be oppositely directed with respect to
the velocity v. We will assume that v < 0 always, hence Ff = +cv 2 .
2
Notice that there
p is a terminal velocity, since setting v = g + (c/m)v = 0 gives v = v ,
where v = mg/c. One can write the equation of motion as
and using
dv
g
2
= 2 (v 2 v
)
dt
v
(1.23)
1
1
1
1
=
2
v 2 v
2v v v v + v
(1.24)
we obtain
v2
dv
1
1
dv
dv
=
2
v
2v v v 2v v + v
v v
1
d ln
=
2v
v + v
g
= 2 dt .
v
(1.25)
1.2. EXAMPLES OF MOTION IN ONE DIMENSION
Assuming v(0) = 0, we integrate to obtain
v v(t)
gt
1
ln
= 2
2v
v + v(t)
v
(1.26)
which may be massaged to give the final result
v(t) = v tanh(gt/v ) .
(1.27)
Recall that the hyperbolic tangent function tanh(x) is given by
tanh(x) =
sinh(x)
ex ex
= x
.
cosh(x)
e + ex
(1.28)
Again, as t one has v(t) v , i.e. v() = v .
Advanced Digression: To gain an understanding of the constant c, consider a flat surface
of area S moving through a fluid at velocity v (v > 0). During a time t, all the fluid
molecules inside the volume V = S v t will have executed an elastic collision with the
moving surface. Since the surface is assumed to be much more massive than each fluid
molecule, the center of mass frame for the surface-molecule collision is essentially the frame
of the surface itself. If a molecule moves with velocity u is the laboratory frame, it moves
with velocity u v in the center of mass (CM) frame, and since the collision is elastic, its
final CM frame velocity is reversed, to v u. Thus, in the laboratory frame the molecules
velocity has become 2v u and it has suffered a change in velocity of u = 2(v u). The
total momentum change is obtained by multiplying u by the total mass M = V , where
is the mass density of the fluid. But then the total momentum imparted to the fluid is
P = 2(v u) S v t
(1.29)
and the force on the fluid is
F =
P
= 2S v(v u) .
t
(1.30)
Now it is appropriate to average this expression over the microscopic distribution of molecular velocities u, and since on average hui = 0, we obtain the result hF i = 2Sv 2 , where
h i denotes a microscopic average over the molecular velocities in the fluid. (There is a
subtlety here concerning the effect of fluid molecules striking the surface from either side
you should satisfy yourself that this derivation is sensible!) Newtons Third Law then states
that the frictional force imparted to the moving surface by the fluid is Ff = hF i = cv 2 ,
where c = 2S. In fact, our derivation is too crude to properly obtain the numerical prefactors, and it is better to write c = S, where is a dimensionless constant which depends
on the shape of the moving object.
1.2.4
Crossed electric and magnetic fields
Consider now a three-dimensional example of a particle of charge q moving in mutually
perpendicular E and B fields. Well throw in gravity for good measure. We take E = E x,
10
CHAPTER 1. INTRODUCTION TO DYNAMICS
and g = gz.
The equation of motion is Newtons 2nd Law again:
B = B z,
m r = mg + qE + qc r B .
(1.31)
The RHS (right hand side) of this equation is a vector sum of the forces due to gravity plus
the Lorentz force of a moving particle in an electromagnetic field. In component notation,
we have
qB
m
x = qE +
y
(1.32)
c
qB
x
(1.33)
m
y=
c
m
z = mg .
(1.34)
The equations for coordinates x and y are coupled, while that for z is independent and may
be immediately solved to yield
z(t) = z(0) + z(0)
t 12 gt2 .
(1.35)
The remaining equations may be written in terms of the velocities vx = x and vy = y:
v x = c (vy + uD )
(1.36)
v y = c vx ,
(1.37)
where c = qB/mc is the cyclotron frequency and uD = cE/B is the drift speed for the
particle. As we shall see, these are the equations for a harmonic oscillator. The solution is
vx (t) = vx (0) cos(c t) + vy (0) + uD sin(c t)
(1.38)
vy (t) = uD + vy (0) + uD cos(c t) vx (0) sin(c t) .
(1.39)
Integrating again, the full motion is given by:
x(t) = x(0) + A sin + A sin(c t )
(1.40)
y(r) = y(0) uD t A cos + A cos(c t ) ,
(1.41)
where
q
2
+ uD
1
1 y(0)
2
.
x (0) + y(0)
+ uD
, = tan
A=
c
x(0)
Thus, in the full solution of the motion there are six constants of integration:
x(0) , y(0) , z(0) , A , , z(0)
(1.42)
(1.43)
Of course instead of A and one may choose as constants of integration x(0)
and y(0).
1.3
Pause for Reflection
In mechanical systems, for each coordinate, or degree of freedom, there exists a corresponding second order ODE. The full solution of the motion of the system entails two
constants of integration for each degree of freedom.
Chapter 2
Systems of Particles
2.1
Work-Energy Theorem
Consider a system of many particles, with positions ri and velocities r i . The kinetic energy
of this system is
X
X
2
1
(2.1)
T =
Ti =
2 mi r i .
i
Now lets consider how the kinetic energy of the system changes in time. Assuming each
mi is time-independent, we have
dTi
= mi r i ri .
dt
Here, weve used the relation
(2.2)
d
dA
.
A2 = 2 A
dt
dt
(2.3)
We now invoke Newtons 2nd Law, mi ri = Fi , to write eqn. 2.2 as Ti = Fi r i . We integrate
this equation from time tA to tB :
(B)
Ti
(A)
Ti
ZtB
dTi
= dt
dt
tA
ZtB
X (AB)
= dt Fi r i
Wi
,
(2.4)
tA
where Wi(AB) is the total work done on particle
i during its motion from state A to state
P
B, Clearly P
the total kinetic energy is T = i Ti and the total work done on all particles is
W (AB) = i Wi(AB) . Eqn. 2.4 is known as the work-energy theorem. It says that
In the evolution of a mechanical system, the change in total kinetic energy is equal to the
total work done: T (B) T (A) = W (AB) .
11
12
CHAPTER 2. SYSTEMS OF PARTICLES
Figure 2.1: Two paths joining points A and B.
2.2
Conservative and Nonconservative Forces
For the sake of simplicity, consider a single particle with kinetic energy T = 12 mr 2 . The
work done on the particle during its mechanical evolution is
W
(AB)
ZtB
= dt F v ,
(2.5)
tA
This is the most general expression for the work done. If the force F depends
where v = r.
only on the particles position r, we may write dr = v dt, and then
W
(AB)
ZrB
= dr F (r) .
(2.6)
rA
Consider now the force
+ K2 x y ,
F (r) = K1 y x
(2.7)
where K1,2 are constants. Lets evaluate the work done along each of the two paths in fig.
2.1:
ZyB
ZxB
(2.8)
W (I) = K1 dx yA + K2 dy xB = K1 yA (xB xA ) + K2 xB (yB yA )
xA
ZxB
yA
ZyB
W (II) = K1 dx yB + K2 dy xA = K1 yB (xB xA ) + K2 xA (yB yA ) .
xA
yA
(2.9)
2.2. CONSERVATIVE AND NONCONSERVATIVE FORCES
13
Note that in general W (I) 6= W (II) . Thus, if we start at point A, the kinetic energy at point
B will depend on the path taken, since the work done is path-dependent.
The difference between the work done along the two paths is
W (I) W (II) = (K2 K1 ) (xB xA ) (yB yA ) .
(2.10)
Thus, we see that if K1 = K2 , the work is the same for the two paths. In fact, if K1 = K2 ,
the work would be path-independent, and would depend only on the endpoints. This is
true for any path, and not just piecewise linear paths of the type depicted in fig. 2.1. The
reason for this is Stokes theorem:
I
Z
F .
(2.11)
d F = dS n
C
Here, C is a connected region in three-dimensional space, C is mathematical notation for
is
the boundary of C, which is a closed path1 , dS is the scalar differential area element, n
the unit normal to that differential area element, and F is the curl of F :
y z
x
F = det x
y
z
Fx Fy Fz
Fx Fz
Fy
Fy
Fx
Fz
+
x
y +
z .
(2.12)
=
y
z
z
x
x
y
+ K2 x y,
the curl is
For the force under consideration, F (r) = K1 y x
F = (K2 K1 ) z ,
(2.13)
which is a constant. The RHS of eqn. 2.11 is then simply proportional to the area enclosed
H
by C. When we compute the work difference in eqn. 2.10, we evaluate the integral d F
C
along the path II1 I , which is to say path I followed by the inverse of path II. In this
= z and the integral of n
F over the rectangle C is given by the RHS of eqn.
case, n
2.10.
When F = 0 everywhere in space, we can always write F = U , where U (r) is the
potential energy. Such forces are called conservative forces because the total energy of the
system, E = T + U , is then conserved during its motion. We can see this by evaluating the
work done,
ZrB
(AB)
W
= dr F (r)
rA
ZrB
= dr U
rA
= U (rA ) U (rB ) .
(2.14)
1
If C is multiply connected, then C is a set of closed paths. For example, if C is an annulus, C is two
circles, corresponding to the inner and outer boundaries of the annulus.
14
CHAPTER 2. SYSTEMS OF PARTICLES
The work-energy theorem then gives
T (B) T (A) = U (rA ) U (rB ) ,
(2.15)
E (B) = T (B) + U (rB ) = T (A) + U (rA ) = E (A) .
(2.16)
which says
Thus, the total energy E = T + U is conserved.
2.2.1
Example : integrating F = U
If F = 0, we can compute U (r) by integrating, viz.
Zr
U (r) = U (0) dr F (r ) .
(2.17)
The integral does not depend on the path chosen connecting 0 and r. For example, we can
take
U (x, y, z) = U (0, 0, 0)
(x,0,0)
Z
dx Fx (x , 0, 0)
(x,y,0)
Z
dy Fy (x, y , 0)
dz Fz (x, y, z ) . (2.18)
(z,y,0)
(x,0,0)
(0,0,0)
(x,y,z)
Z
The constant U (0, 0, 0) is arbitrary and impossible to determine from F alone.
As an example, consider the force
kx y 4bz 3 z ,
F (r) = ky x
(2.19)
where k and b are constants. We have
Fy
Fz
F x =
=0
y
z
Fx Fz
=0
F y =
z
x
Fy
Fx
F z =
=0,
x
y
(2.20)
(2.21)
(2.22)
so F = 0 and F must be expressible as F = U . Integrating using eqn. 2.18, we
have
U (x, y, z) = U (0, 0, 0) +
(x,0,0)
Z
dx k 0 +
(0,0,0)
= U (0, 0, 0) + kxy + bz 4 .
(x,y,0)
Z
dy kxy +
(x,0,0)
(x,y,z)
Z
dz 4bz
(2.23)
(z,y,0)
(2.24)
2.3. CONSERVATIVE FORCES IN MANY PARTICLE SYSTEMS
15
Another approach is to integrate the partial differential equation U = F . This is in fact
three equations, and we shall need all of them to obtain the correct answer. We start with
the x-component,
U
= ky .
(2.25)
x
Integrating, we obtain
U (x, y, z) = kxy + f (y, z) ,
(2.26)
where f (y, z) is at this point an arbitrary function of y and z. The important thing is that
it has no x-dependence, so f /x = 0. Next, we have
U
= kx
y
U (x, y, z) = kxy + g(x, z) .
(2.27)
Finally, the z-component integrates to yield
U
= 4bz 3
z
U (x, y, z) = bz 4 + h(x, y) .
(2.28)
We now equate the first two expressions:
kxy + f (y, z) = kxy + g(x, z) .
(2.29)
Subtracting kxy from each side, we obtain the equation f (y, z) = g(x, z). Since the LHS is
independent of x and the RHS is independent of y, we must have
f (y, z) = g(x, z) = q(z) ,
(2.30)
where q(z) is some unknown function of z. But now we invoke the final equation, to obtain
bz 4 + h(x, y) = kxy + q(z) .
(2.31)
The only possible solution is h(x, y) = C + kxy and q(z) = C + bz 4 , where C is a constant.
Therefore,
U (x, y, z) = C + kxy + bz 4 .
(2.32)
Note that it would be very wrong to integrate U/x = ky and obtain U (x, y, z) = kxy +
C , where C is a constant. As weve seen, the constant of integration we obtain upon
integrating this first order PDE is in fact a function of y and z. The fact that f (y, z) carries
no explicit x dependence means that f /x = 0, so by construction U = kxy + f (y, z) is a
solution to the PDE U/x = ky, for any arbitrary function f (y, z).
2.3
Conservative Forces in Many Particle Systems
T =
2
1
2 mi r i
U=
X
i
V (ri ) +
(2.33)
X
i<j
v |ri rj | .
(2.34)
16
CHAPTER 2. SYSTEMS OF PARTICLES
Here, V (r) is the external (or one-body) potential, and v(rr ) is the interparticle potential,
which we assume to be central, depending only on the distance between any pair of particles.
The equations of motion are
(2.35)
mi ri = Fi(ext) + Fi(int) ,
with
V (ri )
ri
X (int)
X v |ri rj |
Fij .
=
ri
Fi(ext) =
Fi(int)
(2.36)
(2.37)
Here, Fij(int) is the force exerted on particle i by particle j:
(int)
Fij
v |ri rj |
ri rj
=
v |ri rj | .
=
ri
|ri rj |
(2.38)
Note that Fij(int) = Fji(int) , otherwise known as Newtons Third Law. It is convenient to
abbreviate rij ri rj , in which case we may write the interparticle force as
Fij(int) = rij v rij .
2.4
(2.39)
Linear and Angular Momentum
Consider now the total momentum of the system, P =
(int)
Fij
X
X (ext)
dP
=
p i =
Fi
dt
i
(int)
+Fji
pi . Its rate of change is
=0
z }| {
X
(ext)
Fij(int) = Ftot
,
+
(2.40)
i6=j
since the sum over all internal forces cancels as a result of Newtons Third Law. We write
X
(2.41)
P =
mi r i = M R
i
M=
mi
i
P
i mi ri
R= P
i mi
(total mass)
(center-of-mass) .
(2.42)
(2.43)
Next, consider the total angular momentum,
L=
X
i
ri pi =
X
i
mi ri r i .
(2.44)
2.4. LINEAR AND ANGULAR MOMENTUM
17
The rate of change of L is then
dL X
mi r i r i + mi ri ri
=
dt
i
X
X
=
ri Fi(ext) +
ri Fij(int)
i
i6=j
(int)
rij Fij
=0
zX
}|
{
X
(ext)
(int)
1
=
ri Fi + 2
(ri rj ) Fij
i
i6=j
(ext)
= Ntot .
(2.45)
Finally, it is useful to establish the result
T =
1
2
mi r i2 = 12 M R 2 +
1
2
2
mi r i R ,
(2.46)
which says that the kinetic energy may be written as a sum of two terms, those being the
kinetic energy of the center-of-mass motion, and the kinetic energy of the particles relative
to the center-of-mass.
Recall the work-energy theorem for conservative systems,
0=
final
Z
dE =
initial
final
Z
dT +
initial
final
Z
dU
initial
XZ
= T (B) T (A)
dri Fi ,
(2.47)
which is to say
T = T
(B)
(A)
XZ
i
dri Fi = U .
(2.48)
In other words, the total energy E = T + U is conserved:
E=
X
i
2
1
2 mi r i
X
i
V (ri ) +
X
i<j
v |ri rj | .
(2.49)
Note that for continuous systems, we replace sums by integrals over a mass distribution,
viz.
Z
X
mi ri d3r (r) (r) ,
(2.50)
i
where (r) is the mass density, and (r) is any function.
18
2.5
2.5.1
CHAPTER 2. SYSTEMS OF PARTICLES
Scaling of Solutions for Homogeneous Potentials
Eulers theorem for homogeneous functions
In certain cases of interest, the potential is a homogeneous function of the coordinates. This
means
U r1 , . . . , rN = k U r1 , . . . , rN .
(2.51)
Here, k is the degree of homogeneity of U . Familiar examples include gravity,
X mi mj
U r1 , . . . , rN = G
|ri rj |
i<j
k = 1 ,
(2.52)
and the harmonic oscillator,
U q1 , . . . , qn =
1
2
V q q
k = +2 .
(2.53)
The sum of two homogeneous functions is itself homogeneous only if the component functions themselves are of the same degree of homogeneity. Homogeneous functions obey a
special result known as Eulers Theorem, which we now prove. Suppose a multivariable
function H(x1 , . . . , xn ) is homogeneous:
H( x1 , . . . , xn ) = k H(x1 , . . . , xn ) .
Then
n
X
H
d
xi
= kH
H x1 , . . . , xn =
d
xi
(2.55)
i=1
=1
2.5.2
(2.54)
Scaled equations of motion
Now suppose the we rescale distances and times, defining
ri = ri
Then
dri
ri
d
=
dt
dt
t = t .
d2 ri
d2 ri
.
=
dt2
2 dt2
(2.56)
(2.57)
The force Fi is given by
U r1 , . . . , rN
ri
k U r1 , . . . , rN
=
(
ri )
k1
=
F .
Fi =
(2.58)
2.5. SCALING OF SOLUTIONS FOR HOMOGENEOUS POTENTIALS
19
Thus, Newtons 2nd Law says
d2 ri
m
= k1 Fi .
2 i dt2
(2.59)
If we choose such that
We now demand
= k1
2
= 1 2 k ,
(2.60)
then the equation of motion is invariant under the rescaling transformation!
This means
1
k1
t . This gives us
that if r(t) is a solution to the equations of motion, then so is r 2
an entire one-parameter family of solutions, for all real positive .
1
If r(t) is periodic with period T , the ri (t; ) is periodic with period T = 1 2 k T . Thus,
T
T
L
L
1 1 k
2
(2.61)
Here, = L /L is the ratio of length scales. Velocities, energies and angular momenta scale
accordingly:
L
1
v
L T
(2.62)
v =
=
= 2k
T
v
L
T
2 2
M L2
L
E
T
E =
=
= k
(2.63)
2
T
E
L
T
2
M L2
1
L
|L |
T
=
= (1+ 2 k) .
(2.64)
L =
T
|L|
L
T
As examples, consider:
(i) Harmonic Oscillator : Here k = 2 and therefore
q (t) q (t; ) = q (t) .
(2.65)
Thus, rescaling lengths alone gives another solution.
(ii) Kepler Problem : This is gravity, for which k = 1. Thus,
Thus, r 3 t2 , i.e.
r(t) r(t; ) = r 3/2 t .
also known as Keplers Third Law.
L
L
3
T
T
2
(2.66)
(2.67)
20
2.6
CHAPTER 2. SYSTEMS OF PARTICLES
Appendix I : Curvilinear Orthogonal Coordinates
The standard cartesian coordinates are {x1 , . . . , xd }, where d is the dimension of space.
Consider a different set of coordinates, {q1 , . . . , qd }, which are related to the original coordinates x via the d equations
q = q x1 , . . . , xd .
(2.68)
In general these are nonlinear equations.
i be the Cartesian set of orthonormal unit vectors, and define e to be the unit
Let e0i = x
vector perpendicular to the surface dq = 0. A differential change in position can now be
described in both coordinate systems:
ds =
d
X
e0i
dxi =
d
X
e h (q) dq ,
(2.69)
=1
i=1
where each h (q) is an as yet unknown function of all the components q . Finding the
coefficient of dq then gives
h (q) e =
d
X
xi 0
e
q i
i=1
where
e =
d
X
M i e0i ,
1 xi
.
h (q) q
The dot product of unit vectors in the new coordinate system is then
Mi (q) =
e e = M M
(2.70)
i=1
d
X
xi xi
1
.
=
h (q) h (q)
q q
(2.71)
(2.72)
i=1
The condition that the new basis be orthonormal is then
d
X
xi xi
= h2 (q) .
q q
(2.73)
i=1
This gives us the relation
v
u d
uX xi 2
.
h (q) = t
q
(2.74)
i=1
Note that
(ds)2 =
d
X
h2 (q) (dq )2 .
(2.75)
=1
For general coordinate systems, which are not necessarily orthogonal, we have
(ds)2 =
d
X
g (q) dq dq ,
,=1
where g (q) is a real, symmetric, positive definite matrix called the metric tensor .
(2.76)
2.6. APPENDIX I : CURVILINEAR ORTHOGONAL COORDINATES
21
Figure 2.2: Volume element for computing divergences.
2.6.1
Example : spherical coordinates
Consider spherical coordinates (, , ):
x = sin cos ,
y = sin sin ,
z = cos .
(2.77)
It is now a simple matter to derive the results
h2 = 1
Thus,
2.6.2
h2 = 2
h2 = 2 sin2 .
d .
ds = d + d + sin
(2.78)
(2.79)
Vector calculus : grad, div, curl
Here we restrict our attention to d = 3. The gradient U of a function U (q) is defined by
U
U
U
dq1 +
dq2 +
dq
q1
q2
q3 3
U ds .
dU =
Thus,
=
e1
e2
e3
+
+
.
h1 (q) q1 h2 (q) q2 h3 (q) q3
For the divergence, we use the divergence theorem, and we appeal to fig. 2.2:
Z
Z
A ,
dV A = dS n
(2.80)
(2.81)
(2.82)
22
CHAPTER 2. SYSTEMS OF PARTICLES
where is a region of three-dimensional space and is its closed two-dimensional boundary. The LHS of this equation is
LHS = A (h1 dq1 ) (h2 dq2 ) (h3 dq3 ) .
(2.83)
The RHS is
q +dq
q +dq
q +dq
1 3
2 2
1 1
dq1 dq2
dq1 dq3 + A3 h1 h2
dq2 dq3 + A2 h1 h3
RHS = A1 h2 h3
q3
q2
q1
=
A h h +
A h h +
A h h
dq1 dq2 dq3 .
(2.84)
q1 1 2 3
q2 2 1 3
q3 3 1 2
We therefore conclude
1
A=
A h h +
A h h +
A h h
h1 h2 h3 q1 1 2 3
q2 2 1 3
q3 3 1 2
To obtain the curl A, we use Stokes theorem again,
Z
I
A = d A ,
dS n
(2.85)
(2.86)
where is a two-dimensional region of space and is its one-dimensional boundary. Now
consider a differential surface element satisfying dq1 = 0, i.e. a rectangle of side lengths
h2 dq2 and h3 dq3 . The LHS of the above equation is
LHS = e1 A (h2 dq2 ) (h3 dq3 ) .
(2.87)
The RHS is
q +dq
q +dq
3 3
2 2
dq2
dq3 A2 h2
RHS = A3 h3
q3
q2
A3 h3
A h dq2 dq3 .
=
q2
q3 2 2
Therefore
1
h2 h3
This is one component of the full result
( A)1 =
A=
(h3 A3 ) (h2 A2 )
q2
q3
det
h1 h2 h2
h1 e1
q1
h2 e2
q2
h3 e3
q3
h1 A1 h2 A2 h3 A3
(2.88)
(2.89)
(2.90)
The Laplacian of a scalar function U is given by
2 U = U
1
=
h1 h2 h3
)
h1 h3 U
h1 h2 U
h2 h3 U
.
+
+
q1
h1 q1
q2
h2 q2
q3
h3 q3
(2.91)
2.7. COMMON CURVILINEAR ORTHOGONAL SYSTEMS
2.7
2.7.1
23
Common curvilinear orthogonal systems
Rectangular coordinates
In rectangular coordinates (x, y, z), we have
hx = hy = hz = 1 .
(2.92)
dx + y dy + z dz
ds = x
(2.93)
s 2 = x 2 + y 2 + z 2 .
(2.94)
Thus
and the velocity squared is
The gradient is
U = x
U
U
U
+ y
+ z
.
x
y
z
(2.95)
A=
Az
Ax Ay
+
+
.
x
y
z
(2.96)
The divergence is
The curl is
A=
Ay
Ay
Ax Az
Ax
Az
+
x
y +
z .
y
z
z
x
x
y
(2.97)
The Laplacian is
2 U =
2.7.2
2U
2U
2U
+
+
.
x2
y 2
z 2
(2.98)
Cylindrical coordinates
In cylindrical coordinates (, , z), we have
sin
= cos
x
cos
y = sin +
cos + y sin
= x
= x
sin + y cos
d
d =
= d .
d
(2.99)
(2.100)
The metric is given in terms of
h = 1
h =
hz = 1 .
(2.101)
Thus
d + z dz
ds = d +
(2.102)
s 2 = 2 + 2 2 + z 2 .
(2.103)
and the velocity squared is
24
CHAPTER 2. SYSTEMS OF PARTICLES
The gradient is
U
U
+
+ z
.
(2.104)
1 ( A ) 1 A Az
+
+
.
z
(2.105)
U =
The divergence is
A=
The curl is
A=
A
A Az
1 (A ) 1 A
1 Az
+
z .
+
z
z
The Laplacian is
1
2U
U
1 2U
U=
+
.
+ 2
2
z 2
2
2.7.3
(2.106)
(2.107)
Spherical coordinates
In spherical coordinates (r, , ), we have
sin cos + y sin sin + z sin
r = x
cos cos + y cos sin z cos
= x
= x
sin + y cos ,
for which
r =
= r ,
r = .
(2.108)
(2.109)
(2.110)
(2.111)
The inverse is
sin
= r sin cos + cos cos
x
cos
y = r sin sin + cos sin +
z = r cos sin .
(2.112)
(2.113)
(2.114)
The differential relations are
d
d
r = d + sin
d
d = r d + cos
= sin r + cos d
d
(2.115)
(2.116)
(2.117)
The metric is given in terms of
hr = 1
Thus
h = r
h = r sin .
r sin d
ds = r dr + r d +
(2.118)
(2.119)
2.7. COMMON CURVILINEAR ORTHOGONAL SYSTEMS
25
and the velocity squared is
s 2 = r 2 + r 2 2 + r 2 sin2 2 .
The gradient is
(2.120)
U
U
U
+
+
.
r
r sin
(2.121)
1 (r 2 Ar )
1 (sin A )
1 A
+
+
.
r2
r
r sin
r sin
(2.122)
U = r
The divergence is
A=
The curl is
1
A=
r sin
(rA )
(sin A ) A
1 Ar
1
r +
r sin
r
1 (rA ) Ar
+
.
r
r
(2.123)
The Laplacian is
2 U =
2.7.4
2U
1
U
1
1
2 U
.
r
+
sin
+
2
r 2 r
r
r 2 sin
r 2 sin 2
(2.124)
Kinetic energy
Note the form of the kinetic energy of a point particle:
T =
1
2m
ds
dt
2
= 12 m x 2 + y 2 + z 2
= 1 m 2 + 2 2
2
= 21 m 2 + 2 2 + z 2
= 12 m r 2 + r 2 2 + r 2 sin2 2
(3D Cartesian)
(2.125)
(2D polar)
(2.126)
(3D cylindrical)
(2.127)
(3D polar) .
(2.128)
26
CHAPTER 2. SYSTEMS OF PARTICLES
Chapter 3
One-Dimensional Conservative
Systems
3.1
Description as a Dynamical System
For one-dimensional mechanical systems, Newtons second law reads
m
x = F (x) .
(3.1)
A system is conservative if the force is derivable from a potential: F = dU/dx. The total
energy,
(3.2)
E = T + U = 21 mx 2 + U (x) ,
is then conserved. This may be verified explicitly:
i
d h1
dE
2
m
x
+
U
(x)
=
dt
dt 2
h
i
= m
x + U (x) x = 0 .
(3.3)
Conservation of energy allows us to reduce the equation of motion from second order to
first order:
v
!
u
u2
dx
E U (x) .
(3.4)
= t
dt
m
Note that the constant E is a constant of integration. The sign above depends on the
direction of motion. Points x(E) which satisfy
E = U (x)
x(E) = U 1 (E) ,
(3.5)
where U 1 is the inverse function, are called turning points. When the total energy is E,
the motion of the system is bounded by the turning points, and confined to the region(s)
27
28
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
U (x) E. We can integrate eqn. 3.4 to obtain
t(x) t(x0 ) =
m
2
Zx
x0
dx
.
E U (x )
(3.6)
This is to be inverted to obtain the function x(t). Note that there are now two constants
of integration, E and x0 . Since
E = E0 = 12 mv02 + U (x0 ) ,
(3.7)
we could also consider x0 and v0 as our constants of integration, writing E in terms of x0
and v0 . Thus, there are two independent constants of integration.
For motion confined between two turning points x (E), the period of the motion is given
by
xZ
+ (E)
dx
p
.
(3.8)
T (E) = 2m
E U (x )
x (E)
3.1.1
Example : harmonic oscillator
In the case of the harmonic oscillator, we have U (x) = 12 kx2 , hence
dt
=
dx
m
.
2E kx2
(3.9)
p
The turning points are x (E) = 2E/k, for E 0. To solve for the motion, let us
substitute
r
2E
x=
sin .
(3.10)
k
We then find
m
d ,
k
(3.11)
(t) = 0 + t ,
(3.12)
dt =
with solution
p
where = k/m is the harmonic oscillator frequency. Thus, the complete motion of the
system is given by
r
2E
sin(t + 0 ) .
(3.13)
x(t) =
k
Note the two constants of integration, E and 0 .
3.2. ONE-DIMENSIONAL MECHANICS AS A DYNAMICAL SYSTEM
3.2
29
One-Dimensional Mechanics as a Dynamical System
Rather than writing the equation of motion as a single second order ODE, we can instead
write it as two coupled first order ODEs, viz.
dx
=v
dt
(3.14)
dv
1
=
F (x) .
dt
m
(3.15)
This may be written in matrix-vector form, as
d x
v
.
= 1
dt v
m F (x)
(3.16)
This is an example of a dynamical system, described by the general form
d
= V () ,
dt
(3.17)
where = (1 , . . . , N ) is an N -dimensional vector in phase space. For the model of eqn.
3.16, we evidently have N = 2. The object V () is called a vector field. It is itself a vector,
existing at every point in phase space, RN . Each of the components of V () is a function
(in general) of all the components of :
Vj = Vj (1 , . . . , N )
(j = 1, . . . , N ) .
(3.18)
Solutions to the equation = V () are called integral curves. Each such integral curve
(t) is uniquely determined by N constants of integration, which may be taken to be the
initial value (0). The collection of all integral curves is known as the phase portrait of the
dynamical system.
In plotting the phase portrait of a dynamical system, we need to first solve for its motion,
starting from arbitrary initial conditions. In general this is a difficult problem, which can
only be treated numerically. But for conservative mechanical systems in d = 1, it is a trivial
matter! The reason is that energy conservation completely determines theq
phase portraits.
2
E U (x) .
The velocity becomes a unique double-valued function of position, v(x) = m
The phase curves are thus curves of constant energy.
3.2.1
Sketching phase curves
To plot the phase curves,
(i) Sketch the potential U (x).
(ii) Below this plot, sketch v(x; E) =
2
m
E U (x) .
30
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
Figure 3.1: A potential U (x) and the corresponding phase portraits. Separatrices are shown
in red.
(iii) When E lies at a local extremum of U (x), the system is at a fixed point.
(a) For E slightly above Emin , the phase curves are ellipses.
(b) For E slightly below Emax , the phase curves are (locally) hyperbolae.
(c) For E = Emax the phase curve is called a separatrix .
(iv) When E > U () or E > U (), the motion is unbounded.
(v) Draw arrows along the phase curves: to the right for v > 0 and left for v < 0.
The period of the orbit T (E) has a simple geometric interpretation. The area A in phase
space enclosed by a bounded phase curve is
A(E) =
v dx =
8
m
xZ
+ (E)
dx
x (E)
E U (x ) .
(3.19)
Thus, the period is proportional to the rate of change of A(E) with E:
T =m
A
.
E
(3.20)
3.3. FIXED POINTS AND THEIR VICINITY
3.3
31
Fixed Points and their Vicinity
A fixed point (x , v ) of the dynamics satisfies U (x ) = 0 and v = 0. Taylors theorem
then allows us to expand U (x) in the vicinity of x :
U (x) = U (x ) + U (x ) (x x ) + 21 U (x ) (x x )2 + 16 U (x ) (x x )3 + . . . . (3.21)
Since U (x ) = 0 the linear term in x = x x vanishes. If x is sufficiently small, we can
ignore the cubic, quartic, and higher order terms, leaving us with
U (x) U0 + 21 k(x)2 ,
(3.22)
where U0 = U (x ) and k = U (x ) > 0. The solutions to the motion in this potential are:
U (x ) > 0 : x(t) = x0 cos(t) +
v0
sin(t)
U (x ) < 0 : x(t) = x0 cosh(t) +
where =
k/m for k > 0 and =
v0
sinh(t) ,
(3.23)
(3.24)
k/m for k < 0. The energy is
E = U0 + 21 m (v0 )2 + 12 k (x0 )2 .
(3.25)
For a separatrix, we have E = U0 and U (x ) < 0. From the equation for the energy, we
obtain v0 = x0 . Lets take v0 = x0 , so that the initial velocity is directed toward
the unstable fixed point (UFP). I.e. the initial velocity is negative if we are to the right of
the UFP (x0 > 0) and positive if we are to the left of the UFP (x0 < 0). The motion of
the system is then
x(t) = x0 exp(t) .
(3.26)
The particle gets closer and closer to the unstable fixed point at x = 0, but it takes an
infinite amount of time to actually get there. Put another way, the time it takes to get from
x0 to a closer point x < x0 is
x0
1
.
(3.27)
t = ln
x
This diverges logarithmically as x 0. Generically, then, the period of motion along a
separatrix is infinite.
3.3.1
Linearized dynamics in the vicinity of a fixed point
Linearizing in the vicinity of such a fixed point, we write x = x x and v = v v ,
obtaining
d x
x
0
1
+ ...,
(3.28)
=
1
v
m U (x ) 0
dt v
32
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
Figure 3.2: Phase curves in the vicinity of centers and saddles.
This is a linear equation, which we can solve completely.
Consider the general linear equation = A , where A is a fixed real matrix. Now whenever
we have a problem involving matrices, we should start thinking about eigenvalues and
eigenvectors. Invariably, the eigenvalues and eigenvectors will prove to be useful, if not
essential, in solving the problem. The eigenvalue equation is
A = .
(3.29)
Here is the th right eigenvector 1 of A. The eigenvalues are roots of the characteristic
equation P () = 0, where P () = det( I A). Lets expand (t) in terms of the right
eigenvectors of A:
X
(t) =
C (t) .
(3.30)
Assuming, for the purposes of this discussion, that A is nondegenerate, and its eigenvectors
span RN , the dynamical system can be written as a set of decoupled first order ODEs for
the coefficients C (t):
C = C ,
(3.31)
with solutions
C (t) = C (0) exp( t) .
(3.32)
If Re ( ) > 0, C (t) flows off to infinity, while if Re ( ) > 0, C (t) flows to zero. If
| | = 1, then C (t) oscillates with frequency Im ( ).
1
If A is symmetric, the right and left eigenvectors are the same. If A is not symmetric, the right and left
eigenvectors differ, although the set of corresponding eigenvalues is the same.
3.4. EXAMPLES OF CONSERVATIVE ONE-DIMENSIONAL SYSTEMS
33
For a two-dimensional matrix, it is easy to show an exercise for the reader that
P () = 2 T + D ,
where T = Tr(A) and D = det(A). The eigenvalues are then
p
= 12 T 12 T 2 4D .
(3.33)
(3.34)
Well study the general case in Physics 110B. For now, we focus on our conservative mechanical system of eqn. 3.28. The trace and determinant of the above matrix are T = 0 and
1
D=m
U (x ). Thus, there are only two (generic) possibilities: centers, when U (x ) > 0,
and saddles, when U (x ) < 0. Examples of each are shown in Fig. 3.1.
3.4
3.4.1
Examples of Conservative One-Dimensional Systems
Harmonic oscillator
Recall again the harmonic oscillator, discussed in lecture 3. The potential energy is U (x) =
1
2
2 kx . The equation of motion is
m
dU
d2 x
=
= kx ,
2
dt
dx
(3.35)
where m is the mass and k the force constant (of a spring). With v = x,
this may be written
as the N = 2 system,
d x
0
1
x
v
=
=
,
(3.36)
2 0
v
2 x
dt v
p
where = k/m has the dimensions of frequency (inverse time). The solution is well
known:
v0
x(t) = x0 cos(t) +
sin(t)
(3.37)
v(t) = v0 cos(t) x0 sin(t) .
(3.38)
The phase curves are ellipses:
0 x2 (t) + 01 v 2 (t) = C ,
(3.39)
where C is a constant, independent of time. A sketch of the phase curves and of the phase
flow is shown in Fig. 3.3. Note that the x and v axes have different dimensions.
Energy is conserved:
E = 12 mv 2 + 12 kx2 .
(3.40)
Therefore we may find the length of the semimajor and semiminor axes by setting v = 0 or
x = 0, which gives
r
r
2E
2E
, vmax =
.
(3.41)
xmax =
k
m
34
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
Figure 3.3: Phase curves for the harmonic oscillator.
The area of the elliptical phase curves is thus
2E
.
A(E) = xmax vmax =
mk
(3.42)
The period of motion is therefore
A
= 2
T (E) = m
E
m
,
k
(3.43)
which is independent of E.
3.4.2
Pendulum
Next, consider the simple pendulum, composed of a mass point m affixed to a massless rigid
rod of length . The potential is U () = mg cos , hence
m2 =
dU
= mg sin .
d
(3.44)
This is equivalent to
=
,
(3.45)
02 sin
p
where = is the angular velocity, and where 0 = g/ is the natural frequency of small
oscillations.
d
dt
The conserved energy is
E=
1
2
m2 2 + U () .
(3.46)
Assuming the pendulum is released from rest at = 0 ,
2E
= 2 202 cos = 202 cos 0 .
m2
(3.47)
3.4. EXAMPLES OF CONSERVATIVE ONE-DIMENSIONAL SYSTEMS
35
Figure 3.4: Phase curves for the simple pendulum. The separatrix divides phase space into
regions of rotation and libration.
The period for motion of amplitude 0 is then
T 0
Z0
d
4
8
=
=
K sin2 12 0 ,
0
0
cos cos 0
(3.48)
where K(z) is the complete elliptic integral of the first kind. Expanding K(z), we have
2
2 1
4 1
9
1
T 0 =
(3.49)
1 + 4 sin 2 0 + 64 sin 2 0 + . . . .
0
For 0 0, the period approaches the usual result 2/0 , valid for the linearized equation
= 02 . As 0 2 , the period diverges logarithmically.
The phase curves for the pendulum are shown in Fig. 3.4. The small oscillations of the
pendulum are essentially the same as those of a harmonic oscillator. Indeed, within the
small angle approximation, sin , and the pendulum equations of motion are exactly
those of the harmonic oscillator. These oscillations are called librations. They involve
a back-and-forth motion in real space, and the phase space motion is contractable to a
point, in the topological sense. However, if the initial angular velocity is large enough, a
qualitatively different kind of motion is observed, whose phase curves are rotations. In this
case, the pendulum bob keeps swinging around in the same direction, because, as well see
in a later lecture, the total energy is sufficiently large. The phase curve which separates
these two topologically distinct motions is called a separatrix .
36
3.4.3
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
Other potentials
Using the phase plotter application written by Ben Schmidel, available on the Physics 110A
course web page, it is possible to explore the phase curves for a wide variety of potentials.
Three examples are shown in the following pages. The first is the effective potential for the
Kepler problem,
2
k
,
(3.50)
Ueff (r) = +
r
2r 2
about which we shall have much more to say when we study central forces. Here r is the
separation between two gravitating bodies of masses m1,2 , = m1 m2 /(m1 + m2 ) is the
reduced mass, and k = Gm1 m2 , where G is the Cavendish constant. We can then write
1
1
Ueff (r) = U0 + 2 ,
(3.51)
x 2x
where r0 = 2 /k has the dimensions of length, and x r/r0 , and where U0 = k/r0 =
k2 /2 . Thus, if distances are measured in units of r0 and the potential in units of U0 , the
potential may be written in dimensionless form as U(x) = x1 + 2x1 2 .
The second is the hyperbolic secant potential,
U (x) = U0 sech2 (x/a) ,
(3.52)
which, in dimensionless form, is U(x) = sech2 (x), after measuring distances in units of a
and potential in units of U0 .
The final example is
U (x) = U0
x
x
+
cos
a
2a
(3.53)
Again measuring x in units of a and U in units of U0 , we arrive at U(x) = cos(x) + 12 x.
3.4. EXAMPLES OF CONSERVATIVE ONE-DIMENSIONAL SYSTEMS
Figure 3.5: Phase curves for the Kepler effective potential U (x) = x1 + 21 x2 .
37
38
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
Figure 3.6: Phase curves for the potential U (x) = sech2 (x).
3.4. EXAMPLES OF CONSERVATIVE ONE-DIMENSIONAL SYSTEMS
Figure 3.7: Phase curves for the potential U (x) = cos(x) + 21 x.
39
40
CHAPTER 3. ONE-DIMENSIONAL CONSERVATIVE SYSTEMS
Chapter 4
Linear Oscillations
Harmonic motion is ubiquitous in Physics. The reason is that any potential energy function,
when expanded in a Taylor series in the vicinity of a local minimum, is a harmonic function:
U (~
q )=0
z }|
{
N
N
X
X
2 U
U
1
U (~q ) = U (~
q )+
(q qj ) + 2
(q qj ) (qk qk ) + . . . , (4.1)
qj q~=~q j
qj qk q~=~q j
j=1
j,k=1
where the {qj } are generalized coordinates more on this when we discuss Lagrangians. In
one dimension, we have simply
U (x) = U (x ) +
1
2
U (x ) (x x )2 + . . . .
(4.2)
Provided the deviation = x x is small enough in magnitude, the remaining terms in
the Taylor expansion may be ignored. Newtons Second Law then gives
m = U (x ) + O( 2 ) .
(4.3)
This, to lowest order, is the equation of motion for a harmonic oscillator. If U (x ) > 0,
the equilibrium point x = x is stable, since for small deviations from equilibrium the
restoring force pushes the system back toward the equilibrium point. When U (x ) < 0,
the equilibrium is unstable, and the forces push one further away from equilibrium.
4.1
Damped Harmonic Oscillator
In the real world, there are frictional forces, which we here will approximate by F = v.
We begin with the homogeneous equation for a damped harmonic oscillator,
dx
d2x
+ 02 x = 0 ,
+ 2
2
dt
dt
41
(4.4)
42
CHAPTER 4. LINEAR OSCILLATIONS
P
where = 2m. To solve, write x(t) = i Ci eii t . This renders the differential equation
4.4 an algebraic equation for the two eigenfrequencies i , each of which must satisfy
2 + 2i 02 = 0 ,
(4.5)
= i (02 2 )1/2 .
(4.6)
hence
The most general solution to eqn. 4.4 is then
x(t) = C+ ei+ t + C ei t
(4.7)
where C are arbitrary constants. Notice that the eigenfrequencies are in general complex,
with a negative imaginary part (so long as the damping coefficient is positive). Thus
ei t decays to zero as t .
4.1.1
Classes of damped harmonic motion
We identify three classes of motion:
(i) Underdamped (02 > 2 )
(ii) Overdamped (02 < 2 )
(iii) Critically Damped (02 = 2 ) .
Underdamped motion
The solution for underdamped motion is
x(t) = A cos(t + ) et
x(t)
= 0 A cos t + + sin1 (/0 ) et ,
(4.8)
p
where = 02 2 , and where A and are constants determined by initial conditions.
From x0 = A cos and x 0 = A cos A sin , we have x 0 + x0 = A sin , and
s
0 + x0
x 0 + x0 2
1 x
2
A = x0 +
,
= tan
.
(4.9)
x0
Overdamped motion
The solution in the case of overdamped motion is
x(t) = C e()t + D e(+)t
x(t)
= ( ) C e()t ( + ) D e(+)t ,
(4.10)
4.1. DAMPED HARMONIC OSCILLATOR
where =
43
2 02 and where C and D are constants determined by the initial conditions:
1
1
C
x0
.
(4.11)
=
( ) ( + )
D
x 0
Inverting the above matrix, we have the solution
C=
x
( + ) x0
+ 0
2
2
D=
( ) x0
x
0 .
2
2
(4.12)
Critically damped motion
The solution in the case of critically damped motion is
x(t) = E et + F t et
x(t)
= E + (t 1)F et .
(4.13)
Thus, x0 = E and x 0 = F E, so
E = x0
F = x 0 + x0 .
(4.14)
The screen door analogy
The three types of behavior are depicted in fig. 4.1. To concretize these cases in ones mind,
it is helpful to think of the case of a screen door or a shock absorber. If the hinges on the
door are underdamped, the door will swing back and forth (assuming it doesnt have a rim
which smacks into the door frame) several times before coming to a stop. If the hinges are
overdamped, the door may take a very long time to close. To see this, note that for 0
we have
q
02 1/2
2
2
0 = 1 2
02
04
= 1 2 4 + ... ,
(4.15)
2
8
which leads to
q
2
4
2 02 = 0 + 03 + . . .
2 8
q
2
+ 2 02 = 2 0 + . . . .
2
(4.16)
Thus, we can write
x(t) = C et/1 + D et/2 ,
(4.17)
44
CHAPTER 4. LINEAR OSCILLATIONS
Figure 4.1: Three classifications of damped harmonic motion. The initial conditions are
x(0) = 1, x(0)
= 0.
with
1 =
2 =
p
p
02
2
02
(4.18)
02
1
.
2
(4.19)
2
1
Thus x(t) is a sum of exponentials, with decay times 1,2 . For 0 , we have that 1 is
much larger than 2 the ratio is 1 /2 4 2 /02 1. Thus, on time scales on the order of
1 , the second term has completely damped away. The decay time 1 , though, is very long,
since is so large. So a highly overdamped oscillator will take a very long time to come to
equilbrium.
4.1.2
Remarks on the case of critical damping
Define the first order differential operator
Dt =
d
+ .
dt
(4.20)
4.1. DAMPED HARMONIC OSCILLATOR
45
The solution to Dt x(t) = 0 is x
(t) = A et , where A is a constant. Note that the commutator of Dt and t is unity:
Dt , t = 1 ,
(4.21)
where [A, B] AB BA. The simplest way to verify eqn. 4.21 is to compute its action
upon an arbitrary function f (t):
d
d
Dt , t f (t) =
+ t f (t) t
+ f (t)
dt
dt
d
d
(4.22)
t f (t) t f (t) = f (t) .
=
dt
dt
We know that x(t) = x
(t) = A et satisfies Dt x(t) = 0. Therefore
(t)
0 = Dt Dt , t x
0
z }| {
2
(t) Dt t Dt x
= Dt t x
(t)
(t) .
= Dt2 t x
(4.23)
(t) = Dt Dt x
(t) = 0. The above equation establishes that the
We already know that Dt2 x
(t). Indeed,
second independent solution to the second order ODE Dt2 x(t) = 0 is x(t) = t x
we can keep going, and show that
(t) = 0 .
(4.24)
Dtn tn1 x
Thus, the n independent solutions to the nth order ODE
n
d
+ x(t) = 0
dt
are
xk (t) = A tk et , k = 0, 1, . . . , n 1 .
4.1.3
(4.25)
(4.26)
Phase portraits for the damped harmonic oscillator
Expressed as a dynamical system, the equation of motion x
+ 2 x + 02 x = 0 is written as
two coupled first order ODEs, viz.
x = v
v = 02 x 2v .
(4.27)
In the theory of dynamical systems, a nullcline is a curve along which one component of
the phase space velocity vanishes. In our case, there are two nullclines: x = 0 and v = 0.
The equation of the first nullcline, x = 0, is simply v = 0, i.e. the first nullcline is the
x-axis. The equation of the second nullcline, v = 0, is v = (02 /2) x. This is a line which
runs through the origin and has negative slope. Everywhere along the first nullcline x = 0,
we have that lies parallel to the v-axis. Similarly, everywhere along the second nullcline
v = 0, we have that lies parallel to the x-axis. The situation is depicted in fig. 4.2.
46
CHAPTER 4. LINEAR OSCILLATIONS
Figure 4.2: Phase curves for the damped harmonic oscillator. Left panel: underdamped
motion. Right panel: overdamped motion. Note the nullclines along x = 0 and v =
(02 /2)x, which are shown as dashed lines.
4.2
Damped Harmonic Oscillator with Forcing
When forced, the equation for the damped oscillator becomes
dx
d2x
+ 02 x = f (t) ,
+ 2
2
dt
dt
(4.28)
where f (t) = F (t)/m. Since this equation is linear in x(t), we can, without loss of generality,
restrict out attention to harmonic forcing terms of the form
i
h
(4.29)
f (t) = f0 cos(t + 0 ) = Re f0 ei0 eit
where Re stands for real part. Here, is the forcing frequency.
Consider first the complex equation
d2z
dz
+ 2
+ 02 z = f0 ei0 eit .
dt2
dt
(4.30)
We try a solution z(t) = z0 eit . Plugging in, we obtain the algebraic equation
z0 =
02
f0 ei0
A() ei() f0 ei0 .
2i 2
(4.31)
The amplitude A() and phase shift () are given by the equation
A() ei() =
02
1
.
2i 2
(4.32)
4.2. DAMPED HARMONIC OSCILLATOR WITH FORCING
47
A basic fact of complex numbers:
1
1
a + ib
ei tan (b/a)
= 2
=
.
a ib
a + b2
a2 + b2
(4.33)
Thus,
A() =
1/2
(02 2 )2 + 4 2 2
() = tan
2
2
0 2
(4.34)
(4.35)
Now since the coefficients and 02 are real, we can take the complex conjugate of eqn.
4.30, and write
z + 2 z + 02 z = f0 ei0 eit
(4.36)
z
+ 2 z + 02 z = f0 e+i0 e+it ,
(4.37)
where z is the complex conjugate of z. We now add these two equations and divide by two
to arrive at
x
+ 2 x + 02 x = f0 cos(t + 0 ) .
(4.38)
Therefore, the real, physical solution we seek is
i
h
xinh (t) = Re A() ei() f0 ei0 eit
= A() f0 cos t + 0 () .
(4.39)
The quantity A() is the amplitude of the response (in units of f0 ), while () is the
(dimensionless) phase lag (typically expressed in radians).
The maximum of the amplitude A() occurs when A () = 0. From
dA
2
2
2
2
=
3 0 + 2 ,
d
A()
(4.40)
we conclude that A () = 0 for = 0 and for = R , where
R =
02 2 2 .
(4.41)
The solution at = R pertains only if 02 > 2 2 , of course, in which case = 0 is a local
minimum and = R a local maximum. If 02 < 2 2 there is only a local maximum, at
= 0. See Fig. 4.3.
48
CHAPTER 4. LINEAR OSCILLATIONS
Figure 4.3: Amplitude and phase shift versus oscillator frequency (units of 0 ) for /0
values of 0.1 (red), 0.25 (magenta), 1.0 (green), and 2.0 (blue).
Since equation 4.28 is linear, we can add a solution to the homogeneous equation to xinh (t)
and we will still have a solution. Thus, the most general solution to eqn. 4.28 is
x(t) = xinh (t) + xhom (t)
i
h
= Re A() ei() f0 ei0 eit + C+ ei+ t + C ei t
xinh (t)
where =
(t)
hom
z
}|
}|
{
{ z t
=A() f0 cos t + 0 () + C e
cos(t) + D et sin(t) ,
(4.42)
02 2 as before.
The last two terms in eqn. 4.42 are the solution to the homogeneous equation, i.e. with
f (t) = 0. They are necessary to include because they carry with them the two constants
of integration which always arise in the solution of a second order ODE. That is, C and D
are adjusted so as to satisfy x(0) = x0 and x 0 = v0 . However, due to their et prefactor,
these terms decay to zero once t reaches a relatively low multiple of 1 . They are called
transients, and may be set to zero if we are only interested in the long time behavior of the
system. This means, incidentally, that the initial conditions are effectively forgotten over a
time scale on the order of 1 .
For R > 0, one defines the quality factor , Q, of the oscillator by Q = R /2. Q is a rough
measure of how many periods the unforced oscillator executes before its initial amplitude is
damped down to a small value. For a forced oscillator driven near resonance, and for weak
damping, Q is also related to the ratio of average energy in the oscillator to the energy lost
4.2. DAMPED HARMONIC OSCILLATOR WITH FORCING
49
per cycle by the external source. To see this, let us compute the energy lost per cycle,
2/
Z
E = m
dt x f (t)
= m
2/
Z
0
dt A f02 sin(t + 0 ) cos(t + 0 )
= A f02 m sin
= 2 m A2 () f02 ,
(4.43)
since sin () = 2 A(). The oscillator energy, averaged over the cycle, is
E =
2
=
2/
Z
dt 12 m x 2 + 02 x2
0
2
1
4 m (
+ 02 ) A2 () f02 .
(4.44)
Thus, we have
2hEi
2 + 02
=
.
E
4
(4.45)
Thus, for R and 2 02 , we have
Q
4.2.1
0
2hEi
.
E
2
(4.46)
Resonant forcing
When the damping vanishes, the response diverges at resonance. The solution to the
resonantly forced oscillator
x
+ 02 x = f0 cos(0 t + 0 )
(4.47)
is given by
xhom (t)
x(t) =
f0
20
}|
{
z
t sin(0 t + 0 )+ A cos(0 t) + B sin(0 t) .
(4.48)
The amplitude of this solution grows linearly due to the energy pumped into the oscillator by
the resonant external forcing. In the real world, nonlinearities can mitigate this unphysical,
unbounded response.
50
CHAPTER 4. LINEAR OSCILLATIONS
Figure 4.4: An R-L-C circuit which behaves as a damped harmonic oscillator.
4.2.2
R-L-C circuits
Consider the R-L-C circuit of Fig. 4.4. When the switch is to the left, the capacitor is
charged, eventually to a steady state value Q = CV . At t = 0 the switch is thrown to the
right, completing the R-L-C circuit. Recall that the sum of the voltage drops across the
three elements must be zero:
dI
Q
L
+ IR +
=0.
(4.49)
dt
C
We also have Q = I, hence
1
d2 Q R dQ
+
Q=0,
+
2
dt
L dt
LC
(4.50)
which is the equation for a damped harmonic oscillator, with 0 = (LC)1/2 and = R/2L.
The boundary conditions at t = 0 are Q(0) = CV and Q(0)
= 0. Under these conditions,
the full solution at all times is
Q(t) = CV et cos t + sin t
I(t) = CV 0 et sin t ,
again with =
(4.51)
(4.52)
02 2 .
If we put a time-dependent voltage source in series with the resistor, capacitor, and inductor,
we would have
Q
dI
+ IR +
= V (t) ,
(4.53)
L
dt
C
which is the equation of a forced damped harmonic oscillator.
4.2. DAMPED HARMONIC OSCILLATOR WITH FORCING
4.2.3
51
Examples
Third order linear ODE with forcing
The problem is to solve the equation
...
Lt x x + (a + b + c) x + (ab + ac + bc) x + abc x = f0 cos(t) .
(4.54)
The key to solving this is to note that the differential operator Lt factorizes:
d3
d2
d
+
(a
+
b
+
c)
+ (ab + ac + bc)
+ abc
3
2
dt
dt
dt
d
d
d
=
+a
+b
+c ,
dt
dt
dt
Lt =
(4.55)
which says that the third order differential operator appearing in the ODE is in fact a
product of first order differential operators. Since
dx
+ x = 0
dt
x(t) = A ex ,
(4.56)
we see that the homogeneous solution takes the form
xh (t) = A eat + B ebt + C ect ,
(4.57)
where A, B, and C are constants.
To find the inhomogeneous solution, we solve Lt x = f0 eit and take the real part. Writing
x(t) = x0 eit , we have
Lt x0 eit = (a i) (b i) (c i) x0 eit
and thus
x0 =
(4.58)
f0 eit
A() ei() f0 eit ,
(a i)(b i)(c i)
where
h
i1/2
A() = (a2 + 2 ) (b2 + 2 ) (c2 + 2 )
1
1
() = tan1
+ tan
+ tan
.
a
b
c
(4.59)
(4.60)
Thus, the most general solution to Lt x(t) = f0 cos(t) is
x(t) = A() f0 cos t () + A eat + B ebt + C ect .
Note that the phase shift increases monotonically from (0) = 0 to () = 32 .
(4.61)
52
CHAPTER 4. LINEAR OSCILLATIONS
Figure 4.5: A driven L-C-R circuit, with V (t) = V0 cos(t).
Mechanical analog of RLC circuit
Consider the electrical circuit in fig. 4.5. Our task is to construct its mechanical analog.
To do so, we invoke Kirchoffs laws around the left and right loops:
Q1
+ R1 (I1 I2 ) = 0
L1 I1 +
C1
L2 I2 + R2 I2 + R1 (I2 I1 ) = V (t) .
(4.62)
(4.63)
Let Q1 (t) be the charge on the left plate of capacitor C1 , and define
Zt
Q2 (t) = dt I2 (t ) .
(4.64)
Then Kirchoffs laws may be written
1 + R1 (Q 1 Q 2 ) + 1 Q1 = 0
Q
L1
L1 C1
(4.65)
2 + R2 Q 2 + R1 (Q 2 Q 1 ) = V (t) .
Q
L2
L2
L2
(4.66)
Now consider the mechanical system in Fig. 4.6. The blocks have masses M1 and M2 .
The friction coefficient between blocks 1 and 2 is b1 , and the friction coefficient between
block 2 and the floor is b2 . Here we assume a velocity-dependent frictional force Ff = bx,
rather than the more conventional constant Ff = W , where W is the weight of an
object. Velocity-dependent friction is applicable when the relative velocity of an object and
a surface is sufficiently large. There is a spring of spring constant k1 which connects block 1
to the wall. Finally, block 2 is driven by a periodic acceleration f0 cos(t). We now identify
X1 Q1
X2 Q2
b1
R1
L1
b2
R2
L2
k1
1
,
L1 C1
(4.67)
4.2. DAMPED HARMONIC OSCILLATOR WITH FORCING
53
Figure 4.6: The equivalent mechanical circuit for fig. 4.5.
as well as f (t) V (t)/L2 .
The solution again proceeds by Fourier transform. We write
V (t) =
d
V () eit
2
d
2
(4.68)
and
Q1 (t)
I2 (t)
Q1 () it
e
I2 ()
(4.69)
The frequency space version of Kirchoffs laws for this problem is
G()
z
}|
2 i R1 /L1 + 1/L1 C1
i R1 /L2
R1 /L1
i + (R1 + R2 )/L2
()
Q
1
I2 ()
0
V ()/L2
(4.70)
The homogeneous equation has eigenfrequencies given by the solution to det G()
= 0,
which is a cubic equation. Correspondingly, there are three initial conditions to account
for: Q1 (0), I1 (0), and I2 (0). As in the case of the single damped harmonic oscillator, these
transients are damped, and for large times may be ignored. The solution then is
2
()
Q
i R1 /L1 + 1/L1 C1
1
=
i R1 /L2
I ()
R1 /L1
i + (R1 + R2 )/L2
V ()/L2
2
(4.71)
To obtain the time-dependent Q1 (t) and I2 (t), we must compute the Fourier transform back
to the time domain.
54
CHAPTER 4. LINEAR OSCILLATIONS
4.3
General solution by Greens function method
For a general forcing function f (t), we solve by Fourier transform. Recall that a function
F (t) in the time domain has a Fourier transform F () in the frequency domain. The relation
between the two is:1
Z
Z
d it
(4.72)
e
F () F () = dt e+it F (t) .
F (t) =
2
We can convert the differential equation 4.3 to an algebraic equation in the frequency
domain, x
() = G()
f(), where
G()
=
02
1
2i 2
(4.73)
is the Greens function in the frequency domain. The general solution is written
x(t) =
d it
e
G() f() + xh (t) ,
2
(4.74)
where xh (t) = i Ci eii t is a solution to the homogeneous equation. We may also write
the above integral over the time domain:
x(t) =
G(s) =
Z
dt G(t t ) f (t ) + xh (t)
(4.75)
d is
e
G()
2
= 1 exp(s) sin(s) (s)
(4.76)
where (s) is the step function,
(s) =
where once again
1 if s 0
0 if s < 0
(4.77)
(4.78)
02 2 .
Example: force pulse
Consider a pulse force
f (t) = f0 (t) (T t) =
1
f0 if 0 t T
0 otherwise.
Different texts often use different conventions for Fourier and inverse Fourier transforms. Sometimes the
factor of (2)1 is associated with the time integral, and sometimes a factor of (2)1/2 is assigned to both
frequency and time integrals. The convention I use is obviously the best.
4.4. GENERAL LINEAR AUTONOMOUS INHOMOGENEOUS ODES
55
Figure 4.7: Response of an underdamped oscillator to a pulse force.
In the underdamped regime, for example, we find the solution
f
x(t) = 02
0
if 0 t T and
x(t) =
f0
02
t
t
sin t
1e
cos t e
(4.79)
e(tT ) cos (t T ) et cos t
)
(tT )
e
sin (t T ) et sin t
+
(4.80)
if t > T .
4.4
General Linear Autonomous Inhomogeneous ODEs
This method immediately generalizes to the case of general autonomous linear inhomogeneous ODEs of the form
dnx
dn1 x
dx
+
a
+ . . . + a1
+ a0 x = f (t) .
n1
n
n1
dt
dt
dt
(4.81)
Lt x(t) = f (t) ,
(4.82)
We can write this as
56
CHAPTER 4. LINEAR OSCILLATIONS
where Lt is the nth order differential operator
Lt =
dn1
d
dn
+
a
+ . . . + a1
+ a0 .
n1
n
n1
dt
dt
dt
(4.83)
The general solution to the inhomogeneous equation is given by
Z
x(t) = xh (t) + dt G(t, t ) f (t ) ,
(4.84)
where G(t, t ) is the Greens function. Note that Lt xh (t) = 0. Thus, in order for eqns. 4.82
and 4.84 to be true, we must have
this vanishes
Z
z }| {
Lt x(t) = Lt xh (t) + dt Lt G(t, t ) f (t ) = f (t) ,
(4.85)
Lt G(t, t ) = (t t ) ,
(4.86)
which means that
where (t t ) is the Dirac -function. Some properties of (x):
Zb
f (y) if a < y < b
dx f (x) (x y) =
a
0
if y < a or y > b .
X (x x )
g(x) =
i ,
g (x )
i
(4.87)
(4.88)
x with
i
g(x )=0
i
valid for any functions f (x) and g(x). The sum in the second equation is over the zeros xi
of g(x).
Incidentally, the Dirac -function enters into the relation between a function and its Fourier
transform, in the following sense. We have
f (t) =
f() =
d it
e
f ()
2
(4.89)
dt e+it f (t) .
(4.90)
4.4. GENERAL LINEAR AUTONOMOUS INHOMOGENEOUS ODES
57
Substituting the second equation into the first, we have
f (t) =
d it
e
2
dt
( Z
dt eit f (t )
d i(t t)
e
2
f (t ) ,
(4.91)
which is indeed correct because the term in brackets is a representation of (t t ):
Z
d is
e = (s) .
2
(4.92)
If the differential equation Lt x(t) = f (t) is defined over some finite t interval with prescribed
boundary conditions on x(t) at the endpoints, then G(t, t ) will depend on t and t separately.
For the case we are considering, the interval is the entire real line t (, ), and
G(t, t ) = G(t t ) is a function of the single variable t t .
d
d
may be considered a function of the differential operator dt
. If we
Note that Lt = L dt
now Fourier transform the equation Lt x(t) = f (t), we obtain
n
Z
Z
d
dn1
d
it
it
dt e f (t) = dt e
+ a0 x(t)
+ an1 n1 + . . . + a1
dtn
dt
dt
it
dt e
n1
(i) + an1 (i)
(4.93)
+ . . . + a1 (i) + a0 x(t) ,
where we integrate by parts on t, assuming the boundary terms at t = vanish, i.e.
x() = 0, so that, inside the t integral,
"
#
k
k
d
d
x(t)
eit x(t) = (i)k eit x(t) .
eit
(4.94)
dt
dt
Thus, if we define
L()
=
then we have
n
X
ak (i)k ,
(4.95)
k=0
L()
x()
= f() ,
(4.96)
where an 1. According to the Fundamental Theorem of Algebra, the nth degree poly
nomial L()
may be uniquely factored over the complex plane into a product over n
roots:
L()
= (i)n ( 1 )( 2 ) ( n ) .
(4.97)
58
CHAPTER 4. LINEAR OSCILLATIONS
), hence if is a root then so is . Thus,
= L(
If the {ak } are all real, then L()
the roots appear in pairs which are symmetric about the imaginary axis. I.e. if = a + ib
is a root, then so is = a + ib.
The general solution to the homogeneous equation is
xh (t) =
n
X
Ai eii t ,
(4.98)
i=1
which involves n arbitrary complex constants Ai . The susceptibility, or Greens function in
Fourier space, G()
is then
G()
=
L()
in
,
( 1 )( 2 ) ( n )
(4.99)
and the general solution to the inhomogeneous equation is again given by
Z
x(t) = xh (t) + dt G(t t ) f (t ) ,
(4.100)
where xh (t) is the solution to the homogeneous equation, i.e. with zero forcing, and where
G(s) =
=i
d is
e
G()
2
d
eis
2 ( 1 )( 2 ) ( n )
n
X
eij s
j=1
i L (j )
(s) ,
(4.101)
where we assume that Im j < 0 for all j. The integral above was done using Cauchys theorem and the calculus of residues a beautiful result from the theory of complex functions.
As an example, consider the familiar case
L()
= 02 2i 2
= ( + ) ( ) ,
(4.102)
with = i , and = (02 2 )1/2 . This yields
L ( ) = (+ ) = 2 .
(4.103)
4.5. KRAMERS-KRONIG
RELATIONS (ADVANCED MATERIAL)
Then according to equation 4.101,
(
G(s) =
=
ei+ s
iL (+ )
ei s
iL ( )
59
(s)
es eis es eis
+
2i
2i
(s)
= 1 es sin(s) (s) ,
(4.104)
exactly as before.
4.5
Kramers-Kr
onig Relations (advanced material)
Suppose ()
G()
is analytic in the UHP2 . Then for all , we must have
Z
d
()
=0,
2 + i
(4.105)
where is a positive infinitesimal. The reason is simple: just close the contour in the UHP,
assuming ()
vanishes sufficiently rapidly that Jordans lemma can be applied. Clearly
this is an extremely weak restriction on (),
given the fact that the denominator already
1
causes the integrand to vanish as || .
Let us examine the function
i
1
=
.
2
2
+ i
( ) +
( )2 + 2
(4.106)
which we have separated into real and imaginary parts. Under an integral sign, the first
term, in the limit 0, is equivalent to taking a principal part of the integral. That is, for
any function F () which is regular at = ,
lim
d
F () P
2 ( )2 + 2
d F ()
.
2
(4.107)
The principal part symbol P means that the singularity at = is elided, either by
smoothing out the function 1/() as above, or by simply cutting out a region of integration
of width on either side of = .
The imaginary part is more interesting. Let us write
h(u)
2
.
u2 + 2
() for the susceptibility, rather than G()
In this section, we use the notation
(4.108)
60
CHAPTER 4. LINEAR OSCILLATIONS
For |u| , h(u) /u2 , which vanishes as 0. For u = 0, h(0) = 1/ which diverges as
0. Thus, h(u) has a huge peak at u = 0 and rapidly decays to 0 as one moves off the
peak in either direction a distance greater that . Finally, note that
Z
du h(u) = ,
(4.109)
a result which itself is easy to show using contour integration. Putting it all together, this
tells us that
= (u) .
(4.110)
lim 2
0 u + 2
Thus, for positive infinitesimal ,
1
1
= P i(u) ,
u i
u
(4.111)
a most useful result.
We now return to our initial result 4.105, and we separate ()
into real and imaginary
parts:
()
=
() + i
() .
(4.112)
(In this equation, the primes do not indicate differentiation with respect to argument.) We
therefore have, for every real value of ,
0=
ih
i
1
d h
() + i () P
i( ) .
2
(4.113)
Taking the real and imaginary parts of this equation, we derive the Kramers-Kr
onig relations:
() = +P
() = P
()
d
(4.114)
()
d
.
(4.115)
Chapter 5
Calculus of Variations
5.1
Snells Law
Warm-up problem: You are standing at point (x1 , y1 ) on the beach and you want to get to
a point (x2 , y2 ) in the water, a few meters offshore. The interface between the beach and
the water lies at x = 0. What path results in the shortest travel time? It is not a straight
line! This is because your speed v1 on the sand is greater than your speed v2 in the water.
The optimal path actually consists of two line segments, as shown in Fig. 5.1. Let the path
pass through the point (0, y) on the interface. Then the time T is a function of y:
q
q
1
1
2
2
x1 + (y y1 ) +
x22 + (y2 y)2 .
(5.1)
T (y) =
v1
v2
To find the minimum time, we set
1
1
dT
y y1
y2 y
q
q
=0=
dy
v1 x2 + (y y )2 v2 x2 + (y y)2
1
1
2
2
=
sin 1 sin 2
.
v1
v2
Thus, the optimal path satisfies
v1
sin 1
=
,
sin 2
v2
(5.2)
(5.3)
which is known as Snells Law.
Snells Law is familiar from optics, where the speed of light in a polarizable medium is
written v = c/n, where n is the index of refraction. In terms of n,
n1 sin 1 = n2 sin 2 .
(5.4)
If there are several interfaces, Snells law holds at each one, so that
ni sin i = ni+1 sin i+1 ,
61
(5.5)
62
CHAPTER 5. CALCULUS OF VARIATIONS
Figure 5.1: The shortest path between (x1 , y1 ) and (x2 , y2 ) is not a straight line, but rather
two successive line segments of different slope.
at the interface between media i and i + 1.
In the limit where the number of slabs goes to infinity but their thickness is infinitesimal,
we can regard n and as functions of a continuous variable x. One then has
y
sin (x)
= p
=P ,
v(x)
v 1 + y2
(5.6)
p
where P is a constant. Here wve have used the result sin p= y / 1 + y 2 , which follows
from drawing a right triangle with side lengths dx, dy, and dx2 + dy 2 . If we differentiate
the above equation with respect to x, we eliminate the constant and obtain the second order
ODE
v
1
y
=
.
(5.7)
v
1 + y2 y
This is a differential equation that y(x) must satisfy if the functional
T y(x) =
ds
=
v
Zx2 p
1 + y2
dx
v(x)
(5.8)
x1
is to be minimized.
5.2
Functions and Functionals
A function is a mathematical object which takes a real (or complex) variable, or several
such variables, and returns a real (or complex) number. A functional is a mathematical
5.2. FUNCTIONS AND FUNCTIONALS
63
Figure 5.2: The path of shortest length is composed of three line segments. The relation
between the angles at each interface is governed by Snells Law.
object which takes an entire function and returns a number. In the case at hand, we have
T y(x) =
Zx2
dx L(y, y , x) ,
(5.9)
(5.10)
x1
where the function L(y, y , x) is given by
1
L(y, y , x) =
v(x)
1 + y2 .
Here n(x) is a given function characterizing the medium, and y(x) is the path whose time
is to be evaluated.
In ordinary calculus, we extremize a function f (x) by demanding that f not change to
lowest order when we change x x + dx:
f (x + dx) = f (x) + f (x) dx + 12 f (x) (dx)2 + . . . .
(5.11)
We say that x = x is an extremum when f (x ) = 0.
For a functional, the first functional variation is obtained by sending y(x) y(x) + y(x),
64
CHAPTER 5. CALCULUS OF VARIATIONS
Figure 5.3: A path y(x) and its variation y(x) + y(x).
and extracting the variation in the functional to order y. Thus, we compute
T y(x) + y(x) =
=
Zx2
dx L(y + y, y + y , x)
x1
Zx2
x1
L
L
2
y + y + O (y)
dx L +
y
y
= T y(x) +
= T y(x) +
Zx2
L
L d
y +
y
dx
y
y dx
x1
Zx2
x1
x2
#
L
L
d L
dx
y
+
y
.
y
dx y
y
"
(5.12)
x1
Now one very important thing about the variation y(x) is that it must vanish at the
endpoints: y(x1 ) = y(x2 ) = 0. This is because the space of functions under consideration
satisfy fixed boundary conditions y(x1 ) = y1 and y(x2 ) = y2 . Thus, the last term in the
above equation vanishes, and we have
#
Zx2 "
d L
L
y .
(5.13)
T = dx
y
dx y
x1
We say that the first functional derivative of T with respect to y(x) is
"
#
T
d L
L
,
=
y(x)
y
dx y
(5.14)
where the subscript indicates
that
the expression inside the square brackets is to be evaluated
at x. The functional T y(x) is extremized when its first functional derivative vanishes,
5.2. FUNCTIONS AND FUNCTIONALS
65
which results in a differential equation for y(x),
L
d L
=0,
y
dx y
(5.15)
known as the Euler-Lagrange equation.
L(y, y , x) independent of y
Suppose L(y, y , x) is independent of y. Then from the Euler-Lagrange equations we have
that
L
(5.16)
P
y
is a constant.
In classical mechanics, this will turn out to be a generalized momentum. For
p
L = v1 1 + y 2 , we have
y
P = p
.
v 1 + y 2
(5.17)
Setting dP/dx = 0, we recover the second order ODE of eqn. 5.7. Solving for y ,
where v0 = 1/P .
v(x)
dy
,
= p 2
dx
v0 v 2 (x)
(5.18)
L(y, y , x) independent of x
When L(y, y , x) is independent of x, we can again integrate the equation of motion. Consider the quantity
L
H = y L .
(5.19)
y
Then
d L
L
L L
L
dH
L
d
=
y
L
=
y
+
y
y
y
dx
dx
y
y
dx y
y
y
x
d L
L
L
= y
,
dx y
y
x
where we have used the Euler-Lagrange equations to write
we have dH/dx = 0, i.e. H is a constant.
d L
dx y
L
y .
(5.20)
So if L/x = 0,
66
5.2.1
CHAPTER 5. CALCULUS OF VARIATIONS
Functional Taylor series
In general, we may expand a functional F [y + y] in a functional Taylor series,
Z
Z
Z
F [y + y] = F [y] + dx1 K1 (x1 ) y(x1 ) + 21! dx1 dx2 K2 (x1 , x2 ) y(x1 ) y(x2 )
Z
Z
Z
+ 31! dx1 dx2 dx3 K3 (x1 , x2 , x3 ) y(x1 ) y(x2 ) y(x3 ) + . . .
(5.21)
and we write
Kn (x1 , . . . , xn )
for the nth functional derivative.
5.3
nF
y(x1 ) y(xn )
(5.22)
Examples from the Calculus of Variations
Here we present three useful examples of variational calculus as applied to problems in
mathematics and physics.
5.3.1
Example 1 : minimal surface of revolution
Consider a surface formed by rotating the function y(x) about the x-axis. The area is then
s
2
Zx2
dy
,
(5.23)
A y(x) = dx 2y 1 +
dx
x1
p
and is a functional of the curve y(x). Thus we can define L(y, y ) = 2y 1 + y 2 and make
the identification y(x) q(t). Since L(y, y , x) is independent of x, we have
H = y
L
L
y
dH
L
=
,
dx
x
and when L has no explicit x-dependence, H is conserved. One finds
q
2y
y2
H = 2y p
2y 1 + y 2 = p
.
2
1 + y
1 + y2
Solving for y ,
dy
=
dx
s
2y
H
2
1 ,
H
which may be integrated with the substitution y = 2
cosh u, yielding
xa
,
y(x) = b cosh
b
(5.24)
(5.25)
(5.26)
(5.27)
5.3. EXAMPLES FROM THE CALCULUS OF VARIATIONS
67
Figure 5.4: Minimal surface solution, with y(x) = b cosh(x/b) and y(x0 ) = y0 . Top panel:
A/2y02 vs. y0 /x0 . Bottom panel: sech(x0 /b) vs. y0 /x0 . The blue curve corresponds to a
global minimum of A[y(x)], and the red curve to a local minimum or saddle point.
H
are constants of integration. Note there are two such constants, as
where a and b = 2
the original equation was second order. This shape is called a catenary. As we shall later
find, it is also the shape of a uniformly dense rope hanging between two supports, under
the influence of gravity. To fix the constants a and b, we invoke the boundary conditions
y(x1 ) = y1 and y(x2 ) = y2 .
Consider the case where x1 = x2 x0 and y1 = y2 y0 . Then clearly a = 0, and we have
x
0
= 1 cosh ,
(5.28)
y0 = b cosh
b
with y0 /x0 and x0 /b. One finds that for any > 1.5089 there are two solutions,
one of which is a global minimum and one of which is a local minimum or saddle of A[y(x)].
The solution with the smaller value of (i.e. the larger value of sech ) yields the smaller
value of A, as shown in Fig. 5.4. Note that
cosh(x/b)
y
,
=
y0
cosh(x0 /b)
(5.29)
so y(x = 0) = y0 sech(x0 /b).
When extremizing functions that are defined over a finite or semi-infinite interval, one
must take care to evaluate the function at the boundary, for it may be that the boundary
yields a global extremum even though the derivative may not vanish there. Similarly, when
extremizing functionals, one must investigate the functions at the boundary of function
68
CHAPTER 5. CALCULUS OF VARIATIONS
space. In this case, such a function would be the discontinuous solution, with
y1 if x = x1
y(x) = 0 if x1 < x < x2
y2 if x = x2 .
(5.30)
This solution corresponds to a surface consisting of two discs of radii y1 and y2 , joined
by an infinitesimally thin thread. The area functional evaluated for this particular y(x)
is clearly A = (y12 + y22 ). In Fig. 5.4, we plot A/2y02 versus the parameter = y0 /x0 .
For > c 1.564, one of the catenary solutions is the global minimum. For < c , the
minimum area is achieved by the discontinuous solution.
Note that the functional derivative,
(
)
2 1 + y 2 yy
A
d L
L
K1 (x) =
,
=
=
y(x)
y
dx y
(1 + y 2 )3/2
(5.31)
indeed vanishes for the catenary solutions, but does not vanish for the discontinuous solution, where K1 (x) = 2 throughout the interval (x0 , x0 ). Since y = 0 on this interval, y
cannot be decreased. The fact that K1 (x) > 0 means that increasing y will result in an
increase in A, so the boundary value for A, which is 2y02 , is indeed a local minimum.
We furthermore see in Fig. 5.4 that for < 1.5089 the local minimum and saddle
are no longer present. This is the familiar saddle-node bifurcation, here in function space.
Thus, for [0, ) there are no extrema of A[y(x)], and the minimum area occurs for the
discontinuous y(x) lying at the boundary of function space. For ( , c ), two extrema
exist, one of which is a local minimum and the other a saddle point. Still, the area is
minimized for the discontinuous solution. For (c , ), the local minimum is the global
minimum, and has smaller area than for the discontinuous solution.
5.3.2
Example 2 : geodesic on a surface of revolution
We use cylindrical coordinates (, , z) on the surface z = z(). Thus,
ds2 = d2 + 2 d2 + dx2
n
2 o
d + 2 d2 ,
= 1 + z ()
and the distance functional D () is
D () =
Z2
d L(, , ) ,
(5.32)
(5.33)
5.3. EXAMPLES FROM THE CALCULUS OF VARIATIONS
where
L(, , ) =
The Euler-Lagrange equation is
1 + z 2 () + 2 2 () .
d L
L
=0
d
Thus,
69
(5.34)
L
= const.
(5.35)
L
2
p
=
=a,
1 + z 2 + 2 2
where a is a constant. Solving for , we obtain
q
2
a 1 + z ()
p
d =
d ,
2 a2
(5.36)
(5.37)
which we must integrate to find (), subject to boundary conditions (i ) = i , with
i = 1, 2.
On a cone, z() = , and we have
d = a
which yields
1+
d
2 a2
() = +
which is equivalent to
cos
+ 2
1+
d tan
2
1 ,
a2
tan
1 + 2
=a.
2
1 ,
a2
(5.38)
(5.39)
(5.40)
The constants and a are determined from (i ) = i .
5.3.3
Example 3 : brachistochrone
Problem: find the path between (x1 , y1 ) and (x2 , y2 ) which a particle sliding frictionlessly
and under constant gravitational acceleration will traverse in the shortest time. To solve
this we first must invoke some elementary mechanics. Assuming the particle is released
from (x1 , y1 ) at rest, energy conservation says
2
1
2 mv
+ mgy = mgy1 .
Then the time, which is a functional of the curve y(x), is
s
Zx2
Zx2
ds
1
1 + y2
=
T y(x) =
dx
v
y1 y
2g
x1
Zx2
x1
dx L(y, y , x) ,
x1
(5.41)
(5.42)
70
CHAPTER 5. CALCULUS OF VARIATIONS
with
L(y, y , x) =
1 + y 2
.
2g(y1 y)
Since L is independent of x, eqn. 5.20, we have that
i
h
L
2 1/2
H = y L = 2g (y1 y) 1 + y
y
is conserved. This yields
dx =
y1 y
dy ,
2a y1 + y
(5.43)
(5.44)
(5.45)
with a = (4gH 2 )1 . This may be integrated parametrically, writing
y1 y = 2a sin2 ( 12 )
dx = 2a sin2 ( 12 ) d ,
(5.46)
which results in the parametric equations
x x1 = a sin
y y1 = a (1 cos ) .
(5.47)
(5.48)
This curve is known as a cycloid.
5.3.4
Ocean waves
Surface waves in fluids propagate with a definite relation between their angular frequency
and their wavevector k = 2/, where is the wavelength. The dispersion relation is a
function = (k). The group velocity of the waves is then v(k) = d/dk.
In a fluid with a flat bottom at depth h, the dispersion relation turns out to be
gh k shallow (kh 1)
p
(k) = gk tanh kh
gk
deep (kh 1) .
(5.49)
Suppose we are in the shallow case, where the wavelength is significantly greater than
the depth h of the fluid. This is the case for ocean waves which break atthe shore. The
phase velocity and group velocity are then identical, and equal to v(h) = gh. The waves
propagate more slowly as they approach the shore.
Let us choose the following coordinate system: x represents the distance parallel to the
shoreline, y the distance perpendicular to the shore (which lies at y = 0), and h(y) is the
depth profile of the bottom. We assume h(y) to be a slowly varying function of y which
satisfies h(0) = 0. Suppose a disturbance in the ocean at position (x2 , y2 ) propagates until
it reaches the shore at (x1 , y1 = 0). The time of propagation is
s
Zx2
Z
1 + y2
ds
= dx
.
(5.50)
T y(x) =
v
g h(y)
x1
5.3. EXAMPLES FROM THE CALCULUS OF VARIATIONS
71
Figure 5.5: For shallow water waves, v = gh. To minimize the propagation time from a
source to the shore, the waves break parallel to the shoreline.
We thus identify the integrand
L(y, y , x) =
1 + y 2
.
g h(y)
(5.51)
As with the brachistochrone problem, to which this bears an obvious resemblance, L is
cyclic in the independent variable x, hence
H = y
h
i1/2
L
2
L
=
g
h(y)
1
+
y
y
(5.52)
is constant. Solving for y (x), we have
dy
tan =
=
dx
a
1 ,
h(y)
(5.53)
where a = (gH)1 is a constant, and where is the local slope of the function y(x). Thus,
we conclude that near y = 0, where h(y) 0, the waves come in parallel to the shoreline.
If h(y) = y has a linear profile, the solution is again a cycloid, with
x() = b ( sin )
y() = b (1 cos ) ,
(5.54)
(5.55)
where b = 2a/ and where the shore lies at = 0. Expanding in a Taylor series in for
small , we may eliminate and obtain y(x) as
y(x) =
9 1/3 1/3 2/3
b x
2
+ ... .
(5.56)
72
CHAPTER 5. CALCULUS OF VARIATIONS
A tsunami is a shallow water wave that manages propagates in deep water. This requires
> h, as weve seen, which means the disturbance must have a very long spatial extent out
in the open ocean, where h 10 km. An undersea earthquake is the only possible source;
the characteristic length of
earthquake fault lines can be hundreds of kilometers. If we take
h = 10 km, we obtain v = gh 310 m/s or 1100 km/hr. At these speeds, a tsunami can
cross the Pacific Ocean in less than a day.
As the wave approaches the shore, it must slow down, since v = gh is diminishing. But
energy is conserved, which means that the amplitude must concomitantly rise. In extreme
cases, the water level rise at shore may be 20 meters or more.
5.4
Appendix : More on Functionals
We remarked in section 5.2 that a function f is an animal which gets fed a real number x
and excretes a real number f (x). We say f maps the reals to the reals, or
f: R R
(5.57)
Of course we also have functions g : C C which eat and excrete complex numbers,
multivariable functions h : RN R which eat N -tuples of numbers and excrete a single
number, etc.
A functional F [f (x)] eats entire functions (!) and excretes numbers. That is,
n
o
F : f (x) x R R
(5.58)
This says that F operates on the set of real-valued functions of a single real variable, yielding
a real number. Some examples:
Z
2
dx f (x)
F [f (x)] =
1
2
F [f (x)] =
Z
1
2
Z
dx dx K(x, x ) f (x) f (x )
(5.59)
(5.60)
2
Z
df
2
1
1
.
F [f (x)] = dx 2 A f (x) + 2 B
dx
(5.61)
In classical mechanics, the action S is a functional of the path q(t):
Ztb n
o
S[q(t)] = dt 12 mq2 U (q) .
ta
(5.62)
5.4. APPENDIX : MORE ON FUNCTIONALS
73
Figure 5.6: A functional S[q(t)] is the continuum limit of a function of a large number of
variables, S(q1 , . . . , qM ).
We can also have functionals which feed on functions of more than one independent variable,
such as
2 )
Ztb Zxb ( 2
y
y
S[y(x, t)] = dt dx 21
21
,
(5.63)
t
x
ta
xa
which happens to be the functional for a string of mass density under uniform tension .
Another example comes from electrodynamics:
Z
Z
1
1
S[A (x, t)] = d x dt
F F + j A
,
(5.64)
16
c
which is a functional of the four fields {A0 , A1 , A2 , A3 }, where A0 = c. These are the
components of the 4-potential, each of which is itself a function of four independent variables
(x0 , x1 , x2 , x3 ), with x0 = ct. The field strength tensor is written in terms of derivatives of
the A : F = A A , where we use a metric g = diag(+, , , ) to raise and
lower indices. The 4-potential couples linearly to the source term J , which is the electric
4-current (c, J).
We extremize functions by sending the independent variable x to x + dx and demanding
that the variation df = 0 to first order in dx. That is,
f (x + dx) = f (x) + f (x) dx + 12 f (x)(dx)2 + . . . ,
whence df = f (x) dx + O (dx)2 and thus
f (x ) = 0
x an extremum.
(5.65)
(5.66)
We extremize functionals by sending
f (x) f (x) + f (x)
(5.67)
74
CHAPTER 5. CALCULUS OF VARIATIONS
and demanding that the variation F in the functional F [f (x)] vanish to first order in f (x).
The variation f (x) must sometimes satisfy certain boundary conditions. For example, if
F [f (x)] only operates on functions which vanish at a pair of endpoints, i.e. f (xa ) = f (xb ) =
0, then when we extremize the functional F we must do so within the space of allowed
functions. Thus, we would in this case require f (xa ) = f (xb ) = 0. We may expand the
functional F [f + f ] in a functional Taylor series,
Z
Z
Z
F [f + f ] = F [f ] + dx1 K1 (x1 ) f (x1 ) + 21! dx1 dx2 K2 (x1 , x2 ) f (x1 ) f (x2 )
Z
Z
Z
1
+ 3 ! dx1 dx2 dx3 K3 (x1 , x2 , x3 ) f (x1 ) f (x2 ) f (x3 ) + . . .
(5.68)
and we write
Kn (x1 , . . . , xn )
nF
f (x1 ) f (xn )
(5.69)
In a more general case, F = F [{fi (x)} is a functional of several functions, each of which is
a function of several independent variables.1 We then write
Z
F [{fi + fi }] = F [{fi }] + dx1 K1i (x1 ) fi (x1 )
Z
Z
1
+ 2 ! dx1 dx2 K2ij (x1 , x2 ) fi (x1 ) fj (x2 )
Z
Z
Z
1
+ 3 ! dx1 dx2 dx3 K3ijk (x1 , x2 , x3 ) fi (x1 ) fj (x2 ) fk (x3 ) + . . . ,
(5.70)
with
i i2 in
Kn1
(x1 , x2 , . . . , xn ) =
nF
fi (x1 ) fi (x2 ) fi (xn )
1
(5.71)
Another way to compute functional derivatives is to send
f (x) f (x) + 1 (x x1 ) + . . . + n (x xn )
and then differentiate n times with respect to 1 through n . That is,
n
nF
=
F f (x) + 1 (x x1 ) + . . . + n (x xn ) .
f (x1 ) f (xn )
1 n = = =0
1
(5.72)
(5.73)
Lets see how this works. As an example, well take the action functional from classical
mechanics,
Ztb n
o
(5.74)
S[q(t)] = dt 12 mq2 U (q) .
ta
1
It may be also be that different functions depend on a different number of independent variables. E.g.
F = F [f (x), g(x, y), h(x, y, z)].
5.4. APPENDIX : MORE ON FUNCTIONALS
75
To compute the first functional derivative, we replace the function q(t) with q(t)+ (tt1 ),
and expand in powers of :
S q(t) + (t t1 ) = S[q(t)] +
Ztb n
o
dt m q (t t1 ) U (q) (t t1 )
ta
o
= m q(t1 ) + U q(t1 ) ,
n
hence
(5.75)
n
o
S
= m q(t) + U q(t)
q(t)
(5.76)
Z
dt (t t1 ) h(t) = h (t1 ) ,
(5.77)
and setting the first functional derivative to zero yields Newtons Second Law, m
q = U (q),
for all t [ta , tb ]. Note that we have used the result
which is easily established upon integration by parts.
To compute the second functional derivative, we replace
q(t) q(t) + 1 (t t1 ) + 2 (t t2 )
(5.78)
and extract the term of order 1 2 in the double Taylor expansion. One finds this term to
be
Ztb n
o
1 2 dt m (t t1 ) (t t2 ) U (q) (t t1 ) (t t2 ) .
(5.79)
ta
Note that we neednt bother with terms proportional to 21 or 22 since the recipe is to
differentiate once with respect to each of 1 and 2 and then to set 1 = 2 = 0. This
procedure uniquely selects the term proportional to 1 2 , and yields
n
o
2 S
= m (t1 t2 ) + U q(t1 ) (t1 t2 ) .
q(t1 ) q(t2 )
(5.80)
In multivariable calculus, the stability of an extremum is assessed by computing the matrix
of second derivatives at the extremal point, known as the Hessian matrix. One has
2f
f
=0 i
;
Hij =
.
(5.81)
xi x
xi xj x
The eigenvalues of the Hessian Hij determine the stability of the extremum. Since Hij is
a symmetric matrix, its eigenvectors may be chosen to be orthogonal. The associated
eigenvalues , defined by the equation
Hij j = i ,
(5.82)
76
CHAPTER 5. CALCULUS OF VARIATIONS
are the respective curvatures in the directions , where {1, . . . , n} where n is the
number of variables. The extremum is a local minimum if all the eigenvalues are positive,
a maximum if all are negative, and otherwise is a saddle point. Near a saddle point, there
are some directions in which the function increases and some in which it decreases.
In the case of functionals, the second functional derivative K2 (x1 , x2 ) defines an eigenvalue
problem for f (x):
Zxb
dx2 K2 (x1 , x2 ) f (x2 ) = f (x1 ) .
(5.83)
xa
In general there are an infinite number of solutions to this equation which form a basis in
function space, subject to appropriate boundary conditions at xa and xb . For example, in
the case of the action functional from classical mechanics, the above eigenvalue equation
becomes a differential equation,
d2
m 2 + U q (t) q(t) = q(t) ,
dt
(5.84)
where q (t) is the solution to the Euler-Lagrange equations. As with the case of ordinary
multivariable functions, the functional extremum is a local minimum (in function space)
if every eigenvalue is positive, a local maximum if every eigenvalue is negative, and a
saddle point otherwise.
Consider the simple harmonic oscillator, for which U (q) = 12 m02 q 2 . Then U q (t) =
m 02 ; note that we dont even need to know the solution q (t) to obtain the second functional
derivative in this special case. The eigenvectors obey m(q + 02 q) = q, hence
q(t) = A cos
q
02 + (/m) t + ,
(5.85)
where A and are constants. Demanding q(ta ) = q(tb ) = 0 requires
q
02 + (/m) tb ta ) = n ,
where n is an integer. Thus, the eigenfunctions are
t ta
qn (t) = A sin n
,
tb ta
and the eigenvalues are
n = m
n 2
T
m02 ,
(5.86)
(5.87)
(5.88)
where T = tb ta . Thus, so long as T > /0 , there is at least one negative eigenvalue.
(n+1)
there will be n negative eigenvalues. This means the action
Indeed, for n
0 < T <
0
is generally not a minimum, but rather lies at a saddle point in the (infinite-dimensional)
function space.
5.4. APPENDIX : MORE ON FUNCTIONALS
77
To test this explicitly, consider a harmonic oscillator with the boundary conditions q(0) = 0
and q(T ) = Q. The equations of motion, q + 02 q = 0, along with the boundary conditions,
determine the motion,
q (t) =
Q sin(0 t)
sin(0 T )
(5.89)
The action for this path is then
ZT n
o
S[q (t)] = dt 21 m q2 12 m02 q 2
=
=
m 02 Q2
2 sin2 0 T
2
1
2 m0 Q
ZT n
o
dt cos2 0 t sin2 0 t
0
ctn (0 T ) .
(5.90)
Next consider the path q(t) = Q t/T which satisfies the boundary conditions but does not
satisfy the equations of motion (it proceeds with constant velocity). One finds the action
for this path is
!
1
2
1
1
3 0 T .
(5.91)
S[q(t)] = 2 m0 Q
0 T
Thus, provided 0 T 6= n, in the limit T we find that the constant velocity path has
lower action.
Finally, consider the general mechanical action,
S q(t) =
Ztb
dt L(q, q,
t) .
(5.92)
ta
We now evaluate the first few terms in the functional Taylor series:
Ztb (
L
L
(5.93)
dt L(q , q , t) +
qi +
qi
qi
qi
q
q
ta
)
2L
1 2L
1 2L
+
qi qj +
qi qj +
qi qj + . . . .
2 qi qj
qi qj
2 qi qj
S q (t) + q(t) =
To identify the functional derivatives, we integrate by parts. Let ...(t) be an arbitrary
78
CHAPTER 5. CALCULUS OF VARIATIONS
function of time. Then
Ztb
Ztb
i (t) qi (t)
dt i (t) qi (t) = dt
(5.94)
ta
ta
Ztb
dt ij (t) qi (t) qj (t) =
ta
Ztb
ta
Ztb
d
dt dt ij (t) (t t ) qi (t) qj (t )
dt
ta
Ztb Ztb
= dt dt ij (t)) (t t ) qi (t) qj (t )
ta
ta
Ztb
dt ij (t) dqi (t) qj (t) =
ta
Ztb
ta
(5.95)
Ztb
d d
qi (t) qj (t )
dt dt ij (t) (t t )
dt dt
ta
Ztb
ta
Ztb
dt dt ij (t) (t t ) + ij (t) (t t ) qi (t) qj (t ) .
ta
(5.96)
Thus,
"
#
d L
S
L
=
qi (t)
qi dt qi
q (t)
(
2L
2L
2S
=
(t t )
(t t )
qi (t) qj (t )
qi qj
qi qj
q (t)
q (t)
"
#
)
2L
d
2L
+ 2
(t t ) .
qi qj
dt qi qj
q (t)
(5.97)
(5.98)
Chapter 6
Lagrangian Mechanics
6.1
Generalized Coordinates
A set of generalized coordinates q1 , . . . , qn completely describes the positions of all particles
in a mechanical system. In a system with df degrees of freedom and k constraints, n = df k
independent generalized coordinates are needed to completely specify all the positions. A
constraint is a relation among coordinates, such as x2 + y 2 + z 2 = a2 for a particle moving
on a sphere of radius a. In this case, df =
p3 and k = 1. In this case, we could eliminate z
in favor of x and y, i.e. by writing z = a2 x2 y 2 , or we could choose as coordinates
the polar and azimuthal angles and .
For the moment we will assume that n = df k, and that the generalized coordinates are
independent, satisfying no additional constraints among them. Later on we will learn how
to deal with any remaining constraints among the {q1 , . . . , qn }.
The generalized coordinates may have units of length, or angle, or perhaps something totally
different. In the theory of small oscillations, the normal coordinates are conventionally
chosen to have units of (mass)1/2 (length). However, once a choice of generalized coordinate
is made, with a concomitant set of units, the units of the conjugate momentum and force
are determined:
M L2
1
p =
T
q
M L2
1
,
F =
2
T
q
(6.1)
where A means the units of A, and where M , L, and T stand for mass, length, and time,
respectively. Thus, if q has dimensions of length, then p has dimensions of momentum
and F has dimensions of force. If q is dimensionless, as is the case for an angle, p has
dimensions of angular momentum (M L2 /T ) and F has dimensions of torque (M L2 /T 2 ).
79
80
6.2
CHAPTER 6. LAGRANGIAN MECHANICS
Hamiltons Principle
The equations of motion of classical mechanics are embodied in a variational principle,
called Hamiltons principle. Hamiltons principle states that the motion of a system is such
that the action functional
Zt2
t)
(6.2)
S q(t) = dt L(q, q,
t1
is an extremum, i.e. S = 0. Here, q = {q1 , . . . , qn } is a complete set of generalized
coordinates for our mechanical system, and
L=T U
(6.3)
is the Lagrangian, where T is the kinetic energy and U is the potential energy. Setting the
first variation of the action to zero gives the Euler-Lagrange equations,
momentum p
force F
z }| {
z}|{
d L
L
.
=
dt q
q
(6.4)
Thus, we have the familiar p = F , also known as Newtons second law. Note, however,
that the {q } are generalized coordinates, so p may not have dimensions of momentum,
nor F of force. For example, if the generalized coordinate in question is an angle , then
the corresponding generalized momentum is the angular momentum about the axis of s
rotation, and the generalized force is the torque.
6.2.1
Suppose
Then
Invariance of the equations of motion
d
q,
L(q,
t) = L(q, q,
t) + G(q, t) .
dt
(6.5)
S[q(t)]
= S[q(t)] + G(qb , tb ) G(qa , ta ) .
(6.6)
Since the difference S S is a function only of the endpoint values {qa , qb }, their variations
result in the same equations of motion.
are identical: S = S. This means that L and L
Thus, the equations of motion are invariant under a shift of L by a total time derivative of
a function of coordinates and time.
6.2.2
Remarks on the order of the equations of motion
The equations of motion are second order in time. This follows from the fact that L =
L(q, q,
t). Using the chain rule,
2L
2L
2L
d L
q +
q +
.
(6.7)
=
dt q
q q
q q
q t
6.2. HAMILTONS PRINCIPLE
81
That the equations are second order in time can be regarded as an empirical fact. It follows,
as we have just seen, from the fact that L depends on q and on q,
but on no higher time
derivative terms. Suppose the Lagrangian did depend on the generalized accelerations q as
well. What would the equations of motion look like?
Taking the variation of S,
tb
Ztb
L
d L
L
dt L(q, q,
q +
q
q
q, t) =
q
q
dt q
ta
ta
)
Ztb (
d L
d2 L
L
+ dt
q .
+ 2
q
dt q
dt q
(6.8)
ta
The boundary term vanishes if we require q (ta ) = q (tb ) = q (ta ) = q (tb ) = 0 .
The equations of motion would then be fourth order in time.
6.2.3
Lagrangian for a free particle
For a free particle, we can use Cartesian coordinates for each particle as our system of
generalized coordinates. For a single particle, the Lagrangian L(x, v, t) must be a function
solely of v 2 . This is because homogeneity with respect to space and time preclude any
dependence of L on x or on t, and isotropy of space means L must depend on v 2 . We
next invoke Galilean relativity, which says that the equations of motion are invariant under
transformation to a reference frame moving with constant velocity. Let V be the velocity
of the new reference frame K relative to our initial reference frame K. Then x = x V t,
and v = v V . In order that the equations of motion be invariant under the change in
reference frame, we demand
L (v ) = L(v) +
d
G(x, t) .
dt
(6.9)
The only possibility is L = 21 mv 2 , where the constant m is the mass of the particle. Note:
L = 12 m(v V )2 = 21 mv 2 +
d 1
dG
2
.
mV
t
mV
x
=L+
2
dt
dt
(6.10)
For N interacting particles,
L=
1
2
N
X
a=1
ma
dx 2
a
dt
U {xa }, {x a } .
Here, U is the potential energy. Generally, U is of the form
X
X
U=
U1 (xa ) +
v(xa xa ) ,
a
a<a
(6.11)
(6.12)
82
CHAPTER 6. LAGRANGIAN MECHANICS
however, as we shall see, velocity-dependent potentials appear in the case of charged particles interacting with electromagnetic fields. In general, though,
L=T U ,
(6.13)
where T is the kinetic energy, and U is the potential energy.
6.3
Conserved Quantities
A conserved quantity (q, q,
t) is one which does not vary throughout the motion of the
system. This means
d
= 0.
(6.14)
dt
q=q(t)
We shall discuss conserved quantities in detail in the chapter on Noethers Theorem, which
follows.
6.3.1
Momentum conservation
The simplest case of a conserved quantity occurs when the Lagrangian does not explicitly
depend on one or more of the generalized coordinates, i.e. when
F =
L
=0.
q
(6.15)
We then say that L is cyclic in the coordinate q . In this case, the Euler-Lagrange equations
p = F say that the conjugate momentum p is conserved. Consider, for example, the
motion of a particle of mass m near the surface of the earth. Let (x, y) be coordinates
parallel to the surface and z the height. We then have
(6.16)
T = 21 m x 2 + y 2 + z 2
U = mgz
L=T U =
Since
Fx =
1
2m
x + y + z
L
= 0 and
x
(6.17)
mgz .
(6.18)
L
=0,
y
(6.19)
L
= my .
y
(6.20)
Fy =
we have that px and py are conserved, with
px =
L
= mx ,
x
py =
These first order equations can be integrated to yield
x(t) = x(0) +
px
t ,
m
y(t) = y(0) +
py
t.
m
(6.21)
6.3. CONSERVED QUANTITIES
83
The z equation is of course
p z = m
z = mg = Fz ,
(6.22)
z(t) = z(0) + z(0)
t 21 gt2 .
(6.23)
with solution
As another example, consider
moving in the (x, y) plane under the influence of
p a particle
a potential U (x, y) = U
x2 + y 2 which depends only on the particles distance from
p
the origin = x2 + y 2 . The Lagrangian, expressed in two-dimensional polar coordinates
(, ), is
L = 21 m 2 + 2 2 U () .
(6.24)
We see that L is cyclic in the angle , hence
p =
L
= m2
(6.25)
is conserved. p is the angular momentum of the particle about the z axis. In the language
of the calculus of variations, momentum conservation is what follows when the integrand of
a functional is independent of the independent variable.
6.3.2
Energy conservation
When the integrand of a functional is independent of the dependent variable, another conservation law follows. For Lagrangian mechanics, consider the expression
H(q, q,
t) =
n
X
=1
p q L .
Now we take the total time derivative of H:
n
X
L
L
L
dH
=
q
q
.
p q + p q
dt
q
q
t
(6.26)
(6.27)
=1
We evaluate H along the motion of the system, which entails that the terms in the curly
brackets above cancel for each :
p =
Thus, we find
L
q
p =
L
.
q
(6.28)
L
dH
=
,
(6.29)
dt
t
which means that H is conserved whenever the Lagrangian contains no explicit time dependence. For a Lagrangian of the form
X
2
1
(6.30)
L=
2 ma r a U (r1 , . . . , rN ) ,
a
84
CHAPTER 6. LAGRANGIAN MECHANICS
we have that pa = ma r a , and
H = T +U =
1
2
ma r a2 + U (r1 , . . . , rN ) .
(6.31)
However, it is not always the case that H = T + U is the total energy, as we shall see in
the next chapter.
6.4
Choosing Generalized Coordinates
Any choice of generalized coordinates will yield an equivalent set of equations of motion.
However, some choices result in an apparently simpler set than others. This is often true
with respect to the form of the potential energy. Additionally, certain constraints that may
be present are more amenable to treatment using a particular set of generalized coordinates.
The kinetic energy T is always simple to write in Cartesian coordinates, and it is good
practice, at least when one is first learning the method, to write T in Cartesian coordinates
and then convert to generalized coordinates. In Cartesian coordinates, the kinetic energy
of a single particle of mass m is
(6.32)
T = 21 m x 2 + y 2 + x 2 .
If the motion is two-dimensional,
and confined to the plane z = const., one of course has
T = 12 m x 2 + y 2 .
Two other commonly used coordinate systems are the cylindrical and spherical systems. In
cylindrical coordinates (, , z), is the radial coordinate in the (x, y) plane and is the
azimuthal angle:
x = cos sin
y = sin + cos ,
x = cos
y = sin
and the third, orthogonal coordinate is of course z. The kinetic energy is
T = 12 m x 2 + y 2 + x 2
= 1 m 2 + 2 2 + z 2 .
2
(6.33)
(6.34)
(6.35)
When the motion is confined to a plane with z = const., this coordinate system is often
referred to as two-dimensional polar coordinates.
In spherical coordinates (r, , ), r is the radius, is the polar angle, and is the azimuthal
angle. On the globe, would be the colatitude, which is = 2 , where is the latitude.
I.e. = 0 at the north pole. In spherical polar coordinates,
x = r sin cos
y = r sin sin
z = r cos
x = sin cos r + r cos cos r sin sin
y = sin sin r + r cos sin + r sin cos
z = cos r r sin .
(6.36)
(6.37)
(6.38)
6.5. HOW TO SOLVE MECHANICS PROBLEMS
85
The kinetic energy is
T = 12 m x 2 + y 2 + z 2
= 12 m r 2 + r 2 2 + r 2 sin2 2 .
6.5
(6.39)
How to Solve Mechanics Problems
Here are some simple steps you can follow toward obtaining the equations of motion:
1. Choose a set of generalized coordinates {q1 , . . . , qn }.
2. Find the kinetic energy T (q, q,
t), the potential energy U (q, t), and the Lagrangian
L(q, q,
t) = T U . It is often helpful to first write the kinetic energy in Cartesian
coordinates for each particle before converting to generalized coordinates.
3. Find the canonical momenta p =
L
q
and the generalized forces F =
L
q .
4. Evaluate the time derivatives p and write the equations of motion p = F . Be careful
to differentiate properly, using the chain rule and the Leibniz rule where appropriate.
5. Identify any conserved quantities (more about this later).
6.6
6.6.1
Examples
One-dimensional motion
For a one-dimensional mechanical system with potential energy U (x),
L = T U = 21 mx 2 U (x) .
The canonical momentum is
p=
and the equation of motion is
L
d L
=
dt x
x
L
= mx
x
m
x = U (x) ,
(6.40)
(6.41)
(6.42)
which is of course F = ma.
Note that we can multiply the equation of motion by x to get
o dE
n
o
d n1
2
,
m
x
+
U
(x)
=
0 = x m
x + U (x) =
dt 2
dt
where E = T + U .
(6.43)
86
6.6.2
CHAPTER 6. LAGRANGIAN MECHANICS
Central force in two dimensions
Consider next a particle of mass m moving in two dimensions under p
the influence of a
potential U () which is a function of the distance from the origin = x2 + y 2 . Clearly
cylindrical (2d polar) coordinates are called for:
The equations of motion are
L = 12 m 2 + 2 2 U () .
d L
=
dt
d L
=
dt
m
= m 2 U ()
d
m2 = 0 .
dt
(6.44)
(6.45)
(6.46)
Note that the canonical momentum conjugate to , which is to say the angular momentum,
is conserved:
p = m2 = const.
(6.47)
We can use this to eliminate from the first Euler-Lagrange equation, obtaining
m
=
p2
m3
U () .
(6.48)
We can also write the total energy as
E = 12 m 2 + 2 2 + U ()
= 21 m 2 +
p2
2m2
+ U () ,
(6.49)
from which it may be shown that E is also a constant:
p2
dE
= m
+
U
()
= 0 .
dt
m3
(6.50)
We shall discuss this case in much greater detail in the coming weeks.
6.6.3
A sliding point mass on a sliding wedge
Consider the situation depicted in Fig. 6.1, in which a point object of mass m slides
frictionlessly along a wedge of opening angle . The wedge itself slides frictionlessly along a
horizontal surface, and its mass is M . We choose as generalized coordinates the horizontal
position X of the left corner of the wedge, and the horizontal distance x from the left corner
to the sliding point mass. The vertical coordinate of the sliding mass is then y = x tan ,
6.6. EXAMPLES
87
Figure 6.1: A wedge of mass M and opening angle slides frictionlessly along a horizontal
surface, while a small object of mass m slides frictionlessly along the wedge.
where the horizontal surface lies at y = 0. With these generalized coordinates, the kinetic
energy is
2 + 12 my 2
T = 12 M X 2 + 21 m (X + x)
= 21 (M + m)X 2 + mX x + 21 m (1 + tan2 ) x 2 .
(6.51)
The potential energy is simply
U = mgy = mg x tan .
(6.52)
L = 12 (M + m)X 2 + mX x + 21 m (1 + tan2 ) x 2 mg x tan ,
(6.53)
Thus, the Lagrangian is
and the equations of motion are
L
d L
=
dt X
X
L
d L
=
dt x
x
+ mx
(M + m)X
=0
+ m (1 + tan2 ) x = mg tan .
mX
(6.54)
At this point we can use the first of these equations to write
=
X
m
x
.
M +m
(6.55)
Substituting this into the second equation, we obtain the constant accelerations
x
=
(M + m)g sin cos
M + m sin2
= mg sin cos .
X
M + m sin2
(6.56)
88
CHAPTER 6. LAGRANGIAN MECHANICS
Figure 6.2: The springpendulum system.
6.6.4
A pendulum attached to a mass on a spring
Consider next the system depicted in Fig. 6.2 in which a mass M moves horizontally while
attached to a spring of spring constant k. Hanging from this mass is a pendulum of arm
length and bob mass m.
A convenient set of generalized coordinates is (x, ), where x is the displacement of the mass
M relative to the equilibrium extension a of the spring, and is the angle the pendulum
arm makes with respect to the vertical. Let the Cartesian coordinates of the pendulum bob
be (x1 , y1 ). Then
x1 = a + x + sin
y1 = l cos .
(6.57)
The kinetic energy is
T = 12 M x 2 + 12 m (x 2 + y 2 )
i
h
2 + ( sin )
2
= 12 M x 2 + 12 m (x + cos )
= 12 (M + m) x 2 + 12 m2 2 + m cos x ,
(6.58)
and the potential energy is
U = 21 kx2 + mgy1
= 12 kx2 mg cos .
(6.59)
Thus,
L = 12 (M + m) x 2 + 12 m2 2 + m cos x 21 kx2 + mg cos .
(6.60)
6.6. EXAMPLES
89
The canonical momenta are
px =
L
= (M + m) x + m cos
x
p =
L
= m cos x + m2 ,
(6.61)
and the canonical forces are
Fx =
L
= kx
x
F =
L
= m sin x mg sin .
(6.62)
The equations of motion then yield
(M + m) x
+ m cos m sin 2 = kx
m cos x
+ m2 = mg sin .
(6.63)
(6.64)
Small Oscillations : If we assume both x and are small, we may write sin and
cos 1, in which case the equations of motion may be linearized to
(M + m) x
+ m + kx = 0
m x
+ m2 + mg = 0 .
If we define
u
m
M
02
k
M
(6.65)
(6.66)
12
g
,
(6.67)
then
(1 + ) u
+ + 02 u = 0
u
+ + 12 = 0 .
We can solve by writing
in which case
u(t)
(t)
a
eit ,
b
02 (1 + ) 2 2
2
12 2
a
0
=
.
b
0
(6.68)
(6.69)
(6.70)
(6.71)
In order to have a nontrivial solution (i.e. without a = b = 0), the determinant of the above
2 2 matrix must vanish. This gives a condition on 2 , with solutions
q
2
2
02 12 + 2 (02 + 12 ) 12 .
(6.72)
= 12 02 + (1 + ) 12 21
90
CHAPTER 6. LAGRANGIAN MECHANICS
Figure 6.3: The double pendulum, with generalized coordinates 1 and 2 . All motion is
confined to a single plane.
6.6.5
The double pendulum
As yet another example of the generalized coordinate approach to Lagrangian dynamics,
consider the double pendulum system, sketched in Fig. 6.3. We choose as generalized
coordinates the two angles 1 and 2 . In order to evaluate the Lagrangian, we must obtain
the kinetic and potential energies in terms of the generalized coordinates {1 , 2 } and their
corresponding velocities {1 , 2 }.
In Cartesian coordinates,
T = 12 m1 (x 21 + y12 ) + 12 m2 (x 22 + y 22 )
(6.73)
U = m1 g y 1 + m2 g y 2 .
(6.74)
We therefore express the Cartesian coordinates {x1 , y1 , x2 , y2 } in terms of the generalized
coordinates {1 , 2 }:
x1 = 1 sin 1
x2 = 1 sin 1 + 2 sin 2
(6.75)
y1 = 1 cos 1
y2 = 1 cos 1 2 cos 2 .
(6.76)
x 2 = 1 1 cos 1 + 2 2 cos 2
y 2 = 1 1 sin 1 + 2 2 sin 2 .
(6.77)
Thus, the velocities are
x 1 = 1 1 cos 1
y 1 = 1 1 sin 1
(6.78)
Thus,
o
n
T = 12 m1 21 12 + 12 m2 21 12 + 21 2 cos(1 2 ) 1 2 + 22 22
(6.79)
U = m1 g 1 cos 1 m2 g 1 cos 1 m2 g 2 cos 2 ,
(6.80)
6.6. EXAMPLES
91
and
L = T U = 12 (m1 + m2 ) 21 12 + m2 1 2 cos(1 2 ) 1 2 + 12 m2 22 22
+ (m1 + m2 ) g 1 cos 1 + m2 g 2 cos 2 .
(6.81)
The generalized (canonical) momenta are
p1 =
L
= (m1 + m2 ) 21 1 + m2 1 2 cos(1 2 ) 2
1
(6.82)
p2 =
L
= m2 1 2 cos(1 2 ) 1 + m2 22 2 ,
(6.83)
and the equations of motion are
p 1 = (m1 + m2 ) 21 1 + m2 1 2 cos(1 2 ) 2 m2 1 2 sin(1 2 ) (1 2 ) 2
= (m1 + m2 ) g 1 sin 1 m2 1 2 sin(1 2 ) 1 2 =
L
1
(6.84)
and
p2 = m2 1 2 cos(1 2 ) 1 m2 1 2 sin(1 2 ) (1 2 ) 1 + m2 22 2
L
.
= m2 g 2 sin 2 + m2 1 2 sin(1 2 ) 1 2 =
2
(6.85)
We therefore find
1 1 +
m 2 2
m 2 2
cos(1 2 ) 2 +
sin(1 2 ) 22 + g sin 1 = 0
m1 + m2
m1 + m2
1 cos(1 2 ) 1 + 2 2 1 sin(1 2 ) 12 + g sin 2 = 0 .
(6.86)
(6.87)
Small Oscillations : The equations of motion are coupled, nonlinear second order ODEs.
When the system is close to equilibrium, the amplitudes of the motion are small, and we
may expand in powers of the 1 and 2 . The linearized equations of motion are then
1 + 2 + 02 1 = 0
(6.88)
1 + 2 + 02 2 = 0 ,
(6.89)
where we have defined
m2
m1 + m2
2
1
02
g
.
1
(6.90)
92
CHAPTER 6. LAGRANGIAN MECHANICS
We can solve this coupled set of equations by a nifty trick. Lets take a linear combination
of the first equation plus an undetermined coefficient, r, times the second:
(1 + r) 1 + ( + r) 2 + 02 (1 + r 2 ) = 0 .
(6.91)
We now demand that the ratio of the coefficients of 2 and 1 is the same as the ratio of
the coefficients of 2 and 1 :
( + r)
=r
1+r
r = 12 ( 1)
When r = r , the equation of motion may be written
1
2
(1 )2 + 4
d2
02
1 + r 2
1 + r 2 =
2
dt
1 + r
(6.92)
(6.93)
and defining the (unnormalized) normal modes
1 + r 2 ,
we find
(6.94)
2
+
= 0 ,
with
= p
(6.95)
0
.
1 + r
(6.96)
Thus, by switching to the normal coordinates, we decoupled the equations of motion, and
identified the two normal frequencies of oscillation. We shall have much more to say about
small oscillations further below.
For example, with 1 = 2 = and m1 = m2 = m, we have = 21 , and = 1, in which case
r =
12
= 1
1
2
g
.
(6.97)
Note that the oscillation frequency for the in-phase mode + is low, and that for the out
of phase mode is high.
6.6.6
The thingy
Four massless rods of length L are hinged together at their ends to form a rhombus. A
particle of mass M is attached to each vertex. The opposite corners are joined by springs
of spring constant k. In the square configuration, the strings are unstretched. The motion
is confined to a plane, and the particles move only along the diagonals of the rhombus.
Introduce suitable generalized coordinates and find the Lagrangian of the system. Deduce
the equations of motion and find the frequency of small oscillations about equilibrium.
6.6. EXAMPLES
93
Solution
The rhombus is depicted in figure 6.4. Let a be the equilibrium length of the springs; clearly
L = a2 . Let be half of one of the opening angles, as shown. Then the masses are located
at (X, 0) and (0, Y ), with X = a2 cos and Y = a2 sin . The spring extensions are
X = 2X a and Y = 2Y a. The kinetic and potential energies are therefore
Figure 6.4: The thingy: a rhombus with opening angles 2 and 2.
T = M X 2 + Y 2
= 1 M a2 2
2
and
2
2
U = 21 k X + 12 k Y
n
2
2 o
= 12 ka2
2 cos 1 +
2 sin 1
o
n
= 12 ka2 3 2 2(cos + sin ) .
Note that minimizing U () gives sin = cos , i.e. eq = 4 . The Lagrangian is then
L = T U = 12 M a2 2 +
2 ka2 cos + sin + const.
94
CHAPTER 6. LAGRANGIAN MECHANICS
The equations of motion are
d L
L
=
dt
M a2 =
2 ka2 (cos sin )
Its always smart to expand about equilibrium, so lets write =
+ , which leads to
+ 02 sin = 0 ,
p
with 0 = 2k/M . This is the equation of a pendulum! Linearizing gives + 02 = 0, so
the small oscillation frequency is just 0 .
6.7
Appendix : Virial Theorem
The virial theorem is a statement about the time-averaged motion of a mechanical system.
Define the virial,
X
G(q, p) =
p q .
(6.98)
Then
dG X
p q + p q
=
dt
X
X
L
=
q F +
q
.
q
(6.99)
P
Now suppose that T = 12 , T q q is homogeneous of degree k = 2 in q,
and that U
is homogeneous of degree zero in q.
Then
X
X
T
L
=
q
= 2 T,
q
q
(6.100)
which follows from Eulers theorem on homogeneous functions.
Now consider the time average of G over a period :
D dG E
dt
Z
1
dG
=
dt
dt
0
i
1h
G( ) G(0) .
=
(6.101)
= 0. Any bounded motion,
If G(t) is bounded, then in the limit we must have hGi
= 0. But then
such as the orbit of the earth around the Sun, will result in hGi
D dG E
dt
= 2 hT i +
q F = 0 ,
(6.102)
6.7. APPENDIX : VIRIAL THEOREM
95
which implies
hT i =
12
1
2
DX
D X
i
q F = +
X
1
2
U
q
q
E
ri i U r1 , . . . , rN
= 12 k hU i ,
(6.103)
(6.104)
where the last line pertains to homogeneous potentials of degree k. Finally, since T +U = E
is conserved, we have
kE
2E
hT i =
, hU i =
.
(6.105)
k+2
k+2
96
CHAPTER 6. LAGRANGIAN MECHANICS
Chapter 7
Noethers Theorem
7.1
Continuous Symmetry Implies Conserved Charges
Consider a particle moving in two dimensions under the influence of an external potential
U (r). The potential is a function only of the magnitude of the vector r. The Lagrangian is
then
(7.1)
L = T U = 1 m r 2 + r 2 2 U (r) ,
2
where we have chosen generalized coordinates (r, ). The momentum conjugate to is
The generalized force F clearly vanishes, since L does not depend on the
p = m r 2 .
coordinate . (One says that L is cyclic in .) Thus, although r = r(t) and = (t)
will in general be time-dependent, the combination p = m r 2 is constant. This is the
conserved angular momentum about the z axis.
If instead the particle moved in a potential U (y), independent of x, then writing
L = 12 m x 2 + y 2 U (y) ,
(7.2)
we have that the momentum px = L/ x = mx is conserved, because the generalized force
Fx = L/x = 0 vanishes. This situation pertains in a uniform gravitational field, with
U (x, y) = mgy, independent of x. The horizontal component of momentum is conserved.
In general, whenever the system exhibits a continuous symmetry, there is an associated
conserved charge. (The terminology charge is from field theory.) Indeed, this is a rigorous
result, known as Noethers Theorem. Consider a one-parameter family of transformations,
q q (q, ) ,
(7.3)
where is the continuous parameter. Suppose further (without loss of generality) that at
= 0 this transformation is the identity, i.e. q (q, 0) = q . The transformation may be
nonlinear in the generalized coordinates. Suppose further that the Lagrangian L is invariant
97
98
CHAPTER 7. NOETHERS THEOREM
under the replacement q q. Then we must have
d
L
L
0=
q , q, t) =
+
L(
d
q
q
=0
=0
=0
L d q
d L q
+
=
dt q
q dt =0
=0
d L q
=
.
dt q =0
(7.4)
Thus, there is an associated conserved charge
L q
=
q
(7.5)
=0
7.1.1
Examples of one-parameter families of transformations
Consider the Lagrangian
L = 21 m(x 2 + y 2 ) U
In two-dimensional polar coordinates, we have
x2 + y 2 .
(7.6)
L = 12 m(r 2 + r 2 2 ) U (r) ,
(7.7)
r() = r
()
=+ .
(7.8)
and we may now define
(7.9)
Note that r(0) = r and (0)
= , i.e. the transformation is the identity when = 0. We
now have
X L q
r
L
L
=
=
+
= mr 2 .
(7.10)
=0
=0
=0
Another way to derive the same result which is somewhat instructive is to work out the
transformation in Cartesian coordinates. We then have
x
() = x cos y sin
y() = x sin + y cos .
(7.11)
(7.12)
Thus,
x
=
y ,
y
=x
(7.13)
7.2. CONSERVATION OF LINEAR AND ANGULAR MOMENTUM
and
L x
=
x
But
=0
L y
+
y
=0
= m(xy y x)
.
m(xy y x)
= mz r r = mr 2 .
99
(7.14)
(7.15)
As another example, consider the potential
U (, , z) = V (, a + z) ,
(7.16)
where (, , z) are cylindrical coordinates for a particle of mass m, and where a is a constant
with dimensions of length. The Lagrangian is
2
2 2
2
1
V (, a + z) .
(7.17)
2 m + + z
This model possesses a helical symmetry, with a one-parameter family
() =
()
=+
Note that
(7.18)
(7.19)
z() = z a .
(7.20)
a + z = a + z ,
(7.21)
so the potential energy, and the Lagrangian as well, is invariant under this one-parameter
family of transformations. The conserved charge for this symmetry is
L
L z
L
+
+
= m2 maz .
(7.22)
=
=0
=0
=0
We can check explicitly that is conserved, using the equations of motion
L
V
d
d L
= a
m2 =
=
dt
dt
z
d L
L
V
d
=
=
.
= (mz)
dt z
dt
z
z
Thus,
7.2
d
d
=0.
m2 a (mz)
=
dt
dt
(7.23)
(7.24)
(7.25)
Conservation of Linear and Angular Momentum
Suppose that the Lagrangian of a mechanical system is invariant under a uniform translation
direction. Then our one-parameter family of transformations is given
of all particles in the n
by
a = xa + n
,
x
(7.26)
100
CHAPTER 7. NOETHERS THEOREM
and the associated conserved Noether charge is
=
where P =
X L
=n
P ,
n
x
a
a
(7.27)
pa is the total momentum of the system.
then
If the Lagrangian of a mechanical system is invariant under rotations about an axis n,
a = R(, n)
xa
x
xa + O( 2 ) ,
= xa + n
(7.28)
in powers of . The conserved Noether
where we have expanded the rotation matrix R(, n)
charge associated with this symmetry is
=
X
X L
xa = n
L ,
n
xa p a = n
x a
a
a
(7.29)
where L is the total angular momentum of the system.
7.3
Advanced Discussion : Invariance of L vs. Invariance of
S
Observant readers might object that demanding invariance of L is too strict. We should
instead be demanding invariance of the action S 1 . Suppose S is invariant under
t t(q, t, )
q (t) q (q, t, ) .
(7.30)
(7.31)
Then invariance of S means
S=
Ztb
dt L(q, q,
t) =
Ztb
dt L(
q , q, t) .
(7.32)
ta
ta
Note that t is a dummy variable of integration, so it doesnt matter whether we call it t
or t. The endpoints of the integral, however, do change under the transformation.
Now
consider an infinitesimal transformation, for which t = t t and q = q t q(t) are both
small. Thus,
S=
Ztb
ta
1
dt L(q, q,
t) =
tbZ+tb
o
n
L
L
q +
q + . . . ,
dt L(q, q,
t) +
q
q
ta +ta
Indeed, we should be demanding that S only change by a function of the endpoint values.
(7.33)
7.3. ADVANCED DISCUSSION : INVARIANCE OF L VS. INVARIANCE OF S
101
where
(t) q (t) q (t)
q
= q t q t + q (t) q (t)
= q q t + O(q t)
(7.34)
Subtracting eqn. 7.33 from eqn. 7.32, we obtain
tb +tb (
)
Z
L
L
L
d L
(t)
dt
q
q +
q
0 = Lb tb La ta +
q b ,b q a ,a
q
dt q
ta +ta
tb
ta
d
dt
dt
(
)
L
L
,
q t +
q
L
q
q
(7.35)
where La,b is L(q, q,
t) evaluated at t = ta,b . Thus, if is infinitesimal, and
t = A(q, t)
q = B (q, t) ,
(7.36)
(7.37)
then the conserved charge is
=
L
L
L
q A(q, t) +
B (q, t)
q
q
= H(q, p, t) A(q, t) + p B (q, t) .
(7.38)
Thus, when A = 0, we recover our earlier results, obtained by assuming invariance of L.
Note that conservation of H follows from time translation invariance: t t + , for which
A = 1 and B = 0. Here we have written
H = p q L ,
(7.39)
and expressed it in terms of the momenta p , the coordinates q , and time t. H is called
the Hamiltonian.
7.3.1
The Hamiltonian
The Lagrangian is a function of generalized coordinates, velocities, and time. The canonical
momentum conjugate to the generalized coordinate q is
p =
L
.
q
(7.40)
102
CHAPTER 7. NOETHERS THEOREM
The Hamiltonian is a function of coordinates, momenta, and time. It is defined as the
Legendre transform of L:
X
H(q, p, t) =
p q L .
(7.41)
Lets examine the differential of H:
X
L
L
L
dH =
q dp + p dq
dt
dq
dq
q
q
t
X
L
L
dq
dt ,
=
q dp
q
t
(7.42)
where we have invoked the definition of p to cancel the coefficients of dq . Since p =
L/q , we have Hamiltons equations of motion,
q =
Thus, we can write
dH =
Dividing by dt, we obtain
H
p
p =
H
.
q
L
X
dt .
q dp p dq
t
(7.43)
(7.44)
L
dH
=
,
(7.45)
dt
t
which says that the Hamiltonian is conserved (i.e. it does not change with time) whenever
there is no explicit time dependence to L.
Example #1 : For a simple d = 1 system with L = 21 mx 2 U (x), we have p = mx and
H = p x L = 12 mx 2 + U (x) =
p2
+ U (x) .
2m
(7.46)
Example #2 : Consider now the mass point wedge system analyzed above, with
L = 12 (M + m)X 2 + mX x + 21 m (1 + tan2 ) x 2 mg x tan ,
(7.47)
The canonical momenta are
P =
p=
L
= (M + m) X + mx
X
(7.48)
L
= mX + m (1 + tan2 ) x .
x
(7.49)
The Hamiltonian is given by
H = P X + p x L
= 12 (M + m)X 2 + mX x + 21 m (1 + tan2 ) x 2 + mg x tan .
(7.50)
7.3. ADVANCED DISCUSSION : INVARIANCE OF L VS. INVARIANCE OF S
103
However, this is not quite H, since H = H(X, x, P, p, t) must be expressed in terms of the
coordinates and the momenta and not the coordinates and velocities. So we must eliminate
X and x in favor of P and p. We do this by inverting the relations
P
M +m
m
X
=
(7.51)
2
p
m
m (1 + tan )
x
to obtain
1
m (1 + tan2 )
m
P
X
.
=
2
m
M +m
p
x
m M + (M + m) tan
(7.52)
Substituting into 7.50, we obtain
H=
P p cos2
p2
M + m P 2 cos2
+
+ mg x tan .
2m M + m sin2 M + m sin2 2 (M + m sin2 )
(7.53)
L
Notice that P = 0 since X
= 0. P is the total horizontal momentum of the system (wedge
plus particle) and it is conserved.
7.3.2
Is H = T + U ?
The most general form of the kinetic energy is
T = T2 + T1 + T0
(2)
= 21 T (q, t) q q + T(1) (q, t) q + T (0) (q, t) ,
(7.54)
where T (n) (q, q,
t) is homogeneous of degree n in the velocities2 . We assume a potential
energy of the form
U = U1 + U0
= U(1) (q, t) q + U (0) (q, t) ,
(7.55)
which allows for velocity-dependent forces, as we have with charged particles moving in an
electromagnetic field. The Lagrangian is then
(2)
L = T U = 12 T (q, t) q q + T(1) (q, t) q + T (0) (q, t) U(1) (q, t) q U (0) (q, t) . (7.56)
The canonical momentum conjugate to q is
p =
L
(2)
= T q + T(1) (q, t) U(1) (q, t)
q
(7.57)
which is inverted to give
(2) 1
q = T
(1)
(1)
p T + U .
(7.58)
2
A homogeneous
of degree k satisfies f (x1 , . . . , xn ) = k f (x1 , . . . , xn ). It is then easy to prove
P function
f
Eulers theorem, n
x
i=1 i xi = kf .
104
CHAPTER 7. NOETHERS THEOREM
The Hamiltonian is then
H = p q L
=
1
2
(2) 1
p T(1) + U(1)
= T2 T0 + U0 .
(1)
(1)
p T + U
T0 + U0
(7.59)
(7.60)
If T0 , T1 , and U1 vanish, i.e. if T (q, q,
t) is a homogeneous function of degree two in the
generalized velocities, and U (q, t) is velocity-independent, then H = T + U . But if T0 or T1
is nonzero, or the potential is velocity-dependent, then H 6= T + U .
7.3.3
Example: A bead on a rotating hoop
Consider a bead of mass m constrained to move along a hoop of radius a. The hoop is
further constrained to rotate with angular velocity = about the z-axis,
as shown in
Fig. 7.1.
The most convenient set of generalized coordinates is spherical polar (r, , ), in which case
T = 12 m r 2 + r 2 2 + r 2 sin2 2
(7.61)
= 12 ma2 2 + 2 sin2 .
Thus, T2 = 21 ma2 2 and T0 = 12 ma2 2 sin2 . The potential energy is U () = mga(1cos ).
and thus
The momentum conjugate to is p = ma2 ,
H(, p) = T2 T0 + U
= 1 ma2 2 1 ma2 2 sin2 + mga(1 cos )
2
p2
=
1 ma2 2 sin2 + mga(1 cos ) .
2ma2 2
(7.62)
For this problem, we can define the effective potential
Ueff () U T0 = mga(1 cos ) 12 ma2 2 sin2
2
= mga 1 cos 2 sin2 ,
20
(7.63)
where 02 g/a. The Lagrangian may then be written
L = 12 ma2 2 Ueff () ,
(7.64)
and thus the equations of motion are
ma2 =
Ueff
.
(7.65)
7.3. ADVANCED DISCUSSION : INVARIANCE OF L VS. INVARIANCE OF S
105
Figure 7.1: A bead of mass m on a rotating hoop of radius a.
() = 0, which gives
Equilibrium is achieved when Ueff
n
o
2
Ueff
= mga sin 1 2 cos = 0 ,
(7.66)
i.e. = 0, = , or = cos1 (02 / 2 ), where the last pair of equilibria are present
only for 2 > 02 . The stability of these equilibria is assessed by examining the sign of
( ). We have
Ueff
n
o
2
(7.67)
Ueff
() = mga cos 2 2 cos2 1 .
0
Thus,
2
at = 0
mga
1
2
Ueff ( ) = mga 1 + 2
at =
2
= cos1 0
mga 22 02
at
(7.68)
< 02 but
2 > 02 .
Thus,
= 0 is stable for
becomes unstable when the rotation frequency
is sufficiently large, i.e. when
In this regime, there are two new equilibria, at
= cos1 (02 / 2 ), which are both stable. The equilibrium at = is always unstable,
independent of the value of . The situation is depicted in Fig. 7.2.
106
CHAPTER 7. NOETHERS THEOREM
2
2
Figure 7.2: The effective potential Ueff () = mga 1 cos 2
2 sin . (The dimensionless
0
eff (x) = Ueff /mga is shown, where x = /.) Left panels: = 1 3 0 . Right
potential U
2
panels: = 3 0 .
7.4
Charged Particle in a Magnetic Field
Consider next the case of a charged particle moving in the presence of an electromagnetic
field. The particles potential energy is
= q (r, t)
U (r, r)
q
A(r, t) r ,
c
(7.69)
which is velocity-dependent. The kinetic energy is T = 12 m r 2 , as usual. Here (r) is the
scalar potential and A(r) the vector potential. The electric and magnetic fields are given
by
E =
1 A
c t
B =A .
(7.70)
The canonical momentum is
p=
q
L
= m r + A ,
r
c
(7.71)
7.4. CHARGED PARTICLE IN A MAGNETIC FIELD
107
and hence the Hamiltonian is
H(r, p, t) = p r L
q
q
= mr 2 + A r 21 m r 2 A r + q
c
c
= 12 m r 2 + q
2
1
q
=
p A(r, t) + q (r, t) .
2m
c
(7.72)
If A and are time-independent, then H(r, p) is conserved.
Lets work out the equations of motion. We have
!
L
d L
=
dt r
r
(7.73)
which gives
q
q dA
,
= q + (A r)
c dt
c
(7.74)
q Ai
q Aj
q Ai
x +
= q
+
x ,
c xj j c t
xi
c xi j
(7.75)
m r +
or, in component notation,
mx
i +
which is to say
q Ai q
mx
i = q
+
xi
c t
c
Ai
Aj
xi
xj
x j .
(7.76)
It is convenient to express the cross product in terms of the completely antisymmetric tensor
of rank three, ijk :
Ak
,
(7.77)
Bi = ijk
xj
and using the result
ijk imn = jm kn jn km ,
(7.78)
we have ijk Bi = j Ak k Aj , and
mx
i = q
q Ai q
+ ijk x j Bk ,
xi
c t
c
(7.79)
or, in vector notation,
q A q
m r = q
+ r ( A)
c t
c
q
= q E + r B ,
c
which is, of course, the Lorentz force law.
(7.80)
108
7.5
CHAPTER 7. NOETHERS THEOREM
Fast Perturbations : Rapidly Oscillating Fields
Consider a free particle moving under the influence of an oscillating force,
m
q = F sin t .
(7.81)
The motion of the system is then
q(t) = qh (t)
F sin t
,
m 2
(7.82)
where qh (t) = A + Bt is the solution to the homogeneous (unforced) equation of motion.
Note that the amplitude of the response q qh goes as 2 and is therefore small when
is large.
Now consider a general n = 1 system, with
H(q, p, t) = H0 (q, p) + V (q) sin(t + ) .
(7.83)
We assume that is much greater than any natural oscillation frequency associated with
H0 . We separate the motion q(t) and p(t) into slow and fast components:
q(t) = q(t) + (t)
(7.84)
p(t) = p(t) + (t) ,
(7.85)
where (t) and (t) oscillate with the driving frequency . Since and will be small, we
expand Hamiltons equations in these quantities:
1 3H0 2
3H0
2H0
1 3H0 2
H0 2H0
+
+
+ ...
q + =
p
p2
q p
2 q2 p
q p2
2 p3
H0 2H0
2H0
1 3H0 2
3H0
1 3H0 2
p + =
q
q2
q p
2 q3
q2 p
2 q p2
2V
V
sin(t + ) 2 sin(t + ) . . . .
q
q
(7.86)
(7.87)
We now average over the fast degrees of freedom to obtain an equation of motion for the slow
variables q and p, which we here carry to lowest nontrivial order in averages of fluctuating
quantities:
H0 1 3H0
2
3H0
1 3H0
2
+
+
(7.88)
+
2
p
2 q p
q p2
2 p3
H0 1 3H0
2
3H0
1 3H0
2 2V
p =
2 sin(t + ) . (7.89)
3
2
2
q
2 q
q p
2 q p
q
q =
The fast degrees of freedom obey
2H0
2H0
+
=
q p
p2
=
2H0
2H0
V
sin(t + ) .
2
q
q p
q
(7.90)
(7.91)
7.5. FAST PERTURBATIONS : RAPIDLY OSCILLATING FIELDS
109
Let us analyze the coupled equations3
= A + B
(7.92)
it
= C A + F e
The solution is of the form
(7.93)
=
eit .
(7.94)
Plugging in, we find
=
BF
BF
= 2 + O 4
2
2
BC A
(7.95)
(A + i)F
iF
3
+
O
=
.
BC A2 2
(7.96)
Taking the real part, and restoring the phase shift , we have
(t) =
BF
1 V 2H0
sin(t
+
)
=
sin(t + )
2
2 q p2
(t) =
1 V
F
cos(t + ) =
cos(t + ) .
(7.97)
(7.98)
The desired averages, to lowest order, are thus
along with = 0.
2
1
V 2 2H0 2
=
2 4 q
p2
2
1
V 2
=
2 2 q
1 V 2H0
sin(t + ) =
,
2 2 q p2
(7.99)
(7.100)
(7.101)
Finally, we substitute the averages into the equations of motion for the slow variables q and
p, resulting in the time-independent effective Hamiltonian
1 2H0
K(
q , p) = H0 (
q , p) + 2
4 p2
V
q
2
(7.102)
and the equations of motion
q =
K
p
p =
K
.
q
(7.103)
3
With real coefficients A, B, and C, one can always take the real part to recover the fast variable equations
of motion.
110
7.5.1
CHAPTER 7. NOETHERS THEOREM
Example : pendulum with oscillating support
Consider a pendulum with a vertically oscillating point of support. The coordinates of the
pendulum bob are
x = sin , y = a(t) cos .
(7.104)
The Lagrangian is easily obtained:
L = 12 m2 2 + ma sin + mg cos + 21 ma 2 mga
(7.105)
these may be dropped
}|
{
d
= 21 m2 2 + m(g + a
) cos + 12 ma 2 mga
ma sin .
dt
z
(7.106)
Thus we may take the Lagrangian to be
= 1 m2 2 + m(g + a
L
) cos ,
2
(7.107)
from which we derive the Hamiltonian
H(, p , t) =
p2
mg cos m
a cos
2m2
= H0 (, p , t) + V1 () sin t .
(7.108)
(7.109)
We have assumed a(t) = a0 sin t, so
V1 () = ma0 2 cos .
(7.110)
The effective Hamiltonian, per eqn. 7.102, is
p ) =
K(,
p
mg cos + 14 m a20 2 sin2 .
2m2
(7.111)
Lets define the dimensionless parameter
2g
.
2 a20
(7.112)
= mg v(),
with
The slow variable executes motion in the effective potential Veff ()
= cos + 1 sin2 .
v()
2
(7.113)
Differentiating, and dropping the bar on , we find that Veff () is stationary when
v () = 0
sin cos = sin .
(7.114)
Thus, = 0 and = , where sin = 0, are equilibria. When < 1 (note > 0 always),
there are two new solutions, given by the roots of cos = .
7.6. FIELD THEORY: SYSTEMS WITH SEVERAL INDEPENDENT VARIABLES 111
Figure 7.3: Dimensionless potential v() for = 1.5 (black curve) and = 0.5 (blue curve).
To assess stability of these equilibria, we compute the second derivative:
v () = cos +
1
cos 2 .
(7.115)
From this, we see that = 0 is stable (i.e. v ( = 0) > 0) always, but = is stable for
< 1 and unstable for > 1. When < 1, two new solutions appear, at cos = , for
which
1
(7.116)
v (cos1 ()) = ,
which is always negative since < 1 in order for these equilibria to exist. The situation is
sketched in fig. 7.3, showing v() for two representative values of the parameter . For > 1,
the equilibrium at = is unstable, but as decreases, a subcritical pitchfork bifurcation is
encountered at = 1, and = becomes stable, while the outlying = cos1 () solutions
are unstable.
7.6
Field Theory: Systems with Several Independent Variables
Suppose a (x) depends on several independent variables: {x1 , x2 , . . . , xn }. Furthermore,
suppose
Z
S {a (x) = dx L(a a , x) ,
(7.117)
112
CHAPTER 7. NOETHERS THEOREM
i.e. the Lagrangian density L is a function of the fields a and their partial derivatives
a /x . Here is a region in RK . Then the first variation of S is
(
)
Z
L
L a
+
S = dx
a a ( a ) x
(
)
Z
I
L
L
L
a ,
(7.118)
+ dx
= d n
( a ) a
a x ( a )
where is the (n 1)-dimensional boundary of ,
surface area, and
d is the differential
n is the unit normal. If we demand L/( a ) = 0 of a = 0, the surface term
vanishes, and we conclude
S
L
L
.
(7.119)
a (x)
a x ( a )
As an example, consider the case of a stretched string of linear mass density and tension
. The action is a functional of the height y(x, t), where the coordinate along the string, x,
and time, t, are the two independent variables. The Lagrangian density is
2
2
y
y
12
,
(7.120)
L = 21
t
x
whence the Euler-Lagrange equations are
L
L
S
=
0=
y(x, t)
x y
t y
=
2y
2y
,
x2
t2
(7.121)
y
where y = x
and y = y
y = y , which is the Helmholtz equation. Weve
t . Thus,
assumed boundary conditions where y(xa , t) = y(xb , t) = y(x, ta ) = y(x, tb ) = 0.
The Lagrangian density for an electromagnetic field with sources is
1
F F
L = 16
The equations of motion are then
L
L
=0
A x ( A )
1
c j A
F =
(7.122)
4
j ,
c
(7.123)
which are Maxwells equations.
Recall the result of Noethers theorem for mechanical systems:
!
d L q
=0,
dt q
=0
(7.124)
7.6. FIELD THEORY: SYSTEMS WITH SEVERAL INDEPENDENT VARIABLES 113
where q = q (q, ) is a one-parameter () family of transformations of the generalized
coordinates which leaves L invariant. We generalize to field theory by replacing
q (t) a (x, t) ,
(7.125)
where {a (x, t)} are a set of fields, which are functions of the independent variables {x, y, z, t}.
We will adopt covariant relativistic notation and write for four-vector x = (ct, x, y, z). The
generalization of d/dt = 0 is
a
L
( a )
=0,
(7.126)
=0
where there is an implied sum on both and a. We can write this as J = 0, where
a
L
J
( a )
(7.127)
=0
We call = J 0 /c the total charge. If we assume J = 0 at the spatial boundaries of our
system, then integrating the conservation law J over the spatial region gives
d
=
dt
d x 0 J =
d x J =
J =0 ,
d n
(7.128)
assuming J = 0 at the boundary .
As an example, consider the case of a complex scalar field, with Lagrangian density4
L(, , , ) = 12 K ( )( ) U .
(7.129)
This is invariant under the transformation ei , ei . Thus,
= i ei
= i ei ,
(7.130)
and, summing over both and fields, we have
L
L
(i) +
(i )
( )
( )
K
=
.
2i
J =
(7.131)
The potential, which depends on ||2 , is independent of . Hence, this form of conserved
4-current is valid for an entire class of potentials.
4
We raise and lower indices using the Minkowski metric g = diag (+, , , ).
114
CHAPTER 7. NOETHERS THEOREM
7.6.1
Gross-Pitaevskii model
As one final example of a field theory, consider the Gross-Pitaevskii model, with
L = i~
2
~2
g ||2 n0 .
t
2m
(7.132)
This describes a Bose fluid with repulsive short-ranged interactions. Here (x, t) is again
a complex scalar field, and is its complex conjugate. Using the Leibniz rule, we have
S[ , ] = S[ + , + ]
Z Z
~2
~2
d
= dt d x i~
+ i~
t
t
2m
2m
2
2g || n0 ( + )
(
Z Z
~2 2
d
= dt d x
i~
+
2g ||2 n0
t
2m
)
~2 2
2
+ i~
+
2g || n0
,
(7.133)
t
2m
where we have integrated by parts where necessary and discarded the boundary terms.
Extremizing S[ , ] therefore results in the nonlinear Schr
odinger equation (NLSE),
i~
~2 2
=
+ 2g ||2 n0
t
2m
(7.134)
as well as its complex conjugate,
i~
~2 2
=
+ 2g ||2 n0 .
t
2m
Note that these equations are indeed the Euler-Lagrange equations:
L
L
S
=
x
L
S
=
(7.135)
(7.136)
(7.137)
with x = (t, x)5 Plugging in
and
L
= 2g ||2 n0
L
= i~ 2g ||2 n0
L
= i~
t
L
~2
=
2m
(7.138)
L
=0
t
L
~2
=
,
2m
(7.139)
In the nonrelativistic case, there is no utility in defining x0 = ct, so we simply define x0 = t.
7.6. FIELD THEORY: SYSTEMS WITH SEVERAL INDEPENDENT VARIABLES 115
we recover the NLSE and its conjugate.
The Gross-Pitaevskii model also possesses a U(1) invariance, under
t) = ei (x, t)
(x, t) (x,
(x, t) (x, t) = ei (x, t) .
Thus, the conserved Noether current is then
L
L
J =
+
=0
=0
J 0 = ~ ||2
J =
(7.140)
~2
.
2im
(7.141)
(7.142)
Dividing out by ~, taking J 0 ~ and J ~j, we obtain the continuity equation,
+j =0 ,
t
(7.143)
where
~
.
2im
are the particle density and the particle current, respectively.
= ||2
j=
(7.144)
116
CHAPTER 7. NOETHERS THEOREM
Chapter 8
Constraints
A mechanical system of N point particles in d dimensions possesses n = dN degrees of freedom1 . To specify these degrees of freedom, we can choose any independent set of generalized
coordinates {q1 , . . . , qK }. Oftentimes, however, not all n coordinates are independent.
Consider, for example, the situation in Fig. 8.1, where a cylinder of radius a rolls over a halfcylinder of radius R. If there is no slippage, then the angles 1 and 2 are not independent,
and they obey the equation of constraint,
R 1 = a (2 1 ) .
(8.1)
In this case, we can easily solve the constraint equation and substitute 2 = 1 + Ra 1 . In
other cases, though, the equation of constraint might not be so easily solved (e.g. it may be
nonlinear). How then do we proceed?
8.1
Constraints and Variational Calculus
Before addressing the subject of constrained dynamical systems, lets consider the issue of
constraints in the broader context of variational calculus. Suppose we have a functional
Zxb
F [y(x)] = dx L(y, y , x) ,
(8.2)
xa
which we want to extremize subject to some constraints. Here y may stand for a set of
functions {y (x)}. There are two classes of constraints we will consider:
1
For N rigid bodies, the number of degrees of freedom is n = 21 d(d + 1)N , corresponding to d centerof-mass coordinates and 21 d(d 1) angles of orientation for each particle. The dimension of the group of
rotations in d dimensions is 21 d(d 1), corresponding to the number of parameters in a general rank-d
orthogonal matrix (i.e. an element of the group O(d)).
117
118
CHAPTER 8. CONSTRAINTS
Figure 8.1: A cylinder of radius a rolls along a half-cylinder of radius R. When there is no
slippage, the angles 1 and 2 obey the constraint equation R1 = a(2 1 ).
1. Integral constraints: These are of the form
Zxb
dx Nj (y, y , x) = Cj ,
(8.3)
xa
where j labels the constraint.
2. Holonomic constraints: These are of the form
Gj (y, x) = 0 .
(8.4)
The cylinders system in Fig. 8.1 provides an example of a holonomic constraint. There,
G(, t) = R 1 a (2 1 ) = 0. As an example of a problem with an integral constraint,
suppose we want to know the shape of a hanging rope of fixed length C. This means we
minimize the ropes potential energy,
Zxb
Zxb
q
U [y(x)] = g ds y(x) = g dx y 1 + y 2 ,
(8.5)
xa
xa
where is the linear mass density of the rope, subject to the fixed-length constraint
Zxb
Zxb q
C = ds = dx 1 + y 2 .
xa
(8.6)
xa
p
Note ds = dx2 + dy 2 is the differential element of arc length along the rope. To solve
problems like these, we turn to Lagranges method of undetermined multipliers.
8.2. CONSTRAINED EXTREMIZATION OF FUNCTIONS
8.2
119
Constrained Extremization of Functions
Given F (x1 , . . . , xn ) to be extremized subject to k constraints of the form Gj (x1 , . . . , xn ) = 0
where j = 1, . . . , k, construct
F
x1 , . . . , xn ; 1 , . . . , k F (x1 , . . . , xn ) +
k
X
j Gj (x1 , . . . , xn )
(8.7)
j=1
which is a function of the (n + k) variables x1 , . . . , xn ; 1 , . . . , k . Now freely extremize
the extended function F :
dF =
n
X
F
k
X
F
dx +
dj
x
j
j=1
n
k
k
X
X
X
G
F
j
=
j
Gj dj = 0
+
dx +
x
x
=1
=1
j=1
(8.8)
(8.9)
j=1
This results in the (n + k) equations
k
X Gj
F
j
+
=0
x
x
( = 1, . . . , n)
(8.10)
(j = 1, . . . , k) .
(8.11)
j=1
Gj = 0
The interpretation of all this is as follows. The n equations in 8.10 can be written in vector
form as
k
X
j Gj = 0 .
(8.12)
F +
j=1
This says that the (n-component) vector F is linearly dependent upon the k vectors
Gj . Thus, any movement in the direction of F must necessarily entail movement along
one or more of the directions Gj . This would require violating the constraints, since
movement along Gj takes us off the level set Gj = 0. Were F linearly independent of
the set {Gj }, this would mean that we could find a differential displacement dx which
has finite overlap with F but zero overlap with each Gj . Thus x + dx would still satisfy
Gj (x + dx) = 0, but F would change by the finite amount dF = F (x) dx.
8.3
Extremization of Functionals : Integral Constraints
Given a functional
F {y (x)} =
Zxb
dx L {y }, {y }, x
xa
( = 1, . . . , n)
(8.13)
120
CHAPTER 8. CONSTRAINTS
subject to boundary conditions y (xa ) = y (xb ) = 0 and k constraints of the form
Zxb
dx Nl {y }, {y }, x = Cl
(l = 1, . . . , k) ,
(8.14)
xa
construct the extended functional
X
Zxb
k
k
X
F {y (x)}; {j } dx L {y }, {y }, x +
l Nl {y }, {y }, x
l Cl (8.15)
l=1
xa
l=1
and freely extremize over {y1 , . . . , yn ; 1 , . . . , k }. This results in (n + k) equations
(
X
)
k
L
Nl
d L
d Nl
l
=0
( = 1, . . . , n)
(8.16)
y
dx y
y
dx y
l=1
Zxb
dx Nl {y }, {y }, x = Cl
(l = 1, . . . , k) .
(8.17)
xa
8.4
Extremization of Functionals : Holonomic Constraints
Given a functional
F {y (x)} =
Zxb
dx L {y }, {y }, x
( = 1, . . . , n)
(8.18)
xa
subject to boundary conditions y (xa ) = y (xb ) = 0 and k constraints of the form
(j = 1, . . . , k) ,
(8.19)
Gj {y (x)}, x = 0
construct the extended functional
F {y (x)}; {j (x)}
Zxb
k
X
j Gj {y }
dx L {y }, {y }, x +
(8.20)
j=1
xa
and freely extremize over y1 , . . . , yn ; 1 , . . . , k :
)
X
Zxb ( X
k
k
n
X
Gj
d L
L
j
Gj j = 0 ,
+
y +
F = dx
y
dx y
y
=1
j=1
xa
resulting in the (n + k) equations
!
k
X
d L
Gj
L
j
dx y
y
y
(8.21)
j=1
( = 1, . . . , n)
(8.22)
j=1
Gj {y }, x = 0
(j = 1, . . . , k) .
(8.23)
8.4. EXTREMIZATION OF FUNCTIONALS : HOLONOMIC CONSTRAINTS
8.4.1
121
Examples of extremization with constraints
Volume of a cylinder : As a warm-up problem, lets maximize the volume V = a2 h of a
cylinder of radius a and height h, subject to the constraint
G(a, h) = 2a +
h2
=0 .
b
(8.24)
We therefore define
V (a, h, ) V (a, h) + G(a, h) ,
(8.25)
and set
V
= 2ah + 2 = 0
a
(8.26)
h
V
= a2 + 2 = 0
h
b
(8.27)
h2
V
= 2a +
=0 .
(8.28)
Solving these three equations simultaneously gives
r
2
b
2
, h=
, = 3/2 b1/2 3/2
a=
5
5
5
V =
4
55/2
5/2 b1/2 .
(8.29)
Hanging rope : We minimize the energy functional
E y(x) = g
Zx2
q
dx y 1 + y 2 ,
(8.30)
x1
where is the linear mass density, subject to the constraint of fixed total length,
C y(x) =
Zx2 q
dx 1 + y 2 .
x1
Thus,
E y(x), = E y(x) + C y(x) =
with
L (y, y , x) = (gy + )
Since
L
x
= 0 we have that
J = y
(8.31)
Zx2
dx L (y, y , x) ,
(8.32)
x1
1 + y2 .
gy +
L
L = p
y
1 + y2
(8.33)
(8.34)
122
is constant. Thus,
with solution
CHAPTER 8. CONSTRAINTS
p
dy
= J 1 (gy + )2 J 2 ,
dx
g
J
y(x) =
+
cosh
(x a) .
g g
J
(8.35)
(8.36)
Here, J , a, and are constants to be determined by demanding y(xi ) = yi (i = 1, 2), and
that the total length of the rope is C.
Geodesic on a curved surface : Consider next the problem of a geodesic on a curved surface.
Let the equation for the surface be
G(x, y, z) = 0 .
(8.37)
Zb p
Zb
D = ds =
dx2 + dy 2 + dz 2 .
(8.38)
We wish to extremize the distance,
We introduce a parameter t defined on the unit interval: t [0, 1], such that x(0) = xa ,
x(1) = xb , etc. Then D may be regarded as a functional, viz.
D x(t), y(t), z(t) =
Z1
0
dt
x 2 + y 2 + z 2 .
We impose the constraint by forming the extended functional, D :
Z1 p
2
2
2
x + y + z + G(x, y, z) ,
D x(t), y(t), z(t), (t) dt
(8.39)
(8.40)
and we demand that the first functional derivatives of D vanish:
D
x
d
G
p
=
=0
+
x(t)
dt
x
x 2 + y 2 + z 2
G
D
y
d
p
+
=
=0
2
2
2
y(t)
dt
y
x + y + z
z
d
G
D
p
=
=0
+
2
2
2
z(t)
dt
z
x + y + z
D
= G(x, y, z) = 0 .
(t)
Thus,
(t) =
v y y v
v
z z v
v
x x v
= 2
= 2
,
2
v x G
v y G
v z G
p
with v = x 2 + y 2 + z 2 and x
G(x, y, z) = 0, which is the fourth.
x ,
(8.41)
(8.42)
(8.43)
(8.44)
(8.45)
etc. These three equations are supplemented by
8.5. APPLICATION TO MECHANICS
8.5
123
Application to Mechanics
Let us write our system of constraints in the differential form
n
X
gj (q, t) dq + hj (q, t)dt = 0
(j = 1, . . . , k) .
(8.46)
=1
If the partial derivatives satisfy
gj
gj
=
q
q
hj
gj
=
,
t
q
(8.47)
then the differential can be integrated to give dG(q, t) = 0, where
gj =
Gj
q
hj =
Gj
.
t
(8.48)
The action functional is
Ztb
S[{q (t)}] = dt L {q }, {q }, t
( = 1, . . . , n) ,
(8.49)
ta
subject to boundary conditions q (ta ) = q (tb ) = 0. The first variation of S is given by
(
)
Ztb X
n
d L
L
q .
(8.50)
S = dt
q
dt q
=1
ta
Since the {q (t)} are no longer independent, we cannot infer that the term in brackets
vanishes for each . What are the constraints on the variations q (t)? The constraints are
expressed in terms of virtual displacements which take no time: t = 0. Thus,
n
X
gj (q, t) q (t) = 0 ,
(8.51)
=1
where j = 1, . . . , k is the constraint index. We may now relax the constraint by introducing k
undetermined functions j (t), by adding integrals of the above equations with undetermined
coefficient functions to S:
(
)
X
n
k
X
L
d L
j (t) gj (q, t) q (t) = 0 .
(8.52)
+
q
dt
=1
j=1
Now we can demand that the term in brackets vanish for all . Thus, we obtain a set of
(n + k) equations,
k
X
d L
L
j (t) gj (q, t) Q
(8.53)
=
dt q
q
j=1
gj (q, t) q + hj (q, t) = 0 ,
(8.54)
124
CHAPTER 8. CONSTRAINTS
in (n + k) unknowns q1 , . . . , qn , 1 , . . . , k . Here, Q is the force of constraint conjugate
to the generalized coordinate q . Thus, with
p =
L
q
F =
L
q
Q =
k
X
j gj ,
(8.55)
j=1
we write Newtons second law as
Note that we can write
p = F + Q .
(8.56)
L
d L
S
=
q(t)
q
dt q
(8.57)
and that the instantaneous constraints may be written
gj q = 0
(j = 1, . . . , k) .
(8.58)
Thus, by demanding
k
X
S
j g j = 0
+
q(t)
(8.59)
j=1
we require that the functional derivative be linearly dependent on the k vectors gj .
8.5.1
Constraints and conservation laws
We have seen how invariance of the Lagrangian with respect to a one-parameter family of
coordinate transformations results in an associated conserved quantity , and how a lack of
explicit time dependence in L results in the conservation of the Hamiltonian H. In deriving
both these results, however, we used the equations of motion p = F . What happens when
we have constraints, in which case p = F + Q ?
Lets begin with the Hamiltonian. We have H = q p L, hence
L
L
L
dH
= p
q + p
q
dt
q
q
t
= Q q
L
.
t
(8.60)
We now use
Q q = j gj q = j hj
(8.61)
to obtain
dH
L
= j hj
.
(8.62)
dt
t
We therefore conclude that in a system with constraints of the form gj q + hj = 0, the
Hamiltonian is conserved if each hj = 0 and if L is not explicitly dependent on time. In
8.6. WORKED EXAMPLES
125
the case of holonomic constraints, hj =
Gj
t ,
so H is conserved if neither L nor any of the
constraints Gj is explicitly time-dependent.
Next, let us rederive Noethers theorem when constraints are present. We assume a oneparameter family of transformations q q () leaves L invariant. Then
0=
L q
L q
dL
=
+
d
q
q
q
d q
+ p
= p Q
dt
q
q
d
.
p
j gj
=
dt
(8.63)
Now let us write the constraints in differential form as
dt + k d = 0 .
gj d
q + h
j
j
(8.64)
We now have
d
= j kj ,
(8.65)
dt
which says that if the constraints are independent of then is conserved. For holonomic
constraints, this means that
Gj q(), t = 0
Gj
=0,
kj =
(8.66)
i.e. Gj (
q , t) has no explicit dependence.
8.6
Worked Examples
Here we consider several example problems of constrained dynamics, and work each out in
full detail.
8.6.1
One cylinder rolling off another
As an example of the constraint formalism, consider the system in Fig. 8.1, where a cylinder
of radius a rolls atop a cylinder of radius R. We have two constraints:
G1 (r, 1 , 2 ) = r R a = 0
(cylinders in contact)
(8.67)
G2 (r, 1 , 2 ) = R 1 a (2 1 ) = 0
(no slipping) ,
(8.68)
from which we obtain the gj :
gj =
1
0
0
,
0 R + a a
(8.69)
126
CHAPTER 8. CONSTRAINTS
which is to say
G1
=1
r
G1
=0
1
G1
=0
2
(8.70)
G2
=0
r
G2
=R+a
1
G2
= a .
2
(8.71)
The Lagrangian is
L = T U = 12 M r 2 + r 2 12 + 12 I 22 M gr cos 1 ,
(8.72)
where M and I are the mass and rotational inertia of the rolling cylinder, respectively.
Note that the kinetic energy is a sum of center-of-mass translation Ttr = 12 M r 2 + r 2 12
and rotation about the center-of-mass, Trot = 21 I 22 . The equations of motion are
L
d L
= M r M r 12 + M g cos 1 = 1 Qr
(8.73)
dt r
r
d L
L
= M r 2 1 + 2M r r 1 M gr sin 1 = (R + a) 2 Q1
(8.74)
dt 1
1
L
d L
= I 2 = a 2 Q2 .
(8.75)
dt 2
2
To these three
add the two constraints, resulting in five equations in the five
equations we
unknowns r, 1 , 2 , 1 , 2 .
We solve by first implementing
the constraints, which give r = (R + a) a constant (i.e.
r = 0), and 2 = 1 + Ra 1 . Substituting these into the above equations gives
M (R + a) 12 + M g cos 1 = 1
M (R + a)2 1 M g(R + a) sin 1 = (R + a) 2
I
R+a
1 = a2 .
a
(8.76)
(8.77)
(8.78)
From eqn. 8.78 we obtain
R+a
I
2 = 2 = 2 I 1 ,
a
a
which we substitute into eqn. 8.77 to obtain
I
M + 2 (R + a)2 1 M g(R + a) sin 1 = 0 .
a
(8.79)
(8.80)
Multiplying by 1 , we obtain an exact differential, which may be integrated to yield
Mg
Mg
I
1
2
(8.81)
2 M 1 + M a2 1 + R + a cos 1 = R + a cos 1 .
8.6. WORKED EXAMPLES
127
Figure 8.2: Frictionless motion under gravity along a curved surface. The skier flies off the
surface when the normal force vanishes.
Here, we have assumed that 1 = 0 when 1 = 1 , i.e. the rolling cylinder is released from
rest at 1 = 1 . Finally, inserting this result into eqn. 8.76, we obtain the radial force of
constraint,
o
Mg n
(8.82)
(3 + ) cos 1 2 cos 1 ,
Qr =
1+
where = I/M a2 is a dimensionless parameter (0 1). This is the radial component
of the normal force between the two cylinders. When Qr vanishes, the cylinders lose contact
the rolling cylinder flies off. Clearly this occurs at an angle 1 = 1 , where
1 2 cos 1
1 = cos
.
(8.83)
3+
The detachment angle 1 is an increasing function of , which means that larger I delays
detachment. This makes good sense, since when I is larger the gain in kinetic energy is
split between translational and rotational motion of the rolling cylinder.
8.6.2
Frictionless motion along a curve
Consider the situation in Fig. 8.2 where a skier moves frictionlessly under the influence of
gravity along a general curve y = h(x). The Lagrangian for this problem is
L = 12 m(x 2 + y 2 ) mgy
(8.84)
and the (holonomic) constraint is
G(x, y) = y h(x) = 0 .
Accordingly, the Euler-Lagrange equations are
G
L
d L
=
,
dt q
q
q
(8.85)
(8.86)
128
CHAPTER 8. CONSTRAINTS
where q1 = x and q2 = y. Thus, we obtain
m
x = h (x) = Qx
(8.87)
m
y + mg = = Qy .
(8.88)
We eliminate y in favor of x by invoking the constraint. Since we need y, we must differentiate the constraint, which gives
y = h (x) x
y = h (x) x + h (x) x 2 .
(8.89)
Using the second Euler-Lagrange equation, we then obtain
= g + h (x) x + h (x) x 2 .
m
(8.90)
Finally, we substitute this into the first E-L equation to obtain an equation for x alone:
2
x
+ h (x) h (x) x 2 + g h (x) = 0 .
(8.91)
1 + h (x)
Had we started by eliminating y = h(x) at the outset, writing
2 2
x mg h(x) ,
L(x, x)
= 21 m 1 + h (x)
(8.92)
we would also have obtained this equation of motion.
The skier flies off the curve when the vertical force of constraint Qy = starts to become
negative, because the curve can only supply a positive normal force. Suppose the skier
starts from rest at a height y0 . We may then determine the point x at which the skier
detaches from the curve by setting (x) = 0. To do so, we must eliminate x and x
in terms
of x. For x
, we may use the equation of motion to write
gh + h h x 2
x
=
,
(8.93)
1 + h 2
which allows us to write
=m
g + h x 2
1 + h 2
(8.94)
To eliminate x,
we use conservation of energy,
E = mgy0 = 12 m 1 + h
which fixes
2
y0 h
x = 2g
1 + h 2
2
x 2 + mgh ,
(8.95)
(8.96)
Putting it all together, we have
(x) =
mg
1 + h
o
n
2
.
1
+
h
+
2(y
h)
h
0
2 2
(8.97)
8.6. WORKED EXAMPLES
129
Figure 8.3: Finding the local radius of curvature: z = 2 /2R.
The skier detaches from the curve when (x) = 0, i.e. when
2
1 + h + 2(y0 h) h = 0 .
(8.98)
There is a somewhat easier way of arriving at the same answer. This is to note that the
skier must fly off when the local centripetal force equals the gravitational force normal to
the curve, i.e.
m v 2 (x)
= mg cos (x) ,
R(x)
(8.99)
1/2
where R(x) is the local radius of curvature. Now tan = h , so cos = 1 + h 2
. The
2
2
2
2
2
square of the velocity is v = x + y = 1 + h x . What is the local radius of curvature
R(x)? This can be determined from the following argument, and from the sketch in Fig.
8.3. Writing x = x + , we have
y = h(x ) + h (x ) + 21 h (x ) 2 + . . . .
(8.100)
We now drop a perpendicular segment
of length z from the point (x, y) to the line which is
tangent to the curve at x , h(x ) . According to Fig. 8.3, this means
1
h
1
1
.
= 2
z 2
1
y
h
1+h
1+h
(8.101)
130
CHAPTER 8. CONSTRAINTS
Thus, we have
y = h + 12 h 2
+ z h 2
+ z h
1
=h p
+ 2h p
1 + h 2
1 + h 2
h 2
h + z h 2
+ O(z)
+
= p
2 1 + h 2
1 + h 2
from which we obtain
h z
,
=p
1 + h 2
z=
and therefore
R(x) =
Thus, the detachment condition,
h 2
2 1+
3/2
h 2
+ O( 3 )
2 3/2
1
1
+
h (x)
.
h (x)
mv 2
m h x 2
mg
= p
=p
= mg cos
R
1 + h 2
1 + h 2
(8.102)
(8.103)
(8.104)
(8.105)
reproduces the result from eqn. 8.94.
8.6.3
Disk rolling down an inclined plane
A hoop of mass m and radius R rolls without slipping down an inclined plane. The inclined
plane has opening angle and mass M , and itself slides frictionlessly along a horizontal
surface. Find the motion of the system.
Figure 8.4: A hoop rolling down an inclined plane lying on a frictionless surface.
8.6. WORKED EXAMPLES
131
Solution : Referring to the sketch in Fig. 8.4, the center of the hoop is located at
x = X + s cos a sin
y = s sin + a cos ,
where X is the location of the lower left corner of the wedge, and s is the distance
along the wedge to the bottom of the hoop. If the hoop rotates through an angle ,
the no-slip condition is a + s = 0. Thus,
L = 21 M X 2 + 21 m x 2 + y 2 + 12 I 2 mgy
I
1
= 2 m + 2 s 2 + 12 (M + m)X 2 + m cos X s mgs sin mga cos .
a
Since X is cyclic in L, the momentum
PX = (M + m)X + m cos s ,
is preserved: PX = 0. The second equation of motion, corresponding to the generalized
coordinate s, is
I
= g sin .
s + cos X
1+
ma2
and immediately obtain
Using conservation of PX , we eliminate s in favor of X,
The result
=
X
1+
M
m
g sin cos
aX .
I
cos2
1 + ma
2
g 1+ M
m sin
as
s =
I
2
1+ M
cos
1
+
2
m
ma
follows immediately. Thus,
X(t) = X(0) + X(0)
t + 12 aX t2
s(t) = s(0) + s(0)
t + 12 as t2 .
Note that as < 0 while aX > 0, i.e. the hoop rolls down and to the left as the wedge
slides to the right. Note that I = ma2 for a hoop; weve computed the answer here
for general I.
8.6.4
Pendulum with nonrigid support
A particle of mass m is suspended from a flexible string of length in a uniform gravitational
field. While hanging motionless in equilibrium, it is struck a horizontal blow resulting in
an initial angular velocity 0 . Treating the system as one with two degrees of freedom and
a constraint, answer the following:
132
CHAPTER 8. CONSTRAINTS
(a) Compute the Lagrangian, the equation of constraint, and the equations of motion.
Solution : The Lagrangian is
L = 12 m r 2 + r 2 2 + mgr cos .
The constraint is r = . The equations of motion are
m
r mr 2 mg cos =
mr 2 + 2mr r mg sin = 0 .
(b) Compute the tension in the string as a function of angle .
Solution : Energy is conserved, hence
2 2
1
2 m
mg cos = 21 m2 02 mg cos 0 .
We take 0 = 0 and 0 = 0 . Thus,
with =
2 = 02 2 2 1 cos ,
g/. Substituting this into the equation for , we obtain
02
= mg 2 3 cos 2 .
(c) Show that if 02 < 2g/ then the particles motion is confined below the horizontal
and that the tension in the string is always positive (defined such that positive means
exerting a pulling force and negative means exerting a pushing force). Note that the
difference between a string and a rigid rod is that the string can only pull but the rod
can pull or push. Thus, the string tension must always be positive or else the string
goes slack.
Solution : Since 2 0, we must have
02
1 cos .
2 2
The condition for slackness is = 0, or
02
=1
2 2
3
2
cos .
Thus, if 02 < 2 2 , we have
1>
02
> 1 cos > 1 32 cos ,
2 2
and the string never goes slack. Note the last equality follows from cos > 0. The
string rises to a maximum angle
2
max = cos1 1 02 .
2
8.6. WORKED EXAMPLES
133
(d) Show that if 2g/ < 02 < 5g/ the particle rises above the horizontal and the string
becomes slack (the tension vanishes) at an angle . Compute .
Solution : When 2 > 2 2 , the string rises above the horizontal and goes slack at an
angle
2
= cos1 32 02 .
3
This solution craps out when the string is still taut at = , which means 02 = 5 2 .
(e) Show that if 02 > 5g/ the tension is always positive and the particle executes circular
motion.
Solution : For 02 > 5 2 , the string never goes slack. Furthermore, never vanishes.
Therefore, the pendulum undergoes circular motion, albeit not with constant angular
velocity.
8.6.5
Falling ladder
A uniform ladder of length and mass m has one end on a smooth horizontal floor and the
other end against a smooth vertical wall. The ladder is initially at rest and makes an angle
0 with respect to the horizontal.
Figure 8.5: A ladder sliding down a wall and across a floor.
(a) Make a convenient choice of generalized coordinates and find the Lagrangian.
Solution : I choose as generalized coordinates the Cartesian coordinates (x, y) of the
ladders center of mass, and the angle it makes with respect to the floor. The
Lagrangian is then
L = 21 m (x 2 + y 2 ) + 12 I 2 + mgy .
134
CHAPTER 8. CONSTRAINTS
There are two constraints: one enforcing contact along the wall, and the other enforcing contact along the floor. These are written
G1 (x, y, ) = x
1
2
1
2
cos = 0
G2 (x, y, ) = y sin = 0 .
(b) Prove that the ladder leaves the wall when its upper end has fallen to a height 23 L sin 0 .
The equations of motion are
X Gj
d L
L
=
j
.
dt q
q
q
j
Thus, we have
mx
= 1 = Q x
m y + mg = 2 = Qy
I = 12 1 sin 2 cos = Q .
We now implement the constraints to eliminate x and y in terms of . We have
x = 21 sin
y = 1 cos
x
= 12 cos 2 12 sin
y = 1 sin 2 + 1 cos .
We can now obtain the forces of constraint in terms of the function (t):
1 = 21 m sin + cos 2
= + 1 m cos sin 2 + mg .
2
We substitute these into the last equation of motion to obtain the result
I = I0 12 mg cos ,
or
(1 + ) = 202 cos ,
p
with I0 = 14 m2 , I/I0 and 0 = g/. This may be integrated once (multiply by
to convert to a total derivative) to yield
1
2 (1
+ ) 2 + 2 02 sin = 2 02 sin 0 ,
which is of course a statement of energy conservation. This,
4 02 (sin 0 sin )
2 =
1+
2
2 cos
.
= 0
1+
8.6. WORKED EXAMPLES
135
We may now obtain 1 () and 2 ():
mg
3 sin 2 sin 0 cos
1+
o
mg n
2 () =
(3 sin 2 sin 0 sin + .
1+
1 () =
Demanding 1 () = 0 gives the detachment angle = d , where
sin d =
2
3
sin 0 .
Note that 2 (d ) = mg/(1 + ) > 0, so the normal force from the floor is always
positive for > d . The time to detachment is
T1 (0 ) =
d
=
1+
2 0
Z0
d
.
sin 0 sin
(c) Show that the subsequent motion can be reduced to quadratures (i.e. explicit integrals).
Solution : After the detachment, there is no longer a constraint G1 . The equations of
motion are
mx
=0
(conservation of x-momentum)
m y + m g =
I = 21 cos ,
along with the constraint y =
the second equation yields
1
2
sin . Eliminating y in favor of using the constraint,
= mg 12 m sin 2 + 12 m cos .
Plugging this into the third equation of motion, we find
I = 2 I0 02 cos + I0 sin cos 2 I0 cos2 .
Multiplying by one again obtains a total time derivative, which is equivalent to
rediscovering energy conservation:
E = 21 I + I0 cos2 2 + 2 I0 02 sin .
By continuity with the first phase of the motion, we obtain the initial conditions for
this second phase:
= sin1 23 sin 0
s
sin 0
.
= 2 0
3 (1 + )
136
CHAPTER 8. CONSTRAINTS
Figure 8.6: Plot of time to fall for the slipping ladder. Here x = sin 0 .
Thus,
4 02 sin 0
I + I0 49 I0 sin2 0
+
3 (1 + )
2
2
4 sin 0
= 2 I0 0 1 + 27
sin 0 .
1+
E=
1
2
1
3
mg sin 0
(d) Find an expression for the time T (0 ) it takes the
p ladder to smack against the floor.
Note that, expressed in units of the time scale L/g, T is a dimensionless function
of 0 . Numerically integrate this expression and plot T versus 0 .
Solution : The time from detachment to smack is
T2 (0 ) =
d
1
=
2 0
Zd s
d
0
1 + cos2
.
4 sin2 0
1 27
sin 0 sin
1+
The total time is then T (0 ) = T1 (0 ) + T2 (0 ). For a uniformly dense ladder, I =
1
1
1
2
12 m = 3 I0 , so = 3 .
(e) What is the horizontal velocity of the ladder at long times?
8.6. WORKED EXAMPLES
137
Solution : From the moment of detachment, and thereafter,
x =
21
sin =
4g
sin3/2 0 .
27 (1 + )
(f) Describe in words the motion of the ladder subsequent to it slapping against the floor.
Solution : Only a fraction of the ladders initial potential energy is converted into
kinetic energy of horizontal motion. The rest is converted into kinetic energy of
vertical motion and of rotation. The slapping of the ladder against the floor is an
elastic collision. After the collision, the ladder must rise again, and continue to rise
and fall ad infinitum, as it slides along with constant horizontal velocity.
8.6.6
Point mass inside rolling hoop
Consider the point mass m inside the hoop of radius R, depicted in Fig. 8.7. We choose
as generalized coordinates the Cartesian coordinates (X, Y ) of the center of the hoop, the
Cartesian coordinates (x, y) for the point mass, the angle through which the hoop turns,
and the angle which the point mass makes with respect to the vertical. These six coordinates are not all independent. Indeed, there are only two independent coordinates for this
system, which can be taken to be and . Thus, there are four constraints:
X R G1 = 0
(8.106)
Y R G2 = 0
(8.107)
x X R sin G3 = 0
(8.108)
y Y + R cos G4 = 0 .
(8.109)
Figure 8.7: A point mass m inside a hoop of mass M , radius R, and moment of inertia I.
The kinetic and potential energies are easily expressed in terms of the Cartesian coordinates,
aside from the energy of rotation of the hoop about its CM, which is expressed in terms of
138
CHAPTER 8. CONSTRAINTS
:
T = 21 M (X 2 + Y 2 ) + 21 m(x 2 + y 2 ) + 12 I 2
(8.110)
U = M gY + mgy .
(8.111)
The moment of inertia of the hoop about its CM is I = M R2 , but we could imagine a
situation in which I were different. For example, we could instead place the point mass
inside a very short cylinder with two solid end caps, in which case I = 12 M R2 . The
Lagrangian is then
L = 21 M (X 2 + Y 2 ) + 21 m(x 2 + y 2 ) + 12 I 2 M gY mgy .
(8.112)
Note that L as written is completely independent of and !
Continuous symmetry
Note that there is an continuous symmetry to L which is satisfied by all the constraints,
under
X()
=X +
Y () = Y
(8.113)
x
() = x +
y() = y
(8.114)
()
=+
R
=.
()
(8.115)
Thus, according to Noethers theorem, there is a conserved quantity
=
1 L
L
L
+
+
x
R
X
I
= M X + mx + .
R
(8.116)
This means = 0. This reflects the overall conservation of momentum in the x-direction.
Energy conservation
Since neither L nor any of the constraints are explicitly time-dependent, the Hamiltonian
is conserved. And since T is homogeneous of degree two in the generalized velocities, we
have H = E = T + U :
E = 21 M (X 2 + Y 2 ) + 21 m(x 2 + y 2 ) + 12 I 2 + M gY + mgy .
(8.117)
8.6. WORKED EXAMPLES
139
Equations of motion
We have n = 6 generalized coordinates and k = 4 constraints. Thus, there are four undetermined multipliers {1 , 2 , 3 , 4 } used to impose the constraints. This makes for ten
unknowns:
X , Y , x , y , , , 1 , 2 , 3 , 4 .
(8.118)
Accordingly, we have ten equations: six equations of motion plus the four equations of
constraint. The equations of motion are obtained from
k
d L
L X Gj
j
+
.
=
dt q
q
q
(8.119)
j=1
Taking each generalized coordinate in turn, the equations of motion are thus
= 1 3
MX
(8.120)
M Y = M g + 2 4
(8.121)
m
x = 3
(8.122)
m
y = mg + 4
(8.123)
I = R 1
(8.124)
0 = R cos 3 R sin 4 .
(8.125)
Along with the four constraint equations, these determine the motion of the system. Note
that the last of the equations of motion, for the generalized coordinate q = , says that
Q = 0, which means that the force of constraint on the point mass is radial. Were the point
mass replaced by a rolling object, there would be an angular component to this constraint
in order that there be no slippage.
Implementation of constraints
We now use the constraint equations to eliminate X, Y , x, and y in terms of and :
X = R ,
Y =R
x = R + R sin
y = R(1 cos ) .
(8.126)
We also need the derivatives:
and
x = R + R cos
x
= R + R cos R sin 2 ,
(8.127)
y = R sin
x
= R sin + R cos 2 ,
(8.128)
140
CHAPTER 8. CONSTRAINTS
as well as
X = R ,
= R ,
X
Y = 0 ,
Y = 0 .
(8.129)
We now may write the conserved charge as
=
1
(I + M R2 + mR2 ) + mR cos .
R
This, in turn, allows us to eliminate in terms of and the constant :
cos ,
=
1 + mR
where
=
mR2
.
I + M R2
(8.130)
(8.131)
(8.132)
The energy is then
E = 12 (I + M R2 ) 2 + 12 m R2 2 + R2 2 + 2R2 cos + M gR + mgR(1 cos )
(
)
2
2
1
+
sin
2M
g
2g
= 21 mR2
.
(8.133)
+
(1 cos ) +
2 +
1+
R
1 + mR
mR
The last two terms inside the big bracket are constant, so we can write this as
4gk
1 + sin2 2 2g
(1 cos ) =
.
+
1+
R
R
(8.134)
Here, k is a dimensionless measure of the energy of the system, after subtracting the aforementioned constants. If k > 1, then 2 > 0 for all , which would result in loop-the-loop
motion of the point mass inside the hoop provided, that is, the normal force of the hoop
doesnt vanish and the point mass doesnt detach from the hoops surface.
Equation motion for (t)
The equation of motion for obtained by eliminating all other variables from the original
set of ten equations is the same as E = 0, and may be written
sin cos 2
g
1 + sin2
(8.135)
+
= .
1+
1+
R
We can use this to write in terms of 2 , and, after invoking eqn. 17.51, in terms of itself.
We find
4g
1
+
2
=
(8.136)
k sin2 12
2
R
1 + sin
i
(1 + ) sin h
g
21
2
4
k
sin
cos
+
1
+
sin
.
=
2
R 1 + sin2 2
(8.137)
8.6. WORKED EXAMPLES
141
Forces of constraint
We can solve for the j , and thus obtain the forces of constraint Q =
3 = m
x = mR + mR cos mR sin 2
i
mR h
cos 2 sin
=
1+
4 = m
y + mg = mg + mR sin + mR cos 2
h
gi
2
= mR sin + sin +
R
1 =
I (1 + )I
3
=
R
mR2
2 = (M + m)g + m
y = 4 + M g .
Gj
q .
(8.138)
(8.139)
(8.140)
(8.141)
One can check that 3 cos + 4 sin = 0.
The condition that the normal force of the hoop on the point mass vanish is 3 = 0, which
entails 4 = 0. This gives
(8.142)
(1 + sin2 ) cos = 4(1 + ) k sin2 12 .
Note that this requires cos < 0, i.e. the point of detachment lies above the horizontal
diameter of the hoop. Clearly if k is sufficiently large, the equality cannot be satisfied, and
the point mass executes a periodic loop-the-loop motion. In particular, setting = , we
find that
1
.
(8.143)
kc = 1 +
4(1 + )
If k > kc , then there is periodic loop-the-loop motion. If k < kc , then the point mass may
detach at a critical angle , but only if the motion allows for cos < 0. From the energy
conservation equation, we have that the maximum value of achieved occurs when = 0,
which means
cos max = 1 2k .
(8.144)
If 12 < k < kc , then, we have the possibility of detachment. This means the energy must be
large enough but not too large.
142
CHAPTER 8. CONSTRAINTS
Chapter 9
Central Forces and Orbital
Mechanics
9.1
Reduction to a one-body problem
Consider two particles interacting via a potential U (r1 , r2 ) = U |r1 r2 | . Such a potential, which depends only on the relative distance between the particles, is called a central
potential. The Lagrangian of this system is then
L = T U = 12 m1 r 12 + 21 m2 r 22 U |r1 r2 | .
9.1.1
(9.1)
Center-of-mass (CM) and relative coordinates
The two-body central force problem may always be reduced to two independent one-body
problems, by transforming to center-of-mass (R) and relative (r) coordinates (see Fig. 9.1),
viz.
R=
m2
r
m1 + m2
m1
r2 = R
r
m1 + m2
m1 r1 + m2 r2
m1 + m2
r1 = R +
r = r1 r2
(9.2)
(9.3)
We then have
L = 12 m1 r 1 2 + 12 m2 r 2 2 U |r1 r2 |
= 21 M R 2 + 12 r 2 U (r) .
143
(9.4)
(9.5)
144
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Figure 9.1: Center-of-mass (R) and relative (r) coordinates.
where
M = m1 + m2
=
9.1.2
m1 m2
m1 + m2
(total mass)
(9.6)
(reduced mass) .
(9.7)
Solution to the CM problem
= 0 and hence
We have L/R = 0, which gives Rd
R(t) = R(0) + R(0)
t.
(9.8)
Thus, the CM problem is trivial. The center-of-mass moves at constant velocity.
9.1.3
Solution to the relative coordinate problem
Angular momentum conservation: We have that = r p = r r is a constant of the
motion. This means that the motion r(t) is confined to a plane perpendicular to . It is
convenient to adopt two-dimensional polar coordinates (r, ). The magnitude of is
= r 2 = 2A
(9.9)
where dA = 21 r 2 d is the differential element of area subtended relative to the force center.
The relative coordinate vector for a central force problem subtends equal areas in equal times.
This is known as Keplers Second Law.
9.1. REDUCTION TO A ONE-BODY PROBLEM
145
Energy conservation: The equation of motion for the relative coordinate is
d L
U
L
r =
.
=
dt r
r
r
(9.10)
we have
Taking the dot product with r,
U
r
0 = r r +
r
o dE
n
d 1 2
r
+
U
(r)
=
.
dt 2
dt
(9.11)
Thus, the relative coordinate contribution to the total energy is itself conserved. The total
energy is of course Etot = E + 12 M R 2 .
Since is conserved, and since r = 0, all motion is confined to a plane perpendicular to
. Choosing coordinates such that z = , we have
E = 21 r 2 + U (r) = 12 r 2 +
2
+ U (r)
2r 2
= 21 r 2 + Ueff (r)
Ueff (r) =
2
2r 2
(9.12)
+ U (r) .
(9.13)
Integration of the Equations of Motion, Step I: The second order equation for r(t)
is
2
dUeff (r)
dU (r)
dE
= 0
r= 3
=
.
(9.14)
dt
r
dr
dr
However, conservation of energy reduces this to a first order equation, via
q
r
2 dr
2
r =
E Ueff (r)
dt = q
.
U
(r)
E 2r
2
(9.15)
This gives t(r), which must be inverted to obtain r(t). In principle this is possible. Note
that a constant of integration also appears at this stage call it r0 = r(t = 0).
Integration of the Equations of Motion, Step II: After finding r(t) one can integrate to find (t) using the conservation of :
= 2
r
d =
r 2 (t)
dt .
(9.16)
This gives (t), and introduces another constant of integration call it 0 = (t = 0).
Pause to Reflect on the Number of Constants: Confined to the plane perpendicular
to , the relative coordinate vector has two degrees of freedom. The equations of motion
146
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
are second order in time, leading to four constants of integration. Our four constants are
E, , r0 , and 0 .
The original problem involves two particles, hence six positions and six velocities, making
for 12 initial conditions. Six constants are associated with the CM system: R(0) and R(0).
The six remaining constants associated with the relative coordinate system are (three
components), E, r0 , and 0 .
Geometric Equation of the Orbit:
leading to
2
d2r
2
d
r
we have
From = r 2 ,
d
d
= 2
,
dt
r d
(9.17)
(9.18)
dr
d
2
r 4
F (r) + r
2
where F (r) = dU (r)/dr is the magnitude of the central force. This second order equation
may be reduced to a first order one using energy conservation:
E = 12 r 2 + Ueff (r)
2
2
dr
=
+ Ueff (r) .
2r 4 d
(9.19)
Thus,
dr
,
d = p
2 r 2 E Ueff (r)
(9.20)
which can be integrated to yield (r), and then inverted to yield r(). Note that only
one integration need be performed to obtain the geometric shape of the orbit, while two
integrations one for r(t) and one for (t) must be performed to obtain the full motion
of the system.
It is sometimes convenient to rewrite this equation in terms of the variable s = 1/r:
d2s
+ s = 2 2 F s1 .
2
d
s
(9.21)
As an example, suppose the geometric orbit is r() = k e , known as a logarithmic spiral.
What is the force? We invoke (9.18), with s () = 2 s, yielding
with
2
F s1 = (1 + 2 ) s3
2 =
C
1 .
2
F (r) =
C
r3
(9.22)
(9.23)
9.2. ALMOST CIRCULAR ORBITS
147
Figure 9.2: Stable and unstable circular orbits. Left panel: U (r) = k/r produces a stable
circular orbit. Right panel: U (r) = k/r 4 produces an unstable circular orbit.
The general solution for s() for this force law is
A cosh() + B sinh()
s() =
A cos || + B sin ||
if 2 > C
(9.24)
if
< C .
The logarithmic spiral shape is a special case of the first kind of orbit.
9.2
Almost Circular Orbits
(r ) = 0, which says
A circular orbit with r(t) = r0 satisfies r = 0, which means that Ueff
0
2
3
that F (r0 ) = /r0 . This is negative, indicating that a circular orbit is possible only if
the force is attractive over some range of distances. Since r = 0 as well, we must also have
E = Ueff (r0 ). An almost circular orbit has r(t) = r0 + (t), where |/r0 | 1. To lowest
order in , one derives the equations
d2
= 2
dt2
1
U (r ) .
eff 0
2 =
(9.25)
If 2 > 0, the circular orbit is stable and the perturbation oscillates harmonically. If 2 < 0,
the circular orbit is unstable and the perturbation grows exponentially. For the geometric
shape of the perturbed orbit, we write r = r0 + , and from (9.18) we obtain
4
r0
d2
=
F (r0 ) 3 = 2 ,
(9.26)
d2
2
with
d
ln
F
(r)
2 = 3 +
d ln r
r0
(9.27)
148
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
The solution here is
() = 0 cos ( 0 ) ,
(9.28)
where 0 and 0 are initial conditions. Setting = 0 , we obtain the sequence of values
n = 0 +
2n
,
(9.29)
at which () is a local maximum, i.e. at apoapsis, where r = r0 + 0 . Setting r = r0 0
is the condition for closest approach, i.e. periapsis. This yields the identical set if angles,
just shifted by . The difference,
= n+1 n 2 = 2 1 1 ,
(9.30)
is the amount by which the apsides (i.e. periapsis and apoapsis) precess during each cycle.
If > 1, the apsides advance, i.e. it takes less than a complete revolution = 2 between
successive periapses. If < 1, the apsides retreat, and it takes longer than a complete
revolution between successive periapses. The situation is depicted in Fig. 9.3 for the case
= 1.1. Below, we will exhibit a soluble model in which the precessing orbit may be
determined exactly. Finally, note that if = p/q is a rational number, then the orbit is
closed, i.e. it eventually retraces itself, after every q revolutions.
As an example, let F (r) = kr . Solving for a circular orbit, we write
(r) =
Ueff
2
k
=0,
r r 3
(9.31)
which has a solution only for k > 0, corresponding to an attractive potential. We then find
2 1/(3)
,
(9.32)
r0 =
k
and 2 = 3 . The shape of the perturbed orbits follows from = 2 . Thus, while
circular orbits exist whenever k > 0, small perturbations about these orbits are stable only
for 2 > 0, i.e. for < 3. One then has () = A cos ( 0 ). The perturbed orbits are
closed, at least to lowest order in , for = 3 (p/q)2 , i.e. for = p/q. The situation is
depicted in Fig. 9.2, for the potentials U (r) = k/r ( = 2) and U (r) = k/r 4 ( = 5).
9.3
Precession in a Soluble Model
Lets start with the answer and work backwards. Consider the geometrical orbit,
r() =
r0
.
1 cos
(9.33)
Our interest is in bound orbits, for which 0 < 1 (see Fig. 9.3). What sort of potential
gives rise to this orbit? Writing s = 1/r as before, we have
s() = s0 (1 cos ) .
(9.34)
9.4. THE KEPLER PROBLEM: U (R) = K R1
149
Substituting into (9.21), we have
d2s
1
F
s
=
+s
2 s 2
d2
= 2 s0 cos + s
= (1 2 ) s + 2 s0 ,
from which we conclude
F (r) =
with
k = 2 s0
k
C
+
,
r2 r3
C = ( 2 1)
(9.35)
(9.36)
2
.
(9.37)
The corresponding potential is
k
C
U (r) = + 2 + U ,
r
2r
(9.38)
where U is an arbitrary constant, conveniently set to zero. If and C are given, we have
r
C
C
2
+
, = 1+ 2 .
(9.39)
r0 =
k
k
When C = 0, these expressions recapitulate those from the Kepler problem. Note that
when 2 + C < 0 that the effective potential is monotonically increasing as a function of
r. In this case, the angular momentum barrier is overwhelmed by the (attractive, C < 0)
inverse square part of the potential, and Ueff (r) is monotonically increasing. The orbit then
passes through the force center. It is a useful exercise to derive the total energy for the
orbit,
2E(2 + C)
k2
2
=
1
+
.
(9.40)
E = (2 1)
2(2 + C)
k2
9.4
9.4.1
The Kepler Problem: U (r) = k r1
Geometric shape of orbits
The force is F (r) = kr 2 , hence the equation for the geometric shape of the orbit is
k
d2 s
+ s = 2 2 F (s1 ) = 2 ,
d2
s
(9.41)
with s = 1/r. Thus, the most general solution is
s() = s0 C cos( 0 ) ,
(9.42)
150
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Figure 9.3: Precession in a soluble model, with geometric orbit r() = r0 /(1 cos ),
shown here with = 1.1. Periapsis and apoapsis advance by = 2(1 1 ) per cycle.
where C and 0 are constants. Thus,
r() =
r0
1 cos( 0 )
(9.43)
where r0 = 2 /k and where we have defined a new constant Cr0 .
9.4.2
Laplace-Runge-Lenz vector
Consider the vector
A = p k r ,
(9.44)
9.4. THE KEPLER PROBLEM: U (R) = K R1
151
Figure 9.4: The effective potential for the Kepler problem, and associated phase curves.
The orbits are geometrically described as conic sections: hyperbolae (E > 0), parabolae
(E = 0), ellipses (Emin < E < 0), and circles (E = Emin ).
where r = r/|r| is the unit vector pointing in the direction of r. We may now show that A
is conserved:
ro
dn
dA
p k
=
dt
dt
r
r r r r
= p + p k
r2
rr
r
kr
k + k 2
= 3 (r r)
r
r
r
r)
rr
r(r
r
r(r r)
+ k
k + k 2 = 0 .
(9.45)
= k
3
3
r
r
r
r
So A is a conserved vector which clearly lies in the plane of the motion. A points toward
periapsis, i.e. toward the point of closest approach to the force center.
Lets assume apoapsis occurs at = 0 . Then
A r = Ar cos( 0 ) = 2 kr
giving
r() =
where
2
k A cos( 0 )
=
A
k
a(1 2 )
1 cos( 0 )
a(1 2 ) =
2
.
k
(9.46)
(9.47)
(9.48)
152
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
The orbit is a conic section with eccentricity . Squaring A, one finds
A2 = (p )2 2k r p + 2 k2
k
+ 2 k 2
r
2
k k2
k2
2
2 p
+ 2 = 2 E + 2
= 2
2 r
2
2
= p2 2 22
and thus
a=
9.4.3
k
2E
2 = 1 +
2E2
.
k2
(9.49)
(9.50)
Kepler orbits are conic sections
There are four classes of conic sections:
Circle: = 0, E = k2 /22 , radius a = 2 /k. The force center lies at the center of
circle.
2
2
Ellipse:
0 < < 1, k /2 < E < 0, semimajor axis a = k/2E, semiminor axis
b = a 1 2 . The force center is at one of the foci.
Parabola: = 1, E = 0, force center is the focus.
Hyperbola: > 1, E > 0, force center is closest focus (attractive) or farthest focus
(repulsive).
To see that the Keplerian orbits are indeed conic sections, consider the ellipse of Fig. 9.6.
The law of cosines gives
2 = r 2 + 4f 2 4rf cos ,
(9.51)
where f = a is the focal distance. Now for any point on an ellipse, the sum of the distances
to the left and right foci is a constant, and taking = 0 we see that this constant is 2a.
Thus, = 2a r, and we have
(2a r)2 = 4a2 4ar + r 2 = r 2 + 42 a2 4r cos
r(1 cos ) = a(1 2 ) .
Thus, we obtain
r() =
and we therefore conclude that
r0 =
a (1 2 )
,
1 cos
2
= a (1 2 ) .
k
(9.52)
(9.53)
(9.54)
9.4. THE KEPLER PROBLEM: U (R) = K R1
153
Figure 9.5: Keplerian orbits are conic sections, classified according to eccentricity: hyperbola ( > 1), parabola ( = 1), ellipse (0 < < 1), and circle ( = 0). The Laplace-RungeLenz vector, A, points toward periapsis.
Next let us examine the energy,
E = 12 r 2 + Ueff (r)
dr 2
2
k
1
= 2
+
2
2
r d
2r
r
2 2
2 ds 2
+
s ks ,
=
2 d
2
(9.55)
with
s=
1
k
= 2 1 cos .
r
(9.56)
k
ds
= 2 sin ,
d
(9.57)
Thus,
154
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Figure 9.6: The Keplerian ellipse, with the force center at the left focus. The focal distance
is f = a, where a is the semimajor axis length. The length of the semiminor axis is
b = 1 2 a.
and
ds
d
2
2 k 2 2
sin2
4
2
k
2 k2 2
s
=
4
2
2k
2 k 2
= s2 + 2 s + 4 2 1 .
(9.58)
Substituting this into eqn. 9.55, we obtain
E=
k2 2
1
.
22
(9.59)
For the hyperbolic orbit, depicted in Fig. 9.7, we have r = 2a, depending on whether
we are on the attractive or repulsive branch, respectively. We then have
(r 2a)2 = 4a2 4ar + r 2 = r 2 + 42 a2 4r cos
r(1 + cos ) = a(2 1) .
This yields
r() =
9.4.4
a (2 1)
.
1 + cos
(9.60)
(9.61)
Period of bound Kepler orbits
the period is = 2A/, where A = a2 1 2 is the area enclosed
From = r 2 = 2A,
by the orbit. This gives
3 1/2
3 1/2
a
a
= 2
(9.62)
= 2
k
GM
9.4. THE KEPLER PROBLEM: U (R) = K R1
155
Figure 9.7: The Keplerian hyperbolae, with the force center at the left focus. The left (blue)
branch corresponds to an attractive potential, while the right (red) branch corresponds to
a repulsive potential. The equations of these branches are r = = 2a, where the top sign
corresponds to the left branch and the bottom sign to the right branch.
as well as
GM
a3
=
,
2
4 2
(9.63)
where k = Gm1 m2 and M = m1 + m2 is the total mass. For planetary orbits, m1 = M is
the solar mass and m2 = mp is the planetary mass. We then have
a3
mp GM
GM
=
1
+
,
2
2
M 4
4 2
(9.64)
which is to an excellent approximation independent of the planetary mass. (Note that
mp /M 103 even for Jupiter.) This analysis also holds, mutatis mutandis, for the
case of satellites orbiting the earth, and indeed in any case where the masses are grossly
disproportionate in magnitude.
9.4.5
Escape velocity
The threshold for escape from a gravitational potential occurs at E = 0. Since E = T + U
is conserved, we determine the escape velocity for a body a distance r from the force center
by setting
r
GM m
2G(M + m)
2
1
vesc (r) =
.
(9.65)
E = 0 = 2 vesc (t)
r
r
p
When
M m, vesc (r) =
2GM/r. Thus, for an object at the surface of the earth,
vesc = 2gRE = 11.2 km/s.
156
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
9.4.6
Satellites and spacecraft
A satellite in a circular orbit a distance h above the earths surface has an orbital period
=
2
(RE + h)3/2 ,
GME
(9.66)
where we take p
msatellite ME . For low earth orbit (LEO), h RE = 6.37 106 m, in which
case LEO = 2 RE /g = 1.4 hr.
Consider a weather satellite in an elliptical orbit whose closest approach to the earth
(perigee) is 200 km above the earths surface and whose farthest distance (apogee) is 7200
km above the earths surface. What is the satellites orbital period? From Fig. 9.6, we see
that
dapogee = RE + 7200 km = 13571 km
dperigee = RE + 200 km = 6971 km
a = 12 (dapogee + dperigee ) = 10071 km .
(9.67)
a 3/2
LEO 2.65 hr .
RE
(9.68)
We then have
=
What happens if a spacecraft in orbit about the earth fires its rockets? Clearly the energy
and angular momentum of the orbit will change, and this means the shape will change. If
the rockets are fired (in the direction of motion) at perigee, then perigee itself is unchanged,
because v q
r = 0 is left unchanged at this point. However, E is increased, hence the eccen2
increases. This is the most efficient way of boosting a satellite into an
tricity = 1 + 2E
k 2
orbit with higher eccentricity. Conversely, and somewhat paradoxically, when a satellite in
LEO loses energy due to frictional drag of the atmosphere, the energy E decreases. Initially,
because the drag is weak and the atmosphere is isotropic, the orbit remains circular. Since
E decreases, hT i = E must increase, which means that the frictional forces cause the
satellite to speed up!
9.4.7
Two examples of orbital mechanics
Problem #1: At perigee of an elliptical Keplerian orbit, a satellite receives an impulse
p = p0 r. Describe the resulting orbit.
Solution #1: Since the impulse is radial, the angular momentum = r p is unchanged. The energy, however, does change, with E = p20 /2. Thus,
2f
2Ef 2
= 2i +
=1+
k2
p0
k
2
(9.69)
9.4. THE KEPLER PROBLEM: U (R) = K R1
157
Figure 9.8: At perigee of an elliptical orbit ri (), a radial impulse p is applied. The shape
of the resulting orbit rf () is shown.
The new semimajor axis length is
2 /k
1 2i
=
a
i
1 2f
1 2f
ai
.
=
1 (ai p20 /k)
af =
(9.70)
The shape of the final orbit must also be a Keplerian ellipse, described by
rf () =
2
1
,
k 1 cos( + )
f
(9.71)
where the phase shift is determined by setting
ri () = rf () =
Solving for , we obtain
2
1
.
k 1 + i
= cos1 i /f .
The situation is depicted in Fig. 9.8.
(9.72)
(9.73)
158
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Figure 9.9: The larger circular orbit represents the orbit of the earth. The elliptical orbit
represents that for an object orbiting the Sun with distance at perihelion equal to the Suns
radius.
Problem #2: Which is more energy efficient to send nuclear waste outside the solar
system, or to send it into the Sun?
p
Solution #2: Escape velocity for the solar system is vesc, (r) = GM /r. At a
p
distance aE , we then have vesc, (aE ) = 2 vE , where vE = GM /aE = 2aE /E =
29.9 km/s is the velocity of the earth in its orbit. The satellite is launched from earth,
and clearly the most energy efficient launch will be one in the direction ofthe earths
2 1 vE =
motion, in which case the velocity after escape from earth must be u =
12.4 km/s. The speed just above the earths atmosphere must then be u
, where
1
u2
2 m
GME m
= 12 mu2 ,
RE
(9.74)
or, in other words,
2
u
2 = u2 + vesc,
E .
(9.75)
We compute u
= 16.7 km/s.
The second method is to place the trash ship in an elliptical orbit whose perihelion
is the Suns radius, R = 6.98 108 m, and whose aphelion is aE . Using the general
equation r() = (2 /k)/(1 cos ) for a Keplerian ellipse, we therefore solve the
two equations
r( = ) = R =
1
2
1 + k
(9.76)
r( = 0) = aE =
2
1
.
1 k
(9.77)
We thereby obtain
=
aE R
= 0.991 ,
aE + R
(9.78)
9.5. APPENDIX I : MISSION TO NEPTUNE
159
which is a very eccentric ellipse, and
2
a2E v 2
v2
=
aE 2
k
G(M + m)
vE
2aE R
.
= (1 ) aE =
aE + R
Hence,
v2 =
2R
v2 ,
aE + R E
and the necessary velocity relative to earth is
s
!
2R
1 vE 0.904 vE ,
u=
aE + R
(9.79)
(9.80)
(9.81)
i.e. u = 27.0 km/s. Launch is in the opposite direction from the earths orbital mo2
tion, and from u
2 = u2 + vesc,
= 29.2 km/s, which is larger (in magnitude)
E we find u
than in the first scenario. Thus, it is cheaper to ship the trash out of the solar system
than to send it crashing into the Sun, by a factor u
2I /
u2II = 0.327.
9.5
Appendix I : Mission to Neptune
Four earth-launched spacecraft have escaped the solar system: Pioneer 10 (launch 3/3/72),
Pioneer 11 (launch 4/6/73), Voyager 1 (launch 9/5/77), and Voyager 2 (launch 8/20/77).1
The latter two are still functioning, and each are moving away from the Sun at a velocity
of roughly 3.5 AU/yr.
As the first objects of earthly origin to leave our solar system, both Pioneer spacecraft
featured a graphic message in the form of a 6 x 9 gold anodized plaque affixed to the
spacecrafts frame. This plaque was designed in part by the late astronomer and popular
science writer Carl Sagan. The humorist Dave Barry, in an essay entitled Bring Back Carls
Plaque, remarks,
But the really bad part is what they put on the plaque. I mean, if were going to
have a plaque, it ought to at least show the aliens what were really like, right?
Maybe a picture of people eating cheeseburgers and watching The Dukes of
Hazzard. Then if aliens found it, theyd say, Ah. Just plain folks.
But no. Carl came up with this incredible science-fair-wimp plaque that features
drawings of you are not going to believe this a hydrogen atom and naked
people. To represent the entire Earth! This is crazy! Walk the streets of any
town on this planet, and the two things you will almost never see are hydrogen
atoms and naked people.
1
There is a very nice discussion in the Barger and Olsson book on Grand Tours of the Outer Planets.
Here I reconstruct and extend their discussion.
160
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Figure 9.10: The unforgivably dorky Pioneer 10 and Pioneer 11 plaque.
During August, 1989, Voyager 2 investigated the planet Neptune. A direct trip to Neptune
along a Keplerian ellipse with rp = aE = 1 AU and ra = aN = 30.06 AU would take 30.6
years. To see this, note that rp = a (1 ) and ra = a (1 + ) yield
a=
Thus,
1
2
aE + aN = 15.53 AU
=
1
2
aN aE
= 0.9356 .
aN + aE
(9.82)
a 3/2
= 30.6 yr .
aE
(9.83)
GME m
= 12 m ( 1)2 vE2 ,
RE
(9.84)
The energy cost per kilogram of such a mission is computed as follows. Let the speed of
the probe after its escape from earth be vp = vE , and the speed just above the atmosphere
(i.e. neglecting atmospheric friction) is v0 . For the most efficient launch possible, the probe
is shot in the direction of earths instantaneous motion about the Sun. Then we must have
2
1
2 m v0
since the speed of the probe in the frame of the earth is vp vE = ( 1) vE . Thus,
h
i
E
= 12 v02 = 12 ( 1)2 + h vE2
m
vE2 =
GM
= 6.24 107 RJ /kg ,
aE
(9.85)
9.5. APPENDIX I : MISSION TO NEPTUNE
161
Figure 9.11: Mission to Neptune. The figure at the lower right shows the orbits of Earth,
Jupiter, and Neptune in black. The cheapest (in terms of energy) direct flight to Neptune,
shown in blue, would take 30.6 years. By swinging past the planet Jupiter, the satellite can
pick up great speed and with even less energy the mission time can be cut to 8.5 years (red
curve). The inset in the upper left shows the scattering event with Jupiter.
where
ME aE
= 7.050 102 .
M RE
(9.86)
v02
2E
=
= ( 1)2 + 2h .
mvE2
vE2
(9.87)
Therefore, a convenient dimensionless measure of the energy is
As we shall derive below, a direct mission to Neptune requires
r
2aN
= 1.3913 ,
aN + aE
(9.88)
which is close to the criterion for escape from the solar system, esc = 2. Note that about
52% of the energy is expended after the probe escapes the Earths pull, and 48% is expended
in liberating the probe from Earth itself.
This mission can be done much more economically by taking advantage of a Jupiter flyby, as
shown in Fig. 9.11. The idea of a flyby is to steal some of Jupiters momentum and then fly
162
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
away very fast before Jupiter realizes and gets angry. The CM frame of the probe-Jupiter
system is of course the rest frame of Jupiter, and in this frame conservation of energy means
that the final velocity uf is of the same magnitude as the initial velocity ui . However, in
the frame of the Sun, the initial and final velocities are vJ + ui and vJ + uf , respectively,
where vJ is the velocity of Jupiter in the rest frame of the Sun. If, as shown in the inset
to Fig. 9.11, uf is roughly parallel to vJ , the probes velocity in the Suns frame will be
enhanced. Thus, the motion of the probe is broken up into three segments:
I : Earth to Jupiter
II : Scatter off Jupiters gravitational pull
III : Jupiter to Neptune
We now analyze each of these segments in detail. In so doing, it is useful to recall that the
general form of a Keplerian orbit is
r() =
The energy is
d
1 cos
d=
2
= 2 1 a .
k
(9.89)
k2
,
(9.90)
22
with k = GM m, where M is the mass of either the Sun or a planet. In either case, M
dominates, and = M m/(M + m) m to extremely high accuracy. The time for the
trajectory to pass from = 1 to = 2 is
E = (2 1)
T =
dt =
Z2
Z2
Z2
3
d
d
2
=
d r () =
2 .
2
1 cos
1
(9.91)
For reference,
aE = 1 AU
ME = 5.972 1024 kg
aJ = 5.20 AU
MJ = 1.900 1027 kg
aN = 30.06 AU
M = 1.989 1030 kg
with 1 AU = 1.496 108 km. Here aE,J,N and ME,J, are the orbital radii and masses of
Earth, Jupiter, and Neptune, and the Sun. The last thing we need to know is the radius of
Jupiter,
RJ = 9.558 104 AU .
We need RJ because the distance of closest approach to Jupiter, or perijove, must be RJ or
greater, or else the probe crashes into Jupiter!
9.5.1
I. Earth to Jupiter
The probes velocity at perihelion is vp = vE . The angular momentum is = aE vE ,
whence
(aE vE )2
= 2 aE .
(9.92)
d=
GM
9.5. APPENDIX I : MISSION TO NEPTUNE
163
From r() = aE , we obtain
I = 2 1 .
(9.93)
This orbit will intersect the orbit of Jupiter if ra aJ , which means
r
d
2aJ
aJ
= 1.2952 .
1 I
aJ + aE
If this inequality holds, then intersection of Jupiters orbit will occur for
aJ 2 aE
1
J = 2 cos
.
(2 1) aJ
(9.94)
(9.95)
Finally, the time for this portion of the trajectory is
EJ
ZJ
1
d
= E
.
2 1 (2 1) cos 2
3
(9.96)
9.5.2
II. Encounter with Jupiter
We are interested in the final speed vf of the probe after its encounter with Jupiter. We
will determine the speed vf and the angle which the probe makes with respect to Jupiter
after its encounter. According to the geometry of Fig. 9.11,
vf2 = vJ2 + u2 2uvJ cos( + )
cos =
v2
J
Note that
vJ2 =
vf2
+
2vf vJ
(9.97)
u2
(9.98)
GM
aE 2
=
v .
aJ
aJ E
(9.99)
But what are u, , and ?
To determine u, we invoke
u2 = vJ2 + vi2 2vJ vi cos .
(9.100)
The initial velocity (in the frame of the Sun) when the probe crosses Jupiters orbit is given
by energy conservation:
2
1
2 m(vE )
GM m
GM m
= 12 mvi2
,
aE
aJ
(9.101)
which yields
2aE 2
2
vE .
= 2+
aJ
As for , we invoke conservation of angular momentum:
vi2
(vi cos )aJ = (vE )aE
vi cos =
(9.102)
aE
vE .
aJ
(9.103)
164
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
The angle is determined from
vJ = vi cos + u cos .
(9.104)
Putting all this together, we obtain
vi = vE
u = vE
p
2 2 + 2x
p
2 2 + 3x 2x3/2
x x
where
,
cos = p
2 2 + 3x 2x3/2
x
aE
= 0.1923 .
aJ
(9.105)
(9.106)
(9.107)
(9.108)
We next consider the scattering of the probe by the planet Jupiter. In the Jovian frame,
we may write
RJ (1 + J )
r() =
,
(9.109)
1 + J cos
where perijove occurs at
r(0) = RJ .
(9.110)
Here, is a dimensionless quantity, which is simply perijove in units of the Jovian radius.
Clearly we require > 1 or else the probe crashes into Jupiter! The probes energy in this
frame is simply E = 12 mu2 , which means the probe enters into a hyperbolic orbit about
Jupiter. Next, from
E=
we find
k 2 1
2 2 /k
(9.111)
2
= (1 + ) RJ
k
(9.112)
2
RJ
M
u
.
J = 1 +
aE
MJ
vE
(9.113)
The opening angle of the Keplerian hyperbola is then c = cos1 1
, and the angle is
J
related to c through
1
= 2c = 2 cos1
.
(9.114)
J
Therefore, we may finally write
q
vf = x vE2 + u2 + 2 u vE x cos(2c )
cos =
x v2
vf2 u2
+
2 vf vE x
E
(9.115)
(9.116)
9.5. APPENDIX I : MISSION TO NEPTUNE
165
Figure 9.12: Total time for Earth-Neptune mission as a function of dimensionless velocity
at perihelion, = vp /vE . Six different values of , the value of perijove in units of the
Jovian radius, are shown: = 1.0 (thick blue), = 5.0 (red), = 20 (green), = 50
(blue), = 100 (magenta), and = (thick black).
9.5.3
III. Jupiter to Neptune
Immediately after undergoing gravitational scattering off Jupiter, the energy and angular
momentum of the probe are
GM m
E = 12 mvf2
(9.117)
aJ
and
= vf aJ cos .
(9.118)
We write the geometric equation for the probes orbit as
r() =
d
,
1 + cos( J )
where
2
d=
=
k
vf aJ cos
vE aE
2
aE .
(9.119)
(9.120)
Setting E = (k2 /22 )(2 1), we obtain the eccentricity
v
u
u
= t1 +
vf2
2aE
2
vE
aJ
d
.
aE
(9.121)
166
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Note that the orbit is hyperbolic the probe will escape the Sun if vf > vE
condition that this orbit intersect Jupiter at = J yields
1 d
1 ,
cos =
aJ
2x. The
(9.122)
which determines the angle . Interception of Neptune occurs at
d
= aN
1 + cos(N J )
We then have
JN = E
d
aE
N = J + + cos
1 d
1 .
aN
3 ZN
1
d
.
2 1 + cos( J ) 2
(9.123)
(9.124)
The total time to Neptune is then the sum,
EN = EJ + JN .
(9.125)
In Fig. 9.12, we plot the mission time EN versus the velocity at perihelion, vp = vE , for
various values of . The value = corresponds to the case of no Jovian encounter at all.
9.6
Appendix II : Restricted Three-Body Problem
Problem : Consider the restricted three body problem in which a light object of mass m
(e.g. a satellite) moves in the presence of two celestial bodies of masses m1 and m2 (e.g. the
sun and the earth, or the earth and the moon). Suppose m1 and m2 execute stable circular
motion about their common center of mass. You may assume m m2 m1 .
(a) Show that the angular frequency for the motion of masses 1 and 2 is related to their
(constant) relative separation, by
GM
(9.126)
02 = 3 ,
r0
where M = m1 + m2 is the total mass.
Solution : For a Kepler potential U = k/r, the circular orbit lies at r0 = 2 /k, where
= r 2 is the angular momentum and k = Gm1 m2 . This gives
02 =
k
GM
2
= 3 = 3 ,
4
2
r0
r0
r0
(9.127)
with 0 = .
(b) The satellite moves in the combined gravitational field of the two large bodies; the
satellite itself is of course much too small to affect their motion. In deriving the motion
for the satellilte, it is convenient to choose a reference frame whose origin is the CM and
9.6. APPENDIX II : RESTRICTED THREE-BODY PROBLEM
167
Figure 9.13: The Lagrange points for the earth-sun system. Credit: WMAP project.
which rotates with angular velocity 0 . In the rotating frame the masses m1 and m2 lie,
respectively, at x1 = r0 and x2 = r0 , with
=
m2
M
m1
M
(9.128)
and with y1 = y2 = 0. Note + = 1.
Show that the Lagrangian for the satellite in this rotating frame may be written
L = 12 m x 0 y
2
+ 12 m y + 0 x
2
+q
G m1 m
G m2 m
+q
. (9.129)
2
2
2
2
(x + r0 ) + y
(x r0 ) + y
Solution : Let the original (inertial) coordinates be (x0 , y0 ). Then let us define the rotated
coordinates (x, y) as
x = cos(0 t) x0 + sin(0 t) y0
(9.130)
y = sin(0 t) x0 + cos(0 t) y0 .
(9.131)
x = cos(0 t) x 0 + sin(0 t) y0 + 0 y
(9.132)
y = sin(0 t) x0 + cos(0 t) y0 0 x .
(9.133)
(x 0 y)2 + (y + 0 x)2 = x 20 + y02 ,
(9.134)
Therefore,
Therefore
168
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
The Lagrangian is then
L = 21 m x 0 y
2
+ 12 m y + 0 x
2
+p
G m1 m
G m2 m
+p
,
2
2
(x x1 ) + y
(x x2 )2 + y 2
(9.135)
which, with x1 r0 and x2 r0 , agrees with eqn. 9.129
(c) Lagrange discovered that there are five special points where the satellite remains fixed
in the rotating frame. These are called the Lagrange points {L1, L2, L3, L4, L5}. A sketch
of the Lagrange points for the earth-sun system is provided in Fig. 9.13. Observation:
In working out the rest of this problem, I found it convenient to measure all distances in
units of r0 and times in units of 01 , and to eliminate G by writing Gm1 = 02 r03 and
Gm2 = 02 r03 .
Assuming the satellite is stationary in the rotating frame, derive the equations for the
positions of the Lagrange points.
Solution : At this stage it is convenient to measure all distances in units of r0 and times
in units of 01 to factor out a term m r02 02 from L, writing the dimensionless Lagrangian
e L/(m r 2 2 ). Using as well the definition of 2 to eliminate G, we have
L
0 0
0
with
e=
L
1
2
( )2 + 12 ( + )2 + p
x
r0
y
r0
( +
)2
1 dx
0 r0 dt
+p
( )2 + 2
1 dy
.
0 r0 dt
(9.136)
(9.137)
The equations of motion are then
( + ) ( )
2 =
d31
d32
+ 2 = 3 3 ,
d1
d2
where
d1 =
( + )2 + 2
d2 =
( )2 + 2 .
(9.138)
(9.139)
(9.140)
Here, x/r0 , = y/r0 , etc. Recall that + = 1. Setting the time derivatives to zero
yields the static equations for the Lagrange points:
( + ) ( )
+
d31
d32
= 3 + 3 ,
d1
d2
=
(9.141)
(9.142)
(d) Show that the Lagrange points with y = 0 are determined by a single nonlinear equation.
Show graphically that this equation always has three solutions, one with x < x1 , a second
9.6. APPENDIX II : RESTRICTED THREE-BODY PROBLEM
169
Figure 9.14: Graphical solution for the Lagrange points L1, L2, and L3.
with x1 < x < x2 , and a third with x > x2 . These solutions correspond to the points L3,
L1, and L2, respectively.
Solution : If = 0 the second equation is automatically satisfied. The first equation then
gives
+
(9.143)
=
+
.
+ 3
3
The RHS of the above equation diverges to + for = + 0+ and = + 0+ , and
diverges to for = 0+ and = 0+ , where 0+ is a positive infinitesimal. The
situation is depicted in Fig. 9.14. Clearly there are three solutions, one with < , one
with < < , and one with > .
(e) Show that the remaining two Lagrange points, L4 and L5, lie along equilateral triangles
with the two masses at the other vertices.
Solution : If 6= 0, then dividing the second equation by yields
1=
+
.
d31 d32
Substituting this into the first equation,
1
+
+
,
=
d31 d32
d31 d32
(9.144)
(9.145)
170
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
gives
d1 = d2 .
(9.146)
Reinserting this into the previous equation then gives the remarkable result,
d1 = d2 = 1 ,
(9.147)
which says that each of L4 and L5 lies on an equilateral triangle whose two other vertices
are the masses m1 and m2 . The side length of this equilateral triangle is r0 . Thus, the
dimensionless coordinates of L4 and L5 are
,
L5 , L5 = 21 , 23 .
(9.148)
L4 , L4 = 12 , 23
It turns out that L1, L2, and L3 are always unstable. Satellites placed in these positions
must undergo periodic course corrections in order to remain approximately fixed. The SOlar
and Heliopheric Observation satellite, SOHO, is located at L1, which affords a continuous
unobstructed view of the Sun.
(f) Show that the Lagrange points L4 and L5 are stable (obviously you need only consider
one of them) provided that the mass ratio m1 /m2 is sufficiently large. Determine this
critical ratio. Also find the frequency of small oscillations for motion in the vicinity of L4
and L5.
Solution : Now we write
= L4 +
= L4 + ,
(9.149)
and derive the linearized dynamics. Expanding the equations of motion to lowest order in
and , we have
!
!
d
d
d
d
1
2
1
2
23
+ 23
32
2 = 1 + 32
L4
L4
L4
L4
=
3
4
3 3
4
(9.150)
and
+ 2 =
=
3 3
2
3 3
4
d1
+
L4
+ 49 ,
3 3
2
!
d2
+
L4
where we have defined
=
As defined, [0, 1].
3 3
2
m1 m2
.
m1 + m2
d1
+
L4
3 3
2
!
d2
L4
(9.151)
(9.152)
9.6. APPENDIX II : RESTRICTED THREE-BODY PROBLEM
171
Fourier transforming the differential equation, we replace each time derivative by (i),
and thereby obtain
3
2i + 34 3
2 +
4
=0.
(9.153)
3
9
2
2i + 4 3
+4
Nontrivial solutions exist only when the determinant D vanishes. One easily finds
2
,
(9.154)
D( 2 ) = 4 2 + 27
16 1
which yields a quadratic equation in 2 , with roots
p
2 = 21 41 27 2 23 .
(9.155)
These frequencies are dimensionless. To convert to dimensionful units, we simply multiply
the solutions for by 0 , since we have rescaled time by 01 .
Note that the L4 and L5 points are stable only if 2 > 23
27 . If we define the mass ratio
m1 /m2 , the stability condition is equivalent to
27 + 23
m1
= 24.960 ,
>
(9.156)
=
m2
27 23
which is satisfied for both the Sun-Jupiter system ( = 1047) and hence for the Sun and
any planet and also for the Earth-Moon system ( = 81.2).
Objects found at the L4 and L5 points are called Trojans, after the three large asteroids
Agamemnon, Achilles, and Hector found orbiting in the L4 and L5 points of the Sun-Jupiter
system. No large asteroids have been found in the L4 and L5 points of the Sun-Earth system.
Personal aside : David T. Wilkinson
The image in fig. 9.13 comes from the education and outreach program of the Wilkinson
Microwave Anisotropy Probe (WMAP) project, a NASA mission, launched in 2001, which
has produced some of the most important recent data in cosmology. The project is named
in honor of David T. Wilkinson, who was a leading cosmologist at Princeton, and a founder
of the Cosmic Background Explorer (COBE) satellite (launched in 1989). WMAP was sent
to the L2 Lagrange point, on the night side of the earth, where it can constantly scan the
cosmos with an ultra-sensitive microwave detector, shielded by the earth from interfering
solar electromagnetic radiation. The L2 point is of course unstable, with a time scale of
about 23 days. Satellites located at such points must undergo regular course and attitude
corrections to remain situated.
During the summer of 1981, as an undergraduate at Princeton, I was a member of Wilkinsons gravity group, working under Jeff Kuhn and Ken Libbrecht. It was a pretty big
group and Dave everyone would call him Dave used to throw wonderful parties at his
home, where wed always play volleyball. I was very fortunate to get to know David Wilkinson a bit after working in his group that summer I took a class from him the following
year. He was a wonderful person, a superb teacher, and a world class physicist.
172
CHAPTER 9. CENTRAL FORCES AND ORBITAL MECHANICS
Chapter 10
Small Oscillations
10.1
Coupled Coordinates
We assume, for a set of n generalized coordinates {q1 , . . . , qn }, that the kinetic energy is a
quadratic function of the velocities,
T =
1
2
T (q1 , . . . , qn ) q q ,
(10.1)
where the sum on and from 1 to n is implied. For example, expressed in terms of polar
coordinates (r, , ), the matrix Tij is
1
0
0
(10.2)
T = m 0
r2
0 = T = 21 m r 2 + r 2 2 + r 2 sin2 2 .
0
0 r 2 sin2
The potential U (q1 , . . . , qn ) is assumed to be a function of the generalized coordinates alone:
U = U (q). A more general formulation of the problem of small oscillations is given in the
appendix, section 10.8.
The generalized momenta are
L
= T q ,
q
(10.3)
L
1 T
U
=
q q
.
q
2 q
q
(10.4)
p =
and the generalized forces are
F =
The Euler-Lagrange equations are then p = F , or
T
1 T
U
T q +
q q =
2 q
q
q
(10.5)
which is a set of coupled nonlinear second order ODEs. Here we are using the Einstein
summation convention, where we automatically sum over any and all repeated indices.
173
174
10.2
CHAPTER 10. SMALL OSCILLATIONS
Expansion about Static Equilibrium
Small oscillation theory begins with the identification of a static equilibrium {
q1 , . . . , qn },
which satisfies the n nonlinear equations
U
=0.
(10.6)
q q=q
Once an equilibrium is found (note that there may be more than one static equilibrium),
we expand about this equilibrium, writing
q q + .
(10.7)
The coordinates {1 , . . . , n } represent the displacements relative to equilibrium.
We next expand the Lagrangian to quadratic order in the generalized displacements, yielding
L = 12 T 21 V ,
where
T =
2T
q q
V =
q=
q
2U
q q
(10.8)
(10.9)
q=
q
Writing t for the row-vector (1 , . . . , n ), we may suppress indices and write
L=
t T 21 t V ,
1
2
(10.10)
where T and V are the constant matrices of eqn. 10.9.
10.3
Method of Small Oscillations
The idea behind the method of small oscillations is to effect a coordinate transformation
from the generalized displacements to a new set of coordinates , which render the
Lagrangian particularly simple. All that is required is a linear transformation,
= Ai i ,
(10.11)
where both and i run from 1 to n. The n n matrix Ai is known as the modal matrix.
With the substitution = A (hence t = t At , where Ati = Ai is the matrix transpose),
we have
L = 21 t At T A 12 t At V A .
(10.12)
We now choose the matrix A such that
At T A = I
t
A V A = diag
12 ,
... ,
n2
(10.13)
.
(10.14)
10.3. METHOD OF SMALL OSCILLATIONS
175
With this choice of A, the Lagrangian decouples:
L=
1
2
n
X
2 2 2 ,
i
i i
(10.15)
i=1
with the solution
i (t) = Ci cos(i t) + Di sin(i t) ,
(10.16)
where {C1 , . . . , Cn } and {D1 , . . . , Dn } are 2n constants of integration, and where no sum is
implied on i. Note that
= A1 = At T .
(10.17)
In terms of the original generalized displacements, the solution is
(t) =
n
X
i=1
n
o
Ai Ci cos(i t) + Di sin(i t) ,
(10.18)
and the constants of integration are linearly related to the initial generalized displacements
and generalized velocities:
Ci = Ati T (0)
(10.19)
Di = i1 Ati T (0) ,
(10.20)
again with no implied sum on i on the RHS of the second equation, and where we have
used A1 = At T, from eqn. 10.13. (The implied sums in eqn. 10.20 are over and .)
Note that the normal coordinates have unusual dimensions: [] = M L, where L is length
and M is mass.
10.3.1
Can you really just choose an A so that both these wonderful
things happen in 10.13 and 10.14?
Yes.
10.3.2
Er...care to elaborate?
Both T and V are symmetric matrices. Aside from that, there is no special relation between
them. In particular, they need not commute, hence they do not necessarily share any
eigenvectors. Nevertheless, they may be simultaneously diagonalized as per 10.13 and 10.14.
Heres why:
Since T is symmetric, it can be diagonalized by an orthogonal transformation. That
is, there exists a matrix O1 O(n) such that
O1t T O1 = Td ,
where Td is diagonal.
(10.21)
176
CHAPTER 10. SMALL OSCILLATIONS
We may safely assume that T is positive definite. Otherwise the kinetic energy can
become arbitrarily negative, which is unphysical. Therefore, one may form the matrix
1/2
which is the diagonal matrix whose entries are the inverse square roots of the
Td
1/2
corresponding entries of Td . Consider the linear transformation O1 Td . Its effect
on T is
1/2
1/2 t
=1.
(10.22)
O1 T O1 Td
Td
Since O1 and Td are wholly derived from T, the only thing we know about
e T1/2 Ot V O T1/2
V
1
1 d
d
(10.23)
is that it is explicitly a symmetric matrix. Therefore, it may be diagonalized by some
orthogonal matrix O2 O(n). As T has already been transformed to the identity, the
additional orthogonal transformation has no effect there. Thus, we have shown that
there exist orthogonal matrices O1 and O2 such that
1/2
O2t Td
1/2
O1t T O1 Td
O2
1/2
1/2
O2
O1t V O1 Td
O2t Td
=1
(10.24)
= diag (12 , . . . , n2 ) .
(10.25)
1/2
All that remains is to identify the modal matrix A = O1 Td
O2 .
Note that it is not possible to simultaneously diagonalize three symmetric matrices in general.
10.3.3
Finding the modal matrix
While the above proof allows one to construct A by finding the two orthogonal matrices O1
and O2 , such a procedure is extremely cumbersome. It would be much more convenient if
A could be determined in one fell swoop. Fortunately, this is possible.
We start with the equations of motion, T + V = 0. In component notation, we have
T + V = 0 .
(10.26)
We now assume that (t) oscillates with a single frequency , i.e. (t) = eit . This
results in a set of linear algebraic equations for the components :
2 T V = 0 .
(10.27)
These are n equations in n unknowns: one for each value of = 1, . . . , n. Because the
equations are homogeneous and linear, there is always a trivial solution = 0. In fact one
might think this is the only solution, since
2 T V = 0
?
=
= 2 T V
1
0=0.
(10.28)
10.3. METHOD OF SMALL OSCILLATIONS
177
However, this fails when the matrix 2 T V is defective1 , i.e. when
det 2 T V = 0 .
(10.29)
Since T and V are of rank n, the above determinant yields an nth order polynomial in 2 ,
whose n roots are the desired squared eigenfrequencies {12 , . . . , n2 }.
Once the n eigenfrequencies are obtained, the modal matrix is constructed as follows. Solve
the equations
n
X
(i)
i2 T V = 0
(10.30)
=1
which are a set of (n 1) linearly independent equations among the n components of the
eigenvector (i) . That is, there are n equations ( = 1, . . . , n), but one linear dependency
since det (i2 T V) = 0. The eigenvectors may be chosen to satisfy a generalized orthogonality relationship,
(j)
(i) T = ij .
(10.31)
To see this, let us duplicate eqn. 10.30, replacing i with j, and multiply both equations as
follows:
(i)
(10.32)
(j) i2 T V = 0
(j)
(10.33)
(i) j2 T V = 0 .
Using the symmetry of T and V, upon subtracting these equations we obtain
(i2
j2 )
n
X
(j)
(i) T = 0 ,
(10.34)
, =1
where the sums on i and j have been made explicit. This establishes that eigenvectors (i)
and (j) corresponding to distinct eigenvalues i2 6= j2 are orthogonal: ( (i) )t T (j) = 0.
For degenerate eigenvalues, the eigenvectors are not a priori orthogonal, but they may be
orthogonalized via application of the Gram-Schmidt procedure. The remaining degrees of
freedom - one for each eigenvector are fixed by imposing the condition of normalization:
q
(j)
(i)
(i)
(i)
(i)
(10.35)
=
(i) T = ij .
T
(i)
The modal matrix is just the matrix of eigenvectors: Ai = .
(i)
With the eigenvectors thusly normalized, we have
(j)
0 = (i) j2 T V
(j)
= j2 ij (i) V ,
(10.36)
with no sum on j. This establishes the result
At V A = diag 12 , . . . , n2 .
(10.37)
1
The label defective has a distastefully negative connotation. In modern parlance, we should instead refer
to such a matrix as determinantally challenged .
178
CHAPTER 10. SMALL OSCILLATIONS
10.4
Example: Masses and Springs
Two blocks and three springs are configured as in Fig. 17.6. All motion is horizontal. When
the blocks are at rest, all springs are unstretched.
Figure 10.1: A system of masses and springs.
(a) Choose as generalized coordinates the displacement of each block from its equilibrium
position, and write the Lagrangian.
(b) Find the T and V matrices.
(c) Suppose
m1 = 2m
m2 = m
k1 = 4k
k2 = k
k3 = 2k ,
Find the frequencies of small oscillations.
(d) Find the normal modes of oscillation.
(e) At time t = 0, mass #1 is displaced by a distance b relative to its equilibrium position.
I.e. x1 (0) = b. The other initial conditions are x2 (0) = 0, x 1 (0) = 0, and x 2 (0) = 0.
Find t , the next time at which x2 vanishes.
Solution
(a) The Lagrangian is
L = 21 m1 x 21 + 12 m2 x 22 21 k1 x21 12 k2 (x2 x1 )2 12 k3 x22
(b) The T and V matrices are
2T
=
Tij =
x i x j
m1 0
0 m2
2U
=
Vij =
xi xj
k1 + k2
k2
k2
k2 + k3
10.4. EXAMPLE: MASSES AND SPRINGS
179
(c) We have m1 = 2m, m2 = m, k1 = 4k, k2 = k, and k3 = 2k. Let us write 2 02 ,
p
where 0 k/m. Then
2 5
1
2T V = k
.
1
3
The determinant is
det ( 2 T V) = (22 11 + 14) k2
= (2 7) ( 2) k2 .
There are two roots: = 2 and + = 72 , corresponding to the eigenfrequencies
=
2k
m
+ =
7k
2m
~ (a) = 0. Plugging in = 2 we have
(d) The normal modes are determined from (a2 T V)
~ ()
for the normal mode
()
1
1 1
1
()
~
=
C
=
0
1
1 1
2()
Plugging in =
7
2
~ (+)
we have for the normal mode
2 1
1 12
1(+)
2(+)
(a)
The standard normalization i
~ (+) = C
=0
1
2
(b)
Tij j = ab gives
C =
1
3m
C+ =
1
.
6m
(10.38)
(e) The general solution is
!
1
x1
1
1
1
=A
cos( t) + B
cos(+ t) + C
sin( t) + D
sin(+ t) .
1
2
1
2
x2
The initial conditions x1 (0) = b, x2 (0) = x 1 (0) = x 2 (0) = 0 yield
A = 32 b
B = 13 b
C=0 ,
D=0.
Thus,
x1 (t) = 31 b 2 cos( t) + cos(+ t)
x2 (t) = 32 b cos( t) cos(+ t) .
180
CHAPTER 10. SMALL OSCILLATIONS
Figure 10.2: The double pendulum.
Setting x2 (t ) = 0, we find
cos( t ) = cos(+ t )
10.5
t = + t
t =
2
+ +
Example: Double Pendulum
As a second example, consider the double pendulum, with m1 = m2 = m and 1 = 2 = .
The kinetic and potential energies are
T = m2 12 + m2 cos(1 1 ) 1 2 + 21 m2 22
(10.39)
V = 2mg cos 1 mg cos 2 ,
leading to
T=
2m2 m2
m2 m2
Then
2
T V = m
with 0 =
V=
(10.40)
2mg
0
0
mg
2
2 2 202
2
2 02
(10.41)
(10.42)
2) 02 .
(10.43)
g/. Setting the determinant to zero gives
2( 2 02 )2 4 = 0
2 = (2
10.6. ZERO MODES
181
We find the unnormalized eigenvectors by setting (i2 T V ) (i) = 0. This gives
+ = C+
= C
(i)
+ 2
(10.44)
(j)
where C are constants. One can check T vanishes for i 6= j. We then normalize
(i) (i)
by demanding T = 1 (no sum on i), which determines the coefficients C =
q
1
(2
2)/m2 . Thus, the modal matrix is
2
+
1
A=
2+
10.6
2+
= 1
p
2
2 m
2
4+2 2
.
p
+ 42 2
(10.45)
Zero Modes
Recall Noethers theorem, which says that for every continuous one-parameter family of
coordinate transformations,
q q (q, )
q (q, = 0) = q ,
(10.46)
which leaves the Lagrangian invariant, i.e. dL/d = 0, there is an associated conserved
quantity,
X L q
d
satisfies
=0.
(10.47)
=
q
dt
=0
For small oscillations, we write q = q + , hence
X
k =
Ck ,
(10.48)
where k labels the one-parameter families (in the event there is more than one continuous
symmetry), and where
X
q
Ck =
T
.
(10.49)
k
=0
Therefore, we can define the (unnormalized) normal mode
k =
Ck ,
(10.50)
which satisfies k = 0. Thus, in systems with continuous symmetries, to each such continuous symmetry there is an associated zero mode of the small oscillations problem, i.e. a
mode with k2 = 0.
182
CHAPTER 10. SMALL OSCILLATIONS
Figure 10.3: Coupled oscillations of three masses on a frictionless hoop of radius R. All
three springs have the same force constant k, but the masses are all distinct.
10.6.1
Example of zero mode oscillations
The simplest example of a zero mode would be a pair of masses m1 and m2 moving frictionlessly along a line and connected by a spring of force constant k. We know from our
study of central forces that the Lagrangian may be written
L = 12 m1 x 21 + 12 m2 x 22 21 k(x1 x2 )2
= 1 M X 2 + 1 x 2 1 kx2 ,
2
(10.51)
where X = (m1 x1 + m2 x2 )/(m1 + m2 ) is the center of mass position, x = x1 x2 is the
relative coordinate, M = m1 + m2 is the total mass, and = m1 m2 /(m1 + m2 ) is the
reduced
= 02 x, where the oscillation frequency is
p mass. The relative coordinate obeys x
= 0, i.e. its oscillation frequency is
0 = k/. The center of mass coordinate obeys X
zero. The center of mass motion is a zero mode.
Another example is furnished by the system depicted in fig. 10.3, where three distinct masses
m1 , m2 , and m3 move around a frictionless hoop of radius R. The masses are connected to
their neighbors by identical springs of force constant k. We choose as generalized coordinates
the angles ( = 1, 2, 3), with the convention that
1 2 3 2 + 1 .
(10.52)
Let R be the equilibrium length for each of the springs. Then the potential energy is
o
n
U = 12 kR2 (2 1 )2 + (3 2 )2 + (2 + 1 3 )2
n
o
(10.53)
= 12 kR2 (2 1 )2 + (3 2 )2 + (2 + 1 3 )2 + 32 4 .
10.6. ZERO MODES
183
Note that the equilibrium angle enters only in an additive constant to the potential
energy. Thus, for the calculation of the equations of motion, it is irrelevant. It doesnt
matter whether or not the equilibrium configuration is unstretched ( = 2/3) or not
( 6= 2/3).
The kinetic energy is simple:
T = 21 R2 m1 21 + m2 22 + m3 23 .
The T and V matrices are then
m1 R 2
0
0
T= 0
m2 R 2
0
0
0
m3 R 2
2kR2 kR2 kR2
V = kR2 2kR2 kR2 .
kR2 kR2 2kR2
We then have
2 T V = kR2
(10.54)
12
2
22
2
32
(10.55)
(10.56)
We compute the determinant to find the characteristic polynomial:
P () = det( 2 T V)
=
12 22 32
1
12 22
1
22 32
1
12 32
4 + 3
1
1
1
+ 2+ 2
2
1
2
3
(10.57)
2 ,
where i2 k/mi . The equation P () = 0 yields a cubic equation in 2 , but clearly 2 is
a factor, and when we divide this out we obtain a quadratic equation. One root obviously
is 12 = 0. The other two roots are solutions to the quadratic equation:
q
2
2
2
2
2,3
= 12 + 22 + 32 12 12 22 + 12 22 32 + 21 12 32 .
(10.58)
To find the eigenvectors and the modal matrix, we set
2
(j)
j
2
1
1
2
1
1
j2
(j)
1
2
1 2 = 0 ,
22
(j)
2
j
3
2
1
1
2
(10.59)
Writing down the three coupled equations for the components of (j) , we find
2
2
2
j
j
j
(j)
(j)
(j)
3 1 =
3 2 =
3 3 .
2
2
2
1
2
3
We therefore conclude
(j)
2
= Cj
j
12
j2
22
j2
32
1
3
1
.
3
1
3
(10.60)
(10.61)
184
CHAPTER 10. SMALL OSCILLATIONS
(j)
(i)
The normalization condition T = ij then fixes the constants Cj :
"
2
2
2
2
2 #
2
j
j
j2
C = 1 .
+
m
+
m
3
m1
2
3
j
12
22
32
(10.62)
The Lagrangian is invariant under the one-parameter family of transformations
+
(10.63)
for all = 1, 2, 3. The associated conserved quantity is
=
X L
= R2 m1 1 + m2 2 + m3 3 ,
(10.64)
which is, of course, the total angular momentum relative to the center of the ring. Thus,
from = 0 we identify the zero mode as 1 , where
1 = C m1 1 + m2 2 + m3 3 ,
(10.65)
where C is a constant. Recall the relation = Ai i between the generalized displacements
and the normal coordinates i . We can invert this relation to obtain
t
i = A1
i = Ai T .
(10.66)
Here we have used the result At T A = 1 to write
A1 = At T .
(10.67)
This is a convenient result, because it means that if we ever need to express the normal
coordinates in terms of the generalized displacements, we dont have to invert any matrices
we just need to do one matrix multiplication. In our case here, the T matrix is diagonal,
so the multiplication is trivial. From eqns. 10.65 and 10.66, we conclude that the matrix
At T must have a first row which is proportional to (m1 , m2 , m3 ). Since these are the very
diagonal entries of T, we conclude that At itself must have a first row which is proportional
to (1, 1, 1), which means that the first column of A is proportional to (1, 1, 1). But this is
(1)
(1)
(1)
2
confirmed by eqn. 10.60 when we take j = 1, since j=1
= 0: 1 = 2 = 3 .
10.7
Chain of Mass Points
Next consider an infinite chain of identical masses, connected by identical springs of spring
constant k and equilibrium length a. The Lagrangian is
X
X
x 2n 12 k
(xn+1 xn a)2
L = 12 m
n
1
2m
X
n
u 2n
1
2k
X
n
(un+1 un )2 ,
(10.68)
10.7. CHAIN OF MASS POINTS
185
where un xn na b is the displacement from equilibrium of the nth mass. The constant
b is arbitrary. The Euler-Lagrange equations are
L
d L
= m
un =
dt u n
un
= k(un+1 un ) k(un un1 )
= k(un+1 + un1 2un ) .
(10.69)
Now let us assume that the system is placed on a large ring of circumference N a, where
N 1. Then un+N = un and we may shift to Fourier coefficients,
1 X iqan
un =
e
u
q
N q
1 X iqan
u
q =
e
un ,
N n
(10.70)
(10.71)
where qj = 2j/N a, and both sums are over the set j, n {1, . . . , N }. Expressed in terms
of the {
uq }, the equations of motion become
1 X iqna
e
u
n
u
q =
N n
k 1 X iqan
=
e
(un+1 + un1 2un )
m N n
k 1 X iqan iqa
e
(e
+ e+iqa 2) un
=
m N n
2k
q
sin2 12 qa u
=
m
(10.72)
Thus, the {
uq } are the normal modes of the system (up to a normalization constant), and
the eigenfrequencies are
2k
sin 21 qa .
q =
(10.73)
m
This means that the modal matrix is
Anq =
where weve included the
1
m
1
eiqan ,
Nm
(10.74)
factor for a proper normalization. (The normal modes them
= mu
q . For complex A, the normalizations are A TA = I
selves are then q = Aqn Tnn un
2 ).
and A VA = diag(12 , . . . , N
Note that
Tnn = m n,n
(10.75)
Vnn = 2k n,n k n,n +1 k n,n 1
(10.76)
186
CHAPTER 10. SMALL OSCILLATIONS
and that
N X
N
X
(A TA)qq =
Anq Tnn An q
n=1 n =1
N
1 X X iqan
e
m nn eiq an
N m n=1
n =1
1
N
N
X
ei(q q)an = qq ,
(10.77)
n=1
and
(A VA)qq =
N X
N
X
Anq Tnn An q
n=1 n =1
N
N
1 X X iqan
2k n,n k n,n +1 k n,n 1 eiq an
e
Nm
k 1
mN
n=1 n =1
N
X
i(q q)an
n=1
4k
sin2
=
m
1
2 qa
2 eiq a eiq a
qq = q2 qq
(10.78)
Since x
q+G = x
q , where G = 2
a , we may choose any set of q values such that no two are
separated by an integer multiple of G. The set of points {jG} with j Z is called the
reciprocal lattice. For a linear chain, the reciprocal lattice is itself a linear chain2 . One
natural set to choose is q a , a . This is known as the first Brillouin zone of the
reciprocal lattice.
Finally, we can write the Lagrangian itself in terms of the {uq }. One easily finds
L=
1
2
X
q
u
q u
q k
X
q
(1 cos qa) uq u
q ,
(10.79)
where the sum is over q in the first Brillouin zone. Note that
u
q = u
q+G = u
q .
This means that we can restrict the sum to half the Brillouin zone:
X
4k
2 1
1
u
q u
q
q u
q .
L = 2m
sin 2 qa u
m
(10.80)
(10.81)
q[0, a ]
For higher dimensional Bravais lattices, the reciprocal lattice is often different than the real space
(direct) lattice. For example, the reciprocal lattice of a face-centered cubic structure is a body-centered
cubic lattice.
10.7. CHAIN OF MASS POINTS
187
Now u
q and u
q may be regarded as linearly independent, as one regards complex variables
z and z . The Euler-Lagrange equation for u
q gives
d L
L
= 2 u
u
(10.82)
=
q
q q .
dt u
q
q
Extremizing with respect to u
q gives the complex conjugate equation.
10.7.1
Continuum limit
Let us take N , a 0, with L0 = N a fixed. Well write
un (t) u(x = na, t)
(10.83)
in which case
T =
1
2m
u 2n
V = 12 k
dx u 2
a
t
Z
dx u(x + a) u(x) 2 2
1
a
2k
a
a
1
2m
X
(un+1 un )2
n
Recognizing the spatial derivative above, we finally obtain
Z
L = dx L(u, t u, x u)
2
2
u
u
12
,
L = 12
t
x
(10.84)
(10.85)
(10.86)
where = m/a is the linear mass density and = ka is the tension3 . The quantity L is
the Lagrangian density; it depends on the field u(x, t) as well as its partial derivatives t u
and x u4 . The action is
S u(x, t) =
Ztb Zxb
dt dx L(u, t u, x u) ,
ta
(10.87)
xa
where {xa , xb } are the limits on the x coordinate. Setting S = 0 gives the Euler-Lagrange
equations
L
L
L
=0.
(10.88)
u
t (t u)
x (x u)
For our system, this yields the Helmholtz equation,
2u
1 2u
=
,
c2 t2
x2
3
4
For a proper limit, we demand and be neither infinite nor infinitesimal.
L may also depend explicitly on x and t.
(10.89)
188
CHAPTER 10. SMALL OSCILLATIONS
p
where c = / is the velocity of wave propagation. This is a linear equation, solutions of
which are of the form
u(x, t) = C eiqx eit ,
(10.90)
where
= cq .
(10.91)
Note that in the continuum limit a 0, the dispersion relation derived for the chain
becomes
4k
ka2 2
q2 =
sin2 12 qa
q = c2 q 2 ,
(10.92)
m
m
and so the results agree.
10.8
Appendix I : General Formulation
In the development in section 10.1, we assumed that the kinetic energy T is a homogeneous
function of degree 2, and the potential energy U a homogeneous function of degree 0, in
the generalized velocities q . However, weve encountered situations where this is not so:
problems with time-dependent holonomic constraints, such as the mass point on a rotating
hoop, and problems involving charged particles moving in magnetic fields. The general
Lagrangian is of the form
L=
1
2
T2 (q) q q + T1 (q) q + T0 (q) U1 (q) q U0 (q) ,
(10.93)
where the subscript 0, 1, or 2 labels the degree of homogeneity of each term in the generalized
velocities. The generalized momenta are then
p =
L
q
= T2 q + T1 U1
(10.94)
and the generalized forces are
F =
L
q
(T0 U0 )
q
(T1 U1 )
q
q +
1 T2
q q ,
2 q
(10.95)
and the equations of motion are again p = F . Once we solve
In equilibrium, we seek a time-independent solution of the form q (t) = q . This entails
(10.96)
U0 (q) T0 (q) = 0 ,
q q=q
which give us n equations in the n unknowns (q1 , . . . , qn ). We then write q = q + and
expand in the notionally small quantities . It is important to understand that we assume
and all of its time derivatives as well are small. Thus, we can expand L to quadratic order
in (, )
to obtain
L=
1
2
T 12 B
1
2
V ,
(10.97)
10.8. APPENDIX I : GENERAL FORMULATION
189
where
q) ,
T = T2 (
2 U0 T0
=
q q
q=
q
U1 T1
=2
q
. (10.98)
q=
q
Note that the T and V matrices are symmetric. The B term is new.
Now we can always write B = 21 (Bs +Ba ) as a sum over symmetric and antisymmetric parts,
with Bs = B + Bt and Ba = B Bt . Since,
Bs =
d 1 s
B
,
dt 2
(10.99)
any symmetric part to B contributes a total time derivative to L, and thus has no effect on
the equations of motion. Therefore, we can project B onto its antisymmetric part, writing
B =
U1 T1
q
U1 T1
q
!
(10.100)
q=
q
We now have
p =
= T +
1
2
B ,
(10.101)
and
F =
= 12 B V .
(10.102)
The equations of motion, p = F , then yield
T + B + V = 0 .
(10.103)
Let us write (t) = eit . We then have
2 T + i B V = 0 .
To solve eqn. 10.104, we set P () = 0, where P () = det Q() , with
Q() 2 T + i B V .
(10.104)
(10.105)
t ) for any matrix
Since T, B, and V are real-valued matrices, and since det(M ) = det(M
M , we can use Bt = B to obtain P () = P () and P ( ) = P () . This establishes
that if P () = 0, i.e. if is an eigenfrequency, then P () = 0 and P ( ) = 0, i.e.
and are also eigenfrequencies (and hence as well).
190
CHAPTER 10. SMALL OSCILLATIONS
10.9
Appendix II : Additional Examples
10.9.1
Right Triatomic Molecule
A molecule consists of three identical atoms located at the vertices of a 45 right triangle.
Each pair of atoms interacts by an effective spring potential, with all spring constants equal
to k. Consider only planar motion of this molecule.
(a) Find three zero modes for this system (i.e. normal modes whose associated eigenfrequencies vanish).
(b) Find the remaining three normal modes.
Solution
It is useful to choose the following coordinates:
(X1 , Y1 ) = (x1 , y1 )
(10.106)
(X2 , Y2 ) = (a + x2 , y2 )
(10.107)
(X3 , Y3 ) = (x3 , a + y3 ) .
(10.108)
The three separations are then
q
d12 = (a + x2 x1 )2 + (y2 y1 )2
(10.109)
= a + x2 x1 + . . .
q
(a + x3 x2 )2 + (a + y3 y2 )2
= 2 a 12 x3 x2 + 12 y3 y2 + . . .
d23 =
d13 =
(10.110)
(x3 x1 )2 + (a + y3 y1 )2
= a + y3 y1 + . . . .
(10.111)
The potential is then
2
2
2
U = 12 k d12 a + 12 k d23 2 a + 21 k d13 a
= 12 k x2 x1
2
+ 41 k x3 x2
2
+ 14 k y3 y2
2
2
21 k x3 x2 y3 y2 + 12 k y3 y1
(10.112)
(10.113)
10.9. APPENDIX II : ADDITIONAL EXAMPLES
191
Defining the row vector
t x1 , y1 , x2 , y2 , x3 , y3 ,
(10.114)
we have that U is a quadratic form:
U = 12 V = 21 t V ,
with
V = V
0
1
0
3
1 0
2
2 U
=
= k
q q eq.
0 21
0
0
0 21
0 1 21
(10.115)
21
12
1
2
1
2
1
2
1
2
21
12
1
2
21
1
2
(10.116)
3
2
The kinetic energy is simply
T = 12 m x 21 + y 12 + x 22 + y 22 + x 23 + y 32 ,
which entails
(10.117)
T = m .
(10.118)
(b) The three zero modes correspond to x-translation, y-translation, and rotation. Their
eigenvectors, respectively, are
1
0
1
0
1
1
1
1
1
1
0
1 .
=
,
=
(10.119)
1 =
2
3
3m
3m
2 3m
0
1
2
1
0
2
0
1
1
To find the unnormalized rotation vector, we find the CM of the triangle, located at a3 , a3 ,
and sketch orthogonal displacements z (Ri RCM ) at the position of mass point i.
(c) The remaining modes may be determined by symmetry, and are given by
1
1
1
1
1
1
1
2
1
1
1
0
,
, = , =
4 =
5
6
2 m 1
2 m 0
2 3m
1
1
0
1
0
(10.120)
192
CHAPTER 10. SMALL OSCILLATIONS
Figure 10.4: Normal modes of the 45 right triangle. The yellow circle is the location of the
CM of the triangle.
with
1 =
k
m
2 =
2k
m
3 =
3k
.
m
(10.121)
Since T = m1 is a multiple of the unit matrix, the orthogonormality relation ia Tij jb = ab
entails that the eigenvectors are mutually orthogonal in the usual dot product sense, with
a b = m1 ab . One can check that the eigenvectors listed here satisfy this condition.
The simplest of the set {4 , 5 , 6 } to find is the uniform dilation 6 , sometimes called
the breathing mode. This must keep the triangle in the same shape, which means that
the deviations at each mass point are proportional to the distance to the CM. Next, it is
simplest to find 4 , in which the long and short sides of the triangle oscillate out of phase.
Finally, the mode 5 must be orthogonal to all the remaining modes. No heavy lifting (e.g.
Mathematica) is required!
10.9.2
Triple Pendulum
Consider a triple pendulum consisting of three identical masses m and three identical rigid
massless rods of length , as depicted in Fig. 10.5.
(a) Find the T and V matrices.
(b) Find the equation for the eigenfrequencies.
10.9. APPENDIX II : ADDITIONAL EXAMPLES
193
Figure 10.5: The triple pendulum.
(c) Numerically solve the eigenvalue equation for ratios a2 /02 , where 0 =
three normal modes.
g/. Find the
Solution
The Cartesian coordinates for the three masses are
y1 = cos 1
x1 = sin 1
x2 = sin 1 + sin 2
y2 = cos 1 cos 2
x3 = sin 1 + sin 2 + sin 3
y3 = cos 1 cos 2 cos 3 .
By inspection, we can write down the kinetic energy:
T = 12 m x 21 + y 12 + x 22 + y 22 + x 33 + y 32
n
= 12 m 2 3 12 + 2 22 + 32 + 4 cos(1 2 ) 1 2
+ 2 cos(1 3 ) 1 3 + 2 cos(2 3 ) 2 3
The potential energy is
n
o
U = mg 3 cos 1 + 2 cos 2 + cos 3 ,
and the Lagrangian is L = T U :
n
L = 21 m 2 3 12 + 2 22 + 32 + 4 cos(1 2 ) 1 2 + 2 cos(1 3 ) 1 3
o
n
o
+ 2 cos(2 3 ) 2 3 + mg 3 cos 1 + 2 cos 2 + cos 3 .
194
CHAPTER 10. SMALL OSCILLATIONS
The canonical momenta are given by
n
o
L
1 =
= m 2 3 1 + 2 2 cos(1 2 ) + 3 cos(1 3 )
1
n
o
L
= m 2 2 2 + 2 1 cos(1 2 ) + 3 cos(2 3 )
2 =
2
n
o
L
3 =
= m 2 3 + 1 cos(1 3 ) + 2 cos(2 3 ) .
2
The only conserved quantity is the total energy, E = T + U .
(a) As for the T and V matrices, we have
3 2 1
= m2 2 2 1
=
=0
1 1 1
2T
T
and
V
3 0 0
=
= mg 0 2 0 .
=0
0 0 1
2U
(b) The eigenfrequencies are roots of the equation det ( 2 T V) = 0. Defining 0
we have
3( 2 02 )
2 2
2
2( 2 02 )
2
2 T V = m2 2 2
2
2
2
2
( 0 )
g/,
and hence
i
i
h
h
det ( 2 T V) = 3( 2 02 ) 2( 2 02 )2 4 2 2 2 2 ( 2 02 ) 4
h
i
+ 2 2 4 2 2 ( 2 02 )
= 6 ( 2 02 )3 9 4 ( 2 02 ) + 4 6
= 6 9 02 4 + 18 04 2 6 06 .
(c) The equation for the eigenfrequencies is
3 9 2 + 18 6 = 0 ,
(10.122)
22 = 2.29428 02
where 2 = 02 . This is a cubic equation in . Numerically solving for the roots, one finds
12 = 0.415774 02
I find the (unnormalized) eigenvectors to be
1
1
1 = 1.2921 , 2 = 0.35286
2.3981
1.6312
32 = 6.28995 02 .
(10.123)
1
3 = 1.6450 .
0.76690
(10.124)
10.9. APPENDIX II : ADDITIONAL EXAMPLES
10.9.3
195
Equilateral Linear Triatomic Molecule
Consider the vibrations of an equilateral triangle of mass points, depicted in figure 10.6 .
The system is confined to the (x, y) plane, and in equilibrium all the strings are unstretched
and of length a.
Figure 10.6: An equilateral triangle of identical mass points and springs.
(a) Choose as generalized coordinates the Cartesian displacements (xi , yi ) with respect to
equilibrium. Write down the exact potential energy.
(b) Find the T and V matrices.
(c) There are three normal modes of oscillation for which the corresponding eigenfrequencies
all vanish: a = 0. Write down these modes explicitly, and provide a physical interpretation
for why a = 0. Since this triplet is degenerate, there is no unique answer any linear
combination will also serve as a valid zero mode. However, if you think physically, a
natural set should emerge.
(d) The three remaining modes all have finite oscillation frequencies. They correspond to
distortions of the triangular shape. One such mode is the breathing mode in which the
triangle uniformly expands and contracts. Write down the eigenvector associated with this
normal mode and compute its associated oscillation frequency.
(e) The fifth and sixth modes are degenerate. They must be orthogonal (with respect to
the inner product defined by T) to all the other modes. See if you can figure out what these
modes are, and compute their oscillation frequencies. As in (a), any linear combination of
these modes will also be an eigenmode.
(f) Write down your full expression for the modal matrix Aai , and check that it is correct
by using Mathematica.
196
CHAPTER 10. SMALL OSCILLATIONS
Figure 10.7: Zero modes of the mass-spring triangle.
Solution
Choosing as generalized coordinates the Cartesian displacements relative to equilibrium, we
have the following:
#1 : x1 , y1
#2 : a + x2 , y2
#3 : 12 a + x3 , 23 a + y3 .
Let dij be the separation of particles i and j. The potential energy of the spring connecting
them is then
1
2
k (dij a)2 .
2
+ y2 y1
2
2
= 12 a + x3 x2 + 23 a + y3 y2
2
2
= 21 a + x3 x1 + 23 a + y3 y1 .
d212 = a + x2 x1
d223
d213
2
The full potential energy is
U=
1
2
2
2
2
k d12 a + 12 k d23 a + 12 k d13 a .
(10.125)
This is a cumbersome expression, involving square roots.
To find T and V, we need to write T and V as quadratic forms, neglecting higher order
terms. Therefore, we must expand dij a to linear order in the generalized coordinates.
10.9. APPENDIX II : ADDITIONAL EXAMPLES
197
Figure 10.8: Finite oscillation frequency modes of the mass-spring triangle.
This results in the following:
d12 = a + x2 x1 + . . .
d23 = a 12 x3 x2 + 23 y3 y2 + . . .
d13 = a + 12 x3 x1 + 23 y3 y1 + . . . .
Thus,
U=
1
2
Defining
2
+ 81 k x2 x3 3 y2 + 3 y3
2
+ 18 k x3 x1 + 3 y3 3 y1 + higher order terms .
k x2 x1
we may now read off
2
q1 , q2 , q3 , q4 , q5 , q6 = x1 , y1 , x2 , y2 , x3 , y3 ,
5/4
3/4
3/4
3/4
0
0
1
3/4
0
5/4
= k
=
0
3/4
3/4
q q q
0
1/4 3/4 1/4
3/4
3/4 3/4
3/4
3/4
2U
The T matrix is trivial. From
T = 12 m x 21 + y 12 + x 22 + y 22 + x 23 + y 32 .
1/4 3/4
3/4 3/4
1/4
3/4
3/4
3/4
1/2
0
0
3/2
198
CHAPTER 10. SMALL OSCILLATIONS
Figure 10.9: John Henry, statue by Charles O. Cooper (1972). Now the man that invented
the steam drill, he thought he was mighty fine. But John Henry drove fifteen feet, and the
steam drill only made nine. - from The Ballad of John Henry.
we obtain
Tij =
2T
= m ij ,
qi qj
and T = m I is a multiple of the unit matrix.
The zero modes are depicted graphically in figure 10.7. Explicitly, we have
1
1
x =
3m 0
1
0
1
0
y =
3m 1
rot
1/2
3/2
=
3m
1/2
3/2
1
0
That these are indeed zero modes may be verified by direct multiplication:
V x.y = V rot = 0 .
(10.126)
The three modes with finite oscillation frequency are depicted graphically in figure 10.8.
10.10. ASIDE : CHRISTOFFEL SYMBOLS
Explicitly, we have
1/2
3/2
A =
3m
1/2
3/2
1
0
B =
3m
199
3/2
1/2
3/2
1/2
0
1
3/2
dil
=
3m
1/2
3/2
1/2
0
1
The oscillation frequencies of these modes are easily checked by multiplying the eigenvectors
by the matrix V. Since T = m I is diagonal, we have V a = ma2 a . One finds
r
r
3k
3k
A = B =
, dil =
.
2m
m
Mathematica? I dont need no stinking Mathematica.
10.10
Aside : Christoffel Symbols
The coupled equations in eqn. 10.5 may be written in the form
q + q q = F ,
with
1
2
1
T
T
T
T
+
q
q
q
and
1
F = T
(10.127)
U
.
q
(10.128)
(10.129)
The components of the rank-three tensor are known as Christoffel symbols, in the case
where T (q) defines a metric on the space of generalized coordinates.
200
CHAPTER 10. SMALL OSCILLATIONS
Chapter 11
Elastic Collisions
11.1
Center of Mass Frame
A collision or scattering event is said to be elastic if it results in no change in the internal
state of any of the particles involved. Thus, no internal energy is liberated or captured in
an elastic process.
Consider the elastic scattering of two particles. Recall the relation between laboratory
coordinates {r1 , r2 } and the CM and relative coordinates {R, r}:
R=
m2
r
m1 + m2
m1
r2 = R
r
m1 + m2
m1 r1 + m2 r2
m1 + m2
r1 = R +
r = r1 r2
(11.1)
(11.2)
If external forces are negligible, the CM momentum P = M R is constant, and therefore the
frame of reference whose origin is tied to the CM position is an inertial frame of reference.
In this frame,
m1 v
m2 v
,
v2CM =
,
(11.3)
v1CM =
m1 + m2
m1 + m2
where v = v1 v2 = v1CM v2CM is the relative velocity, which is the same in both L and
CM frames. Note that the CM momenta satisfy
CM
pCM
1 = m1 v1 = v
(11.4)
CM
pCM
2 = m2 v2 = v ,
(11.5)
CM
= 0 and the total
where = m1 m2 /(m1 + m2 ) is the reduced mass. Thus, pCM
1 + p2
momentum in the CM frame is zero. We may then write
pCM
1 p0 n
pCM
2 p0 n
201
E CM =
p20
p2
p2
+ 0 = 0 .
2m1 2m2
2
(11.6)
202
CHAPTER 11. ELASTIC COLLISIONS
Figure 11.1: The scattering of two hard spheres of radii a and b The scattering angle is .
The energy is evaluated when the particles are asymptotically far from each other, in which
case the potential energy is assumed to be negligible. After the collision, energy and momentum conservation require
p1
CM
p0 n
p2
CM
p0 n
CM
= E CM =
p20
.
2
(11.7)
The angle between n and n is the scattering angle :
n n cos .
(11.8)
The value of depends on the details of the scattering process, i.e. on the interaction
potential U (r). As an example, consider the scattering of two hard spheres, depicted in
Fig. 11.1. The potential is
(
if r a + b
(11.9)
U (r) =
0
if r > a + b .
Clearly the scattering angle is = 20 , where 0 is the angle between the initial
momentum of either sphere and a line containing their two centers at the moment of contact.
There is a simple geometric interpretation of these results, depicted in Fig. 11.2. We have
p 1 = m1 V + p 0 n
p1 = m1 V + p0 n
(11.10)
p 2 = m2 V p 0 n
.
p2 = m2 V p0 n
(11.11)
and p0 n
must
So draw a circle of radius p0 whose center is the origin. The vectors p0 n
both lie along this circle. We define the angle between V and n:
V n = cos .
(11.12)
11.1. CENTER OF MASS FRAME
203
Figure 11.2: Scattering of two particles of masses m1 and m2 . The scattering angle is the
and n
.
angle between n
It is now an exercise in geometry, using the law of cosines, to determine everything of
interest in terms of the quantities V , v, , and . For example, the momenta are
p1 =
p1 =
p2 =
p2
q
q
m21 V 2 + 2 v 2 + 2m1 V v cos
(11.13)
m21 V 2 + 2 v 2 + 2m1 V v cos( )
(11.14)
m22 V 2 + 2 v 2 2m2 V v cos
(11.15)
m22 V 2 + 2 v 2 2m2 V v cos( ) ,
(11.16)
and the scattering angles are
1 = tan
2 = tan1
v sin
v cos + m1V
v sin
v cos m2 V
!
!
+ tan
+ tan1
v sin( )
v cos( ) + m1V
v sin( )
v cos( ) m2 V
!
!
(11.17)
(11.18)
204
CHAPTER 11. ELASTIC COLLISIONS
Figure 11.3: Scattering when particle 2 is initially at rest.
If particle 2, say, is initially at rest, the situation is somewhat simpler. In this case, V =
m1 V /(m1 + m2 ) and m2 V = v, which means the point B lies on the circle in Fig. 11.3
(m1 6= m2 ) and Fig. 11.4 (m1 = m2 ). Let 1,2 be the angles between the directions of
motion after the collision and the direction V of impact. The scattering angle is the
angle through which particle 1 turns in the CM frame. Clearly
tan 1 =
m1
m2
sin
+ cos
2 = 21 ( ) .
We can also find the speeds v1 and v2 in terms of v and , from
2
2
2
1
1
2m
p1 = p20 + m
m2 p0
m2 p0 cos( )
(11.19)
(11.20)
and
p22 = 2 p20 (1 cos ) .
These equations yield
p
m21 + m22 + 2m1 m2 cos
v
v1 =
m1 + m2
v2 =
(11.21)
2m1 v
sin( 21 ) .
m1 + m2
(11.22)
Figure 11.4: Scattering of identical mass particles when particle 2 is initially at rest.
11.2. CENTRAL FORCE SCATTERING
205
Figure 11.5: Repulsive (A,C) and attractive (B,D) scattering in the lab (A,B) and CM
(C,D) frames, assuming particle 2 starts from rest in the lab frame. (From Barger and
Olsson.)
2
The angle max from Fig. 11.3(b) is given by sin max = m
m1 . Note that when m1 = m2
we have 1 + 2 = . A sketch of the orbits in the cases of both repulsive and attractive
scattering, in both the laboratory and CM frames, in shown in Fig. 11.5.
11.2
Central Force Scattering
Consider a single particle of mass movng in a central potential U (r), or a two body central
force problem in which is the reduced mass. Recall that
dr
d dr
dr
=
= 2
,
dt
dt d
r d
(11.23)
and therefore
2
E = 12 r 2 +
+ U (r)
2r 2
2
2
dr
2
=
+
+ U (r) .
2r 4 d
2r 2
Solving for
dr
d ,
(11.24)
we obtain
dr
=
d
2r 4
E
U
(r)
r2 ,
2
(11.25)
206
CHAPTER 11. ELASTIC COLLISIONS
Figure 11.6: Scattering in the CM frame. O is the force center and P is the point of
periapsis. The impact parameter is b, and is the scattering angle. 0 is the angle through
which the relative coordinate moves between periapsis and infinity.
Consulting Fig. 11.6, we have that
0 =
2
rp
r2
dr
,
E Ueff (r)
(11.26)
where rp is the radial distance at periapsis, and where
Ueff (r) =
2
+ U (r)
2r 2
(11.27)
is the effective potential, as before. From Fig. 11.6, we conclude that the scattering angle
is
(11.28)
= 20 .
It is convenient to define the impact parameter b as the distance of the asymptotic trajectory
from a parallel line containing the force center. The geometry is shown again in Fig. 11.6.
Note that the energy and angular momentum, which are conserved, can be evaluated at
infinity using the impact parameter:
2
E = 12 v
= v b .
(11.29)
Substituting for (b), we have
0 (E, b) =
rp
dr
q
r2
b
1
b2
r2
U (r)
E
(11.30)
In physical applications, we are often interested in the deflection of a beam of incident
particles by a scattering center. We define the differential scattering cross section d by
d =
# of particles scattered into solid angle d per unit time
.
incident flux
(11.31)
11.2. CENTRAL FORCE SCATTERING
207
Figure 11.7: Geometry of hard sphere scattering.
Now for particles of a given energy E there is a unique relationship between the scattering
angle and the impact parameter b, as we have just derived in eqn. 11.30. The differential
solid angle is given by d = 2 sin d, hence
b db d ( 21 b2 )
d
=
=
.
d
sin d d cos
(11.32)
d
d
Note that d
has dimensions of area. The integral of d
over all solid angle is the total
scattering cross section,
Z
d
.
(11.33)
T = 2 d sin
d
0
11.2.1
Hard sphere scattering
Consider a point particle scattering off a hard sphere of radius a, or two hard spheres of
radii a1 and a2 scattering off each other, with a a1 + a2 . From the geometry of Fig. 11.7,
we have b = a sin 0 and 0 = 12 ( ), so
b2 = a2 sin2
1
2
12 ) = 21 a2 (1 + cos ) .
(11.34)
We therefore have
d ( 12 b2 )
d
=
=
d
d cos
1
4
a2
(11.35)
and T = a2 . The total scattering cross section is simply the area of a sphere of radius a
projected onto a plane perpendicular to the incident flux.
208
11.2.2
CHAPTER 11. ELASTIC COLLISIONS
Rutherford scattering
Consider scattering by the Kepler potential U (r) = kr . We assume that the orbits are
unbound, i.e. they are Keplerian hyperbolae with E > 0, described by the equation
r() =
a (2 1)
1 + cos
cos 0 =
1
.
(11.36)
Recall that the eccentricity is given by
2E2
=1+
=1+
k2
2
bv
k
2
(11.37)
We then have
bv
k
2
= 2 1
= sec2 0 1 = tan2 0 = ctn2
Therefore
b() =
We finally obtain
which is the same as
k
ctn
2
v
1
2
1
2
k
4E
2
csc4
1
2
(11.38)
(11.39)
d ( 12 b2 )
k 2 d ctn2 12
1
d
=
=
2
d
d cos
2 v
d cos
2
d
k
1 + cos
1
=
2
2 v
d cos 1 cos
2
k
=
csc4 12 ,
2
2v
d
=
d
(11.40)
(11.41)
d
Since d
4 as 0, the total cross section T diverges! This is a consequence of
the long-ranged nature of the Kepler/Coulomb potential. In electron-atom scattering, the
Coulomb potential of the nucleus is screened by the electrons of the atom, and the 1/r
behavior is cut off at large distances.
11.2.3
Transformation to laboratory coordinates
We previously derived the relation
tan =
sin
,
+ cos
(11.42)
11.2. CENTRAL FORCE SCATTERING
209
1
where 1 is the scattering angle for particle 1 in the laboratory frame, and = m
m2
is the ratio of the masses. We now derive the differential scattering cross section in the
laboratory frame. To do so, we note that particle conservation requires
d
d
2 sin d =
2 sin d ,
(11.43)
d L
d CM
which says
d
d
=
L
d
d
CM
d cos
.
d cos
(11.44)
From
cos =
1 + tan2
+ cos
=p
,
1 + 2 + 2 cos
we derive
and, accordingly,
d
d
1 + cos
d cos
=
3/2
d cos
1 + 2 + 2 cos
L
1 + 2 + 2 cos
=
1 + cos
3/2
d
.
d CM
(11.45)
(11.46)
(11.47)
210
CHAPTER 11. ELASTIC COLLISIONS
Chapter 12
Noninertial Reference Frames
12.1
Accelerated Coordinate Systems
A reference frame which is fixed with respect to a rotating rigid body is not inertial. The
parade example of this is an observer fixed on the surface of the earth. Due to the rotation of
the earth, such an observer is in a noninertial frame, and there are corresponding corrections
to Newtons laws of motion which must be accounted for in order to correctly describe
mechanical motion in the observers frame. As is well known, these corrections involve
fictitious centrifugal and Coriolis forces.
Consider an inertial frame with a fixed set of coordinate axes e , where runs from 1 to
d, the dimension of space. Any vector A may be written in either basis:
X
X
(12.1)
A e ,
A=
A e =
where A = A e and A = A e are projections onto the different coordinate axes. We
may now write
X dA
dA
e
=
dt inertial
dt
X dA
i
dt
e +
d
e
.
dt
(12.2)
The first term on the RHS is (dA/dt)body , the time derivative of A along body-fixed axes,
i.e. as seen by an observer rotating with the body. But what is d
ei /dt? Well, we can always
expand it in the {
ei } basis:
X
e e .
(12.3)
d
e =
d e d d
j
Note that d = d is antisymmetric, because
0 = d e e = d + d ,
211
(12.4)
212
CHAPTER 12. NONINERTIAL REFERENCE FRAMES
Figure 12.1: Reference frames related by both translation and rotation.
because e e = is a constant. Now we may define d12 d3 , et cyc., so that
d =
which yields
d
,
dt
d
e
= e .
dt
Finally, we obtain the important result
dA
dt
=
inertial
dA
dt
body
(12.5)
(12.6)
+A
(12.7)
which is valid for any vector A.
Applying this result to the position vector r, we have
dr
dr
=
+r .
dt inertial
dt body
Applying it twice,
2
d r
=
dt2 inertial
!
!
d
d
+
+ r
dt body
dt body
2
d r
dr
d
=
r
+
2
+
+ ( r) .
dt2 body
dt
dt body
(12.8)
(12.9)
Note that d/dt appears with no inertial or body label. This is because, upon invoking
eq. 12.7,
d
d
=
+ ,
(12.10)
dt inertial
dt body
12.1. ACCELERATED COORDINATE SYSTEMS
213
and since = 0, inertial and body-fixed observers will agree on the value of inertial =
body .
12.1.1
Translations
Suppose that frame K moves with respect to an inertial frame K 0 , such that the origin of
K lies at R(t). Suppose further that frame K rotates with respect to K, but shares the
same origin (see Fig. 12.1). Consider the motion of an object lying at position relative
to the origin of K 0 , and r relative to the origin of K/K . Thus,
=R+r ,
(12.11)
and
d
dt
d2
dt2
inertial
inertial
dr
=
+
+r
dt body
inertial
2
2
dR
d r
d
r
=
+
+
2
2
dt inertial
dt body
dt
dr
+ 2
+ ( r) .
dt body
dR
dt
(12.12)
(12.13)
Here, is the angular velocity in the frame K or K .
12.1.2
Motion on the surface of the earth
The earth both rotates about its axis and orbits the Sun. If we add the infinitesimal effects
of the two rotations,
dr1 = 1 r dt
dr2 = 2 (r + dr1 ) dt
dr = dr1 + dr2
= (1 + 2 ) dt r + O (dt)2 .
Thus, infinitesimal rotations add. Dividing by dt, this means that
X
=
i ,
(12.14)
(12.15)
where the sum is over all the rotations. For the earth, = rot + orb .
The rotation about earths axis, rot has magnitude rot = 2/(1 day) = 7.29
105 s1 . The radius of the earth is Re = 6.37 103 km.
214
CHAPTER 12. NONINERTIAL REFERENCE FRAMES
The orbital rotation about the Sun, orb has magnitude orb = 2/(1 yr) = 1.99
107 s1 . The radius of the earth is ae = 1.50 108 km.
Thus, rot /orb = Torb /Trot = 365.25, which is of course the number of days (i.e. rotational
periods) in a year (i.e. orbital period). There is also a very slow precession of the earths
axis of rotation, the period of which is about 25,000 years, which we will ignore. Note = 0
for the earth. Thus, applying Newtons second law and then invoking eq. 12.14, we arrive
at
2
2
d R
dr
d r
(tot)
=F
m
2m
m ( r) , (12.16)
m
2
2
dt earth
dt
dt earth
Sun
where = rot + orb , and where R
Sun is the acceleration of the center of the earth around
the Sun, assuming the Sun-fixed frame to be inertial. The force F (tot) is the total force on
the object, and arises from three parts: (i) gravitational pull of the Sun, (ii) gravitational
pull of the earth, and (iii) other earthly forces, such as springs, rods, surfaces, electric fields,
etc.
On the earths surface, the ratio of the Suns gravity to the earths is
GM m
F
=
Fe
a2e
GMe m
M
=
Re2
Me
Re
ae
2
6.02 104 .
(12.17)
In fact, it is clear that the Suns field precisely cancels with the term m R
Sun at the earths
center, leaving only gradient contributions of even lower order, i.e. multiplied by Re /ae
4.25 105 . Thus, to an excellent approximation, we may neglect the Sun entirely and
write
F
dr
d2 r
=
+ g 2
( r)
(12.18)
2
dt
m
dt
Note that weve dropped the earth label here and henceforth. We define g = GMe r/r 2 ,
the acceleration due to gravity; F is the sum of all earthly forces other than the earths
gravity. The last two terms on the RHS are corrections to mr = F due to the noninertial
frame of the earth, and are recognized as the Coriolis and centrifugal acceleration terms,
respectively.
12.2
Spherical Polar Coordinates
}
varies with position. In terms of the body-fixed
The locally orthonormal triad {
r , ,
y,
z},
we have
triad {x,
+ sin sin y + cos z
r = sin cos x
+ cos sin y sin z
= cos cos x
= sin x
+ cos y .
(12.19)
(12.20)
(12.21)
12.2. SPHERICAL POLAR COORDINATES
215
}.
Figure 12.2: The locally orthonormal triad {
r , ,
}
and {x,
y,
z},
we obtain
Inverting the relation between the triads {
r , ,
= sin cos r + cos cos sin
x
y = sin sin r + cos sin + cos
(12.22)
z = cos r sin .
(12.24)
(12.23)
The differentials of these unit vectors are
d
d
r = d + sin
d
d = r d + cos
= sin r d cos d .
d
(12.25)
(12.26)
(12.27)
Thus,
d
r r = r r + r r
dt
.
= r r + r + r sin
r =
(12.28)
If we differentiate a second time, we find, after some tedious accounting,
r = r r 2 r sin2 2 r + 2 r + r r sin cos 2
.
+ 2 r sin + 2 r cos + r sin
(12.29)
216
12.3
CHAPTER 12. NONINERTIAL REFERENCE FRAMES
Centrifugal Force
One major distinction between the Coriolis and centrifugal forces is that the Coriolis force
acts only on moving particles, whereas the centrifugal force is present even when r = 0.
Thus, the equation for stationary equilibrium on the earths surface is
mg + F m ( r) = 0 ,
(12.30)
involves the centrifugal term. We can write this as F + me
g = 0, where
GMe r
( r)
r2
= g0 2 Re sin2 r + 2 Re sin cos ,
ge =
(12.31)
(12.32)
where g0 = GMe /Re2 = 980 cm/s2 . Thus, on the equator, g = g0 2 Re r, with
2 Re 3.39 cm/s2 , a small but significant correction. Thus, you weigh less on the equator.
This means that a plumb bob suspended from a general
e along .
Note also the term in g
point above the earths surface wont point exactly toward the earths center. Moreover, if
the earth were replaced by an equivalent mass of fluid, the fluid would rearrange itself so
e. Indeed, the earth (and Sun) do exhibit
as to make its surface locally perpendicular to g
quadrupolar distortions in their mass distributions both are oblate spheroids. In fact, the
observed difference g( = 2 ) g( = 0) 5.2 cm/s2 , which is 53% greater than the navely
expected value of 3.39 cm/s2 . The earths oblateness enhances the effect.
12.3.1
Rotating tube of fluid
Consider a cylinder filled with a liquid, rotating with angular frequency about its sym In steady state, the fluid is stationary in the rotating frame, and we may
metry axis z.
write, for any given element of fluid
0 = f + g 2 z (z r) ,
(12.33)
where f is the force per unit mass on the fluid element. Now consider a fluid element on
the surface. Since there is no static friction to the fluid, any component of f parallel to
the fluids surface will cause the fluid to flow in that direction. This contradicts the steady
where n
is the local unit normal
state assumption. Therefore, we must have f = f n,
to the fluid surface. We write the equation for the fluids surface as z = z(). Thus, with
Newtons second law yields
r = + z() z,
= g z 2 ,
f n
(12.34)
where g = g z is assumed. From this, we conclude that the unit normal to the fluid surface
and the force per unit mass are given by
g z 2
() = p
n
g2 + 4 2
f () =
g2 + 4 2 .
(12.35)
12.4. THE CORIOLIS FORCE
217
Figure 12.3: A rotating cylinder of fluid.
Now suppose r(, ) = + z() z is a point on the surface of the fluid. We have that
d ,
dr = d + z () z d +
(12.36)
dz
d, which follows from eqn. 12.25 after
, and where we have used d =
where z = d
dr = 0, which says
setting = 2 . Now dr must lie along the surface, therefore n
dz
= 2 .
d
(12.37)
Integrating this equation, we obtain the shape of the surface:
z() = z0 +
12.4
2 2
.
2g
(12.38)
The Coriolis Force
According to (12.18), the acceleration
The Coriolis force is given by FCor = 2m r.
e an orthogonal component is generated by the
of a free particle (F = 0) isnt along g
Coriolis force. To actually solve the coupled equations of motion is difficult because the
}
change with position, and hence with time. The following standard
unit vectors {
r , ,
problem highlights some of the effects of the Coriolis and centrifugal forces.
PROBLEM: A cannonball is dropped from the top of a tower of height h located at a northerly
latitude of . Assuming the cannonball is initially at rest with respect to the tower, and
neglecting air resistance, calculate its deflection (magnitude and direction) due to (a) centrifugal and (b) Coriolis forces by the time it hits the ground. Evaluate for the case h = 100
m, = 45 . The radius of the earth is Re = 6.4 106 m.
SOLUTION: The equation of motion for a particle near the earths surface is
r = 2 r g0 r ( r) ,
(12.39)
218
CHAPTER 12. NONINERTIAL REFERENCE FRAMES
with = 2/(24 hrs) = 7.3 105 rad/s. Here, g0 = GMe /Re2 = 980 cm/s2 .
where = z,
}
and write
We use a locally orthonormal coordinate system {
r , ,
+ (Re + z) r ,
r = x + y
(12.40)
where Re = 6.4 106 m is the radius of the earth. Expressing z in terms of our chosen
orthonormal triad,
z = cos r sin ,
(12.41)
where = 2 is the polar angle, or colatitude. Since the height of the tower and
the deflections are all very small on the scale of Re , we may regard the orthonormal triad
as fixed and time-independent. (In general, these unit vectors change as a function of r.)
+ z r, and we find
Thus, we have r x + y
y sin r
z r = y cos + (x cos + z sin )
(12.42)
( r) = 2 Re sin cos 2 Re sin2 r ,
(12.43)
where we neglect the O(z) term in the second equation, since z Re .
The equation of motion, written in components, is then
x
= 2 cos y + 2 Re sin cos
(12.44)
y = 2 cos x 2 sin z
(12.45)
z = g0 + 2 sin y + 2 Re sin2 .
(12.46)
While these (inhomogeneous) equations are linear, they also are coupled, so an exact analytical solution is not trivial to obtain (but see below). Fortunately, the deflections are
small, so we can solve this perturbatively. We write x = x(0) + x, etc., and solve to lowest
order by including only the g0 term on the RHS. This gives z (0) (t) = z0 12 g0 t2 , along
with x(0) (t) = y (0) (t) = 0. We then substitute this solution on the RHS and solve for the
deflections, obtaining
x(t) = 21 2 Re sin cos t2
(12.47)
y(t) = 31 g0 sin t3
(12.48)
z(t) = 21 2 Re sin2 t2 .
(12.49)
is due to
The deflection along and r is due to the centrifugal term, while that along
the Coriolis term. (At higher order, the two terms interact and the deflection in any given
direction cant uniquely be associated to a single fictitious force.) To p
find the deflection of
an object dropped from a height h, solve z (0) (t ) = 0 to obtain t = 2h/g0 for the drop
time, and substitute. For h = 100 m and = 2 , find x(t ) = 17 cm south (centrifugal)
and y(t ) = 1.6 cm east (Coriolis).
In fact, an exact solution to (12.46) is readily obtained, via the following analysis. The
equations of motion may be written v = 2iJ v + b, or
b
J
}|
{
z
}|
z
{
v x
vx
g1 sin cos
0
i cos
0
0
0
i sin vy +
v y = 2i i cos
0
i sin
0
g0 + g1 sin2
v x
vx
(12.50)
12.4. THE CORIOLIS FORCE
219
with g1 2 Re . Note that J = J , i.e. J is a Hermitian matrix. The formal solution is
2iJ t
v(t) = e
v(0) +
e2iJ t 1
J 1 b .
2i
(12.51)
When working with matrices, it is convenient to work in an eigenbasis. The characteristic
polynomial for J is P () = det ( 1 J ) = (2 1), hence the eigenvalues are 1 = 0,
2 = +1, and 3 = 1. The corresponding eigenvectors are easily found to be
sin
1 = 0
cos
cos
1
2 = i
2 sin
cos
1
3 = i .
2 sin
(12.52)
Note that a a = aa .
v and
Expanding v and b in this eigenbasis, we have v a = 2ia va + ba , where va = ia
i
b . The solution is
ba = ia
i
2ia t
va (t) = va (0) e
which entails
vi (t) =
ia
e2ia t 1
2ia
e2ia t 1
2ia
ba ,
(12.53)
bj ,
ja
(12.54)
where we have taken v(0) = 0, i.e. the object is released from rest. Doing the requisite
matrix multiplications,
t sin2 + sin22t cos2
vx (t)
2
sin t cos
vy (t) =
vz (t)
12 t sin 2 + sin42t sin 2
sin2 t
cos
sin 2t
2
sin2 t
sin
12 t sin 2 + sin42t sin 2
g1 sin cos
2
,
0
sin t sin
2
sin 2t
2
2
g0 + g1 sin
t cos + 2 sin
(12.55)
which says
vx (t) =
vy (t) =
1
2
sin 2 +
sin2 t
t
sin 2t
4t
g0 t
vz (t) = cos2 +
sin 2 g0 t +
sin2 t
t
sin2t
2t
sin 2t
4t
sin 2 g1 t
sin g1 t
sin2 g0 t +
(12.56)
sin2 t
2t
g1 t .
Why is the deflection always to the east? The earth rotates eastward, and an object starting
from rest in the earths frame has initial angular velocity equal to that of the earth. To
conserve angular momentum, the object must speed up as it falls.
220
CHAPTER 12. NONINERTIAL REFERENCE FRAMES
Figure 12.4: Foucaults pendulum.
12.4.1
Foucaults pendulum
A pendulum swinging over one of the poles moves in a fixed inertial plane while the earth
rotates underneath. Relative to the earth, the plane of motion of the pendulum makes
one revolution every day. What happens at a general latitude? Assume the pendulum is
located at colatitude and longitude . Assuming the length scale of the pendulum is small
,
r} as fixed. The situation is depicted
compared to Re , we can regard the local triad {,
in Fig. 12.4. We write
+ z r ,
r = x + y
(12.57)
with
x = sin cos ,
y = sin sin
z = (1 cos ) .
(12.58)
In our analysis we will ignore centrifugal effects, which are of higher order in , and we
take g = g r. We also idealize the pendulum, and consider the suspension rod to be of
negligible mass.
The total force on the mass m is due to gravity and tension:
F = mg + T
= T sin cos , T sin sin , T cos mg
= T x/, T y/, T M g T z/ .
(12.59)
12.4. THE CORIOLIS FORCE
221
The Coriolis term is
FCor = 2m r
(12.60)
+ z r
= 2m cos r sin x + y
= 2m y cos , x cos z sin , y sin .
(12.61)
m
x = T x/ + 2m cos y
(12.62)
m
y = T y/ 2m cos x 2m sin z
(12.63)
m
z = T mg T z/ + 2m sin y .
(12.64)
The equations of motion are m r = F + FCor :
These three equations are to be solved for the three unknowns x, y, and T . Note that
x2 + y 2 + ( z)2 = 2 ,
(12.65)
so z = z(x, y) is not an independent degree of freedom. This equation may be recast in the
form z = (x2 + y 2 + z 2 )/2 which shows that if x and y are both small, then z is at least of
second order in smallness. Therefore, we will approximate z 0, in which case z may be
neglected from the second equation of motion. The third equation is used to solve for T :
T mg 2m sin y .
(12.66)
Adding the first plus i times the second then gives the complexified equation
T
2i cos
=
m
02 2i cos
(12.67)
p
where x + iy, and where 0 = g/. Note that we have approximated T mg in
deriving the second line.
It is now a trivial matter to solve the homogeneous linear ODE of eq. 12.67. Writing
= 0 eit
(12.68)
2 2 02 = 0 ,
(12.69)
and plugging in to find , we obtain
with cos . The roots are
=
hence the most general solution is
2 ,
02 +
(t) = A+ ei+ t + A ei t .
(12.70)
(12.71)
222
CHAPTER 12. NONINERTIAL REFERENCE FRAMES
Finally, if we take as initial conditions x(0) = a, y(0) = 0, x(0)
= 0, and y(0)
= 0, we
obtain
a n
o
x(t) =
sin( t) sin(t) + cos( t) cos(t)
(12.72)
o
a n
cos( t) sin(t) sin( t) cos(t) ,
(12.73)
y(t) =
with =
2 . Typically , since = 7.3 105 s1 . In the limit ,
02 +
0
0
then, we have 0 and
x(t) a cos( t) cos(0 t) ,
y(t) a sin( t) cos(0 t) ,
(12.74)
and the plane of motion rotates with angular frequency , i.e. the period is | sec | days.
Viewed from above, the rotation is clockwise in the northern hemisphere, where cos > 0
and counterclockwise in the southern hemisphere, where cos < 0.
Chapter 13
Rigid Body Motion and Rotational
Dynamics
13.1
Rigid Bodies
A rigid body consists of a group of particles whose separations are all fixed in magnitude. Six
independent coordinates are required to completely specify the position and orientation of a
rigid body. For example, the location of the first particle is specified by three coordinates. A
second particle requires only two coordinates since the distance to the first is fixed. Finally,
a third particle requires only one coordinate, since its distance to the first two particles
is fixed (think about the intersection of two spheres). The positions of all the remaining
particles are then determined by their distances from the first three. Usually, one takes
these six coordinates to be the center-of-mass position R = (X, Y, Z) and three angles
specifying the orientation of the body (e.g. the Euler angles).
As derived previously, the equations of motion are
X
P =
mi r i , P = F (ext)
L=
X
i
(13.1)
mi ri r i
L = N (ext) .
(13.2)
These equations determine the motion of a rigid body.
13.1.1
Examples of rigid bodies
Our first example of a rigid body is of a wheel rolling with constant angular velocity = ,
and without slipping, This is shown in Fig. 13.1. The no-slip condition is dx = R d, so
x = VCM = R. The velocity of a point within the wheel is
v = VCM + r ,
223
(13.3)
224
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.1: A wheel rolling to the right without slipping.
where r is measured from the center of the disk. The velocity of a point on the surface is
+
r).
then given by v = R x
As a second example, consider a bicycle wheel of mass M and radius R affixed to a light, firm
rod of length d, as shown in Fig. 13.2. Assuming L lies in the (x, y) plane, one computes
The angular momentum vector then
the gravitational torque N = r (M g) = M gd .
rotates with angular frequency . Thus,
d =
dL
L
M gd
=
.
L
(13.4)
But L = M R2 , so the precession frequency is
p = =
gd
.
R2
(13.5)
For R = d = 30 cm and /2 = 200 rpm, find p /2 15 rpm. Note that we have here
resulting in the
ignored the contribution to L from the precession itself, which lies along z,
nutation of the wheel. This is justified if Lp /L = (d2 /R2 ) (p /) 1.
13.2
The Inertia Tensor
Suppose first that a point within the body itself is fixed. This eliminates the translational
degrees of freedom from consideration. We now have
dr
dt
inertial
=r ,
(13.6)
13.2. THE INERTIA TENSOR
225
Figure 13.2: Precession of a spinning bicycle wheel.
since r body = 0. The kinetic energy is then
T =
1
2
1
2
X
i
mi
dri
dt
2
inertial
1
2
X
i
mi ( ri ) ( ri )
mi 2 ri2 ( ri )2 21 I ,
(13.7)
where is the component of along the body-fixed axis e . The quantity I is the
inertia tensor,
I =
mi ri2 ri, ri,
(13.8)
Zi
(continuous media) .
= dd r (r) r 2 r r
(13.9)
The angular momentum is
L=
X
i
X
i
mi ri
dri
dt
inertial
mi ri ( ri ) = I .
(13.10)
The diagonal elements of I are called the moments of inertia, while the off-diagonal
elements are called the products of inertia.
226
13.2.1
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Coordinate transformations
Consider the basis transformation
e = R e .
(13.11)
We demand e e = , which means R O(d) is an orthogonal matrix, i.e. Rt = R1 .
Thus the inverse transformation is e = Rt e . Consider next a general vector A = A e .
Expressed in terms of the new basis {
e }, we have
z }| { z }| {
A = A Rt e = R A e
(13.12)
Thus, the components of A transform as A = R A . This is true for any vector.
Under a rotation, the density (r) must satisfy (r ) = (r). This is the transformation
rule for scalars. The inertia tensor therefore obeys
Z
i
h
2
I = d3 r (r ) r r r
=
h
i
d3 r (r) r 2 R r R r
= R I Rt .
(13.13)
I.e. I = RIRt , the transformation rule for tensors. The angular frequency is a vector, so
= R . The angular momentum L also transforms as a vector. The kinetic energy
is T = 12 t I , which transforms as a scalar.
13.2.2
The case of no fixed point
If there is no fixed point, we can let r denote the distance from the center-of-mass (CM),
which will serve as the instantaneous origin in the body-fixed frame. We then adopt the
notation where R is the CM position of the rotating body, as observed in an inertial frame,
and is computed from the expression
Z
1
1 X
mi i =
R=
d3 r (r) r ,
(13.14)
M
M
i
where the total mass is of course
M=
X
i
mi =
d3 r (r) .
(13.15)
The kinetic energy and angular momentum are then
T = 21 M R 2 + 12 I
L = M R R + I ,
where I is given in eqs. 13.8 and 13.9, where the origin is the CM.
(13.16)
(13.17)
13.3. PARALLEL AXIS THEOREM
227
Figure 13.3: Application of the parallel axis theorem to a cylindrically symmetric mass
distribution.
13.3
Parallel Axis Theorem
Suppose I is given in a body-fixed frame. If we displace the origin in the body-fixed frame
by d, then let I (d) be the inertial tensor with respect to the new origin. If, relative to
the origin at 0 a mass element lies at position r, then relative to an origin at d it will lie at
r d. We then have
o
X n
(13.18)
I (d) =
mi (ri2 2d ri + d2 ) (ri, d )(ri, d ) .
i
If ri is measured with respect to the CM, then
X
mi ri = 0
(13.19)
and
I (d) = I (0) + M d2 d d ,
(13.20)
a result known as the parallel axis theorem.
As an example of the theorem, consider the situation depicted in Fig. 13.3, where a cylindrically symmetric mass distribution is rotated about is symmetry axis, and about an axis
tangent to its side. The component Izz of the inertia tensor is easily computed when the
origin lies along the symmetry axis:
Izz =
Za
3
d3 r (r) (r 2 z 2 ) = L 2 dr r
= 2 La4 = 12 M a2 ,
(13.21)
228
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
where M = a2 L is the total mass. If we compute Izz about a vertical axis which is
tangent to the cylinder, the parallel axis theorem tells us that
Izz
= Izz + M a2 = 32 M a2 .
R
2 would be tedious!
Doing this calculation by explicit integration of dm r
13.3.1
(13.22)
Example
Problem:
Compute the CM and the inertia tensor for the planar right triangle of Fig.
13.4, assuming it to be of uniform two-dimensional mass density .
Solution:
The total mass is M = 21 ab. The x-coordinate of the CM is then
x
Z a)
Za b(1
Za
1
X=
dx dy x =
dx b 1 xa x
M
M
0
Z1
a2 b
a2 b
du u(1 u) =
=
M
6M
1
3
a.
(13.23)
Clearly we must then have Y =
1
3
b, which may be verified by explicit integration.
We now compute the inertia tensor, with the origin at (0, 0, 0). Since the figure is planar,
z = 0 everywhere, hence Ixz = Izx = 0, Iyz = Izy = 0, and also Izz = Ixx + Iyy . We now
compute the remaining independent elements:
x
Ixx
Za b(1
Z a)
Za
3
2
= dx dy y = dx 13 b3 1 xa
0
Z1
= 13 ab3 du (1 u)3 =
3
1
12 ab
= 16 M b2
(13.24)
and
x
Ixy
Za b(1
Z a)
Za
2
2
1
= dx dy x y = 2 b dx x 1 xa
0
= 21 a2 b2
Thus,
Z1
0
1
1
a2 b2 = 12
M ab .
du u (1 u)2 = 24
0
b2
12 ab
M 1
I=
2 ab
a2
0 .
6
2
0
0
a + b2
(13.25)
(13.26)
13.3. PARALLEL AXIS THEOREM
229
Figure 13.4: A planar mass distribution in the shape of a triangle.
Suppose we wanted the inertia tensor relative in a coordinate system where the CM lies at
3b y.
Thus,
the origin. What we computed in eqn. 13.26 is I(d), with d = 3a x
2
b
ab
0
1
0 .
(13.27)
d2 d d = ab a2
9
0
0
a2 + b2
Since
we have that
I(d) = I CM + M d2 d d ,
I CM = I(d) M d2 d d
2 1
b
ab
0
2
M 1
ab a2
0 .
=
18 2
2
0
0 a + b2
13.3.2
(13.28)
(13.29)
(13.30)
General planar mass distribution
For a general planar mass distribution,
(x, y, z) = (x, y) (z) ,
which is confined to the plane z = 0, we have
Z Z
Ixx = dx dy (x, y) y 2
Z Z
Iyy = dx dy (x, y) x2
Z Z
Ixy = dx dy (x, y) xy
and Izz = Ixx + Iyy , regardless of the two-dimensional mass distribution (x, y).
(13.31)
(13.32)
(13.33)
(13.34)
230
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
13.4
Principal Axes of Inertia
We found that an orthogonal transformation to a new set of axes e = R e entails
I = RIRt for the inertia tensor. Since I = I t is manifestly a symmetric matrix, it can
be brought to diagonal form by such an orthogonal transformation. To find R, follow this
recipe:
1. Find the diagonal elements of I by setting P () = 0, where
P () = det 1 I ,
(13.35)
2. For each eigenvalue a , solve the d equations
X
I a = a a .
(13.36)
is the characteristic polynomial for I, and 1 is the unit matrix.
Here, a is the th component of the ath eigenvector. Since ( 1 I) is degenerate,
these equations are linearly dependent, which means that the first d 1 components
may be determined in terms of the dth component.
3. Because I = I t , eigenvectors corresponding to different eigenvalues are orthogonal.
In cases of degeneracy, the eigenvectors may be chosen to be orthogonal, e.g. via the
Gram-Schmidt procedure.
4. Due to the underdetermined aspect to step 2, we may choose an arbitrary normalization for
eigenvector. It is conventional to choose the eigenvectors to be orthonorP each
a b = ab .
mal:
5. The matrix R is explicitly given by Ra = a , the matrix whose row vectors are the
eigenvectors a . Of course Rt is then the corresponding matrix of column vectors.
6. The eigenvectors form a complete basis. The resolution of unity may be expressed as
X
a a = .
(13.37)
a
As an example, consider the inertia tensor for a general planar mass distribution, which is
of the form
Ixx Ixy 0
I = Iyx Iyy 0 ,
(13.38)
0
0 Izz
where Iyx = Ixy and Izz = Ixx + Iyy . Define
A=
1
2
Ixx + Iyy
B=
1
4
= tan
Ixx Iyy
(13.39)
2
2
+ Ixy
2Ixy
Ixx Iyy
(13.40)
(13.41)
13.5. EULERS EQUATIONS
so that
231
A + B cos
B sin
0
I = B sin
A B cos 0 ,
0
0
2A
(13.42)
i
h
P () = ( 2A) ( A)2 B 2 ,
(13.43)
The characteristic polynomial is found to be
which gives 1 = A, 2,3 = A B. The corresponding normalized eigenvectors are
and therefore
13.5
0
1
= 0
1
cos 12
2 = sin 12
0
sin 12
3 = cos 21
0
0
0
1
R = cos 12 sin 12 0 .
sin 12 cos 21 0
(13.44)
(13.45)
Eulers Equations
Let us now choose our coordinate axes to
the origin. We may then write
1
I1
= 2
, I= 0
0
3
The equations of motion are
N ext =
=
be the principal axes of inertia, with the CM at
0 0
I2 0
0 I3
dL
dt
dL
dt
I1 1
L = I2 2 .
I3 3
(13.46)
inertial
body
+L
= I + (I ) .
Thus, we arrive at Eulers equations:
I1 1 = (I2 I3 ) 2 3 + N1ext
I2 2 = (I3 I1 ) 3 1 +
I3 3 = (I1 I2 ) 1 2 +
N2ext
N3ext
(13.47)
(13.48)
.
(13.49)
These are coupled and nonlinear. Also note the fact that the external torque must be
evaluated along body-fixed principal axes. We can however make progress in the case
232
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.5: Wobbling of a torque-free symmetric top.
where N ext = 0, i.e. when there are no external torques. This is true for a body in free
space, or in a uniform gravitational field. In the latter case,
X
X
(13.50)
mi ri g ,
N ext =
ri (mi g) =
i
where g is theP
uniform gravitational acceleration. In a body-fixed frame whose origin is the
CM, we have i mi ri = 0, and the external torque vanishes!
Precession of torque-free symmetric tops:
Consider a body which has a symmetry axis e3 . This guarantees I1 = I2 , but in general we still have I1 6= I3 . In the absence
of external torques, the last of Eulers equations says 3 = 0, so 3 is a constant. The
remaining two equations are then
I1 I3
I3 I1
1 =
3 2 , 2 =
3 1 .
(13.51)
I1
I1
I.e. 1 = 2 and 2 = + 1 , with
=
I3 I1
I1
3 ,
which are the equations of a harmonic oscillator. The solution is easily obtained:
1 (t) = cos t +
, 2 (t) = sin t + , 3 (t) = 3 ,
(13.52)
(13.53)
2 + 2 )1/2 . This motion is
where and are constants of integration, and where || = (
3
sketched in Fig. 13.5. Note that the perpendicular components of oscillate harmonically,
and that the angle makes with respect to e3 is = tan1 ( /3 ).
1
For the earth, (I3 I1 )/I1 305
, so 3 , and /305, yielding a precession period
of 305 days, or roughly 10 months. Astronomical observations reveal such a precession,
13.5. EULERS EQUATIONS
233
known as the Chandler wobble. For the earth, the precession angle is Chandler 6 107
rad, which means that the North Pole moves by about 4 meters during the wobble. The
Chandler wobble has a period of about 14 months, so the nave prediction of 305 days is
off by a substantial amount. This discrepancy is attributed to the mechanical properties of
the earth: elasticity and fluidity. The earth is not solid!1
Asymmetric tops:
Next, consider the torque-free motion of an asymmetric top, where
I1 6= I2 6= I3 6= I1 . Unlike the symmetric case, there is no conserved component of . True,
we can invoke conservation of energy and angular momentum,
E = 21 I1 12 + 12 I2 22 + 12 I3 32
L2 = I12 12 + I22 22 + I32 32 ,
(13.54)
(13.55)
and, in principle, solve for 1 and 2 in terms of 3 , and then invoke Eulers equations
(which must honor these conservation laws). However, the nonlinearity greatly complicates
matters and in general this approach is a dead end.
We can, however, find a particular solution quite easily one in which the rotation is about
a single axis. Thus, 1 = 2 = 0 and 3 = 0 is indeed a solution for all time, according to
Eulers equations. Let us now perturb about this solution, to explore its stability. We write
= 0 e3 + ,
(13.56)
and we invoke Eulers equations, linearizing by dropping terms quadratic in . This yield
I1 1 = (I2 I3 ) 0 2 + O(2 3 )
I2 2 = (I3 I1 ) 0 1 + O(3 1 )
I3 3 = 0 + O(1 2 ) .
(13.57)
(13.58)
(13.59)
Taking the time derivative of the first equation and invoking the second, and vice versa,
yields
1 = 2 1 ,
2 = 2 2 ,
(13.60)
with
2 =
(I3 I2 )(I3 I1 ) 2
0 .
I1 I2
(13.61)
The solution is then 1 (t) = C cos(t + ).
If 2 > 0, then is real, and the deviation results in a harmonic precession. This occurs
if I3 is either the largest or the smallest of the moments of inertia. If, however, I3 is the
middle moment, then 2 < 0, and is purely imaginary. The perturbation will in general
increase exponentially with time, which means that the initial solution to Eulers equations
is unstable with respect to small perturbations. This result can be vividly realized using a
tennis racket, and sometimes goes by the name of the tennis racket theorem.
1
The earth is a layered like a Mozartkugel, with a solid outer shell, an inner fluid shell, and a solid (iron)
core.
234
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
13.5.1
Example
PROBLEM: A unsuspecting solid spherical planet of mass M0 rotates with angular velocity
0 . Suddenly, a giant asteroid of mass M0 smashes into and sticks to the planet at a
location which is at polar angle relative to the initial rotational axis. The new mass
distribution is no longer spherically symmetric, and the rotational axis will precess. Recall
Eulers equation
dL
+ L = N ext
dt
(13.62)
for rotations in a body-fixed frame.
(a) What is the new inertia tensor I along principle center-of-mass frame axes? Dont
forget that the CM is no longer at the center of the sphere! Recall I = 25 M R2 for a solid
sphere.
(b) What is the period of precession of the rotational axis in terms of the original length
of the day 2/0 ?
SOLUTION: Lets choose body-fixed axes with z pointing from the center of the planet to
the smoldering asteroid. The CM lies a distance
d=
M0 R + M0 0
R
=
(1 + )M0
1+
(13.63)
from the center of the sphere. Thus, relative to the center of the sphere, we have
1 0 0
1 0 0
I = 52 M0 R2 0 1 0 + M0 R2 0 1 0 .
0 0 0
0 0 1
(13.64)
Now we shift to a frame with the CM at the origin, using the parallel axis theorem,
Thus, with d = dz,
CM
I
CM
+ M d2 d d .
I (d) = I
1 0 0
1
2
2
2
= 5 M0 R 0 1 0 + M0 R 0
0 0 1
0
2
0
0
5 + 1+
2
= M0 R
0
5 + 1+ 0
2
0
0
5
0 0
1 0 0
1 0 (1 + )M0 d2 0 1 0
0 0
0 0 0
.
(13.65)
(13.66)
(13.67)
13.6. EULERS ANGLES
235
In the absence of external torques, Eulers equations along principal axes read
d1
= (I2 I3 ) 2 3
dt
d2
= (I3 I1 ) 3 1
I2
dt
d3
= (I1 I2 ) 1 2
I3
dt
I1
(13.68)
Since I1 = I2 , 3 (t) = 3 (0) = 0 cos is a constant. We then obtain 1 = 2 , and
2 = 1 , with
I2 I3
5
=
.
(13.69)
3 =
I1
7 + 2 3
The period of precession in units of the pre-cataclysmic day is
7 + 2
=
=
.
T
5 cos
13.6
(13.70)
Eulers Angles
In d dimensions, an orthogonal matrix R O(d) has 21 d(d 1) independent parameters.
To see this, consider the constraint Rt R = 1. The matrix Rt R is manifestly symmetric,
so it has 21 d(d + 1) independent entries (e.g. on the diagonal and above the diagonal).
This amounts to 12 d(d + 1) constraints on the d2 components of R, resulting in 12 d(d 1)
freedoms. Thus, in d = 3 rotations are specified by three parameters. The Euler angles
{, , } provide one such convenient parameterization.
A general rotation R(, , ) is built up in three steps. We start with an orthonormal triad
e0 of body-fixed axes. The first step is a rotation by an angle about e03 :
e = R
e03 , e0
cos sin 0
R e03 , = sin cos 0
0
0
1
(13.71)
This step is shown in panel (a) of Fig. 13.6. The second step is a rotation by about the
new axis e1 :
1
0
0
e = R e1 , e , R e1 , = 0 cos sin
(13.72)
0 sin cos
This step is shown in panel (b) of Fig. 13.6. The third and final step is a rotation by
about the new axis e3 :
cos sin 0
3 , e , R e3 , = sin cos 0
e
(13.73)
= R e
0
0
1
236
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.6: A general rotation, defined in terms of the Euler angles {, , }. Three
successive steps of the transformation are shown.
This step is shown in panel (c) of Fig. 13.6. Putting this all together,
R(, , ) = R e3 , R e1 , R e03 ,
(13.74)
cos sin 0
1
0
0
cos sin 0
= sin cos 0 0 cos sin sin cos 0
0
0
1
0 sin cos
0
0
1
cos cos sin cos sin
cos sin + sin cos cos sin sin
= sin cos cos cos sin sin sin + cos cos cos cos sin .
sin sin
sin cos
cos
Next, wed like to relate the components = e (with e e
) of the rotation in the
,
and .
To do this, we write
body-fixed frame to the derivatives ,
= e + e + e ,
(13.75)
where
e03 = e = sin sin e1 + sin cos e2 + cos e3
e = cos e1 sin e2
e = e3 .
(line of nodes)
(13.76)
(13.77)
(13.78)
13.6. EULERS ANGLES
237
This gives
1 = e1 = sin sin + cos
2 = e2 = sin cos sin
= e = cos + .
3
(13.79)
(13.80)
(13.81)
Note that
precession
nutation
axial rotation .
(13.82)
The general form of the kinetic energy is then
T = 12 I1 sin sin + cos
2
+ 21 I2 sin cos sin
Note that
2
+ 12 I3 cos +
2
L = p e + p e + p e ,
(13.83)
(13.84)
which may be verified by explicit computation.
13.6.1
Torque-free symmetric top
A body falling in a gravitational field experiences no net torque about its CM:
X
X
N ext =
ri (mi g) = g
mi ri = 0 .
i
(13.85)
For a symmetric top with I1 = I2 , we have
2
T = 12 I1 2 + 2 sin2 + 12 I3 cos + .
(13.86)
The potential is cyclic in the Euler angles, hence the equations of motion are
T
d
T
.
=
dt (, , )
(, , )
(13.87)
Since and are cyclic in T , their conjugate momenta are conserved:
L
cos
= I1 sin2 + I3 ( cos + )
L
.
p =
= I3 ( cos + )
p =
(13.88)
(13.89)
Note that p = I3 3 , hence 3 is constant, as we have already seen.
To solve for the motion, we first note that L is conserved in the inertial frame. We are
= e0 = e . Thus, p = L. Since e e = cos , we have
therefore permitted to define L
3
238
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.7: A dreidl is a symmetric top. The four-fold symmetry axis guarantees I1 = I2 .
The blue diamond represents the center-of-mass.
that p = L e = L cos . Finally, e e = 0, which means p = L e = 0. From the
equations of motion,
p = I1 = I1 cos p sin ,
(13.90)
hence we must have
= 0 ,
p
.
I1 cos
(13.91)
Note that = 0 follows from conservation of p = L cos . From the equation for p , we
may now conclude
p
p
I3 I1
=
3 ,
(13.92)
I3
I1
I3
which recapitulates (13.52), with = .
13.6.2
Symmetric top with one point fixed
Consider the case of a symmetric top with one point fixed, as depicted in Fig. 13.7. The
Lagrangian is
2
L = 12 I1 2 + 2 sin2 + 21 I3 cos + M g cos .
(13.93)
Here, is the distance from the fixed point to the CM, and the inertia tensor is defined along
principal axes whose origin lies at the fixed point (not the CM!). Gravity now supplies a
13.6. EULERS ANGLES
239
Figure 13.8: The effective potential of eq. 13.102.
torque, but as in the torque-free case, the Lagrangian is still cyclic in and , so
p = (I1 sin2 + I3 cos2 ) + I3 cos
p = I cos + I
(13.94)
(13.95)
are each conserved. We can invert these relations to obtain and in terms of {p , p , }:
p p cos
=
I1 sin2
p
(p p cos ) cos
=
.
I3
I1 sin2
(13.96)
In addition, since L/t = 0, the total energy is conserved:
Ueff ()
}|
{
z
2
2
p
(p
p
cos
)
E = T + U = 21 I1 2 +
+
+ M g cos ,
2I3
2I1 sin2
(13.97)
where the term under the brace is the effective potential Ueff ().
The problem thus reduces to the one-dimensional dynamics of (t), i.e.
I1 =
Ueff
,
(13.98)
with
p2
(p p cos )2
+
+ M g cos .
2I3
2I1 sin2
Using energy conservation, we may write
r
I1
d
p
.
dt =
2
E Ueff ()
Ueff () =
(13.99)
(13.100)
240
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
and thus the problem is reduced to quadratures:
t() = t(0 )
I1
2
Z
d p
1
.
E Ueff ()
(13.101)
We can gain physical insight into the motion by examining the shape of the effective potential,
p2
(p p cos )2
+
M
g
cos
+
,
(13.102)
Ueff () =
2I3
2I1 sin2
over the interval [0, ]. Clearly Ueff (0) = Ueff () = , so the motion must be bounded.
What is not yet clear, but what is nonetheless revealed by some additional analysis, is that
Ueff () has a single minimum on this interval, at = 0 . The turning points for the motion
are at = a and = b , where Ueff (a ) = Ueff (b ) = E. Clearly if we expand about 0 and
write = 0 + , the motion will be harmonic, with
s
( )
Ueff
0
.
(13.103)
(t) = 0 cos(t + ) , =
I1
To prove that Ueff () has these features, let us define u cos . Then u = sin , and
from E = 12 I1 2 + Ueff () we derive
2
u =
p p u 2
p2
2E
2M g
2
2
f (u) .
(1 u ) u
(1 u )
I1
I1 I3
I1
I1
(13.104)
The turning points occur at f (u) = 0. The function f (u) is cubic, and the coefficient of the
cubic term is 2M g/I1 , which is positive. Clearly f (u = 1) = (p p )2 /I12 is negative,
so there must be at least one solution to f (u) = 0 on the interval u (1, ). Clearly there
can be at most three real roots for f (u), since the function is cubic in u, hence there are at
most two turning points on the interval u [1, 1]. Thus, Ueff () has the form depicted in
fig. 13.8.
To apprehend the full motion of the top in an inertial frame, let us follow the symmetry
axis e3 :
e3 = sin sin e01 sin cos e02 + cos e03 .
(13.105)
Once we know (t) and (t) were done. The motion (t) is described above: oscillates
between turning points at a and b . As for (t), we have already derived the result
p p cos
.
=
I1 sin2
(13.106)
Thus, if p > p cos a , then will remain positive throughout the motion. If, on the other
hand, we have
p cos b < p < p cos a ,
(13.107)
then changes sign at an angle = cos1 p /p . The motion is depicted in Fig. 13.9.
An extensive discussion of this problem is given in H. Goldstein, Classical Mechanics.
13.7. ROLLING AND SKIDDING MOTION OF REAL TOPS
241
Figure 13.9: Precession and nutation of the symmetry axis of a symmetric top.
13.7
Rolling and Skidding Motion of Real Tops
The material in this section is based on the corresponding sections from V. Barger and
M. Olsson, Classical Mechanics: A Modern Perspective. This is an excellent book which
contains many interesting applications and examples.
13.7.1
Rolling tops
In most tops, the point of contact rolls or skids along the surface. Consider the peg end top
of Fig. 13.10, executing a circular rolling motion, as sketched in Fig. 13.11. There are three
components to the force acting on the top: gravity, the normal force from the surface, and
friction. The frictional force is perpendicular to the CM velocity, and results in centripetal
acceleration of the top:
f = M 2 M g ,
(13.108)
where is the frequency of the CM motion and is the coefficient of friction. If the above
inequality is violated, the top starts to slip.
The frictional and normal forces combine to produce a torque N = M g sin f cos
about the CM2 . This torque is tangent to the circular path of the CM, and causes L to
precess. We assume that the top is spinning rapidly, so that L very nearly points along
the symmetry axis of the top itself. (As well see, this is true for slow precession but not
for fast precession, where the precession frequency is proportional to 3 .) The precession is
then governed by the equation
N = M g sin f cos
= L = L I3 3 sin ,
Gravity of course produces no net torque about the CM.
(13.109)
242
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.10: A top with a peg end. The frictional forces f and fskid are shown. When the
top rolls without skidding, fskid = 0.
3 is the instantaneous symmetry axis of the top. Substituting f = M 2 ,
where e
2
M g
ctn = ,
(13.110)
1
I3 3
g
which is a quadratic equation for . We supplement this with the no slip condition,
3 = + sin ,
(13.111)
resulting in two equations for the two unknowns and .
Substituting for () and solving for , we obtain
s
)
(
2
I3 3
M g
4M 2 M g
M g
=
.
1+
1+
ctn
ctn
2M 2 cos
I3
I3
I3
I3 32
This in order to have a real solution we must have
2M 2 sin
3
I3 sin + M g cos
g
.
(13.112)
(13.113)
If the inequality is satisfied, there are two possible solutions for , corresponding to fast
and slow precession. Usually one observes slow precession. Note that it is possible that
< 0, in which case the CM and the peg end lie on opposite sides of a circle from each
other.
13.7.2
Skidding tops
A skidding top experiences a frictional force which opposes the skidding velocity, until
vskid = 0 and a pure rolling motion sets in. This force provides a torque which makes the
13.7. ROLLING AND SKIDDING MOTION OF REAL TOPS
243
Figure 13.11: Circular rolling motion of the peg top.
top rise:
M g
Nskid
=
.
(13.114)
=
L
I3 3
Suppose 0, in which case + sin = 0, from eqn. 13.111, and the point of contact
remains fixed. Now recall the effective potential for a symmetric top with one point fixed:
Ueff () =
p2
(p p cos )2
+
+ M g cos .
2I3
2I1 sin2
(13.115)
( ) = 0, which yields
We demand Ueff
0
where
cos 0 2 p sin2 0 + M gI1 sin4 0 = 0 ,
(13.116)
p p cos 0 = I1 sin2 0 .
(13.117)
Solving the quadratic equation for , we find
s
!
I
4M
gI
cos
3
3
1
0
=
1 1
.
I32 32
2I1 cos 0
(13.118)
This is simply a recapitulation of eqn. 13.112, with = 0 and with M 2 replaced by I1 .
Note I1 = M 2 by the parallel axis theorem if I1CM = 0. But to the extent that I1CM 6= 0, our
treatment of the peg top was incorrect. It turns out to be OK, however, if the precession is
slow, i.e. if /3 1.
On a level surface, cos 0 > 0, and therefore we must have
q
2
M gI1 cos 0 .
3
I3
(13.119)
244
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Figure 13.12: The tippie-top behaves in a counterintuitive way. Once started spinning with
the peg end up, the peg axis rotates downward. Eventually the peg scrapes the surface and
the top rises to the vertical in an inverted orientation.
Thus, if the top spins too slowly, it cannot maintain precession. Eqn. 13.118 says that there
are two possible precession frequencies. When 3 is large, we have
M g
+ O(31 )
slow =
I3 3
fast =
I3 3
+ O(33 ) .
I1 cos 0
(13.120)
Again, one usually observes slow precession.
A top with 3 > I23 M gI1 may sleep in the vertical position with 0 = 0. Due to the
constant action of frictional forces, 3 will eventually drop below this value, at which time
the vertical position is no longer stable. The top continues to slow down and eventually
falls.
13.7.3
Tippie-top
A particularly nice example from the Barger and Olsson book is that of the tippie-top, a
truncated sphere with a peg end, sketched in Fig. 13.12 The CM is close to the center
of curvature, which means that there is almost no gravitational torque acting on the top.
The frictional force f opposes slipping, but as the top spins f rotates with it, and hence
the time-averaged frictional force hf i 0 has almost no effect on the motion of the CM.
A similar argument shows that the frictional torque, which is nearly horizontal, also time
averages to zero:
dL
0.
(13.121)
dt inertial
In the body-fixed frame, however, N is roughly constant, with magnitude N M gR,
13.7. ROLLING AND SKIDDING MOTION OF REAL TOPS
245
where R is the radius of curvature and the coefficient of sliding friction. Now we invoke
dL
+ L .
(13.122)
N=
dt body
The second term on the RHS is very small, because the tippie-top is almost spherical, hence
inertia tensor is very nearly diagonal, and this means
L I = 0 .
(13.123)
we
Thus, L body N , and taking the dot product of this equation with the unit vector k,
obtain
N = d k
L
N sin = k
(13.124)
body = L sin .
dt
Thus,
N
M gR
=
.
(13.125)
L
I
Once the stem scrapes the table, the tippie-top rises to the vertical just like any other rising
top.
246
CHAPTER 13. RIGID BODY MOTION AND ROTATIONAL DYNAMICS
Chapter 14
Continuum Mechanics
14.1
Strings
Consider a string of linear mass density (x) under tension (x).1 Let the string move in a
plane, such that its shape is described by a smooth function y(x), the vertical displacement
of the string at horizontal position x, as depicted in fig. 14.1. The action is a functional
of the height y(x, t), where the coordinate along the string, x, and time, t, are the two
independent variables. Consider a differential element of the string extending from x to
x + dx. The change in length relative to the unstretched (y = 0) configuration is
p
1 y 2
2
2
dx + O dx2 .
(14.1)
d = dx + dy dx =
2 x
The differential potential energy is then
dU = (x) d =
1
2
(x)
2
y
x
2
dx .
(14.2)
The differential kinetic energy is simply
1
2
dT =
(x)
We can then write
L=
where the Lagrangian density L is
L(y, y,
y ; x, t) =
1
2
y
t
dx .
dx L ,
2
2
y
y
1
(x)
2 (x)
.
t
x
(14.3)
(14.4)
(14.5)
1
As an example of a string with a position-dependent tension, consider a string of length freely suspended
from one end at z = 0 in a gravitational field. The tension is then (z) = g ( z).
247
248
CHAPTER 14. CONTINUUM MECHANICS
Figure 14.1: A string is described by the vertical displacement field y(x, t).
The action for the string is now a double integral,
Ztb Zxb
S = dt dx L(y, y,
y ; x, t) ,
ta
(14.6)
xa
where y(x, t) is the vertical displacement field. Typically, we have L = 12 y 2 21 y 2 . The
first variation of S is
#
Zxb Ztb "
L
L
L
S = dx dt
y
x y
t y
(14.7)
t=t
x=xa
Ztb
Zxb
b
L
L
,
y
y
+ dt
+ dx
y
y
x=x
t=ta
(14.8)
xa
ta
xa
ta
which simply recapitulates the general result from eqn. 14.203. There are two boundary
terms, one of which is an integral over time and the other an integral over space. The first
boundary term vanishes provided y(x, ta ) = y(x, tb ) = 0. The second boundary term
vanishes provided (x) y (x) y(x) = 0 at x = xa and x = xb , for all t. Assuming (x) does
not vanish, this can happen in one of two ways: at each endpoint either y(x) is fixed or
y (x) vanishes.
Assuming that either y(x) is fixed or y (x) = 0 at the endpoints x = xa and x = xb , the
Euler-Lagrange equations for the string are obtained by setting S = 0:
L
L
L
S
=
0=
y(x, t)
y
t y
x y
y
2y
=
(14.9)
(x)
(x) 2 ,
x
x
t
y
where y = x
and y =
Helmholtz equation,
y
t .
When (x) = and (x) = are both constants, we obtain the
2y
1 2y
=0,
c2 t2
x2
(14.10)
14.2. DALEMBERTS SOLUTION TO THE WAVE EQUATION
249
p
which is the wave equation for the string, where c = / has dimensions of velocity. We
will now see that c is the speed of wave propagation on the string.
14.2
dAlemberts Solution to the Wave Equation
Let us define two new variables,
u x ct
v x + ct .
(14.11)
We then have
u
v
=
+
x
x u x v
+
u v
(14.12)
1 u
1 v
1
=
+
=
+
.
c t
c t u c t v
u v
(14.13)
1 2
2
2
=
4
.
c2 t2 x2
u v
(14.14)
Thus,
Thus, the wave equation may be solved:
2y
=0
u v
y(u, v) = f (u) + g(v) ,
(14.15)
where f (u) and g(v) are arbitrary functions. For the moment, we work with an infinite
string, so we have no spatial boundary conditions to satisfy. Note that f (u) describes a
right-moving disturbance, and g(v) describes a left-moving disturbance:
y(x, t) = f (x ct) + g(x + ct) .
(14.16)
We do, however, have boundary conditions in time. At t = 0, the configuration of the string
is given by y(x, 0), and its instantaneous vertical velocity is y(x,
0). We then have
y(x, 0) = f (x) + g(x)
(14.17)
y(x,
0) = c f (x) + c g (x) ,
(14.18)
hence
f (x) =
1
2
y (x, 0)
1
2c
y(x,
0)
(14.19)
g (x) =
1
2
y (x, 0) +
1
2c
y(x,
0) ,
(14.20)
250
CHAPTER 14. CONTINUUM MECHANICS
and integrating we obtain the right and left moving components
f () =
1
2
Z
y(, 0)
d y(
, 0) C
1
2c
(14.21)
Z
1
g() = 12 y(, 0) + 2c
d y(
, 0) + C ,
(14.22)
where C is an arbitrary constant. Adding these together, we obtain the full solution
y(x, t) =
1
2
i
y(x ct, 0) + y(x + ct, 0) +
x+ct
Z
1
2c
d y(,
0) ,
(14.23)
xct
valid for all times.
14.2.1
Energy density and energy current
The Hamiltonian density for a string is
H = y L ,
(14.24)
L
= y
y
(14.25)
2 1 2
+ 2 y .
2
(14.26)
where
=
is the momentum density. Thus,
H=
Expressed in terms of y rather than , this is the energy density E,
2
E = 21 y 2 + 12 y .
(14.27)
We now evaluate E for a solution to the equations of motion:
y 2 y
y 2 y
E
=
+
t
t t2
x t x
y
y
y 2 y
=
+
t x
x
x t x
j
y y
=
E ,
x x t
x
(14.28)
where the energy current density (or energy flux) is
jE =
y y
.
x t
(14.29)
14.2. DALEMBERTS SOLUTION TO THE WAVE EQUATION
251
Figure 14.2: Reflection of a pulse at an interface at x = 0, with y(0, t) = 0.
We therefore have that solutions of the equation of motion also obey the energy continuity
equation
j
E
+ E =0.
(14.30)
t
x
Let us integrate the above equation between points x1 and x2 . We obtain
Zx2
Zx2
j (x, t)
= jE (x1 , t) jE (x2 , t) ,
dx E(x, t) = dx E
x
x1
(14.31)
x1
which says that the time rate of change of the energy contained in the interval x1 , x2 is
equal to the difference between the entering and exiting energy flux.
When (x) = and (x) = , we have
y(x, t) = f (x ct) + g(x + ct)
(14.32)
and we find
2
2
E(x, t) = [f (x ct) + g (x + ct)
2
2
jE (x, t) = c f (x ct) c g (x + ct) ,
(14.33)
(14.34)
which are each sums over right-moving and left-moving contributions.
14.2.2
Reflection at an interface
Consider a semi-infinite string on the interval 0, , with y(0, t) = 0. We can still invoke
dAlemberts solution, y(x, t) = f (x ct) + g(x + ct), but we must demand
y(0, t) = f (ct) + g(ct) = 0
f () = g() .
(14.35)
Thus,
y(x, t) = g(ct + x) g(ct x) .
(14.36)
Now suppose g() describes a pulse, and is nonzero only within a neighborhood of = 0.
For large negative values of t, the right-moving part, g(ct x), is negligible everywhere,
252
CHAPTER 14. CONTINUUM MECHANICS
Figure 14.3: Reflection of a pulse at an interface at x = 0, with y (0, t) = 0.
since x > 0 means that the argument ct x is always large and negative. On the other
hand, the left moving part g(ct + x) is nonzero for x ct > 0. Thus, for t < 0 we have
a left-moving pulse incident from the right. For t > 0, the situation is reversed, and the
left-moving component is negligible, and we have a right moving reflected wave. However,
the minus sign in eqn. 14.35 means that the reflected wave is inverted.
If instead of fixing the endpoint at x = 0 we attach this end of the string to a massless ring
which frictionlessly slides up and down a vertical post, then we must have y (0, t) = 0, else
there is a finite vertical force on the massless ring, resulting in infinite acceleration. We
again write y(x, t) = f (x ct) + g(x + ct), and we invoke
y (0, t) = f (ct) + g (ct)
f () = g () ,
(14.37)
which, upon integration, yields f () = g(), and therefore
y(x, t) = g(ct + x) + g(ct x) .
(14.38)
The reflected pulse is now right-side up, in contrast to the situation with a fixed endpoint.
14.2.3
Mass point on a string
Next, consider the case depicted in Fig. 14.4, where a point mass m is affixed to an infinite
string at x = 0. Let us suppose that at large negative values of t, a right moving wave
f (ct x) is incident from the left. The full solution may then be written as a sum of
incident, reflected, and transmitted waves:
x<0
y(x, t) = f (ct x) + g(ct + x)
(14.39)
x>0
y(x, t) = h(ct x) .
(14.40)
At x = 0, we invoke Newtons second Law, F = ma:
m y(0, t) = y (0+ , t) y (0 , t) .
(14.41)
Any discontinuity in the derivative y (x, t) at x = 0 results in an acceleration of the point
mass. Note that
y (0 , t) = f (ct) + g (ct)
y (0+ , t) = h (ct) .
(14.42)
14.2. DALEMBERTS SOLUTION TO THE WAVE EQUATION
253
Figure 14.4: Reflection and transmission at an impurity. A point mass m is affixed to an
infinite string at x = 0.
Further invoking continuity at x = 0, i.e. y(0 , t) = y(0+ , t), we have
h() = f () + g() ,
(14.43)
and eqn. 14.41 becomes
g () +
2
g () = f () .
mc2
(14.44)
We solve this equation by Fourier analysis:
f () =
dk
f (k) eik
2
f(k) =
Z
d f () eik .
(14.45)
Defining 2 /mc2 = 2/m, we have
k2 + ik g(k) = k2 f(k) .
(14.46)
We then have
g(k) =
k
f(k) r(k) f(k)
k i
f (k) t(k) f(k) ,
h(k)
=
k i
(14.47)
(14.48)
where r(k) and t(k) are the reflection and transmission amplitudes, respectively. Note that
t(k) = 1 + r(k) .
(14.49)
254
CHAPTER 14. CONTINUUM MECHANICS
In real space, we have
h() =
dk
t(k) f(k) eik
2
" Z
#
dk
t(k) eik( ) f ( )
2
d T ( ) f ( ) ,
(14.50)
(14.51)
(14.52)
where
T ( ) =
dk
t(k) eik( ) ,
2
(14.53)
is the transmission kernel in real space. For our example with r(k) = i/(k i), the
integral is done easily using the method of contour integration:
T ( ) =
dk i ik( )
e
= e( ) ( ) .
2 k i
(14.54)
Therefore,
Z
h() = d e( ) f ( ) ,
(14.55)
and of course g() = h() f (). Note that m = means = 0, in which case r(k) = 1
and t(k) = 0. Thus we recover the inversion of the pulse shape under reflection found
earlier.
For example, let the incident pulse shape be f () = b a || . Then
Z
h() = d e( ) b (a ) (a + )
Taking cases,
i
h
= b e e min(a,) ea ( + a) .
0
if < a
(a+)
if a < < a
h() = b 1 e
2b e sinh(a) if > a .
(14.56)
(14.57)
In Fig. 14.5 we show the reflection and transmission of this square pulse for two different
values of a.
14.2. DALEMBERTS SOLUTION TO THE WAVE EQUATION
255
Figure 14.5: Reflection and transmission of a square wave pulse by a point mass at x = 0
The configuration of the string is shown for six different times, for a = 0.5 (left panel)
and a = 5.0 (right panel). Note that the a = 0.5 case, which corresponds to a large mass
m = 2/, results in strong reflection with inversion, and weak transmission. For large ,
corresponding to small mass m, the reflection is weak and the transmission is strong.
14.2.4
Interface between strings of different mass density
Consider the situation in fig. 14.6, where the string for x < 0 is of density L and for x > 0
is of density R . The dAlembert solution in the two regions, with an incoming wave from
the left, is
x < 0:
y(x, t) = f (cL t x) + g(cL t + x)
(14.58)
x > 0:
y(x, t) = h(cR t x) .
(14.59)
f (cL t) + g(cL t) = h(cR t)
(14.60)
(14.61)
At x = 0 we have
f (cL t) + g (cL t) = h (cR t) ,
y (0+ , t)
y (0 , t),
where the second equation follows from
=
so there is no finite vertical force on the infinitesimal interval bounding x = 0, which contains infinitesimal mass.
Defining cR /cL , we integrate the second of these equations and have
f () + g() = h( )
(14.62)
f () g() = 1 h( ) .
(14.63)
256
CHAPTER 14. CONTINUUM MECHANICS
Figure 14.6: String formed from two semi-infinite regions of different densities..
Note that y(, 0) = 0 fixes the constant of integration. The solution is then
g() =
1
f ()
+1
(14.64)
h() =
2
f (/) .
+1
(14.65)
Thus,
x < 0:
x > 0:
y(x, t) = f cL t x +
y(x, t) =
1
+1
2
f (cR t x)/ .
+1
f cL t + x
(14.66)
(14.67)
It is instructive to compute the total energy in the string. For large negative values of the
time t, the entire disturbance is confined to the region x < 0. The energy is
Z
2
E() = d f () .
(14.68)
For large positive times, the wave consists of the left-moving reflected g() component in
the region x < 0 and the right-moving transmitted component h() in the region x > 0.
The energy in the reflected wave is
1
EL (+) =
+1
2 Z
2
d f () .
(14.69)
For the transmitted portion, we use
y (x > 0, t) =
2
f (cR t x)/
+1
(14.70)
14.3. FINITE STRINGS : BERNOULLIS SOLUTION
257
to obtain
4
ER () =
( + 1)2
=
Z
2
d f (/)
2
d f () .
4
( + 1)2
(14.71)
Thus, EL () + ER () = E(), and energy is conserved.
14.3
Finite Strings : Bernoullis Solution
Suppose xa = 0 and xb = L are the boundaries of the string, where y(0, t) = y(L, t) = 0.
Again we write
y(x, t) = f (x ct) + g(x + ct) .
(14.72)
Applying the boundary condition at xa = 0 gives, as earlier,
y(x, t) = g(ct + x) g(ct x) .
(14.73)
Next, we apply the boundary condition at xb = L, which results in
g(ct + L) g(ct L) = 0
g() = g( + 2L) .
(14.74)
Thus, g() is periodic, with period 2L. Any such function may be written as a Fourier sum,
(
)
X
n
n
An cos
g() =
.
(14.75)
+ Bn sin
L
L
n=1
The full solution for y(x, t) is then
y(x, t) = g(ct + x) g(ct x)
=
where An =
2
L
1/2 X
sin
n=1
nx
L
(
)
nct
nct
An cos
,
+ Bn sin
L
L
(14.76)
2L Bn and Bn = 2L An . This is known as Bernoullis solution.
We define the functions
n (x)
We also write
kn
nx
L
2
L
1/2
nc
L
sin
(14.77)
n = 1, 2, 3, . . . , .
(14.78)
nx
L
258
CHAPTER 14. CONTINUUM MECHANICS
Thus, n (x) =
that
p
2/L sin(kn x) has (n + 1) nodes at x = jL/n, for j {0, . . . , n}. Note
m n
ZL
dx m (x) n (x) = mn .
(14.79)
Furthermore, this basis is complete:
n=1
n (x) n (x ) = (x x ) .
(14.80)
Our general solution is thus equivalent to
y(x, 0) =
y(x,
0) =
n=1
n=1
An n (x)
(14.81)
nc
Bn n (x) .
L
(14.82)
The Fourier coefficients {An , Bn } may be extracted from the initial data using the orthonormality of the basis functions and their associated resolution of unity:
An =
ZL
dx n (x) y(x, 0)
(14.83)
L
Bn =
nc
ZL
dx n (x) y(x,
0) .
(14.84)
As an example, suppose our initial configuration is a triangle, with
2b
if 0 x 21 L
L x
y(x, 0) =
2b
1
L (L x) if 2 L x L ,
(14.85)
and y(x,
0) = 0. Then Bn = 0 for all n, while
An =
2
L
1/2
2b
( ZL/2
ZL
)
nx
nx
+ dx (L x) sin
dx x sin
L
L
0
4b
= (2L)1/2 2 2 sin
n
L/2
1
2 n
n,odd ,
(14.86)
after changing variables to x = L/n and using sin d = d sin cos . Another
way to write this is to separately give the results for even and odd coefficients:
A2k = 0
A2k+1 =
(1)k
4b
1/2
(2L)
.
2
(2k + 1)2
(14.87)
14.4. STURM-LIOUVILLE THEORY
259
Figure 14.7: Evolution of a string with fixed ends starting from an isosceles triangle shape.
Note that each 2k (x) = 2k (L x) is antisymmetric about the midpoint x = 12 L, for
all k. Since our initial conditions are that y(x, 0) is symmetric about x = 12 L, none of the
even order eigenfunctions can enter into the expansion, precisely as we have found. The
dAlembert solution to this problem is particularly simple and is shown in Fig. 14.7. Note
that g(x) = 21 y(x, 0) must be extended to the entire real line. We know that g(x) = g(x+2L)
is periodic with
spatial
period 2L, but how to we extend g(x) from the interval 0, L to
the interval L, 0 ? To do this, we use y(x, 0) = g(x) g(x), which says that g(x) must
be antisymmetric, i.e. g(x) = g(x). Equivalently, y(x,
0) = cg (x) cg (x) = 0, which
integrates to g(x) = g(x).
14.4
Sturm-Liouville Theory
Consider the Lagrangian density
L=
1
2
(x) y 2 21 (x) y
1
2
v(x) y 2 .
(14.88)
The last term is new and has the physical interpretation of a harmonic potential which
attracts the string to the line y = 0. The Euler-Lagrange equations are then
y
2y
(14.89)
(x)
+ v(x) y = (x) 2 .
x
x
t
This equation is invariant under time translation. Thus, if y(x, t) is a solution, then so is
y(x, t + t0 ), for any t0 . This means that the solutions can be chosen to be eigenstates of
the operator t , which is to say y(x, t) = (x) eit . Because the coefficients are real, both
y and y are solutions, and taking linear combinations we have
y(x, t) = (x) cos(t + ) .
Plugging this into eqn. 14.89, we obtain
i
d h
(x) (x) + v(x) (x) = 2 (x) (x) .
dx
(14.90)
(14.91)
260
CHAPTER 14. CONTINUUM MECHANICS
This is the Sturm-Liouville equation. There are four types of boundary conditions that we
shall consider:
1. Fixed endpoint: (x) = 0, where x = xa,b .
2. Natural: (x) (x) = 0, where x = xa,b .
3. Periodic: (x) = (x + L), where L = xb xa .
4. Mixed homogeneous: (x) + (x) = 0, where x = xa,b .
The Sturm-Liouville equation is an eigenvalue equation. The eigenfunctions {n (x)} satisfy
i
d h
(x) n (x) + v(x) n (x) = n2 (x) n (x) .
dx
(14.92)
i
dh
2
(x) m
(x) + v(x) m (x) = m
(x) m (x) .
dx
(14.93)
Now suppose we a second solution m (x), satisfying
Now multiply (14.92) by m (x) and (14.93) by n (x) and subtract, yielding
n
d h i
dh i
2
m n
m m
n = n 2 m
dx
dx
i
d h
=
n m m n .
dx
(14.94)
(14.95)
We integrate this equation over the length of the string, to get
2
n 2 m
Zxb
h
ix=xb
dx (x) n (x) m (x) = (x) n (x) m
(x) (x) m (x) n (x)
x=xa
xa
=0.
(14.96)
The RHS vanishes for any of the four types of boundary conditions articulated above.
Thus, we have
2
n 2 m
where the inner product is defined as
n m = 0 ,
Zxb
dx (x) (x) (x) .
(14.97)
(14.98)
xa
Note that the distribution (x) is non-negative definite. Setting m = n, we have n n
2 6= 2 , the eigenfunctions are
0, and hence n 2 = n2 , which says that n2 R. When m
n
orthogonal with respect to the above inner product. In the case of degeneracies, we may
invoke the Gram-Schmidt procedure, which orthogonalizes the eigenfunctions within a given
14.4. STURM-LIOUVILLE THEORY
261
degenerate subspace. Since the Sturm-Liouville equation is linear, we may normalize the
eigenfunctions, taking
m n = mn .
(14.99)
Finally, since the coefficients in the Sturm-Liouville equation are all real, we can and henceforth do choose the eigenfunctions themselves to be real.
Another important result, which we will not prove here, is the completeness of the eigenfunction basis. Completeness means
X
(x)
n (x) n (x ) = (x x ) .
(14.100)
n
Thus, any function can be expanded in the eigenbasis, viz.
X
(x) =
Cn n (x)
,
Cn = n .
(14.101)
14.4.1
Variational method
Consider the functional
2 (x) =
1
2
o
Rxb n
dx (x) 2 (x) + v(x) 2 (x)
xa
1
2
Rxb
dx (x) 2 (x)
N
.
D
(14.102)
xa
The variation is
2 =
=
N D
N
D
D2
N 2 D
.
D
(14.103)
Thus,
2 = 0
which says
N = 2 D ,
d
d(x)
(x)
+ v(x) (x) = 2 (x) (x) ,
dx
dx
(14.104)
(14.105)
which is the Sturm-Lioiuville equation. In obtaining this equation, we have dropped a
boundary term, which is correct provided
h
(x) (x) (x)
ix=x
x=xa
=0.
(14.106)
This condition is satisfied for any of the first three classes of boundary conditions: = 0
(fixed endpoint), = 0 (natural), or (xa ) = (xb ), (xa ) = (xb ) (periodic). For
262
CHAPTER 14. CONTINUUM MECHANICS
the fourth class of boundary conditions, + = 0 (mixed homogeneous), the SturmLiouville equation may still be derived, provided one uses a slightly different functional,
e
N
2 (x) =
D
h
i
e = N + x 2 x xa 2 xa ,
N
b
b
2
with
since then
(14.107)
)
Zxb (
d
d(x)
2
e N
e D = dx
N
(x)
+ v(x) (x) (x) (x) (x)
dx
dx
xa
#x=x
b
,
+ (x) (x) + (x) (x)
"
(14.108)
x=xa
and the last term vanishes as a result of the boundary conditions.
For all four classes of boundary conditions we may write
K
}|
{
i
d
d
(x)
+ v(x) (x)
dx (x)
dx
dx
xa
2
(x) =
Rxb
dx (x) 2 (x)
z
h
Rxb
(14.109)
xa
If we expand (x) in the basis of eigenfunctions of the Sturm-Liouville operator K,
(x) =
n=1
we obtain
Cn n (x) ,
(x) = 2 (C1 , . . . , C ) =
2
(14.110)
Pj=1
|Cj |2 j2
2
k=1 |Ck |
(14.111)
If 12 22 . . ., then we see that 2 12 , so an arbitrary function (x) will always yield
an upper bound to the lowest eigenvalue.
As an example, consider a violin string (v = 0) with a mass m affixed in the center. We
write (x) = + m (x 21 L), hence
2 (x) =
RL
dx 2 (x)
0
m 2 ( 21 L) +
Now consider a trial function
(x) =
A x
A (L x)
if
if
RL
dx 2 (x)
(14.112)
0 x 12 L
1
2L
xL.
(14.113)
14.4. STURM-LIOUVILLE THEORY
263
Figure 14.8: One-parameter variational solution for a string with a mass m affixed at
x = 21 L.
The functional 2 (x) now becomes an ordinary function of the trial parameter , with
2
() =
m
R L/2
2 22
0 dx x
2
1
2L
L/2
R
dx x2
+ 2
2c
L
2
2 (2 + 1)
m ,
(2 1) 1 + (2 + 1) M
(14.114)
where M = L is the mass of the string alone. We minimize 2 () to obtain the optimal
solution of this form:
d 2
() = 0
d
42 2 1 + (2 + 1)2 ( 1)
m
=0.
M
(14.115)
For m/M 0, we obtain = 14 1 + 5 0.809. The variational estimate for the
eigenvalue is then 6.00% larger than the exact answer 10 = c/L. In the opposite limit,
m/M , the inertia of the string may be neglected. The normal mode is then piecewise
linear, in the shape of an isosceles triangle with base L and height y. The
p equation of
motion is then m
y = 2 (y/ 21 L), assuming |y/L| 1. Thus, 1 = (2c/L) M/m. This
is reproduced exactly by the variational solution, for which 1 as m/M .
264
14.5
CHAPTER 14. CONTINUUM MECHANICS
Continua in Higher Dimensions
In higher dimensions, we generalize the operator K as follows:
K=
(x) + v(x) .
x
x
(14.116)
The eigenvalue equation is again
K(x) = 2 (x) (x) ,
and the Greens function again satisfies
h
i
K 2 (x) G (x, x ) = (x x ) ,
(14.117)
(14.118)
and has the eigenfunction expansion,
X
n (x) n (x )
.
G (x, x ) =
n2 2
(14.119)
n=1
The eigenfunctions form a complete and orthonormal basis:
(x)
Z
n=1
n (x) n (x ) = (x x )
dx (x) m (x) n (x) = mn ,
(14.120)
(14.121)
where is the region of space in which the continuous medium
exists. For purposes of
simplicity, we consider here fixed boundary conditions u(x, t) = 0, where is the
boundary of . The general solution to the wave equation
2
(x) 2 (x) + v(x) u(x, t) = 0
(14.122)
t
x
x
is
u(x, t) =
n=1
Cn n (x) cos(n t + n ) .
The variational approach generalizes as well. We define
Z
2
+v
N (x) = dx
x x
Z
D (x) = dx 2 ,
(14.123)
(14.124)
(14.125)
and
N (x)
.
(x) =
D (x)
2
Setting the variation 2 = 0 recovers the eigenvalue equation K = 2 .
(14.126)
14.5. CONTINUA IN HIGHER DIMENSIONS
14.5.1
265
Membranes
Consider a surface where the height z is a function of the lateral coordinates x and y:
z = u(x, y) .
(14.127)
F (x, y, z) = z u(x, y) = 0 .
(14.128)
The equation of the surface is then
Let the differential element of surface area be dS. The projection of this element onto the
(x, y) plane is
dA = dx dy
z dS .
=n
(14.129)
is given by
The unit normal n
Thus,
F
z u
p
=
n
F = 1 + (u)2 .
dS =
dx dy p
= 1 + (u)2 dx dy .
z
n
(14.130)
(14.131)
The potential energy for a deformed surface can take many forms. In the case we shall
consider here, we consider only the effect of surface tension , and we write the potential
energy functional as
Z
U u(x, y, t) = dS
Z
1
= U0 + 2 dA (u)2 + . . . .
(14.132)
The kinetic energy functional is
Thus, the action is
T u(x, y, t) =
S u(x, t) =
where the Lagrangian density is
1
2
dA (x) (t u)2 .
d2x L(u, u, t u, x) ,
L = 21 (x) (t u)2 12 (x) (u)2 ,
(14.133)
(14.134)
(14.135)
where here we have allowed both (x) and (x) to depend on the spatial coordinates. The
equations of motion are
L
L
L
+
t t u
u u
n
o
2
u
= (x) 2 (x) u .
t
0=
(14.136)
(14.137)
266
14.5.2
CHAPTER 14. CONTINUUM MECHANICS
Helmholtz equation
When and are each constant, we obtain the Helmholtz equation:
1 2
c2 t2
u(x, t) = 0 ,
(14.138)
p
/. The dAlembert solution still works waves of arbitrary shape can
with c =
propagate in a fixed direction k:
x ct) .
u(x, t) = f (k
(14.139)
This is called a plane wave because the three dimensional generalization of this wave has
wavefronts which are planes. In our case, it might better be called a line wave, but people
will look at you funny if you say that, so well stick with plane wave. Note that the locus
of points of constant f satisfies
x ct = constant ,
(x, t) = k
(14.140)
dx = c ,
k
dt
(14.141)
and setting d = 0 gives
is c. The component of x perpendicular to k
is
which means that the velocity along k
arbitrary, hence the regions of constant correspond to lines which are orthogonal to k.
Owing to the linearity of the wave equation, we can construct arbitrary superpositions of
plane waves. The most general solution is written
u(x, t) =
i
d2k h
i(kxckt)
i(kx+ckt)
.
A(k) e
+ B(k) e
(2)2
(14.142)
The first term in the bracket on the RHS corresponds to a plane wave moving in the +k
direction.
direction, and the second term to a plane wave moving in the k
14.5.3
Rectangles
Consider a rectangular membrane where x [0, a] and y [0, b], and subject to the boundary conditions u(0, y) = u(a, y) = u(x, 0) = u(x, b) = 0. We try a solution of the form
u(x, y, t) = X(x) Y (y) T (t) .
(14.143)
This technique is known as separation of variables. Dividing the Helmholtz equation by u
then gives
1 2Y
1 1 2T
1 2X
+
=
.
(14.144)
X x2
Y y 2
c2 T t2
14.5. CONTINUA IN HIGHER DIMENSIONS
267
The first term on the LHS depends only on x. The second term on the LHS depends only
on y. The RHS depends only on t. Therefore, each of these terms must individually be
constant. We write
1 2X
= kx2
X x2
1 2Y
= ky2
Y y 2
1 2T
= 2 ,
T t2
(14.145)
with
2
.
c2
Thus, = c|k|. The most general solution is then
kx2 + ky2 =
(14.146)
X(x) = A cos(kx x) + B sin(kx x)
(14.147)
Y (y) = C cos(ky y) + D sin(ky y)
(14.148)
T (t) = E cos(t) + B sin(t) .
(14.149)
The boundary conditions now demand
A=0
C=0 ,
sin(kx a) = 0 ,
sin(ky b) = 0 .
Thus, the most general solution subject to the boundary conditions is
X
X
mx
ny
Amn sin
u(x, y, t) =
sin
cos mn t + mn ,
a
b
(14.150)
(14.151)
m=1 n=1
where
mn =
14.5.4
s
mc
a
2
nc
b
2
(14.152)
Circles
For a circular membrane, such as a drumhead, it is convenient to work in two-dimensional
polar coordinates (r, ). The Laplacian is then
2 =
1 2
r
+ 2
.
r r r r 2
(14.153)
We seek a solution to the Helmholtz equation which satisfies the boundary conditions u(r =
a, , t) = 0. Once again, we invoke the separation of variables method, writing
resulting in
u(r, , t) = R(r) () T (t) ,
(14.154)
1 1 2T
1 1
R
1 1 2
=
.
r
+ 2
R r r
r
r 2
c2 T t2
(14.155)
The azimuthal and temporal functions are
() = eim
T (t) = cos(t + ) ,
(14.156)
268
CHAPTER 14. CONTINUUM MECHANICS
where m is an integer in order that the function u(r, , t) be single-valued. The radial
equation is then
2
2R 1 R
m2
+
+
2 R=0.
(14.157)
r 2
r r
c2
r
This is Bessels equation, with solution
R(r) = A Jm
r
c
+ B Nm
r
c
(14.158)
where Jm (z) and Nm (z) are the Bessel and Neumann functions of order m, respectively.
Since the Neumann functions diverge at r = 0, we must exclude them, setting B = 0 for
each m.
We now invoke the boundary condition u(r = a, , t) = 0. This requires
a
c
=0
=
= m = xm ,
Jm
c
a
(14.159)
where Jm (xm ) = 0, i.e. xm is the th zero of Jm (x). The mose general solution is therefore
u(r, , t) =
m=0 =1
14.5.5
Am Jm xm r/a cos m + m cos(m t + m .
(14.160)
Sound in fluids
Let (x, t) and v(x, t) be the density and velocity fields in a fluid. Mass conservation
requires
+ ( v) = 0 .
(14.161)
t
This is the continuity equation for mass.
Focus now on a small packet of fluid
of infinitesimal volume dV . The total force on this
fluid element is dF = p + g dV . By Newtons Second Law,
dF = dV
Note that the chain rule gives
dv
dt
v
dv
=
+ v v .
dt
t
Thus, dividing eqn, 14.162 by dV , we obtain
v
+ v v = p + g .
This is the inviscid (i.e. zero viscosity) form of the Navier-Stokes equation.
(14.162)
(14.163)
(14.164)
14.5. CONTINUA IN HIGHER DIMENSIONS
269
Locally the fluid can also be described in terms of thermodynamic variables p(x, t) (pressure)
and T (x, t) (temperature). For a one-component fluid there is necessarily an equation of
state of the form p = p(, T ). Thus, we may write
p
p
dp =
d +
dT .
(14.165)
T
T
We now make the following approximations. First, we assume that the fluid is close to
equilibrium at v = 0, meaning we write p = p + p and = + , and assume that p, ,
and v are small. The smallness of v means we can neglect the nonlinear term (v )v in
eqn. 14.164. Second, we neglect gravity (more on this later). The continuity equation then
takes the form
+ v = 0 ,
(14.166)
t
and the Navier-Stokes equation becomes
v
= p .
t
(14.167)
Taking the time derivative of the former, and then invoking the latter of these equations
yields
p
2
2
= p=
2 c2 2 .
(14.168)
2
t
The speed of wave propagation, i.e. the speed of sound, is given by
s
p
.
c=
(14.169)
Finally, we must make an assumption regarding the conditions under which the derivative
p/ is computed. If the fluid is an excellent conductor of heat, then the temperature will
equilibrate quickly and it is a good approximation to take the derivative at fixed temperature. The resulting value of c is called the isothermal sound speed cT . If, on the other
hand, the fluid is a poor conductor of heat, as is the case for air, then it is more appropriate
to take the derivative at constant entropy, yielding the adiabatic sound speed. Thus,
s
s
p
p
cT =
, cS =
.
(14.170)
T
S
In an ideal gas, cS /cT = , where = cp /cV is the ratio of the specific heat at constant
pressure to that at constant volume. For a (mostly) diatomic gas like air (comprised of N2
and O2 and just a little Ar), = 57 . Note that one can write c2 = 1/, where
1
=
(14.171)
p
is the compressibility, which is the inverse of the bulk modulus. Again, one must specify
whether one is talking about T or S . For reference in air at T = 293 K, using M =
270
CHAPTER 14. CONTINUUM MECHANICS
28.8 g/mol, one obtains cT = 290.8 m/s and cS = 344.0 m/s. In H2 O at 293 K, c = 1482 m/s.
In Al at 273 K, c = 6420 m/s.
If we retain gravity, the wave equation becomes
2
= c2 2 g .
t2
(14.172)
The dispersion relation is then
(k) =
c2 k2 + ig k .
(14.173)
We are permitted to ignore the effects of gravity so long as c2 k2 gk. In terms of the
wavelength = 2/k, this requires
14.6
2c2
= 75.9 km (at T = 293 K) .
g
(14.174)
Dispersion
The one-dimensional Helmholtz equation x2 y = c2 t2 y is solved by a plane wave
y(x, t) = A eikx eit ,
(14.175)
provided = ck. We say that there are two branches to the dispersion relation (k)
for this equation. In general, we may add solutions, due to the linearity of the Helmholtz
equation. The most general solution is then
y(x, t) =
i
dk h
f (k) eik(xct) + g(k) eik(x+ct)
2
= f (x ct) + g(x + ct) ,
(14.176)
which is consistent with dAlemberts solution.
Consider now the free particle Schr
odinger equation in one space dimension,
i~
~2 2
=
.
t
2m x2
(14.177)
The function (x, t) is the quantum mechanical wavefunction for a particle of mass m
moving freely along a one-dimensional line. The probability density for finding the particle
at position x at time t is
2
(x, t) = (x, t) .
(14.178)
Conservation of probability therefore requires
Z
2
dx (x, t) = 1 .
(14.179)
14.6. DISPERSION
271
This condition must hold at all times t.
As is the case with the Helmholtz equation, the Schr
odinger equation is solved by a plane
wave of the form
(x, t) = A eikx eit ,
(14.180)
where the dispersion relation now only has one branch, and is given by
(k) =
~k2
.
2m
(14.181)
The most general solution is then
(x, t) =
dk
2
(k) eikx ei~k t/2m .
2
(14.182)
Lets suppose we start at time t = 0 with a Gaussian wavepacket,
(x, 0) = 20
1/4
ex
2 /22
0
eik0 x .
(14.183)
To find the amplitude (k),
we perform the Fourier transform:
(k)
=
Z
dx (x, 0) eikx
1/4 (kk )2 2 /2
0
0
2 20
e
.
(14.184)
We now compute (x, t) valid for all times t:
1/4
20
dk ikx (kk0 )2 20 /2 ikx i~k2 t/2m
e e
e e
2
"
2 #
1/2
x ~k0 t/m
2 1/4
= 0
exp 2
1 + it/
2 0 1 + t2 / 2
"
#
i 2k0 20 x + x2 t/ k02 40 t/
exp
,
2 20 1 + t2 / 2
(x, t) =
(14.185)
(14.186)
where m20 /~. The probability density is then the normalized Gaussian
where v0 = ~k0 /m and
(x, t) = p
2 /2 (t)
2 (t)
(t) = 0
e(xv0 t)
1 + t2 / 2 .
(14.187)
(14.188)
Note that (t) gives the width of the wavepacket, and that this width increases as a function
of time, with (t ) 0 t/ .
272
CHAPTER 14. CONTINUUM MECHANICS
Figure 14.9: Wavepacket spreading for k0 0 = 2 with t/ = 0, 2, 4, 6, and 8.
Unlike the case of the Helmholtz equation, the solution to the Schr
odinger equation does not
retain its shape as it moves. This phenomenon is known as the spreading of the wavepacket.
In fig. 14.9, we show the motion and spreading of the wavepacket.
For a given plane wave eikx ei(k)t , the wavefronts move at the phase velocity
vp (k) =
(k)
.
k
The center of the wavepacket, however, travels at the group velocity
d
vg (k) =
,
dk k
(14.189)
(14.190)
2 .
where k = k0 is the maximum of (k)
14.7
Appendix I : Three Strings
Problem: Three identical strings are connected to a ring of mass m as shown in fig. 14.10.
The linear mass density of each string is and each string is under identical tension . In
equilibrium, all strings are coplanar. All motion on the string is in the z-direction,
which is
perpendicular to the equilibrium plane. The ring slides frictionlessly along a vertical pole.
It is convenient to describe each string as a half line [, 0]. We can choose coordinates
x1 , x2 , and x3 for the three strings, respectively. For each string, the ring lies at xi = 0.
14.7. APPENDIX I : THREE STRINGS
273
A pulse is sent down the first string. After a time, the pulse arrives at the ring. Transmitted
waves are sent down the other two strings, and a reflected wave down the first string. The
solution to the wave equation in the strings can be written as follows. In string #1, we have
z = f (ct x1 ) + g(ct + x1 ) .
(14.191)
In the other two strings, we may write z = hA (ct + x2 ) and z = hB (ct + x3 ), as indicated in
the figure.
Figure 14.10: Three identical strings arranged symmetrically in a plane, attached to a
common end. All motion is in the direction perpendicular to this plane. The red ring,
whose mass is m, slides frictionlessly in this direction along a pole.
(a) Write the wave equation in string #1. Define all constants.
(b) Write the equation of motion for the ring.
(c) Solve for the reflected wave g() in terms of the incident wave f (). You may write this
relation in terms of the Fourier transforms f(k) and g(k).
(d) Suppose a very long wavelength pulse of maximum amplitude A is incident on the ring.
What is the maximum amplitude of the reflected pulse? What do we mean by very long
wavelength?
Solution:
274
CHAPTER 14. CONTINUUM MECHANICS
(a) The wave equation is
2z
1 2z
=
,
(14.192)
x2
c2 t2
p
where x is the coordinate along the string, and c = / is the speed of wave propagation.
(b) Let Z be the vertical coordinate of the ring. Newtons second law says mZ = F , where
the force on the ring is the sum of the vertical components of the tension in the three strings
at x = 0:
h
i
F = f (ct) + g (ct) + hA (ct) + hB (ct) ,
(14.193)
where prime denotes differentiation with respect to argument.
(c) To solve for the reflected wave, we must eliminate the unknown functions hA,B and
then obtain g in terms of f . This is much easier than it might at first seem. We start by
demanding continuity at the ring. This means
Z(t) = f (ct) + g(ct) = hA (ct) = hB (ct)
(14.194)
for all t. We can immediately eliminate hA,B :
hA () = hB () = f () + g() ,
for all . Newtons second law from part (b) may now be written as
mc2 f () + g () = f () + 3g () .
(14.195)
(14.196)
This linear ODE becomes a simple linear algebraic equation for the Fourier transforms,
f () =
etc. We readily obtain
dk
f (k) eik ,
2
k iQ
g(k) =
k 3iQ
f(k) ,
(14.197)
(14.198)
where Q /mc2 has dimensions of inverse length. Since hA,B = f + g, we have
2iQ
hA (k) = hB (k) =
f(k) .
(14.199)
k 3iQ
(d) For a very long wavelength pulse, composed of plane waves for which |k| Q, we
have g(k) 13 f(k). Thus, the reflected pulse is inverted, and is reduced by a factor 13 in
amplitude. Note that for a very short wavelength pulse, for which k Q, we have perfect
reflection with inversion, and no transmission. This is due to the inertia of the ring.
It is straightforward to generalize this problem to one with n strings. The transmission
into each of the (n 1) channels is of course identical (by symmetry). One then finds the
reflection and transmission amplitudes
2iQ
k i(n 2)Q
,
t(k) =
.
(14.200)
r(k) =
k inQ
k inQ
14.8. APPENDIX II : GENERAL FIELD THEORETIC FORMULATION
275
Conservation of energy means that the sum of the squares of the reflection amplitude and
all the (n 1) transmission amplitudes must be unity:
r(k) 2 + (n 1) t(k) 2 = 1 .
(14.201)
14.8
Appendix II : General Field Theoretic Formulation
Continuous systems possess an infinite number of degrees of freedom. They are described
by a set of fields a (x, t) which depend on space and time. These fields may represent
local displacement, pressure, velocity, etc. The equations of motion of the fields are again
determined by extremizing the action, which, in turn, is an integral of the Lagrangian
density over all space and time. Extremization yields a set of (generally coupled) partial
differential equations.
14.8.1
Euler-Lagrange equations for classical field theories
Suppose a (x) depends on n independent variables, {x1 , x2 , . . . , xn }. Consider the functional
Z
S {a (x) = dx L(a a , x) ,
(14.202)
i.e. the Lagrangian density L is a function of the fields a and their partial derivatives
a /x . Here is a region in Rn . Then the first variation of S is
)
(
Z
L a
L
+
S = dx
a a ( a ) x
(
)
Z
I
L
L
L
a ,
(14.203)
+ dx
= d n
( a ) a
a x ( a )
where is the (n 1)-dimensional boundary of ,
surface area, and
d is the differential
n is the unit normal. If we demand L/( a ) = 0 of a = 0, the surface term
vanishes, and we conclude
"
#
S
L
L
=
,
(14.204)
a (x)
a x ( a )
x
where the subscript means we are to evaluate the term in brackets at x. In a mechanical
system, one of the n independent variables (usually x0 ), is the time t. However, we may
be interested in a time-independent context in which we wish to extremize the energy
functional, for example. In any case, setting the first variation of S to zero yields the
Euler-Lagrange equations,
L
=0
(14.205)
S = 0
a x ( a )
276
CHAPTER 14. CONTINUUM MECHANICS
The Lagrangian density for an electromagnetic field with sources is
1
F F J A .
L = 16
The equations of motion are then
L
L
=0
A
x ( A )
F = 4J ,
(14.206)
(14.207)
which are Maxwells equations.
14.8.2
Conserved currents in field theory
Recall the result of Noethers theorem for mechanical systems:
!
d L q
=0,
dt q
(14.208)
=0
where q = q (q, ) is a one-parameter () family of transformations of the generalized
coordinates which leaves L invariant. We generalize to field theory by replacing
q (t) a (x, t) ,
(14.209)
where {a (x, t)} are a set of fields, which are functions of the independent variables {x, y, z, t}.
We will adopt covariant relativistic notation and write for four-vector x = (ct, x, y, z). The
generalization of dQ/dt = 0 is
!
a
L
=0,
(14.210)
x ( a )
=0
where there is an implied sum on both and a. We can write this as J = 0, where
a
L
.
(14.211)
J
( a )
=0
We call Q = J 0 /c the total charge. If we assume J = 0 at the spatial boundaries of our
system, then integrating the conservation law J over the spatial region gives
Z
Z
I
dQ
3
0
3
J =0 ,
= d x 0 J = d x J = d n
(14.212)
dt
assuming J = 0 at the boundary .
As an example, consider the case of a complex scalar field, with Lagrangian density2
(14.213)
L(, , , , ) = 12 K ( )( ) U .
2
We raise and lower indices using the Minkowski metric g = diag (+, , , ).
14.8. APPENDIX II : GENERAL FIELD THEORETIC FORMULATION
277
This is invariant under the transformation ei , ei . Thus,
= i ei
= i ei ,
(14.214)
and, summing over both and fields, we have
L
L
(i) +
(i )
( )
( )
K
=
.
2i
J =
(14.215)
The potential, which depends on ||2 , is independent of . Hence, this form of conserved
4-current is valid for an entire class of potentials.
14.8.3
Gross-Pitaevskii model
As one final example of a field theory, consider the Gross-Pitaevskii model, with
L = i~
2
~2
g ||2 n0 .
t
2m
(14.216)
This describes a Bose fluid with repulsive short-ranged interactions. Here (x, t) is again
a complex scalar field, and is its complex conjugate. Using the Leibniz rule, we have
S[ , ] = S[ + , + ]
Z Z
~2
~2
d
= dt d x i~
+ i~
t
t
2m
2m
2g ||2 n0 ( + )
(
Z Z
~2 2
d
2
= dt d x
i~
+
2g || n0
t
2m
)
~2 2
+ i~
+
2g ||2 n0 ,
(14.217)
t
2m
where we have integrated by parts where necessary and discarded the boundary terms.
Extremizing S[ , ] therefore results in the nonlinear Schr
odinger equation (NLSE),
i~
~2 2
=
+ 2g ||2 n0
t
2m
(14.218)
as well as its complex conjugate,
i~
~2 2
=
+ 2g ||2 n0 .
t
2m
(14.219)
278
CHAPTER 14. CONTINUUM MECHANICS
Note that these equations are indeed the Euler-Lagrange equations:
S
L
S
=
(14.220)
(14.221)
with x = (t, x)3 Plugging in
L
= 2g ||2 n0
L
= i~
t
L
~2
=
2m
(14.222)
L
~2
,
=
2m
(14.223)
and
L
= i~ 2g ||2 n0
L
=0 ,
t
we recover the NLSE and its conjugate.
The Gross-Pitaevskii model also possesses a U(1) invariance, under
t) = ei (x, t)
(x, t) (x,
(x, t) (x, t) = ei (x, t) .
(14.224)
Thus, the conserved Noether current is then
L
J =
J 0 = ~ ||2
J =
=0
L
+
=0
~2
.
2im
(14.225)
(14.226)
Dividing out by ~, taking J 0 ~ and J ~j, we obtain the continuity equation,
+j =0 ,
t
where
= ||2
j=
~
.
2im
are the particle density and the particle current, respectively.
3
In the nonrelativistic case, there is no utility in defining x0 = ct, so we simply define x0 = t.
(14.227)
(14.228)
14.9. APPENDIX III : GREENS FUNCTIONS
14.9
279
Appendix III : Greens Functions
Suppose we add a forcing term,
i
h
2y
y
(x) 2
(x)
+ v(x) y = Re (x) f (x) eit .
t
x
x
We write the solution as
where
or
i
h
y(x, t) = Re y(x) eit ,
(14.229)
(14.230)
dy(x)
d
(x)
+ v(x) y(x) 2 (x) y(x) = (x) f (x) ,
dx
dx
h
i
K 2 (x) y(x) = (x) f (x) ,
(14.231)
(14.232)
where K is a differential operator,
d
d
(x)
+ v(x) .
dx
dx
(14.233)
Note that the eigenfunctions of K are the {n (x)}:
K n (x) = n2 (x) n (x) .
(14.234)
The formal solution to equation 14.232 is then
h
i1
y(x) = K 2
(x ) f (x )
x,x
Zxb
= dx (x ) G (x, x ) f (x ).
(14.235)
(14.236)
xa
What do we mean by the term in brackets? If we define the Greens function
h
i1
,
G (x, x ) K 2
x,x
what this means is
i
K 2 (x) G (x, x ) = (x x ) .
(14.237)
(14.238)
Note that the Greens function may be expanded in terms of the (real) eigenfunctions, as
G (x, x ) =
X n (x) n (x )
n
n2 2
(14.239)
which follows from completeness of the eigenfunctions:
(x)
n=1
n (x) n (x ) = (x x ) .
(14.240)
280
CHAPTER 14. CONTINUUM MECHANICS
The expansion in eqn. 14.239 is formally exact, but difficult to implement, since it requires
summing over an infinite set of eigenfunctions. It is more practical to construct the Greens
function from solutions to the homogeneous Sturm Liouville equation, as follows. When
x 6= x , we have that (K 2 ) G (x, x ) = 0, which is a homogeneous ODE of degree two.
Consider first the interval x [xa , x ]. A second order homogeneous ODE has two solutions,
and further invoking the boundary condition at x = xa , there is a unique solution, up to
a multiplicative constant. Call this solution y1 (x). Next, consider the interval x [x , xb ].
Once again, there is a unique solution to the homogeneous Sturm-Liouville equation, up
to a multiplicative constant, which satisfies the boundary condition at x = xb . Call this
solution y2 (x). We then can write
A(x ) y1 (x) if xa x < x
(14.241)
G (x, x ) =
B(x ) y2 (x) if x < x xb .
Here, A(x ) and B(x ) are undetermined functions. We now invoke the inhomogeneous
Sturm-Liouville equation,
dG (x, x )
d
(x)
+ v(x) G (x, x ) 2 (x) G (x, x ) = (x x ) .
(14.242)
dx
dx
We integrate this from x = x to x = x + , where is a positive infinitesimal. This
yields
h
i
(x ) A(x ) y1 (x ) B(x ) y2 (x ) = 1 .
(14.243)
Continuity of G (x, x ) itself demands
A(x ) y1 (x ) = B(x ) y2 (x ) .
(14.244)
Solving these two equations for A(x ) and B(x ), we obtain
A(x ) =
y2 (x )
(x ) Wy1 ,y2 (x )
B(x ) =
y1 (x )
,
(x ) Wy1 ,y2 (x )
(14.245)
where Wy1 ,y2 (x) is the Wronskian
y1 (x) y2 (x)
Wy1 ,y2 (x) = det
y1 (x) y2 (x)
= y1 (x) y2 (x) y2 (x) y1 (x) .
(14.246)
Now it is easy to show that Wy1 ,y2 (x) (x) = W is a constant. This follows from the fact
that
0 = y2 K y1 y2 K y1
i
h
d
.
(x) y1 y2 y2 y1
=
dx
(14.247)
14.9. APPENDIX III : GREENS FUNCTIONS
Thus, we have
G (x, x ) =
or, in compact form,
281
y1 (x) y2 (x )/W
y1
(x ) y
2 (x)/W
G (x, x ) =
where x< = min(x, x ) and x> = max(x, x ).
if xa x < x
if
(14.248)
< x xb ,
y1 (x< ) y2 (x> )
,
W
(14.249)
As an example, consider a uniform string (i.e. and constant, v = 0) with fixed endpoints
at xa = 0 and xb = L. The normalized eigenfunctions are
n (x) =
2
sin
L
nx
L
(14.250)
and the eigenvalues are n = nc/L. The Greens function is
G (x, x ) =
2 X sin(nx/L) sin(nx /L)
.
L
(nc/L)2 2
(14.251)
n=1
Now construct the homogeneous solutions:
2
(K ) y1 = 0
y1 (0) = 0
(K 2 ) y2 = 0 ,
y2 (L) = 0
The Wronskian is
W = y1 y2 y2 y1 =
x
y1 (x) = sin
c
(L x)
.
y2 (x) = sin
c
sin
c
L
c
(14.252)
(14.253)
(14.254)
Therefore, the Greens function is
sin x< /c sin (L x> )/c
G (x, x ) =
.
( /c) sin(L/c)
14.9.1
(14.255)
Perturbation theory
Suppose we have solved for the Greens function for the linear operator K0 and mass density
0 (x). I.e. we have
K0 2 0 (x) G0 (x, x ) = (x x ) .
(14.256)
We now imagine perturbing 0 0 + 1 , v0 v0 + v2 , 0 0 + 1 . What is the new
Greens function G (x, x )? We must solve
L0 + L1 G (x, x ) = (x x ) ,
(14.257)
282
CHAPTER 14. CONTINUUM MECHANICS
Figure 14.11: Diagrammatic representation of the perturbation expansion in eqn. 14.260..
where
L0 K0 2 0
(14.258)
L1 K1 2 1 .
(14.259)
Dropping the subscript for simplicity, the full Greens function is then given by
i1
h
G = L0 + L1
i1
h
1
= G0
+ L1
i1
h
G0
= 1 + G0 L1
= G0 G0 L1 G0 + 2 G0 L1 G0 L1 G0 + . . . .
(14.260)
The matrix multiplication is of course a convolution, i.e.
G (x, x ) =
G0 (x, x )
Zxb
0
d
G (x1 , x ) + . . . .
dx1 G0 (x, x1 ) L1 x1 , dx
1
(14.261)
xa
Each term in the perturbation expansion of eqn. 14.260 may be represented by a diagram,
as depicted in Fig. 14.11.
As an example, consider a string with xa = 0 and xb = L with a mass point m affixed at
the point x = d. Thus, 1 (x) = m (x d), and L1 = m 2 (x d), with = 1. The
perturbation expansion gives
G (x, x ) = G0 (x, x ) + m 2 G0 (x, d) G0 (d, x ) + m2 4 G0 (x, d) G0 (d, d) G0 (d, x ) + . . .
= G0 (x, x ) +
m 2 G0 (x, d) G0 (d, x )
.
1 m 2 G0 (d, d)
(14.262)
14.9. APPENDIX III : GREENS FUNCTIONS
283
Note that the eigenfunction expansion,
G (x, x ) =
X n (x) n (x )
n2 2
(14.263)
says that the exact eigenfrequencies are poles of G (x, x ), and furthermore the residue at
each pole is
1
Res G (x, x ) =
n (x) n (x ) .
(14.264)
=n
2n
According to eqn. 14.262, the poles of G (x, x ) are located at solutions to4
m 2 G0 (d, d) = 1 .
For simplicity let us set d =
according to eqn. 14.255,
1
2 L,
G0
(14.265)
so the mass point is in the middle of the string. Then
1
1
2 L, 2 L
sin2 (L/2c)
( /c) sin(L/c)
L
c
tan
=
.
2
2c
(14.266)
The eigenvalue equation is therefore
tan
L
2c
2
,
mc
(14.267)
which can be manipulated to yield
m
= ctn ,
M
(14.268)
where = L/2c and M = L is the total mass of the string. When m = 0, the LHS
vanishes, and the roots lie at = (n + 12 ), which gives = 2n+1 . Why dont we see the
poles at the even mode eigenfrequencies 2n ? The answer is that these poles are present
in the Greens function. They do not cancel for d = 12 L because the perturbation does
not couple to the even modes, which all have 2n ( 21 L) = 0. The case of general d may be
instructive in this regard. One finds the eigenvalue equation
sin(2)
m
=
,
M
2 sin 2 sin 2(1 )
(14.269)
where = d/L. Now setting m = 0 we recover 2 = n, which says = n , and all the
modes are recovered.
4
Note in particular that there is no longer any divergence at the location of the original poles of G0 (x, x ).
These poles are cancelled.
284
14.9.2
CHAPTER 14. CONTINUUM MECHANICS
Perturbation theory for eigenvalues and eigenfunctions
We wish to solve
K0 + K1 = 2 0 + 1 ,
which is equivalent to
0 1
Multiplying by L
L0 = L1 .
(14.270)
(14.271)
= G0 on the left, we have
Zxb
(x) = dx G (x, x ) L1 (x )
(14.272)
xa
m (x)
2
2 m
m=1
Zxb
dx m (x ) L1 (x ) .
(14.273)
xa
We are free to choose any normalization we like for (x). We choose
which entails
n =
2
Zxb
dx 0 (x) n (x) (x) = 1 ,
(14.274)
xa
n2
Zxb
= dx n (x) L1 (x)
(14.275)
xa
as well as
(x) = n (x) +
k
(k6=n)
k (x)
2 k2
Zxb
dx k (x ) L1 (x ) .
(14.276)
xa
By expanding and 2 in powers of , we can develop an order by order perturbation
series.
To lowest order, we have
2
n2
Zxb
+ dx n (x) L1n n (x) .
(14.277)
xa
For the case L1 = m 2 (x d), we have
2
n
= 21 m n (d)
n
nd
m
.
sin2
=
M
L
(14.278)
For d = 21 L, only the odd n modes are affected, as the even n modes have a node at x = 21 L.
14.9. APPENDIX III : GREENS FUNCTIONS
285
Carried out to second order, one obtains for the eigenvalues,
2
n2
Zxb
+ dx n (x) L1n n (x)
xa
+ 2
2
R
xb
1 (x)
dx
(x)
L
X xa
k
n n
n2 k2
k
(k6=n)
+ O(3 )
Zxb
Zxb
2
1
dx n (x) Ln n (x) dx 1 (x ) n (x ) + O(3 ) .
2
xa
xa
(14.279)
286
CHAPTER 14. CONTINUUM MECHANICS
Chapter 15
Special Relativity
For an extraordinarily lucid, if characteristically brief, discussion, see chs. 1 and 2 of L. D.
Landau and E. M. Lifshitz, The Classical Theory of Fields (Course of Theoretical Physics,
vol. 2).
15.1
Introduction
All distances are relative in physics. They are measured with respect to a fixed frame of
reference. Frames of reference in which free particles move with constant velocity are called
inertial frames. The principle of relativity states that the laws of Nature are identical in
all inertial frames.
15.1.1
Michelson-Morley experiment
We learned how sound waves in a fluid, such as air, obey the Helmholtz equation. Let us
restrict our attention for the moment to solutions of the form (x, t) which do not depend
on y or z. We then have a one-dimensional wave equation,
1 2
2
=
.
(15.1)
x2
c2 t2
The fluid in which the sound propagates is assumed to be at rest. But suppose the fluid
is not at rest. We can investigate this by shifting to a moving frame, defining x = x ut,
with y = y, z = z and of course t = t. This is a Galilean transformation. In terms of the
new variables, we have
=
= u + .
,
x
x
t
x
t
The wave equation is then
1 2
2u 2
u2 2
.
=
1 2
c x 2
c2 t 2
c2 x t
287
(15.2)
(15.3)
288
CHAPTER 15. SPECIAL RELATIVITY
Clearly the wave equation acquires a different form when expressed in the new variables
(x , t ), i.e. in a frame in which the fluid is not at rest. The general solution is then of the
modified dAlembert form,
(x , t ) = f (x cR t ) + g(x + cL t ) ,
(15.4)
where cR = c u and cL = c + u are the speeds of rightward and leftward propagating
disturbances, respectively. Thus, there is a preferred frame of reference the frame in
which the fluid is at rest. In the rest frame of the fluid, sound waves travel with velocity c
in either direction.
Light, as we know, is a wave phenomenon in classical physics. The propagation of light is
described by Maxwells equations,
1 B
c t
4
1 E
B =
j+
,
c
c t
E = 4
E =
B =0
(15.5)
(15.6)
where and j are the local charge and current density, respectively. Taking the curl of
Faradays law, and restricting to free space where = j = 0, we once again have (using a
Cartesian system for the fields) the wave equation,
2E =
1 2E
.
c2 t2
(15.7)
(We shall discuss below, in section 15.8, the beautiful properties of Maxwells equations
under general coordinate transformations.)
In analogy with the theory of sound, it was assumed prior to Einstein that there was in
fact a preferred reference frame for electromagnetic radiation one in which the medium
which was excited during the EM wave propagation was at rest. This notional medium was
called the lumineferous ether . Indeed, it was generally assumed during the 19th century
that light, electricity, magnetism, and heat (which was not understood until Boltzmanns
work in the late 19th century) all had separate ethers. It was Maxwell who realized that
light, electricity, and magnetism were all unified phenomena, and accordingly he proposed
a single ether for electromagnetism. It was believed at the time that the earths motion
through the ether would result in a drag on the earth.
In 1887, Michelson and Morley set out to measure the changes in the speed of light on earth
due to the earths movement through the ether (which was generally assumed to be at rest
in the frame of the Sun). The Michelson interferometer is shown in fig. 15.1, and works as
through the ether. Then the
follows. Suppose the apparatus is moving with velocity u x
time it takes a light ray to travel from the half-silvered mirror to the mirror on the right
and back again is
2c
+
= 2
.
(15.8)
tx =
c+u cu
c u2
15.1. INTRODUCTION
289
Figure 15.1: The Michelson-Morley experiment (1887) used an interferometer to effectively
measure the time difference for light to travel along two different paths. Inset: analysis for
the y-directed path.
For motion q
along the other arm of the interferometer, the geometry in the inset of fig. 15.1
shows = 2 + 41 u2 t2y , hence
ty =
2
2
=
c
c
2 + 14 u2 t2y
ty =
2
.
u2
c2
(15.9)
Thus, the difference in times along these two paths is
t = tx ty =
2
u2
2c
c2
c c2
c2 u2
(15.10)
Thus, the difference in phase between the two paths is
u2
= t 2 ,
2
c
(15.11)
where is the wavelength of the light. We take u 30 km/s, which is the earths orbital
velocity, and 5000
A. From this we find that 0.02 2 if = 1 m. Michelson
and Morley found that the observed fringe shift /2 was approximately 0.02 times the
expected value. The inescapable conclusion was that the speed of light did not depend on
the motion of the source. This was very counterintuitive!
290
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.2: Experimental setup of Alvager et al. (1964), who used the decay of high energy
neutral pions to test the source velocity dependence of the speed of light.
The history of the development of special relativity is quite interesting, but we shall not have
time to dwell here on the many streams of scientific thought during those exciting times.
Suffice it to say that the Michelson-Morley experiment, while a landmark result, was not the
last word. It had been proposed that the ether could be dragged, either entirely or partially,
by moving bodies. If the earth dragged the ether along with it, then there would be no
ground-level ether wind for the MM experiment to detect. Other experiments, however,
such as stellar aberration, in which the apparent position of a distant star varies due to the
earths orbital velocity, rendered the ether drag theory untenable the notional ether
bubble dragged by the earth could not reasonably be expected to extend to the distant
stars.
A more recent test of the effect of a moving source on the speed of light was performed
by T. Alv
ager et al., Phys. Lett. 12, 260 (1964), who measured the velocity of -rays
(photons) emitted from the decay of highly energetic neutral pions ( 0 ). The pion energies
were in excess of 6 GeV, which translates to a velocity of v = 0.99975 c, according to special
relativity. Thus, photons emitted in the direction of the pions should be traveling at close
to 2c, if the source and photon velocities were to add. Instead, the velocity of the photons
was found to be c = 2.9977 0.0004 1010 cm/s, which is within experimental error of the
best accepted value.
15.1.2
Einsteinian and Galilean relativity
The Principle of Relativity states that the laws of nature are the same when expressed in
any inertial frame. This principle can further be refined into two classes, depending on
whether one takes the velocity of the propagation of interactions to be finite or infinite.
15.1. INTRODUCTION
291
Figure 15.3: Two reference frames.
The interaction of matter in classical mechanics is described by a potential function U (r1 , . . . , rN ).
P
Typically, one has two-body interactions in which case one writes U = i<j U (ri , rj ). These
interactions are thus assumed to be instantaneous, which is unphysical. The interaction of
particles is mediated by the exchange of gauge bosons, such as the photon (for electromagnetic interactions), gluons (for the strong interaction, at least on scales much smaller
than the confinement length), or the graviton (for gravity). Their velocity of propagation,
according to the principle of relativity, is the same in all reference frames, and is given by
the speed of light, c = 2.998 108 m/s.
Since c is so large in comparison with terrestrial velocities, and since d/c is much shorter
than all other relevant time scales for typical interparticle separations d, the assumption
of an instantaneous interaction is usually quite accurate. The combination of the principle
of relativity with finiteness of c is known as Einsteinian relativity. When c = , the
combination comprises Galilean relativity:
c<
c=
Einsteinian relativity
Galilean relativity .
Consider a train moving at speed u. In the rest frame of the train track, the speed of
the light beam emanating from the trains headlight is c + u. This would contradict the
principle of relativity. This leads to some very peculiar consequences, foremost among them
being the fact that events which are simultaneous in one inertial frame will not in general
be simultaneous in another. In Newtonian mechanics, on the other hand, time is absolute,
and is independent of the frame of reference. If two events are simultaneous in one frame
then they are simultaneous in all frames. This is not the case in Einsteinian relativity!
We can begin to apprehend this curious feature of simultaneity by the following Gedankenexperiment (a long German word meaning thought experiment)1 . Consider the case in
with respect to frame K. Let a source
fig. 15.3 in which frame K moves with velocity u x
at S emit a signal (a light pulse) at t = 0. In the frame K the signals arrival at equidistant
locations A and B is simultaneous. In frame K, however, A moves toward left-propagating
1
Unfortunately, many important physicists were German and we have to put up with a legacy of long
German words like Gedankenexperiment, Zitterbewegung, Brehmsstrahlung, Stosszahlansatz , Kartoffelsalat,
etc.
292
CHAPTER 15. SPECIAL RELATIVITY
emitted wavefront, and B moves away from the right-propagating wavefront. For classical
sound, the speed of the left-moving and right-moving wavefronts is cu, taking into account
the motion of the source, and thus the relative velocities of the signal and the detectors
remain at c. But according to the principle of relativity, the speed of light is c in all frames,
and is so in frame K for both the left-propagating and right-propagating signals. Therefore,
the relative velocity of A and the left-moving signal is c + u and the relative velocity of B
and the right-moving signal is c u. Therefore, A closes in on the signal and receives it
before B, which is moving away from the signal. We might expect the arrival times to be
tA = d/(c + u) and tB = d/(c u), where d is the distance between the source S and either
detector A or B in the K frame. Later on we shall analyze this problem and show that
r
r
cu d
c+u d
,
tB =
.
(15.12)
tA =
c+u c
cu c
Our nave analysis has omitted an important detail the Lorentz contraction of the distance
d as seen by an observer in the K frame.
15.2
Intervals
Now let us express mathematically the constancy of c in all frames. An event is specified
by the time and place where it occurs. Thus, an event is specified by four coordinates,
(t, x, y, z). The four-dimensional space spanned by these coordinates is called spacetime.
The interval between two events in spacetime at (t1 , x1 , y1 , z1 ) and (t2 , x2 , y2 , z2 ) is defined
to be
q
(15.13)
s12 = c2 (t1 t2 )2 (x1 x2 )2 (y1 y2 )2 (z1 z2 )2 .
For two events separated by an infinitesimal amount, the interval ds is infinitesimal, with
ds2 = c2 dt2 dx2 dy 2 dz 2 .
(15.14)
Now when the two events denote the emission and reception of an electromagnetic signal,
we have ds2 = 0. This must be true in any frame, owing to the invariance of c, hence since
ds and ds are differentials of the same order, we must have ds 2 = ds2 . This last result
requires homogeneity and isotropy of space as well. Finally, if infinitesimal intervals are
invariant, then integrating we obtain s = s , and we conclude that the interval between two
space-time events is the same in all inertial frames.
When s212 > 0, the interval is said to be time-like. For timelike intervals, we can always
find a reference frame in which the two events occur at the same locations. As an example,
consider a passenger sitting on a train. Event #1 is the passenger yawning at time t1 . Event
#2 is the passenger yawning again at some later time t2 . To an observer sitting in the train
station, the two events take place at different locations, but in the frame of the passenger,
they occur at the same location.
p
When s212 < 0, the interval is said to be space-like. Note that s12 = s212 iR is pure
imaginary, so one says that imaginary intervals are spacelike. As an example, at this
15.2. INTERVALS
293
Figure 15.4: A (1 + 1)dimensional light cone. The forward light cone consists of timelike
events with t > 0. The backward light cone consists of timelike events with t < 0. The
causally disconnected regions are time-like, and intervals connecting the origin to any point
on the light cone itself are light-like.
moment, in the frame of the reader, the North and South poles of the earth are separated
by a space-like interval. If the interval between two events is space-like, a reference frame
can always be found in which the events are simultaneous.
An interval with s12 = 0 is said to be light-like.
This leads to the concept of the light cone, depicted in fig. 15.4. Consider an event E. In the
frame of an inertial observer, all events with s2 > 0 and t > 0 are in Es forward light cone
and are part of his absolute future. Events with s2 > 0 and t < 0 lie in Es backward light
cone are are part of his absolute past. Events with spacelike separations s2 < 0 are causally
disconnected from E. Two events which are causally disconnected can not possible influence
each other. Uniform rectilinear motion is represented by a line t = x/v with constant slope.
If v < c, this line is contained within Es light cone. E is potentially influenced by all events
in its backward light cone, i.e. its absolute past. It is impossible to find a frame of reference
which will transform past into future, or spacelike into timelike intervals.
294
15.2.1
CHAPTER 15. SPECIAL RELATIVITY
Proper time
Proper time is the time read on a clock traveling with a moving observer. Consider two
observers, one at rest and one in motion. If dt is the differential time elapsed in the rest
frame, then
ds2 = c2 dt2 dx2 dy 2 dz 2
2
= c dt
(15.15)
(15.16)
where dt is the differential time elapsed on the moving clock. Thus,
r
v2
dt = dt 1 2 ,
c
(15.17)
and the time elapsed on the moving observers clock is
Zt2 r
v 2 (t)
t2 t1 = dt 1 2 .
c
(15.18)
t1
Thus, moving clocks run slower . This is an essential feature which is key to understanding
many important aspects of particle physics. A particle with a brief lifetime can, by moving
at speeds close to c, appear to an observer in our frame to be long-lived. It is customary to
define two dimensionless measures of a particles velocity:
v
c
As v c, we have 1 and .
1
1 2
(15.19)
Suppose we wish to compare the elapsed time on two clocks. We keep one clock at rest in
an inertial frame, while the other executes a closed path in space, returning to its initial
location after some interval of time. When the clocks are compared, the moving clock will
show a smaller elapsed time. This is often stated as the twin paradox. The total elapsed
time on a moving clock is given by
1
=
c
Zb
ds ,
(15.20)
where the integral is taken over the world line of the moving clock. The elapsed time
takes on a minimum
value when the path from a to b is a straight line. To see this, one can
= 0.
express x(t) as a functional of the path x(t) and extremize. This results in x
15.2.2
Irreverent problem from Spring 2002 final exam
Flowers for Algernon Bobs beloved hamster, Algernon, is very ill. He has only three hours
to live. The veterinarian tells Bob that Algernon can be saved only through a gallbadder
15.2. INTERVALS
295
transplant. A suitable donor gallbladder is available from a hamster recently pronounced
brain dead after a blender accident in New York (miraculously, the gallbladder was unscathed), but it will take Life Flight five hours to bring the precious rodent organ to San
Diego.
Bob embarks on a bold plan to save Algernons life. He places him in a cage, ties the cage to
the end of a strong meter-long rope, and whirls the cage above his head while the Life Flight
team is en route. Bob reasons that if he can make time pass more slowly for Algernon, the
gallbladder will arrive in time to save his life.
(a) At how many revolutions per second must Bob rotate the cage in order that the gallbladder arrive in time for the life-saving surgery? What is Algernons speed v0 ?
Solution : We have (t) = 0 R/c is constant, therefore, from eqn. 15.18,
t = t .
Setting t = 3 hr and t = 5 hr, we have = 53 , which entails =
v0 = 45 c, which requires a rotation frequency of 0 /2 = 38.2 MHz.
(15.21)
1 2 = 54 . Thus,
(b) Bob finds that he cannot keep up the pace! Assume Algernons speed is given by
r
t
v(t) = v0 1
(15.22)
T
where v0 is the speed from part (a), and T = 5 h. As the plane lands at the pet hospitals
emergency runway, Bob peers into the cage to discover that Algernon is dead! In order to
fill out his death report, the veterinarian needs to know: when did Algernon die? Assuming
he died after his own hamster watch registered three hours, derive an expression for the
elapsed time on the veterinarians clock at the moment of Algernons death.
1/2
Solution : hSnifflei. We have (t) = 45 1 Tt
. We set
ZT p
T = dt 1 2 (t)
(15.23)
where T = 3 hr and T is the time of death in Bobs frame. We write 0 =
p
0 = (1 02 )1/2 = 53 . Note that T /T = 1 02 = 01 .
4
5
and
Rescaling by writing = t/T , we have
T
= 01 =
T
TZ /T
1 02 + 02
"
#
3/2
T
1 02 + 02
(1 02 )3/2
T
"
#
3/2
1
T
2
1 + (02 1)
1 .
=
30 02 1
T
2
= 2
30
(15.24)
296
CHAPTER 15. SPECIAL RELATIVITY
Solving for T /T we have
T
T
With 0 =
5
3
3
2
02
1
2
2/3
02 1
(15.25)
we obtain
i
h
T
11 2/3
9
1
= 0.77502 . . .
= 16
3
T
Thus, T = 3.875 hr = 3 hr 52 min 50.5 sec after Bob starts swinging.
(15.26)
(c) Identify at least three practical problems with Bobs scheme.
Solution : As you can imagine, student responses to this part were varied and generally
sarcastic. E.g. the atmosphere would ignite, or Bobs arm would fall off, or Algernons
remains would be found on the inside of the far wall of the cage, squashed flatter than a
coat of semi-gloss paint, etc.
15.3
Four-Vectors and Lorentz Transformations
We have spoken thus far about different reference frames. So how precisely do the coordinates (t, x, y, z) transform between frames K and K ? In classical mechanics, we have t = t
and x = x + u t, according to fig. 15.3. This yields the Galilean transformation,
1 0 0 0
t
t
ux 1 0 0 x
=
.
(15.27)
y uy 0 1 0
y
z
z
uz 0 0 1
Such a transformation does not leave intervals invariant.
Let us define the four-vector x as
ct
x
ct
x =
.
y
x
z
(15.28)
Thus, x0 = ct, x1 = x, x2 = y, and x3 = z. In order for intervals to be invariant, the
transformation between x in frame K and x in frame K must be linear:
x = L x ,
(15.29)
where we are using the Einstein convention of summing over repeated indices. We define
the Minkowski metric tensor g as follows:
1 0
0
0
0 1 0
0
g = g =
(15.30)
0 0 1 0 .
0 0
0 1
15.3. FOUR-VECTORS AND LORENTZ TRANSFORMATIONS
297
Clearly g = gt is a symmetric matrix.
Note that the matrix L has one raised index and one lowered index. For the notation we
are about to develop, it is very important to distinguish raised from lowered indices. To
raise or lower an index, we use the metric tensor. For example,
ct
x
x = g x =
y .
z
(15.31)
The act of summing over an identical raised and lowered index is called index contraction.
Note that
1 0 0 0
0 1 0 0
g = g g = =
(15.32)
0 0 1 0 .
0 0 0 1
Now lets investigate the invariance of the interval. We must have x x = x x . Note
that
from which we conclude
x x = L x L x
= L g L x x ,
(15.33)
L g L = g .
(15.34)
This result also may be written in other ways:
L g L = g
Lt g L = g
(15.35)
Another way to write this equation is Lt g L = g. A rank-4 matrix which satisfies this
constraint, with g = diag(+, , , ) is an element of the group O(3, 1), known as the
Lorentz group.
Let us now count the freedoms in L. As a 4 4 real matrix, it contains 16 elements. The
matrix Lt g L is a symmetric 4 4 matrix, which contains 10 independent elements: 4 along
the diagonal and 6 above the diagonal. Thus, there are 10 constraints on 16 elements of L,
and we conclude that the group O(3, 1) is 6-dimensional. This is also the dimension of the
four-dimensional orthogonal group O(4), by the way. Three of these six parameters may
be taken to be the Euler angles. That is, the group O(3) constitutes a three-dimensional
subgroup of the Lorentz group O(3, 1), with elements
1 0
0
0
0 R
11 R12 R13
L =
(15.36)
,
0 R21 R22 R23
0 R31 R32 R33
298
CHAPTER 15. SPECIAL RELATIVITY
where Rt R = M I, i.e. R O(3) is a rank-3 orthogonal matrix, parameterized by the three
Euler angles (, , ). The remaining three parameters form a vector = (x , y , z ) and
define a second class of Lorentz transformations, called boosts:2
x
y
z
( 1) x y
( 1) x z
1 + ( 1) x x
(15.37)
L = x
,
y
( 1) x y
1 + ( 1) y y
( 1) y z
z
( 1) x z
( 1) y z
1 + ( 1) z z
where
b =
||
= 1 2
1/2
(15.38)
IMPORTANT : Since the components of are not the spatial components of a four
vector, we will only write these components with a lowered index, as i , with i = 1, 2, 3. We
will not write i with a raised index, but if we did, wed mean the same thing, i.e. i = i .
Note that for the spatial components of a 4-vector like x , we have xi = xi .
Lets look at a simple example, where x = and y = z = 0. Then
=
0
0
0
0
0
0
.
0
1
0
0
1
0
(15.39)
The effect of this Lorentz transformation x = L x is thus
ct = ct + x
x = ct + x .
(15.40)
(15.41)
How fast is the origin of K moving in the K frame? We have dx = 0 and thus
c dt
1 dx
=
= .
c dt
c dt
(15.42)
Thus, u = c, i.e. = u/c.
It is convenient to take advantage of the fact that Pij i j is a projection operator , which
2
satisfies P = P . The action of Pij on any vector is to project that vector onto the
direction:
P = ( ) .
(15.43)
We may now write the general Lorentz boost, with = u/c, as
t
L=
,
I + ( 1) P
2
Unlike rotations, the boosts do not themselves define a subgroup of O(3, 1).
(15.44)
15.3. FOUR-VECTORS AND LORENTZ TRANSFORMATIONS
299
where I is the 3 3 unit matrix, and where we write column and row vectors
x
= y
,
t = x y z
z
(15.45)
as a mnemonic to help with matrix multiplications. We now have
ct
t
ct
ct + x
=
=
.
I + ( 1) P
x
ct + x + ( 1) P x
x
(15.46)
Thus,
ct = ct + x
(15.47)
) .
x = ct + x + ( 1) (x
(15.48)
If we resolve x and x into components parallel and perpendicular to , writing
xk = x
x = x (x)
,
(15.49)
with corresponding definitions for xk and x , the general Lorentz boost may be written as
ct = ct + xk
(15.50)
xk = ct + xk
(15.51)
x = x .
(15.52)
Thus, the components of x and x which are parallel to enter into a one-dimensional
Lorentz boost along with t and t , as described by eqn. 15.41. The components of x and
x which are perpendicular to are unaffected by the boost.
Finally, the Lorentz group O(3, 1) is a group under multiplication, which means that if La
and Lb are elements, then so is the product La Lb . Explicitly, we have
(La Lb )t g La Lb = Ltb (Lta g La ) Lb = Ltb g Lb = g .
15.3.1
(15.53)
Covariance and contravariance
Note that
Lt g L
0
0
=
0
0 1
0
0 0
1 0
0
0 1 0
=
0 0 1
0 0
0
0
1 0
0 1
0
0 0 0
1
0 0
0
0
=g ,
0
1
0
0
0
0
1 0 0
0
0 1
0
0
0
0
1
0
0
0
0
1
(15.54)
300
CHAPTER 15. SPECIAL RELATIVITY
since 2 (1 2 ) = 1. This is in fact the general way that tensors transform under a Lorentz
transformation:
covariant vectors : x = L x
covariant tensors : F
L L
(15.55)
F
Lt
(15.56)
Note how index contractions always involve one raised index and one lowered index. Raised
indices are called contravariant indices and lowered indiced are called covariant indices.
The transformation rules for contravariant vectors and tensors are
contravariant vectors : x = L x
(15.57)
contravariant tensors : F = L L F = L F Lt
(15.58)
A Lorentz scalar has no indices at all. For example,
ds2 = g dx dx ,
(15.59)
is a Lorentz scalar. In this case, we have contracted a tensor with two four-vectors. The
dot product of two four-vectors is also a Lorentz scalar:
a b a b = g a b
= a0 b0 a1 b1 a2 b2 a3 b3
= a0 b0 a b .
(15.60)
Note that the dot product a b of four-vectors is invariant under a simultaneous Lorentz
transformation of both a and b , i.e. a b = a b . Indeed, this invariance is the very
definition of what it means for something to be a Lorentz scalar. Derivatives with respect
to covariant vectors yield contravariant vectors:
f
f
x
A
= A B
x
B
= B C
x
et cetera. Note that differentiation with respect to the covariant vector x is expressed by
the contravariant differential operator :
=
,
,
,
(15.61)
x
c t x y z
,
, ,
.
(15.62)
=
c t
x
y
z
x
The contraction
is a Lorentz scalar differential operator, called the DAlembertian:
=
2
2
2
1 2
.
c2 t2 x2 y 2 z 2
(15.63)
The Helmholtz equation for scalar waves propagating with speed c can thus be written in
compact form as = 0.
15.3. FOUR-VECTORS AND LORENTZ TRANSFORMATIONS
15.3.2
301
What to do if you hate raised and lowered indices
Admittedly, this covariant and contravariant business takes some getting used to. Ultimately, it helps to keep straight which indices transform according to L (covariantly) and
which transform according to Lt (contravariantly). If you find all this irksome, the raising
and lowering can be safely ignored. We define the position four-vector as before, but with no
difference between raised and lowered indices. In fact, we can just represent all vectors and
tensors with lowered indices exclusively, writing e.g. x = (ct, x, y, z). The metric tensor is
g = diag(+, , , ) as before. The dot product of two four-vectors is
x y = g x y .
(15.64)
x = L x .
(15.65)
The Lorentz transformation is
Since this preserves intervals, we must have
which entails
g x y = g L x L y
= Lt g L x y ,
(15.66)
Lt g L = g .
(15.67)
In terms of the quantity L defined above, we have L = L . In this convention, we could
completely avoid raised indices, or we could simply make no distinction, taking x = x
and L = L = L , etc.
15.3.3
Comparing frames
Suppose in the K frame we have a measuring rod which is at rest. What is its length as
with respect to K. From
measured in the K frame? Recall K moves with velocity u = u x
the Lorentz transformation in eqn. 15.41, we have
x1 = (x1 + c t1 )
(15.68)
x2 = (x2 + c t2 ) ,
(15.69)
where x1,2 are the positions of the ends of the rod in frame K. The rods length in any
frame is the instantaneous spatial separation of its ends. Thus, we set t1 = t2 and compute
the separation x = x2 x1 :
x = x
x = 1 x = 1 2
1/2
x .
(15.70)
The proper length 0 of a rod is its instantaneous end-to-end separation in its rest frame.
We see that
1/2
() = 1 2
0 ,
(15.71)
302
CHAPTER 15. SPECIAL RELATIVITY
so the length is always greatest in the rest frame. This is an example of a Lorentz-Fitzgerald
contraction. Note that the transverse dimensions do not contract:
y = y
z = z .
(15.72)
Thus, the volume contraction of a bulk object is given by its length contraction: V = 1 V.
A striking example of relativistic issues of length, time, and simultaneity is the famous
pole and the barn paradox, described in the Appendix (section ). Here we illustrate some
essential features via two examples.
15.3.4
Example I
Next, lets analyze the situation depicted in fig. 15.3. In the K frame, well denote the
following spacetime points:
ct
ct
ct
ct
, S =
, S =
.
(15.73)
, B =
A =
ct
+ct
+d
d
Note that the origin in K is given by O = (ct , 0). Here we are setting y = y = z = z = 0
denote the left-moving (S )
and dealing only with one spatial dimension. The points S
and right-moving (S+ ) wavefronts. We see that the arrival of the signal S1 at A requires
S1 = A , hence ct = d. The same result holds when we set S2 = B for the arrival of the
right-moving wavefront at B .
We now use the Lorentz transformation
L
to transform to the K frame. Thus,
1
ctA
d
1
= LA =
A=
= (1 )d
xA
1
d
1
1
ctB
d
1
= LB =
B=
= (1 + )d
.
xB
1
+d
1
(15.74)
(15.75)
(15.76)
Thus, tA = (1 )d/c and tB = (1 + )d/c. Thus, the two events are not simultaneous
in K. The arrival at A is first.
15.3.5
Example II
at rest in frame K.
Consider a rod of length 0 extending from the origin to the point 0 x
In the frame K, the two ends of the rod are located at spacetime coordinates
ct
ct
A=
and B =
,
(15.77)
0
0
15.3. FOUR-VECTORS AND LORENTZ TRANSFORMATIONS
303
Figure 15.5: A rectangular plate moving at velocity V = V x.
respectively. Now consider the origin in frame K . Its spacetime coordinates are
ct
.
C =
0
To an observer in the K frame, we have
ct
ct
C=
=
.
ct
0
(15.78)
(15.79)
Now consider two events. The first event is the coincidence of A with C, i.e. the origin of
K instantaneously coincides with the origin of K. Setting A = C we obtain t = t = 0.
The second event is the coincidence of B with C. Setting B = C we obtain t = l0 /c and
t = 0 /c. Note that t = ()/c, i.e. due to the Lorentz-Fitzgerald contraction of the
rod as seen in the K frame, where () = 0 /.
15.3.6
Deformation of a rectangular plate
Problem: A rectangular plate of dimensions a b moves at relativistic velocity V = V x
as shown in fig. 15.5. In the rest frame of the rectangle, the a side makes an angle with
axis. Describe in detail and sketch the shape of the plate as measured by
respect to the x
an observer in the laboratory frame. Indicate the lengths of all sides and
q the values of all
interior angles. Evaluate your expressions for the case = 41 and V =
2
3
c.
to be
Solution: An observer in the laboratory frame will measure lengths parallel to x
Lorentz contracted by a factor 1 , where = (1 2 )1/2 and = V /c. Lengths perpen remain unaffected. Thus, we have the situation depicted in fig. 15.6. Simple
dicular to x
trigonometry then says
tan = tan
tan = 1 tan ,
304
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.6: Relativistic deformation of the rectangular plate.
as well as
q
p
a = a 2 cos2 + sin2 = a 1 2 cos2
q
q
b = b 2 sin2 + cos2 = b 1 2 sin2 .
The plate deforms to a parallelogram, with internal angles
= 21 + tan1 ( tan ) tan1 ( 1 tan )
= 12 tan1 ( tan ) + tan1 ( 1 tan ) .
Note that the area of the plate as measured in the laboratory frame is
= a b sin = a b cos( )
= 1 ,
where = ab is the proper area. The area contraction factor is 1 and not 2 (or 3
in a three-dimensional system) because only the parallel dimension gets contracted.
q
Setting V = 23 c gives = 3, and with = 41 we have = 13 and = 61 . The interior
q
q
= 13 . The side lengths are a = 23 a and b = 23 b.
angles are then = 23 and
15.3. FOUR-VECTORS AND LORENTZ TRANSFORMATIONS
15.3.7
305
Transformation of velocities
Let K move at velocity u = c relative to K. The transformation from K to K is given
by the Lorentz boost,
x
y
z
( 1) x y
( 1) x z
1 + ( 1) x x
(15.80)
L = x
.
y
( 1) x y
1 + ( 1) y y
( 1) y z
z
( 1) x z
( 1) y z
1 + ( 1) z z
Applying this, we have
dx = L dx .
(15.81)
This yields
dx0 = dx 0 + dx
dx = dx
(15.82)
+ dx + ( 1) dx
.
(15.83)
We then have
V =c
dx
c dx 0 + c dx + c ( 1) dx
=
dx0
dx 0 + dx
uV
u + 1 V + (1 1 ) u
.
1 + uV /c2
(15.84)
The second line is obtained by dividing both numerator and denominator by dx 0 , and then
writing V = dx /dx 0 . There are two special limiting cases:
(u + V ) u
1 + u V /c2
V = 1)
velocities parallel u
V =
V = 0)
velocities perpendicular u
V = u + 1 V .
(15.85)
(15.86)
Note that if either u or V is equal to c, the resultant expression has |V | = c as well. One
cant boost the speed of light!
Lets revisit briefly the example in section 15.3.4. For an observer, in the K frame, the
relative velocity of S and A is c + u, because even though we must boost the velocity
of the left-moving light wave by u x,
the result is still c x,
according to our velocity
c x
addition formula. The distance between the emission and detection points is d() = d/.
Thus,
d
1
d 1
d
d()
=
=
(15.87)
= (1 ) .
tA =
c+u
c+u
c 1 2
c
This result is exactly as found in section 15.3.4 by other means. A corresponding analysis
yields tB = (1 + ) d/c. again in agreement with the earlier result. Here, it is crucial to
account for the Lorentz contraction of the distance between the source S and the observers
A and B as measured in the K frame.
306
15.3.8
CHAPTER 15. SPECIAL RELATIVITY
Four-velocity and four-acceleration
In nonrelativistic mechanics, the velocity V = dx
dt is locally tangent to a particles trajectory.
In relativistic mechanics, one defines the four-velocity,
dx
dx
=p
=
,
(15.88)
u
ds
1 2 c dt
which is locally tangent to the world line of a particle. Note that
g u u = 1 .
(15.89)
The four-acceleration is defined as
w
du
d2x
=
.
ds
ds2
(15.90)
Note that u w = 0, so the 4-velocity and 4-acceleration are orthogonal with respect to the
Minkowski metric.
15.4
Three Kinds of Relativistic Rockets
15.4.1
Constant acceleration model
Clearly the rocket has
Consider a rocket which undergoes constant acceleration along x.
no rest frame per se, because its velocity is changing. However, this poses no serious
obstacle to discussing its relativistic motion. We consider a frame K in which the rocket
is instantaneously at rest. In such a frame, the rockets 4-acceleration is w = (0, a/c2 ),
where we suppress the transverse coordinates y and z. In an inertial frame K, we have
=
.
(15.91)
w =
ds
c +
Transforming w into the K frame, we have
0
a/c2
w =
.
=
a/c2
a/c2
(15.92)
Taking the upper component, we obtain the equation
a
=
c
d
dt
the solution of which, with (0) = 0, is
at
(t) =
2
c + a2 t2
(t) =
1+
a
,
c
at
c
2
(15.93)
(15.94)
15.4. THREE KINDS OF RELATIVISTIC ROCKETS
307
The proper time for an observer moving with the rocket is thus
=
Zt
0
c dt1
c2 + a2 t21
at
c
.
sinh1
a
c
For large times t c/a, the proper time grows logarithmically in t, which is parametrically
slower. To find the position of the rocket, we integrate x = c, and obtain, with x(0) = 0,
x(t) =
Zt
0
a ct dt
c p 2
2 t2 c .
p 1 1 =
c
+
a
a
c2 + a2 t21
(15.95)
It is interesting to consider the situation in the frame K . We then have
( ) = tanh(a /c)
( ) = cosh(a /c) .
(15.96)
For an observer in the frame K , the distance he has traveled
is x ( ) = x( )/( ), as
2
we found in eqn. 15.70. Now x( ) = (c /a) cosh(a /c) 1 , hence
x ( ) =
c2
1 sech(a /c) .
a
(15.97)
For c/a, we expand sech(a /c) 1 12 (a /c)2 and find x ( ) = 21 a 2 , which clearly is
the nonrelativistic limit. For , however, we have x ( ) c2 /a is finite! Thus, while
the entire Universe is falling behind the accelerating observer, it all piles up at a horizon a
distance c2 /a behind it, in the frame of the observer. The light from these receding objects
is increasingly red-shifted (see section 15.6 below), until it is no longer visible. Thus, as
John Baez describes it, the horizon is a dark plane that appears to be swallowing the
entire Universe! In the frame of the inertial observer, however, nothing strange appears to
be happening at all!
15.4.2
Constant force with decreasing mass
Suppose instead the rocket is subjected to a constant force F0 in its instantaneous rest
frame, and furthermore that the rockets mass satisfies m( ) = m0 (1 ), where is the
proper time for an observer moving with the rocket. Then from eqn. 15.93, we have
F0
m0 (1 )
d()
d()
= 1
dt
d
=
1
d
d
=
1 2 d
d
1
2
ln
1+
1
after using the chain rule, and with d /dt = 1 . Integrating, we find
1+
1 (1 )r
2F0
ln 1
= ( ) =
ln
=
1
1 + (1 )r
m0 c
(15.98)
(15.99)
308
CHAPTER 15. SPECIAL RELATIVITY
with r = 2F0 /m0 c. As 1 , the rocket loses all its mass, and it asymptotically
approaches the speed of light.
It is convenient to write
"
r
( ) = tanh
ln
2
1
1
#
(15.100)
in which case
"
#
1
r
dt
= cosh
ln
=
d
2
1
"
#
1 dx
1
r
= sinh
ln
.
c d
2
1
(15.101)
(15.102)
Integrating the first of these from = 0 to = 1 , we find t t = 1 is
h
2
Z1
1
d r/2 + r/2 =
t =
i
F0 2 1
mc
if >
F0
mc
if
F0
mc
(15.103)
.
Since ( = 1 ) = 1, this is the time in the K frame when the rocket reaches the speed of
light.
15.4.3
Constant ejecta velocity
Our third relativistic rocket model is a generalization of what is commonly known as the
rocket equation in classical physics. The model is one of a rocket which is continually
ejecting burnt fuel at a velocity u in the instantaneous rest frame of the rocket. The
nonrelativistic rocket equation follows from overall momentum conservation:
dprocket + dpfuel = d(mv) + (v u) (dm) = 0 ,
(15.104)
since if dm < 0 is the differential change in rocket mass, the differential ejecta mass is dm.
This immediately gives
m dv + u dm = 0
v = u ln
m0
m
(15.105)
where the rocket is assumed to begin at rest, and where m0 is the initial mass of the rocket.
Note that as m 0 the rockets speed increases without bound, which of course violates
special relativity.
In relativistic mechanics, as we shall see in section 15.5, the rockets momentum, as described
by an inertial observer, is p = mv, and its energy is mc2 . We now write two equations
15.5. RELATIVISTIC MECHANICS
309
for overall conservation of momentum and energy:
d(mv) + e ve dme = 0
(15.106)
d(mc2 ) + e (dme c2 ) = 0 ,
(15.107)
where ve is the velocity of the ejecta in the inertial frame, dme is the differential mass of
2 1/2
the ejecta, and e = 1 vce2
. From the second of these equations, we have
e dme = d(m) ,
(15.108)
which we can plug into the first equation to obtain
(v ve ) d(m) + m dv = 0 .
(15.109)
Before solving this, we remark that eqn. 15.108 implies that dme < |dm| the differential
mass of the ejecta is less than the mass lost by the rocket! This is Einsteins famous equation
E = mc2 at work more on this later.
To proceed, we need to use the parallel velocity addition formula of eqn. 15.85 to find ve :
2
u 1 vc2
vu
.
ve =
(15.110)
=
v ve =
1 uv
1 uv
c2
c2
We now define u = u/c, in which case eqn, 15.109 becomes
u (1 2 ) d(m) + (1 u ) m d = 0 .
(15.111)
Using d = 3 d, we observe a felicitous cancellation of terms, leaving
u
dm
d
+
=0.
m
1 2
Integrating, we obtain
m
= tanh u ln 0
m
(15.112)
(15.113)
Note that this agrees with the result of eqn. 15.100, if we take u = F0 /mc.
15.5
Relativistic Mechanics
Relativistic particle dynamics follows from an appropriately extended version of Hamiltons
principle S = 0. The action S must be a Lorentz scalar. The action for a free particle is
Ztb r
Zb
v2
2
S x(t) = mc ds = mc dt 1 2 .
c
ta
(15.114)
310
CHAPTER 15. SPECIAL RELATIVITY
Thus, the free particle Lagrangian is
r
2 2
v2
2
2
2 v
2
1
1
L = mc 1 2 = mc + 2 mv + 8 mc
+ ... .
c
c2
(15.115)
Thus, L can be written as an expansion in powers of v 2 /c2 . Note that L(v = 0) = mc2 .
We interpret this as U0 , where U0 = mc2 is the rest energy of the particle. As a constant,
it has no consequence for the equations of motion. The next term in L is the familiar
nonrelativistic kinetic energy, 21 mv 2 . Higher order terms are smaller by increasing factors
of 2 = v 2 /c2 .
We can add a potential U (x, t) to obtain
2
t) = mc
L(x, x,
x 2
U (x, t) .
c2
(15.116)
The momentum of the particle is
p=
L
= mx .
x
(15.117)
The force is F = U as usual, and Newtons Second Law still reads p = F . Note that
v v 2
p = m v + 2 v .
(15.118)
c
The
Thus, the force F is not necessarily in the direction of the acceleration a = v.
Hamiltonian, recall, is a function of coordinates and momenta, and is given by
p
(15.119)
H = p x L = m2 c4 + p2 c2 + U (x, t) .
Since L/t = 0 for our case, H is conserved by the motion of the particle. There are two
limits of note:
p2
+ U + O(p4 /m4 c4 )
2m
: H = c|p| + U + O(mc/p) .
: H = mc2 +
|p| mc (non-relativistic)
|p| mc (ultra-relativistic)
(15.120)
(15.121)
Expressed in terms of the coordinates and velocities, we have H = E, the total energy, with
E = mc2 + U .
(15.122)
In particle physics applications, one often defines the kinetic energy T as
T = E U mc2 = ( 1)mc2 .
(15.123)
When electromagnetic fields are included,
r
x 2
q
q + A x
2
c
c
q
dx
2
= mc A
,
c
dt
t) = mc2
L(x, x,
(15.124)
15.5. RELATIVISTIC MECHANICS
311
where the electromagnetic 4-potential is A = ( , A). Recall A = g A has the sign of
its spatial components reversed. One the has
p=
and the Hamiltonian is
H=
15.5.1
q
L
= mx + A ,
x
c
(15.125)
q 2
m2 c4 + p A + q .
c
(15.126)
Relativistic harmonic oscillator
From E = mc2 + U , we have
"
x 2 = c2 1
mc2
E U (x)
2 #
Consider the one-dimensional harmonic oscillator potential U (x) =
turning points as x = b, satisfying
E mc2 = U (b) = 21 kb2 .
(15.127)
1
2
2 kx .
We define the
(15.128)
Now define the angle via x b cos , and further define the dimensionless parameter
= kb2 /4mc2 . Then, after some manipulations, one obtains
p
1 + sin2
,
(15.129)
= 0
1 + 2 sin2
p
with 0 = k/m as in the nonrelativistic case. Hence, the problem is reduced to quadratures (a quaint way of saying doing an an integral):
t() t0 =
Z
1 + 2 sin2
.
d p
1 + sin2
01
(15.130)
While the result can be expressed in terms of elliptic integrals, such an expression is not
particularly illuminating. Here we will content ourselves with computing the period T ():
Z2
1 + 2 sin2
4
T () =
d p
0
1 + sin2
(15.131)
Z2
4
=
d 1 + 32 sin2 85 2 sin4 + . . .
0
0
o
2 n
15 2
+ ... .
1 + 34 64
=
0
(15.132)
Thus, for the relativistic harmonic oscillator, the period does depend on the amplitude,
unlike the nonrelativistic case.
312
CHAPTER 15. SPECIAL RELATIVITY
15.5.2
Energy-momentum 4-vector
Lets focus on the case where U (x) = 0. This is in fact a realistic assumption for subatomic
particles, which propagate freely between collision events.
The differential proper time for a particle is given by
d =
ds
= 1 dt ,
c
(15.133)
where x = (ct, x) are coordinates for the particle in an inertial frame. Thus,
dx
d
p = mx = m
E
dx0
= mc = m
,
c
d
with x0 = ct. Thus, we can write the energy-momentum 4-vector as
E/c
dx
px
p = m
=
py .
d
pz
(15.134)
(15.135)
Note that p = mcu , where u is the 4-velocity of eqn. 15.88. The four-momentum satisfies
the relation
E2
(15.136)
p p = 2 p2 = m2 c2 .
c
The relativistic generalization of force is
f =
where F = dp/dt as usual.
dp
= F v/c , F ,
d
(15.137)
The energy-momentum four-vector transforms covariantly under a Lorentz transformation.
This means
p = L p .
(15.138)
relative to frame K, then
If frame K moves with velocity u = c x
E
c1 E + px
= p
c
1 2
px =
px + c1 E
p
1 2
py = py
pz = pz .
(15.139)
In general, from eqns. 15.50, 15.51, and 15.52, we have
E
E
=
+ pk
c
c
(15.140)
E
+ pk
c
(15.141)
pk =
p = p
and p = p (p)
where pk = p
.
(15.142)
15.6. RELATIVISTIC DOPPLER EFFECT
15.5.3
313
4-momentum for massless particles
For a massless particle, such as a photon, we have p p = 0, which means E 2 = p2 c2 . The
4-momentum may then be written p = |p| , p . We define the 4-wavevector k by the
relation p = ~k , where ~ = h/2 and h is Plancks constant. We also write = ck, with
E = ~.
15.6
Relativistic Doppler Effect
The 4-wavevector k = /c , k for electromagnetic radiation satisfies k k = 0. The
energy-momentum 4-vector is p = ~k . The phase (x ) = k x = k x t of a plane
wave is a Lorentz scalar. This means that the total number of wave crests (i.e. = 2n)
emitted by a source will be the total number observed by a detector.
Suppose a moving source emits radiation of angular frequency in its rest frame. Then
k = L () k
x
y
z
/c
x
( 1) x y
( 1) x z
x 1 + ( 1) x x
k .
=
y
( 1) y z ky
1 + ( 1) y y
( 1) x y
kz
z
( 1) x z
( 1) y z
1 + ( 1) z z
(15.143)
This gives
= k = (1 cos ) ,
(15.144)
c
c
c
is the angle measured in K between and k.
Solving for , we have
where = cos1 ( k)
p
1 2
=
,
(15.145)
1 cos 0
where 0 = is the angular frequency in the rest frame of the moving source. Thus,
s
1+
(15.146)
= 0 source approaching
=
1 0
= 12
=
source perpendicular
source receding
1 2 0
(15.147)
1
.
1+ 0
(15.148)
Recall the non-relativistic Doppler effect:
=
0
.
1 (V /c) cos
(15.149)
314
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.7: Alices big adventure.
We see that approaching sources have their frequencies shifted higher; this is called the
blue shift, since blue light is on the high frequency (short wavelength) end of the optical
spectrum. By the same token, receding sources are red-shifted to lower frequencies.
15.6.1
Romantic example
Alice and Bob have a May-December thang going on. Bob is May and Alice December,
if you get my drift. The social stigma is too much to bear! To rectify this, they decide
that Alice should take a ride in a space ship. Alices itinerary takes her along a sector of
a circle of radius R and angular span of = 1 radian, as depicted in fig. 15.7. Define
O (r = 0), P (r = R, = 12 ), and Q (r = R, = 12 ). Alices speed along the
first leg (straight from O to P) is va = 35 c. Her speed along the second leg (an arc from
12
P to Q) is vb = 13
c. The final leg (straight from Q to O) she travels at speed vc = 45 c.
Remember that the length of an circular arc of radius R and angular spread (radians) is
= R.
(a) Alice and Bob synchronize watches at the moment of Alices departure. What is the
elapsed time on Bobs watch when Alice returns? What is the elapsed time on Alices
watch? What must R be in order for them to erase their initial 30 year age difference?
Solution : In Bobs frame, Alices trip takes a time
R
R
R
+
+
ca
cb
cc
R 5 13 5 4R
=
+ +
=
.
c 3 12 4
c
The elapsed time on Alices watch is
t =
R
R
R
+
+
ca a cb b cc c
5R
R 5 4 13 5
5 3
.
=
=
c 3 5 12 13 4 5
2c
(15.150)
t =
(15.151)
15.6. RELATIVISTIC DOPPLER EFFECT
315
Thus, T = t t = 3R/2c and setting T = 30 yr, we find R = 20 ly. So Bob will
have aged 80 years and Alice 50 years upon her return. (Maybe this isnt such a good plan
after all.)
(b) As a signal of her undying love for Bob, Alice continually shines a beacon throughout
her trip. The beacon produces monochromatic light at wavelength 0 = 6000
A (frequency
14
f0 = c/0 = 5 10 Hz). Every night, Bob peers into the sky (with a radiotelescope),
hopefully looking for Alices signal. What frequencies fa , fb , and fc does Bob see?
Solution : Using the relativistic Doppler formula, we have
fa =
fb =
fc =
1 a
f0 = 12 f0
1 + a
1 b2 f0 =
5
13 f0
1 + c
f0 = 3f0 .
1 c
(15.152)
(c) Show that the total number of wave crests counted by Bob is the same as the number
emitted by Alice, over the entire trip.
Solution : Consider first the OP leg of Alices trip. The proper time elapsed on Alices
watch during this leg is ta = R/ca a , hence she emits Na = Rf0 /ca a wavefronts during
this leg. Similar considerations hold for the PQ and QO legs, so Nb = Rf0 /cb b and
Nc = Rf0 /cc c .
Although the duration of the OP segment of Alices trip takes a time ta = R/ca in Bobs
frame, he keeps receiving the signal at the Doppler-shifted frequency fa until the wavefront
emitted when Alice arrives at P makes its way back to Bob. That takes an extra time R/c,
hence the number of crests emitted for Alices OP leg is
Na =
R
R
+
ca
c
s
1 a
Rf0
= Na ,
f0 =
1 + a
ca a
(15.153)
since the source is receding from the observer.
During the PQ leg, we have = 12 , and Alices velocity is orthogonal to the wavevector k,
which is directed radially inward. Bobs first signal at frequency fb arrives a time R/c after
Alice passes P, and his last signal at this frequency arrives a time R/c after Alice passes Q.
Thus, the total time during which Bob receives the signal at the Doppler-shifted frequency
fb is tb = R/c, and
R
Nb =
cb
1 b2 f0 =
Rf0
cb b
= Nb .
(15.154)
316
CHAPTER 15. SPECIAL RELATIVITY
Finally, during the QO home stretch, Bob first starts to receive the signal at the Dopplershifted frequency fc a time R/c after Alice passes Q, and he continues to receive the signal
until the moment Alice rushes into his open and very flabby old arms when she makes it
back to O. Thus, Bob receives the frequency fc signal for a duration tc R/c, where
tc = R/cc . Thus,
s
1 + c
R
R
Rf0
= Nc ,
(15.155)
Nc =
f0 =
cc
c
1 c
cc c
since the source is approaching.
Therefore, the number of wavelengths emitted by Alice will be precisely equal to the number
received by Bob none of the waves gets lost.
15.7
Relativistic Kinematics of Particle Collisions
As should be expected, special relativity is essential toward the understanding of subatomic
particle collisions, where the particles themselves are moving at close to the speed of light. In
our analysis of the kinematics of collisions, we shall find it convenient to adopt the standard
convention on units, where we set c 1. Energies will typically be given in GeV, where
1 GeV = 109 eV = 1.602 1010 RJ . Momenta will then be in units of GeV/c, and masses
in units of GeV/c2 . With c 1, it is then customary to quote masses in energy units. For
example, the mass of the proton in these units is mp = 938 MeV, and m = 140 MeV.
For a particle of mass M , its 4-momentum satisfies P P = M 2 (remember c = 1). Consider
now an observer with 4-velocity U . The energy of the particle, in the rest frame of the
observer is E = P U . For example, if P = (M, 0, 0, 0) is its rest frame, and U = ( , ),
then E = M , as we have already seen.
Consider next the emission of a photon of 4-momentum P = (~/c, ~k) from an object
with 4-velocity V , and detected in a frame with 4-velocity U . In the frame of the detector,
the photon energy is E = P U , while in the frame of the emitter its energy is E = P V .
If U = (1, 0, 0, 0) and V = ( , ), then E = ~ and E = ~ = ~( k) =
. Thus, = 1 /(1 cos ). This recapitulates
~(1 cos ), where = cos1 k
our earlier derivation in eqn. 15.144.
Consider next the interaction of several particles. If in a given frame the 4-momenta of the
reactants are Pi , where n labels the reactant species, and the 4-momenta of the products
are Qj , then if the collision is elastic, we have that total 4-momentum is conserved, i.e.
N
X
Pi
i=1
N
X
Qj ,
(15.156)
j=1
products. For massive particles, we can write
where there are N reactants and N
Pi = i mi 1 , vi ) ,
Qj = j m
j 1 , vj ) ,
(15.157)
15.7. RELATIVISTIC KINEMATICS OF PARTICLE COLLISIONS
317
Figure 15.8: Spontaneous decay of a single reactant into two products.
while for massless particles,
Pi = ~ki 1 , k
15.7.1
.
Qj = ~kj 1 , k
(15.158)
Spontaneous particle decay into two products
Consider first the decay of a particle of mass M into two particles. We have P = Q1 + Q2 ,
hence in the rest frame of the (sole) reactant, which is also called the center of mass
(CM) frame since the total 3-momentum vanishes therein, we have M = E1 + E2 . Since
EiCM = CM mi , and i 1, clearly we must have M > m1 + m2 , or else the decay cannot
possibly conserve energy. To analyze further, write P Q1 = Q2 . Squaring, we obtain
M 2 + m21 2P Q1 = m22 .
(15.159)
The dot-product P Q1 is a Lorentz scalar, and hence may be evaluated in any frame.
Let us first consider the CM frame, where P = M (1, 0, 0, 0), and P Q1 = M E1CM , where
E1CM is the energy of n = 1 product in the rest frame of the reactant. Thus,
E1CM =
M 2 + m21 m22
2M
E2CM =
M 2 + m22 m21
,
2M
(15.160)
where the second result follows merely from switching the product labels. We may now
write Q1 = (E1CM , pCM ) and Q2 = (E2CM , pCM ), with
(pCM )2 = (E1CM )2 m21 = (E2CM )2 m22
2
M m21 m22 2
m1 m2 2
=
.
2M
M
(15.161)
In the laboratory frame, we have P = M (1 , V ) and Qi = i mi (1 , Vi ). Energy and
momentum conservation then provide four equations for the six unknowns V1 and V2 . Thus,
there is a two-parameter family of solutions, assuming we regard the reactant velocity V K as
318
CHAPTER 15. SPECIAL RELATIVITY
fixed, corresponding to the freedom to choose pCM in the CM frame solution above. Clearly
the three vectors V , V1 , and V2 must lie in the same plane, and with V fixed, only one
additional parameter is required to fix this
plane. The other free parameter may be taken
to be the relative angle 1 = cos1 V V1 (see fig. 15.8). The angle 2 as well as the speed
V2 are then completely determined. We can use eqn. 15.159 to relate 1 and V1 :
M 2 + m21 m22 = 2M m1 1 1 V V1 cos 1 .
(15.162)
It is convenient to express both 1 and V1 in terms of the energy E1 :
s
q
m2
E1
2
,
V1 = 1 1 = 1 21 .
1 =
E1
m1
(15.163)
This results in a quadratic equation for E1 , which may be expressed as
p
(1 V 2 cos2 1 )E12 2 1 V 2 E1CM E1 + (1 V 2 )(E1CM )2 + m21 V 2 cos2 1 = 0 , (15.164)
the solutions of which are
q
CM
2
1 V E1 V cos 1 (1 V 2 )(E1CM )2 (1 V 2 cos2 1 )m21
E1 =
.
1 V 2 cos2 1
The discriminant is positive provided
CM 2
E1
m1
which means
sin2 1 <
where
1 V 2 cos2 1
,
1V2
V 2 1
sin2 1 ,
(V1CM )2 1
CM
V1
>
m1
E1CM
2
(15.165)
(15.166)
(15.167)
(15.168)
is the speed of product 1 in the CM frame. Thus, for V < V1CM < 1, the scattering angle 1
may take on any value, while for larger reactant speeds V1CM < V < 1 the quantity sin2 1
cannot exceed a critical value.
15.7.2
Miscellaneous examples of particle decays
Let us now consider some applications of the formulae in eqn. 15.160:
Consider the decay 0 , for which m1 = m2 = 0. We then have E1CM = E2CM =
1
CM
= E2CM = 67.5 MeV for the
2 M . Thus, with M = m 0 = 135 MeV, we have E1
photon energies in the CM frame.
15.7. RELATIVISTIC KINEMATICS OF PARTICLE COLLISIONS
319
For the reaction K + + + we have M = mK + = 494 MeV and m1 = m =
106 MeV. The neutrino mass is m2 0, hence E2CM = 236 MeV is the emitted neutrinos energy in the CM frame.
A 0 hyperon with a mass M = m0 = 1116 MeV decays into a proton (m1 = mp =
938 MeV) and a pion m2 = m = 140 MeV). The CM energy of the emitted proton
is E1CM = 943 MeV and that of the emitted pion is E2CM = 173 MeV.
15.7.3
Threshold particle production with a stationary target
incident upon a
Consider now a particle of mass M1 moving with velocity V1 = V1 x,
stationary target particle of mass M2 , as indicated in fig. 15.9. Let the product masses be
m1 , m2 , . . . , mN . The 4-momenta of the reactants and products are
P1 = E1 , P1
P2 = M2 1 , 0
Qj = j , pj .
(15.169)
Note that E12 P12 = M12 and 2j p2j = m2j , with j {1, 2, . . . , N }.
Conservation of momentum means that
P1
P2
N
X
Qj .
(15.170)
j=1
In particular, taking the = 0 component, we have
N
X
j ,
(15.171)
mj M2
(15.172)
E1 + M2 =
j=1
which certainly entails
E1
N
X
j=1
since j = j mj mj . But can the equality ever be achieved? This would only be the case
if j = 1 for all j, i.e. the final velocities are all zero. But this itself is quite impossible,
since the initial state momentum is P .
To determine the threshold energy E1thr , we compare the length of the total momentum
vector in the LAB and CM frames:
(P1 + P2 )2 = M12 + M22 + 2E1 M2
!2
N
X
CM
j
=
j=1
(LAB)
(15.173)
(CM) .
(15.174)
320
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.9: A two-particle initial state, with a stationary target in the LAB frame, and an
N -particle final state.
Thus,
E1 =
P
and we conclude
THR
E1 E1
N
CM
j=1 j
2
M12 M22
(15.175)
2M2
P
N
j=1 mj
2
M12 M22
2M2
(15.176)
Note that in the CM frame it is possible for each CM
j = mj .
Finally, we must have E1THR
PN
j=1 mj
M2 . This then requires
M1 + M2
15.7.4
N
X
mj .
(15.177)
j=1
Transformation between frames
Consider a particle with 4-velocity u in frame K and consider a Lorentz transformation
between this frame and a frame K moving relative to K with velocity V . We may write
.
, u = , v cos , v sin n
u = , v cos , v sin n
(15.178)
According to the general transformation rules of eqns. 15.50, 15.51, and 15.52, we may
write
= + V v cos
v cos = V + v cos
v sin = v sin
=
n
(15.179)
(15.180)
(15.181)
(15.182)
axis is taken to be V , and where (1 V 2 )1/2 . Note that the last two of
where the x
these equations may be written as a single vector equation for the transverse components.
15.7. RELATIVISTIC KINEMATICS OF PARTICLE COLLISIONS
321
Dividing the eqn. 15.181 by eqn. 15.180, we obtain the result
tan =
sin
V
v
+ cos
We can then use eqn. 15.179 to relate v and cos :
1 v 2 =
.
(15.183)
1 + V v cos .
Squaring both sides, we obtain a quadratic equation whose roots are
p
2 V cos 4 2 2 (1 V 2 cos2 )
v =
.
2 + 2 V 2 cos2
(15.184)
(15.185)
CM frame mass and velocity
To find the velocity of the CM frame, simply write
Ptot
=
N
X
Pi =
N
X
i mi ,
M (1 , V ) .
Then
N
X
M =
i mi
i=1
and
!2
PN
i=1
V = P
N
N
X
i mi vi
i=1
i mi vi
i=1 i mi
15.7.5
i mi vi
i=1
i=1
i=1
N
X
(15.186)
(15.187)
!2
(15.188)
(15.189)
Compton scattering
An extremely important example of relativistic scattering occurs when a photon scatters
off an electron: e + e + (see fig. 15.10). Let us work in the rest frame of the
reactant electron. Then we have
Pe = me (1, 0)
Pee = me ( , V )
(15.190)
for the initial and final 4-momenta of the electron. For the photon, we have
P = ( , k)
e ,
Pe = (e
, k)
(15.191)
where weve set ~ = 1 as well. Conservation of 4-momentum entails
P Pe = Pee Pe .
(15.192)
322
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.10: Compton scattering of a photon and an electron.
Thus,
= m 1 , V .
, kk
e
(15.193)
Squaring each side, we obtain
2
2 = 2
kk
(cos 1)
= m2e ( 1)2 2 V 2
= 2m2e (1 )
= 2me
) .
(15.194)
Here we have used |k| = for photons, and also ( 1) me =
, from eqn. 15.193.
Restoring the units ~ and c, we find the Compton formula
1
1
~
=
1
cos
.
(15.195)
me c2
This is often expressed in terms of the photon wavelengths, as
= 4~ sin2 1 ,
(15.196)
2
me c
showing that the wavelength of the scattered light increases with the scattering angle in the
rest frame of the target electron.
15.8
Covariant Electrodynamics
We begin with the following expression for the Lagrangian density of charged particles
coupled to an electromagnetic field, and then show that the Euler-Lagrange equations recapitulate Maxwells equations. The Lagrangian density is
1
1
F F j A .
(15.197)
L=
16
c
15.8. COVARIANT ELECTRODYNAMICS
323
Here, A = ( , A) is the electromagnetic 4-potential, which combines the scalar field
and the vector field A into a single 4-vector. The quantity F is the electromagnetic field
strength tensor and is given by
F = A A .
(15.198)
Note that as defined F = F is antisymmetric. Note that, if i = 1, 2, 3 is a spatial
index, then
F0i =
Fij =
1 Ai A0
= Ei
c t
xi
(15.199)
Ai Aj
= ijk Bk .
xj
xi
(15.200)
Here we have used A = (A0 , A) and A = (A0 , A), as well as = (c1 t , ).
IMPORTANT : Since the electric and magnetic fields E and B are not part of a 4-vector,
we do not use covariant / contravariant notation for their components. Thus, Ei is the ith
component of the vector E. We will not write E i with a raised index, but if we did, wed
mean the same thing: E i = Ei . By contrast, for the spatial components of a four-vector
like A , we have Ai = Ai .
Explicitly, then, we have
0
Ex
Ey
Ez
0
Bz By
Ex
F =
Ey Bz
0
Bx
Ez By Bx
0
0 Ex Ey Ez
0
Bz By
E
,
F = x
,
Ey Bz
0
Bx
Ez By Bx
0
(15.201)
where F = g g F . Note that when comparing F and F , the components with
one space and one time index differ by a minus sign. Thus,
1
E2 B2
F F =
,
16
8
(15.202)
which is the electromagnetic Lagrangian density. The j A term accounts for the interaction
between matter and electromagnetic degrees of freedom. We have
1
1
j A = j A ,
c
c
where
c
j =
j
A =
,
A
(15.203)
(15.204)
where is the charge density and j is the current density. Charge conservation requires
j =
+ j = 0 .
t
(15.205)
324
CHAPTER 15. SPECIAL RELATIVITY
We shall have more to say about this further on below.
Let us now derive the Euler-Lagrange equations for the action functional,
Z
1
1
4
S = c
F F + c j A
dx
.
16
(15.206)
We first vary with respect to A . Clearly
F = A A .
We then have
L =
1
F c1 j
4
(15.207)
1
F A
4
Ignoring the boundary term, we obtain Maxwells equations,
F = 4c1 j
(15.208)
(15.209)
The = k component of these equations yields
0 F 0k + i F jk = 0 Ek jkl j Bl = 4c1 j k ,
(15.210)
which is the k component of the Maxwell-Amp`ere law,
B =
4
1 E
j+
.
c
c t
(15.211)
The = 0 component reads
i F i0 =
4 0
j
c
E = 4
(15.212)
which is Gausss law. The remaining two Maxwell equations come for free from the very
definitions of E and B:
E = A0
B =A
1 A
c t
,
(15.213)
(15.214)
which imply
1 B
c t
B =0 .
E =
15.8.1
(15.215)
(15.216)
Lorentz force law
This has already been worked out in chapter 7. Here we reiterate our earlier derivation.
The 4-current may be written as
X Z
dXn (4)
(x X) .
(15.217)
j (x, t) = c
qn d
d
n
15.8. COVARIANT ELECTRODYNAMICS
325
Thus, writing Xn = ct , Xn (t) , we have
X
j 0 (x, t) =
qn c x Xn (t)
n
j(x, t) =
X
n
(15.218)
qn X n (t) x Xn (t) .
(15.219)
The Lagrangian for the matter-field interaction term is then
Z
L = c1 d3x j 0 A0 j A
X
qn
=
qn (Xn , t)
A(Xn , t) Xn ,
c
n
(15.220)
where = A0 . For each charge qn , this is equivalent to a particle with velocity-dependent
potential energy
q
U (x, t) = q (x, t) A(r, t) x ,
(15.221)
c
where x = Xn .
Lets work out the equations of motion. We assume a kinetic energy T =
charge. We then have
L
d L
=
dt x
x
1
2
2 mx
for the
(15.222)
with L = T U , which gives
q
q dA
,
= q + (A x)
c dt
c
(15.223)
q Ai j q Ai
q Aj j
x
+
=
q
+
x ,
c xj
c t
xi
c xi
(15.224)
+
mx
or, in component notation,
mx
i +
which is to say
q Ai q
mx
= q i
+
x
c t
c
i
Aj
Ai
xi
xj
x j .
(15.225)
It is convenient to express the cross product in terms of the completely antisymmetric tensor
of rank three, ijk :
Ak
,
(15.226)
Bi = ijk
xj
and using the result
ijk imn = jm kn jn km ,
(15.227)
we have ijk Bi = j Ak k Aj , and
mx
i = q
q Ai q
+ ijk x j Bk ,
xi
c t
c
(15.228)
326
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.11: Homer celebrates the manifest gauge invariance of classical electromagnetic
theory.
or, in vector notation,
q A q
= q
+ x ( A)
mx
c t
c
q
= q E + x B ,
c
(15.229)
which is, of course, the Lorentz force law.
15.8.2
Gauge invariance
R
The action S = c1 d4x L admits a gauge invariance. Let A A + , where (x, t)
is an arbitrary scalar function of spacetime coordinates. Clearly
F F + = F ,
(15.230)
and hence the fields E and B remain invariant under the gauge transformation, even though
the 4-potential itself changes. What about the matter term? Clearly
c1 j A c1 j A c1 j
= c1 j A + c1 j c1 j .
(15.231)
Once again we ignore the boundary term. We may now invoke charge conservation to write
j = 0, and we conclude that the action is invariant! Woo hoo! Note also the very deep
connection
gauge invariance
charge conservation .
(15.232)
15.8. COVARIANT ELECTRODYNAMICS
15.8.3
327
Transformations of fields
One last detail remains, and that is to exhibit explicitly the Lorentz transformation properties of the electromagnetic field. For the case of vectors like A , we have
A = L A .
(15.233)
The E and B fields, however, appear as elements in the field strength tensor F . Clearly
this must transform as a tensor:
F = L L F = L F Lt .
(15.234)
We can write a general Lorentz transformation as a product of a rotation Lrot and a boost
Lboost . Lets first see how rotations act on the field strength tensor. We take
!
111 013
L = Lrot =
,
(15.235)
031 R33
where Rt R = I, i.e. R O(3) is an orthogonal matrix. We must compute
!
!
0
Ek
1 0
1 0
t
L F
L =
t
0 Rkl
Ej jkm Bm
0 Rij
!
t
0
Ek Rkl
=
.
Rij Ej jkm Rij Rlk Bm
(15.236)
Thus, we conclude
El = Rlk Ek
iln Bn =
jkm Rij Rlk Bm
(15.237)
.
(15.238)
Now for any 3 3 matrix R we have
jks Rij Rlk Rrs = det(R) ilr ,
(15.239)
and therefore
jkm Rij Rlk Bm
= jkm Rij Rlk Rnm Rns Bs
= det(R) iln Rns Bs ,
(15.240)
Therefore,
Ei = Rij Ej
Bi = det(R) Rij Bj .
(15.241)
For any orthogonal matrix, Rt R = I gives that det(R) = 1. The extra factor of det(R)
in the transformation properties of B is due to the fact that the electric field transforms as
a vector , while the magnetic field transforms as a pseudovector . Under space inversion, for
example, where R = I, the electric field is odd under this transformation (E E) while
328
CHAPTER 15. SPECIAL RELATIVITY
the magnetic field is even (B +B). Similar considerations hold in particle mechanics for
the linear momentum, p (a vector) and the angular momentum L = r p (a pseudovector).
The analogy is not complete, however, because while both p and L are odd under the
operation of time-reversal, E is even while B is odd.
OK, so how about boosts? We can write the general boost, from eqn. 15.37, as
L=
I + ( 1)P
(15.242)
where Pij = i j is the projector onto the direction of . We now compute
!
t
t
0
E
t
. (15.243)
L F
L =
I + ( 1)P
I + ( 1)P
E jkm Bm
Carrying out the matrix multiplications, we obtain
E = (E B ) ( 1)( E )
B = (B + E ) ( 1)( B ) .
Expressed in terms of the components Ek , E , Bk , and B , one has
B
Ek = Ek
,
E = E
+ E
Bk = Bk
,
B = B
.
(15.244)
(15.245)
(15.246)
(15.247)
Recall that for any vector , we write
k =
(15.248)
= ( ) ,
(15.249)
so that = 0.
15.8.4
Invariance versus covariance
We saw that the laws of electromagnetism were gauge invariant. That is, the solutions to
the field equations did not change under a gauge transformation A A + . With
respect to Lorentz transformations, however, the theory is Lorentz covariant. This means
that Maxwells equations in different inertial frames take the exact same form, F =
4c1 j , but that both the fields and the sources transform appropriately under a change
in reference frames. The sources are described by the current 4-vector j = (c , j) and
transform as
c = c + jk
(15.250)
jk = c + jk
(15.251)
j = j
(15.252)
15.8. COVARIANT ELECTRODYNAMICS
329
The fields transform according to eqns. 15.246 and 15.247.
Consider, for example, a static point charge q located at the origin in the frame K , which
with respect to K. An observer in K measures a charge density
moves with velocity u x
(x , t ) = q (x ). The electric and magnetic fields in the K frame are then E = q r /r 2
and B = 0. For an observer in the K frame, the coordinates transform as
ct = ct + x
ct = ct x
x = ct + x ,
x = ct + x
(15.253)
(15.254)
as well as y = y and z = z . The observer in the K frame sees instead a charge at
x = (ct , ut , 0 , 0) and both a charge density as well as a current density:
(x, t) = (x , t ) = q (x ut) (y) (z)
(15.255)
= u q (x ut) (y) (z) x
.
j(x, t) = c (x , t ) x
(15.256)
OK, so much for the sources. How about the fields? Expressed in terms of Cartesian
coordinates, the electric field in K is given by
E (x , t ) = q
+ y y + z z
x x
3/2 .
x 2 + y 2 + z 2
(15.257)
From eqns. 15.246 and 15.247, we have Ex = Ex and Bx = Bx = 0. Furthermore, we have
Ey = Ey , Ez = Ez , By = Ez , and Bz = Ey . Thus,
+ y y + z z
(x ut)x
E(x, t) = q
3/2
2 (x ut)2 + y 2 + z 2
B(x, t) =
Let us define
y z z y
u
q
3/2 .
c
2 (x ut)2 + y 2 + z 2
+ y y + z z .
R(t) = (x ut) x
. We may then write
We further define the angle cos1 R
E(x, t) =
B(x, t) =
qR
1 2
3/2
R3
1 2 sin2
1 2
q R
3/2 .
R3
1 2 sin2
The fields are therefore enhanced in the transverse directions: E /Ek = 3 .
(15.258)
(15.259)
(15.260)
(15.261)
330
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.12: A relativistic runner carries a pole of proper length and runs into a barn of
proper length .
15.9
Appendix I : The Pole, the Barn, and Rashoman
Akira Kurosawas 1950 cinematic masterpiece, Rashoman, describes a rape, murder, and
battle from four different and often contradictory points of view. It poses deep questions
regarding the nature of truth. Psychologists sometimes refer to problems of subjective
perception as the Rashoman effect. In literature, William Faulkners 1929 novel, The Sound
and the Fury, which describes the tormented incestuous life of a Mississippi family, also is
told from four points of view. Perhaps Faulkner would be a more apt comparison with
Einstein, since time plays an essential role in his novel. For example, Quentins watch,
given to him by his father, represents time and the sweep of lifes arc (Quentin, I give you
the mausoleum of all hope and desire...). By breaking the watch, Quentin symbolically
attempts to escape time and fate. One could draw an analogy to Einstein, inheriting a watch
from those who came before him, which he too broke and refashioned. Did Faulkner know
of Einstein? But I digress.
Consider a relativistic runner carrying a pole of proper length , as depicted in fig. 15.12.
He runs toward a barn of proper length at velocity u = c. Let the frame of the barn
be K and the frame of the runner be K . Recall that the Lorentz transformations between
frames K and K are given by
ct = ct + x
ct = ct x
x = ct + x .
x = ct + x
(15.262)
(15.263)
We define the following points. Let A denote the left door of the barn and B the right
door. Furthermore, let P denote the left end of the pole and Q its right end. The spacetime
coordinates for these points in the two frames are clearly .
P = (ct , 0)
A = (ct , 0)
B = (ct , )
Q = (ct , )
(15.264)
(15.265)
We now compute A and B in frame K , as well as P and Q in frame K:
A = (ct , ct)
(ct , ct )
B = (ct , ct + )
(ct , ct +
) .
(15.266)
(15.267)
15.9. APPENDIX I : THE POLE, THE BARN, AND RASHOMAN
331
Similarly,
P = (ct , ct )
Q = (ct + , ct + )
(ct , ct + 1 ) .
(ct , ct)
(15.268)
(15.269)
We now define four events, by the coincidences of A and B with P and Q:
Event I : The right end of the pole enters the left door of the barn. This is described
by Q = A in frame K and by Q = A in frame K .
Event II : The right end of the pole exits the right door of the barn. This is described
by Q = B in frame K and by Q = B in frame K .
Event III : The left end of the pole enters the left door of the barn. This is described
by P = A in frame K and by P = A in frame K .
Event IV : The left end of the pole exits the right door of the barn. This is described
by P = B in frame K and by P = B in frame K .
Mathematically, we have in frame K that
I:
Q=A
tI =
II :
Q=B
tII = ( 1)
III :
P =A
IV :
P =B
tIII = 0
tIV =
u
(15.270)
(15.271)
(15.272)
(15.273)
In frame K , however
I:
Q = A
tI =
II :
Q = B
tII = ( 1)
III :
P = A
IV :
P = B
tIII = 0
tIV =
u
(15.274)
(15.275)
(15.276)
(15.277)
Thus, to an observer in frame K, the order of events is I, III, II, and IV, because
tI < tIII < tII < tIV .
(15.278)
For tIII < t < tII , he observes that the pole is entirely in the barn. Indeed, the right door can
start shut and the left door open, and sensors can automatically and, for the purposes of
argument, instantaneously trigger the closing of the left door immediately following event
332
CHAPTER 15. SPECIAL RELATIVITY
Figure 15.13: An object of proper length and moving with velocity u, when photographed
from an angle , appears to have a length .
III and the opening of the right door immediately prior to event II. So the pole can be
inside the barn with both doors shut!
But now for the Rashoman effect: according to the runner, the order of events is I, II, III,
and IV, because
tI < tII < tIII < tIV .
(15.279)
At no time does the runner observe the pole to be entirely within the barn. Indeed, for
tII < t < tIII , both ends of the pole are sticking outside of the barn!
15.10
Appendix II : Photographing a Moving Pole
What is the length of a moving pole of proper length 0 as measured by an observer at
rest? The answer would appear to be 1 0 , as we computed in eqn. 15.71. However, we
should be more precise when we we speak of length. The relation () = 1 0 tells us
the instantaneous end-to-end distance as measured in the observers rest frame K. But an
actual experiment might not measure this quantity.
For example, suppose a relativistic runner carrying a pole of proper length 0 runs past a
measuring rod which is at rest in the rest frame K of an observer. The observer takes a
photograph of the moving pole as it passes by. Suppose further that the angle between the
observers line of sight and the velocity u of the pole is , as shown in fig. 15.13. What is
the apparent length (, u) of the pole as observed in the photograph? (I.e. the pole will
appear to cover a portion of the measuring rod which is of length .)
15.10. APPENDIX II : PHOTOGRAPHING A MOVING POLE
333
The point here is that the shutter of the camera is very fast (otherwise the image will appear
blurry). In our analysis we will assume the shutter opens and closes instantaneously. Lets
define two events:
Event 1 : photon 1 is emitted by the rear end of the pole.
Event 2 : photon 2 is emitted by the front end of the pole.
Both photons must arrive at the cameras lens simultaneously. Since, as shown in the figure,
the path of photon #1 is longer by a distance cos , where is the apparent length of the
pole, 2 must be emitted a time t = c1 cos after 1 . Now if we Lorentz transform from
frame K to frame K , we have
x = x t .
(15.280)
But x = 0 is the proper length of the pole, and x = is the apparent length. With
ct = cos , then, we have
1 0
=
.
(15.281)
1 cos
When = 90 , we recover the familiar Lorentz-Fitzgerald contraction () = 1 0 . This
is because the photons 1 and 2 are then emitted simultaneously, and the photograph
measures the instantaneous end-to-end distance of the pole as measured in the observers
rest frame K. When cos 6= 0, however, the two photons are not emitted simultaneously,
and the apparent length is given by eqn. 15.281.
334
CHAPTER 15. SPECIAL RELATIVITY
Chapter 16
Hamiltonian Mechanics
16.1
The Hamiltonian
Recall that L = L(q, q,
t), and
p =
L
.
q
(16.1)
The Hamiltonian, H(q, p) is obtained by a Legendre transformation,
H(q, p) =
n
X
=1
p q L .
(16.2)
Note that
n
X
L
L
dH =
p dq + q dp
dq
dq
q
q
=1
n
X
L
L
dt .
dq
=
q dp
q
t
=1
L
dt
t
(16.3)
Thus, we obtain Hamiltons equations of motion,
H
= q
p
and
H
L
=
= p
q
q
dH
H
L
=
=
.
dt
t
t
(16.4)
(16.5)
Some remarks:
As an example, consider a particle moving in three dimensions, described by spherical
polar coordinates (r, , ). Then
(16.6)
L = 21 m r 2 + r 2 2 + r 2 sin2 2 U (r, , ) .
335
336
CHAPTER 16. HAMILTONIAN MECHANICS
We have
pr =
L
= mr
r
p =
L
= mr 2
p =
L
= mr 2 sin2 ,
(16.7)
and thus
H = pr r + p + p L
=
p2
p2
p2r
+
+
+ U (r, , ) .
2m 2mr 2 2mr 2 sin2
Note that H is time-independent, hence
of the motion.
H
t
dH
dt
(16.8)
= 0, and therefore H is a constant
In order to obtain H(q, p) we must invert the relation p =
L
q
= p (q, q)
to obtain
q (q, p). This is possible if the Hessian,
2L
p
=
q
q q
(16.9)
is nonsingular. This is the content of the inverse function theorem of multivariable
calculus.
Define the rank 2n vector, , by its components,
(
qi
i =
pin
if 1 i n
if n < i 2n .
(16.10)
Then we may write Hamiltons equations compactly as
H
i = Jij
,
j
where
J=
Onn Inn
Inn Onn
(16.11)
(16.12)
is a rank 2n matrix. Note that J t = J, i.e. J is antisymmetric, and that J 2 =
I2n2n . We shall utilize this symplectic structure to Hamiltons equations shortly.
16.2. MODIFIED HAMILTONS PRINCIPLE
16.2
337
Modified Hamiltons Principle
We have that
Ztb
Ztb
0 = dt L = dt p q H
ta
(16.13)
ta
Ztb
H
H
= dt p q + q p
q
p
q p
ta
)
Ztb (
tb
H
H
q + q
p + p q ,
= dt p +
q
p
ta
ta
assuming q (ta ) = q (tb ) = 0. Setting the coefficients of q and p to zero, we recover
Hamiltons equations.
16.3
Phase Flow is Incompressible
A flow for which v = 0 is incompressible we shall see why in a moment. Lets check
that the divergence of the phase space velocity does indeed vanish:
=
=
n
X
q
=1
2n
X
i=1
p
+
q
p
i X
2H
=
Jij
=0.
i
i j
(16.14)
i,j
Now let (, t) be a distribution on phase space. Continuity implies
=0.
+ ( )
t
(16.15)
D
=
+ = 0 ,
Dt
t
(16.16)
Invoking = 0, we have that
where D/Dt is sometimes
called the convective derivative it is the total derivative of the
function (t), t , evaluated at a point (t) in phase space which moves according to the
dynamics. This says that the density in the comoving frame is locally constant.
338
16.4
CHAPTER 16. HAMILTONIAN MECHANICS
Poincar
e Recurrence Theorem
Let g be the -advance mapping which evolves points in phase space according to Hamiltons equations
H
H
qi = +
,
p i =
(16.17)
pi
qi
for a time interval t = . Consider a region in phase space. Define gn to be the
nth image of under the mapping g . Clearly g is invertible; the inverse is obtained by
integrating the equations of motion backward in time. We denote the inverse of g by g1 .
By Liouvilles theorem, g is volume preserving when acting on regions in phase space, since
the evolution of any given point is Hamiltonian. This follows from the continuity equation
for the phase space density,
+ (u) = 0
(16.18)
t
p}
is the velocity vector in phase space, and Hamiltons equations, which
where u = {q,
say that the phase flow is incompressible, i.e. u = 0:
n
X
qi pi
u =
+
qi pi
i=1
(
)
n
X
H
H
=0.
(16.19)
=
+
qi pi
pi
qi
i=1
Thus, we have that the convective derivative vanishes, viz.
+ u = 0 ,
Dt
t
(16.20)
which guarantees that the density remains constant in a frame moving with the flow.
The proof of the recurrence theorem is simple. Assume that g is invertible and volumepreserving, as is the case for Hamiltonian flow. Further assume that phase space volume
is finite. Since the energy is preserved in the case of time-independent Hamiltonians, we
simply ask that the volume of phase space at fixed total energy E be finite, i.e.
Z
d E H(q, p) < ,
(16.21)
where d = dq dp is the phase space uniform integration measure.
Theorem: In any finite neighborhood of phase space there exists a point 0 which will
return to after n applications of g , where n is finite.
Proof: Assume the theorem fails; we will show this assumption results in a contradiction.
Consider the set formed from the union of all sets gm for all m:
=
m=0
gm
(16.22)
16.5. POISSON BRACKETS
339
We assume that the set {gm | m Z , m 0} is disjoint. The volume of a union of disjoint
sets is the sum of the individual volumes. Thus,
vol() =
vol(gm )
m=0
= vol()
m=1
1=,
(16.23)
since vol(gm ) = vol() from volume preservation. But clearly is a subset of the entire
phase space, hence we have a contradiction, because by assumption phase space is of finite
volume.
Thus, the assumption that the set {gm | m Z , m 0} is disjoint fails. This means that
there exists some pair of integers k and l, with k 6= l, such that gk gl 6= . Without
loss of generality we may assume k > l. Apply the inverse g1 to this relation l times to get
gkl 6= . Now choose any point gn , where n = k l, and define 0 = gn .
Then by construction both 0 and gn 0 lie within and the theorem is proven.
Each of the two central assumptions invertibility and volume preservation is crucial.
Without either of them, the proof fails. Consider, for example, a volume-preserving map
which is not invertible. An example might be a mapping f : R R which takes any real
number to its fractional part. Thus, f () = 0.14159265 . . .. Let us restrict our attention
to intervals of width less than unity. Clearly f is then volume preserving. The action of f
on the interval [2, 3) is to map it to the interval [0, 1). But [0, 1) remains fixed under the
action of f , so no point within the interval [2, 3) will ever return under repeated iterations
of f . Thus, f does not exhibit Poincare recurrence.
Consider next the case of the damped harmonic oscillator. In this case, phase space volumes
2
contract. For a one-dimensional oscillator obeying x
+2 x+
0 x = 0 one has u = 2 <
0 ( > 0 for damping). Thus the convective derivative obeys Dt = (u) = +2 which
says that the density increases exponentially in the comoving frame, as (t) = e2t (0).
Thus, phase space volumes collapse, and are not preserved by the dynamics. In this case, it
is possible for the set to be of finite volume, even if it is the union of an infinite number of
sets gn , because the volumes of these component sets themselves decrease exponentially,
as vol(gn ) = e2n vol(). A damped pendulum, released from rest at some small angle
0 , will not return arbitrarily close to these initial conditions.
16.5
Poisson Brackets
The time evolution of any function F (q, p) over phase space is given by
n
F X
F
F
d
F q(t), p(t), t =
+
q +
p
dt
t
q
p
=1
F
+ F, H ,
(16.24)
340
CHAPTER 16. HAMILTONIAN MECHANICS
where the Poisson bracket { , } is given by
n
X
A B
A B
A, B
q p
p q
=1
2n
X
i,j=1
Jij
A B
.
i j
(16.25)
(16.26)
Properties of the Poisson bracket:
Antisymmetry:
f, g = g, f .
(16.27)
Bilinearity: if is a constant, and f , g, and h are functions on phase space, then
f + g, h = f, h + {g, h .
(16.28)
Linearity in the second argument follows from this and the antisymmetry condition.
Associativity:
f g, h = f g, h + g f, h .
Jacobi identity:
f, {g, h} + g, {h, f } + h, {f, g} = 0 .
(16.29)
(16.30)
Some other useful properties:
If {A, H} = 0 and
A
t
= 0, then
dA
dt
= 0, i.e. A(q, p) is a constant of the motion.
If {A, H} = 0 and {B, H} = 0, then {A, B}, H = 0. If in addition A and B have
no explicit time dependence, we conclude that {A, B} is a constant of the motion.
It is easily established that
{q , q } = 0 ,
{p , p } = 0
{q , p } = .
16.6
Canonical Transformations
16.6.1
Point transformations in Lagrangian mechanics
(16.31)
In Lagrangian mechanics, we are free to redefine our generalized coordinates, viz.
Q = Q (q1 , . . . , qn , t) .
(16.32)
16.6. CANONICAL TRANSFORMATIONS
341
This is called a point transformation. The transformation is invertible if
Q
6= 0 .
det
q
(16.33)
written as a function of the new coordinates Q and velocThe transformed Lagrangian, L,
is
ities Q,
Q, Q,
t) = L q(Q, t), q(Q,
t) .
L
Q,
(16.34)
Finally, Hamiltons principle,
Ztb
t) = 0
dt L(Q,
Q,
(16.35)
t1
with Q (ta ) = Q (tb ) = 0, still holds, and the form of the Euler-Lagrange equations
remains unchanged:
d L
L
=0.
(16.36)
Q
dt Q
The invariance of the equations of motion under a point transformation may be verified
explicitly. We first evaluate
d L q
d L q
d L
,
=
=
dt Q
dt q Q
dt q Q
(16.37)
q
q
=
Q
Q
(16.38)
where the relation
follows from
q =
q
q
Q +
.
Q
t
(16.39)
Now we compute
L
L q
L q
=
+
Q
q Q
q Q
L q
2 q
L
2 q
=
+
Q +
q Q
q Q Q
Q t
L d q
d L q
+
=
dt q Q
q dt Q
d L q
d L
=
=
,
dt q Q
dt Q
where the last equality is what we obtained earlier in eqn. 16.37.
(16.40)
342
16.6.2
CHAPTER 16. HAMILTONIAN MECHANICS
Canonical transformations in Hamiltonian mechanics
In Hamiltonian mechanics, we will deal with a much broader class of transformations ones
which mix all the q s and p s. The general form for a canonical transformation (CT) is
q = q Q1 , . . . , Qn ; P1 , . . . , Pn ; t
(16.41)
p = p Q1 , . . . , Qn ; P1 , . . . , Pn ; t ,
(16.42)
with {1, . . . , n}. We may also write
i = i 1 , . . . , 2n ; t ,
(16.43)
with i {1, . . . , 2n}. The transformed Hamiltonian is H(Q,
P, t).
What sorts of transformations are allowed? Well, if Hamiltons equations are to remain
invariant, then
H
H
, P =
,
(16.44)
Q =
P
Q
which gives
P
i
Q
+
=0=
.
Q
P
i
(16.45)
I.e. the flow remains incompressible in the new (Q, P ) variables. We will also require that
phase space volumes are preserved by the transformation, i.e.
(Q, P )
i
= 1 .
det
(16.46)
=
j
(q, p)
Additional conditions will be discussed below.
16.6.3
Hamiltonian evolution
Hamiltonian evolution itself defines a canonical transformation. Let i = i (t) and i =
H
i (t + dt). Then from the dynamics i = Jij
, we have
j
i (t + dt) = i (t) + Jij
Thus,
Now, using the result
H
dt + O dt2 .
j
i
H
2
dt + O dt
=
+ Jik
j
j i
k
2H
dt + O dt2 .
= ij + Jik
j k
det 1 + M = 1 + Tr M + O(2 ) ,
(16.47)
(16.48)
(16.49)
16.6. CANONICAL TRANSFORMATIONS
343
we have
i
2H
2
j = 1 + Jjk j k dt + O dt
= 1 + O dt2 .
16.6.4
(16.50)
(16.51)
Symplectic structure
We have that
H
.
i = Jij
j
(16.52)
Suppose we make a time-independent canonical transformation to new phase space coordinates, a = a (). We then have
a
a
H
a =
j =
Jjk
.
j
j
k
(16.53)
But if the transformation is canonical, then the equations of motion are preserved, and we
also have
k H
H
= Jab
.
(16.54)
a = Jab
b
b k
Equating these two expressions, we have
Maj Jjk
H
1 H
= Jab Mkb
,
k
k
(16.55)
a
j
(16.56)
where
Maj
is the Jacobian of the transformation. Since the equality must hold for all , we conclude
MJ = J Mt
1
M JM t = J .
(16.57)
A matrix M satisfying M M t = I is of course an orthogonal matrix. A matrix M satisfying
M JM t = J is called symplectic. We write M Sp(2n), i.e. M is an element of the group
of symplectic matrices 1 of rank 2n.
The symplectic property of M guarantees that the Poisson brackets are preserved under a
1
Note that the rank of a symplectic matrix is always even. Note also M JM t = J implies M t JM = J.
344
CHAPTER 16. HAMILTONIAN MECHANICS
canonical transformation:
16.6.5
A, B
A B
i j
A a B b
= Jij
a i b j
A B
t
= Mai Jij Mjb
a b
A B
= Jab
a b
= A, B .
= Jij
(16.58)
Generating functions for canonical transformations
For a transformation to be canonical, we require
Ztb n
Ztb n
o
o
dt p q H(q, p, t) = 0 = dt P Q H(Q,
P, t) .
ta
(16.59)
ta
This is satisfied provided
n
o
dF
,
p q H(q, p, t) = P Q H(Q,
P, t) +
dt
(16.60)
where is a constant. For canonical transformations, = 1.2 Thus,
F
F
Q
H(Q,
P, t) = H(q, p, t) + P Q p q +
q +
q
Q
+
F
F
F
P +
p +
.
p
P
t
(16.61)
Thus, we require
F
= p
q
F
= P
Q
F
=0
p
F
=0.
P
(16.62)
The transformed Hamiltonian is
F
H(Q,
P, t) = H(q, p, t) +
.
t
(16.63)
2
Solutions of eqn. 16.60 with 6= 1 are known as extended canonical transformations. We can always
rescale coordinates and/or momenta to achieve = 1.
16.6. CANONICAL TRANSFORMATIONS
345
There are four possibilities, corresponding to the freedom to make Legendre transformations
with respect to each of the arguments of F (q, Q) :
F1
1
F1 (q, Q, t)
; p = + F
, P = Q
(type I)
F2
2
; p = + F
, Q = + P
(type II)
F2 (q, P, t) P Q
q
F (q, Q, t) =
F3
3
, P = Q
(type III)
; q = F
F3 (p, Q, t) + p q
F4
4
, Q = + P
(type IV)
F4 (p, P, t) + p q P Q ; q = F
p
In each case ( = 1, 2, 3, 4), we have
H(Q,
P, t) = H(q, p, t) +
.
t
Lets work out some examples:
(16.64)
Consider the type-II transformation generated by
F2 (q, P ) = A (q) P ,
(16.65)
where A (q) is an arbitrary function of the {q }. We then have
Q =
F2
= A (q)
P
p =
F2
A
=
P .
q
q
(16.66)
Thus,
q
p .
(16.67)
Q
This is a general point transformation of the kind discussed in eqn. 16.32. For a general
1
, i.e. Q = M q,
linear point transformation, Q = M q , we have P = p M
Q = A (q)
P =
P = p M 1 . If M = , this is the identity transformation. F2 = q1 P3 + q3 P1
interchanges labels 1 and 3, etc.
Consider the type-I transformation generated by
F1 (q, Q) = A (q) Q .
(16.68)
We then have
p =
F1
A
=
Q
q
q
P =
F1
= A (q) .
Q
Note that A (q) = q generates the transformation
q
P
.
p
+Q
(16.69)
(16.70)
(16.71)
346
CHAPTER 16. HAMILTONIAN MECHANICS
A mixed transformation is also permitted. For example,
F (q, Q) = q1 Q1 + (q3 Q2 ) P2 + (q2 Q3 ) P3
(16.72)
is of type-I with respect to index = 1 and type-II with respect to indices = 2, 3.
The transformation effected is
Q1 = p 1
Q2 = q 3
Q3 = q 2
(16.73)
P1 = q1
P2 = p3
P3 = p2 .
(16.74)
Consider the harmonic oscillator,
H(q, p) =
p2
+ 12 kq 2 .
2m
(16.75)
If we could find a time-independent canonical transformation such that
r
p
2 f (P )
p = 2mf (P ) cos Q
sin Q ,
,
q=
k
(16.76)
where f (P ) is some function of P , then wed have H(Q,
P ) = f (P ), which is cyclic in
Q. To find this transformation, we take the ratio of p and q to obtain
p = mk q ctn Q ,
(16.77)
which suggests the type-I transformation
F1 (q, Q) =
1
2
mk q 2 ctn Q .
(16.78)
This leads to
F1
p=
= mk q ctn Q
q
Thus,
where =
2P
sin Q
q=
4
mk
mk q 2
F1
P =
=
.
Q
2 sin2 Q
f (P ) =
k
P = P ,
m
(16.79)
(16.80)
k/m is the oscillation frequency. We therefore have
H(Q,
P ) = P ,
(16.81)
whence P = E/. The equations of motion are
H
P =
=0 ,
Q
which yields
Q(t) = t + 0
q(t) =
H
Q =
= ,
P
(16.82)
(16.83)
2E
sin t + 0 .
2
m
16.7. HAMILTON-JACOBI THEORY
16.7
347
Hamilton-Jacobi Theory
Weve stressed the great freedom involved in making canonical transformations. Coordinates and momenta, for example, may be interchanged the distinction between them is
purely a matter of convention! We now ask: is there any specially preferred canonical trans
formation? In this regard, one obvious goal is to make the Hamiltonian H(Q,
P, t) and the
corresponding equations of motion as simple as possible.
Recall the general form of the canonical transformation:
F
,
H(Q,
P ) = H(q, p) +
t
(16.84)
with
F
= p
q
F
=0
p
(16.85)
F
= P
Q
F
=0.
P
(16.86)
We now demand that this transformation result in the simplest Hamiltonian possible, that
is, H(Q,
P, t) = 0. This requires we find a function F such that
F
= H
t
F
q
= p .
(16.87)
The remaining functional dependence may be taken to be either on Q (type I) or on P
(type II). As it turns out, the generating function F we seek is in fact the action, S, which
is the integral of L with respect to time, expressed as a function of its endpoint values.
16.7.1
The action as a function of coordinates and time
We have seen how the action S[( )] is a functional of the path ( ) and a function of the
endpoint values {qa , ta } and {qb , tb }. Let us define the action function S(q, t) as
Zt
S(q, t) = d L , ,
) ,
(16.88)
ta
where ( ) starts at (qa , ta ) and ends at (q, t). We also require that ( ) satisfy the EulerLagrange equations,
d L
L
=0
(16.89)
d
Let us now consider a new path, ( ), also starting at (qa , ta ), but ending at (q + dq, t + dt),
348
CHAPTER 16. HAMILTONIAN MECHANICS
and also satisfying the equations of motion. The differential of S is
dS = S ( ) S ( )
=
t+dt
Z
ta
Zt
Zt
d L(
, , ) d L , ,
)
ta
(16.90)
ta
i
i
L h
L h
( ) ( ) +
( ) ( )
)h
Zt (
i
d L
L
= d
( ) ( )
d
ta
h
i
L
(t)
(t)
+ L (t), (t), t dt
+
+ L (t), (t), t dt
= 0 + (t) (t) + L (t), (t),
t dt + O(q dt) ,
where we have defined
=
L
,
(16.91)
(16.92)
and
( ) ( ) ( ) .
(16.93)
Note that the differential dq is given by
dq = (t + dt) (t)
(16.94)
= (t + dt) (t) + (t) (t)
= (t) dt + (t)
= q (t) dt + (t) + O(q dt) .
(16.95)
Thus, with (t) p , we have
dS = p dq + L p q dt
= p dq H dt .
We therefore obtain
S
= p
q
S
= H
t
dS
=L.
dt
(16.96)
(16.97)
What
about the lower limit at ta ? Clearly there are n + 1 constants associated with this
limit: q1 (ta ), . . . , qn (ta ); ta . Thus, we may write
S = S(q1 , . . . , qn ; 1 , . . . , n , t) + n+1 ,
(16.98)
16.7. HAMILTON-JACOBI THEORY
349
Figure 16.1: A one-parameter family of paths q(s; ).
where our n + 1 constants are {1 , . . . , n+1 }. If we regard S as a mixed generator, which
is type-I in some variables and type-II in others, then each for 1 n may be chosen
to be either Q or P . We will define
(
+Q if = P
S
(16.99)
=
=
P if = Q
For each , the two possibilities = Q or = P are of course rendered equivalent by
a canonical transformation (Q , P ) (P , Q ).
16.7.2
The Hamilton-Jacobi equation
Since the action S(q, , t) has been shown to generate a canonical transformation for which
H(Q,
P ) = 0. This requirement may be written as
S
S S
H q1 , . . . , qn ,
=0.
(16.100)
,...,
,t +
q1
qn
t
This is the Hamilton-Jacobi equation (HJE). It is a first order partial differential equation
in n + 1 variables, and in general is nonlinear (since kinetic energy is generally a quadratic
function of momenta). Since H(Q,
P, t) = 0, the equations of motion are trivial, and
Q (t) = const.
P (t) = const.
(16.101)
Once the HJE is solved, one must invert the relations = S(q, , t)/ to obtain
q(Q, P, t). This is possible only if
2S
6= 0 ,
(16.102)
det
q
350
CHAPTER 16. HAMILTONIAN MECHANICS
which is known as the Hessian condition.
It is worth noting that the HJE may have several solutions. For example, consider the case
of the free particle, with H(q, p) = p2 /2m. The HJE is
1 S 2 S
=0.
(16.103)
+
2m q
t
One solution of the HJE is
S(q, , t) =
m (q )2
.
2t
(16.104)
For this we find
S
= (q ) q(t) = t .
t
m
Here = q(0) is the initial value of q, and = p is minus the momentum.
=
(16.105)
Another equally valid solution to the HJE is
S(q, , t) = q
2m t .
(16.106)
This yields
r
r
2m
S
=q
t q(t) =
(t + ) .
(16.107)
=
2m
For
p this solution, is the energy and may be related to the initial value of q(t) =
/2m.
16.7.3
Time-independent Hamiltonians
When H has no explicit time dependence, we may reduce the order of the HJE by one,
writing
S(q, , t) = W (q, ) + T (, t) .
(16.108)
The HJE becomes
W
T
H q,
.
=
q
t
(16.109)
S(q, , t) = W (q, ) 1 t .
(16.110)
Note that the LHS of the above equation is independent of t, and the RHS is independent
of q. Therefore, each side must only depend on the constants , which is to say that each
side must be a constant, which, without loss of generality, we take to be 1 . Therefore
The function W (q, ) is called Hamiltons characteristic function. The HJE now takes the
form
W
W
,...,
= 1 .
(16.111)
H q1 , . . . , qn ,
q1
qn
Note that adding an arbitrary constant C to S generates the same equation, and simply
shifts the last constant n+1 n+1 + C. This is equivalent to replacing t by t t0 with
t0 = C/1 , i.e. it just redefines the zero of the time variable.
16.7. HAMILTON-JACOBI THEORY
16.7.4
351
Example: one-dimensional motion
As an example of the method, consider the one-dimensional system,
H(q, p) =
The HJE is
(16.112)
1 S 2
+ U (q) = .
2m q
(16.113)
q
2m U (q) ,
(16.114)
which may be recast as
S
=
q
with solution
p2
+ U (q) .
2m
Z
p
S(q, , t) = 2m dq U (q ) t .
We now have
S
=
p=
q
as well as
S
=
=
m
2
2m U (q) ,
q(t)
Thus, the motion q(t) is given by quadrature:
+t=
m
2
dq
t .
U (q )
Zq(t)
dq
p
,
U (q )
(16.115)
(16.116)
(16.117)
(16.118)
where and are constants. The lower limit on the integral is arbitrary and merely shifts
t by another constant. Note that is the total energy.
16.7.5
Separation of variables
It is convenient to first work an example before discussing the general theory. Consider the
following Hamiltonian, written in spherical polar coordinates:
potential U (r,,)
}|
{
p2
B()
1
C()
2
H=
+ 2 2 + A(r) + 2 + 2 2 .
p +
2m r r 2
r
r sin
r sin
p2
(16.119)
We seek a solution with the characteristic function
W (r, , ) = Wr (r) + W () + W () .
(16.120)
352
CHAPTER 16. HAMILTONIAN MECHANICS
The HJE is then
W 2
1 Wr 2
W 2
1
1
+
+
2m r
2mr 2
2mr 2 sin2
+ A(r) +
C()
B()
+ 2 2 = 1 = E .
2
r
r sin
(16.121)
Multiply through by r 2 sin2 to obtain
(
)
2
1 W 2
W
1
+ C() = sin2
+ B()
2m
2m
(
)
1 Wr 2
2
2
r sin
+ A(r) 1 .
2m r
(16.122)
The LHS is independent of (r, ), and the RHS is independent of . Therefore, we may set
1 W 2
+ C() = 2 .
2m
(16.123)
Proceeding, we replace the LHS in eqn. 16.122 with 2 , arriving at
(
)
2
1
W
2
1 W 2
r
= r 2
+ B() +
+ A(r) 1 .
2m
2m r
sin2
(16.124)
The LHS of this equation is independent of r, and the RHS is independent of . Therefore,
1 W 2
2
+ B() +
= 3 .
2m
sin2
(16.125)
3
1 Wr 2
+ A(r) + 2 = 1 .
2m r
r
(16.126)
Were left with
The full solution is therefore
S(q, , t) =
Zr r
3
2m dr 1 A(r ) 2
r
+
(16.127)
Z r
2
2m d 3 B( )
sin2
+
2m
2 C( ) 1 t .
(16.128)
16.7. HAMILTON-JACOBI THEORY
We then have
S
=
1 =
1
r(t)
pm
2
353
dr
t
1 A(r ) 3 r 2
pm
Z (t) p m
Z (t)
d
S
2 d
q
q 2
2 =
+
=
2
sin2 3 B( ) 2 csc2
2 C( )
pm
pm
Z r(t)
Z (t)
S
2 dr
2 d
q
=
+ q
.
3 =
3
r 2 1 A(r ) 3 r 2
3 B( ) 2 csc2
q
(16.129)
(16.130)
(16.131)
The game plan here is as follows. The first of the above trio of equations is inverted to yield
r(t) in terms of t and constants. This solution is then invoked in the last equation (the upper
limit on the first integral on the RHS) in order to obtain an implicit equation for (t), which
is invoked in the second equation to yield an implicit equation for (t). The net result is
the motion of the system in terms of time t and the six constants (1 , 2 , 3 , 1 , 2 , 3 ). A
seventh constant, associated with an overall shift of the zero of t, arises due to the arbitrary
lower limits of the integrals.
In general, the separation of variables method begins with3
W (q, ) =
n
X
W (q , ) .
(16.132)
=1
Each W (q , ) may be regarded as a function of the single variable q , and is obtained by
satisfying an ODE of the form4
dW
= .
(16.133)
H q ,
dq
We then have
p =
W
q
W
+ ,1 t .
(16.134)
Note that while each W depends on only a single q , it may depend on several of the .
16.7.6
Example #2 : point charge plus electric field
Consider a potential of the form
k
Fz ,
(16.135)
r
which corresponds to a charge in the presence of an external point charge plus an external
electric field. This problem is amenable to separation in parabolic coordinates, (, , ):
p
p
x = cos , y = sin , z = 21 ( ) .
(16.136)
U (r) =
Here we assume complete separability. A given system may only be partially separable.
H
` (q , p ) may also depend on several of the . See e.g. eqn. 16.126, which is of the form
Hr r, r Wr , 3 = 1 .
4
354
CHAPTER 16. HAMILTONIAN MECHANICS
Note that
p
p
x2 + y 2 =
p
r = 2 + z 2 = 21 ( + ) .
The kinetic energy is
T = 12 m 2 + 2 2 + z 2
2
2
1
= 8 m ( + )
+
+ 21 m 2 ,
(16.137)
(16.138)
(16.139)
and hence the Lagrangian is
2 2
L = 18 m ( + )
+
+ 12 m 2
2k
+ 1 F ( ) .
+ 2
(16.140)
Thus, the conjugate momenta are
p =
L
= 41 m ( + )
(16.141)
p =
L
= 14 m ( + )
(16.142)
p =
L
= m ,
(16.143)
and the Hamiltonian is
H = p + p + p
2
=
m
p2 + p2
+
(16.144)
+
p2
2k
+
1 F ( ) .
2m + 2
(16.145)
Notice that H/t = 0, which means dH/dt = 0, i.e. H = E 1 is a constant of the
motion. Also, is cyclic in H, so its conjugate momentum p is a constant of the motion.
We write
S(q, ) = W (q, ) Et
= W (, ) + W (, ) + W (, ) Et .
(16.146)
(16.147)
with E = 1 . Clearly we may take
W (, ) = P ,
(16.148)
16.7. HAMILTON-JACOBI THEORY
355
where P = 2 . Multiplying the Hamilton-Jacobi equation by 21 m ( + ) then gives
dW
2
P2
+ mk 41 F 2 12 mE
4
dW 2 P2 1 2 1
4 F + 2 mE ,
=
d
4
(16.149)
where = 3 is the third constant: = (E, P , ). Thus,
q
Z s
z }| {
P2
mk 1
S , , ; E, P , = d 12 mE +
mF
+
4
4 2
| {z }
Z s
P2
+ d 12 mE 41 mF 2
+ P Et .
16.7.7
(16.150)
Example #3 : Charged Particle in a Magnetic Field
The Hamiltonian is
e 2
1
p A .
2m
c
and we write
We choose the gauge A = Bxy,
(16.151)
H=
S(x, y, P1 , P2 ) = Wx (x, P1 , P2 ) + Wy (y, P1 , P2 ) P1 t .
(16.152)
Note that here we will consider S to be a function of {q } and {P }.
The Hamilton-Jacobi equation is then
Wx
x
2
Wy
eBx
y
c
2
= 2mP1 .
(16.153)
2
(16.154)
We solve by writing
Wy = P2 y
dWx
dx
2
eBx
+ P2
c
= 2mP1 .
This equation suggests the substitution
x=
in which case
cP2
c p
2mP1 sin .
+
eB
eB
c p
x
=
2mP1 cos
eB
(16.155)
(16.156)
356
CHAPTER 16. HAMILTONIAN MECHANICS
and
Wx
1 Wx
Wx
eB
=
=
.
x
x
c 2mP1 cos
Substitution this into eqn. 16.154, we have
2mcP1
Wx
=
cos2 ,
eB
(16.157)
(16.158)
with solution
mcP1
mcP1
+
sin(2) .
eB
2eB
Wx
Wx x p
px =
=
= 2mP1 cos
x
Wx =
We then have
and
py =
Wy
= P2 .
y
(16.159)
(16.160)
(16.161)
The type-II generator we seek is then
mcP1
mcP1
+
sin(2) + P2 y P1 t ,
eB
2eB
where
cP2
eB
1
sin
.
x
=
eB
c 2mP1
Note that, from eqn. 16.155, we may write
c
1
mc
c p
dx =
2mP1 cos d ,
dP2 +
sin dP1 +
eB
eB 2mP1
eB
S(q, P, t) =
(16.162)
(16.163)
(16.164)
from which we derive
tan
=
P1
2P1
1
.
=
P2
2mP1 cos
(16.165)
These results are useful in the calculation of Q1 and Q2 :
Q1 =
S
P1
mcP1
mc
mcP1
mc
+
+
sin(2) +
cos(2)
t
eB
eB P1
2eB
eB
P1
mc
=
t
eB
(16.166)
and
Q2 =
S
P2
mcP1
1 + cos(2)
eB
P2
p
c
=y
2mP1 cos .
eB
=y+
(16.167)
16.8. ACTION-ANGLE VARIABLES
357
Now since H(P,
Q) = 0, we have that Q = 0, which means that each Q is a constant. We
therefore have the following solution:
x(t) = x0 + A sin(c t + )
(16.168)
y(t) = y0 + A cos(c t + ) ,
(16.169)
where c = eB/mc is the cyclotron frequency, and
x0 =
cP2
eB
y 0 = Q2
c Q1
A=
c p
2mP1 .
eB
16.8
Action-Angle Variables
16.8.1
Circular Phase Orbits: Librations and Rotations
(16.170)
In a completely integrable system, the Hamilton-Jacobi equation may be solved by separation of variables. Each momentum p is a function of only its corresponding coordinate q
plus constants no other coordinates enter:
p =
W
= p (q , ) .
q
(16.171)
The motion satisfies
H (q , p ) = .
(16.172)
The level sets of H are curves C . In general, these curves each depend on all of the
constants , so we write C = C (). The curves C are the projections of the full motion
onto the (q , p ) plane. In general we will assume the motion, and hence the curves C ,
is bounded. In this case, two types of projected motion are possible: librations and rotations. Librations are periodic oscillations about an equilibrium position. Rotations involve
the advancement of an angular variable by 2 during a cycle. This is most conveniently
illustrated in the case of the simple pendulum, for which
H(p , ) =
p2
2I
+ 21 I 2 1 cos .
(16.173)
When E < I 2 , the momentum p vanishes at = cos1 (2E/I 2 ). The system
executes librations between these extreme values of the angle .
When E > I 2 , the kinetic energy is always positive, and the angle advances monotonically, executing rotations.
In a completely integrable system, each C is either a libration or a rotation5 . Both librations
and rotations are closed curves. Thus, each C is in general homotopic to (= can be
5
C may correspond to a separatrix, but this is a nongeneric state of affairs.
358
CHAPTER 16. HAMILTONIAN MECHANICS
Figure 16.2: Phase curves for the simple pendulum, showing librations (in blue), rotations
(in green), and the separatrix (in red). This phase flow is most correctly viewed as taking
place on a cylinder, obtained from the above sketch by identifying the lines = and
= .
continuously distorted to yield) a circle, S1 . For n freedoms, the motion is therefore
confined to an n-torus, Tn :
n times
Tn
}|
{
z
= S1 S1 S1 .
(16.174)
These are called invariant tori (or invariant manifolds). There are many such tori, as there
are many C curves in each of the n two-dimensional submanifolds.
Invariant tori never intersect! This is ruled out by the uniqueness of the solution to the
dynamical system, expressed as a set of coupled ordinary differential equations.
Note also that phase space is of dimension 2n, while the invariant tori are of dimension n.
Phase space is covered by the invariant tori, but it is in general difficult to conceive of how
this happens. Perhaps the most accessible analogy is the n = 1 case, where the 1-tori are
just circles. Two-dimensional phase space is covered noninteracting circular orbits. (The
orbits are topologically equivalent to circles, although geometrically they may be distorted.)
It is challenging to think about the n = 2 case, where a four-dimensional phase space is
filled by nonintersecting 2-tori.
16.8.2
Action-Angle Variables
For a completely integrable system, one can transform canonically from (q, p) to new coordinates (, J) which specify a particular n-torus Tn as well as the location on the torus,
which is specified by n angle variables. The {J } are momentum variables which specify
the torus itself; they are constants of the motion since the tori are invariant. They are
16.8. ACTION-ANGLE VARIABLES
359
called action variables. Since J = 0, we must have
H
J =
=0
H = H(J) .
(16.175)
The { } are the angle variables.
The coordinate describes the projected motion along C , and is normalized by
I
d = 2
(once around C ) .
(16.176)
C
The dynamics of the angle variables are given by
H
(J) .
=
J
(16.177)
Thus,
(t) = (0) + (J) t .
(16.178)
The (J) are frequencies describing the rate at which the C are traversed; T (J) =
2/ (J) is the period.
16.8.3
Canonical Transformation to Action-Angle Variables
The {J } determine the {C }; each q determines a point on C . This suggests a type-II
transformation, with generator F2 (q, J):
p =
F2
q
F2
.
J
(16.179)
Note that6
2 =
d =
I
I
I
2F2
F2
p dq ,
dq =
=
d
J
J q
J
which suggests the definition
(16.180)
1
J =
2
p dq .
(16.181)
I.e. J is (2)1 times the area enclosed by C .
If, separating variables,
W (q, ) =
W (q , )
(16.182)
` F2
2
2
In general, we should write d J
= JFq
dq with a sum over . However, in eqn. 16.180 all
coordinates and momenta other than q and p are held fixed. Thus, = is the only term in the sum
which contributes.
6
360
CHAPTER 16. HAMILTONIAN MECHANICS
is Hamiltons characteristic function for the transformation (q, p) (Q, P ), then
I
W
1
dq = J ()
(16.183)
J =
2
q
C
is a function only of the { } and not the { }. We then invert this relation to obtain
(J), to finally obtain
X
F2 (q, J) = W q, (J) =
W q , (J) .
(16.184)
Thus, the recipe for canonically transforming to action-angle variable is as follows:
(1) Separate and solve the Hamilton-Jacobi equation for W (q, ) =
(2) Find the orbits C the level sets of satisfying H (q , p ) = .
H W
1
(3) Invert the relation J () = 2
q dq to obtain (J).
W (q , ).
(4) F2 (q, J) =
16.8.4
W q , (J) is the desired type-II generator7 .
Example : Harmonic Oscillator
The Hamiltonian is
H=
p2
+ 1 m02 q 2 ,
2m 2
(16.185)
+ m2 02 q 2 = 2m .
(16.186)
hence the Hamilton-Jacobi equation is
Thus,
2
q
dW
p=
= 2m m2 02 q 2 .
dq
We now define
q
2
m02
in which case
1
J=
2
7
dW
dq
1/2
sin
1 2
p dq =
2 0
p=
(16.187)
2m cos ,
(16.188)
Z2
.
d cos2 =
0
(16.189)
= 0, but rather to H
= H(J).
Note that F2 (q, J) is time-independent. I.e. we are not transforming to H
16.8. ACTION-ANGLE VARIABLES
Solving the HJE, we write
361
q dW
dW
=
= 2J cos2 .
d
dq
(16.190)
W = J + 12 J sin 2 ,
(16.191)
Integrating,
up to an irrelevant constant. We then have
W
1
.
=
= + 2 sin 2 + J 1 + cos 2
J q
J q
p
To find (/J)q , we differentiate q = 2J/m0 sin :
r
sin
1
2J
dq =
dJ +
tan .
cos d
=
m
J
2J
2m0 J
0
q
(16.192)
(16.193)
Plugging this result into eqn. 16.192, we obtain = . Thus, the full transformation is
p
2J 1/2
sin , p = 2m0 J cos .
q=
(16.194)
m0
The Hamiltonian is
H = 0 J ,
hence =
16.8.5
H
J
(16.195)
= 0 and J = H
= 0, with solution (t) = (0) + 0 t and J(t) = J(0).
Example : Particle in a Box
Consider a particle in an open box of dimensions Lx Ly moving under the influence of
gravity. The bottom of the box lies at z = 0. The Hamiltonian is
H=
p2y
p2
p2x
+
+ z + mgz .
2m 2m 2m
(16.196)
Step one is to solve the Hamilton-Jacobi equation via separation of variables. The HamiltonJacobi equation is written
1 Wx 2
1 Wy 2
1 Wz 2
+
+
+ mgz = E z .
(16.197)
2m x
2m y
2m z
We can solve for Wx,y by inspection:
p
Wx (x) = 2mx x
We then have8
Wy (y) =
2my y .
q
Wz (z) = 2m z x y mgz
3/2
2 2
Wz (z) =
.
z x y mgz
3 mg
Our choice of signs in taking the square roots for Wx , Wy , and Wz is discussed below.
(16.198)
(16.199)
(16.200)
362
CHAPTER 16. HAMILTONIAN MECHANICS
Figure 16.3: The librations Cz and Cx . Not shown is Cy , which is of the same shape as Cx .
p
Step two is to find the C . Clearly px,y = 2mx,y . For fixed px , the x motion proceeds
from x = 0 to x = Lx and back, with corresponding motion for y. For x, we have
q
pz (z) = Wz (z) = 2m z x y mgz ,
(16.201)
and thus Cz is a truncated parabola, with zmax = (z x y )/mg.
Step three is to compute J() and invert to obtain (J). We have
1
Jx =
2
1
px dx =
1
py dy =
(16.202)
ZLy
p
Ly p
dy 2my =
2my
(16.203)
Cx
1
Jy =
2
ZLx p
Lx p
dx 2mx =
2mx
Cy
and
1
Jz =
2
Cz
zZmax
q
1
dx 2m z x y mgz
pz dz =
3/2
2 2
=
z x y
.
3 m g
(16.204)
We now invert to obtain
2
2
2
J
,
=
J2
y
2mL2x x
2mL2y y
2
2
3 m g 2/3 2/3
2
J
+
J2 .
Jz +
z =
x
2
2 y
2mL
2mL
2 2
x
y
x =
(16.205)
(16.206)
16.8. ACTION-ANGLE VARIABLES
F2 x, y, z, Jx , Jy , Jz
363
x
y
2m2/3 g1/3 z 3/2
2/3
=
.
J +
J + Jz
Lx x Ly y
(3)2/3
We now find
x =
F2
x
=
Jx
Lx
and
F2
=
z =
Jz
2m2/3 g1/3 z
(3Jz )
where
zmax (Jz ) =
The momenta are
px =
Jx
F2
=
x
Lx
and
F2
pz =
= 2m
z
y =
2/3
F2
y
=
Jy
Ly
(16.208)
z
,
zmax
(3Jz )2/3
.
2m2/3 g1/3
,
py =
(16.207)
F2
Jy
=
y
Ly
!1/2
3 m g 2/3 2/3
Jz mgz
.
2 2
(16.209)
(16.210)
(16.211)
(16.212)
We note that the angle variables x,y,z seem to be restricted to the range [0, ], which
seems to be at odds with eqn. 16.180. Similarly, the momenta px,y,z all seem to be positive,
whereas we know the momenta reverse sign when the particle bounces off a wall. The origin
of the apparent discrepancy is that when we solved for the functions Wx,y,z , we had to take
a square root in
each case, and we chose a particular branch of the square root. So rather
than Wx (x) = 2mx x, we should have taken
(
2mx x
if px > 0
Wx (x) =
2mx (2Lx x) if px < 0 .
(16.213)
The relation Jx = (Lx /) 2mx is unchanged, hence
(
(x/Lx ) Jx
Wx (x) =
2Jx (x/Lx ) Jx
and
x =
x/Lx
(2Lx x)/Lx
if px > 0
if px < 0 .
if px > 0
if px < 0 .
(16.214)
(16.215)
Now the angle variable x advances by 2 during the cycle Cx . Similar considerations apply
to the y and z sectors.
364
CHAPTER 16. HAMILTONIAN MECHANICS
16.8.6
Kepler Problem in Action-Angle Variables
This is discussed in detail in standard texts, such as Goldstein. The potential is V (r) =
k/r, and the problem is separable. We write9
W (r, , ) = Wr (r) + W () + W () ,
(16.216)
hence
W 2
1 Wr 2
W 2
1
1
+
+
+ V (r) = E r .
2m r
2mr 2
2mr 2 sin2
(16.217)
Separating, we have
1 dW 2
=
2m d
J =
p
dW
= 2 2m .
d
(16.218)
Next we deal with the coordinate:
1 dW 2
=
2m d
sin2
Z0 q
p
J = 4 2m d 1 / csc2
= 2 2m
p
,
(16.219)
where 0 = sin1 ( / ). Finally, we have10
k
1 dWr 2
=E+ 2
2m dr
r
r
s
I
k
Jr = dr 2m E + 2
r
r
Cr
s
2m
= (J + J ) + k
,
|E|
(16.220)
where weve assumed E < 0, i.e. bound motion.
Thus, we find
H =E=
2 2 mk2
Jr + J + J
Note that the frequencies are completely degenerate:
r,,
9
10
2 .
4 2 mk2
H
=
=
3 =
Jr,,
Jr + J + J
(16.221)
2 mk2
2|E|3
!1/2
We denote the azimuthal angle by to distinguish it from the AA variable .
The details of performing the integral around Cr are discussed in e.g. Goldstein.
(16.222)
16.8. ACTION-ANGLE VARIABLES
365
This threefold degeneracy may be removed by a transformation to new AA variables,
n
o
n
o
(r , Jr ), ( , J ), ( , J ) (1 , J1 ), (2 , J2 ), (3 , J3 ) ,
(16.223)
using the type-II generator
F2 (r , , ; J1 , J2 , J3 ) = ( ) J1 + ( r ) J2 + r J3 ,
(16.224)
which results in
F2
=
J1
F2
= r
2 =
J2
F2
= r
3 =
J3
1 =
F2
= J3 J2
r
F2
= J2 J1
J =
F2
J =
= J1 .
(16.226)
2 2 mk2
,
J32
(16.228)
Jr =
The new Hamiltonian is
H(J1 , J2 , J3 ) =
(16.225)
(16.227)
whence 1 = 2 = 0 and 3 = .
16.8.7
Charged Particle in a Magnetic Field
For the case of the charged particle in a magnetic field, studied above in section 16.7.7, we
found
c p
cP2
2mP1 sin
(16.229)
+
x=
eB
eB
and
p
px = 2mP1 cos
,
py = P2 .
(16.230)
The action variable J is then
J=
Z2
2mcP1
mcP1
px dx =
.
d cos2 =
eB
eB
(16.231)
W = J + 21 J sin(2) + P y ,
(16.232)
We then have
where P P2 . Thus,
=
W
J
= + 12 sin(2) + J 1 + cos(2)
J
tan
2
1
= + 2 sin(2) + 2J cos
2J
=.
(16.233)
366
CHAPTER 16. HAMILTONIAN MECHANICS
The other canonical pair is (Q, P ), where
W
Q=
=y
P
2cJ
cos .
eB
(16.234)
Therefore, we have
cP
x=
+
eB
and
px =
2cJ
sin
eB
r
y =Q+
2eBJ
cos
c
2cJ
cos
eB
py = P .
(16.235)
(16.236)
The Hamiltonian is
H=
=
p2x
1
eBx 2
+
py
2m 2m
c
eBJ
eBJ
cos2 +
sin2
mc
mc
= c J ,
(16.237)
where c = eB/mc. The equations of motion are
H
=
= c
J
and
H
Q =
=0
P
H
J =
=0
(16.238)
H
P =
=0.
Q
(16.239)
Thus, Q, P , and J are constants, and (t) = 0 + c t.
16.8.8
Motion on Invariant Tori
The angle variables evolve as
(t) = (J) t + (0) .
(16.240)
Thus, they wind around the invariant torus, specified by {J } at constant rates. In general,
while each executed periodic motion around a circle, the motion of the system as a whole
is not periodic, since the frequencies (J) are not, in general, commensurate. In order for
the motion to be periodic, there must exist a set of integers, {l }, such that
n
X
=1
l (J) = 0 .
(16.241)
16.9. CANONICAL PERTURBATION THEORY
367
This means that the ratio of any two frequencies / must be a rational number. On a
given torus, there are several possible orbits, depending on initial conditions (0). However,
since the frequencies are determined by the action variables, which specify the tori, on a
given torus either all orbits are periodic, or none are.
In terms of the original coordinates q, there are two possibilities:
q (t) =
1 =
or
n =
A ei(t)
()
A1 2 n ei1 1 (t) ein n (t)
(libration)
(16.242)
q (t) = q (t) +
B ei(t)
(rotation) .
(16.243)
For rotations, the variable q (t) increased by q = 2 q . R
16.9
Canonical Perturbation Theory
16.9.1
Canonical Transformations and Perturbation Theory
Suppose we have a Hamiltonian
H(, t) = H0 (, t) + H1 (, t) ,
(16.244)
where is a small dimensionless parameter. Lets implement a type-II transformation,
generated by S(q, P, t):11
S(q, P, t) .
H(Q,
P, t) = H(q, p, t) +
t
Lets expand everything in powers of :
(16.245)
q = Q + q1, + 2 q2, + . . .
(16.246)
p = P + p1, + 2 p2, + . . .
(16.247)
=H
+ H
+ 2 H
+ ...
H
0
1
2
(16.248)
S = q|{zP} + S1 + 2 S2 + . . . .
(16.249)
identity
transformation
Then
Q =
11
S
S1
S2
= q +
+ 2
+ ...
P
P
P
S2 2
S1
+ q2, +
+ ...
= Q + q1, +
P
P
Here, S(q, P, t) is not meant to signify Hamiltons principal function.
(16.250)
368
CHAPTER 16. HAMILTONIAN MECHANICS
and
p =
S1
S2
S
= P +
+ 2
+ ...
q
q
q
= P + p1, + 2 p2, + . . . .
(16.251)
(16.252)
We therefore conclude, order by order in ,
qk, =
Sk
P
pk, = +
Sk
.
q
(16.253)
Now lets expand the Hamiltonian:
S
H(Q,
P, t) = H0 (q, p, t) + H1 (q, p, t) +
t
H0
H0
= H0 (Q, P, t) +
(q Q ) +
(p P )
Q
P
+ H1 (Q, P, t) + S1 (Q, P, t) + O(2 )
t
(16.254)
!
H0 S1
S1
H0 S1
+ H1 + O(2 )
+
+
= H0 (Q, P, t) +
Q P
P Q
t
S1
+ O(2 ) .
(16.255)
= H0 (Q, P, t) + H1 + S1 , H0 +
t
In the above expression, we evaluate Hk (q, p, t) and Sk (q, P, t) at q = Q and p = P and
expand in the differences q Q and p P . Thus, we have derived the relation
(Q, P, t) + H
(Q, P, t) + . . .
H(Q,
P, t) = H
0
1
(16.256)
(Q, P, t) = H (Q, P, t)
H
0
0
(Q, P, t) = H + S , H + S1 .
H
1
1
1
0
t
(16.257)
with
(16.258)
The problem, though, is this: we have one equation, eqn, 16.258, for the two unknowns H
1
= 0, which
and S1 . Thus, the problem is underdetermined. Of course, we could choose H
1
basically recapitulates standard Hamilton-Jacobi theory. But we might just as well demand
satisfy some other requirement, such as that H
+ H
being integrable.
that H
1
0
1
Incidentally, this treatment is paralleled by one in quantum mechanics, where a unitary
transformation may be implemented to eliminate a perturbation to lowest order in a small
parameter. Consider the Schr
odinger equation,
i~
= (H0 + H1 ) ,
t
(16.259)
16.9. CANONICAL PERTURBATION THEORY
369
and define by
eiS/~ ,
(16.260)
S = S 1 + 2 S 2 + . . . .
(16.261)
with
As before, the transformation U exp(iS/~) collapses to the identity in the 0 limit.
Now lets write the Schr
odinger equation for . Expanding in powers of , one finds
S1
1
+ ... H
= H0 + H1 +
S ,H +
(16.262)
i~
t
i~ 1 0
t
where [A, B] = AB BA is the commutator. Note the classical-quantum correspondence,
{A, B}
1
[A, B] .
i~
(16.263)
Again, what should we choose for S1 ? Usually the choice is made to make the O() term
vanish. But this is not the only possible simplifying choice.
in H
16.9.2
Canonical Perturbation Theory for n = 1 Systems
Henceforth we shall assume H(, t) = H() is time-independent, and we write the perturbed
Hamiltonian as
H() = H0 () + H1 () .
(16.264)
Let (0 , J0 ) be the action-angle variables for H0 . Then
We define
( , J ) = H q( , J ), p( , J ) = H
(J ) .
H
0 0 0
0
0 0
0 0
0 0
( , J ) = H q( , J ), p( , J ) .
H
0 0
0 0
1 0 0
1
(16.265)
(16.266)
= H
+ H
is integrable12 , so it, too, possesses action-angle variWe assume that H
0
1
ables, which we denote by (, J)13 . Thus, there must be a canonical transformation taking
(0 , J0 ) (, J), with
(, J), J (, J) K(J) = E(J) .
H
(16.267)
0
0
We solve via a type-II canonical transformation:
S(0 , J) = 0 J + S1 (0 , J) + 2 S2 (0 , J) + . . . ,
(16.268)
where 0 J is the identity transformation. Then
S
S1
S2
=J +
+ 2
+ ...
0
0
0
S2
S1
S
= 0 +
+ 2
+ ... ,
=
J
J
J
J0 =
12
13
This is always true, in fact, for n = 1.
We assume the motion is bounded, so action-angle variables may be used.
(16.269)
(16.270)
370
CHAPTER 16. HAMILTONIAN MECHANICS
and
E(J) = E0 (J) + E1 (J) + 2 E2 (J) + . . .
( , J ) + H
( , J ) .
=H
0
(16.271)
(16.272)
We now expand H(
0 , J0 ) in powers of J0 J:
H(
0 , J0 ) = H0 (0 , J0 ) + H1 (0 , J0 )
(J) + H0 (J J) +
=H
0
0
J
(16.273)
2H
0
(J J)2 + . . .
J 2 0
( , J ) + H1 (J J) + . . .
+ H
1 0 0
0
J
S1
H
0
(16.274)
= H0 (J) + H1 (0 , J0 ) +
J 0
!
S2 1 2H
S1
S1 2 H
H
0
1
0
+
+
+
2 + . . . .
J 0
2 J 2 0
J 0
1
2
Equating terms, then,
(J)
E0 (J) = H
0
(16.275)
( , J) + H0 S1
E1 (J) = H
1 0
J 0
S1
S1 2 H
S2 1 2H
H
1
0
0
+
E2 (J) =
+
.
J 0 2 J 2 0
J 0
(16.276)
(16.277)
How, one might ask, can we be sure that the LHS of each equation in the above hierarchy
depends only on J when each RHS seems to depend on 0 as well? The answer is that we
use the freedom to choose each Sk to make this so. We demand each RHS be independent
of 0 , which means it must be equal to its average, h RHS(0 ) i, where
f 0
Z2
d0
f 0 .
=
2
(16.278)
The average is performed at fixed J and not at fixed J0 . In this regard, we note that holding
J constant and increasing 0 by 2 also returns us to the same starting point. Therefore,
J is a periodic function of 0 . We must then be able to write
Sk (0 , J) =
Sk (J; m) eim0
(16.279)
m=
for each k > 0, in which case
Sk
0
1
Sk (2) Sk (0) = 0 .
2
(16.280)
16.9. CANONICAL PERTURBATION THEORY
371
(J) is
Lets see how this averaging works to the first two orders of the hierarchy. Since H
0
independent of 0 and since S1 /0 is periodic, we have
this vanishes!
( , J) + H0
E1 (J) = H
1 0
J
and hence S1 must satisfy
z }| {
S1
0
(16.281)
H
S1
1
1
,
=
0
0 (J)
(16.282)
/J. Clearly the RHS of eqn. 16.282 has zero average, and must be a
where 0 (J) = H
0
periodic function of 0 . The solution is S1 = S1 (0 , J) + g(J), where g(J) is an arbitrary
function of J. However, g(J) affects only the difference 0 , changing it by a constant
value g (J). So there is no harm in taking g(J) = 0.
Next, lets go to second order in . We have
this vanishes!
z }| {
S1 2
S2
H1 S1
1 0
+2
+ 0 (J)
.
E2 (J) =
J 0
J
1
0
The equation for S2 is then
(
H
H
1
H
S2
H
1
1
1
1
= 2
H0
H0
H1
H1 +
0
J
J
J
J
0 (J)
)
1 ln 0
2
H
12
+
H1 2 H
+2 H
.
1
1
2 J
(16.283)
(16.284)
The expansion for the energy E(J) is then
(J) + H
E(J) = H
0
1
2
+
0 (J)
(
H
1
J
H
1
H1
H1
J
)
1 ln 0
2
2
+
+ O(3 ) .
H1 H1
2 J
(16.285)
Note that we dont need S to find E(J)! The perturbed frequencies are
(J) =
E
.
J
(16.286)
Sometimes the frequencies are all that is desired. However, we can of course obtain the full
motion of the system via the succession of canonical transformations,
(, J) (0 , J0 ) (q, p) .
(16.287)
372
CHAPTER 16. HAMILTONIAN MECHANICS
Figure 16.4: Action-angle variables for the harmonic oscillator.
16.9.3
Example : Nonlinear Oscillator
Consider the nonlinear oscillator with Hamiltonian
H0
z
}|
{
p2
2
H(q, p) =
+ 1 m q 2 + 14 q 4 .
2m 2 0
(16.288)
The action-angle variables for the harmonic oscillator Hamiltonian H0 are
0 = tan1 mvq/p)
J0 =
p2
+ 1 m q 2 ,
2m0 2 0
(16.289)
and the relation between (0 , J0 ) and (q, p) is further depicted in fig. 16.4. Note H0 = 0 J0 .
For the full Hamiltonian, we have
r
4
2J0
1
H(0 , J0 ) = 0 J0 + 4
sin 0
m0
= 0 J0 + 2 2 J02 sin4 0 .
(16.290)
m 0
We may now evaluate
E1 (J) = H
1
J 2
= 2 2
m 0
Z2
3J 2
d0
sin4 0 =
.
2
8m2 02
(16.291)
The frequency, to order , is
(J) = 0 +
3 J
.
4m2 02
(16.292)
Now to lowest order in , we may replace J by J0 = 12 m0 A2 , where A is the amplitude of
the q motion. Thus,
3
.
(16.293)
(A) = 0 +
8m0
16.9. CANONICAL PERTURBATION THEORY
373
This result agrees with that obtained via heavier lifting, using the Poincare-Lindstedt
method.
Next, lets evaluate the canonical transformation (0 , J0 ) (, J). We have
S1
J 2
0
= 2 2 83 sin4 0
0
m 0
J 2
3 + 2 sin2 0 sin 0 cos 0 + O(2 ) .
S(0 , J) = 0 J +
3
2
8m 0
(16.294)
Thus,
S
J
= 0 +
3 + 2 sin2 0 sin 0 cos 0 + O(2 )
3
2
J
4m 0
J 2
S
=J+
J0 =
+ O(2 ) .
4
cos
2
cos
4
0
0
23
8m
0
0
=
(16.295)
(16.296)
Again, to lowest order, we may replace J by J0 in the above, whence
J02
4
cos
2
cos
4
+ O(2 )
0
0
3
8m2 0
J0
= 0 +
3 + 2 sin2 0 sin 20 + O(2 ) .
3
2
8m 0
J = J0
(16.297)
(16.298)
To obtain (q, p) in terms of (, J) is not analytically tractable the relations cannot be
analytically inverted.
16.9.4
n > 1 Systems : Degeneracies and Resonances
Generalizing the procedure we derived for n = 1, we obtain
S
S2
S1
= J + + 2
+ ...
0
0
0
S1
S2
S
= 0 +
+ 2
+ ...
=
J
J
J
J0 =
(16.299)
(16.300)
and
(J)
E0 (J) = H
0
(16.301)
( , J) + (J) S1
E1 (J) = H
0
0
0
0
S2
1 0 S1 S1
H
S1
0
+
.
E2 (J) =
+ 0
J 0
2 J 0 0
0
(16.302)
(16.303)
We now implement the averaging procedure, with
f (J , . . . , J ) =
1
Z2 1
Z2 n
d0
d0
f 10 , . . . , n0 , J 1 , . . . , J n .
2
2
0
(16.304)
374
CHAPTER 16. HAMILTONIAN MECHANICS
The equation for S1 is
0
S1
H
= H
V ei ,
1
1
(16.305)
where = {1 , 2 , . . . , n }, with each an integer, and with 6= 0. The solution is
S1 (0 , J) = i
X V
ei .
0
(16.306)
where 0 = l 0 . When two or more of the frequencies (J) are commensurate, there
exists a set of integers l such that the denominator of D(l) vanishes. But even when the
frequencies are not rationally related, one can approximate the ratios 0 /0 by rational
numbers, and for large enough l the denominator can become arbitrarily small.
Periodic time-dependent perturbations present a similar problem. Consider the system
H(, J, t) = H0 (J) + V (, J, t) ,
where V (t + T ) = V (t). This means we may write
X
V (, J, t) =
Vk (, J) eikt
(16.307)
(16.308)
XX
k
Vk,(J) ei eikt .
(16.309)
by Fourier transforming from both time and angle variables; here = 2/T . Note that
=V
V (, J, t) is real if Vk,
k,l . The equations of motion are
H
J = = i
l Vk,(J) ei eikt
(16.310)
k,
H
= + = 0 (J) +
J
X Vk,(J)
J
k,
ei eikt .
(16.311)
We now expand in :
= 0 + 1 + 2 2 + . . .
J =
J0
J1
J2
+ ... .
To order 0 , J = J0 and 0 = 0 t + 0 . To order 1 ,
X
J1 = i
l Vk,(J0 ) ei(0 k)t ei0
(16.312)
(16.313)
(16.314)
k,l
and
X Vk,(J)
ei(0 k)t ei0 ,
1 = 0 J1 +
J
J
k,
(16.315)
16.9. CANONICAL PERTURBATION THEORY
375
where derivatives are evaluated at J = J0 . The solution is:
X l Vk,(J0 )
ei(0 k)t ei0
k 0
k,
)
(
Vk,(J)
1
0 l Vk,(J0 )
+
ei(0 k)t ei0 .
1 =
J (k 0 )2
J k 0
J1 =
(16.316)
(16.317)
When the resonance condition,
k = 0 (J0 ) ,
(16.318)
holds, the denominators vanish, and the perturbation theory breaks down.
16.9.5
Particle-Wave Interaction
Consider a particle of charge e moving in the presence of a constant magnetic field B = B z
and a space- and time-varying electric field E(x, t), described by the Hamiltonian
2
1
(16.319)
H=
p ec A + eV0 cos(k x + kz z t) ,
2m
from our
where is a dimensionless expansion parameter. Working in the gauge A = Bxy,
earlier discussions in section 16.7.7, we may write
r
2J
p2z
k P
H = c J +
+ eV0 cos kz z +
+ k
sin t .
(16.320)
2m
mc
mc
Here,
r
r
P
2J
2J
+
sin
,
y =Q+
cos ,
(16.321)
x=
mc
mc
mc
with c = eB/mc, the cyclotron frequency. We now make a mixed canonical transformation,
generated by
k P
t K P Q ,
(16.322)
F = J + kz z +
mc
where the new sets of conjugate variables are ( , J ) , (Q , P ) , ( , K ) . We then have
=
F
=
J
Q=
=
k K
F
=
+ Q
P
mc
F
k P
= kz z +
t
K
mc
J=
F
= J
F
=P
Q
(16.324)
F
= kz K .
z
(16.325)
P =
pz =
(16.323)
The transformed Hamiltonian is
F
H = H +
t
r
2J
kz2 2
K K + eV0 cos + k
sin .
= c J +
2m
mc
(16.326)
376
CHAPTER 16. HAMILTONIAN MECHANICS
We will now drop primes and simply write H = H0 + H1 , with
kz2 2
K K
2m
r
2J
H1 = eV0 cos + k
sin .
mc
H0 = c J +
(16.327)
(16.328)
When = 0, the frequencies associated with the and motion are
0 =
H0
= c
0 =
H0
k2 K
= z = kz vz ,
(16.329)
where vz = pz /m is the z-component of the particles velocity. Now let us solve eqn. 16.305:
0
S1
S1
+ 0
= h H1 i H1 .
(16.330)
This yields
S1
+
c
r
kz2 K
S1
2J
= eA0 cos + k
sin
m
mc
!
r
X
2J
Jn k
= eA0
cos( + n) ,
mc
n=
(16.331)
where we have used the result
eiz sin =
Jn (z) ein .
(16.332)
n=
The solution for S1 is
S1 =
X
n
eV0
Jn k
nc kz2 K/m
2J
mc
sin( + n) .
(16.333)
where
We then have new action variables J and K,
S1
+ O(2 )
J = J +
+ S1 + O(2 ) .
K=K
(16.334)
(16.335)
Defining the dimensionless variable
k
2J
,
mc
(16.336)
we obtain the result
X nJn () cos( + n)
mc2
mc2
2
+ O(2 ) ,
=
2
2
2
kz2 K
2eV0 k
2eV0 k
n
n
c
mc
(16.337)
16.9. CANONICAL PERTURBATION THEORY
377
Figure 16.5: Plot of versus for = 0 (Poincare section) for = 30.11 c Top panels are
nonresonant invariant curves calculated to first order. Bottom panels are exact numerical
dynamics, with x symbols marking the initial conditions. Left panels: weak amplitude
(no trapping). Right panels: stronger amplitude (shows trapping). From Lichtenberg and
Lieberman (1983).
=k
where
2J/mc .14
We see that resonances occur whenever
k2 K
z =n,
c mc
(16.338)
for any integer n. Let us consider the case kz = 0, in which the resonance condition is
= nc . We then have
2
2 X n Jn () cos( + n)
2
2
c
n
where
=
E0 ck
B c
(16.339)
(16.340)
14
This arises because we are computing the
Note that the argument of Jn in eqn. 16.337 is and not .
new action J in terms of the old variables (, J) and (, K).
378
CHAPTER 16. HAMILTONIAN MECHANICS
is a dimensionless measure of the strength of the perturbation, with E0 k V0 . In Fig.
16.5 we plot the level sets for the RHS of the above equation () for = 0, for two different
values of the dimensionless amplitude , for /c = 30.11 (i.e. off resonance). Thus, when
the amplitude is small, the level sets are far from a primary resonance, and the analytical and
numerical results are very similar (left panels). When the amplitude is larger, resonances
may occur which are not found in the lowest order perturbation treatment. However, as
is apparent from the plots, the gross features of the phase diagram are reproduced by
perturbation theory. What is missing is the existence of chaotic islands which initially
emerge in the vicinity of the trapping regions.
16.10
Adiabatic Invariants
Adiabatic perturbations are slow, smooth, time-dependent perturbations to a dynamical
system. A classic example: a pendulum with a slowly varying length l(t). Suppose (t)
is the adiabatic parameter. We write H = H q, p; (t) . All explicit time-dependence to
H comes through (t). Typically, a dimensionless parameter may be associated with the
perturbation:
1 d ln
,
(16.341)
=
0 dt
where 0 is the natural frequency of the system when is constant. We require 1 for
adiabaticity. In adiabatic processes, the action variables are conserved to a high degree of
accuracy. These are the adiabatic invariants. For example, for the harmonix oscillator, the
action is J = E/. While E and may vary considerably during the adiabatic process,
their ratio is very nearly fixed. As a consequence, assuming small oscillations,
E = J = 21 mgl 02
0 (l)
2J
3/2 ,
m gl
(16.342)
so 0 () l3/4 .
Suppose that for fixed the Hamiltonian is transformed to action-angle variables via the
generator S(q, J; ). The transformed Hamiltonian is
S d
,
H(,
J, t) = H(, J; ) +
dt
(16.343)
H(, J; ) = H q(, J; ), p(, J; ); ) .
(16.344)
where
We assume n = 1 here. Hamiltons equations are now
H
2S d
= +
= (J; ) +
J
J dt
2S d
H
=
.
J =
dt
(16.345)
(16.346)
16.10. ADIABATIC INVARIANTS
379
Figure 16.6: A mechanical mirror.
The second of these may be Fourier decomposed as
J = i
Sm (J; ) im
e
,
(16.347)
hence
J = J(t = +) J(t = ) = i
X
m
Z
Sm (J; ) d im
m dt
e
.
dt
(16.348)
Since is small, we have (t) = t + , to lowest order. We must therefore evaluate
integrals such as
Z
Sm (J; ) d imt
e
.
(16.349)
Im = dt
dt
The term in curly brackets is a smooth, slowly varying function of t. Call it f (t). We
presume f (t) can be analytically continued off the real t axis, and that its closest singularity
in the complex t plane lies at t = i , in which case I behaves as exp(|m| ). Consider,
for example, the Lorentzian,
C
f (t) =
1 + (t/ )2
Z
dt f (t) eimt = e|m| ,
(16.350)
which is exponentially small in the time scale . Because of this, only m = 1 need be
considered. What this tells us is that the change J may be made arbitrarily small by a
sufficiently slowly varying (t).
16.10.1
Example: mechanical mirror
Consider a two-dimensional version of a mechanical mirror, depicted in fig. 16.6. A particle
bounces between two curves, y = D(x), where |D (x)| << 1. The bounce time is b =
2D/vy . We assume L/vx , where vx,y are the components of the particles velocity, and
L is the total length of the system. There are, therefore, many bounces, which means the
particle gets to sample the curvature in D(x).
380
CHAPTER 16. HAMILTONIAN MECHANICS
The adiabatic invariant is the action,
1
J=
2
ZD
ZD
1
2
dy m vy +
dy m (vy ) = mvy D(x) .
2
(16.351)
Thus,
E = 12 m vx2 + vy2 ) = 21 mvx2 +
or
2 J 2
,
8mD2 (x)
2
.
(16.352)
J
2E
(16.353)
=
m
2mD(x)
The particle is reflected in the throat of the device at horizontal coordinate x , where
vx2
D(x ) =
16.10.2
J
.
8mE
(16.354)
Example: magnetic mirror
Consider a particle of charge e moving in the presence of a uniform magnetic field B = B z.
Recall the basic physics: velocity in the parallel direction vz is conserved, while in the plane
perpendicular to B the particle executes circular cyclotron orbits, satisfying
2
mv
e
= v B
mcv
,
eB
(16.355)
where is the radial coordinate in the plane perpendicular to B. The period of the orbits
is T = 2.v = 2mc/eB, hence their frequency is the cyclotron frequency c = eB/mc.
Now assume that the magnetic field is spatially dependent. Note that a spatially varying
B-field cannot be unidirectional:
Bz
=0.
(16.356)
B = B +
z
The non-collinear nature of B results in the drift of the cyclotron orbits. Nevertheless, if
the field B felt by the particle varies slowly on the time scale T = 2/c , then the system
possesses an adiabatic invariant:
I
I
1
1
J=
(16.357)
p d =
mv + ec A d
2
2
C
C
I
I
m
e
d .
=
v d +
Bn
(16.358)
2
2c
C
int(C)
The last two terms are of opposite sign, and one has
e
m eBz
2 +
Bz 2
2 mc
2c
2c
m2 v
e
eBz 2
=
B (C) =
,
=
2c
2c
2eBz
J =
(16.359)
(16.360)
16.10. ADIABATIC INVARIANTS
381
Figure 16.7: B field lines in a magnetic bottle.
where B (C) is the magnetic flux enclosed by C.
The energy is
2
E = 12 mv
+ 12 mvz2 ,
hence we have
vz =
where
M
2
E MB .
m
e2
e
J=
B (C)
mc
2mc2
(16.361)
(16.362)
(16.363)
is the magnetic moment. Note that vz vanishes when B = Bmax = E/M . When this limit
is reached, the particle turns around. This is a magnetic mirror . A pair of magnetic mirrors
may be used to confine charged particles in a magnetic bottle, depicted in fig. 16.7.
Let vk,0 , v,0 , and Bk,0 be the longitudinal particle velocity, transverse particle velocity,
and longitudinal component of the magnetic field, respectively, at the point of injection.
Our two conservation laws (J and E) guarantee
2
2
2
(z) = vk,0
+ v,0
vk2 (z) + v
2
v,0
v (z)2
=
.
Bk (z)
Bk,0
(16.364)
(16.365)
This leads to reflection at a longitudinal coordinate z , where
v
u
2
u
vk,0
Bk (z ) = Bk,0 t1 + 2 .
v,0
The physics is quite similar to that of the mechanical mirror.
(16.366)
382
16.10.3
CHAPTER 16. HAMILTONIAN MECHANICS
Resonances
When n > 1, we have
J = i
J = i
X
m
Sm (J; ) im
e
Z
Sm (J; ) d imt im
dt
m
e
e
.
dt
(16.367)
(16.368)
Therefore, when m (J) = 0 we have a resonance, and the integral grows linearly with
time a violation of the adiabatic invariance of J .
16.11
Appendix : Canonical Perturbation Theory
Consider the Hamiltonian
H=
q3
p2
+ 21 m 02 q 2 + 13 m 02
,
2m
a
where is a small dimensionless parameter.
(a) Show that the oscillation frequency satisfies (J) = 0 + O(2 ). That is, show that the
first order (in ) frequency shift vanishes.
Solution: It is good to recall the basic formulae
r
p
2J0
q=
sin 0
,
p = 2m 0 J0 cos 0
m0
(16.369)
as well as the results
J0 =
=
S
S2
S1
=J +
+ 2
+ ...
0
0
0
(16.370)
S
S2
S1
= 0 +
+ 2
+ ... ,
J
J
J
(16.371)
and
(J)
E0 (J) = H
0
( , J) + H0 S1
E1 (J) = H
1 0
J 0
S1 2 H
S1
S2 1 2H
H
0
1
0
+
+
.
E2 (J) =
J 0 2 J 2 0
J 0
(16.372)
(16.373)
(16.374)
16.11. APPENDIX : CANONICAL PERTURBATION THEORY
383
Expressed in action-angle variables,
( , J) = J
H
0 0
0
r
2 20 3/2
H1 (0 , J) =
J
sin3 0 .
3 ma2
Thus, 0 =
H
0
J
(16.375)
(16.376)
= 0 .
Averaging the equation for E1 (J) yields
( , J) = 2
E1 (J) = H
1 0
3
20 3/2
3
J
sin 0 = 0 .
ma2
(16.377)
(b) Compute the frequency shift (J) to second order in .
Solution : From the equation for E1 , we also obtain
S1
1
H1 H1 .
=
0
0
Inserting this into the equation for E2 (J) and averaging then yields
1 H
1 H
1
1
H1 H1
H1
=
E2 (J) =
0 J
0
J
40 J 2
6
=
sin 0
3ma2
(16.378)
(16.379)
(16.380)
In computing the average of sin6 0 , it is good to recall the binomial theorem, or the Fibonacci tree. The sixth order coefficents are easily found to be {1, 6, 15, 20, 15, 6, 1}, whence
sin6 0 =
=
Thus,
6
1
ei0 ei0
6
(2i)
1
64
(16.381)
2 sin 60 + 12 sin 40 30 sin 20 + 20 .
whence
sin6 0 =
E(J) = 0 J
and
(J) =
5
16
5 2
12
(16.382)
(16.383)
J2
ma2
E
J
= 0 65 2
.
J
ma2
(16.384)
(16.385)
(c) Find q(t) to order . Your result should be finite for all times.
Solution : From the equation for E1 (J), we have
s
2
2J 3
S1
=
sin3 0 .
0
3 m0 a2
(16.386)
384
CHAPTER 16. HAMILTONIAN MECHANICS
Integrating, we obtain
2
S1 (0 , J) =
3
2J 3
cos 0
m0 a2
J 3/2
cos 0
2m0 a2
Thus, with
1
3
1
9
cos3 0
cos 30 .
S(0 , J) = 0 J + S1 (0 , J) + . . . ,
(16.387)
(16.388)
(16.389)
we have
S
3 J 1/2
1
cos
cos
3
= 0 +
0
0
9
J
2 2m0 a2
S
J 3/2
1
J0 =
sin
sin
3
.
=J
0
0
3
0
2m0 a2
=
(16.390)
(16.391)
Inverting, we may write 0 and J0 in terms of and J:
0 = +
3 J 1/2
2 2m0 a2
J0 = J +
J 3/2
2m0 a2
1
9
1
3
cos 3 cos
sin 3 sin .
(16.392)
(16.393)
Thus,
q(t) =
=
=
with
2J0
sin 0
m0
(16.394)
J
2J
+ ...
sin + cos + . . .
sin 1 +
m0
2J
(16.395)
J
2J
sin
1 + 13 cos 2 + O 2 ,
m0
m0 a
(t) = (0) + (J) t .
(16.396)
(16.397)
Chapter 17
Physics 110A-B Exams
The following pages contain problems and solutions from midterm and final exams in Physics
110A-B.
385
386
17.1
CHAPTER 17. PHYSICS 110A-B EXAMS
F05 Physics 110A Midterm #1
[1] A particle of mass m moves in the one-dimensional potential
U (x) = U0
x2 x/a
e
.
a2
(17.1)
(a) Sketch U (x). Identify the location(s) of any local minima and/or maxima, and be sure
that your sketch shows the proper behavior as x .
(b) Sketch a representative set of phase curves. Identify and classify any and all fixed points.
Find the energy of each and every separatrix.
(c) Sketch all the phase curves for motions with total energy E =
E = U0 . (Recall that e = 2.71828 . . . .)
2
5
U0 . Do the same for
(d) Derive and expression for the period T of the motion when |x| a.
Solution:
(a) Clearly U (x) diverges to + for x , and U (x) 0 for x +. Setting
U (x) = 0, we obtain the equation
x2
U0
U (x) = 2 2x
ex/a = 0 ,
(17.2)
a
a
with (finite x) solutions at x = 0 and x = 2a. Clearly x = 0 is a local minimum and x = 2a
a local maximum. Note U (0) = 0 and U (2a) = 4 e2 U0 0.541 U0 .
Figure 17.1: The potential U (x). Distances are here measured in units of a, and the
potential in units of U0 .
(b) Local minima of a potential U (x) give rise to centers in the (x, v) plane, while local
maxima give rise to saddles. In Fig. 17.2 we sketch the phase curves. There is a center at
17.1. F05 PHYSICS 110A MIDTERM #1
387
Figure 17.2: Phase curves for the potential U (x). The red curves show phase curves for
E = 25 U0 (interior, disconnected red curves, |v| < 1) and E = U0 (outlying red curve). The
separatrix is the dark blue curve which forms a saddle at (x, v) = (2, 0), and corresponds
to an energy E = 4 e2 U0 .
(0, 0) and a saddle at (2a, 0). There is one separatrix, at energy E = U (2a) = 4 e2 U0
0.541 U0 .
(c) Even without a calculator, it is easy to verify that 4 e2 > 52 . One simple way is to
multiply both sides by 52 e2 to obtain 10 > e2 , which is true since e2 < (2.71828 . . .)2 < 10.
Thus, the energy E = 25 U0 lies below the local maximum value of U (2a), which means that
there are two phase curves with E = 52 U0 .
It is also quite obvious that the second energy value given, E = U0 , lies above U (2a), which
means that there is a single phase curve for this energy. One finds bound motions only for
x < 2 and 0 E < U (2a). The phase curves corresponding to total energy E = 25 U0 and
E = U0 are shown in Fig. 17.2.
(d) Expanding U (x) in a Taylor series about x = 0, we have
x4
U0 2 x3
+ 2 + ... .
U (x) = 2 x
a
a
2a
(17.3)
The leading order term is sufficient for |x| a. The potential energy is then equivalent to
that of a spring, with spring constant k = 2U0 /a2 . The period is
T = 2
m
= 2
k
ma2
2U0
(17.4)
388
CHAPTER 17. PHYSICS 110A-B EXAMS
[2] A forced, damped oscillator obeys the equation
x
+ 2 x + 02 x = f0 cos(0 t) .
(17.5)
You may assume the oscillator is underdamped.
(a) Write down the most general solution of this differential equation.
(b) Your solution should involve two constants. Derive two equations relating these constants to the initial position x(0) and the initial velocity x(0).
You do not have to solve
these equations.
(c) Suppose 0 = 5.0 s1 , = 4.0 s1 , and f0 = 8 cm s2 . Suppose further you are told that
x(0) = 0 and x(T ) = 0, where T = 6 s. Derive an expression for the initial velocity x(0).
Solution: (a) The general solution with forcing f (t) = f0 cos(t) is
x(t) = xh (t) + A() f0 cos t () ,
with
i1/2
h
A() = (02 2 )2 + 4 2 2
and
with =
() = tan
(17.6)
2
2
0 2
(17.7)
xh (t) = C et cos(t) + D et sin(t) ,
(17.8)
1
2 .
Thus, the most general
02 2 .
In our case, = 0 , in which case A = (20 )1 and =
solution is
x(t) = C et cos(t) + D et sin(t) +
f0
sin(0 t)
20
(17.9)
(b) We determine the constants C and D by the boundary conditions on x(0) and x(0):
x(0) = C
x(0)
= C + D +
Thus,
C = x(0)
D=
f0
2
1
f0
x(0) + x(0)
(17.10)
(17.11)
(c) From x(0) = 0 we obtain C = 0. The constant D is then determined by the condition
at time t = T = 16 .
p
Note that = 02 2 = 3.0 s1 . Thus, with T = 61 , we have T = 12 , and
x(T ) = D eT +
f0
sin(0 T ) .
20
(17.12)
17.1. F05 PHYSICS 110A MIDTERM #1
This determines D:
D=
389
f0 T
e sin(0 T ) .
20
(17.13)
We now can write
x(0)
= D +
f0
2
(17.14)
f0
T
=
e sin(0 T )
1
2
0
=
1
3
10
Numerically, the value is x(0)
0.145 cm/s .
e2/3 cm/s
(17.15)
(17.16)
390
17.2
CHAPTER 17. PHYSICS 110A-B EXAMS
F05 Physics 110A Midterm #2
[1] Two blocks connected by a spring of spring constant k are free to slide frictionlessly
along a horizontal surface, as shown in Fig. 17.3. The unstretched length of the spring is a.
Figure 17.3: Two masses connected by a spring sliding horizontally along a frictionless
surface.
(a) Identify a set of generalized coordinates and write the Lagrangian.
[15 points]
Solution : As generalized coordinates I choose X and u, where X is the position of the
right edge of the block of mass M , and X + u + a is the position of the left edge of the
block of mass m, where a is the unstretched length of the spring. Thus, the extension of
the spring is u. The Lagrangian is then
2 21 ku2
L = 21 M X 2 + 12 m(X + u)
= 12 (M + m)X 2 + 21 mu 2 + mX u 12 ku2 .
(17.17)
(b) Find the equations of motion.
[15 points]
Solution : The canonical momenta are
pX
L
= (M + m)X + mu
X
pu
L
= m(X + u)
.
u
(17.18)
The corresponding equations of motion are then
L
X
L
pu = Fu =
u
p X = FX =
+ m
(M + m)X
u=0
+u
m(X
) = ku .
(17.19)
(17.20)
(c) Find all conserved quantities.
[10 points]
Solution : There are two conserved quantities. One is pX itself, as is evident from the
fact that L is cyclic in X. This is the conserved charge associated with the continuous
symmetry X X + . i.e. = pX . The other conserved quantity is the Hamiltonian H,
since L is cyclic in t. Furthermore, because the kinetic energy is homogeneous of degree
two in the generalized velocities, we have that H = E, with
E = T + U = 21 (M + m)X 2 + 21 mu 2 + mX u + 12 ku2 .
(17.21)
17.2. F05 PHYSICS 110A MIDTERM #2
391
using the conservation of :
It is possible to eliminate X,
mu
.
X =
M +m
(17.22)
This allows us to write
E=
M m u 2
2
+
+ 1 ku2 .
2(M + m) 2(M + m) 2
(17.23)
(d) Find a complete solution to the equations of motion. As there are two degrees of
freedom, your solution should involve 4 constants of integration. You need not match initial
conditions, and you need not choose the quantities in part (c) to be among the constants.
[10 points]
in terms of x
Solution : Using conservation of , we may write X
, in which case
Mm
u
= ku
M +m
where
u(t) = A cos(t) + B sin(t) ,
r
(M + m)k
.
Mm
For the X motion, we integrate eqn. 17.22 above, obtaining
m
t
A cos(t) A + B sin(t) .
X(t) = X0 +
M +m M +m
=
(17.24)
(17.25)
(17.26)
There are thus four constants: X0 , , A, and B. Note that conservation of energy says
E=
2
+ 1 k(A2 + B 2 ) .
2(M + m) 2
(17.27)
Alternate solution : We could choose X as the position of the left block and x as the
position of the right block. In this case,
L = 12 M X 2 + 21 mx 2 12 k(x X b)2 .
(17.28)
Here, b includes the unstretched length a of the spring, but may also include the size of
the blocks if, say, X and x are measured relative to the blocks midpoints. The canonical
momenta are
L
L
= mx .
(17.29)
= M X , px =
pX =
x
X
The equations of motion are then
L
X
L
p x = Fx =
x
pX = FX =
= k(x X b)
MX
m
x = k(x X b) .
(17.30)
(17.31)
392
CHAPTER 17. PHYSICS 110A-B EXAMS
The one-parameter family which leaves L invariant is X X + and x x + , i.e.
simultaneous and identical displacement of both of the generalized coordinates. Then
= M X + mx ,
(17.32)
which is simply the x-component of the total momentum. Again, the energy is conserved:
E = 21 M X 2 + 12 mx 2 + 21 k (x X b)2 .
(17.33)
We can combine the equations of motion to yield
Mm
which yields
d2
x X b = k (M + m) x X b ,
2
dt
x(t) X(t) = b + A cos(t) + B sin(t) ,
(17.34)
(17.35)
From the conservation of , we have
M X(t) + m x(t) = t + C ,
(17.36)
were C is another constant. Thus, we have the motion of the system in terms of four
constants: A, B, , and C:
t + C
X(t) = Mm
+m b + A cos(t) + B sin(t) + M + m
x(t) =
M
M +m
t + C
b + A cos(t) + B sin(t) +
.
M +m
(17.37)
(17.38)
17.2. F05 PHYSICS 110A MIDTERM #2
393
[2] A uniformly dense ladder of mass m and length 2 leans against a block of mass M ,
as shown in Fig. 17.4. Choose as generalized coordinates the horizontal position X of the
right end of the block, the angle the ladder makes with respect to the floor, and the
coordinates (x, y) of the ladders center-of-mass. These four generalized coordinates are not
all independent, but instead are related by a certain set of constraints.
Recall that the kinetic energy of the ladder can be written as a sum TCM + Trot , where
TCM = 21 m(x 2 + y 2 ) is the kinetic energy of the center-of-mass motion, and Trot = 12 I 2 ,
where I is the moment of inertial. For a uniformly dense ladder of length 2, I = 13 m2 .
Figure 17.4: A ladder of length 2 leaning against a massive block. All surfaces are frictionless..
(a) Write down the Lagrangian for this system in terms of the coordinates X, , x, y, and
their time derivatives.
[10 points]
Solution : We have L = T U , hence
L = 12 M X 2 + 12 m(x 2 + y 2 ) + 21 I 2 mgy .
(17.39)
(b) Write down all the equations of constraint.
[10 points]
Solution : There are two constraints, corresponding to contact between the ladder and the
block, and contact between the ladder and the horizontal surface:
G1 (X, , x, y) = x cos X = 0
(17.40)
G2 (X, , x, y) = y sin = 0 .
(17.41)
(c) Write down all the equations of motion.
[10 points]
Solution : Two Lagrange multipliers, 1 and 2 , are introduced to effect the constraints.
We have for each generalized coordinate q ,
k
X
Gj
L
d L
j
=
Q ,
(17.42)
dt q
q
q
j=1
394
CHAPTER 17. PHYSICS 110A-B EXAMS
where there are k = 2 constraints. We therefore have
=
MX
1
(17.43)
m
x = +1
(17.44)
m
y = mg + 2
I = sin 1 cos 2 .
(17.45)
(17.46)
These four equations of motion are supplemented by the two constraint equations, yielding
six equations in the six unknowns {X, , x, y, 1 , 2 }.
(d) Find all conserved quantities.
[10 points]
Solution : The Lagrangian and all the constraints are invariant under the transformation
X X +
x x+
yy
(17.47)
The associated conserved charge is
L q
=
= M X + mx .
q =0
(17.48)
Using the first constraint to eliminate x in terms of X and , we may write this as
= (M + m)X m sin .
(17.49)
The second conserved quantity is the total energy E. This follows because the Lagrangian
and all the constraints are independent of t, and because the kinetic energy is homogeneous
of degree two in the generalized velocities. Thus,
E = 21 M X 2 + 12 m(x 2 + y 2 ) + 21 I 2 + mgy
(17.50)
2
2
2
2
(17.51)
+ 1 I + m2 Mm
=
+m m sin + mg sin ,
2(M + m) 2
where the second line is obtained by using the constraint equations to eliminate x and y in
terms of X and .
(e) What is the condition that the ladder detaches from the block? You do not have to solve
for the angle of detachment! Express the detachment condition in terms of any quantities
you find convenient.
[10 points]
Solution : The condition for detachment from the block is simply 1 = 0, i.e. the normal
force vanishes.
Further analysis : It is instructive to work this out in detail (though this level of analysis
was not required for the exam). If we eliminate x and y in terms of X and , we find
x = X + cos
x = X sin
sin cos 2
x
=X
y = sin
(17.52)
y = cos
y = cos sin 2 .
(17.53)
(17.54)
17.2. F05 PHYSICS 110A MIDTERM #2
395
Figure 17.5: Plot of versus 0 for the ladder-block problem (eqn. 17.64). Allowed
solutions, shown in blue, have 1, and thus 0 . Unphysical solutions, with < 1,
are shown in magenta. The line = 0 is shown in red.
We can now write
m sin m cos 2 = M X
,
1 = m
x = mX
which gives
and hence
= m sin + cos 2 ,
(M + m)X
Q x = 1 =
We also have
Mm
sin + cos 2 .
m+m
(17.55)
(17.56)
(17.57)
Qy = 2 = mg + m
y
= mg + m cos sin 2 .
(17.58)
This comes from the last of the equations of
We now need an equation relating and .
396
CHAPTER 17. PHYSICS 110A-B EXAMS
motion:
I = sin 1 cos 2
m 2
= MM+m
sin2 + sin cos 2 mg cos m2 cos2 sin cos 2
= mg cos m2 1
sin2 +
m2 sin cos 2 .
(17.59)
we obtain
Collecting terms proportional to ,
2
2
2
I + m2 Mm
sin
= Mm
+m
+m m sin cos mg cos .
(17.60)
m
M +m
m
M +m
We are now ready to demand Qx = 1 = 0, which entails
cos 2
=
.
sin
Substituting this into eqn. 17.60, we obtain
I + m2 2 = mg sin .
(17.61)
(17.62)
Finally, we substitute this into eqn. 17.51 to obtain an equation for the detachment angle,
m
m2
2
2
(17.63)
= 3
sin 21 mg sin .
E
2(M + m)
M + m I + m2
If our initial conditions are that the system starts from rest1 with an angle of inclination
0 , then the detachment condition becomes
m2
sin 0 = 32 sin 12 Mm
sin3
2
+m
I+m
=
3
2
sin 12 1 sin3 ,
where
M
1+
m
I
.
1+
m2
(17.64)
(17.65)
Note that 1, and that when M/m = 2 , we recover = sin1 32 sin 0 . For finite
, the ladder detaches at a larger value of . A sketch of versus 0 is provided in Fig.
17.5. Note that, provided 1, detachment always occurs for some unique value for
each 0 .
1
2
Rest means that the initial velocities are X = 0 and = 0, and hence = 0 as well.
I must satisfy I m2 .
17.3. F05 PHYSICS 110A FINAL EXAM
17.3
397
F05 Physics 110A Final Exam
[1] Two blocks and three springs are configured as in Fig. 17.6. All motion is horizontal.
When the blocks are at rest, all springs are unstretched.
Figure 17.6: A system of masses and springs.
(a) Choose as generalized coordinates the displacement of each block from its equilibrium
position, and write the Lagrangian.
[5 points]
(b) Find the T and V matrices.
[5 points]
(c) Suppose
m1 = 2m
m2 = m
k1 = 4k
k2 = k
k3 = 2k ,
Find the frequencies of small oscillations.
[5 points]
(d) Find the normal modes of oscillation.
[5 points]
(e) At time t = 0, mass #1 is displaced by a distance b relative to its equilibrium position.
I.e. x1 (0) = b. The other initial conditions are x2 (0) = 0, x 1 (0) = 0, and x 2 (0) = 0.
Find t , the next time at which x2 vanishes.
[5 points]
Solution
(a) The Lagrangian is
L = 21 m1 x21 + 12 m2 x22 21 k1 x21 12 k2 (x2 x1 )2 12 k3 x22
(b) The T and V matrices are
2T
=
Tij =
x i x j
m1 0
0 m2
2U
=
Vij =
xi xj
k1 + k2
k2
k2
k2 + k3
398
CHAPTER 17. PHYSICS 110A-B EXAMS
(c) We have m1 = 2m, m2 = m, k1 = 4k, k2 = k, and k3 = 2k. Let us write 2 02 ,
p
where 0 k/m. Then
2 5
1
2T V = k
.
1
3
The determinant is
det ( 2 T V) = (22 11 + 14) k2
= (2 7) ( 2) k2 .
There are two roots: = 2 and + = 72 , corresponding to the eigenfrequencies
=
2k
m
+ =
7k
2m
~ (a) = 0. Plugging in = 2 we have
(d) The normal modes are determined from (a2 T V)
~ ()
for the normal mode
()
1
1 1
1
()
~
=
C
=
0
1
1 1
2()
Plugging in =
7
2
~ (+)
we have for the normal mode
2 1
1 12
1(+)
2(+)
(a)
The standard normalization i
~ (+) = C
=0
1
2
(b)
Tij j = ab gives
C =
1
3m
1
C2 =
.
6m
(17.66)
(e) The general solution is
!
1
x1
1
1
1
=A
cos( t) + B
cos(+ t) + C
sin( t) + D
sin(+ t) .
1
2
1
2
x2
The initial conditions x1 (0) = b, x2 (0) = x 1 (0) = x 2 (0) = 0 yield
A = 23 b
B = 13 b
C=0 ,
D=0.
Thus,
x1 (t) = 31 b 2 cos( t) + cos(+ t)
x2 (t) = 32 b cos( t) cos(+ t) .
17.3. F05 PHYSICS 110A FINAL EXAM
399
Setting x2 (t ) = 0, we find
cos( t ) = cos(+ t )
t = + t
t =
2
+ +
[2] Two point particles of masses m1 and m2 interact via the central potential
r2
U (r) = U0 ln
,
r 2 + b2
where b is a constant with dimensions of length.
(a) For what values of the relative angular momentum does a circular orbit exist? Find
the radius r0 of the circular orbit. Is it stable or unstable?
[7 points]
(c) For the case where a circular orbit exists, sketch the phase curves for the radial motion
in the (r, r)
half-plane. Identify the energy ranges for bound and unbound orbits.
[5 points]
(c) Suppose the orbit is nearly circular, with r = r0 +, where || r0 . Find the equation
for the shape () of the perturbation.
[8 points]
(d) What is the angle through which periapsis changes each cycle? For which value(s)
of does the perturbed orbit not precess?
[5 points]
400
CHAPTER 17. PHYSICS 110A-B EXAMS
Solution
(a) The effective potential is
2
+ U (r)
2r 2
2
r2
=
+ U0 ln
.
2r 2
r 2 + b2
Ueff (r) =
where = m1 m2 /(m1 + m1 ) is the reduced mass. For a circular orbit, we must have
(r) = 0, or
Ueff
l2
2rU0 b2
.
= U (r) = 2 2
3
r
r (r + b2 )
The solution is
r02 =
b2 2
2b2 U0 2
Since r02 > 0, the condition on is
< c
For large r, we have
Ueff (r) =
2b2 U0
2
U0 b2
2
1
+ O(r 4 ) .
r2
Thus, for < c the effective potential is negative for sufficiently large values of r. Thus,
over the range < c , we must have Ueff,min < 0, which must be a global minimum, since
Ueff (0+ ) = and Ueff () = 0. Therefore, the circular orbit is stable whenever it exists.
(b) Let = c . The effective potential is then
Ueff (r) = U0 f (r/b) ,
where the dimensionless effective potential is
f (s) =
2
ln(1 + s2 ) .
s2
The phase curves are plotted in Fig. 17.7.
(c) The energy is
E = 21 r 2 + Ueff (r)
2
dr
2
=
+ Ueff (r) ,
4
2r
d
17.3. F05 PHYSICS 110A FINAL EXAM
401
Figure 17.7: Phase curves for the scaled effective potential f (s) = p
s2 ln(1 + s2 ), with
= 12 . Here, = /c . The dimensionless time variable is = t U0 /mb2 .
Writing r = r + and differentiating E
where weve used r = r along with = r 2 .
0
with respect to , we find
= 2
2 =
r04
U (r ) .
2 eff 0
For our potential, we have
2
2
=2 1 2
2 = 2 2
b U0
c
The solution is
() = A cos( + )
where A and are constants.
(d) The change of periapsis per cycle is
= 2 1 1
(17.67)
402
CHAPTER 17. PHYSICS 110A-B EXAMS
If > 1 then < 0 and periapsis advances each cycle (i.e.it comes sooner with
p every
cycle). If < 1 then > 0 and periapsis recedes. For = 1, which means = b2 U0 ,
there is no precession and = 0.
[3] A particle of charge e moves in three dimensions in the presence of a uniform magnetic
The potential energy is
field B = B0 z and a uniform electric field E = E0 x.
e
= e E0 x B0 x y ,
U (r, r)
c
where we have chosen the gauge A = B0 x y.
(a) Find the canonical momenta px , py , and pz .
[7 points]
(b) Identify all conserved quantities.
[8 points]
(c) Find a complete, general solution for the motion of the system x(t), y(t), x(t) .
[10 points]
Solution
(a) The Lagrangian is
L = 21 m(x 2 + y 2 + z 2 ) +
e
B x y + e E0 x .
c 0
The canonical momenta are
px =
py =
L
= mx
x
e
L
= my + B0 x
y
c
px =
L
= mz
z
(b) There are three conserved quantities. First is the momentum py , since Fy =
Second is the momentum pz , since Fy =
Hamiltonian, since L
t = 0. We have
L
z
= 0.
= 0. The third conserved quantity is the
H = px x + py y + pz z L
L
y
H = 21 m(x 2 + y 2 + z 2 ) e E0 x
17.3. F05 PHYSICS 110A FINAL EXAM
403
(c) The equations of motion are
e
E
m 0
y + c x = 0
x
c y =
z = 0 .
The second equation can be integrated once to yield y = c (x0 x), where x0 is a constant.
Substituting this into the first equation gives
x
+ c2 x = c2 x0 +
e
E .
m 0
This is the equation of a constantly forced harmonic oscillator. We can therefore write the
general solution as
x(t) = x0 +
eE0
+ A cos c t +
2
mc
y(t) = y0
eE0
t A sin c t +
mc
z(t) = z0 + z0 t
Note that there are six constants, A, , x0 , y0 , z0 , z0 , are are required for the general
solution of three coupled second order ODEs.
[4] An N = 1 dynamical system obeys the equation
du
= ru + 2bu2 u3 ,
dt
where r is a control parameter, and where b > 0 is a constant.
(a) Find and classify all bifurcations for this system.
[7 points]
(b) Sketch the fixed points u versus r.
[6 points]
Now let b = 3. At time t = 0, the initial value of u is u(0) = 1. The control parameter
r is then increased very slowly from r = 20 to r = +20, and then decreased very
slowly back down to r = 20.
(c) What is the value of u when r = 5 on the increasing part of the cycle?
[3 points]
404
CHAPTER 17. PHYSICS 110A-B EXAMS
(d) What is the value of u when r = +16 on the increasing part of the cycle?
[3 points]
(e) What is the value of u when r = +16 on the decreasing part of the cycle?
[3 points]
(f) What is the value of u when r = 5 on the decreasing part of the cycle?
[3 points]
Solution
(a) Setting u = 0 we obtain
(u2 2bu r) u = 0 .
The roots are
p
u = 0 , u = b b2 + r .
The roots at u = u = b b2 + r are only present when r > b2 . At r = b2 there
is a saddle-node bifurcation. The fixed point u = u crosses the fixed point at u = 0 at
r = 0, at which the two fixed points exchange stability. This corresponds to a transcritical
bifurcation. In Fig. 17.8 we plot u/b
3 versus u/b for several representative values of r/b2 .
Note that, defining u
= u/b, r = r/b2 , and t = b2 t that our N = 1 system may be written
d
u
= r + 2
uu
2 u
,
dt
which shows that it is only the dimensionless combination r = r/b2 which enters into the
location and classification of the bifurcations.
(b) A sketch of the fixed points u versus r is shown in Fig. 17.9. Note the two bifurcations
at r = b2 (saddle-node) and r = 0 (transcritical).
(c) For r = 20 < b2 = 9, the initial condition u(0) = 1 flows directly toward the stable
fixed point at u = 0. Since the approach to the FP is asymptotic, u remains slightly positive
even after a long time. When r = 5, the FP at u = 0 is still stable. Answer: u = 0.
(d) As soon as r becomes positive, the FP at u =0 becomes unstable, and u flows to the
upper branch u+ . When r = 16, we have u = 3 + 32 + 16 = 8. Answer: u = 8.
(e) Coming back down from larger r, the upper FP branch remains stable, thus, u = 8 at
r = 16 on the way down as well. Answer: u = 8.
(f) Now when r first becomes negative on the way down, the upper branch u+ remains
stable. Indeed it remains stable all the way down to r = b2 , the location of the saddlenode bifurcation, at which point the solution u = u+ simply vanishes and the flow is toward
u = 0 again. Thus, forr = 5 on the way down, the system remains on the upper branch,
in which case u = 3 + 32 5 = 5. Answer: u = 5.
17.4. F07 PHYSICS 110A MIDTERM #1
405
Figure 17.8: Plot of dimensionless velocity u/b
3 versus dimensionless coordinate u/b for
several values of the dimensionless control parameter r = r/b2 .
17.4
F07 Physics 110A Midterm #1
[1] A particle of mass m moves in the one-dimensional potential
U (x) =
2
U0 2
x a2 .
4
a
(17.68)
(a) Sketch U (x). Identify the location(s) of any local minima and/or maxima, and be sure
that your sketch shows the proper behavior as x .
[15 points]
Solution : Clearly the minima lie at x = a and there is a local maximum at x = 0.
(b) Sketch a representative set of phase curves. Be sure to sketch any separatrices which
exist, and identify their energies. Also sketch all the phase curves for motions with total
energy E = 12 U0 . Do the same for E = 2 U0 .
[15 points]
Solution : See Fig. 17.10 for the phase curves. Clearly U (a) = 0 is the minimum of the
406
CHAPTER 17. PHYSICS 110A-B EXAMS
Figure 17.9: Fixed points and their stability versus control parameter for the N = 1 system
u = ru + 2bu2 u3 . Solid lines indicate stable fixed points; dashed lines indicate unstable
fixed points. There is a saddle-node bifurcation at r = b2 and a transcritical bifurcation
at r = 0. The hysteresis loop in the upper half plane u > 0 is shown. For u < 0 variations
of the control parameter r are reversible and there is no hysteresis.
potential, and U (0) = U0 is the local maximum and the energy of the separatrix. Thus,
E = 12 U0 cuts through the potential in both wells, and the phase curves at this energy form
two disjoint sets. For E < U0 there are four turning points, at
s
x1,< = a
and
x2,< = a
1+
E
U0
E
U0
x1,> = a
x2,> = a
1+
E
U0
E
U0
For E = 2 U0 , the energy is above that of the separatrix, and there are only two turning
points, x1,< and x2,> . The phase curve is then connected.
17.4. F07 PHYSICS 110A MIDTERM #1
407
Figure 17.10: Sketch of the double well potential U (x) = (U0 /a4 )(x2 a2 )2 , here with
distances in units of a, and associated phase curves. The separatrix is the phase curve
which runs through the origin. Shown in red is the phase curve for U = 21 U0 , consisting
of two deformed ellipses. For U = 2 U0 , the phase curve is connected, lying outside the
separatrix.
x
. Find the upper
(c) The phase space dynamics are written as = V (), where =
x
and lower components of the vector field V .
[10 points]
Solution :
d
dt
x
x
x
=
.
=
1
0
m
x
U (x)
x (x2 a2 )
4U
a2
(17.69)
(d) Derive and expression for the period T of the motion when the system exhibits small
oscillations about a potential minimum.
[10 points]
408
CHAPTER 17. PHYSICS 110A-B EXAMS
Solution : Set x = a + and Taylor expand:
U (a + ) =
4U0 2
+ O( 3 ) .
a2
(17.70)
Equating this with 12 k 2 , we have the effective spring constant k = 8U0 /a2 , and the small
oscillation frequency
r
r
k
8 U0
0 =
=
.
(17.71)
m
ma2
The period is 2/0 .
[2] An R-L-C circuit is shown in fig. 17.11. The resistive element is a light bulb. The
inductance is L = 400 H; the capacitance is C = 1 F; the resistance is R = 32 . The
voltage V (t) oscillates sinusoidally, with V (t) = V0 cos(t), where V0 = 4 V. In this problem,
you may neglect all transients; we are interested in the late time, steady state operation of
this circuit. Recall the relevant MKS units:
1 = 1V s/C
1F = 1C/V
1 H = 1 V s2 / C .
Figure 17.11: An R-L-C circuit in which the resistive element is a light bulb.
(a) Is this circuit underdamped or overdamped?
[10 points]
Solution : We have
0 = (LC)1/2 = 5 104 s1
Thus, 02 > 2 and the circuit is underdamped.
R
= 4 104 s1 .
2L
17.4. F07 PHYSICS 110A MIDTERM #1
409
(b) Suppose the bulb will only emit light when the average power dissipated by the bulb is
greater than a threshold Pth = 92 W . For fixed V0 = 4 V, find the frequency range for over
which the bulb emits light. Recall that the instantaneous power dissipated by a resistor is
PR (t) = I 2 (t)R. (Average this over a cycle to get the average power dissipated.)
[20 points]
Solution : The charge on the capacitor plate obeys the ODE
+ R Q +
LQ
Q
= V (t) .
C
The solution is
Q(t) = Qhom (t) + A()
with
i1/2
h
A() = (02 2 )2 + 4 2 2
V0
cos t () ,
L
2
1
, () = tan
.
02 2
Thus, ignoring the transients, the power dissipated by the bulb is
PR (t) = Q 2 (t) R
= 2 A2 ()
V02 R
sin2 t () .
2
L
Averaging over a period, we have h sin2 (t ) i = 12 , so
h PR i = 2 A2 ()
Now V02 /2R =
1
4
V02 R
V02
4 2 2
.
=
2
2L2
2 R (0 2 )2 + 4 2 2
W. So Pth = V02 /2R, with = 98 . We then set hPR i = Pth , whence
(1 ) 4 2 2 = (02 2 )2 .
The solutions are
=
1
+
s
1 2
+ 02 = 3 3 2 1000 s1 .
(c) Compare the expressions for the instantaneous power dissipated by the voltage source,
PV (t), and the power dissipated by the resistor PR (t) = I 2 (t)R. If PV (t) 6= PR (t), where
does the power extra power go or come from? What can you say about the averages of PV
and PR (t) over a cycle? Explain your answer.
[20 points]
Solution : The instantaneous power dissipated by the voltage source is
V0
sin(t ) cos(t)
L
V0
sin sin(2t ) .
= A
2L
PV (t) = V (t) I(t) = A
410
CHAPTER 17. PHYSICS 110A-B EXAMS
As we have seen, the power dissipated by the bulb is
PR (t) = 2 A2
V02 R
sin2(t ) .
L2
These two quantities are not identical, but they do have identical time averages over one
cycle:
V2
h PV (t) i = h PR (t) i = 0 4 2 2 A2 () .
2R
Energy conservation means
PV (t) = PR (t) + E(t)
,
where
E(t) =
LQ 2
Q2
+
2
2C
is the energy in the inductor and capacitor. Since Q(t) is periodic, the average of E over a
cycle must vanish, which guarantees h PV (t) i = h PR (t) i.
What was not asked:
(d) What is the maximum charge Qmax on the capacitor plate if = 3000 s1 ?
[10 points]
Solution : Kirchoffs law gives for this circuit the equation
+ 2 Q + 02 Q = V0 cos(t) ,
Q
L
with the solution
Q(t) = Qhom (t) + A()
V0
cos t () ,
L
where Qhom (t) is the homogeneous solution, i.e. the transient which we ignore, and
i1/2
h
2
1
2
2 2
2 2
, () = tan
A() = (0 ) + 4
.
02 2
Then
Qmax = A()
V0
.
L
Plugging in = 3000 s1 , we have
1/2
1
A() = (52 42 )2 + 4 42 32
103 s2 = 103 s2 .
8 13
Since V0 /L = 104 C/s2 , we have
5
Qmax = Coul .
4 13
17.5. F07 PHYSICS 110A MIDTERM #2
17.5
411
F07 Physics 110A Midterm #2
[1] A point mass m slides frictionlessly, under the influence of gravity, along a massive ring
of radius a and mass M . The ring is affixed by horizontal springs to two fixed vertical
surfaces, as depicted in fig. 17.12. All motion is within the plane of the figure.
Figure 17.12: A point mass m slides frictionlessly along a massive ring of radius a and mass
M , which is affixed by horizontal springs to two fixed vertical surfaces.
(a) Choose as generalized coordinates the horizontal displacement X of the center of the
ring with respect to equilibrium, and the angle a radius to the mass m makes with respect
to the vertical (see fig. 17.12). You may assume that at X = 0 the springs are both
t).
,
unstretched. Find the Lagrangian L(X, , X,
[15 points]
The coordinates of the mass point are
x = X + a sin
y = a cos .
The kinetic energy is
2
T = 21 M X 2 + 12 m X + a cos + 21 ma2 sin2 2
= 1 (M + m)X 2 + 1 ma2 2 + ma cos X .
2
The potential energy is
U = kX 2 mga cos .
Thus, the Lagrangian is
L = 12 (M + m)X 2 + 21 ma2 2 + ma cos X kX 2 + mga cos .
(b) Find the generalized momenta pX and p , and the generalized forces FX and F
[10 points]
We have
pX =
L
= (M + m)X + ma cos
X
p =
L
= ma2 + ma cos X .
412
CHAPTER 17. PHYSICS 110A-B EXAMS
For the forces,
FX =
L
= 2kX
X
F =
L
= ma sin X mga sin .
(c) Derive the equations of motion.
[15 points]
The equations of motion are
d L
L
,
=
dt q
q
for each generalized coordinate q . For X we have
+ ma cos ma sin 2 = 2kX .
(M + m)X
For ,
= mga sin .
ma2 + ma cos X
(d) Find expressions for all conserved quantities.
[10 points]
Horizontal and vertical translational symmetries are broken by the springs and by gravity,
L
respectively. The remaining symmetry is that of time translation. From dH
dt = t , we have
P
that H = p q L is conserved. For this problem, the kinetic energy is a homogeneous
function of degree 2 in the generalized velocities, and the potential is velocity-independent.
Thus,
H = T + U = 12 (M + m)X 2 + 21 ma2 2 + ma cos X + kX 2 mga cos .
17.5. F07 PHYSICS 110A MIDTERM #2
413
[2] A point particle of mass m moves in three dimensions in a helical potential
2z
U (, , z) = U0 cos
.
b
We call b the pitch of the helix.
(a) Write down the Lagrangian, choosing (, , z) as generalized coordinates.
[10 points]
The Lagrangian is
L=
1
2m
2 2
+ + z
2z
U0 cos
b
(b) Find the equations of motion.
[20 points]
Clearly
p = m
p = m2 ,
pz = mz ,
and
2z
2
F = m U0 cos
b
2z
F = U0 sin
b
2z
2U0
sin
Fz =
.
b
b
Thus, the equation of motion are
2z
2
m
= m U0 cos
b
2z
m2 + 2m = U0 sin
b
2U0
2z
m
z=
sin
.
b
b
(c) Show that there exists a continuous one-parameter family of coordinate transformations
which leaves L invariant. Find the associated conserved quantity, . Is anything else
conserved?
[20 points]
Due to the helical symmetry, we have that
+
zz+
is such a continuous one-parameter family of coordinate transformations. Since it leaves
414
the combination
CHAPTER 17. PHYSICS 110A-B EXAMS
2z
b
unchanged, we have that
dL
d
= 0, and
z
= p
+ p
+ pz
=0
=0
=0
b
p
2 z
mb
z
= m2 +
2
= p +
is the conserved Noether charge. The other conserved quantity is the Hamiltonian,
2z
2
2 2
2
1
.
H = 2 m + + z + U0 cos
b
Note that H = T + U , because T is homogeneous of degree 2 and U is homogeneous of
degree 0 in the generalized velocities.
17.6. F07 PHYSICS 110A FINAL EXAM
17.6
415
F07 Physics 110A Final Exam
[1] Two masses and two springs are configured linearly and externally driven to rotate with
angular velocity about a fixed point on a horizontal surface, as shown in fig. 17.13. The
unstretched length of each spring is a.
Figure 17.13: Two masses and two springs rotate with angular velocity .
(a) Choose as generalized coordinates the radial distances r1,2 from the origin. Find the
Lagrangian L(r1 , r2 , r1 , r2 , t).
[5 points]
The Lagrangian is
L = 12 m r12 + r22 + 2 r12 + 2 r22 12 k (r1 a)2 12 k (r2 r1 a)2 .
(17.72)
(b) Derive expressions for all conserved quantities.
[5 points]
The Hamiltonian is conserved. Since the kinetic energy is not homogeneous of degree 2 in
the generalized velocities, H 6= T + U . Rather,
X
H=
p q L
(17.73)
=
1
2m
r12 + r22 12 m 2 r12 + r22 ) + 12 k (r1 a)2 + 21 k (r2 r1 a)2 .
(17.74)
We could define an effective potential
Ueff (r1 , r2 ) = 12 m 2 r12 + r22 ) + 21 k (r1 a)2 + 12 k (r2 r1 a)2 .
(17.75)
Note the first term, which comes from the kinetic energy, has an interpretation of a fictitious
potential which generates a centrifugal force.
416
CHAPTER 17. PHYSICS 110A-B EXAMS
(c) What equations determine the equilibrium radii r10 and r20 ? (You do not have to solve
these equations.)
[5 points]
The equations of equilibrium are F = 0. Thus,
L
= m 2 r1 k (r1 a) + k (r2 r1 a)
r1
L
0 = F2 =
= m 2 r2 k (r2 r1 a) .
r2
0 = F1 =
(17.76)
(17.77)
(d) Suppose now that the system is not externally driven, and that the angular coordinate
t).
is a dynamical variable like r1 and r2 . Find the Lagrangian L(r1 , r2 , , r1 , r2 , ,
[5 points]
Now we have
L = 12 m r12 + r22 + r12 2 + r22 2 21 k (r1 a)2 12 k (r2 r1 a)2 .
(17.78)
(e) For the system described in part (d), find expressions for all conserved quantities.
[5 points]
There are two conserved quantities. One is p , owing to the fact the is cyclic in the
Lagrangian. I.e. + is a continuous one-parameter coordinate transformation which
leaves L invariant. We have
p =
L
= m r12 + r22 .
(17.79)
The second conserved quantity is the Hamiltonian, which is now H = T + U , since T is
homogeneous of degree 2 in the generalized velocities. Using conservation of momentum,
we can write
H=
1
2m
r12
r22
p2
2m(r12 + r22 )
+ 21 k (r1 a)2 + 12 k (r2 r1 a)2 .
(17.80)
Once again, we can define an effective potential,
Ueff (r1 , r2 ) =
p2
2m(r12 + r22 )
+ 21 k (r1 a)2 + 21 k (r2 r1 a)2 ,
(17.81)
which is different than the effective potential from part (b). However in both this case and
in part (b), we have that the radial coordinates obey the equations of motion
m
rj =
Ueff
,
rj
for j = 1, 2. Note that this equation of motion follows directly from H = 0.
(17.82)
17.6. F07 PHYSICS 110A FINAL EXAM
417
Figure 17.14: A mass point m rolls inside a hoop of mass M and radius R which rolls
without slipping on a horizontal surface.
[2] A point mass m slides inside a hoop of radius R and mass M , which itself rolls without
slipping on a horizontal surface, as depicted in fig. 17.14.
Choose as general coordinates (X, , r), where X is the horizontal location of the center of
the hoop, is the angle the mass m makes with respect to the vertical ( = 0 at the bottom
of the hoop), and r is the distance of the mass m from the center of the hoop. Since the
mass m slides inside the hoop, there is a constraint:
G(X, , r) = r R = 0 .
Nota bene: The kinetic energy of the moving hoop, including translational and rotational
components (but not including the mass m), is Thoop = M X 2 (i.e. twice the translational
contribution alone).
r,
(a) Find the Lagrangian L(X, , r, X , ,
t).
[5 points]
The Cartesian coordinates and velocities of the mass m are
x = X + r sin + r cos
y = r cos + r sin
x = X + r sin
y = R r cos
(17.83)
(17.84)
The Lagrangian is then
z
z
}|
{
}|
{
2
2
2
2
1
1
r sin + r cos ) mg(R r cos ) (17.85)
L =(M + 2 m)X + 2 m(r + r ) + mX(
Note that we are not allowed to substitute r = R and hence r = 0 in the Lagrangian prior
to obtaining the equations of motion. Only after the generalized momenta and forces are
computed are we allowed to do so.
(b) Find all the generalized momenta p , the generalized forces F , and the forces of
constraint Q .
[10 points]
418
CHAPTER 17. PHYSICS 110A-B EXAMS
The generalized momenta are
L
= mr + mX sin
r
(17.86)
pX =
L
= (2M + m)X + mr sin + mr cos
X
(17.87)
p =
L
= mr 2 + mr X cos
(17.88)
pr =
The generalized forces and the forces of constraint are
Fr =
L
= mr 2 + mX cos + mg cos
r
Qr =
G
=
r
(17.89)
FX =
L
=0
X
QX =
G
=0
X
(17.90)
F =
L
= mX r cos mX sin mgr sin
Q =
G
=0.
(17.91)
The equations of motion are
p = F + Q .
(17.92)
At this point, we can legitimately invoke the constraint r = R and set r = 0 in all the p
and F .
(c) Derive expressions for all conserved quantities.
[5 points]
There are two conserved quantities, which each derive from continuous invariances of the
Lagrangian which respect the constraint. The first is the total momentum pX :
FX = 0
P pX = constant .
(17.93)
The second conserved quantity is the Hamiltonian, which in this problem turns out to be
the total energy E = T + U . Incidentally, we can use conservation of P to write the energy
in terms of the variable alone. From
X =
mR cos
P
,
2M + m
2M + m
(17.94)
we obtain
E = 21 (2M + m)X 2 + 12 mR2 2 + mRX cos + mgR(1 cos )
2
P 2
2 1 + sin
1
+ mR
2 + mgR(1 cos ) ,
=
2m(1 + ) 2
1+
(17.95)
where weve defined the dimensionless ratio m/2M . It is convenient to define the
quantity
1 + sin2 2
2
+ 202 (1 cos ) ,
(17.96)
1+
17.6. F07 PHYSICS 110A FINAL EXAM
with 0
419
g/R. Clearly 2 is conserved, as it is linearly related to the energy E:
E=
P 2
+ 1 mR2 2 .
2m(1 + ) 2
(17.97)
(d) Derive a differential equation of motion involving the coordinate (t) alone. I.e. your
equation should not involve r, X, or the Lagrange multiplier .
[5 points]
From conservation of energy,
d( 2 )
=0
dt
1 + sin2
1+
sin cos
1+
2 + 02 sin = 0 ,
(17.98)
again with = m/2M . Incidentally, one can use these results in eqns. 17.96 and 17.98 to
eliminate and in the expression for the constraint force, Qr = = pr Fr . One finds
2 + 02 cos
1 + sin2
)
(
2
mR 02
4 sin2 ( 12 ) + (1 + sin2 ) cos .
=
(1 + )
02
(1 + sin2 )2
= mR
(17.99)
This last equation can be used to determine the angle of detachment, where vanishes and
the mass m falls off the inside of the hoop. This is because the hoop can only supply a
repulsive normal force to the mass m. This was worked out in detail in my lecture notes on
constrained systems.
[3] Two objects of masses m1 and m2 move under the influence of a central potential
1/4
U = k r1 r2 .
(a) Sketch the effective potential Ueff (r) and the phase curves for the radial motion. Identify
for which energies the motion is bounded.
[5 points]
The effective potential is
Ueff (r) =
2
+ kr n
2r 2
(17.100)
with n = 41 . In sketching the effective potential, I have rendered it in dimensionless form,
Ueff (r) = E0 Ueff (r/r0 ) ,
1
where r0 = (2 /nk)(n+2) and E0 =
of part (b). One then finds
1
2
Ueff (x) =
1
n
(17.101)
2 /r02 , which are obtained from the results
n x2 + 2 xn
.
n+2
(17.102)
420
CHAPTER 17. PHYSICS 110A-B EXAMS
Figure 17.15: The effective Ueff (r) = E0 Ueff (r/r0 ), where r0 and E0 are the radius and
energy of the circular orbit.
Although it is not obvious from the detailed sketch in fig. 17.15, the effective potential does
diverge, albeit slowly, for r . Clearly it also diverges for r 0. Thus, the relative
coordinate motion is bounded for all energies; the allowed energies are E E0 .
(b) What is the radius r0 of the circular orbit? Is it stable or unstable? Why?
[5 points]
For the general power law potential U (r) = kr n , with nk > 0 (attractive force), setting
(r ) = 0 yields
Ueff
0
2
(17.103)
3 + nkr0n1 = 0 .
r0
Thus,
r0 =
2
nk
1
n+2
42
k
4
9
(17.104)
The orbit r(t) = r0 is stable because the effective potential has a local minimum at r = r0 ,
17.6. F07 PHYSICS 110A FINAL EXAM
421
(r ) > 0. This is obvious from inspection of the graph of U (r) but can also be
i.e. Ueff
eff
0
computed explicitly:
32
+ n(n 1)kr0n
r04
2
= (n + 2) 4 .
r0
Ueff
(r0 ) =
(17.105)
(r ) > 0.
Thus, provided n > 2 we have Ueff
0
(c) For small perturbations about a circular orbit, the radial coordinate oscillates between
two values. Suppose we compare two systems, with / = 2, but = and k = k. What
is the ratio / of their frequencies of small radial oscillations?
[5 points]
(r), we expand r = r + and find
From the radial coordinate equation
r = Ueff
0
= Ueff
(r0 ) + O( 2 ) .
(17.106)
The radial oscillation frequency is then
= (n + 2)1/2
n2
2
2
n
1/2 n+2
n+2 n+2 n+2 .
k
=
(n
+
2)
n
r02
(17.107)
The dependence is what is key here. Clearly
n2
n+2
.
(17.108)
In our case, with n = 41 , we have 7/9 and thus
= 27/9 .
(17.109)
(d) Find the equation of the shape of the slightly perturbed circular orbit: r() = r0 +().
That is, find (). Sketch the shape of the orbit.
[5 points]
We have that () = 0 cos( + 0 ), with
=
r 2
= 0 = n+2 .
(17.110)
With n = 14 , we have = 32 . Thus, the radial coordinate makes three oscillations for every
two rotations. The situation is depicted in fig. 17.21.
(e) What value of n would result in a perturbed orbit shaped like that in fig. 17.22?
[5 points]
422
CHAPTER 17. PHYSICS 110A-B EXAMS
Figure 17.16: Radial oscillations with = 23 .
Figure 17.17: Closed precession in a central potential U (r) = kr n .
Clearly = n + 2 = 4, in order that () = 0 cos(+0 ) executes four complete periods
over the interval [0, 2]. This means n = 14.
[4] Two masses and three springs are arranged as shown in fig. 17.18. You may assume
that in equilibrium the springs are all unstretched with length a. The masses and spring
constants are simple multiples of fundamental values, viz.
m1 = m
m2 = 4m
k1 = k
k2 = 4k
k3 = 28k .
Figure 17.18: Coupled masses and springs.
(a) Find the Lagrangian.
[5 points]
(17.111)
17.6. F07 PHYSICS 110A FINAL EXAM
423
Choosing displacements relative to equilibrium as our generalized coordinates, we have
T = 12 m 12 + 2m 22
(17.112)
U = 12 k 12 + 2k (2 1 )2 + 14k 22 .
(17.113)
and
Thus,
L = T U = 12 m 12 + 2m 22 21 k 12 2k (2 1 )2 14k 22 .
(17.114)
You are not required to find the equilibrium values of x1 and x2 . However, suppose all the
unstretched spring lengths are a and the total distance between the walls is L. Then, with
x1,2 being the location of the masses relative to the left wall, we have
U = 12 k1 (x1 a)2 + 21 k2 (x2 x1 a)2 + 12 k3 (L x2 a)2 .
(17.115)
Differentiating with respect to x1,2 then yields
U
= k1 (x1 a) k2 (x2 x1 a)
x1
U
= k2 (x2 x1 a) k3 (L x2 a) .
x2
(17.116)
(17.117)
Setting these both to zero, we obtain
(k1 + k2 ) x1 k2 x2 = (k1 k2 ) a
(17.118)
k2 x1 + (k2 + k3 ) x2 = (k2 k3 ) a + k3 L .
(17.119)
Solving these two inhomogeneous coupled linear equations for x1,2 then yields the equilibrium positions. However, we dont need to do this to solve the problem.
(b) Find the T and V matrices.
[5 points]
We have
T
and
V
2T
=
=
2U
=
=
m 0
0 4m
5k 4k
4k 32k
(17.120)
(17.121)
(c) Find the eigenfrequencies 1 and 2 .
[5 points]
We have
m 2 5k
4k
Q() T V =
2
4k
4m 32k
5
4
=k
,
4
4 32
2
(17.122)
424
where = 2 /02 , with 0 =
CHAPTER 17. PHYSICS 110A-B EXAMS
k/m. Setting det Q() = 0 then yields
2 13 + 36 = 0 ,
(17.123)
the roots of which are = 4 and + = 9. Thus, the eigenfrequencies are
= 2 0
+ = 3 0 .
(17.124)
(d) Find the modal matrix Ai .
[5 points]
To find the normal modes, we set
5
4
4
4 32
()
1
()
2
=0.
(17.125)
This yields two linearly dependent equations, from which we can determine only the ratios
()
()
2 /1 . Plugging in for , we find
!
!
()
(+)
1
4
1
1
= C
,
= C+
.
(17.126)
()
(+)
1
1
2
2
(i)
(j)
We then normalize by demanding T = ij . We can practically solve this by
inspection:
20m |C |2 = 1
,
5m |C+ |2 = 1 .
(17.127)
We may now write the modal matrix,
1
A=
5m
2
1
2
1
.
1
(17.128)
(e) Write down the most general solution for the motion of the system.
[5 points]
The most general solution is
!
1 (t)
4
1
= B
cos(20 t + ) + B+
cos(30 t + + ) .
1
1
2 (t)
Note that there are four constants of integration: B and .
(17.129)
17.7. W08 PHYSICS 110B MIDTERM EXAM
17.7
425
W08 Physics 110B Midterm Exam
[1] Two identical semi-infinite lengths of string are joined at a point of mass m which
moves vertically along a thin wire, as depicted in fig. 17.21. The mass moves with friction
coefficient , i.e. its equation of motion is
m
z + z = F ,
(17.130)
where z is the vertical displacement of the mass, and F is the force on the mass due to
the string segments on either side. In this problem, gravity is to be neglected. It may be
convenient to define K 2 /mc2 and Q /mc.
Figure 17.19: A point mass m joining two semi-infinite lengths of identical string moves
vertically along a thin wire with friction coefficient .
(a) The general solution with an incident wave from the left is written
(
f (ct x) + g(ct + x) (x < 0)
y(x, t) =
h(ct x)
(x > 0) .
Find two equations relating the functions f (), g(), and h().
[20 points]
The first equation is continuity at x = 0:
f () = g() + h()
where = ct ranges over the real line [, ]. The second equation comes from Newtons
2nd law F = ma applied to the mass point:
m y(0, t) + y(0,
t) = y (0+ , t) y (0 , t) .
Expressed in terms of the functions f (), g(), and h(), and dividing through by mc2 ,
this gives
f () + g () + Q f () + Q g () = 12 K h () + 12 K f () 21 K g ().
426
CHAPTER 17. PHYSICS 110A-B EXAMS
Integrating once, and invoking h = f + g, this second equation becomes
f () + Q f () = g () (K + Q) g()
(b) Solve for the reflection amplitude r(k) = g(k)/f(k) and the transmission amplitude
t(k) = h(k)/
f(k). Recall that
f () =
dk
f (k) eik
2
f (k) = d f () eik ,
et cetera for the Fourier
Also compute the sum of the reflection and transmission
2 transforms.
2
coefficients, r(k) + t(k) . Show that this sum is always less than or equal to unity, and
interpret this fact.
[20 points]
Using d/d ik, we have
(Q + ik) f(k) = (K + Q + ik) g(k) .
(17.131)
Therefore,
r(k) =
Q + ik
g(k)
=
Q
+
K + ik
f (k)
(17.132)
To find the transmission amplitude, we invoke h() = f () + g(), in which case
t(k) =
K
h(k)
=
Q + K + ik
f(k)
(17.133)
The sum of reflection and transmission coefficients is
2
2
2
r(k) 2 + t(k) 2 = Q + K + k
(Q + K)2 + k2
(17.134)
Clearly the RHS of this equation is bounded from above by unity, since both Q and K are
nonnegative.
(c) Find an expression in terms of the functions f , g, and h (and/or their derivatives) for
the rate E at which energy is lost by the string. Do this by evaluating the energy current
on either side of the point mass. Your expression should be an overall function of time t.
[10 points]
Recall the formulae for the energy density in a string,
E(x, t) =
1
2
y 2 (x, t) + 12 y (x, t)
(17.135)
17.7. W08 PHYSICS 110B MIDTERM EXAM
427
and
jE (x, t) = y(x,
t) y (x, t) .
(17.136)
The energy continuity equation is t E + x jE = 0. Assuming jE (, t) = 0, we have
dE
=
dt
Z0
Z
E
E
+ dx
dx
t
t
0+
= jE (, t) + jE (0+ , t) + jE (, t) jE (0 , t) .
(17.137)
Thus,
2
2
2
dE
= c g (ct) + h (ct) f (ct)
dt
(17.138)
Incidentally, if we integrate over all time, we obtain the total energy change in the string:
Z
2
2
2
E = d g () + h () f ()
2QK k2
dk
f(k) 2 .
2
2
2 (Q + K) + k
(17.139)
Note that the initial energy in the string, at time t = , is
E0 =
dk 2 2
k f (k) .
2
(17.140)
If the incident wave packet is very broad, say described by a Gaussian f () = A exp(x2 /2 2 )
with K 1 and Q 1, then k2 may be neglected in the denominator of eqn. 17.139,
in which case
2QK
E 12 E0 .
E
(17.141)
(Q + K)2 0
428
CHAPTER 17. PHYSICS 110A-B EXAMS
[2] Consider a rectangular cube of density and dimensions a b c, as depicted in fig.
17.22.
Figure 17.20: A rectangular cube of dimensions a b c. In part (c), a massless torsional
fiber is attached along the diagonal of one of the b c faces.
(a) Compute the inertia tensor I along body-fixed principle axes, with the origin at the
center of mass.
[15 points]
We first compute Izz :
CM
Izz
Za/2 Zb/2 Zc/2
= dx dy dz x2 + y 2 ) =
a/2 b/2 c/2
1
12
M a2 + b2 ,
(17.142)
where M = abc. Corresponding expressions hold for the other moments of inertia. Thus,
I CM
b2 + c2
0
0
1
a2 + c2
0
= 12
M 0
2
0
0
a + b2
(17.143)
(b) Shifting the origin to the center of either of the bc faces, and keeping the axes parallel,
compute the new inertia tensor.
[15 points]
and use the parallel axis theorem,
We shift the origin by a distance d = 12 a x
I (d) = I (0) + M d2 d d ,
resulting in
b + c2
0
0
I = 0
4a2 + c2
0
2
2
0
0
4a + b
(17.144)
(17.145)
(c) A massless torsional fiber is (masslessly) welded along the diagonal of either b c face.
The potential energy in this fiber is given by U () = 12 Y 2 , where Y is a constant and is
17.7. W08 PHYSICS 110B MIDTERM EXAM
429
the angle of rotation of the fiber. Neglecting gravity, find an expression for the oscillation
frequency of the system.
[20 points]
Let be the twisting angle of the fiber. The kinetic energy in the fiber is
T =
=
where
1
2 I
1
2
2 n I n
(17.146)
c z
b y
=
+
n
.
2
2
2
b +c
b + c2
We then find
Iaxis n I n = 31 M a2 + 16 M
The frequency of oscillation is then =
b2 c2
.
b2 + c2
(17.147)
(17.148)
Y /Iaxis , or
6Y
b2 + c2
2 2
M 2a b + c2 + b2 c2
(17.149)
430
CHAPTER 17. PHYSICS 110A-B EXAMS
17.8
W08 Physics 110B Final Exam
[1] Consider a string with uniform mass density and tension . At the point x = 0, the
string is connected to a spring of force constant K, as shown in the figure below.
Figure 17.21: A string connected to a spring.
(a) The general solution with an incident wave from the left is written
(
f (ct x) + g(ct + x) (x < 0)
y(x, t) =
h(ct x)
(x > 0) .
Find two equations relating the functions f (), g(), and h().
[10 points]
SOLUTION : The first equation is continuity at x = 0:
f () + g() = h()
where = ct ranges over the real line [, ]. The second equation comes from Newtons
2nd law F = ma applied to the mass point:
y (0+ , t) y (0 , t) K y(0, t) = 0 ,
or
h () + f () g () K f () + g() = 0
(b) Solve for the reflection amplitude r(k) = g(k)/f(k) and the transmission amplitude
t(k) = h(k)/
f(k). Recall that
f () =
dk
f (k) eik
2
Z
f(k) = d f () eik ,
17.8. W08 PHYSICS 110B FINAL EXAM
431
et cetera for the Fourier transforms.
Also
compute the sum of the reflection and
2
2
[10 points]
transmission coefficients, r(k) + t(k) .
SOLUTION : Taking the Fourier transform of the two equations from part (a), we have
f(k) + g(k) = h(k)
i k
f (k) + g(k) =
f (k) g(k) h(k) .
K
Solving for g(k) and h(k)
in terms of f(k), we find
g(k) = r(k) f(k)
h(k) = t(k) f(k)
where the reflection coefficient r(k) and the transmission coefficient t(k) are given by
r(k) =
K
K + 2i k
t(k) =
2i k
K + 2i k
Note that
r(k) 2 + t(k) 2 = 1
which says that the energy flux is conserved.
(c) For the Lagrangian density
L=
1
2
y
t
2
1
2
y
x
2
find the Euler-Lagrange equations of motion.
1
4
y
x
4
,
[7 points]
SOLUTION : For a Lagrangian density L(y, y,
y ), the Euler-Lagrange equations are
L
L
L
=
.
+
y
t y
x y
Thus, the wave equation for this system is
y = y + 3 y
2
432
CHAPTER 17. PHYSICS 110A-B EXAMS
(d) For the Lagrangian density
L=
1
2
y
t
2
2
1
2 y
1
2
y
x
2
find the Euler-Lagrange equations of motion.
1
4
2y
x2
2
,
[7 points]
SOLUTION : For a Lagrangian density L(y, y,
y , y ), the Euler-Lagrange equations are
L
L
2 L
L
=
2
.
+
y
t y
x y
x y
The last term arises upon integrating by parts twice in the integrand of the variation of the
action S. Thus, the wave equation for this system is
y = y + y y
17.8. W08 PHYSICS 110B FINAL EXAM
433
[2] Consider single species population dynamics governed by the differential equation
dN
N2
HN
= N
,
dt
K
N +L
where , K, L, and H are constants.
(a) Show that by rescaling N and t that the above ODE is equivalent to
du
hu
= r u u2
.
ds
u+1
Give the definitions of u, s, r, and h.
[5 points]
SOLUTION : From the denominator u + 1 in the last term of the scaled equation, we see that
we need to define N = Lu. We then write t = s, and substituting into the original ODE
yields
L2 2
Hu
L du
= Lu
u
.
ds
K
u+1
Multiplying through by /L then gives
L 2 H u
du
= u
u
.
ds
K
L u+1
We set the coefficient of the second term on the RHS equal to 1 to obtain the desired
form. Thus, = K/L and
u=
N
L
s=
Lt
K
r=
K
L
h=
KH
L2
(b) Find and solve the equation for all fixed points u (r, h).
[10 points]
SOLUTION : In order for u to be a fixed point, we need u = 0, which requires
h
u ru
u+1
=0
One solution is always u = 0 . The other roots are governed by the quadratic equation
(u r)(u + 1) + h = 0 ,
with roots at
u =
1
2
p
r 1 (r + 1)2 4h
434
CHAPTER 17. PHYSICS 110A-B EXAMS
Figure 17.22: Bifurcation curves for the equation u = ru u2 hu/(u + 1). Red curve:
hSN (r) = 14 (r + 1)2 , corresponding to saddle-node bifurcation. Blue curve: hT (r) = r,
corresponding to transcritical bifurcation.
(c) Sketch the upper right quadrant of the (r, h) plane. Show that there are four distinct
regions:
Region I
3 real fixed points (two negative)
Region II
3 real fixed points (one positive, one negative)
Region III
3 real fixed points (two positive)
Region IV
1 real fixed point
Find the equations for the boundaries of these regions. These boundaries are the
locations of bifurcations. Classify the bifurcations. (Note that negative values of u
are unphysical in the context of population dynamics, but are legitimate from a purely
[10 points]
mathematical standpoint.)
SOLUTION : From the quadratic equation for the non-zero roots, we see the discriminant
vanishes for h = 14 (r + 1)2 . For h > 14 (r + 1)2 , the discriminant is negative, and there is
one real root at u = 0. Thus, the curve hSN (r) = 41 (r + 1)2 corresponds to a curve of
saddle-node bifurcations. Clearly the largest value of u must be a stable node, because
for large u the u2 dominates on the RHS of u = f (u). In cases where there are three
fixed points, the middle one must be unstable, and the smallest stable. There is another
bifurcation, which occurs when the root at u = 0 is degenerate. This occurs at
p
=
h=r .
r 1 = (r + 1)2 4h
17.8. W08 PHYSICS 110B FINAL EXAM
435
Figure 17.23: Examples of phase flows for the equation u = ru u2 hu/(u + 1). (a) r = 1,
h = 0.22 (region I) ; (b) r = 1, h = 0.5 (region II) ; (c) r = 3, h = 3.8 (region III) ; (d)
r = 1, h = 1.5 (region IV).
This defines the curve for transcritical bifurcations: hT (r) = r . Note that hT (r) hSN (r),
since hSN (r) hT (r) = 41 (r 1)2 0. For h < r, one root is positive and one negative,
corresponding to region II.
The (r, h) control parameter space is depicted in fig. 17.22, with the regions I through IV
bounded by sections of the bifurcation curves, as shown.
(d) Sketch the phase flow for each of the regions I through IV.
SOLUTION : See fig. 17.23.
[8 points]
436
CHAPTER 17. PHYSICS 110A-B EXAMS
[3] Two brief relativity problems:
(a) A mirror lying in the (x, y) plane moves in the z direction with speed u. A monochromatic ray of light making an angle with respect to the z axis in the laboratory
frame reflects off the moving mirror. Find (i) the angle of reflection, measured in the
[17 points]
laboratory frame, and (ii) the frequency of the reflected light.
SOLUTION : The reflection is simplest to consider in the frame of the mirror, where pz
pz
upon reflection. In the laboratory frame, the 4-momentum of a photon in the beam is
P = E , 0 , E sin , E cos ,
where, without loss of generality, we have taken the light ray to lie in the (y, z) plane, and
where we are taking c = 1. Lorentz transforming to the frame of the mirror, we have
P = E(1 u cos ) , 0 , E sin , E(u + cos ) .
which follows from the general Lorentz boost of a 4-vector Q ,
0 = Q0 uQ
Q
k
= uQ0 + Q
Q
k
k
=Q ,
Q
moves with velocity u with respect to frame K.
where frame K
Upon reflection, we reverse the sign of P 3 in the frame of the mirror:
P = E(1 u cos ) , 0 , E sin , E(u cos ) .
Transforming this back to the laboratory frame yields
E = P 0 = 2 E (1 u cos ) + 2 E u (u cos )
= 2 E 1 2u cos + u2
P 1 = 0
P 2 = E sin
P 3 = 2 E u (1 u cos ) + 2 E (u cos )
= 2 E (1 + u2 ) cos 2u
Thus, the angle of reflection is
3
P (1 + u2 ) cos 2u
cos = 0 =
P
1 2u cos + u2
17.8. W08 PHYSICS 110B FINAL EXAM
437
and the reflected photon frequency is = E /h, where
E =
1 2u cos + u2
E
1 u2
(b) Consider the reaction + + n K+ + 0 . What is the threshold kinetic energy of the
pion to create kaon at an angle of 90 in the rest frame of the neutron? Express your
answer in terms of the masses m , mn , mK , and m .
[16 points]
SOLUTION : We have conservation of 4-momentum, giving
P + Pn = PK + P .
Thus,
P2 = (E + En EK )2 (P + Pn PK )2
2
= (E2 P2 ) + (En2 Pn2 ) + (EK
PK2 )
+ 2E En 2E EK 2En EK 2P Pn + 2P PK + 2Pn PK
= E2 P2 = m2 .
Now in the laboratory frame the neutron is at rest, so
Pn = (mn , 0) .
Thus, P Pn = Pn PK = 0. We are also told that the pion and the kaon make an angle
of 90 in the laboratory frame, hence P PK = 0. And of course for each particle we have
E 2 P2 = m2 . Thus, we have
m2 = m2 + m2n + m2K 2mn EK + 2(mn EK ) E ,
or, solving for E ,
E =
m2 m2 m2n m2K + 2mn EK
2(mn EK )
The threshold pion energy is the minimum value of E , which must occur when EK takes
its minimum allowed value, EK = mK . Thus,
T = E m
m2 m2 m2n m2K + 2mn mK
m
2(mn mK )
438
CHAPTER 17. PHYSICS 110A-B EXAMS
[4] Sketch what a bletch might look like.
[10,000 quatloos extra credit]
[-50 points if it looks like your professor]
Figure 17.24: The putrid bletch, from the (underwater) Jkroo forest, on planet Barney.
You might also like
- Leinaas J. Classical Mechanics and Electrodynamics 2ed 2024No ratings yetLeinaas J. Classical Mechanics and Electrodynamics 2ed 2024424 pages
- Optimal Control Engineering Matlab: - V-PublishersNo ratings yetOptimal Control Engineering Matlab: - V-Publishers4 pages
- Lecture Notes On Classical Mechanics (PDFDrive)No ratings yetLecture Notes On Classical Mechanics (PDFDrive)463 pages
- Master's Thesis - Quaternion-Based Control For Aggressive Trajectory Tracking With A Micro-Quadrotor UAVNo ratings yetMaster's Thesis - Quaternion-Based Control For Aggressive Trajectory Tracking With A Micro-Quadrotor UAV113 pages
- Partial Differential Equations: Analytical and Numerical MethodsNo ratings yetPartial Differential Equations: Analytical and Numerical Methods136 pages
- Introduction To Quantum Field Theory: Marina Von Steinkirch State University of New York at Stony Brook March 3, 2011No ratings yetIntroduction To Quantum Field Theory: Marina Von Steinkirch State University of New York at Stony Brook March 3, 2011121 pages
- Analytical Classical Dynamics An Intermediate Level Course PDFNo ratings yetAnalytical Classical Dynamics An Intermediate Level Course PDF315 pages
- Solving ODEs With Matlab Instructors ManualNo ratings yetSolving ODEs With Matlab Instructors Manual35 pages
- Classical and Quantum Mechanics of The Damped Harmonic Oscillator - DekkerNo ratings yetClassical and Quantum Mechanics of The Damped Harmonic Oscillator - Dekker110 pages
- Elementary Treatise On Rigid Body MechanicsNo ratings yetElementary Treatise On Rigid Body Mechanics396 pages
- Classical Mechanics Probs Solns 01 Dewar P103 PirxNo ratings yetClassical Mechanics Probs Solns 01 Dewar P103 Pirx103 pages
- An Introduction To FISIKA STATISTIK WORDNo ratings yetAn Introduction To FISIKA STATISTIK WORD396 pages
- Classical and Quantum Mechanics of The Damped Harmonic Oscillator (Physrep80 (1981) 1)No ratings yetClassical and Quantum Mechanics of The Damped Harmonic Oscillator (Physrep80 (1981) 1)110 pages
- Differential Geometry Syed Hassan Waqas PDF100% (1)Differential Geometry Syed Hassan Waqas PDF69 pages
- Classical Physics With Geometric AlgebraNo ratings yetClassical Physics With Geometric Algebra116 pages
- Differential Geometry: Computer Vision #8No ratings yetDifferential Geometry: Computer Vision #849 pages
- Design of Feedback Control Systems 4th eNo ratings yetDesign of Feedback Control Systems 4th e860 pages
- Physical and Computational Aspects of Convective Heat Transfer - CompressNo ratings yetPhysical and Computational Aspects of Convective Heat Transfer - Compress496 pages
- Denn Optimization by Variational MethodsNo ratings yetDenn Optimization by Variational Methods431 pages
- Introductory Applications of Partial Differential Equations: With Emphasis on Wave Propagation and DiffusionFrom EverandIntroductory Applications of Partial Differential Equations: With Emphasis on Wave Propagation and DiffusionNo ratings yet
- 1.2 Continuacion Resumen Principios FundamentalesNo ratings yet1.2 Continuacion Resumen Principios Fundamentales453 pages
- Shen S.-Q. Lecture Notes On Quantum Mechanics (Hong Kong, Web Draft, 2004) (265s) - PQMNo ratings yetShen S.-Q. Lecture Notes On Quantum Mechanics (Hong Kong, Web Draft, 2004) (265s) - PQM265 pages
- Lecture Notes On Nonlinear Dynamics PDFNo ratings yetLecture Notes On Nonlinear Dynamics PDF345 pages
- University Physics I - Classical Mechanics100% (1)University Physics I - Classical Mechanics342 pages
- Thesis Architectural Intelligence - Bo Westerlinck - 2MA AR (Compressed)100% (1)Thesis Architectural Intelligence - Bo Westerlinck - 2MA AR (Compressed)152 pages
- Animation: Where Codes Collide: The Emergent Ecology ofNo ratings yetAnimation: Where Codes Collide: The Emergent Ecology of15 pages
- Instance Normalization: The Missing Ingredient For Fast StylizationNo ratings yetInstance Normalization: The Missing Ingredient For Fast Stylization6 pages
- Controlling Perceptual Factors in Neural Style TransferNo ratings yetControlling Perceptual Factors in Neural Style Transfer9 pages
- Gatys Image Style Transfer CVPR 2016 PaperNo ratings yetGatys Image Style Transfer CVPR 2016 Paper10 pages
- Anterior Cross Bites in Primary Mixed Dentition PedoNo ratings yetAnterior Cross Bites in Primary Mixed Dentition Pedo32 pages
- Luciferian Witchcraft - The Mystery RevealedNo ratings yetLuciferian Witchcraft - The Mystery Revealed3 pages
- Dedekind'S Proof of Euler'S Reflection Formula Via OdesNo ratings yetDedekind'S Proof of Euler'S Reflection Formula Via Odes3 pages
- Net-Analyser Na96 and Na96+: General InformationNo ratings yetNet-Analyser Na96 and Na96+: General Information31 pages
- (Ebook) Julius Caesar and the transformation of the Roman Republic by Tom Stevenson ISBN 9781315746173, 1315746174 - Download the ebook with all fully detailed chapters100% (1)(Ebook) Julius Caesar and the transformation of the Roman Republic by Tom Stevenson ISBN 9781315746173, 1315746174 - Download the ebook with all fully detailed chapters47 pages
- Costache - Christian Worldview: Understandings From ST Basil The Great (Cappadocian Legacy)100% (1)Costache - Christian Worldview: Understandings From ST Basil The Great (Cappadocian Legacy)34 pages
- Prologue Magazine - Thanksgiving: Another FDR ExperimentNo ratings yetPrologue Magazine - Thanksgiving: Another FDR Experiment7 pages
- Download Complete (Ebook) Rapid Quantitative Aptitude - With Shortcuts & Tricks for Competitive Exams by Deepak Agarwal; D.P. Gupta PDF for All Chapters100% (4)Download Complete (Ebook) Rapid Quantitative Aptitude - With Shortcuts & Tricks for Competitive Exams by Deepak Agarwal; D.P. Gupta PDF for All Chapters67 pages
- Leinaas J. Classical Mechanics and Electrodynamics 2ed 2024Leinaas J. Classical Mechanics and Electrodynamics 2ed 2024
- Optimal Control Engineering Matlab: - V-PublishersOptimal Control Engineering Matlab: - V-Publishers
- Master's Thesis - Quaternion-Based Control For Aggressive Trajectory Tracking With A Micro-Quadrotor UAVMaster's Thesis - Quaternion-Based Control For Aggressive Trajectory Tracking With A Micro-Quadrotor UAV
- Partial Differential Equations: Analytical and Numerical MethodsPartial Differential Equations: Analytical and Numerical Methods
- Introduction To Quantum Field Theory: Marina Von Steinkirch State University of New York at Stony Brook March 3, 2011Introduction To Quantum Field Theory: Marina Von Steinkirch State University of New York at Stony Brook March 3, 2011
- Analytical Classical Dynamics An Intermediate Level Course PDFAnalytical Classical Dynamics An Intermediate Level Course PDF
- Classical and Quantum Mechanics of The Damped Harmonic Oscillator - DekkerClassical and Quantum Mechanics of The Damped Harmonic Oscillator - Dekker
- Classical Mechanics Probs Solns 01 Dewar P103 PirxClassical Mechanics Probs Solns 01 Dewar P103 Pirx
- Classical and Quantum Mechanics of The Damped Harmonic Oscillator (Physrep80 (1981) 1)Classical and Quantum Mechanics of The Damped Harmonic Oscillator (Physrep80 (1981) 1)
- Physical and Computational Aspects of Convective Heat Transfer - CompressPhysical and Computational Aspects of Convective Heat Transfer - Compress
- Introductory Applications of Partial Differential Equations: With Emphasis on Wave Propagation and DiffusionFrom EverandIntroductory Applications of Partial Differential Equations: With Emphasis on Wave Propagation and Diffusion
- Shen S.-Q. Lecture Notes On Quantum Mechanics (Hong Kong, Web Draft, 2004) (265s) - PQMShen S.-Q. Lecture Notes On Quantum Mechanics (Hong Kong, Web Draft, 2004) (265s) - PQM
- Thesis Architectural Intelligence - Bo Westerlinck - 2MA AR (Compressed)Thesis Architectural Intelligence - Bo Westerlinck - 2MA AR (Compressed)
- Animation: Where Codes Collide: The Emergent Ecology ofAnimation: Where Codes Collide: The Emergent Ecology of
- Instance Normalization: The Missing Ingredient For Fast StylizationInstance Normalization: The Missing Ingredient For Fast Stylization
- Controlling Perceptual Factors in Neural Style TransferControlling Perceptual Factors in Neural Style Transfer
- Anterior Cross Bites in Primary Mixed Dentition PedoAnterior Cross Bites in Primary Mixed Dentition Pedo
- Dedekind'S Proof of Euler'S Reflection Formula Via OdesDedekind'S Proof of Euler'S Reflection Formula Via Odes
- (Ebook) Julius Caesar and the transformation of the Roman Republic by Tom Stevenson ISBN 9781315746173, 1315746174 - Download the ebook with all fully detailed chapters(Ebook) Julius Caesar and the transformation of the Roman Republic by Tom Stevenson ISBN 9781315746173, 1315746174 - Download the ebook with all fully detailed chapters
- Costache - Christian Worldview: Understandings From ST Basil The Great (Cappadocian Legacy)Costache - Christian Worldview: Understandings From ST Basil The Great (Cappadocian Legacy)
- Prologue Magazine - Thanksgiving: Another FDR ExperimentPrologue Magazine - Thanksgiving: Another FDR Experiment
- Download Complete (Ebook) Rapid Quantitative Aptitude - With Shortcuts & Tricks for Competitive Exams by Deepak Agarwal; D.P. Gupta PDF for All ChaptersDownload Complete (Ebook) Rapid Quantitative Aptitude - With Shortcuts & Tricks for Competitive Exams by Deepak Agarwal; D.P. Gupta PDF for All Chapters