(Baker M., Sutlief S.) Green's Functions in Physics

Greens Functions in Physics
Version 1
M. Baker, S. Sutlief
Revision:
December 19, 2003
Contents
1 The Vibrating String
1.1 The String . . . . . . . . . . . . . . . . . . . . . .
1.1.1 Forces on the String . . . . . . . . . . . .
1.1.2 Equations of Motion for a Massless String
1.1.3 Equations of Motion for a Massive String .
1.2 The Linear Operator Form . . . . . . . . . . . . .
1.3 Boundary Conditions . . . . . . . . . . . . . . . .
1.3.1 Case 1: A Closed String . . . . . . . . . .
1.3.2 Case 2: An Open String . . . . . . . . . .
1.3.3 Limiting Cases . . . . . . . . . . . . . . .
1.3.4 Initial Conditions . . . . . . . . . . . . . .
1.4 Special Cases . . . . . . . . . . . . . . . . . . . .
1.4.1 No Tension at Boundary . . . . . . . . . .
1.4.2 Semi-infinite String . . . . . . . . . . . . .
1.4.3 Oscillatory External Force . . . . . . . . .
1.5 Summary . . . . . . . . . . . . . . . . . . . . . .
1.6 References . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
1
2
2
3
4
5
5
6
6
7
8
8
9
9
9
10
11
2 Greens Identities
2.1 Greens 1st and 2nd Identities . . . . . . .
2.2 Using G.I. #2 to Satisfy R.B.C. . . . . . .
2.2.1 The Closed String . . . . . . . . . .
2.2.2 The Open String . . . . . . . . . .
2.2.3 A Note on Hermitian Operators . .
2.3 Another Boundary Condition . . . . . . .
2.4 Physical Interpretations of the G.I.s . . . .
2.4.1 The Physics of Greens 2nd Identity
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
13
14
15
15
16
17
17
18
18
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
ii
CONTENTS
2.5
2.6
2.4.2 A Note on Potential Energy . . . .

2.4.3 The Physics of Greens 1st Identity
Summary . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . .
3 Greens Functions
3.1 The Principle of Superposition .
3.2 The Dirac Delta Function . . .
3.3 Two Conditions . . . . . . . . .
3.3.1 Condition 1 . . . . . . .
3.3.2 Condition 2 . . . . . . .
3.3.3 Application . . . . . . .
3.4 Open String . . . . . . . . . . .
3.5 The Forced Oscillation Problem
3.6 Free Oscillation . . . . . . . . .
3.7 Summary . . . . . . . . . . . .
3.8 Reference . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4 Properties of Eigen States

4.1 Eigen Functions and Natural Modes . . . . . . . .
4.1.1 A Closed String Problem . . . . . . . . . .
4.1.2 The Continuum Limit . . . . . . . . . . .
4.1.3 Schrodingers Equation . . . . . . . . . . .
4.2 Natural Frequencies and the Greens Function . .
4.3 GF behavior near = n . . . . . . . . . . . . . .
4.4 Relation between GF & Eig. Fn. . . . . . . . . . .
4.4.1 Case 1: Nondegenerate . . . . . . . . . .
4.4.2 Case 2: n Double Degenerate . . . . . . .
4.5 Solution for a Fixed String . . . . . . . . . . . . .
4.5.1 A Non-analytic Solution . . . . . . . . . .
4.5.2 The Branch Cut . . . . . . . . . . . . . . .
4.5.3 Analytic Fundamental Solutions and GF .
4.5.4 Analytic GF for Fixed String . . . . . . .
4.5.5 GF Properties . . . . . . . . . . . . . . . .
4.5.6 The GF Near an Eigenvalue . . . . . . . .
4.6 Derivation of GF form near E.Val. . . . . . . . . .
4.6.1 Reconsider the Gen. Self-Adjoint Problem
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
18
19
20
21
.
.
.
.
.
.
.
.
.
.
.
23
23
24
28
28
28
28
29
31
32
32
34
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
35
37
37
38
39
40
41
42
43
44
45
45
46
46
47
49
50
51
51
CONTENTS
4.7
4.8
4.9
4.10
4.11
4.12
4.13
4.14
iii
4.6.2 Summary, Interp. & Asymptotics . .

General Solution form of GF . . . . . . . . .
4.7.1 -fn Representations & Completeness
Extension to Continuous Eigenvalues . . . .
Orthogonality for Continuum . . . . . . . .
Example: Infinite String . . . . . . . . . . .
4.10.1 The Greens Function . . . . . . . . .
4.10.2 Uniqueness . . . . . . . . . . . . . .
4.10.3 Look at the Wronskian . . . . . . . .
4.10.4 Solution . . . . . . . . . . . . . . . .
4.10.5 Motivation, Origin of Problem . . . .
Summary of the Infinite String . . . . . . . .
The Eigen Function Problem Revisited . . .
Summary . . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . .
5 Steady State Problems

5.1 Oscillating Point Source . . . . . . .
5.2 The Klein-Gordon Equation . . . . .
5.2.1 Continuous Completeness . .
5.3 The Semi-infinite Problem . . . . . .
5.3.1 A Check on the Solution . . .
5.4 Steady State Semi-infinite Problem .
5.4.1 The Fourier-Bessel Transform
5.5 Summary . . . . . . . . . . . . . . .
5.6 References . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
52
53
57
58
59
62
62
64
64
65
65
67
68
69
71
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
73
73
74
76
78
80
80
82
83
84
6 Dynamic Problems
6.1 Advanced and Retarded GFs . . . . .
6.2 Physics of a Blow . . . . . . . . . . . .
6.3 Solution using Fourier Transform . . .
6.4 Inverting the Fourier Transform . . . .
6.4.1 Summary of the General IVP .
6.5 Analyticity and Causality . . . . . . .
6.6 The Infinite String Problem . . . . . .
6.6.1 Derivation of Greens Function
6.6.2 Physical Derivation . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
85
86
87
88
90
92
92
93
93
96
iv
CONTENTS
6.7
6.8
6.9
6.10
Semi-Infinite String with Fixed End . . . . .

Semi-Infinite String with Free End . . . . .
Elastically Bound Semi-Infinite String . . . .
Relation to the Eigen Fn Problem . . . . . .
6.10.1 Alternative form of the GR Problem
6.11 Comments on Greens Function . . . . . . .
6.11.1 Continuous Spectra . . . . . . . . . .
6.11.2 Neumann BC . . . . . . . . . . . . .
6.11.3 Zero Net Force . . . . . . . . . . . .
6.12 Summary . . . . . . . . . . . . . . . . . . .
6.13 References . . . . . . . . . . . . . . . . . . .
7 Surface Waves and Membranes
7.1 Introduction . . . . . . . . . . .
7.2 One Dimensional Surface Waves
7.2.1 The Physical Situation .
7.2.2 Shallow Water Case . . .
7.3 Two Dimensional Problems . .
7.3.1 Boundary Conditions . .
7.4 Example: 2D Surface Waves . .
7.5 Summary . . . . . . . . . . . .
7.6 References . . . . . . . . . . . .
. . . . . .
on Fluids
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
. . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
97
97
99
99
101
102
102
102
104
104
105
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
107
107
108
108
108
109
111
112
113
113
.
.
.
.
.
.
.
.
.
.
.
.
.
115
. 115
. 116
. 117
. 117
. 117
. 118
. 118
. 119
. 119
. 119
. 119
. 119
. 120
8 Extension to N -dimensions
8.1 Introduction . . . . . . . . . . . . . . . . . . . . .
8.2 Regions of Interest . . . . . . . . . . . . . . . . .
8.3 Examples of N -dimensional Problems . . . . . . .
8.3.1 General Response . . . . . . . . . . . . . .
8.3.2 Normal Mode Problem . . . . . . . . . . .
8.3.3 Forced Oscillation Problem . . . . . . . . .
8.4 Greens Identities . . . . . . . . . . . . . . . . . .
8.4.1 Greens First Identity . . . . . . . . . . . .
8.4.2 Greens Second Identity . . . . . . . . . .
8.4.3 Criterion for Hermitian L0 . . . . . . . . .
8.5 The Retarded Problem . . . . . . . . . . . . . . .
8.5.1 General Solution of Retarded Problem . .
8.5.2 The Retarded Greens Function in N -Dim.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CONTENTS
8.5.3 Reduction to Eigenvalue Problem
Region R . . . . . . . . . . . . . . . . .
8.6.1 Interior . . . . . . . . . . . . . .
8.6.2 Exterior . . . . . . . . . . . . . .
8.7 The Method of Images . . . . . . . . . .
8.7.1 Eigenfunction Method . . . . . .
8.7.2 Method of Images . . . . . . . . .
8.8 Summary . . . . . . . . . . . . . . . . .
8.9 References . . . . . . . . . . . . . . . . .
8.6
v
.
.
.
.
.
.
.
.
.
9 Cylindrical Problems
9.1 Introduction . . . . . . . . . . . . . . . . .
9.1.1 Coordinates . . . . . . . . . . . . .
9.1.2 Delta Function . . . . . . . . . . .
9.2 GF Problem for Cylindrical Sym. . . . . .
9.3 Expansion in Terms of Eigenfunctions . . .
9.3.1 Partial Expansion . . . . . . . . . .
9.3.2 Summary of GF for Cyl. Sym. . . .
9.4 Eigen Value Problem for L0 . . . . . . . .
9.5 Uses of the GF Gm (r, r0 ; ) . . . . . . . . .
9.5.1 Eigenfunction Problem . . . . . . .
9.5.2 Normal Modes/Normal Frequencies
9.5.3 The Steady State Problem . . . . .
9.5.4 Full Time Dependence . . . . . . .
9.6 The Wedge Problem . . . . . . . . . . . .
9.6.1 General Case . . . . . . . . . . . .
9.6.2 Special Case: Fixed Sides . . . . .
9.7 The Homogeneous Membrane . . . . . . .
9.7.1 The Radial Eigenvalues . . . . . . .
9.7.2 The Physics . . . . . . . . . . . . .
9.8 Summary . . . . . . . . . . . . . . . . . .
9.9 Reference . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
121
122
122
122
122
123
123
124
125
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
127
. 127
. 128
. 129
. 130
. 131
. 131
. 132
. 133
. 134
. 134
. 134
. 135
. 136
. 136
. 137
. 138
. 138
. 140
. 141
. 141
. 142
10 Heat Conduction
143
10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 143
10.1.1 Conservation of Energy . . . . . . . . . . . . . . . 143
10.1.2 Boundary Conditions . . . . . . . . . . . . . . . . 145
vi
CONTENTS
10.2 The Standard form of the Heat Eq. . . . . . . .
10.2.1 Correspondence with the Wave Equation
10.2.2 Greens Function Problem . . . . . . . .
10.2.3 Laplace Transform . . . . . . . . . . . .
10.2.4 Eigen Function Expansions . . . . . . . .
10.3 Explicit One Dimensional Calculation . . . . . .
10.3.1 Application of Transform Method . . . .
10.3.2 Solution of the Transform Integral . . . .
10.3.3 The Physics of the Fundamental Solution
10.3.4 Solution of the General IVP . . . . . . .
10.3.5 Special Cases . . . . . . . . . . . . . . .
10.4 Summary . . . . . . . . . . . . . . . . . . . . .
10.5 References . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
159
. 160
. 162
. 164
. 164
. 165
. 169
. 170
. 171
. 172
. 172
. 173
. 174
. 177
. 177
. 178
. 180
. 181
12 Steady State Scattering

12.1 Spherical Waves . . . . . . . . . . . . . . . . . . . . . .
12.2 Plane Waves . . . . . . . . . . . . . . . . . . . . . . . .
12.3 Relation to Potential Theory . . . . . . . . . . . . . . .
183
. 183
. 185
. 186
11 Spherical Symmetry
11.1 Spherical Coordinates . . . . . . . . . . . . . . . .
11.2 Discussion of L . . . . . . . . . . . . . . . . . .
11.3 Spherical Eigenfunctions . . . . . . . . . . . . . .
11.3.1 Reduced Eigenvalue Equation . . . . . . .
11.3.2 Determination of um
l (x) . . . . . . . . . .
11.3.3 Orthogonality and Completeness of um
l (x)
11.4 Spherical Harmonics . . . . . . . . . . . . . . . .
11.4.1 Othonormality and Completeness of Ylm .
11.5 GFs for Spherical Symmetry . . . . . . . . . . .
11.5.1 GF Differential Equation . . . . . . . . . .
11.5.2 Boundary Conditions . . . . . . . . . . . .
11.5.3 GF for the Exterior Problem . . . . . . . .
11.6 Example: Constant Parameters . . . . . . . . . .
11.6.1 Exterior Problem . . . . . . . . . . . . . .
11.6.2 Free Space Problem . . . . . . . . . . . . .
11.7 Summary . . . . . . . . . . . . . . . . . . . . . .
11.8 References . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
146
146
146
147
148
150
151
151
154
154
155
156
157
CONTENTS
vii
12.4 Scattering from a Cylinder . . . . . . . . . . . . . . . . . 189

12.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 190
12.6 References . . . . . . . . . . . . . . . . . . . . . . . . . . 190
13 Kirchhoff s Formula
191
13.1 References . . . . . . . . . . . . . . . . . . . . . . . . . . 194
14 Quantum Mechanics
14.1 Quantum Mechanical Scattering . . . .
14.2 Plane Wave Approximation . . . . . .
14.3 Quantum Mechanics . . . . . . . . . .
14.4 Review . . . . . . . . . . . . . . . . . .
14.5 Spherical Symmetry Degeneracy . . . .
14.6 Comparison of Classical and Quantum
14.7 Summary . . . . . . . . . . . . . . . .
14.8 References . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
195
. 197
. 199
. 200
. 201
. 202
. 202
. 204
. 204
15 Scattering in 3-Dim
15.1 Angular Momentum . . . . . . . . . . . . . . . . . . .
15.2 Far-Field Limit . . . . . . . . . . . . . . . . . . . . . .
15.3 Relation to the General Propagation Problem . . . . .
15.4 Simplification of Scattering Problem . . . . . . . . . .
15.5 Scattering Amplitude . . . . . . . . . . . . . . . . . . .
15.6 Kinematics of Scattered Waves . . . . . . . . . . . . .
15.7 Plane Wave Scattering . . . . . . . . . . . . . . . . . .
15.8 Special Cases . . . . . . . . . . . . . . . . . . . . . . .
15.8.1 Homogeneous Source; Inhomogeneous Observer
15.8.2 Homogeneous Observer; Inhomogeneous Source
15.8.3 Homogeneous Source; Homogeneous Observer .
15.8.4 Both Points in Interior Region . . . . . . . . . .
15.8.5 Summary . . . . . . . . . . . . . . . . . . . . .
15.8.6 Far Field Observation . . . . . . . . . . . . . .
15.8.7 Distant Source: r0 . . . . . . . . . . . . .
15.9 The Physical significance of Xl . . . . . . . . . . . . . .
15.9.1 Calculating l (k) . . . . . . . . . . . . . . . . .
15.10Scattering from a Sphere . . . . . . . . . . . . . . . . .
15.10.1 A Related Problem . . . . . . . . . . . . . . . .
205
. 207
. 208
. 210
. 210
. 211
. 212
. 213
. 214
. 214
. 215
. 216
. 217
. 218
. 218
. 219
. 219
. 222
. 223
. 224
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
viii
CONTENTS
15.11Calculation of Phase for a Hard Sphere . .
15.12Experimental Measurement . . . . . . . .
15.12.1 Cross Section . . . . . . . . . . . .
15.12.2 Notes on Cross Section . . . . . . .
15.12.3 Geometrical Limit . . . . . . . . .
15.13Optical Theorem . . . . . . . . . . . . . .
15.14Conservation of Probability Interpretation:
15.14.1 Hard Sphere . . . . . . . . . . . . .
15.15Radiation of Sound Waves . . . . . . . . .
15.15.1 Steady State Solution . . . . . . . .
15.15.2 Far Field Behavior . . . . . . . . .
15.15.3 Special Case . . . . . . . . . . . . .
15.15.4 Energy Flux . . . . . . . . . . . . .
15.15.5 Scattering From Plane Waves . . .
15.15.6 Spherical Symmetry . . . . . . . .
15.16Summary . . . . . . . . . . . . . . . . . .
15.17References . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
225
226
227
229
230
231
231
231
232
234
235
236
237
240
241
242
243
16 Heat Conduction in 3D
16.1 General Boundary Value Problem
16.2 Time Dependent Problem . . . .
16.3 Evaluation of the Integrals . . . .
16.4 Physics of the Heat Problem . . .
16.4.1 The Parameter . . . . .
16.5 Example: Sphere . . . . . . . . .
16.5.1 Long Times . . . . . . . .
16.5.2 Interior Case . . . . . . .
16.6 Summary . . . . . . . . . . . . .
16.7 References . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
245
245
247
248
251
251
252
253
254
255
256
17 The Wave Equation

17.1 introduction . . . . . . .
17.2 Dimensionality . . . . .
17.2.1 Odd Dimensions
17.2.2 Even Dimensions
17.3 Physics . . . . . . . . . .
17.3.1 Odd Dimensions
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
257
257
259
259
260
260
260
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CONTENTS
17.3.2 Even Dimensions . . . . . . .
17.3.3 Connection between GFs in 2
17.4 Evaluation of G2 . . . . . . . . . . .
17.5 Summary . . . . . . . . . . . . . . .
17.6 References . . . . . . . . . . . . . . .
18 The
18.1
18.2
18.3
18.4
ix
. . . . .
& 3-dim
. . . . .
. . . . .
. . . . .
Method of Steepest Descent

Review of Complex Variables . . . . . .
Specification of Steepest Descent . . . .
Inverting a Series . . . . . . . . . . . . .
Example 1: Expansion of function . .
18.4.1 Transforming the Integral . . . .
18.4.2 The Curve of Steepest Descent .
18.5 Example 2: Asymptotic Hankel Function
18.6 Summary . . . . . . . . . . . . . . . . .
18.7 References . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19 High Energy Scattering

19.1 Fundamental Integral Equation of Scattering
19.2 Formal Scattering Theory . . . . . . . . . .
19.2.1 A short digression on operators . . .
19.3 Summary of Operator Method . . . . . . . .
19.3.1 Derivation of G = (E H)1 . . . .
19.3.2 Born Approximation . . . . . . . . .
19.4 Physical Interest . . . . . . . . . . . . . . .
19.4.1 Satisfying the Scattering Condition .
19.5 Physical Interpretation . . . . . . . . . . . .
19.6 Probability Amplitude . . . . . . . . . . . .
19.7 Review . . . . . . . . . . . . . . . . . . . . .
19.8 The Born Approximation . . . . . . . . . . .
19.8.1 Geometry . . . . . . . . . . . . . . .
19.8.2 Spherically Symmetric Case . . . . .
19.8.3 Coulomb Case . . . . . . . . . . . . .
19.9 Scattering Approximation . . . . . . . . . .
19.10Perturbation Expansion . . . . . . . . . . .
19.10.1 Perturbation Expansion . . . . . . .
19.10.2 Use of the T -Matrix . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
260
261
263
264
264
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
265
266
269
270
273
273
274
276
280
280
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
281
. 283
. 285
. 287
. 288
. 289
. 289
. 290
. 291
. 292
. 292
. 293
. 294
. 296
. 296
. 297
. 298
. 299
. 300
. 301
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
CONTENTS
19.11Summary . . . . . . . . . . . . . . . . . . . . . . . . . . 302
19.12References . . . . . . . . . . . . . . . . . . . . . . . . . . 302
A Symbols Used
303
List of Figures
1.1 A string with mass points attached to springs. . . . . . .
1.2 A closed string, where a and b are connected. . . . . . .
1.3 An open string, where the endpoints a and b are free. . .
2
6
7
3.1
The pointed string . . . . . . . . . . . . . . . . . . . . . 27
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
The closed string with discrete

Negative energy levels . . . .
The -convention . . . . . . .
The contour of integration . .
Circle around a singularity. . .
Division of contour. . . . . . .
near the branch cut. . . . .
specification. . . . . . . . .
Geometry in -plane . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
37
40
46
54
55
56
61
63
69
6.1 The contour L in the -plane. . . . . . . . . . .

6.2 Contour LC1 = L + LU HP closed in UH -plane.
6.3 Contour closed in the lower half -plane. . . . .
6.4 An illustration of the retarded Greens Function.
6.5 GR at t1 = t0 + 21 x0 /c and at t2 = t0 + 32 x0 /c. . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
92
93
95
96
98
7.1
7.2
mass points.
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
. . . . . . . .
.
.
.
.
.
.
.
.
.
Water waves moving in channels. . . . . . . . . . . . . . 108

The rectangular membrane. . . . . . . . . . . . . . . . . 111
9.1 The region R as a circle with radius a. . . . . . . . . . . 130

9.2 The wedge. . . . . . . . . . . . . . . . . . . . . . . . . . 137
10.1 Rotation of contour in complex plane. . . . . . . . . . . . 148
xi
xii
LIST OF FIGURES
10.2 Contour closed in left half s-plane. . . . . . . . . . . . . 149
10.3 A contour with Branch cut. . . . . . . . . . . . . . . . . 152
11.1 Spherical Coordinates. . . . . . . . . . . . . . . . . . . . 160
11.2 The general boundary for spherical symmetry. . . . . . . 174
12.1 Waves scattering from an obstacle. . . . . . . . . . . . . 184
12.2 Definition of and .. . . . . . . . . . . . . . . . . . . . 186
13.1 A screen with a hole in it. . . . . . . . . . . . . . . . . . 192
13.2 The source and image source. . . . . . . . . . . . . . . . 193
13.3 Configurations for the Gs. . . . . . . . . . . . . . . . . . 194
14.1 An attractive potential. . . . . . . . . . . . . . . . . . . . 196
14.2 The complex energy plane. . . . . . . . . . . . . . . . . . 197
15.1
15.2
15.3
15.4
15.5
15.6
15.7
The schematic representation of a scattering experiment.

The geometry defining and . . . . . . . . . . . . . . .
Phase shift due to potential. . . . . . . . . . . . . . . . .
A repulsive potential. . . . . . . . . . . . . . . . . . . . .
The potential V and Veff for a particular example. . . . .
An infinite potential wall. . . . . . . . . . . . . . . . . .
Scattering with a strong forward peak. . . . . . . . . . .
208
212
221
223
225
227
232
16.1 Closed contour around branch cut. . . . . . . . . . . . . 250

17.1 Radial part of the 2-dimensional Greens function. . . . . 261
17.2 A line source in 3-dimensions. . . . . . . . . . . . . . . . 263
18.1
18.2
18.3
18.4
18.5
18.6
Contour C & deformation C0 with point z0 .

Gradients of u and v. . . . . . . . . . . . . .
f (z) near a saddle-point. . . . . . . . . . . .
Defining Contour for the Hankel function. .
Deformed contour for the Hankel function. .
Hankel function contours. . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
266
267
268
277
278
280
19.1 Geometry of the scattered wave vectors. . . . . . . . . . 296
Preface
This manuscript is based on lectures given by Marshall Baker for a class
on Mathematical Methods in Physics at the University of Washington
in 1988. The subject of the lectures was Greens function techniques in
Physics. All the members of the class had completed the equivalent of
the first three and a half years of the undergraduate physics program,
although some had significantly more background. The class was a
preparation for graduate study in physics.
These notes develop Greens function techiques for both single and
multiple dimension problems, and then apply these techniques to solving the wave equation, the heat equation, and the scattering problem.
Many other mathematical techniques are also discussed.
To read this manuscript it is best to have Arfkens book handy
for the mathematics details and Fetter and Waleckas book handy for
the physics details. There are other good books on Greens functions
available, but none of them are geared for same background as assumed
here. The two volume set by Stakgold is particularly useful. For a
strictly mathematical discussion, the book by Dennery is good.
Here are some notes and warnings about this revision:
Text This text is an amplification of lecture notes taken of the
Physics 425-426 sequence. Some sections are still a bit rough. Be
alert for errors and omissions.
List of Symbols A listing of mostly all the variables used is included. Be warned that many symbols are created ad hoc, and
thus are only used in a particular section.
Bibliography The bibliography includes those books which have
been useful to Steve Sutlief in creating this manuscript, and were
xiii
xiv
LIST OF FIGURES
not necessarily used for the development of the original lectures.
Books marked with an asterisk are are more supplemental. Comments on the books listed are given above.
Index The index was composed by skimming through the text
and picking out places where ideas were introduced or elaborated
upon. No attempt was made to locate all relevant discussions for
each idea.
A Note About Copying:

These notes are in a state of rapid transition and are provided so as
to be of benefit to those who have recently taken the class. Therefore,
please do not photocopy these notes.
Contacting the Authors:
A list of phone numbers and email addresses will be maintained of
those who wish to be notified when revisions become available. If you
would like to be on this list, please send email to
sutlief@u.washington.edu
before 1996. Otherwise, call Marshall Baker at 206-543-2898.
Acknowledgements:
This manuscript benefits greatly from the excellent set of notes
taken by Steve Griffies. Richard Horn contributed many corrections
and suggestions. Special thanks go to the students of Physics 425-426
at the University of Washington during 1988 and 1993.
This first revision contains corrections only. No additional material
has been added since Version 0.
Steve Sutlief
Seattle, Washington
16 June, 1993
4 January, 1994
Chapter 1
The Vibrating String
4 Jan p1
p1prv.yr.
Chapter Goals:
Construct the wave equation for a string by identifying forces and using Newtons second law.
Determine boundary conditions appropriate for a
closed string, an open string, and an elastically
bound string.
Determine the wave equation for a string subject to
an external force with harmonic time dependence.
The central topic under consideration is the branch of differential equation theory containing boundary value problems. First we look at an pr:bvp1
example of the application of Newtons second law to small vibrations:
transverse vibrations on a string. Physical problems such as this and
those involving sound, surface waves, heat conduction, electromagnetic
waves, and gravitational waves, for example, can be solved using the
mathematical theory of boundary value problems.
Consider the problem of a string embedded in a medium with a pr:string1
restoring force V (x) and an external force F (x, t). This problem covers pr:V1
most of the physical interpretations of small vibrations. In this chapter pr:F1
we will investigate the mathematics of this problem by determining the
equations of motion.
1
CHAPTER 1. THE VIBRATING STRING
ui+1
ui
ui1
mi
mi+1
i+1
Fiy !u
X
X

!
!!
!
u
X
X

Fiyi""!

X
X

X
X

X
mi1 ""
X

X
X

"

X
X

u
"

X
X
P
P

#P
X
X
P

X
X
# P
P

X
X ki
P
P

k
#

i1
X
X
P
P

xi1
xi
X
X

X
X

X
X

X
X

X
X

X
X

X
X

X
X

X
X

X
X

X
X

X
X

X
X

k

X
X
i+1

xi+1
Figure 1.1: A string with mass points attached to springs.
1.1
pr:N1
pr:mi1
pr:tau1
fig1.1
pr:eom1
The String
We consider a massless string with equidistant mass points attached. In

the case of a string, we shall see (in chapter 3) that the Greens function
corresponds to an impulsive force and is represented by a complete set
of functions. Consider N mass points of mass mi attached to a massless
string, which has a tension between mass points. An elastic force at
each mass point is represented by a spring. This problem is illustrated
in figure 1.1 We want to find the equations of motion for transverse
vibrations of the string.
1.1.1
Forces on the String
For the massless vibrating string, there are three forces which are included in the equation of motion. These forces are the tension force,
elastic force, and external force.
4 Jan p2
pr:tension1
pr:ui1
pr:a1
pr:transvib1
Tension Force
For each mass point there are two force contributions due to the tension
on the string. We call i the tension on the segment between mi1
and mi , ui the vertical displacement of the ith mass point, and a the
horizontal displacement between mass points. Since we are considering
transverse vibrations (in the u-direction) , we want to know the tension
1.1. THE STRING
force in the u-direction, which is i+1 sin . From the figure we see that pr:theta1
(ui+1 ui )/a for small angles and we can thus write
Fiyi+1 = i+1
(ui+1 ui )
a
and
(ui ui1 )
Fiyi = i
.
a
Note that the equations agree with dimensional analysis:
Fiyi = dim(m l/t2 ),
ui = dim(l),
i = dim(m l/t2 ),
and
a = dim(l).
Elastic Force
We add an elastic force with spring constant ki :
Fielastic
pr:Fiyt1
Grifs
uses
Taylor exp
pr:m1
pr:l1
pr:t1
pr:elastic1
pr:ki1
= ki ui ,
where dim(ki ) = (m/t2 ). This situation can be visualized by imagining pr:Fel1

vertical springs attached to each mass point, as depicted in figure 1.1.
A small value of ki corresponds to an elastic spring, while a large value
of ki corresponds to a rigid spring.
External Force
We add the external force Fiext . This force depends on the nature of pr:ExtForce1
the physical problem under consideration. For example, it may be a pr:Fext1
transverse force at the end points.
1.1.2
Equations of Motion for a Massless String
The problem thus far has concerned a massless string with mass points
attached. By summing the above forces and applying Newtons second
law, we have
pr:Newton1
(ui+1 ui )
(ui ui1 )
d2
i
ki ui + Fiext = mi 2 ui . (1.1)
a
a
dt
This gives us N coupled inhomogeneous linear ordinary differential
equations where each ui is a function of time. In the case that Fiext
is zero we have free vibration, otherwise we have forced vibration.
Ftot = i+1
pr:t2
eq1force
pr:diffeq1
pr:FreeVib1
pr:ForcedVib1
1.1.3
Equations of Motion for a Massive String
4 Jan p3
pr:deltax1
pr:deltau1
For a string with continuous mass density, the equidistant mass points
on the string are replaced by a continuum. First we take a, the separation distance between mass points, to be small and redefine it as
a = x. We correspondingly write ui ui1 = u. This allows us to
write

(ui ui1 )
u
=
.
(1.2)
a
x i
The equations of motion become (after dividing both sides by x)
"
1
u
i+1
x
x
eq1deltf
i+1
u
x
#
ki
F ext
mi d2 ui
ui + i =
.
x
x
x dt2
In the limit we take a 0, N , and define their product to be

lim N a L.
a0
N
pr:sigmax1
(1.3)
(1.4)
The limiting case allows us to redefine the terms of the equations of

motion as follows:
mi 0
ki 0
Fiext 0
mi
x
ki
x
F ext
x
(xi )
mass
length
= mass density;
V (xi ) = coefficient of elasticity of the media;

F ext
)
mi
mi
= ( x
(xi )f (xi )
(1.5)
where
f (xi ) =
F ext
external force
=
.
mi
mass
(1.6)
Since
xi = x
xi1 = x x
xi+1 = x + x
pr:x1
we have

u
x
=
i
ui ui1
u(x, t)
xi xi1
x
(1.7)
1.2. THE LINEAR OPERATOR FORM
so that
"
#
1
u
u
i+1
i
x
x i+1
x i
"
#
1
u(x + x)
u(x)
=
(x + x)
(x)
x
x
x
"
#
u
=
(x)
.
x
x

This allows us to write 1.3 as

"
#
u
2u
(x)
V (x)u + (x)f (x, t) = (x) 2 .
x
x
t
(1.8)
4 Jan p4
(1.9)
This is a partial differential equation. We will look at this problem in eq1diff

detail in the following chapters. Note that the first term is net tension pr:pde1
force over dx.
1.2
The Linear Operator Form
We define the linear operator L0 by the equation
pr:LinOp1
L0
(x)
+ V (x).
x
x
(1.10)
We can now write equation (1.9) as

"
2
u(x, t) = (x)f (x, t)
t2
eq1LinOp
L0 + (x)
on a < x < b.
(1.11)
This is an inhomogeneous equation with an external force term. Note eq1waveone

that each term in this equation has units of m/t2 . Integrating this
equation over the length of the string gives the total force on the string.
1.3
Boundary Conditions
pr:bc1
To obtain a unique solution for the differential equation, we must place

restrictive conditions on it. In this case we place conditions on the ends
of the string. Either the string is tied together (i.e. closed), or its ends
are left apart (open).

'
ar
br
&
Figure 1.2: A closed string, where a and b are connected.
1.3.1
Case 1: A Closed String
fig1loop
A closed string has its endpoints a and b connected. This case is illustrated in figure 2. This is the periodic boundary condition for a closed
string. A closed string must satisfy the following equations:
pr:pbc1
u(a, t) = u(b, t)
pr:ClStr1
pr:a2
eq1pbc1
(1.12)
which is the condition that the ends meet, and

u(x, t)
u(x, t)
=

x x=a
x x=b

eq1pbc2
pr:ebc1
pr:OpStr1
(1.13)
which is the condition that the ends have the same declination (i.e.,
the string must be smooth across the end points).
1.3.2
sec1-c2
4 Jan p5
Case 2: An Open String
For an elastically bound open string we have the boundary condition

that the total force must vanish at the end points. Thus, by multiplying
equation 1.3 by x and setting the right hand side equal to zero, we
have the equation

u(x, t)

ka u(a, t) + Fa (t) = 0.
a
x x=a
pr:ha1
The homogeneous terms of this equation are a u

|
and ka u(a, t), and
x x=a
the inhomogeneous term is Fa (t). The term ka u(a) describes how the
string is bound. We now define
ha (t)
Fa
a
and
ka
.
a
1.3. BOUNDARY CONDITIONS
b
n
r
ar n
Figure 1.3: An open string, where the endpoints a and b are free.
pr:EffFrc1
pr:esc1
The term ha (t) is the effective force and a is the effective spring constant.
u
+ a u(x) = ha (t)
for x = a.
(1.14)
x
We also define the outward normal, n
, as shown in figure 1.3. This eq1bound
allows us to write 1.14 as
pr:OutNorm1
n
u(x) + a u(x) = ha (t)
fig1.2
for x = a.
The boundary condition at b can be similarly defined:

u
+ b u(x) = hb (t)
x
for x = b,
where
hb (t)
Fb
b
and
kb
.
b
For a more compact notation, consider points a and b to be elements

of the surface of the one dimensional string, S = {a, b}. This gives pr:S1
us
n
S u(x) + S u(x) = hS (t)
In this case n
a = ~lx and n
b = ~lx .
for x on S, for all t.
(1.15)
eq1osbc
pr:lhat1
1.3.3
Limiting Cases
6 Jan p2.1
It is also worthwhile to consider the limiting cases for an elastically

bound string. These cases may be arrived at by varying a and b . The pr:ebc2
terms a and b signify how rigidly the strings endpoints are bound.
The two limiting cases of equation 1.14 are as follows:
pr:ga1

a 0
a
pr:nbc1
pr:dbc1
pr:hS1
u

= ha (t)
x x=a
u(x, t)|x=a = ha /a = Fa /ka .

pr:rbc1
see
Stakgold
p269
(1.17)
The boundary condition a 0 corresponds to an elastic media, and pr:ElMed1

is called the Neumann boundary condition. The case a corresponds to a rigid medium, and is called the Dirichlet boundary condition.
If hS (t) = 0 in equation 1.15, so that
[
nS + S ]u(x, t) = hS (t) = 0
eq1RBC
(1.16)
for x on S,
(1.18)
then the boundary conditions are called regular boundary conditions.

Regular boundary conditions are either
d
1. u(a, t) = u(b, t), dx
u(a, t) =
d
u(b, t)
dx
2. [
nS + S ]u(x, t) = 0
for x on S.
(periodic), or
Thus regular boundary conditions correspond to the case in which there

is no external force on the end points.
pr:ic1
6 Jan p2
pr:u0.1
1.3.4
Initial Conditions
The complete description of the problem also requires information about

the string at some reference point in time:
u(x, t)|t=0 = u0 (x)
for a < x < b
(1.19)
and
u(x, t)|t=0 = u1 (x)

for a < x < b.
(1.20)
t
Here we claim that it is sufficient to know the position and velocity of
the string at some point in time.
1.4
Special Cases
This material
was originally We now consider two singular boundary conditions and a boundary
condition leading to the Helmholtz equation. The conditions first two
in chapter 3
8 Jan p3.3
pr:sbc1
cases will ensure that the right-hand side of Greens second identity
(introduced in chapter 2) vanishes. This is necessary for a physical
system.
1.4. SPECIAL CASES
1.4.1
No Tension at Boundary
For the case in which (a) = 0 and the regular boundary conditions
hold, the condition that u(a) be finite is necessary. This is enough to
ensure that the right hand side of Greens second identity is zero.
1.4.2
Semi-infinite String
In the case that a , we require that u(x) have a finite limit as

x . Similarly, if b , we require that u(x) have a finite limit
as x . If both a and b , we require that u(x) have
finite limits as either x or x .
1.4.3
Oscillatory External Force

sec1helm
In the case in which there are no forces at the boundary we have

ha = hb = 0.
(1.21)
The terms ha , hb are extra forces on the boundaries. Thus the condition
of no forces on the boundary does not imply that the internal forces
are zero. We now treat the case where the interior force is oscillatory
and write
pr:omega1
it
f (x, t) = f (x)e
.
(1.22)
In this case the physical solution will be
Re f (x, t) = f (x) cos t.
(1.23)
We look for steady state solutions of the form

u(x, t) = eit u(x)
for all t.
pr:sss1
(1.24)
This gives us the equation

"
2
L0 + (x) 2 eit u(x) = (x)f (x)eit .
t
#
(1.25)
If u(x, ) satisfies the equation

[L0 2 (x)]u(x) = (x)f (x)
with R.B.C. on u(x)
(1.26)
(the Helmholtz equation), then a solution exists. We will solve this eq1helm
equation in chapter 3.
pr:Helm1
10
1.5
Summary
In this chapter the equations of motion have been derived for the small
oscillation problem. Appropriate forms of the boundary conditions and
initial conditions have been given.
The general string problem with external forces is mathematically
the same as the small oscillation (vibration) problem, which uses vectors
and matrices. Let ui = u(xi ) be the amplitude of the string at the point
xi . For the discrete case we have N component vectors ui = u(xi ), and
for the continuum case we have a continuous function u(x). These
considerations outline the most general problem.
The main results for this chapter are:
1. The equation of motion for a string is
"
2
L0 + (x) 2 u(x, t) = (x)f (x, t)
t
#
where
"
on a < x < b
L0 u =
(x)
+ V (x) u.
x
x
2. Regular boundary conditions refer to the boundary conditions for
either
(a) a closed string:
u(a, t) = u(b, t)

(continuous)

u(a, t)
u(b, t)

=

x x=a
x x=b
(no bends)
or
(b) an open string:
[
nS + S ]u(x, t) = hS (t) = 0
x on S, all t.
3. The initial conditions are given by the equations

u(x, t)|t=0 = u0 (x)
for a < x < b
(1.27)
1.6. REFERENCES
11
and

u(x, t) = u1 (x)

t
t=0
for a < x < b.
(1.28)
4. The Helmholtz equation is

[L0 2 (x)]u(x) = (x)f (x).
1.6
References
See any book which derives the wave equation, such as [Fetter80, p120ff],
[Griffiths81, p297], [Halliday78, pA5].
A more thorough definition of regular boundary conditions may be
found in [Stakgold67a, p268ff].
12
Chapter 2
Greens Identities
Chapter Goals:
Derive Greens first and second identities.
Show that for regular boundary conditions, the linear operator is hermitian.
In this chapter, appropriate tools and relations are developed to solve
the equation of motion for a string developed in the previous chapter.
In order to solve the equations, we will want the function u(x) to take
on complex values. We also need the notion of an inner product. The note
inner product of S and u is defined as
pr:InProd1
pr:S2
( P
n
S u
hS, ui = R bi=1 i i
for the discrete case

a dxS (x)u(x) for the continuous case.
(2.1)
In the uses of the inner product which will be encountered here, for the eq2.2
continuum case, one of the variables S or u will be a length (amplitude
of the string), and the other will be a force per unit length. Thus the
inner product will have units of force times length, which is work.
13
14
CHAPTER 2. GREENS IDENTITIES
2.1
Greens 1st and 2nd Identities

6 Jan p2.4
In the definition of the inner product we make the substitution of L0 u

for u, where
!
"
d
d
(x)
+ V (x) u(x).
L0 u(x)
dx
dx
eq2.3
(2.2)
This substitution gives us

"
d
d
hS, L0 ui =
dxS (x)
(x)
+ V (x) u(x)
dx
dx
a
!!
Z b
d
d
=
dxS (x)
(x) u
dx
dx
a
Z
dxS (x)V (x)u(x).
We now integrate twice by parts ( ud

v = uv vd
u), letting
R
u = S (x) = d
u = dS (x) = dx
dS (x)
dx
and
!
d
d
d
d
d
v = dx
(x) u = d (x) u = v = (x) u
dx
dx
dx
dx
so that
b

hS, L0 ui = S (x) (x)
d
u(x)
dx
a
! Z
Z b
b
dS
d
+
dx
(x)
u(x) +
dxS (x)V (x)u(x)
dx
dx
a
a
b

d
= S (x) (x) u(x)
dx !
a
"
#
Z b
d
d
+
dx
S (x) u(x) + S (x)V (x)u(x) .
dx
dx
a
pr:G1Id1
Note that the final integrand is symmetric in terms of S (x) and u(x).
This is Greens First Identity:
2.2. USING G.I. #2 TO SATISFY R.B.C.
15
b

d
hS, L0 ui = S (x) (x) u(x)
dx !
a
"
Z b
(2.3)
#
d
d
S (x) u(x) + S (x)V (x)u(x) .
dx
dx
dx
Now interchange S and u to get
eq2G1Id
hu, L0 Si = hL0 S, ui
b

d
= u(x) (x) S (x)
dx !
a
"
Z b
dx
(2.4)
#
d
d
u (x) S (x) + u(x)V (x)S (x) .
dx
dx
When the difference of equations 2.3 and 2.4 is taken, the symmetric eq2preG2Id
terms cancel. This is Greens Second Identity:
pr:G2Id1
"
#
b

d
d

hS, L0 ui hL0 S, ui = (x) u(x) S (x) S (x) u(x) .
a
dx
dx
(2.5)
In the literature, the expressions for the Greens identities take = 1 eq2G2Id
and V = 0 in the operator L0 . Furthermore, the expressions here are
for one dimension, while the multidimensional generalization is given
in section 8.4.1.
2.2
Using G.I. #2 to Satisfy R.B.C.

6 Jan p2.5
The regular boundary conditions for a string (either equations 1.12 and
1.13 or equation 1.18) can simplify Greens 2nd Identity. If S and u
correspond to physical quantities, they must satisfy RBC. We will
verify this statement for two special cases: the closed string and the
open string.
2.2.1
The Closed String
For a closed string we have (from equations 1.12 and 1.13)

u(a, t) = u(b, t),
S (a, t) = S (b, t),
16

(a) = (b),
d
d
S
=
S ,
dx x=a
dx x=b
d
d
u
=
u .
dx x=a
dx x=b
By plugging these equalities into Greens second identity, we find that

hS, L0 ui = hL0 S, ui.
(2.6)
eq2twox
2.2.2
The Open String
For an open string we have

u
+ Ka u
x
S
+ Ka S
x
u
+ Kb u
x
S
+ Kb S
x
eq21osbc
= 0
for x = a,
= 0
for x = a,
= 0
for x = b,
= 0
for x = b.
(2.7)
These are the conditions for RBC from equation 1.14. Plugging these
expressions into Greens second identity gives
"
#

dS
du
(x) u
S
= (a)[uKa S S Ka u] = 0

a
dx
dx
and
"
#
b

dS
du
(x) u
S
= (b)[uKb S S Kb u] = 0.

dx
dx
Thus from equation 2.5 we find that

hS, L0 ui = hL0 S, ui,
eq2twox2
just as in equation 2.6 for a closed string.
(2.8)
2.3. ANOTHER BOUNDARY CONDITION
2.2.3
pr:HermOp1
17
A Note on Hermitian Operators
The equation hS, L0 ui = hL0 S, ui, which we have found to hold for
both a closed string and an open string, is the criterion for L0 to be a
Hermitian operator. By using the definition 2.1, this expression can be
rewritten as
hS, L0 ui = hu, L0 Si .
(2.9)
Hermitian operators are generally generated by nondissipative physical problems. Thus Hermitian operators with Regular Boundary Conditions are generated by nondissipative mechanical systems. In a dissipative system, the acceleration cannot be completely specified by the
position and velocity, because of additional factors such as heat, friction, and/or other phenomena.
2.3
Another Boundary Condition

6 Jan p2.6
If the ends of an open string are free of horizontal forces, the tension
at the end points must be zero. Since
lim (x) = 0
xa,b
we have
lim (x)u(x)
xa,b

S (x) = 0
x
and
lim (x)S (x)
xa,b
u(x) = 0.
x
In the preceding equations, the abbreviated notation lim is introduced

xa,b
to represent either the limit as x approaches the endpoint a or the limit

as x approaches the endpoint b. These equations allow us to rewrite
Greens second identity (equation 2.5) as
hS, L0 ui = hL0 S, ui
(2.10)
for the case of zero tension on the end points This is another way of eq2G2Id
getting at the result in equation 2.8 for the special case of free ends.
18
2.4
Physical Interpretations of the G.I.s
sec2.4
Certain qualities of the Greens Identities correspond to physical situations and constraints.
2.4.1
The Physics of Greens 2nd Identity
6 Jan p2.6
The right hand side of Greens 2nd Identity will always vanish for physically realizable systems. Thus L0 is Hermitian for any physically realizable system.
We could extend the definition of regular boundary conditions by
letting them be those in which the right-hand side of Greens second
identity vanishes. This would allow us to include a wider class of problems, including singular boundary conditions, domains, and operators.
This will be necessary to treat Bessels equation. For now, however, we
only consider problems whose boundary conditions are periodic or of
the form of equation 1.18.
2.4.2
A Note on Potential Energy
The potential energy of an element dx of the string has two contributions. One
is the spring
potential energy 12 V (x)(u(x))2 (c.f., 12 kx2 in
R
R
U = F dx = (kx)dx = 12 kx2 [Halliday76, p141]). The other is
the tension potential energy, which comes from the tension force in
section 1.1.3, dF = x
[ (x) x
u(x)]dx, and thus Utension is
U =
so
dx,
x
x
Z
dU
d Z
u
u
=
dx =
dt
dt x
x
x
x x
u
t
dx,
and so the change in potential energy in a time interval dt is

U dt
a
parts
x
x
u
t
dt dx
u
dt dx
t x
"
#b
u
dt
t
2.4. PHYSICAL INTERPRETATIONS OF THE G.I.S

Z
Zb1
t a 2
19
u
dt dx
t x
u
x
!2
t+dt
dx
The second term in the second equality vanishes. We may now sum
the differentials of U in time to obtain the potential energy:
U=
2.4.3
t0 =0
Z
U dt =
u
x
!2
dx =
0
u
x
!2
dx.
The Physics of Greens 1st Identity

sec2.4.2
Let S = u. Then 2.3 becomes
6 Jan p2
b

d
hu, L0 ui = u (x) (x) u(x)
dx !
a
"
Z b
(2.11)
#
d
d
u (x) u(x) + u (x)V (x)u(x) .
dx
dx
dx
For a closed string we have

hu, L0 ui =
du
dx (x)
dx
!2
+ V (x)(u(x))2 = 2U
(2.12)
since each quantity is the same at a and b. For an open string we found eq2x
(equation 1.15)
8 Jan p3.2

du
= Ka u
(2.13)
dx x=a
and

du
= Kb u
(2.14)
dx x=b
so that
hu, L0 ui = (a)Ka |u(a)|2 + (b)Kb |u(b)|2
+
= 2U,
du
dx (x)
dx
!2
+ V (x)(u(x))2
20
pr:pdo1
twice the potential energy. The term 12 (a)Ka |u(a)|2 + 12 (b)Kb |u(b)|2
is the potential energy due to two discrete springs at the end points,
and is simply the spring constant times the displacement squared.
The term (x) (du/dx)2 is the tension potential energy. Since du/dx
represents the string stretching in the transverse direction, (x) (du/dx)2
is a potential due to the stretching of the string. V (x)(u(x))2 is the
elastic potential energy.
For the case of the closed string, equation 2.12, and the open string,
equation 2.15, the right hand side is equal to twice the potential energy.
If Ka , Kb , and V are positive for the open string, the potential energy
U is also positive. Thus hu, L0 ui > 0, which implies that L0 is a positive
definite operator.
2.5
Summary
1. The Greens identities are:

(a) Greens first identity:
b

d
hS, L0 ui = S (x) (x) u(x)
dx !
a

Z b
d
d
S (x) u(x)
dx
dx
dx
+ S (x)V (x)u(x) ,
(b) Greens second identity:
"
#
b

d
d

hS, L0 uihL0 S, ui = (x) u(x) S (x) S (x) u(x) .
a
dx
dx
2. For a closed string and an open string (i.e., RBC) the linear operator L0 is Hermitian:
hS, L0 ui = hu, L0 Si .
see FW p207
expl.
p10
p126
eq2y
pr:pe1
2.6. REFERENCES
2.6
21
References
Greens formula is described in [Stakgold67, p70] and [Stakgold79,

p167].
The derivation of the potential energy of a string was inspired by
[Simon71,p390].
22
Chapter 3
Greens Functions
Chapter Goals:
Show that an external force can be written as a
sum of -functions.
Find the Greens function for an open string with
no external force on the endpoints.
In this chapter we want to solve the Helmholtz equation, which was
obtained in section 1.4.3. First we will develop some mathematical
principles which will facilitate the derivation.
8 Jan p3.4
3.1
Lagrangian
stuff
commented out
The Principle of Superposition
Suppose that
pr:a1.1
f (x) = a1 f1 (x) + a2 f2 (x).
(3.1)
If u1 and u2 are solutions to the equations (c.f., 1.26)

[L0 2 (x)]u1 (x) = (x)f1 (x)
(3.2)
[L0 2 (x)]u2 (x) = (x)f2 (x)
(3.3)
with RBC and such that (see equation 1.15)

(
nS + S )u1 = 0
(
nS + S )u2 = 0
23
for x on S
eq3q
24
CHAPTER 3. GREENS FUNCTIONS
then their weighted sum satisfies the same equation of motion

[L0 2 (x)](a1 u1 (x) + a2 u2 (x))
= a1 [L0 2 (x)]u1 (x) +a2 [L0 2 (x)]u2 (x)
|
{z
(x)f1 (x)
{z
(x)f2 (x)
= (x)f (x).
and boundary condition
[
nS + S ][a1 u1 (x) + a2 u2 (x)]
= a1 [
nS + S ]u1 + a2 [
nS + S ]u2
= a1 (0) + a2 (0) = 0.
We have thus shown that
L0 [a1 u1 + a2 u2 ] = a1 L0 u1 + a2 L0 u2 .
pr:pos1
(3.4)
This is called the principle of superposition, and it is the defining property of a linear operator.
3.2
The Dirac Delta Function
11 Jan p4.1
We now develop a tool to solve the Helmholtz equation (which is also

called the steady state equation), equation 1.26:
[L0 2 (x)]u(x) = (x)f (x).
pr:DeltaFn1
The delta function is defined by the equation
pr:Fcd
Fcd =
eq3deltdef
pr:Fcd1
Z
c
dx(x xk ) =
1 if c < xk < d
0 otherwise.
(3.5)
where Fcd represents the total force over the interval [c, d]. Thus we see
that the appearance of the delta function is equivalent to the application
of a unit force at xk . The Dirac delta function has units of force/length.
On the right-hand side of equation 1.26 make the substitution
(x)f (x) = (x xk ).
(3.6)
3.2. THE DIRAC DELTA FUNCTION
25
Integration gives us
d
Z
c
eq3fdc
(x)f (x)dx = Fcd ,
(3.7)
which is the total force applied over the domain. This allows us to write
[L0 2 ]u(x, ) = (x xk )
a < x < b, RBC
(3.8)
where we have written RBC to indicate that the solution of this equation must also satisfy regular boundary conditions. We may now use 11 Jan p2
the principle of superposition to get an arbitrary force. We define an
element of such an arbitrary force as
Fk =
xk +x
dx(x)f (x)
(3.9)
xk
= the force on the interval x.
(3.10)
eq3Fsubk
We now prove that

(x)f (x) =
N
X
Fk (x x0k )
(3.11)
k=1
where xk < x0k < xk + x. We first integrate both sides to get

Z
dx(x)f (x) =
dx
N
X
Fk (x x0k ).
k=1
eq3sfd
x0 replaces x
(3.12) so -fn isnt on
boundary
By definition (equation 3.7), the left-hand side is the total force applied
over the domain, Fcd . The right-hand side is
Z
N
dX
c k=1
Fk (x
x0k )dx
N Z
X
k=1 c
dxFk (x x0k )
(3.13)
Fk
(3.14)
c<xk <d
=
N
xk +x
c<xk <d xk
Z d
dx(x)f (x) (3.15)
dx(x)f (x)
(3.16)
Fcd .
(3.17)
26
pr:Helm2
In the first equality, 3.13, switching the sum and integration holds for eq3sum1-5
all well behaved Fk . Equality 3.14 follows from the definition of the
delta function in equation 3.5. Equality 3.15 follows from equation 3.9.
By taking the continuum limit, equality 3.16 completes the proof.
The Helmholtz equation 3.2 can now be rewritten (using 3.11) as
2
[L0 (x) ]u(x, ) =
N
X
Fk (x xk ).
(3.18)
k=1
By the principle of superposition we can write

u(x) =
N
X
Fk uk (x)
(3.19)
k=1
11 Jan p3
where uk (x) is the solution of [L0 (x) 2 ]uk (x, 2 ) = (x xk ). Thus,

if we know the response of the system to a localized force, we can find
the response of the system to a general force as the sum of responses
to localized forces.
We now introduce the following notation
uk (x) G(x, xk ; 2 )
pr:Gxxo1
(3.20)
where G is the Greens function, xk signifies the location of the disturbance, and corresponds to frequency. This allows us to write
u(x)
=
=
N
N
X
Fk uk (x)
k=1
N Z xk +x
X
k=1 xk
Z b
0
dx0 (x0 )f (x0 )G(x, xk ; 2 )
dx G(x, x0 ; 2 )(x0 )f (x0 ).
We have defined the Greens function by

[L0 (x) 2 ]G(x, x0 ; 2 ) = (x x0 )
11 Jan p4
pr:NatFreq1
where will nat
freq be defined
a < x, x0 < b, RBC. (3.21)
The solution will explode for 2 when is a natural frequency of the

system, as will be seen later.
3.2. THE DIRAC DELTA FUNCTION
27
G(x, x0 ; 2 )
d
G|x=x0
dx
d
G|x=x0 +
dx

%e
A

%
e
AA
l
U ,
,
l
%
e
%
eQ

Q

HH
%
e
%
e

PPX

X``
%
e
%
e
x0 x0 x0 +
Figure 3.1: The pointed string
pr:lambda1
fix this
Let = 2 be an arbitrary complex number. Since the squared

frequency 2 cannot be complex, we relabel it . So now we want to
solve
"
d
d
(x)
+ V (x) (x) G(x, x0 ; 2 ) = (x x0 )
dx
dx
a < x, x0 < b, RBC
(3.22)
Note that G will have singularities when is a natural frequency. To eq3.19a

obtain a condition which connects solutions on either side of the singularity, we integrate equation 3.22. Consider figure 3.1. In this case
Z
x0 +
x0
"
fig3.1
d
d
dx
(x)
+ V (x) (x) G(x, x0 ; )
dx
dx
=
which becomes
x0 +
x0
(x x0 )dx
pr:epsilon1
x0 +

d
(x) G(x, x0 ; )
=1
0
dx
x
(3.23)
since the integrals over V (x) and (x) vanish as 0. Note that in
this last expression 1 has units of force.
28
3.3
Two Conditions
11 Jan p5
3.3.1
Condition 1
The previous equation can be written as

d
d
1
G
G
= .
dx x=x0 + dx x=x0
eq3other
eq3b
(3.24)
This makes sense after considering that a larger tension implies a smaller
kink (discontinuity of first derivative) in the string.
3.3.2
Condition 2
We also require that the string doesnt break:

G(x, x0 )|x=x0 + = G(x, x0 )|x=x0 .
eq3d
(3.25)
This is called the continuity condition.
pr:ContCond1
3.3.3
pr:homog1
Application
To find the Greens function for equation 3.22 away from the point x0 ,
we study the homogeneous equation
[L0 (x)]u(x, ) = 0
pr:efp1
x 6= x0 , RBC.
This is called the eigen function problem. Once we specify G(x0 , x0 ; )

d
and dx
G(x0 , x0 ; ), we may use this equation to get all higher derivatives
and thus determine G(x, x0 ; ).
We know, from differential equation theory, that two fundamental
solutions must exist. Let u1 and u2 be the solutions to
[L0 (x)]u1,2 (x, ) = 0
pr:AABB
(3.27)
where u1,2 denotes either solution. Thus

G(x, x0 ; ) = A1 u1 (x, ) + A2 u2 (x, )
eq3ab1
(3.26)
and
for x < x0 ,
(3.28)
3.4. OPEN STRING
29
G(x, x0 ; ) = B1 u1 (x, ) + B2 u2 (x, )

eq3ab2
11 Jan p6
for x > x0 .
(3.29)
We have now defined the Greens function in terms of four constants.

We have two matching conditions and two R.B.C.s which determine
these four constants.
3.4
Open String
13 Jan p2
We will solve for an open string with no external force h(x), which was where is 13 Jan
first discussed in section 1.3.2. G(x, x0 ; ) must satisfy the boundary p1
condition 1.18. Choose u1 such that it satisfies the boundary condition
for the left end

u1

+ Ka u1 (a) = 0.
(3.30)
x x=a
This determines u1 up to an arbitrary constant. Choose u2 such that
it satisfies the right end boundary condition

u2
+ Kb u2 (b) = 0.

x x=b
(3.31)
We find that in equations 3.28 and 3.29, A2 = B1 = 0. Thus we

have two remaining conditions to satisfy.
13 Jan p1
We now have
G(x, x0 ; ) = A1 (x0 )u1 (x, )
for x < x0 .
(3.32)
Note that only the boundary condition at a applies since the behavior
of u1 (x) does not matter at b (since b > x0 ). This gives G determined
up to an arbitrary constant. We can also write
G(x, x0 ; ) = B2 (x0 )u2 (x, )
for x > x0 .
(3.33)
We also note that A and B are constants determined by x0 only.

Thus we can write the previous expressions in a more symmetric form: pr:CD
G(x, x0 ; ) = Cu1 (x, )u2 (x0 , )
for x < x0 ,
(3.34)
G(x, x0 ; ) = Du1 (x0 , )u2 (x, )
for x > x0 .
(3.35)
30
13 Jan p3
In one of the problem sets we prove that G(x, x0 ; ) = G(x0 , x; ).

This can also be stated as Greens Reciprocity Principle: The ampli- pr:grp1
tude of the string at x subject to a localized force applied at x0 is
equivalent to the amplitude of the string at x0 subject to a localized
force applied at x.
We now apply the continuity condition. Equation 3.25 implies that
C = D.
Now we have a function symmetric in x and x0 , which verifies the
Greens Reciprocity Principle. By imposing the condition in equation
3.24 we will be able to determine C:

dG
du1

=C
u2 (x0 )
dx x=x0 +
dx x0
eq3trione
eq3tritwo
dG
du2
= Cu1 (x0 )

dx x=x0
dx x0
Combining equations (3.24), (3.36), and (3.37) gives us

"
(3.37)

du2 du1
1
C u1
u2
=
.
dx
dx
(x0 )
x=x0
pr:wronsk1
(3.36)
(3.38)
The Wronskian is defined as

du2
du1
u2
.
dx
dx
(3.39)
1
.
0
0
1 (x , ), u2 (x , ))
(3.40)
W (u1 , u2 ) u1
This allows us to write
C=
13 Jan p4
(x0 )W (u
Thus
G(x, x0 ; ) =
eq3.39
u1 (x< , )u2 (x> , )

,
(x0 )W (u1 (x0 , ), u2 (x0 , ))
where we define
pr:xless1
u(x) if x < x0
u(x0 ) if x0 < x
u(x) if x > x0
u(x0 ) if x0 > x.
u(x< )
and
u(x> )
(3.41)
3.5. THE FORCED OSCILLATION PROBLEM
31
The us are two different solutions to the differential equation:

[L0 (x)]u1 = 0
[L0 (x)]u2 = 0.
(3.42)
Multiply the first equation by u2 and the second by u1 . Subtract one FW p249
equation from the other to get u2 ( u01 )0 + u1 ( u02 )0 = 0 (where we
( x
) + V ). Rewriting this as a
have used equation 1.10, L0 = x
total derivative gives
d
[ (x)W (u1 , u2 )] = 0.
dx
(3.43)
This implies that the expression (x)W (u1 (x, ), u2 (x, )) is independent of x. Thus G is symmetric in x and x0 .
The case in which the Wronskian is zero implies that u1 = u2 ,
since then 0 = u1 u02 u2 u01 , or u02 /u2 = u01 /u1 , which is only valid for all
x if u1 is proportional to u2 . Thus if u1 and u2 are linearly independent,
the Wronskian is non-zero.
3.5
but
isnt
W 0 (x) = 0 also
true?
pr:LinIndep1
The Forced Oscillation Problem

13 Jan p5
The general forced harmonic oscillation problem can be expanded into pr:fhop1
equations having forces internally and on the boundary which are simple time harmonic functions. Consider the effect of a harmonic forcing
term
"
#
2
L0 + 2 u(x, t) = (x)f (x)eit .
(3.44)
t
We apply the following boundary conditions:
u(x, t)
+ a u(x, t) = ha eit
x
eq3ss
for x = a,
(3.45)
and
u(x, t)
+ b u(x, t) = hb eit .
for x = b.
(3.46)
x
We want to find the steady state solution. First, we assume a steady
state solution form, the time dependence of the solution being
u(x, t) = eit u(x).
(3.47)
32
After making the substitution we get an ordinary differential equation

in x. Next determine G(x, x0 ; = 2 ) to obtain the general steady state
solution. In the second problem set we use Greens Second Identity to
solve this inhomogeneous boundary value problem. All the physics of
the exciting system is given by the Greens function.
3.6
pr:fop1
pr:NatFreq2
eq3fo
Free Oscillation
Another kind of problem is the free oscillation problem. In this case

f (x, t) = 0 and ha = hb = 0. The object of this problem is to find the
natural frequencies and normal modes. This problem is characterized
by the equation:
"
#
2
L0 + 2 u(x, t) = 0
(3.48)
t
with the Regular Boundary Conditions:
u is periodic. (Closed string)
[
n + KS ]u = 0
pr:NatMode1
for x in S. (Open string)
The goal is to find normal mode solutions u(x, t) = ein t un (x). The
natural frequencies are the n and the natural modes are the un (x).
We want to solve the eigenvalue equation
[L0 n2 ]un (x) = 0
eq3.48
pr:EVect1
with R.B.C.
(3.49)
The variable n2 is called the eigenvalue of L0 . The variable un (x) is

called the eigenvector (or eigenfunction) of the operator L0 .
3.7
Summary
1. The Principle of superposition is

L0 [a1 u1 + a2 u2 ] = a1 L0 u1 + a2 L0 u2 ,
where L0 is a linear operator, u1 and u2 are functions, and a1 and
a2 are constants.
3.7. SUMMARY
33
2. The Dirac Delta Function is defined as

(
Z
c
dx(x xk ) =
1 if c < xk < d
0 otherwise.
3. Force contributions can be constructed by superposition.

(x)f (x) =
N
X
Fk (x x0k ).
k=1
4. The Greens Function is the solution to to an equation whose

inhomogeneous term is a -function. For the Helmholtz equation,
the Greens function satisfies:
[L0 (x) 2 ]G(x, x0 ; 2 ) = (x x0 )
a < x, x0 < b, RBC.
d
G|x=x0 +
5. At the source point x0 , the Greens function satisfies dx
d
1
0
0
G|x=x0 = and G(x, x )|x=x0 + = G(x, x )|x=x0 .
dx
6. Greens Reciprocity Principle is The amplitude of the string at

x subject to a localized force applied at x0 is equivalent to the
amplitude of the string at x0 subject to a localized force applied
at x.
7. The Greens function for the 1-dimensional wave equation is given
by
u1 (x< , )u2 (x> , )
G(x, x0 ; ) =
.
(x0 )W (u1 (x0 , ), u2 (x0 , ))
8. The forced oscillation problem is
"
2
L0 + 2 u(x, t) = (x)f (x)eit ,
t
#
with periodic boundary conditions or the elastic boundary conditions with harmonic forcing.
9. The free oscillation problem is
"
2
L0 + 2 u(x, t) = 0.
t
#
34
3.8
Reference
See [Fetter81, p249] for the derivation at the end of section 3.4.
A more complete understanding of the delta function requires knowledge of the theory of distributions, which is described in [Stakgold67a,
p28ff] and [Stakgold79, p86ff].
The Greens function for a string is derived in [Stakgold67a, p64ff].
Chapter 4
Properties of Eigen States
13 Jan p7
Chapter Goals:
Show that for the Helmholtz equation, n2 > 0, n2
is real, and the eigen functions are orthogonal.
Derive the dispersion relation for a closed massless
string with discrete mass points.
Show that the Greens function obeys Hermitian
analyticity.
Derive the form of the Greens function for near
an eigen value n .
Derive the Greens function for the fixed string
problem.
2 > 0
By definition 2.1
hS, ui =
dxS (x)u(x).
(4.1)
In section 2.4 we saw (using Greens first identity) for L0 as defined

in equation 2.2, and for all u which satisfy equation 1.26, that V > 0
implies hu, L0 ui > 0 . We choose u = un and use equation 3.49 so that
0 < hun , L0 un i = hun , un in2 .
35
(4.2)
36
13 Jan p8
2 real
CHAPTER 4. PROPERTIES OF EIGEN STATES
Remember that signifies the mass density, and thus > 0. So we

conclude
hun , L0 un i
n2 =
> 0.
(4.3)
hun , un i
This all came from Greens first identity.
Next we apply Greens second identity 2.5,
hS, L0 ui = hL0 S, ui
for S, u satisfying RBC
(4.4)
Let S = u = un . This gives us

n2 hun , un i = hun , L0 un i
= hL0 un , un i
= (n2 ) hun , un i.
orthogonality
(4.5)
(4.6)
(4.7)
We used equation 3.49 in the first equality, 2.5 in the second equality,
and both in the third equality. From this we can conclude that n2 is
real.
Now let us choose u = un and S = um . This gives us
hum , L0 un i = hL0 um , un i.
(4.8)
Extracting n2 gives (note that (x) is real)

2
2
n2 hum , un i = m
hum , un i = m
hum , un i.
(4.9)
So
2
(n2 m
)hum , un i = 0.
(4.10)
2
then hum , un i = 0:
Thus if n2 6= m
eq4.11
13 Jan p8
pr:ortho1
pr:orthon1
2
if n2 6= m
.
(4.11)
That is, two eigen vectors um and un of L0 corresponding to different

eigenvalues are orthogonal with respect to the weight function . If
the eigen vectors um and un are normalized, then the orthonormality
condition is
Z
eq4.11p
dxum (x)(x)un (x) = 0
dxum (x)(x)un (x) = mn
2
,
if n2 6= m
(4.12)
where the Kronecker delta function is 1 if m = n and 0 otherwise.
4.1. EIGEN FUNCTIONS AND NATURAL MODES

((
u

u

u

u
LLu
Z
Z
Zu
P
PPu
hh
37
u hhu
PP
u
1 a 2 PZ
Z
Zu
LLu
u

u

u
(
u (
Figure 4.1: The closed string with discrete mass points.
4.1
Eigen Functions and Natural Modes

15 Jan p1
We now examine the natural mode problem given by equation 3.49. To

find the natural modes we must know the natural frequencies n and pr:NatMode2
the normal modes un . This is equivalent to the problem
pr:lambdan1
L0 un (x) = (x)n un (x),
RBC.
(4.13)
To illustrate this problem we look at a discrete problem.
4.1.1
eq4A
A Closed String Problem

pr:dcs1
This problem is illustrated in figure 4.11 . In this problem the mass 15 Jan p2
density and the tension are constant, and the potential V is zero.
fig4w
The term u(xi ) represents the perpendicular displacement of the ith
mass point. The string density is given by = m/a where m is the
mass of each mass
point and a is a unit of length. We also make the
q
definition c = /. Under these conditions equation 1.1 becomes
(ui+1 + ui1 2ui ).

a
Substituting the solution form ui = eikxi eit into this equation gives
m
ui = Ftot =
eika + eika
ka
m 2 = 2
+ 1 = 2 (1 cos ka) = 4 sin2 .

a
2
a
a
2
!
But continuity implies u(x) = u(x + N a), so eikN a = 1, or kN a = 2n,

so that

2
n,
n = 1, . . . , N.
k=
Na
1
See FW p115.
38

The natural frequencies for this system are then
n2 =
eq4r
pr:DispRel1
c2 sin2 (kn a/2)

a2 /4
where kn = 2
n and n can take on the values 0, 1, . . . , N 21 for odd
L
N , and 0, 1, . . . , N2 1, + N2 for even N . The constant is c2 = a /m.
Equation 4.14 is called the dispersion relation. If n is too large, the un
take on duplicate values. The physical reason that we are restricted to a
finite number of natural modes is because we cannot have a wavelength
< a. The corresponding normal modes are given by
n (xi , t) = ein t un (xi )
= ei[n tkn xi ]
2
= ei[n t L nxi ] .
eq4giver
pr:travel1
frequency
expression
(4.15)
(4.16)
(4.17)
The normal modes correspond to traveling waves. Note that n is

doubly degenerate in equation 4.14. Solutions n (x) for n which are
larger than allowed give the same displacement of the mass points, but
with some nonphysical wavelength. Thus we are restricted to N modes
and a cutoff frequency.
4.1.2
pr:cutoff1
(4.14)
The Continuum Limit
We now let a become increasingly small so that N becomes large for L

fixed. This gives us kn = 2/L for L = N a. In the continuum limit,
the number of normal modes becomes infinite. Shorter and shorter
wavelengths become physically relevant and there is no cutoff frequency.
Letting a approach zero while L remains fixed gives
n2 =
c2 sin2
a2
4
kn a
2
N a=L
a0
c2 kn2
(4.18)
and so
n = c|kn |
(4.19)

2
n = c|kn | = c
(4.20)
4.1. EIGEN FUNCTIONS AND NATURAL MODES

n = c
2
n.
L
39
(4.21)
Equation (4.17) gives us the un s for all n.

We have found
characteristics
For a closed string, the two eigenvectors for every eigenvalue

(called degeneracy) correspond to the two directions in which a pr:degen1
wave can move. The eigenvalues are n2 .
The natural frequencies n are always discrete, with a separation
distance proportional to 1/L.
For open strings there is no degeneracy. This is because the position and slope of the Greens function at the ends is fixed by
the open string boundary conditions, whereas the closed string
boundary conditions do not determine the Greens function at
any particular point.
For the discrete closed case, the n s are discrete with double degeneracy, 15 Jan p4
giving
u . We also find the correspondence n c/L where c
q
/. We also found that there is no degeneracy for the open discrete
case.
4.1.3
Schr
odingers Equation
pr:Schro1
Consider again equation 4.13

L0 un (x) = (x)n un (x)
RBC
(4.22)
where
L0 =
d
d
(x) + V (x).
dx
dx
(4.23)
We now consider the case in which (x) = h

2 /2m and = 1, both
quantities being numerical constants. The linear operator now becomes
h2 d2
+ V (x).
L0 =
2m dx2
(4.24)
40

V (x)
E
E
EE
D
D
D

C
CC
B
B

A

QQ
Figure 4.2: Negative energy levels

This is the linear operator for the Schrodinger equation for a particle
of mass m in a potential V :
"
h2 d2
+ V (x) un (x) = n un (x)
2m dx2
#
+ RBC
(4.25)
In this case gives the allowed energy values.

The potential V (x) can be either positive or negative. It needs to
be positive for L0 to be positive definite, in which case 0 > 0. For
V < 0 we can have a finite number of of eigenvalues less that zero.
verify
this On a physical string the condition that V > 0 is necessary.
paragraph
One can prove that negative energy levels are discrete and bounded
from below. The bound depends on the nature of V (x) (Rayleigh quo15 Jan p5
tient idea). Suppose that V has a minimum, as shown in figure 4.2 for
pr:RayQuo1
example. By Greens first identity the quantity L0 Vmin gives a new
fig4neg
operator which is positive definite.
4.2
pr:Fred1
Natural Frequencies and the Greens

Function
We now look at a Greens function problem in the second problem set.

Under consideration is the Fredholm equation
[L0 (x)n2 ]u(x) = (x)f (x),
RBC.
(4.26)
4.3. GF BEHAVIOR NEAR = N
41
In problem 2.2 one shows that the solution un (x) for this equation only
exists if
Z
dxun (x)(x)f (x) = 0.
(4.27)
This is the condition that the eigenvectors un (x) are orthogonal to
the function f (x). We apply this to the Greens function. We choose
= n = n2 and evaluate the Greens function at this point. Thus for
the equation
[L0 (x)n ]G(x, x0 ; ) = (x x0 )
(4.28)
there will be a solution G(x, x0 ; ) only if (using 4.27)
G(x, x; ) =
dx0 G(x, x0 ; )(x x0 ) = 0.
eq4.26
Ask Baker
Isnt this just
hitting a stationary point?
eq4.26p
(4.29) clarify this.
The result G(x, x; ) = 0 implies that un (x0 ) = 0 (using equation 3.41). eq4.26b
There will be no solution unless x0 is a node. In physical terms, this 15 Jan p6
means that a natural frequency can only be excited at a node.
4.3
GF behavior near = n
15 Jan p6
From the result of the previous section, we expect that if the driving
frequency is not a natural frequency, everything will be well behaved.
So we show that G(x, x0 ; ) is good everywhere (in the finite interval
[a, b]) except for a finite number of points. The value of G(x, x0 ; )
becomes infinite near n , that is, as n . For near n we can
write the Greens function as
pr:gnxx1
1
G(x, x0 ; )
gn (x, x0 ) + finite
n
(4.30)
n
where finite is a value always of finite magnitude. We want to find gn ,
so we put [L0 (x)] in front of each side of the equation and then
add and subtract n (x)G(x, x0 ; ) on the the right-hand side. This
gives us (using 4.28)

1
(x x0 ) = [L0 n (x)]
n

1
0
gn (x, x ) + finite
n
+ (n )(x)
n

1
n .
= [L0 n (x)]
n
42
15 Jan p7
The left-hand side is also finite if we exclude x = x0 . This can only

occur if
[L0 n (x)]gn (x, x0 ) = 0
x=
6 x0 .
(4.31)
From this we can conclude that gn has the form
gn (x, x0 ) = un (x)f (x0 ),
eq4guf
pr:fxprime1
pr:NatFreq3
x 6= x0 ,
(4.32)
where f (x0 ) is a finite term and un (x) satisfies [L0 n (x)]un (x) = 0
with RBC. Note that here the eigen functions un (x) are not yet normalized. This is the relation between the natural frequency and the
Greens function.
4.4
Relation between GF & Eig. Fn.
20 Jan p2
pr:SpecThy1
We continue developing the relation between the Greens function and

spectral theory. So far we have discussed the one dimensional problem.
This problem was formulated as
[L0 (x)]G(x, x0 ; ) = (x x0 ),
eq4.31m
RBC.
(4.33)
To solve this problem we first solved the corresponding homogeneous

problem
[L0 (x)n ]un (x) = 0,
RBC.
(4.34)
The eigenvalue n is called degenerate if there is more than one un per
n . We note the following properties in the Greens function:
1. G (x, x0 ; ) = G(x, x0 ; ). Recall that , L0 , and the boundary
condition terms are real. First we take the complex conjugate of
equation 4.33,
[L0 (x) ]G (x, x0 ; ) = (x x0 ),
RBC
(4.35)
and then we take the complex conjugate of to get

[L0 (x)]G (x, x0 ; ) = (x x0 ),
RBC
(4.36)
which gives us
G (x, x0 ; ) = G(x, x0 ; ).
(4.37)
4.4. RELATION BETWEEN GF & EIG. FN.
43
2. G is symmetric. In the second problem set it was seen that

G(x, x0 ; ) = G(x0 , x; ).
(4.38)
3. The Greens function G has the property of Hermitian analyticity. 18 Jan p3

By combining the results of 1 and 2 we get
pr:HermAn1
G (x, x0 ; ) = G(x0 , x; ).
(4.39)
This may be called the property of Hermitian analyticity.
eq4ha
In the last section we saw that
G(x, x0 ; ) n
gn (x, x0 )
n
(4.40)
for g such that
eq4cA
0
[L0 (x)n ]gn (x, x ) = 0.
(4.41)
For the open string there is no degeneracy and for the closed string
there is double degeneracy. (There is also degeneracy for the 2- and give ref
3-dimensional cases.)
4.4.1
Case 1: Nondegenerate
Assume that n is non-degenerate. In this case we can write (using

equation 4.32)
gn (x, x0 ) = un (x)fn (x0 ).
(4.42)
Hermitian analyticity and the complex conjugate of equation 4.40 give eq4cB
(note that n R)
gn (x, x0 )
gn (x0 , x)
=
(4.43)
n
n
as n . This implies that
gn (x, x0 ) = gn (x0 , x).
(4.44)
eq4cC
So now 4.42 becomes
20 Jan p4
44

un (x)fn (x0 ) = fn (x)un (x0 )
(4.45)
so that (since x and x0 are independent) if un (x) is normalized (according to equation 4.12)
(4.46)
fn (x) = un (x).
In the non-degenerate case we have (from 4.40 and 4.42)
G(x, x0 ; ) n
eq4.47
un (x)un (x0 )
.
n
(4.47)
where un (x) are normalized eigen functions.
pr:normal1
4.4.2
is it +?
Case 2: n Double Degenerate
In the second case, the eigenvalue n has double degeneracy like the
closed string. The homogeneous closed string equation is
(L0 + n )u()
n (x) = 0.
why?
The eigenfunctions corresponding to n are u+

n (x) and un (x). By using
the same reasoning that lead to equation 4.47 we can write
G(x, x0 ; )
1
()
[u(+) (x)u(+)
(x) + u()
(x)]
n
n (x)un
n n
(4.48)
for the equation

[L0 (x)n ]u()
n (x) = 0,
RBC.
(4.49)
How does this The eigenfunction un may be written as un = A+ u+ + A u . Double

fit in?
degeneracy is the maximum possible degeneracy in one dimension.
20 Jan p5
In the general case of -fold degeneracy
G(x, x0 ; ) n
1 X
0
[un (x)u
n (x )]
n
(4.50)
where un (x) solves the equation

[L0 (x)n ]un (x) = 0,
pr:poles1
RBC.
(4.51)
The mathematical relation between the Greens function and the

eigen functions is the following: The eigenvalues n are the poles of G.
P
0
The sum of bilinear products un (x)u
n (x ) is the residue of the pole
= n .
4.5. SOLUTION FOR A FIXED STRING
4.5
45
Solution for a Fixed String
We want to solve equation 3.22. Further, we take V = 0, and

constant, and a = 0, b = L.
"
d2
2 G(x, x0 ; ) = (x x0 )
dx
#
for 0 < x, x0 < L
(4.52)
for the case of a fixed end string. Our boundary conditions are
G(x, x0 ; ) = 0
4.5.1
for x = a, b.
A Non-analytic Solution
We know from equation 3.41 the solution is

G(x, x0 ; ) =
20 Jan p6
u1 (x< , )u2 (x> , )

.
W (u1 , u2 )
(4.53)
This solution only applies to the one dimensional case. This is because eq4.54
the solution was obtained using the theory of ordinary differential equations. The corresponding homogeneous equations are given by
pr:homog2
"
d2
+ 2 u1,2 (x, ) = 0.
2
dx
c
#
(4.54)
In this equation we have used the definition 1/c2 / . The variables eq4lcu
u1 and u2 also satisfy the conditions
pr:c1
u1 (0, ) = 0
and
u2 (L, ) = 0.
(4.55)
The solution to this homogeneous problem can be found to be

s
u1 = sin
x
and
u2 = sin
(L x).
(4.56)
2
c
c2
In these solutions appears. Since can be complex, we must define

a branch cut2 .
pr:branch1
46

Im
u

Re
Figure 4.3: The -convention
4.5.2
fig4one
The Branch Cut
Since G n1 (see 4.40) it follows (from n > 0) that Gn is analytic pr:analytic1

for Re () < 0. As a convention, we choose such that 0 < < 2.
This is illustrated in figure4.3.
Using this convention, can be represented by
q
=
||ei/2
(4.57)
=
eq4.58-59
20 Jan p7
.
|| cos + i sin
2
2
Analytic Fundamental Solutions and GF
We now look back at the fixed string problem. We found that

s
u1 (x, ) = sin
so that
du1
=
cos
dx
c2
but only if x = This gives us the boundary value
a = 0.
(4.58)
Note that has a discontinuity along the positive real axis. The
function G is analytic in the complex plane if the positive real axis is
removed. This can be expressed mathematically as the condition that
(4.59)
Im > 0.
4.5.3
pr:u1.1
"
See also FW p485.
x
c2
x.
c2

du1

=
dx x=a
47
.
c2
(4.60)
Because of the , this is not analytic over the positive real axis. We eq4fs
choose instead the solution
pr:baru1
1
u1
= q sin
u1 du1
|
dx x=a
2
c
x u1 .
c2
(4.61)
The function u1 has the properties

u1 (a) = 0
du1

= 1.
dx x=a
and
This satisfies the differential equation.

"
18 Jan p8
d2
+
u1 (x, ) = 0.
dx2 c2
#
(4.62)
So u1 is analytic for with no branch cut. Similarly, for the substitution justify this
2
u2 u2 /( du
| ) we obtain
dx x=b
u2 = (/c2 )1/2 sin
/c2 (L x).
One can always find u1 and u2 as analytic functions of with no branch

cut.
Is this always
true?
4.5.4
Analytic GF for Fixed String
We have been considering the Greens function equation

[L0 (x)]G(x, x0 ; ) = (x x0 )
for 0 < x, x0 < b
(4.63)
with the open string RBC (1.18)

[
nS + S ]G(x, x0 ; ) = 0
for x on S
(4.64)
and linear operator L0 defined as

L0 =
d
d
(x) + V (x).
dx
dx
(4.65)
48
We found that the solution to this equation can be written (3.41)

G(x, x0 ; ) =
eq4cD
u1 (x< , )u2 (x> , )

(x)W (u1 , u2 )
(4.66)
where u1 and u2 are solutions to the homogeneous equation with the

same boundary conditions as (1.18)
[L0 (x)]u1,2 (x, ) = 0
for a < x < b
(4.67)
u1 (x, ) + ka u1 (x, ) = 0
for x = a
(4.68)
x
for x = b.
(4.69)
+ u2 (x, ) + kb u2 (x, ) = 0
x
We have been calculating the Greens function for a string with
fixed tension ( = constant) and fixed string density ( = constant) in
the absence of a potential field (V = 0) and fixed end points. This last
condition implies that the Greens function is restricted to the boundary
condition that G = 0 at x = a = 0 and x = b = L. We saw that
u1 = sin
and
x
c2
(L x).
c2
We also assigned the convention that
q
= ||ei/2 .
u2 = sin
22 Jan p2
(4.70)
(4.71)
(4.72)
This is shown in figure 4.3.

The Wronskian in equation 4.53 for this problem can be simplified
as
u2
u1
u2
x
x
s
s
s
=
sin
x cos
(L x)
2
2
c
c
c2
W (u1 , u2 ) = u1

s
sin
49
(L x) cos
c2
= 2 sin
c

x
c2
L.
c2
(4.73)
Thus we can write the full solution for the fixed string problem as
0
G(x, x ; ) =
sin
x
c2 <
sin
c2
sin
(L
c2
L
c2
x> )
eq4.75b
(4.74)
eq4ss
4.5.5
GF Properties
We may now summarize the properties of G in the complex plane.

Branch Cut: G has no branch cut. It is analytic except at isolated
simple poles. The 1/2 branch vanishes if the eigen functions are
properly chosen. This is a general result for discrete spectrum. pr:DiscSpec1
Justify
Asymptotic limit: G goes to zero as goes to infinity. If is

pr:asymp1
real and we do not go through the poles, this result can be seen
immediately from equation 4.74. For complex , we let ||
and use the definition stated in equation
4.57 which is valid for
0 < < 2. This definition gives Im > 0. We can then write
i /c2 x
i /c2 x
q
e
e
sin /c2 x
=
(4.75)
2i
i /c2 x
|| e
(4.76)
2i
for > 0. Thus from equation 3.44 we get
i /c2 x< i /c2 (Lx> )

1e
e
||
G(x, x0 ; )
q
i /c2 L
2i
2
/c e
2
1 e+i /c (x> x< )
q
=
.
2i
/c2
23 Jan p3
(4.77)
(4.78)
50

By convention x> x< > 0, and thus we conclude
||
G(x, x0 ; ) 0.
(4.79)
eq4.81a
Poles: The Greens function can have poles. The Greens function
is a ratio of analytic functions. Thus the poles occur at the zeros
of the denominator.
We now look at sin
n
L
c2
= 0 from the denominator of equation
3.44. The poles are at = n . We can write

cn
n =
L

eq4et
23 Jan p4
n /c2 L = n or
2
for n = 1, 2, . . .
(4.80)
We delete the case n = 0 since we have a removable singularity at

= 0. Equation 4.80 occurs when the Wronskian vanishes. This
happens when u1 = constant u2 (not linearly independent). Both
u1 and u2 satisfy the boundary conditions at both boundaries and are
therefore eigenfunctions. Thus the un s are eigenfunctions and the n s
are the eigenvalues. So
[L0 n ]un (x) = 0
eq4star
RBC
(4.81)
is satisfied for n by un .
4.5.6
The GF Near an Eigenvalue
We now look at equation 4.74 near an eigenvalue. First we expand the

denominator in a power series about = n :
s
( n ) Lc cos n Lc
sin
L=
+ O( n )2 .
(4.82)
c2
2 n
eq4.85
So for near an eigenvalue we have

s
sin
c2
n L
L 2 ( n ) cos n.
c2
2c
(4.83)
4.6. DERIVATION OF GF FORM NEAR E.VAL.
51
eq4.86
Now we look at the numerator of 4.74. We can rewrite

sin
eq4.87
/c2 (L x> ) = sin nx> cos /c2 L.
(4.84)
Note that f (x< )f (x> ) = f (x)f (x0 ). So, with = /c2 , and substituting 4.83 and 4.84, equation 4.74 becomes
0
sin nx
2 sin nx
L
L
.
G(x, x ; )
L
n
0
(4.85)
So we conclude that
un (x)un (x0 )
n
as in 4.47 where the eigenfunction is
G n
un (x) =
2
nx
sin
,
L
L
which satisfies the completeness relation
RL
0
(4.86)
eq4cC2
22 Jan p5
(4.87)
um (x)un (x)dx = mn .
eq4.91
pr:CompRel1
4.6
4.6.1
Derivation of GF form near E.Val.

Reconsider the Gen. Self-Adjoint Problem
We now give an indirect proof of equation 4.86 based on the specific

Greens function defined in equation 4.53,
G(x, x0 ; ) =
u1 (x< , )u2 (x> , )

.
(x)W (u1 , u2 )
(4.88)
The boundary conditions are (see 1.18)
u1
+ ka u1 = 0
x
eq4nt
for x = a = 0
(4.89)
u2
+ kb u2 = 0
for x = b = L.
(4.90)
x
The function u1 (respectively u2 ) may be any solution which is an
analytic function of , and independent of at x = a (respectively
+
52
x = b). Thus both the numerator and the denominator of equation

4.88 are analytic functions of , so there is no branch cut. Note that 22 Jan p6
G(x, x0 ; ) may only have poles when W (u1 (x, )u2 (x, )) = 0, which
only occurs when
u1 (x, n ) = dn u2 (x, n ).
(4.91)
where the dn are constants.
Look at the Greens function near = n . Finding the residue will
give the correct normalization. We have (using 4.73 and 4.82)
(x)W (u1 , u2 ) n ( n )cn ,
(4.92)
where cn is some normalization constant. In this limit equation 4.88

becomes
1
n (x< )
un (x> )
n dn u
0
(4.93)
G(x, x ; )
( n )cn
where dn is some constant. So un un
function. Equation 4.88 then implies
0
1
cn dn
is the normalized eigen-
0
n un (x)un (x )
G(x, x ; )
( n )
(4.94)
where un satisfies equation 4.81.
4.6.2
Summary, Interp. & Asymptotics
25 Jan p1
In the previous sections we looked at the eigenvalue problem

[L0 n (x)]un (x) = 0
for a < x < b, RBC
(4.95)
and the Greens function problem

[L0 n (x)]G(x, x0 ; ) = (x x0 )
eq4fooo
where
for a < x, x0 < b, RBC (4.96)

!
d
d
(x)
+ V (x),
L0 =
dx
dx
pr:selfadj1
(4.97)
which is a formally self-adjoint operator. The general problem requires
4.7. GENERAL SOLUTION FORM OF GF
53
finding an explicit expression for the Greens function for a force localized at x0 .
For n we found
G(x, x0 ; )
un (x)un (x0 )
.
n
(4.98)
This equation shows the contribution of the nth eigenfunction. We saw eq4.102
that G is an analytic function of (with poles at n ) where G 0 as
|| . We can think of G as the inverse operator of L0 n :
G=
1
.
L 0 n
(4.99)
Thus the poles are at L0 = n .

d2
For large the behavior of G is determined by the dx
2 G term since
it brings down the highest power of . Thus for our simple example of
constant,
d2
L0 2
(4.100)
dx
and for large
q
G exp(i /c2 x),

G0 G,
G00 ( )2 G
(4.101)
where the derivatives are taken with respect to x.
4.7
General Solution form of GF

25 Jan p2
In this section we obtain a general form (equation 4.108) for the Greens fig4ff
function which is constructed using the solutions to the corresponding
eigen value equation. This is done by evaluating a particular complex
integral. We have seen that G(x, x0 ; ) is analytic in the complex plane except for poles on the real axis at the eigen values n .
We consider the following complex integral
I
c1 +c2
d0 G(x, x0 ; 0 ) I
d0 F (0 )
0
c1 +c2
(4.102)
where we have defined F (0 ) G(x, x0 ; 0 )/(0 ). Let the contour of

integration be the contour illustrated in figure 4.4. This equation has
54

Im 0
0 -plane

PP
P
Q
Q
k
Q
Q C2
S
J
q
J
J
B
C1
B
B Re 0

B
B
B
J
J
J
QQ
QP
PP

3

Figure 4.4: The contour of integration
pr:Cauchy1
a singularity only at . Thus we need only integrate on the contour pr:singular1

around . This is accomplished by deforming the contour C1 + C2 to
the contour S (following Cauchys theorem)
I
d0 F (0 ) =
C1 +C2
d0 F (0 ).
(4.103)
See figure 4.5. Note that although F (0 ) blows up as 0 approaches ,

G(x, x0 ; 0 ) G(x, x0 ; ) in this limit. Thus the integration about the
small circle around can be written
I
I
0
d0
0
0
0
d F ( ) G(x, x ; )
.
S
S 0
Now make the substitution
0 = ei
(4.104)
d0 = iei d
d0
= id.
0
(4.105)
This allows us to write

I
S
Z 2
d0
=
i
lim
d = 2i.
0 0
0
(4.106)
55
#

q
"!
Figure 4.5: Circle around a singularity.

We conclude
d0 F (0 ) = 2iG(x, x0 ; )
and thus
I
2iG(x, x ; ) =
d0
C1 +C2
G(x, x0 ; 0 )
.
0
(4.107)
eq4tpig
We now assume that G(x, x ; ) 0 as . We must check

this for each example we consider. An intuitive reason for this limit is
the following. The Greens function is like the inverse of the differential
operator: G 1/(L0 ). Thus as becomes large, G must vanish.
This assumption allows us to evaluate the integral around the large
circle C2 . We parameterize 0 along this contour as
0 = Rei ,
d0 Rei id
as R .
So
lim
R C2
Z 2
G(x, x0 ; 0 ) 0
Rei d
d
=
lim
G(x, x0 ; Rei ) 0
R 0
0
Rei
since G(x, x0 ; Rei ) 0 as R .

We now only need to evaluate the integral for the contour C1 .
I
C1
fig4fof
d0 G(x, x0 ; 0 ) X I d0 G(x, x0 ; 0 )
=
.
0
0
cn
n
In this equation we replaced the contour C1 by a sum of contours around

the poles, as shown in figure 4.6. Recall that
fig4.6
0
0
0 n
G(x, x ; )
0
un (x)u
n (x )
.
n 0
56

q1

-C1
q2

q3
q4
...
c1
c2
c3
c4

q2
q3
q4
...
q1
=

Figure 4.6: Division of contour.

We note that
d0
1 I
d0
=
(0 )(n 0 )
n cn n 0
1 I
d0
=
n 0 n
1
2i.
=
n
I
cn
The first equality is valid since 1/(0 ) is well behaved as 0 n .

The last equality follows from the same change of variables performed
above. The integral along the small circle containing n is thus
I
cn
d0
2i X
0
0
0
G(x,
x
;
)
=
u (x)u
n (x ).
0
n n
The integral along the contour C1 (and thus the closed contour C1 +C2 )
is then
I
C1
X
d0
0
0
G(x,
x
;
)
=
2i
0
n
0
un (x)u
n (x )
.
n
Substituting equation 4.107 gives the result

!
G(x, x ; ) =
X
n
eq4.124
X
1
0
un (x)u
n (x ) ,
n
(4.108)
where the indices n sum n = 1, 2, . . . and sums over the degeneracy.
4.7.1
57
-fn Representations & Completeness
Using the above result (equation 4.108) we can write

(x x0 ) = [L0 (x)]G(x, x0 ; )
X
un (x)un (x0 )
= [L0 (x)]
n
n=1
n=1
[(L0 n ) + (n )]
(4.109)
(4.110)
un (x)un (x0 )
(4.111)
n
un (x)un (x0 ).
(4.112)
n=1
In the last equality we used
X
n=1
(L0 n )
un (x)un (x0 ) X
un (x0 )
=
(L0 n )un (x) = 0
n
n=1 n
since L0 is a differential operator in terms of x, From this we get the

completeness relation
pr:CompRel2
(x x ) =
(x)un (x)un (x0 )
(4.113)
n=1
or
(x x0 ) X
=
un (x)un (x0 ).
(x)
n=0
(4.114)
This is called the completeness relation because it is only true if the eq4.128b
un are a complete orthonormal set of eigenfunctions, which means that
any f (x) can be written as a sum of the un s weighted by the projection
of f (x) onto them. This notion is expressed by the expansion theorem
(4.117 and 4.118).
We now derive the expansion theorem. Consider
pr:ExpThm1
f (x) =
a
b
So
25 Jan p6
dx0 f (x0 )(x x0 )

0
dx f (x )(x )
for a < x < b

un (x)un (x0 )
(4.115)
(4.116)
n=1
eq4.129
58

f (x) =
un (x)fn ,
(4.117)
n=1
where
fn =
eq4.132
pr:FourCoef1
pr:normal2
eq4fo32
0 0
dx u (x )(x )f (x )
(4.118)
is the generalized nth Fourier coefficient for f (x). Equation 4.117 represents the projection of f (x) onto the un (x) normal modes. This was
obtained using the completeness relation.
Now we check normalization. The Greens function G is normalized
because the us are normalized. We check the normalization of the us
by looking at the completeness relation
(x x0 ) = (x0 )
un (x)un (x0 ).
(4.119)
Integrate both
sides by dx0 um (x0 ). On the left hand side we immediR
ately obtain dx0 um (x0 )(x x0 ) = um (x). On the right hand side
R
dx um (x )(x )
un (x)un (x0 )
un (x)
dx0 un (x0 )(x0 )um (x0 )
where we used 4.11 in the equality. But

um (x) =
un (x)n,m .
Thus we conclude that normalized eigen functions are used in the completeness relation:
Z
pr:orthon2
dxum (x)(x)un (x) = n,m .
(4.120)
This is the condition for orthonormality.
4.8
Extension to Continuous Eigenvalues
27 Jan p1
27 Jan p2
As L (the length of the string) becomes large, the eigen values become
closer together. The normalized eigen functions un (x) and eigen values
n for the fixed string problem, equation 4.54, can be written
1
2
(4.121)
un = ei n /c x
L
4.9. ORTHOGONALITY FOR CONTINUUM

factor of 2?
59
and
2n
n = 0, 1, 2, . . . .
L
The separation between the eigen values is then
n = (ckn )2
kn =
n = n n1 c2
2
L
2
n2 (n 1)2 0.
(4.122)
(4.123)
We now consider the case of continuous eigen values. Let be

complex (as before) and let n 0. This limit exists as long as
is not on the positive real axis, which means that the denominator
will not blow up as n 0. In the continuum case equation 4.108
becomes
0
lim G(x, x ; ) =
n 0
dn X
un (x)u
n (x).
n
(4.124)
eq4.124ab
The completeness relation, equation 4.114, becomes

X
(x x0 ) Z
=
d
un (x)un (x0 ).
n
(x0 )
(4.125)
Now we take any function f (x) and express it as a superposition using

the -function representation (in direct analogy with equations 4.115
and 4.118 in the discrete case)
f (x) =
dn
fn un (x).
(4.126)
This is the generalized Fourier integral, with generalized Fourier coef- eq4.175
ficients
Z
pr:GenFourInt1
fn = dxu
(x)(x)f
(x).
(4.127)
n
The coefficients fn may be interpreted as the projection of f (x) with eq4.176
respect to (x) onto the eigenfunction un (x).
4.9
Orthogonality for Continuum

27 Jan p3
We now give the derivation of the orthogonality of the eigenfunctions

for the continuum case. The method of derivation is the same as we
60
used in the discrete spectrum case. First we choose f (x) = um (x) for
f (x) in equation 4.126. Equation 4.127 then becomes
0
f0n
dxu0n (x)um (x).
(4.128)
The form of equation 4.126 corresponding to this is

um (x)
XZ
d0m f0n u0n (x).
(4.129)
This equation can only be true if

0
f0n = 0 (0n m ).
(4.130)
So we conclude that
Z
dxu0n (x)um (x) = f0n = 0 (0n m ).
(4.131)
This is the statement of orthogonality, analogous to equation 4.114. All

of these results come from manipulations on equation 4.108. We now
have both a Fourier sum theorem and a Fourier integral theorem.
We now investigate equation 4.124 in more detail
0
G(x, x ; ) =
27 Jan p4
fig4555
pr:branch2
X
dn
u (x)un (x0 ) .
n n
(Generally, the integration is over the interval from zero to infinity.)

Where are the singularities? Consider approaching the positive real
axis. It cant ever get there. The value approaching the negative side
may be different from the value approaching the positive side. Therefore this line must be a branch cut corresponding to a continuous spectrum. Note that generally there is only a positive continuous spectrum,
although we may have a few negative bound states. So G(x, x0 ; ) is
analytic on the entire complex cut -plane. It is in the region of nonanalyticity that all the physics occurs. The singular difference of a
branch cut is the difference in the value of G above and below. See
figure 4.7.
We now examine the branch cut in more detail. Using equation
4.9. ORTHOGONALITY FOR CONTINUUM
61
0
r + i
r
0
r i
0
Figure 4.7: near the branch cut.

4.108 we can write
G(x, x0 ; 0 + i) G(x, x0 ; 0 i)
0
2i

X
1 Z
1
1
0
=
dn
un (x)u
(x
)
n
2i 0
n 0 i n 0 + i
lim
where
1
1
1
0
2i n i n 0 + i

1
2i
2i (n 0 )2 + 2

1
.
=
(n 0 )2 + 2
In the first problem set we found that

1
= (0 ).
0 ( 0 )2 + 2
lim
(4.132)
So
G(x, x0 ; 0 + i) G(x, x0 ; 0 i)
lim
0
2i
Z
=
=
dn
0
0
un (x)u
n (x )( n )
u0 (x)u0 (x0 ).
Therefore the discontinuity gives the product of the eigenfunctions.

We derived in the second problem set the property
G (x, x0 ; ) = G(x, x0 ; ).
(4.133)
27 Jan p5
62
Now take = 0 + i. This allows us to write

G (x, x0 ; 0 + i) = G(x, x0 ; 0 i)
(4.134)
and
G(x, x0 ; 0 + i) G(x, x0 ; 0 i)
2i
G(x, x0 ; 0 + i) G (x, x0 ; 0 + i)
=
2i
1
=
ImG(x, x0 ; 0 + i)
X
0
=
u0 (x)u
0 (x ).
(4.135)
(4.136)
(4.137)
(4.138)
eq4.151
So we can say that the sum over degeneracy of the bilinear product
of the eigen function u is proportional to the imaginary part of the
Greens function.
4.10
Example: Infinite String
29 Jan p1
pr:InfStr1
Consider the case of an infinite string. In this case we take the end
points a , b and the density , tension , and potential V
as constants. The term V is the elastic constant of media.
4.10.1
The Greens Function
In this case the Greens function is defined as the solution to the equation
"
d2
2 + V G(x, x0 ; ) = (x x0 )
dx
#
for < x, x0 < .
(4.139)
To get the solution we must take (= 2 ) to be imaginary. The solution
for the Greens function can be written in terms of the normal modes
(3.41)
u1 (x< )u2 (x> )
G(x, x0 ; ) =
(4.140)
W (u1 , u2 )
4.10. EXAMPLE: INFINITE STRING

Im

u
63
u

Re
2 2
ck
Figure 4.8: specification.

where u1 and u2 satisfy the equation
"
d2
2 + V u1,2 = 0.
dx
#
(4.141)
The boundary conditions are that u1 is bounded and converges as x

and that u2 is bounded and converges as x . Divide both
sides of equation 4.141 by and substitute the definitions / = 1/c2
and V / = k 2 . This gives us the equation
"
d2
c2 k 2
+
u(x) = 0.
dx2
c2
eq4.139
and converges
- R. Horn
pr:k2.1
(4.142)
The general solution to this equation can be written as
u(x) = Aei
c2 k2 x/c
+ Bei
c2 k2 x/c
eq4.132a
(4.143)
We specify the root by the angle extending around the point c2 k 2 on eq4.132
the real axis, as shown in figure 4.8. This is valid for 0 < < 2. The fig4.6a
correspondence of is as follows:
c2 k 2 = | c2 k 2 |ei .
(4.144)
Thus
c2 k 2 =
| c2 k 2 |ei/2
(4.145)
!
| c2 k 2 | cos
+ i sin
.
2
2
(4.146)
64
So that
q
c2 |k 2 |
(4.147)
i c2 |k 2 |
(4.148)
2 c2 |k 2 |.
(4.149)
This is good everywhere except for values on the real line greater than
c2 k 2 .
4.10.2
Uniqueness
29 Jan p2
The Greens function is unique since it was found using the theory of
ordinary differential equations. We identify the fundamental solutions
u1 , u2 by looking at the large x behavior of 4.143.
ei
c2 k2 x/c x
and
ei
and
i c2 k2 x/c x +
so
c2 k2 x/c x +
(4.150)
(4.151)
i c2 k2 x/c x
u1 (x) = ei
c2 k2 xc
i c2 k2 xc
u2 (x) = e
(4.152)
(4.153)
The boundary condition has an explicit dependence on so the solution

is not analytic. Notice that this time there is no way to get rid of the
branch cut. The branch cut that comes in the solution is unavoidable
because satisfaction of the boundary condition depends on the value of
.
4.10.3
Look at the Wronskian
In the problem we are considering we have

W (u1 , u2 ) = u1 u02 u2 u01

i
i
= + c2 k 2 c2 k 2
c
c
2i
=
c2 k 2 .
c
(4.154)
(4.155)
(4.156)
4.10. EXAMPLE: INFINITE STRING
4.10.4
65
Solution
This gives the Greens function
G(x, x0 ; ) =
icei
c2 k2 x< /c i c2 k2 x> /c
e
2 c2 k 2
2 2
(4.157)
ic ei c k |xx |/c
=
.
2
c2 k 2
(4.158)
We now have a branch cut for Re () c2 k 2 with branch point at

= c2 k 2 . Outside of this, G is analytic with no poles. This is analytic
for in the cut -plane > c2 k 2 . Consider the special case of large 29 Jan p3
ic ei |xx |/c
0.
G
2
(4.159)
Notice that this is the same asymptotic form we obtained in the discrete
case when we looked at G as (see equation 4.79). Now consider
the case < c2 k 2 on the real axis. This corresponds to the case that
= . So
c2 k|xx0 |
c
e
G(x, x0 ; ) =
.
(4.160)
2
c2 k 2
This function is real and exponentially decreasing. For > c2 k 2 this eq4.159
function oscillates. If = 0, it oscillates one way, and if = 2 it
oscillates the other way, so the solution lacks uniqueness. The solutions
correspond to different directions of traveling waves.
4.10.5
Motivation, Origin of Problem
We want to understand the degeneracy in the Greens function for an

infinite string. So we take a look at the physics behind the problem.
How did the problem arise? It came from the time dependent problem
with forced oscillation imposed by an impulsive force at x0 :
"
d2
2
2 + V + 2 u(x, t) = (x x0 )eit
dx
t
#
where 2 < c2 k 2 .
(4.161)
66
We wanted the steady state solution:

u(x, t) = eit G(x, x0 ; ).
(4.162)
By substituting 4.160 this can be rewritten as
2 2
c eit e c k |xx |
u(x, t) =
.
2
c2 k 2 2
(4.163)
We consider the cases 2 < c2 k 2 and 2 > c2 k 2 separately.

If 2 < c2 k 2 , then there is a unique solution and the exponential
pr:WaveProp1 dies off. This implies that there is no wave propagation. This agrees
with what the physics tells us intuitively: k 2 c2 > 2 implies large k,
which corresponds to a large elastic constant V , which in turn means
there will not be any waves.
29 Jan p4
For > c2 k 2 there is propagation of waves. For > c2 k 2 we denote
the two solutions:
eq4.163ab
u (x, t) = eit G(x, x0 ; = 2 i)

pr:cutoff2
as 0.
(4.164)
As increases, it approaches the branch point. We define the cutoff

frequency as being at the branch point, c2 = c2 k 2 . We have seen that
for < c there is no wave propagation, and for > c there is
propagation, but we dont know the direction of propagation, so there
is no unique solution.
The natural appearance of a branch cut with two solutions means
that all the physics has not yet been given. We may rewrite equation
4.163 as:
0
2
2 2
eiti( c k /c)|xx |
u (x, t) =
.
(4.165)
2
2 c2 k 2
c
These solutions to the steady state problem can be interpreted as
follows. The solution u+ represents a wave traveling to the right for
points to the right of (i.e. on the positive side of) the source and a
wave traveling to the left for points to the left of (i.e. on the negative
side of) the source. Mathematically this means
(
u+ =
2 2
ei(tx c k /c) for x > x0

2
2 2
ei(t+x c k /c) for x < x0 .
4.11. SUMMARY OF THE INFINITE STRING
67
Similarly, the solution u represents a wave traveling to the left for

points to the right of the source and a wave traveling to the right for
points to the left of the source. Mathematically this means
(
u =
2 2
ei(t+x c k /c) for x > x0

2
2 2
ei(tx c k /c) for x < x0 .
These results can be rephrased by saying that u+ is a steady state

solution having only waves going out from the source, and u is a
steady state solution having only waves going inward from the outside
absorbed by the point. So the equation describes two situations, and
the branch cut corresponds to the ambiguity in the situation.
4.11
Summary of the Infinite String

2 Feb p1
We have considered the equation

(L0 )G = (x x0 )
where
L0 =
(4.166)
d2
+ V.
dx2
(4.167)
We found that
is V 6= 0?
G(x, x ; ) =
c2
k2
k2 |xx0 |
c2
(4.168)
where we have introduced the substitutions V / k 2 and / 1/c2 .

The time dependent response is
u (x, t) =
1
2
c2
iti
k2
k2
|xx0 |
c2
(4.169)
If 2 < c2 k 2 c2 there is exponential decay, in which case there is no

singularity of G at = 2 .
The other case is that
u (x, t) = eit G(x, x0 ; = 2 i).
(4.170)
68
1 Feb p2
This case occurs when 2 > c2 , for which there is a branch cut across
the real axis. In this case we have traveling waves.
Note that in the case that k = 0 we always have traveling waves.
The relevance of the equation k 2 = V / is that the resistance of the
medium to propagation determines whether waves are produced. When
k = 0 there is no static solution wave propagation always occurs. If
2 < c2 , then the period of the external force is small with respect to
the response of the system, so that the media has no time to respond
the system doesnt know which way to go, so it exponentially decays.
Recall that the solution for the finite string (open or closed) allowed
incoming and outgoing waves corresponding to reflections (for the open
string) or different directions ( for the closed string). Recall the periodic
boundary condition problem:
u(x, t) = eit G(x, x0 ; )
c cos[ c ( L2 |x x0 |)]
= eit
.
2
sin[ c 2l ]
(4.171)
(4.172)
This is equal to the combination of incoming and outgoing waves, which

can be seen by expanding the cosine. We need the superposition to
satisfy the boundary conditions and physically correspond to reflections
at the boundaries. The sum of the two waves superimpose to satisfy
the boundary conditions.
4.12
The Eigen Function Problem Revisited
1 Feb p3
pr:efp2
We now return to the the connection with the eigen function problem.
We have seen before that the expression
G( = 0 + i) G( = 0 i)
2i
(4.173)
vanishes if 0 < c2 k 2 . In the case that 0 > c2 k 2 we have

G( = 0 + i) G( = 0 i) X
=
u0 (x)(u0 (x0 )) .
2i
(4.174)
4.13. SUMMARY
69
q
q
c2 k 2
/c2 k 2 +|/c2 k 2 |1/2

q
BMB
/c2 k 2 |/c2 k 2 |1/2
Figure 4.9: Geometry in -plane

eq4777
The geometry of this on the -plane is shown in figure 4.9. This gives
us
G( = 0 + i) G( = 0 i)
2i
=
1
1
i| 2 k2 |1/2 |xx0 |
i| 2 k2 |1/2 |xx0 |
c
c
e
)
(e
2i 2 | c2 k 2 |1/2

1/2

1
1
k 2
=
sin
c2

2 | c2 k 2 |1/2
|x x0 |
1
ImG( = 0 + i).
For 0 > c2 k 2 we can write (using equation 4.138)
u
0 (x) = r
4
0
c2
ei
0 c2 k2 x/c
(4.175)
k2
for 0 > c2 k 2 . We now see that no eigen functions exist for 0 < c2 k 2 eq4.188
since it exponentially increases as 0 and we must kill both terms.
The Greens function is all right since C.
4.13
Summary
1. For the Helmholtz equation, n2 > 0, n2 is real, and the eigen

functions are real.
2. The dispersion relation for a closed massless string with discrete
mass points is
c2 sin2 (kn a/2)
n2 =
.
a2 /4
70

3. The Greens function obeys Hermitian analyticity:
G (x, x0 ; ) = G(x0 , x; ).
4. The form of the Greens function for near an eigen value n is
G(x, x0 ; ) n
un (x)un (x0 )
.
n
5. The Greens function for the fixed string problem is

sin
G(x, x0 ) =
x
c2 <
sin
c2
sin
c
q
(L x> )
L
c2
6. The completeness relation is

(x x0 ) =
(x)un (x)un (x).
n=1
7. The expansion theorem is

f (x) =
un (x)fn
n=1
where
fn =
dx0 un (x0 )(x0 )f (x0 ).
8. The Greens function near the branch cut is related to the eigen
functions by
X
1
ImG(x, x0 ; 0 + i) =
u0 (x)u0 (x0 ).
9. The Greens function solution for an infinite string is
2 2
ic ei c k |xx |/c
.
G(x, x0 ; ) =
2
c2 k 2
4.14. REFERENCES
4.14
71
References
The Rayleigh quotient is described in [Stakgold67a, p226ff] and [Stakgold79, p339ff].

For other ideas in this chapter, see Fetter and Stakgold.
A discussion of the discrete closed string is given in [Fetter80, p115].
The material in this chapter is also in [Fetter81, p245ff].
72
Chapter 5
Steady State Problems
Chapter Goals:
Interpret the effect of an oscillating point source on
an infinite string.
Construct the Klein-Gordon equation and interpret
its steady state solutions.
Write the completeness relation for a continuous
eigenvalue spectrum and apply it to the KleinGordon problem.
Show that the solutions for the string problem with
= x, = x, and V = m/x2 on the interval
0 < x < are Bessel functions.
moved
from
few pages later
Construct the Greens function for this problem.

Construct and interpret the steady state solutions
for this problem with a source point.
Derive the Fourier-Bessel transform.
5.1
Oscillating Point Source

pr:ops1
We now look at the problem with an oscillating point source. In the

notation of the previous chapter this is
3 Feb p1
73
74
"
CHAPTER 5. STEADY STATE PROBLEMS
2
L0 + (x) 2 u(x, t) = (x x0 )eit
t
#
< x, x0 < + R.B.C.
(5.1)
and can be written in terms of the Green function G(x, x0 ; ) which
satisfies
[L0 (x)]G(x, x0 ; ) = (xx0 )
< x, x0 < + R.B.C. (5.2)
The steady state solution corresponding to energy radiated outward to

infinity is
u(x, t) = eit G(x, x0 , = 2 + i).
(5.3)
1 Feb p5
The solution for energy radiated inward from infinity is the same equation with = 2 i, but this is generally not a physical solution.
This contrasts with the case of a finite region. In that case there are
no branch cuts and there is no radiation.
5.2
The Klein-Gordon Equation
pr:kge1
We now apply the results of the previous chapter to another physical

problem. Consider the equations of relativistic quantum mechanics. In
the theory of relativity we have the energy relation
E 2 = m2 c4 + p2 c2
(5.4)
In the theory of quantum mechanics we treat momentum and energy

as operators
p i
h
E i
h
t
(5.5)
(5.6)
where
dim[
h] = Action
(5.7)
We want to derive the appropriate wave equation, so we start with

E 2 (m2 c4 + p2 c2 ) = 0
(5.8)
5.2. THE KLEIN-GORDON EQUATION
75
Now substitute the operators into the above equation to get
i
h
t
!2
m2 c4 +
!2
c 2 = 0
(5.9)
This is the KleinGordon equation, which is a relativistic form of the pr:phi1

Schrodinger Equation. Note that ||2 still has a probability interpretation, as it does in non-relativistic quantum mechanics.
Now specialize this equation to one dimension.
1 Feb p6
"
2
d2
2
2 4
hc
+
m
c
+
h
(x, t) = 0
dx2
t2
2 2
(5.10)
This is like the equation of a string (c.f., 1.11). In the case of the eq5.energy
string the parameters were tension = dim[E/t], coefficient of elasticity
V = dim[E/l3 ], and mass density = [t2 E/l3 ], where E is energy, t
is time, and l is length. The overall equation has units of force over
length (since it is the derivative of Newtons second law). By comparing
equation 1.11 with equation 5.10 we note the correspondence
V m2 c4
h
2
h
2 c2
and
f (x, t) 0. (5.11)
Note that V / = mc/

h is a fundamental length known as the Compton wavelength, c . It represents the intrinsic size of the particle.
The expression expression / = 1/c2 shows that particle inertia
corresponds to elasticity, which prevents the particle from responding
quickly. Equation 5.10 has dimensions of energy squared (since it came ?
from an energy equation).
We look again for steady state solutions
0
(x, t) = eiE t/h E 0 (x)
(5.12)
to get the eigen value problem

"
d2
0
hc
+ m2 c4 E 2 E 0 (x) = 0
2
dx
#
2 2
(5.13)
Thus we quote the previous result (equation 4.175 which solves equation
4.142)
(5.14)
E 0 (x) = u (x)
76
where we let k 2 ( mc
)2 and 0 E 02 /
h2 . As a notational shorthand,
h
let p0 = E 02 m2 c4 /c. Thus we write
i( E 0 2 m2 c4 /
hc)x
e
q
(x)
=
E0
4
hc E 2 m2 c4
0
ep /h
=
.
4
hp0 c2
pr:rad1
The cut-off energy is mc2 . We have the usual condition on the solution
that 0 > c2 k 2 . This eigen value condition implies E 2 > m2 c4 .
In relativistic quantum mechanics, if E < mc2 , then no free particle
is emitted at large distances. In the case that E > mc2 there is radiation. Also note that m becomes inertia. For the case that m 0 there
is always radiation. This corresponds to V 0 in the elastic string
analogy. The potential V acts as an elastic resistance.
The Greens function has the form
G(x, x0 ; E) exp{( m2 c4 E 2 /
hc)|x x0 |}.
So G(x, x0 ; E) has a characteristic half-width of
|x x0 | h
c/ m2 c4 E 2 = h
/p.
This is a manifestation of the uncertainty principle. As x , for
E < mc2 , the Greens function vanishes and no particle is radiated,
while for E > mc2 the Greens function remains finite at large distances
which corresponds to the radiation of a particle of mass m.
5.2.1
Continuous Completeness
pr:CompRel3
3 Feb p2
Recall that the completeness condition in the discrete case is

(x x0 ) X
0
=
u0 (x)u
0 (x ).
(x)
0 ,
(5.15)
R. Horn says The corresponding equation for the case of continuous eigenvalues is
no 0 in sum.
Z
0
X
(x x )
=
(x)
0 ,=
c2
0
d0 u0 (x)u
0 (x ).
(5.16)
5.2. THE KLEIN-GORDON EQUATION
77
Recall that in this case the condition for an eigen function to exist is
0 > c2 . In this case
q
0
k2 x
c2
.
u
0 (x) = q
0
4 ( c2 k 2 )1/2
(5.17)
Substituting the u into the continuous completeness relation gives

0
0
c Z
d0
0
i 0 c2 ( xx
)
i 0 c2 ( xx
)
c
c
e
+e
(x x ) = (x)
0
4 c2 ( c2 k 2 )1/2
(5.18)
2
2 2
where c = k c . We now make a change of variables. We define the
wave number as
q
0 c2
k=
(5.19)
c
It follows that the differential of the wave number is given by
1
d0
q
dk =
.
2c 0 c2
(5.20)
With this definition we can write

(x x0 ) =
3 Feb p3
i
2c2 Z h ik(xx0 )
0
dk e
+ eik(xx ) .
4 0
(5.21)
Note the symmetry of the transformation k k. This property

allows us to write
1 Z
0
0
dkeik(xx ) .
(5.22)
(x x ) =
2
This is a Fourier integral. In our problem this has a wave interpretation.
We now apply this to the quantum problem just studied. In this
case 0 = (E/
h)2 and c2 = m2 c4 /
h2 . With these substitutions we can
write
2
k =
=
( Eh2
m2 c4 1/2
)
h2
c
1
h
E2
p
m2 c2 =
2
c
h
(5.23)
(5.24)
78
so
1 Z 0 i(xx0 )p0 /h
(x x ) =
dp e
.
2
0
pr:FI1
This Fourier integral has a particle interpretation.
5.3
3 Feb p4
(5.25)
The Semi-infinite Problem
Consider the following linear operator in the semi-infinite region

d
d
m2
x
+
L0 =
dx
dx
x
!
0 < x < .
(5.26)
Here we have let the string tension be = x and the potential be V =

m2 /x. We will let the density be (x) = x. This is like a centrifugal
potential. The region of consideration is 0 < x < .
This gives us the following Greens function equation (from 3.22)
"
d
d
m2
x
+
x G(x, x0 ; ) = (x x0 )
dx
dx
x
!
(5.27)
defined on the interval 0 < x, x0 < .

We now discuss the boundary conditions appropriate for the semiinfinite problem. We require that the solution be bounded at infinity,
as was required in the infinite string problem. Note that in the above
equation = 0 at x = 0. Then at that point the right hand side of
Greens second identity vanishes, as long as the amplitude at x = 0
is finite. Physically, 0 at x = 0 means that the string has a free
end. So there is a solution which becomes infinite at x = 0. This
non-physical solution is eliminated by the boundary condition that the
amplitude of that end is finite. Under these boundary conditions L0 is
hermitian for 0 x < .
The solution can be written in the form (using 3.41)
G(x, x0 ; ) =
eq5.29
u1 (x< , )u2 (x> , )

xW (u1 , u2 )
The function u1 and u2 are the solutions to the equation
(5.28)
5.3. THE SEMI-INFINITE PROBLEM
79
(L0 x)u1,2 = 0
(5.29)
where we restrict u1 to be that function which is regular at x = 0, and

u2 to be that function which is bounded at infinity.
We note that the equations are Bessels equations of order m:
pr:Bessel1
"
d2
d
+
y
+ (y 2 m2 ) Xm (y) = 0
y
dy 2
dy
#
(5.30)
where y = x and Xm (y) is any solution. The solutions are then
u1 (x) = Jm (x )
(5.31)
and
(1)
u2 (x) = Hm
(x ).
(5.32)
(1)
The function Hm
(x ) is known as the Hankel function. For large x 3 Feb p5
it may be approximated as
(1)
Hm
(x
s
x
m
2
ei(x 2 4 )
x
(5.33)
and since Im( ) > 0, we have decay as well as out going waves.
Now we get the Wronskian:
Justify this
(1)
W (u1 , u2 ) = W (Jm (x ), Hm
(x ))
(5.34)
= iW (Jm (x ), Nm (x ))
(5.35)
!
2
2i
= i
=
.
(5.36)
x
x
(1)
In the second equality we used the definition Hm
(x) = Jm (X)+iNm (x),
where
Jm (x) cos m Jm (x)
.
Nm (x)
sin m
The third equality is verified in the problem set. So W is independent
of x, as expected.
Therefore (using equation 5.28)
1 x
(1)
G(x, x0 ; ) =
Jm (x< )Hm
(x> )
(5.37)
x 2i
i
(1)
=
Jm (x< )Hm
(x> )
(5.38)
2
80
5.3.1
A Check on the Solution
Suppose that < 0. In this case we should have G real, since there is
no branch cut. We have
= i||1/2
at
= .
(5.39)
pr:Bessel2
We can use the definitions

Im (x) eim/2 Jm (ix),
(1)
Km (x) (i/2)eim/2 Hm
(ix)
to write
(1)
G(x, x0 ; ) = i Jm (i||1/2 x< )Hm
(i||1/2 x> )
(5.40)
2

i m
2
=
i Im (x< ||1/2 ) im Km (x> ||1/2
(5.41)
2
i
= Im (x< ||1/2 )Km (x> ||1/2 ) R
(5.42)
Note also that G 0 as x , so we have decay, and therefore no
propagation. This is because asymptotically
Im (z) (2z)1/2 ez
Km (z) (2z/)1/2 ez
5.4
| arg z| < 1/2, |z|

| arg z| < 3/2, |z| .
Steady State Semi-infinite Problem
5 Feb p1
For the equation

"
d
d
m2
x
+
x G(x, x0 ; ) = (x x0 )
dx
dx
x
!
for 0 < x, x0 <

(5.43)
the solution we obtained was

G(x, x0 ; ) =
i
(1)
Jm ( x< )Hm
( x> ).
2
(5.44)
We now look at the steady state solution for the wave equation,
"
2
L0 + x 2 u(x, t) = (x x0 )eit .
t
#
(5.45)
5.4. STEADY STATE SEMI-INFINITE PROBLEM

pr:rad2
81
We only consider outgoing radiation ( 2 2 + i). We take > 0,

since < 0 corresponds to incoming radiation. So (x x0 ) acts as a
point source, but not as a sink.
u(x, t) = eit G(x, x0 ; = 2 + i)
i
(1)
= eit Jm (x< )Hm
(x> )
2
i
(1)
= eit Jm (x0 )Hm
(x> )
2
(5.46)
(5.47)
for x > x0 .
(5.48)
Next let x 1, so
i(tx)
i
0 e
u(x, t) = Jm (x )
2
x
for x 1.
The condition x 1 allows us to use the asymptotic form of the 5 Feb p2

Hankel function.
In this case we have outgoing (right moving) waves. These waves
are composed of radiation reflected from the boundary x = 0 and from
direct radiation. If in addition to x 1 we take x0 0, then we Explain why?
have
i
ei(tx)
(5.49)
u(x, t) = (x0 )m
x
2
We now look at the case x < x0 with large. In this case we have
i (1)
H (x0 )Jm (x)
2 m
i (1)
1 (1)
(2)
(x0 ) [Hm
(x) + Hm
(x)]
= eit Hm
2
2
"
#
eix
eix
it (1)
0
e
Hm (x )
+
x 1
x
x
u(x, t) = eit
(1)
(x0 )[ei(tx) + ei(t+x) ].
Hm
We now look at the Greens function as a complete set of eigen

functions. First we consider
5 Feb p3
1
[G(x, x0 ; 0 + i) G(x, x0 ; 0 i)]
2i
82
1
(1)
(1)
[Jm ( 0 x< )Hm
( 0 x> ) Jm ( 0 x< )Hm
( 0 x> )]
2i
1
=
[Jm ( 0 x> )][Hn(1) ( 0 x) + Hn(2) ( 0 x)]
4
1
=
Jm ( 0 x)Jm ( 0 x0 )
2
1
Im G(x, x0 ; 0 + i).
=
5.4.1
The Fourier-Bessel Transform
The eigen functions u0 satisfy

"
5 Feb p4
d
d
m2
x
+
x u0 = 0
dx
dx
x
!
for 0 < x < .
(5.50)
In this case since there is a boundary at the origin, waves move only to
the right. There is no degeneracy, just one eigen function:
s
u0 =
1
Jm ( 0 x).
2
(5.51)
We know from the general theory that if there is no degeneracy, then

Z
1
(x x0 ) =
d0 u0 (x)u0 (x0 )
(x)
0
(5.52)
1
1Z 0
0
(x x ) =
d Jm ( 0 x)Jm ( 0 x0 )
x
Z2
(5.53)
so
0 d 0 Jm ( 0 x)Jm ( 0 x0 ).
(5.54)
This is valid for 0 < x, x0 < . Thus for f (x) on 0 < x < we have
f (x) =
Z
0
=
=
Z0
0
dx0 f (x0 )(x x0 )

dx0 f (x0 )x0
0 d 0 Jm ( 0 x)Jm ( 0 x0 )
d 0 Jm ( 0 x)
(5.55)
Z
0
dx0 x0 f (x0 )Jm ( 0 x0 ).
(5.56)
(5.57)
5.5. SUMMARY
5 Feb p5
83
Thus for a given f (x) on 0 < x < we can write

f (x) =
0 d 0 Jm ( 0 x)Fm ( 0 ).
(5.58)
This is the inversion theorem.

Fm () =
Z
0
x0 dx0 f (x0 )Jm (x0 ).
(5.59)
This is the Fourier-Bessel transform of order m.
5.5
Summary
Explain why
this is useful?
pr:FBt1
1. The string equation of an oscillating point source on an infinite

string has solutions corresponding to energy radiated in from or
out to infinity.
2. The Klein-Gordon equation is
"
2
d2
2
2 4
hc
+
m
c
+
h
(x, t) = 0.
dx2
t2
2 2
Steady-state solutions for a point source with |E| > mc2 correspond to a mass m particle radiated to () infinity, where as
solutions with |E| < mc2 die off with a characteristic range of
xh
/p.
pr:chRan1
3. The string problem with = x, = x, and V = m/x2 on the
interval 0 < x < corresponds the Bessels equation
d2
d
y
+ y + (y 2 m2 ) Xm (y) = 0
2
dy
dy
where y = x . The linearly independent pairs of solutions

to this equation are the various Bessel functions: (i) Jm (y) and
(1)
(2)
Nm (y), and (ii) Hm
(y) and Hm
(y).
"
4. The Greens function for this problem is
i
(1)
G(x, x0 ; ) = Jm ( x< )Hm
( x> ).
2
84

5. The steady state solutions for this problem with point source are
u(x, t) = eit
i
(1)
Jm (x< )Hm
(x> ).
2
The outgoing solutions consist of direct radiation and radiation

reflected from the x = 0 boundary.
6. The Fourier-Bessel transform is
Fm ( 0 ) =
Z
0
x0 dx0 f (x0 )Jm (x0 ).
The inversion theorem for this transform is

f (x) =
Z
0
5.6
0 d 0 Jm ( 0 x)Fm ( 0 ).
References
The Greens function related to Bessels equation is given in [Stakgold67a, p75].
Chapter 6
Dynamic Problems
8 Feb p1
Chapter Goals:
State the problem which the retarded Greens function GR solves, and the problem which the advanced Greens function GA solves. Give a physical
interpretation for GR and GA .
Show how the retarded Greens function can be
written in terms of the Greens function which
solves the steady state problem.
Find the retarded Greens function for an infinite
string with and constant, and V = 0.
Find the retarded Greens function for a semiinfinite string with a fixed end, and constant,
and V = 0.
Find the retarded Greens function for a semiinfinite string with a free end, and constant,
and V = 0.
85
4 Jan p1
p1prv.yr.
86
CHAPTER 6. DYNAMIC PROBLEMS

Explain how to find the retarded Greens function
for an elastically bound semi-infinite string with
and constant, and V = 0.
Find an expression for the retarded Greens function in terms of the eigen functions.
Show how the retarded boundary value problem
can be restated as an initial value problem.
6.1
pr:impf1
Advanced and Retarded GFs
Consider an impulsive force, a force applied at a point in space along

the string at an instant in time. This force is represented by
(x)f (x, t) = (x x0 )(t t0 ).
(6.1)
As with the steady state problem we considered in chapter 5, we apply

no external forces on the boundary. If we can solve this problem, then
we can solve the problem for a general time dependent force density
f (x, t).
We now examine this initial value problem (in contrast to the steady
state problems considered in the previous chapter). We begin with the
string at rest. Then we apply a blow at the point x0 at the time t0 . For
this physical situation we want to find the solution
u(x, t) = GR (x, t; x0 , t0 )
pr:GR1
(6.2)
where GR stands for the retarded Greens function:

"
2
L0 + 2 GR (x, t; x0 , t0 ) = (x x0 )(t t0 )
t
for a < x, x0 < b; all t, t0 .
#
Now we look at the form of the two possible Regular Boundary

Conditions. These two sets of conditions correspond to the case of an
open and closed string. In the case of an open string the boundary
condition is characterized by the equation
[
ns + s ]GR (x, t; x0 , t0 ) = 0
x S, a < x0 < b; t, t0
(6.3)
6.2. PHYSICS OF A BLOW

eq6RBC
87
where S is the set of end points {a, b}. In the case of a closed string
the boundary condition is characterized by the equations
GR (x, t; x0 , t0 )|x=a = GR (x, t; x0 , t0 )|x=b

for a < x0 < b, t, t0 , (6.4)

GR (x, t; x0 , t0 )
=
GR (x, t; x0 , t0 )
x
x
x=a
x=b
for a < x0 < b, t, t0 .

(6.5)
We now apply the condition that the string begins at rest:

GR (x, t; x0 , t0 ) = 0
for t < t0 .
8 Feb p2
(6.6)
This is called the retarded Greens function since the motionless string
becomes excited as a result of the impulse. This causeeffect relationship is called causality. The RBCs are the same as in previous chapters pr:caus1
but now apply to all times.
Another Greens function is GA , which satisfies the same differential
equation as GR with RBC with the definition
GA (x, t; x0 , t0 ) = 0
for t > t0
(6.7)
This is called the advanced Greens function since the string is in an pr:AGF1
excited state until the impulse is applied, after which it is at rest. In
what follows we will usually be concerned with the retarded Green
function, and thus write G for GR (suppressing the R) except when
contrasting the advanced and retarded Green functions.
6.2
Physics of a Blow
We now look at the physics of a blow. Consider a string which satisfies

the inhomogeneous wave equation with arbitrary force (x)f (x). The
momentum applied to the string, over time t, is then
pr:momch1
p = p(t2 ) p(t1 )
Z t2
dp
=
dt
dt
t1
=
t2
t1
dt
x2
x1
dx(x)f (x, t).
88
8 Feb p3
The
third equality holds because dp/dt is the force, which in this case
R
is xx12 dx(x)f (x, t). We now look at the special case where the force is
the function. In this case
p =
t2
t1
dt
x2
(t t0 )(x x0 )dx = 1
(6.8)
x1
for x1 < x0 < x2 , t < t0 < t2 . Thus a delta force imparts one unit of
momentum. Therefore we find that GR is the response of our system
to a localized blow at x = x0 , t = t0 which imparts a unit impulse of
momentum to the string.
6.3
Solution using Fourier Transform
We consider the Greens function which solves the following problem

given by a differential equation and an initial condition:
"
2
L0 + 2 G(x, t; x0 , t0 ) = (t t0 )(x x0 )
t
#
+ RBC,
(6.9)
eq6de1
G(x, t; x0 , t0 ) = 0
for t < t0 .
For fixed x we note that G(x, t; x0 , t0 ) is a function of t t0 and not t

and t0 separately, since only 2 /t2 and t t0 appear in the equation.
Thus the transformation t t + a and t0 t0 + a does not change
anything. This implies that the Greens function can be written
G(x, t; x0 , t0 ) = G(x, x0 ; t t0 )
= G(x, x0 ; ),
pr:FTrans1
8 Feb p4
eq6Ftran
where we define t t0 . By this definition, G = 0 for < 0.

This problem can be solved by taking the complex Fourier transform:
Z
x0 ; ) =
G(x,
d ei G(x, x0 ; ).
(6.10)
x0 ; ) is convergent everywhere in the upper half Note that G(x,

plane, which we now show. The complex frequency can be written as
= R + iI . Thus
ei = eiR eI .
6.3. SOLUTION USING FOURIER TRANSFORM
89
x0 ; ) to exist, the integral must converge. Thus we require

For G(x,
ei 0 as , which means we must have eI 0 as .
This is only true when is in the upper half plane, I > 0. Thus G
A everything is
exists for all such that Im > 0. Note that for G
reversed and is defined in the lower half plane.
x0 ; ) in the Fourier
By taking the derivative of both sides of G(x,
transform, equation 6.10, we have
Z
Z
d
d
G(x, x0 ; ) =
d ei G(x, x0 ; ) = i
d G(x, x0 ; )ei .
d
d
which is finite. Therefore the derivative exists everywhere in the upper

is analytic in the upper half -plane. We have
half -plane. Thus G
thus seen that the causality condition allows us to use the Fourier transform to show analyticity and pick the correct solution. The condition
that G = 0 for < 0 (causality) was only needed to show analyticity;
it is not needed anymore.
We now Fourier transform the boundary condition of an open string
6.3:
0 =
d ei ((
nS + S )G)
= (
nS + S )
d eit G
= (
nS + S )G.
Similarly, in the periodic case we regain the periodic boundary condia = G
b and smoothness G
0a = G
0 . So G
satisfies
tions of continuity G
b
the same boundary conditions as G since the boundary conditions do
not involve any time derivatives.
Consider equation 6.9 rewritten as
"
2
L0 + 2 G(x, x0 ; ) = (x x0 )( ).
t
#
The Fourier transform of this equation is

x0 ; ) + (x)
L0 G(x,
d ei
2
G(x, x0 ; ) = (x x0 ).
2
(6.11)
Ask Baker
but how do we
know
R
i d G( )
converges?
8 Feb p5
90
Using the product rule for differentiation, we can pull out a divergence
term:
i
2
G=
2
2 i
e
G+
2
G G ei .
Thus our equation 6.11 becomes

x0 ; ) + (x)( 2 G(x,
x0 ; ))
(x x0 ) = L0 G(x,
"
i
+ (x) e
8 Feb p6
G G ei
t
t
.
t=
Now we evaluate the surface term. Note that G = 0 for < 0 implies
G/t = 0, and thus | = 0. Similarly, as , ei 0 since
Im > 0, and thus | = 0. So we can drop the boundary term.
We thus find that GR satisfies the differential equation
x0 ; ) = (x x0 )
[L0 2 (x)]G(x,
eq6stst
#t=+
RBC,
(6.12)
with Im > 0. We now recognize that the Green function must be the
same as in the steady state case:
x0 ; ) = G(x, x0 ; = 2 ).
G(x,
this is unclear
Recall that from our study of the steady state problem we know that
the function G(x, x0 ; ) is analytic in the cut -plane. Thus by analytic
x0 ; ) is analytic in the whole cut plane.
continuation we knowthat G(x,
The convention = compresses the region of interest to the upper
half plane, where satisfies
[L0 (x)]G(x, x0 ; ) = (x x0 )
eq6ststb
+ RBC,
(6.13)
All that is left is to invert the Fourier transform.
6.4
Inverting the Fourier Transform
In the previous section we showed that for the Greens function G we

have the Fourier Transform
x0 ; ) =
G(x,
d ei G(x, x0 ; )
6.4. INVERTING THE FOURIER TRANSFORM
91
where = t t0 , and we also found

x0 ; ) = G(x, x0 ; = 2 )
G(x,
where = R + iI with I > 0, and G(x, x0 ; = 2 ) is the solution
of the steady state problem. Now we only need to invert the Fourier but
didnt
we use analytic
Transform to get the retarded Greens function. We write
continuation to
ge the whole
plane.
iR I
=e
so that
x0 ; R + iI ) =
G(x,
|
{z
F (R )
d eiR [eI GR (x, x0 ; )] .

|
{z
F ( )
This is a real Fourier Transform in terms of F ( ). We now apply the

Fourier Inversion Theorem:
pr:FIT1
I
F ( ) = e
1 Z
x0 ; R + iI )
GR (x, x ; ) =
dR eiR G(x,
2
0
so
1 Z
x0 ; )eI
dR eiR G(x,
2
fix I = and integrate over R
1 Z
x0 ; )
=
dei G(x,
2 L
GR (x, x0 ; ) =
(6.14)
(6.15)
(6.16)
where the contour L is a line in the upper half plane parallel to the R 10 Feb p2
axis, as shown in figure 6.1.a. The contour is off the real axis because of fig6Lcont
the branch cut. We note that any line in the upper half plane parallel
to the real axis may be used as the contour of integration. This can
be seen by considering the rectangular integral shown in figure 6.1.b.
= 0 as R , we know that the sides LS1 and LS2 vanish.
Because G
are analytic in the upper half plane, Cauchys
And since ei and G
theorem tells use that the integral over the closed contour is zero. Thus
the integrals over path L1 and path L2 must be equal.
92

I
LS1 6
L1 R
(a) The line L1

-
L2
?
LS2
L1
(b) Closed contour with L1
Figure 6.1: The contour L in the -plane.
6.4.1
Summary of the General IVP

pr:IVP1
We have considered the problem of a string hit with a blow of unit

momentum. This situation was described by the equation
"
eq6GFxt
2
L0 + 2 GR (x, t; x0 , t0 ) = (x x0 )(t t0 ) + RBC
t
#
with the condition that GR (x, t; x0 , t0 ) = 0 for t < t0 . The Greens

function which satisfies this equation was found to be
GR (x, t; x0 t0 ) =
eq6GFT
(6.17)
Z
L
d i(tt0 )
e
G(x, x0 ; ).
2
(6.18)
x0 ; ) satisfies the steady state Greens function problem.

where G(x,
6.5
Analyticity and Causality
To satisfy the physical constraints on the problem, we need to have

GR = 0 for t < t0 . This condition is referred to as causality. This con0
x0 ; 2 )
dition is obtained due to the fact that the product ei(tt ) G(x,
Need to show appearing in equation 6.17 is analytic. In this way we see that the anasomewhere
lyticity of the solution allows it to satisfy the causality condition. As a
||
that G 0. check, for the case t < t0 we write ei(tt0 ) = eiR (tt0 ) eI (tt0 ) and close
0
x0 ; 2 )
fig6Luhp
the contour as shown in figure 6.2. The quantity ei(tt ) G(x,
6.6. THE INFINITE STRING PROBLEM
93
0 -plane

PP
P
Q

Q
k
Q
Q LU HP
J
J
J
B
L
B
B R
-

Figure 6.2: Contour LC1 = L + LU HP closed in UH -plane.

0
vanishes on the contour LC1 = L + LU HP since eiI (tt ) 0 as

0
x0 ; 2 )| 0
I and |eiR (tt ) | = 1, while we required |G(x,
as || .
6.6
The Infinite String Problem

pr:ISP1
We now consider an infinite string where we take and to be a 10 Feb p3

constant, and V = 0. Thus our linear operator (cf 1.10) is given by
d2
L0 = 2 .
dx
6.6.1
Derivation of Greens Function
We want to solve the equation

"
2
2
2 + 2 GR (x, t; x0 , t0 ) = (xx0 )(tt0 )
x
t
#
for < x, x0 <

(6.19)
with the initial condition

GR (x, t; x0 t0 ) = 0
for t < t0 .
94
We write the Fourier transform of the Greens function in terms of :

x0 ; ) = G(x, x0 ; = 2 ).
G(x,
From 6.13, we know that G(x, x0 ; ) satisfies
"
d2
2 G(x, x0 ; ) = (x x0 )
dx
#
for < x, x0 < .
Actually since For the case of largewe found the solution (4.159)
V =0
i c 2
0
G = ei /c |xx |
2
or ( 2 )
x0 ; ) = i c ei|xx0 |/c .
G(x,
2
This gives us the retarded Greens function
1 Z
i c i |xx0 |
0
GR (x, t; x , t ) =
dei(tt )
ec
.
2 L
2
0
eq6GRInfStr
(6.20)
Now consider the term

h
i (tt0 )
|xx0 |
c
We treat this term in two cases:

t t0 <
|xx0 |
.
c
In this case
h
i (tt0 )
|xx0 |
c
as I .
But the term (i/2) + as R 0 in equation 6.20. Thus

the integral vanishes along the contour LU HP shown in figure 6.2
so (using Cauchys theorem) equation 6.20 becomes
GR (x, t; x0 , t0 ) =
I
L
I
L+LU HP
= 0.
6.6. THE INFINITE STRING PROBLEM
95
0 -plane

LLHP
B
B
J
J
J
Q

+

QQ
P

PP
Figure 6.3: Contour closed in the lower half -plane.

10 Feb p4
t t0 >
|xx0 |
.
c
In this case
h
i (tt0 )
e
fig6Llhp
|xx0 |
c
as I
so we close the contour below as shown in figureR 6.3. Since

the
R
integral vanishes along LLHP , we have GR = L = L+LLHP .
Cauchys theorem says that the integral around the closed contour is 2i times the sum of the residues of the enclosed poles.
1 ic
The only pole is at = 0 and its residue is 2
. For this case
2
we obtain

1 ic
c
0 0
GR (x, t; x , t ) = 2i
=
,
2 2
2
which is constant.
From these two cases we conclude that
c
|x x0 |
GR (x, t; x , t ) =
t t0
.
2
c
!
The function is defined by the equation

(
(u) =
0
1
for u < 0
for u > 0.
(6.21)
96

GR (x, t; x0 , t0 )
c
2
G=0
0
x c(t t )
G=0
x + c(t t0 )
0
Figure 6.4: An illustration of the retarded Greens Function.

fig6retGF
The situation is illustrated in figure 6.4. This solution displays some

interesting physical properties.
The function is zero for x < x0 c(t t0 ) and for x > x0 + c(t t0 ),
so it represents an expanding pulse.
The amplitude of the string is c/2 , which makes sense since for a
smaller string tension we expect a larger transverse amplitude.
is this right?
The traveling pulse does not damp out since V = 0.
6.6.2
Physical Derivation
10 Feb p5
12 Feb p1
We now explain how to get the solution from purely physical grounds.
Consider an impulse p applied at position x0 and time t0 . Applying
symmetry, at the first instant py = 1/2 for movement to the left and
py = 1/2 for movement to the right. We may also write the velocity
vy = py /m = 21 /dx since py = 1/2 and m = dx. By substituting dx = cdt, we find that in the time dt a velocity vy = 1/2cdt
is imparted to the string. This vy is the velocity of the string portion
at dx. By conservation of momentum, the previous string portion must
now be stationary. In time dt the disturbance moves in the y direction
1
an amount y = vy dt = 2c
= 2c . In these equalities we have used
2
the identity 1/c = / . Thus momentum is continually transferred
from point to point (which satisfies the condition of conservation of
momentum).
6.7. SEMI-INFINITE STRING WITH FIXED END
6.7
97
Semi-Infinite String with Fixed End
We now consider the problem of an infinite string with one end fixed.
We will get the same form of Greens function. The defining equation
is (c.f. 6.19)
"
2
2
2 + 2 GR (x, t; x0 , t0 ) = (x x0 )(t t0 )
(6.22)
x
t
for < t, t0 < ; 0 < x, x0 <
#
with the further condition
eq6qdblst
GR (x, t; x0 , t0 ) = 0
for x = 0.
(6.23)
This is called the Dirichlet boundary condition. We could use transform eq6qdblstbc
methods to solve this problem, but it is easier to use the method of
images and the solution 6.21 to the infinite string problem.
To solve this problem we consider an infinite string with sources at 12 Feb p2
0
x and x0 . This gives us a combined force
f (x, t) = [(x x0 ) (x + x0 )](t t0 )
and the principle of superposition allows us to write the solution of the
problem as the sum of the solutions for the forces separately:
c
|x x0 |
|x + x0 |
GR (x, t; x0 , t0 ) =
t t0
t t0
2
c
c
(6.24)
where u is the solution (c.f. 6.21) of the infinite string with sources at eq6qsst
x and x0 . This solution is shown in figure 6.5. Since u satisfies 6.22 fig6LandG
and 6.23, we can identify GR = u for x 0. The case of a finite string
leads to an infinite number of images to solve (c.f., section 8.7).
"
6.8
!#
Semi-Infinite String with Free End
We now consider a new problem, that of a string with two free ends.
The free end Greens function is
c
|x x0 |
|x + x0 |
0
0
GR =
tt
+ tt
2
c
c
"
!#
98

GR
GR
x
x0

x0
(a) GR at time t
(a) GR at time t
Figure 6.5: GR at t1 = t0 + 12 x0 /c and at t2 = t0 + 32 x0 /c.

This satisfies the equation
"
eq6qdblst2
2
2
(6.25)
2 + 2 GR (x, t; x0 , t0 ) = (x x0 )(t t0 )
x
t
for < t, t0 < ; a < x, x0 < b
#
with the boundary condition

d
GR (x, t; x0 , t0 )
=0

dx
x=0
12 Feb p3
which corresponds to a = 0 and ha = 0 is equation 6.3 (c.f., section

1.3.3).
The derivative of GR (x = 0) is always zero. Note that
Z
d
(x) = (b) (a) =
dx
1 for a < 0 < b

0 otherwise.
(6.26)
which implies
ask Baker
d
(x) = (x).
dx
But for any fixed t t0 we can chose an such that the interval [0, ] is
is flat. Therefore
d free end
G
= 0.
dx R
6.9. ELASTICALLY BOUND SEMI-INFINITE STRING
99
Notes about the physics: For a string with a free end, the force on the
end point is Fy = dG
= 0 at x = 0 which implies dG
= 0 at x = 0
dx
dx
if the tension does not vanish. If the tension does vanish at x = 0,
then we have a singular point at the origin and do not restrict dG
=0
dx
at x = 0.
6.9
Elastically Bound Semi-Infinite String
We now consider the problem with boundary condition

"
d
+ GR = 0
dx
for x = 0.
The solution can be found using the standard transform method. Do

an inverse Fourier transform of the Greens function in eq. 6.13 for the
= 0. The frequency space part of this
related problem [d/dx + ]G
problem is done in problem 4.3.
6.10
Relation to the Eigen Fn Problem
We now look at the relation between the general problem and the eigen
function problem (normal modes and natural frequencies). The normal
mode problem is used in solving
[L0 ]G(x, x0 ; ) = (x x0 )
+ RBC
for which G(x, x0 ; ) has poles at the eigen values of L0 . We found in

chapter 4 that G(x, x0 ; ) can be written as a bilinear summation
G(x, x0 ; ) =
X un (x)un (x0 )
n
(6.27)
where the un (x) solve the normal mode problem:

[L0 n ]un (x) = 0
+ RBC.
eq6qA
(6.28)
Here we made the identification n = n2 where the n s are the natural eq6qE
100
frequencies and the un s are the normal modes.

Recall also that the steady state solution for the force (x x0 )eit
is u(x, t) = G(x, x0 ; = 2 + i)eit . The non-steady state response is
GR (x, t; x0 , t0 ) which is given by
GR (x, t; x0 , t0 ) =
12 Feb p4
d i(tt0 )
e
G(x, x0 ; )
2
x0 ; ) = G(x, x0 ; = 2 ). Plug G(x, x0 ; ) (from equation

where G(x,
6.27) into the Fourier transformed expression (equation 6.18). This
gives
X un (x)un (x0 )
1 Z
i(tt0 )
de
GR (x, t; x , t ) =
2 L
n 2
n
0
(6.29)
In this equation can be arbitrarily complex. (This equation is very

different (c.f. section 4.6) from the steady state problem G(x, x0 ; =
2 + i)eit where was real.) Note that we are only interested in
t > t0 , since we have shown already that GR = 0 for t < t0 .
0
This is backNow we add the lower contour since ei(tt ) is small for t0 > t and
wards
I < 0. This contour is shown in figure 6.3. The integral vanishes over
the curved path, so we can useCauchys theorem to solve 6.29. The
poles are at n = 2 , or = n .
We now perform an evaluation of the integral for one of the terms
12 Feb p5
of the summation in equation 6.29.
eq6qC
Z
L+LLHP
Z
d ei(tt )
d
ei(tt )
(6.30)
=
2 n 2
L+LLHP 2 (
n )( + n )
2i ei n (tt ) ei n (tt )
=
2
2 n
2 n
sin n (t t0 )
=
.
n
"
eq6qFa
By equation 6.29 we get

0
GR (x, t; x , t ) =
X
n
eq6qF
sin
un (x)un (x0 )
(t t0 )
n
n
(6.31)
where n = n and t > t0 . This general solution gives the relationship

between the retarded Greens function problem (equation 6.17) and the
eigen function problem (eq. 6.28).
6.10. RELATION TO THE EIGEN FN PROBLEM
6.10.1
101
Alternative form of the GR Problem
For t t0 small, eq. 6.31 becomes

GR (x, t; x0 , t0 )
X un (x)un (x0 ) q
= (t t0 )
n (t t0 )
un (x)un (x0 )
(x x0 )
= (t t0 )
.
(x)
Where we used the completeness relation 4.113. Thus for t t0 small, 12 Feb 6
the GR has the form
GR (x, t; x0 , t0 )|tt0 (t t0 )
(x x0 )
.
(x)
Differentiating, this equation gives

(x x0 )

GR (x, t; x0 , t0 )
=
.
t
(x)
tt0 +

Also, as t approaches t0 from the right hand side

GR (x, t; x0 , t0 ) = 0
for t t .
These results allow us to formulate an alternative statement of the GR

problem in terms of an initial value problem. The GR is specified by
the following three equations:
"
2
L0 + (x) 2 GR (x, t; x0 , t0 ) = 0
t
#
GR (x, t; x0 , t0 ) = 0
for t > t0 + RBC

for t = t0
GR = (x x0 )
for t = t0
t
Here (t) t
GR (x, t; x0 , t0 ) represents a localized unit of impulse at x0 , t0
(like p = 1). Thus we have the solution to the initial value problem
for which the string is at rest and given a unit of momentum at t0 .
(x)
102
We have now cast the statement of the GR problem in two forms,

as a retarded boundary value problem (RBVP) and as an initial value
problem (IVP):
( h
RBVP =
0 0
0
0
L0 + t
2 GR (x, t; x , t ) = (x x )(t t )
GR (x, t; x0 , t0 ) = 0
for t < t0
h
i
2
0 0
L0 + (x) t2 GR (x, t; x , t ) = 0
+ RBC,
for t > t0 , RBC,
IVP = GR (x, t; x0 , t0 ) = 0
for t = t0 ,
(x) G = (x x0 )
for t = t0 .
t R
6.11
Comments on Greens Function
6.11.1
Continuous Spectra
17 Feb p1
In the previous section we obtained the spectral expansion for discrete

eigenvalues:
GR (x, t; x0 , t0 ) =
X
un (x)un (x0 )
n=1
eq6rst
sin
n (t t0 )
(6.32)
This gives us an expansion of the Greens function in terms of the

natural frequencies.
For continuous spectra the sum is replaced by an integral
Z
X
(t t0 )
sin
n
GR = dn
un un
n
where we have included a sum over degeneracy index (c.f. 4.108).

Note that this result follows directly because the derivation in the previous section did not refer to whether we had a discrete or continuous
spectrum.
6.11.2
Neumann BC
Recall that the RBC for an open string, equation 1.18, is

"
d
+ a GR = 0
dx
for x = a.
6.11. COMMENTS ON GREENS FUNCTION
103
If a 0 (Neumann boundary condition) then the boundary condition

for the normal mode problem will be (d/dx)u(x) = 0 which will have
a constant solution, i.e., 1 = 0. We cannot substitute 1 = 0 into Ask why
equation 6.32, but instead must take the limit as 1 approaches zero.
Physically, this corresponds to taking the elasticity a as a small quantity, and then letting it go to zero. In this case we can write equation
6.32 with the 1 eigen value separated out:
0
1
GR (x, t; x0 , t0 )
u1 (x)u1 (x0 )(t t0 )
q
X un (x)un (x0 )
+
sin n (t t0 )
n
n
t
0
u1 (x)u1 (x )(t t0 ).
(6.33)
The last limit is true because the sum oscillates in t. The Greens
function represents the response to a unit momentum, but a = 0
which means there is no restoring force. Thus a change in momentum
p = 1 is completely imparted to the string, which causes the string
to acquire a constant velocity, so its amplitude increases linearly with
time.
Note that equation 6.33 would still be valid if we had taken 1 = 0
in our derivation of the Greens function as a bilinear sum. In this
case equation 6.30 would have a double pole for 1 = 0, so the residue
would involve the derivative of the numerator, which would give the
linear factor of t t0 .
Consider a string subject to an arbitrary force (x)f (x, t). Remember 17 Feb p2
that (x)f (x, t) = (t t0 )(x x0 ) gives GR (x, t; x0 , t0 ). A general force
(x)f (x, t) gives a response u(x, t) which is a superposition of Greens pr:GenResp1
functions:
u(x, t) =
dt0
dx0 G(x, t; x0 , t0 )(x0 )f (x0 , t0 )
with no boundary terms (u(x, 0) = 0 =

Greens function expansion 6.33 to get
u(x, t) = u1 (x)
+
dt (t t )
n=2
X
un (x) Z
dt0 sin
d
u(x, 0)).
dt
Now plug in the
dx0 u1 (x0 )(x0 )f (x0 , t0 )

q
n (t t0 )
dx0 un (x0 )(x0 )f (x0 , t0 ).
104
Note that again the summation terms oscillate with frequency n . The
spatial dependence is given by the un (x). The coefficients give the
projection of (x)f (x, t) onto un (x). In the 1 = 0 case the u1 (x) term
is constant.
6.11.3
Zero Net Force
Now let
F (t) (const.)
ask baker
about last part
not included
here
dt0 (x0 )f (x0 , t0 )
where F (t0 ) represents the total applied force at time t0 . If F (t0 ) = 0,

then there are no terms which are linearly increasing in time contributing to the response u(x, t). This is a meaningful situation, corresponding to a disturbance which sums to zero. The response is purely oscillatory; there is no growth or decay.
6.12
Summary
1. The retarded Greens function GR solves

"
2
L0 + 2 GR (x, t; x0 , t0 ) = (x x0 )(t t0 )
+ RBC
t
for a < x, x0 < b; all t, t0 .
#
with the condition

GR (x, t; x0 , t0 ) = 0
for t < t0 .
The advanced Greens function GA solves the same equation, but

with the condition
GA (x, t; x0 , t0 ) = 0
for t > t0 .
The retarded Greens function gives the response of the string

(initially at rest) to a unit of momentum applied to the string at
a point in time t0 at a point x0 along the string. The advanced
Greens function gives the initial motion of the string such that
a unit of momentum applied at x0 , t0 causes it to come to rest.
6.13. REFERENCES
105
2. The retarded Greens function can be written in terms of the

steady state Greens function:
GR (x, t; x0 t0 ) =
Z
L
d i(tt0 )
e
G(x, x0 ; ).
2
3. The retarded Greens function for an infinite string with and

constant, and V = 0 is
c
|x x0 |
t t0
.
GR (x, t; x , t ) =
2
c
!
4. The retarded Greens function for a semi-infinite string with a

fixed end, and constant, and V = 0 is
c
|x x0 |
|x + x0 |
t t0
t t0
GR (x, t; x , t ) =
2
c
c
"
!#
5. The retarded Greens function for a semi-infinite string with a

free end, and constant, and V = 0 is
c
|x x0 |
|x + x0 |
GR =
t t0
+ t t0
2
c
c
"
!#
6. The retarded Greens function can be written in terms of the eigen

functions as
X
n (t t0 )
0 0
0 sin
.
GR (x, t; x , t ) =
un (x)un (x )
n
n
6.13
References
A good reference is [Stakgold67b, p246ff].

This material is developed in three dimensions in [Fetter80, p311ff].
106
Chapter 7
Surface Waves and
Membranes
Chapter Goals:
Show how the equation describing shallow water
surface waves is related to our most general differential equation.
Derive the equation of motion for a 2-dimensional
membrane and state the corresponding regular
boundary conditions.
7.1
Introduction
In this chapter we formulate physical problems which correspond to

equations involving more than one dimension. This serves to motivate the mathematical study of N -dimensional equations in the next
chapter.
107
108
CHAPTER 7. SURFACE WAVES AND MEMBRANES

h(x)

x
b(x)
Figure 7.1: Water waves moving in channels.
7.2
7.2.1
pr:surf1
fig:7.1
pr:hww1
One Dimensional Surface Waves on

Fluids
The Physical Situation
Consider the physical situation of a surface wave moving in a channel1 .

This situation is represented in figure 7.1. The height of equilibrium
is h(x) and the width of the channel is b(x). The height of the wave
z(x, t) can then be written as
z(x, t) = h(x) + u(x, t)
where u(x, t) is the deviation from equilibrium. We now assume the
shallow wave case u(x, t) h(x). This will allow us to linearize the
NavierStokes equation.
pr:shal1
pr:lambda2
7.2.2
Shallow Water Case
This is the case in which the height satisfies the condition h(x)
where is the wavelength. In this case the motion of the water
is approximately horizontal. Let S(x) = h(x)b(x). The equation of
continuity and Newtons law (i.e., the NavierStokes equation) then
give
!
gS(x)
u + b(x) 2 u(x, t) = 0,
x
x
t
1
This material corresponds to FW p. 357363.
17 Feb p3
7.3. TWO DIMENSIONAL PROBLEMS
109
which is equivalent to the 1-dimensional string, where (x) b(x) and 17 Feb p4
(x) gS(x).
Consider the case in which b(x) is independent of x:
gh(x)
u + 2 u(x, t) = 0.
x
x
t
!
(7.1)
This corresponds to = 1 and = gh(x).

eq7shallow
Propagation of shallow water waves looks identical to waves on a
string. For example, in problem 3.5, h(x) = x gives the Bessels equation, with the identification (x) = x and V (x) = m2 /x.
As another example,
q take h(x) to be constant. This gives us wave
propagation with c = /, = gh, and = 1. So the velocity of a
water wave is c = gh. The deeper the channel, the faster the velocity.
This partially explains wave breaking: The crest sees more depth than
the trough.
7.3
Two Dimensional Problems

17 Feb p5
2
We now look at the 2-dimensional problem, that of an elastic membrane .

We denote the region of the membrane by R and the perimeter (1- pr:elmem1
dimensional surface) by S. The potential energy differential for an
element of a 1-dimensional string is
1
du
dU = (x)
2
dx
!2
dx.
In the case of a 2-dimensional membrane we replace u(x) with u(x, y) =

u(x). In this case the potential energy difference is (see section 2.4.2)
1
du
dU =
(x, y)
2
dx
=
2
1
(x)(u)2 dx
2
This is discussed on p. 271288 of FW.
!2
du
+
dy
!2
dxdy
(7.2)
(7.3)
110
where is the tension of the membrane. Note that there is no mixed

d d
term dx
u since the medium is homogeneous. The total potential
dy
energy is given by the equation
U=
1
dx (x)(u)2 ,
2
R
where (x) is the surface tension. In this equation dx = dx dy and

(u)2 = (u) (u). The total kinetic energy is
Z
1
1
u
T = mv 2 = dx (x)
2
2
t
R
!2
Now think of the membrane as inserted in an elastic media. We then

get an addition to the U (x) energy due to elasticity, 12 V (x)u(x, t)2 . We
also add an additional force which will add to the potential energy:
f (x, t)(x)dxu(x, t) =
pr:lagr1
force
mass
The Lagrangian is thus

L=
x
u
t
19 Feb p1
19 Feb p2
mass
(length) (displacement) .
length
1Z
dx
2 R
!2
(x)(u)2 V (x)u(x)2 f (x, t)(x)dxu(x, t) .
Notice the resemblance of this Lagrangian to the one for a one dimensional string (see section 2.4.2).
We apply Hamiltonian Dynamics to get the equation of motion:
"
2
L0 + 2 u(x, t) = (x)f (x, t)
t
#
where L0 = ( (x)) V (x). This is identical to the equation of

motion for a string except now in two dimensions. It is valid everywhere
for x inside the region R.
7.3. TWO DIMENSIONAL PROBLEMS
111
(0, b)
(a, b)
(0, y) r
r(a, y)
(0, 0)
(a, 0)
Figure 7.2: The rectangular membrane.
7.3.1
Boundary Conditions
pr:bc2
Elastically Bound Surface

The most general statement of the boundary condition for an elastically
bound surface is
[
n + (x)]u(x, t) = h(x, t)
for x on S.
(7.4)
In this equation the surface S is the perimeter of the membrane, n

is sone
the outward normal for a point on the perimeter, (x) = k(x)/ (x) is
the effective spring constant at a point on the boundary, and h(x, t) =
f (x, t)/ (x) is an external force acting on the boundary S.
Periodic Boundary Conditions
19 Feb p3
We now consider the case of a rectangular membrane, illustrated in pr:pbc2

figure 7.2, with periodic boundary conditions:
fig:7.2
and
u(0, y) = u(a, y)
for 0 y b,
u(x, 0) = u(x, b)
for 0 x a,

2
2

u(x,
y)

=
u(x,
y)

2
2

x
x
x=0
x=a
for 0 y b,
(7.5)
112

2
2

u(x,
y)

=
u(x,
y)

2

x2
x
y=0
y=b
for 0 x a.
In this case we can consider the region R to be a torus.
7.4
19 Feb p4
Example: 2D Surface Waves
We now give one last 2-dimensional example3 . We consider a tank of

water whose bottom has arbitrary height h(x) and look at the surface
waves. This example connects the 1-dimensional surface wave problem
and the 2-dimensional membrane problem.
For this problem the vertical displacement is given by
z(x, t) = h(x) + u(x, t),
with h(x) for the shallow water case and u h(x). Thus (using
7.1) our equation of motion is
"
2
(gh(x)) + 2 u(x, t) = f (x, t).
t
#
Note that in this equation we have = 1. For this problem we take

the Neumann natural boundary condition:
n
u(x, t) = 0
and
19 Feb p5
u = g u|S
t
for x on S. This is the case of rigid walls. The latter equation just
means that there is no perpendicular velocity at the surface.
The case of membranes for a small displacement is the same as for
surface waves. We took = 1 and (x) = gh(x).
The formula for all these problems is just
"
2
L0 + (x) 2 u(x, t) = (x)f (x, t)
t
#
for x in R
where L0 = ( (x)) + V (x). (Note that (x) is not necessarily

tension.) We let x = (x1 , . . . , xn ) and = (/x1 , . . . , /xn ). The
boundary conditions can be elastic or periodic.
3
This one comes from FWp. 363366.
7.5. SUMMARY
7.5
113
Summary
1. The equation for shallow water surface waves and the equation
for string motion are the same if we identify gravity times the
cross-sectional area with tension, and the width of the channel
with mass density.
2. The two-dimensional membrane problem is characterized by the
wave equation
"
2
L0 + 2 u(x, t) = (x)f (x, t)
t
#
where L0 = ( (x)) V (x), subject to either an elastic

boundary condition,
[
n + (x)]u(x, t) = h(x, t)
for x on S,
or a periodic boundary condition.
7.6
References
The material on surface waves is covered in greater depth in [Fetter80,

p357ff], while the material on membranes can be found in [Fetter80,
p271].
114
Chapter 8
Extension to N -dimensions
Chapter Goals:
Describe the different sorts of boundaries and
boundary conditions which can occur for the N dimensional problem.
Derive the Greens identities for the N -dimensional
case.
22 Feb p1
Write the solution for the N -dimensional problem

in terms of the Greens function.
Describe the method of images.
8.1
Introduction
In the previous chapter we obtained the general equation in two dimensions:

"
2
L0 + (x) 2 u(x, t) = (x)f (x, t)
t
#
for x in R
(8.1)
where
eq8usf
L0 = (x) + V (x).
This is immediately generalizable to N -dimensions. We simply let
x = (x1 , . . . , xn ) and = (/x1 , . . . , /xn ). and we introduce the
115
116
pr:bc3
CHAPTER 8. EXTENSION TO N -DIMENSIONS
notation n
u u/n. R is now a region in N -dimensional space
and S is the (N 1)-dimensional surface of R.
The boundary conditions can either be elastic or periodic:
1. Elastic: The equation for this boundary condition is
[
n + K(x)]u(x, t) = h(x, t)
(8.2)
The term K(x) is like a spring constant which determines the

properties of the medium on the surface, and h(x, t) is an external force on the boundary. These terms determine the outward
gradient of u(x, t).
eq8.1b
pr:pbc3
for x on S.
2. Periodic: For the two dimensional case the region looks like 7.2
and the periodic boundary conditions are 7.5 and following. In
the N -dimensional case the region R is an N -cube. Connecting
matching periodic boundaries of S yields an N -torus in (N + 1)dimensional space.
To uniquely specify the time dependence of u(x, t) we must specify the
initial conditions
u(x, t)|t=0 = u0 (x)
for x in R
(8.3)
(x, t)|t=0 = u1 (x)

for x in R.
(8.4)
t
For the 1-dimensional case we solved this problem using Greens Identities. In section 8.3 we will derive the Greens Identities for N -dimensions.
22 Feb p2
8.2
Regions of Interest
There are three types of regions of interest: the Interior problem, the
Exterior problem, and the All-space problem.
pr:intprob1
1. Interior problem: Here R is enclosed in a finite region bounded by

S. In this case we expect a discrete spectrum of eigenvalues, like
one would expect for a quantum mechanical bound state problem
or for pressure modes in a cavity.
8.3. EXAMPLES OF N -DIMENSIONAL PROBLEMS

pr:extprob1
117
2. Exterior problem: Here R extend to infinity in all directions but is

excluded by a finite region bounded by S. In this case we expect
a continuum spectrum if V > 0 and a mixed spectrum is V < 0.
This is similar to what one would expect for quantum mechanical
scattering.
3. All-space problem: Here R extends to infinity in all directions pr:allprob1
and is not excluded from any region. This can be considered a
degenerate case of the Exterior problem.
8.3
8.3.1
Examples of N -dimensional Problems

General Response
pr:GenResp2
In the following sections we will show that the N -dimensional general

response problem can be solved using the Greens function solution to
the steady state problem. The steps are identical to those for the single
dimension case covered in chapter 6.
x0 ; )
G(x, x0 ; = 2 ) = G(x,
GR (x, t; x0 , t0 )
retarded
u(x, t)
General Response.
8.3.2
Normal Mode Problem
The normal mode problem is given by the homogeneous differential

equation
!
2
L0 + (x) 2 u(x, t) = 0.
t
Look for solutions of the form
u(x, t) = ein t un (x).
The natural frequencies are n = n , where n is an eigen value. The

normal modes are eigen functions of L0 . Note: we need RBC to ensure
that L0 is Hermitian.
pr:NormMode3
118
24 Feb p2
pr:fhop2
8.3.3
Forced Oscillation Problem
The basic problem of steady state oscillation is given by the equation

2
L0 + 2 u(x, t) = eit (x)f (x)
t
!
for x R
and (for example) the elastic boundary condition

(
n + K)u = h(x)eit
for x S.

u(x, t) = eit u(x, ).
The value of is chosen, so this is not an eigen value problem. We
assert
u(x, ) =
Z
x0 R
dx0 G(x, x0 ; = + i)(x0 )f (x0 )
Z
x0 S
dx0 (x0 )G(x, x0 ; = + i)(x)h(x0 ).
The first term gives the contribution due to forces on the volume and
the second term gives the contribution due to forces on the surface.
In the special case that (x)f (x) = (x x0 ), we have
u(x, t) = eit G(x, x0 ; = 2 ).
8.4
Greens Identities
In this section we will derive Greens 1st and 2nd identities for the
N -dimensional case. We will use the general linear operator for N dimensions
L0 = ( (x)) + V (x)
and the inner product for N -dimensions
hS, L0 ui =
eq8tst
dxS (x)L0 u(x).
(8.5)
8.5. THE RETARDED PROBLEM
8.4.1
119
Greens First Identity
pr:G1Id2
The derivation here generalizes the derivation given in section 2.1.

hS, L0 ui =
dxS (x)[ ( (x)) + V (x)]u(x)
(8.6)
integrate 1st term by parts

=
dx[ (S (x)u) + (S ) (x)u + S V u]
integrate 1st term using Gauss Theorem

=
dS n
(S (x)u) +
[S V u + ( S (x)u)]dx
This is Greens First Identity generalized to N -dimensions:

hS, L0 ui =
dS n
(S (x)u)+
[S V u+(S (x)u)]dx. (8.7)
Compare this with 2.3.
8.4.2
Greens Second Identity

pr:G2Id2
We now interchange S and u. In the quantity hS, L0 ui hL0 S, ui the 22 Feb p3

symmetric terms will drop out, i.e., the second integral in 8.7 is cancelled. We are left with
hS, L0 ui hL0 S, ui =
8.4.3
dS n
[S (x)u + u (x)S ].
Criterion for Hermitian L0
If u, S satisfy the RBC, then the surface integral in Greens second

identity vanishes. This leaves hS, L0 ui hL0 S, ui, which means that L0
is a hermitian (or self-adjoint) operator: L = L .
8.5
8.5.1
pr:HermOp2
The Retarded Problem

General Solution of Retarded Problem
We now reduce 8.1 to a simpler problem. If this is an initial value

problem, then u(x, t) is completely determined by equations 8.1, 8.2, pr:IVP2
120
eq8tB
22 Feb p4
8.3, and 8.4. We look again at GR which is the response of a system to

a unit force:
"
#
2
L0 + (x) 2 GR (x, t; x0 , t0 ) = (x x0 )(t t0 )
for x, x0 in R.
t
(8.8)
We also require the retarded Greens function to satisfy RBC and the
initial condition GR = 0 for t < t0 . We now use the result from problem
4.2:
u(x, t) =
Z
x0 R
+
+
dx0
Z
x0 S
Z
x0 R
Z
0
dt0 GR (x, t; x0 , t0 )(x0 )f (x0 , t0 )
dx (x )
Z
0
dt0 GR (x, t; x0 , t0 )h(x0 , t0 )
dx (x ) GR (x, t; x0 , 0)u1 (x0 )
0 GR (x, t; x0 , t0 )u0 (x0 )

(8.9)
t
eq8ugen
The first line gives the volume sources, the second gives the surface
R Horn says sources, and the third and fourth gives the contribution from the initial
eval G R at x conditions. Since the defining equations for GR are linear in volume,
surface, and initial condition terms, we were able to write down the
=0
solution u(x, t) as a linear superposition of the GR .
ask Baker
Note that we recover the initial value of u(x, t) in the limit t t0 .
about limits
This is true since in the above equation we can substitute

lim GR (x, t; x0 , t0 ) = 0
tt0
and
lim0 GR (x, t; x0 , t0 ) =
tt
pr:GR2
8.5.2
(x x0 )
.
(x0 )
The Retarded Greens Function in N -Dim.
By using 8.9 we need only solve 8.8 to solve 8.1. In section 6.3 we found
that GR could by determined by using a Fourier Transform. Here we
follow the same procedure generalized to N -dimensions. The Fourier
transform of GR in N -dimensions is
Z
d i(tt0 )
e
G(x, x0 ; )
GR (x, t; x0 , t0 ) =
2
L
8.5. THE RETARDED PROBLEM
121
where L is a line in the upper half plane parallel to the real axis (c.f., 22 Feb p5
is analytic in the upper half plane (which is due
section 6.4), since G
to the criterion GR = 0 for t < t0 ).
Now the problem is simply to evaluate the Fourier Transform. By
the same reasoning in section 6.3 the Fourier transform of GR is identical to the Greens function for the steady state case:
x0 ; ) = G(x, x; = 2 ).
G(x,
Recall that the Greens function for the steady state problem satisfies
[L0 ]G(x, x0 ; ) = (x x0 )
for x, x0 R, RBC
(8.10)
We have reduced the general problem in N dimensions (equation8.1)

to the steady state Greens function problem in N -dimensions.
22 Feb p6
8.5.3
Reduction to Eigenvalue Problem

pr:efp3
The eigenvalue problem (i.e., the homogeneous equation) in N -dimensions

is
L0 un (x) = n un (x)
x R, RBC
(8.11)
The n s are the eigen values of L0 . Since L0 is hermitian, the n s Lots of work
are real. The un (x)s are the corresponding eigenfunctions of L0 . We
can also prove orthonormality (using the same method as in the single
dimension case)
Z
R
dx
u
n (x)um (x) = 0 if n 6= m .
This follows from the hermiticity of L0 . Note that because we are now
in N -dimensions, the degeneracy may now be infinite.
22 Feb p7
By using the same procedure as in chapter 4, we can write a solution
of eq. 8.8 expanded in terms of a solution of 8.11:
G(x, x0 ; ) =
X un (x)un (x0 )
n
Note that the sum would become an integral for a continuous spectrum.
The methods of chapter 4 also allow us to construct the -function 22 Feb p8
122
representation
(x x0 ) = (x0 )
un (x)un (x0 ),
24 Feb p1
which is also called the completeness relation. All we have left is to

discuss the physical interpretation of G.
8.6
Region R
24 Feb p3
pr:intprob2
8.6.1
Interior
In the interior problem the Greens function can be written as a discrete

spectrum of eigenvalues. In the case of a discrete spectrum we have
G(x, x0 ; = 2 ) =
X un (x)un (x0 )
n 2
pr:extprob2
8.6.2
Exterior
For the exterior problem the sums become integrals and we have a
continuous spectrum:
0
G(x, x ; ) =
Z
n
ask Baker
about omitted
material
24 Feb p4
24 Feb p5
pr:MethIm1
dn
un (x)un (x0 )
.
n
In this case we take = 2 + i. G now has a branch cut for all real
s, which means that there will be two linearly independent solutions
which correspond to whether we approach the real axis from above or
below. We choose > 0 to correspond to the physical out going wave
solution.
8.7
The Method of Images
We now present an alternative method for solving N -dimensional problem which is sometimes useful when the problem exhibits sufficient
8.7. THE METHOD OF IMAGES
123
symmetry. It is called the Method of Images. For simplicity we consider a one dimensional problem. Consider the GR problem for periodic
boundary conditions with constant coefficients.
2
L0 2 GR = (x x0 )(t t0 )
t
!
8.7.1
0 x l.
Eigenfunction Method
We have previously solved this problem by using an eigen function

expansion solution (equation 6.31)
GR =
X un (x)un (x0 )
sin
n (t t0 ).
For this problem the eigen functions and eigen values are
1/2
un (x) =
1
l
n =
8.7.2
2in
l
e2inx/l
2
n = 0, 1, 2, . . .
n = 0, 1, 2, . . .
Method of Images
The method of images solution uses the uniqueness theorem. Put images over to in region of length .
c
|x x0 |
=
t t0
.
2
c

This is not periodic over 0 to l. Rather, it is over all space. Our GR is
c X
|x x0 nl|
GR =
t t0
.
2 n=
c
Notice that this solution satisfies
X
2
L0 + 2 G R =
(x x0 nl)
t
n=
< x, x0 < .
124
However, we only care about 0 < x < l
X
2
L0 + 2 G R =
(x x0 nl)
t
n=
0 < x, x0 < l.
Since the other sources are outside the region of interest they do not
affect this equation. Our Greens function is obviously periodic.
The relation between these solution forms is a Fourier series.
8.8
Summary
1. For the exterior problem, the region is outside the boundary and
extends to the boundary. The the interior problem, the boundary
is inside the boundary and has finite extent. For the all-space
problem, there is no boundary. The boundary conditions can be
either elastic or periodic, or in the case that there is no boundary,
the function must be regular at large and/or small values of its
parameter.
2. The Greens identities for the N -dimensional case are
hS, L0 ui =
dS n
(S (x)u)+
[S V u+(S (x)u)]dx,
hS, L0 ui hL0 S, ui =
dS n
[S (x)u + u (x)S ].
3. The solution for the N -dimensional problem in terms of the Greens

function is
u(x, t) =
x0 R
+
+
dx
Z
0
dx (x )
x0 S
Z
0
x0 R
dt0 GR (x, t; x0 , t0 )(x0 )f (x0 , t0 )
dt0 GR (x, t; x0 , t0 )h(x0 , t0 )
dx (x ) GR (x, t; x0 , 0)u1 (x0 )
GR (x, t; x0 , t0 )u0 (x0 ) .

t0

8.9. REFERENCES
125
4. The method of images is applicable if the original problem exhibits enough symmetry. The method is to replace the original
problem, which has a boundary limiting region of the solution,
with a new problem in which the boundary is taken away and
sources are placed in the region which was excluded by the boundary such that the solution will satisfy the boundary conditions of
the original problem.
8.9
References
The method of images is covered in most electromagnetism books, for

example [Jackson75, p54ff], [Griffiths81, p106ff]; a Greens function
application is given in [Fetter80, p317]. The other material in this
chapter is a generalization of the results from the previous chapters.
126
Chapter 9
Cylindrical Problems
Chapter Goals:
Define the coordinates for cylindrical symmetry
and obtain the appropriate -function.
Write down the Greens function equation for the
case of circular symmetry.
29 Feb p1
Use a partial expansion for the Greens function to

obtain the radial Greens function equation for the
case of cylindrical symmetry.
Find the Greens function for the case of a circular
wedge and for a circular membrane.
9.1
Introduction
In the previous chapter we considered the Greens function equation

(L0 (x))G(x, x0 ; ) = (x x0 )
for x, x0 R
where
L0 = ( (x)) + V (x)
subject to RBC, which are either for the elastic case
(
n + (S))G(x, x0 ; ) = 0
127
for x S
(9.1)
128
CHAPTER 9. CYLINDRICAL PROBLEMS
or the periodic case. In this chapter we want to systematically solve this

problem for 2-dimensional cases which exhibit cylindrical symmetry.
9.1.1
Coordinates
A point in space can be represented in cartesian coordinates as

x = ix + jy.
pr:CartCoord1 Instead of the coordinate pair (x, y) we may choose polar coordinates
(r, ). The transformation to cartesian coordinates is

x = r cos
pr:r1
y = r sin
while the transformation to polar coordinates is (for tan defined on

the interval /2 < < /2)
r=
x2 + y 2
tan (y/x)
for x > 0, y > 0

= tan1 (y/x) + for x < 0
tan1 (y/x) + 2 for x > 0, y < 0

pr:jak1
A differential of area for polar coordinates is related by that for cartesian coordinates by a Jacobian (see Boas, p220):
dxdy = dA
=
=
=
=

!

x, y

J
drd

r,

!
(x, y)

drd
(r, )

(x/r) (x/)

drd
(y/r) (y/)

cos r sin

drd
sin r cos
= rdrd
9.1. INTRODUCTION
129
By expanding dx and dy in terms of dr and d, we can write the

differential of arc length in polar coordinates (see Boas, p224)
ds =
dx2 + dy 2 =
dr2 + r2 d2 .
The differential operator becomes (see Boas, p252,431)
pr:grad1
u 1 u
+
,
r
r
1
1
B=
(rBr ) +
(B ).
r r
r
u = r
gradient
divergence
Let B = (x)x, then
1
u
1
u
r (x)
+
(x)
.
( (x)u(x)) =
r r
r
r
9.1.2
Delta Function
29 Feb p2
The N -dimensional -function is defined by the property

f (x) =
(9.2)
pr:DeltaFn2
d2 xf (x)(x x0 ).
In polar form we have (since dxdy = rdrd)

f (x) = f (r, ) =
r0 dr0 d0 f (r0 , 0 )(x x0 ).
By comparing this with

f (r, ) =
dr0 d0 f (r0 , 0 )(r r0 )( 0 )
we identify that the delta function can be written in polar coordinates

in the form
pr:delta1
0
(r
r
)
( 0 ).
(x x0 ) =
r
29
February
1988
130
#
S J
]J

3 r
n
a
"!
Figure 9.1: The region R as a circle with radius a.
9.2
GF Problem for Cylindrical Sym.
The analysis in the previous chapters may be carried into cylindrical

coordinates. For simplicity we consider cylindrical symmetry: (x) =
(r), (x) = (r), and V (x) = V (r). Thus 9.2 becomes
1
u
(r) 2 u
( (r)u(x)) =
r (r)
+ 2
.
r r
r
r 2
!
The equation for the Greens function

1
(L0 )G(r, ; r0 , 0 ) = (r r0 )( 0 )
r
becomes (for r, R)
(r) 2
1
2
+ V (r) (r) G = (rr0 )(0 ).
2
r
r
(9.3)
Here R may be the interior or the exterior of a circle. It could also
be a wedge of a circle, or an annulus, or anything else with circular
symmetry.
For definiteness, take the region R to be the interior of a circle of
radius a (see figure 9.1) and apply the elastic boundary condition 9.1.
We now define the elasticity on the boundary S, (S) = (). We
must further specify () = , a constant, since if = (), then we
might not have cylindrical symmetry. Cylindrical symmetry implies
n
= /r so that the boundary condition
"
fig10a
r (r)
r r
r
(
n + (S))G(r, ; r0 , 0 ) = 0
is now
for r = a
+ G(r, ; r0 , 0 ) = 0
r
for r = a.
9.3. EXPANSION IN TERMS OF EIGENFUNCTIONS
131
We also need to have G periodic under + 2. So
29 Feb p3
G(r, 0; r0 , 0 ) = G(r, 2; r0 , 0 )
and
G
G

=

.
=0
=2
We have now completely respecified the Greens function for the case
of cylindrical symmetry.
9.3
Expansion in Terms of Eigenfunctions
Since the Greens function is periodic in and since only appears in

the operator as 2 /2 , we use an eigenfunction expansion to separate
out the -dependence. Thus we look for a complete set of eigenfunctions pr:efexp1
which solve
2
2 um () = m um ()
(9.4)
for um () periodic. The solutions of this equation are

1
um () = eim
2
for m = 0, 1, 2, . . .
and the eigenvalues are

m = m 2
for m = 0, 1 2, . . ..
(Other types of regions would give different eigenvalues m ).

this set of eigenfunctions is complete it satisfies the expansion
( 0 ) =
Since pr:CompRel4
um ()um (0 ).
9.3.1
Partial Expansion
We now want to find Gm (r, r0 ; ) which satisfies the partial expansion

(using the principle of superposition)
pr:partExp1
G(r, ; r0 , 0 ) =
X
m
um ()Gm (r, r0 ; )um (0 ).
(9.5)
132
We plug this and 9.4 into the partial differential equation 9.3:
"
X
um ()
1 d
d
r2
r dr
dr
ask Baker
where
thi
comes from
29 Feb p4
m
+ V (r) (r) Gm um (0 )
2
r
X
1
= (r r0 )
um ()um (0 )
r
m
pr:rlo1
We now define the reduced linear operator

L0 m
eq9rLo
d
d
rL0 =
r (r)
dr
dr
"
m (r)
+r
+ V (r)
r2
(9.6)
so Gm (r, r0 ; ) must satisfy

(L0 m r(r))Gm (r, r0 ; ) = (r r0 ),
for 0 < r, r0 < a
and the boundary condition

!
+ Gm (r, r0 ; ) = 0
r
29 Feb p5
for r = a, 0 < r0 < a.
Comments on the eigenvalues m : The RBC will always lead to

n > 0. If m < 0 then the term m (r)/r2 in 9.6 would act like an
attractive sink and there would be no stable solution. Since m > 0,
this term instead looks like a centrifugal barrier at the origin.
Note that the effective tension in this case is r (r), so r = 0 is
a singular point. Thus we must impose regularity at r = 0: |G(r =
0, r0 ; )| < .
9.3.2
Summary of GF for Cyl. Sym.
We have reduced the Greens function for cylindrical symmetry to the

1-dimensional problem:
(L0 m r(r))Gm (r, r0 ; ) = (r r0 ),
for 0 < r, r0 < a,
+ Gm (r, r0 ; ) = 0
r
for r = a, 0 < r0 < a,
|G(0, r0 ; )| < .
9.4. EIGEN VALUE PROBLEM FOR L0
9.4
133
Eigen Value Problem for L0
To solve the reduced Greens function problem which we have just obtained, we must solve the reduced eigen value problem
pr:efp4
(m)
(m)
L0 m u(m)
n (r) = n r(r)un (r)
for 0 < r < a,
du(m)
n
+ u(m)
for r = a,
n (r) = 0
dr
|u(m)
at r = 0.
n (r)| <
In these equation (m)
is the nth eigenvalue of the reduced operator
n
(m )
( )
(m)
L0
and un (r) is the nth eigenfunction of L0 m . From the general
theory of 1-dimensional problems (c.f., chapter 4) we know that
Gm (r, r0 ; ) =
(m) 0
X u(m)
(r )
n (r)un
(m)
for m = 0, 1, 2, . . .
It follows that (using 9.5)

0
G(r, , r , ; ) =
=
29 Feb p6
X
um ()
(m) 0
X u(m)
(r )
n (r)un
(m)
n
m
n
(m) 0
X u(m)
(r , 0 )
n (r, )un
(m)
n
n,m
um (0 )
(m)
where u(m)
n (r, ) = um ()un (r). Recall that G satisfies (L0 )G =
(x x0 ). with RBC. Thus we can conclude
(m)
(m)
L0 u(m)
n (r, ) = n (r)un (r, )
RBC.
These u(m)
n (r, ) also satisfy a completeness relation
X
m,n
(m) 0
u(m)
(r , 0 )
n (r, )un
(x x0 )
=
(x)
=
figure this part

out
29 Feb p7
(r r0 )( 0 )
r0 (r0 )
The radial part of the Greens function, Gm , may also be constructed

directly if solutions satisfying the homogeneous equation are known,
2 Mar p1
134
where one of them also satisfies the r = 0 boundary condition and the
other also satisfies the r = a boundary condition. The method from
chapter 3 (which is valid for 1-dimensional problems) gives
Gm (r, r0 ; ) =
(r)?
u1 (r< )u2 (r> )

r (r)W (u1 , u2 )
where
(L0 m r)u1,2 = 0
|u1 | <
at r = 0
u2
+ u2 = 0
for r = a.
r
The effective mass density is r( ), the effective tension is r (r), and
the effective potential is r(m /r2 + V (r)).
9.5
Uses of the GF Gm(r, r0; )
2 Mar p3
9.5.1
pr:efp5
Eigenfunction Problem
Once Gm (r, r0 ; ) is known, the eigenvalues and normalized eigenfunctions can be found using the relation
(m)
Gm (r, r0 ; ) n
(m) 0
u(m)
n (r)un (r )
(m)
The eigen values come from the poles, the eigen functions come from
the residues.
9.5.2
Normal Modes/Normal Frequencies
pr:NormMode4
In the general problem with no external forces the equation of motion

is homogeneous
2
L0 + 2 u(x, t) = 0
t
!
this section is We look for natural mode solutions:

still rough
+ RBC.
9.5. USES OF THE GF GM (R, R0 ; )

(m)
u(x, t) = ein
135
t (m)
un (r, ).
The natural frequencies are given by

n(m)
(m)
n .
The eigen functions (natural modes) are (cf section 9.4)

(m)
u(m)
n (r, ) = un (r)um ().
The normal modes are

(m)
in
u(m)
n (x, t) = e
t (m)
un (r, ).
The normalization of the factored eigenfunctions u(m)

n (r) and um () is
a
Z
0
(m)
0
drr(r)u(m)
n (r)un0 (r ) = n,n0
Z
0
for n = 1, 2, . . .
dum ()um0 () = m,m0 .
The overall normalization of the (r, ) eigen functions is

2
Z
0
(m0 )
dr(r(r))u(m)
n (r, )un0
pr:normal3
2 Mar p4
(r, ) = n,n0 m,m0
or
Z
Z
0
(m0 )
rdrd(r)u(m)
(r, )
n (r, )un0
d3 x(x)u (x)u(x)
= n,n0 m,m0 .
9.5.3
The Steady State Problem

pr:sss2
This is the case of a periodic driving force:

2
L0 + 2 u(r, , t) = f (r, )eit
t
!
+ u(r, , t) = h()eit
r
136
Note: As long as the normal mode solution has circular symmetry, we

may perturb it with forces f (r, ) and h(). It is not necessary to have
circularly symmetric forces.
The solution is (using 9.5)
u(r, ) =
um ()
Z
0
r (r )dr Gm (r, r ; = + i)
2
+Gm (r, a; = + i)
ad
Z
0
d0 um (0 )f (r0 , 0 )

(a)um (0 )f (r0 , 0 )
0
In this equation 02 d0 um (0 )f (r
, 0 ) is the mth Fourier coefficient of
R 2
0
0
the interior force f (r , ) and 0 ad0 (a)um (0 )f (r0 , 0 ) is the mth
Fourier coefficient of the surface force (a)h(0 ).
9.5.4
Full Time Dependence
2 Mar p5
pr:GR3
For the retarded Greens function we have

GR (r, , t; r0 , 0 , t0 ) =
um ()GmR (r, t, r0 , t0 )um (0 )
where
d i(tt0 )
e
Gm (r, r0 ; = 2 )
L 2
1 u1 (r< )u2 (r> )
Gm (r, r0 ; = 2 ) =
.
r (r) W (u1 , u2 )
GmR (r, t; r0 , t0 ) =
Note: In the exterior case, the poles coalesce to a branch

cut. All space
R
P
has circular symmetry. All the normal limits ( , n,n0 (nn0 ),
etc.,) hold.
9.6
The Wedge Problem
pr:wedge1
We now consider the case of a wedge. The equations are similar for the
internal and external region problems. We consider the internal region
problem. The region R is now 0 < r < a, 0 < < and its boundary
is formed by = 0, 2, and r = a.
9.6. THE WEDGE PROBLEM
137

AA

B

B
B
Figure 9.2: The wedge.
9.6.1
General Case
See the figure 9.2. The angular eigenfunction equation is again 9.4:
2
um () = m um ()
2
RBC.
Note that the operator 2 /2 is positive definite by Greens 1st identity. The angular eigenvalues are completely determined by the angular
boundary conditions. For RBC it is always true the m > 0. This is
physically important since if it were negative, the solutions to 9.4 would
be real exponentials, which would not satisfy the case of periodic boundary conditions.
The boundary condition is now
(
n + )G = 0
x S.
This is satisfied if we choose 1 (r) = 1 /r, 2 (r) = 2 /r, and 3 () =

3 , with 1 , 2 0. The boundary condition (
n + )G = 0 becomes
!
+ 1 G = 0
for = 0
+ 2 G = 0
for =
+ 3 G = 0
r
for r = a, 0 < < .
We now choose the um () to satisfy the first two boundary conditions.

The rest of the problem is the same, except that L0 m gives different m
eigenvalues.
fig10.3
138
9.6.2
Special Case: Fixed Sides
2 Mar p6
The case 1 and 2 corresponds to fixed sides. We thus

have G = 0 for = 0 and = . So the um eigenvalues must satisfy
2
u m = m u m
2
and
um = 0
for = 0, .
The solution to this problem is

s
um () =
with
m =
2
m
sin
!2
m = 1, 2, . . . .
The case m = 0 is excluded because its eigenfunction is trivial. As

2 we recover the full circle case.
ask Baker
why not
2?
2 Mar p7
9.7 The Homogeneous
pr:membrane1
4 Mar p1
Recall the general Greens function problem for circular symmetry. By

substituting the completeness relation for um (), our differential equation becomes
(L0 )G(x, x0 ; ) = (x x0 ) =
where
(r r0 ) X
um ()um (0 )
r
m
1
L0 um () = L0 m um (),
r
L0 m
4 Mar p2
Membrane
d
d
=
r (r)
dr
dr
m (r)
+r
+ V (r) .
r2
We now consider the problem of a complete circle and a wedge.
9.7. THE HOMOGENEOUS MEMBRANE
139
We look at the case of a circular membrane or wedge with V = 0, =

constant, = constant. This corresponds to a homogeneous membrane.
We separate the problem into radial and angular parts.
First we consider the radial part. To find Gm (r, r0 ; ), we want to
solve the problem
"
d
d
r
dr
dr
m r
1
+
2 Gm (r, r0 ; ) = (r r0 )
r
c
with G = 0 and r = a, which corresponds to fixed ends. This problem

was solved in problem set 3:
q
Gm

q
q
Jm (r< /c2 )
q
J m (r> /c2 )Nm (a /c2 )
=
2 Jm (a /c2 )
m (a/c
)N
m (r>
/c2 )
(9.7)
Using 9.5, this provides an explicit solution of the full Greens function
problem. Now we consider the angular part, where we have = 2, so 4 Mar p3
that m = m which means the angular eigenfunctions are the same

as for the circular membrane problem considered before:
1
um = eim
2
for m = m2 , m = 0, 1, 2, . . ..
The total answer is thus a sum over both positive and negative m
G(r, , r0 , 0 ; ) =
um ()Gm (r, r0 ; )um (0 ).
m=
We now redo this with and arbitrary . This implies that

the eigen functions are the same as the wedge problem considered before
s
um () =
m =
2
m
sin
,
!2
140

q
We now get Jm/ (r /c2 ) and Nm/ (r /c2 ). We also get the orig- Show
why
(new Bessel op
inal expansion for G:
G(r, ; r0 , 0 ; ) =
um ()Gm (r, r0 ; )um (0 )
m=1
9.7.1
The Radial Eigenvalues
pr:efp6
The poles of 9.7 occur when

q
Jm (a /c2 ) = 0.
We denote the nth zero of Jm by xm n This gives us

mn =
x m n c 2
a
for n = 1, 2, . . .
where Jm (xm ,n ) = 0 is the nth root of the m Bessel function. To

find the normalized eigenfunctions, we look at the residues of
Gm n
4 Mar p4
(m) 0
u(m)
n (r)un (r )
(m)
We find
s
um
n (r) =
2 Jm (xm n ar )
.
a2 J0 m (xm n )
Thus the normalized eigen functions of the overall operator

(m)
L0 u(m)
n (r, ) = n un (r, )
are
(m)
u(m)
n (r, ) = un (r)um ()
where the the form of um () depends on whether we are considering a

wedge or circular membrane.
9.8. SUMMARY
9.7.2
141
The Physics
The normal mode frequencies are given by the radial eigenvalues

q
m,n =
(m)
c
= x,n .
a
The eigen values increase in two ways: as n increases and as m increases.

For small x (i.e., x 1), Jm (x) m which implies that for larger
m the rise is slower.
As m increases, m increases, so the first root occurs at larger x. As
we increase m, we also increase the number of angular nodes in eim or
sin(mn/). This also increases the centrifugal potential. Thus m,2 is this true?
increases with m. The more angular modes that are present, the more
angular kinetic energy contributes to the potential barrier in the radial
equation.
Now consider behavior with varying for a fixed m. m increases 4 Mar p5
as we decrease , so that m,n increases. Thus the smaller the wedge,
the larger the first frequency. The case 0 means the angular eigenfunctions oscillate very quickly and this angular energy gets thrown
into the radial operator and adds to the centrifugal barrier.
9.8
Summary
1. Whereas cartesian coordinates measure the perpendicular distance from two lines, cylindrical coordinates measure the length
of a line from some reference point in its angle from some reference
line.
2. The -function for circular coordinates is
(x x0 ) =
(r r0 )
( 0 ).
r
3. The Greens function equation for circular coordinates is

"
r (r)
r r
r
(r) 2 u
2
+ V (r) (r) G = (x x0 ).
2
r
#
142
4. The partial expansion of the Greens function for the circular

problem is
G(r, ; r0 , 0 ) =
um ()Gm (r, r0 ; )um (0 ).
5. The radial Greens function for circular coordinates satisfies

(L0 m r(r))Gm (r, r0 ; ) = (r r0 ),
for 0 < r, r0 < a,
where the reduced linear operator is

L0 m
d
d
rL0 =
r (r)
dr
dr
"
m (r)
+r
+ V (r) ,
r2

!
+ Gm (r, r0 ; ) = 0
r
9.9
for r = a, 0 < r0 < a.
Reference
The material in this chapter can be also found in various parts of [Fetter80] and [Stakgold67].
The preferred special functions reference for physicists seems to be
[Jackson75].
Chapter 10
Heat Conduction
Chapter Goals:
Derive the conservation law and boundary conditions appropriate for heat conduction.
7 Mar p1
Construct the heat equation and the Greens function equation for heat conduction.
Solve the heat equation and interpret the solution.
10.1
Introduction
pr:heat1
1
We now turn to the problem of heat conduction. The following physical parameters will be used: mass density , specific heat per unit mass
cp , temperature T , and energy E. Again we consider a region R with
boundary S and outward normal n
.
10.1.1
Conservation of Energy
The specific heat, cp , gives the additional amount of thermal energy

which is stored in a unit of mass of a particular material when its
temperature is raised by one unit: E = cp T . Thus the total energy
can be expressed as
Etotal = E0 +
1
Z
R
d3 xcp T.
The corresponding material in FW begins on page 408
143
144
CHAPTER 10. HEAT CONDUCTION
Differentiating with respect to time gives

T
dE Z 3
= d xcp
dt
t
R
There are two types of energy flow: from across the boundary S
and from sources/sinks in R.
1. Energy flow into R across S. This gives
dE
dt
pr:heatcur1
n
jn dS =
boundary
Z
R
dx jn
where the heat current is defined

jn = kT T.
kT is the thermal conductivity. Note that since T points toward
the hot regions, the minus sign in the equation defining heat flow
indicates that heat flows from hot to cold regions.
2. Energy production in R due to sources or sinks,
dE
dt
=
sources
d3 xq
where q is the rate of energy production per unit mass by sources

inside S.
Thus the total energy is given by
Z
R
d3 xcp
T
t
dE
dt !
!
dE
dE
+
=
dt boundary
dt sources
=
Z
R
d3 x(q jn ).
By taking an arbitrary volume, we get the relation

cp
T
= (kT T ) + q.
(10.1)
10.1. INTRODUCTION
10.1.2
145
Boundary Conditions
pr:bc4
7 Mar p2
There are three types of boundary conditions which we will encounter:

1. T given on S. This is the case of a region surrounded by a heat
bath.
2. n
T for x S given. This means that the heat current normal
to the boundary, n
jn , is specified. In particular, if the boundary
is insulated, then n
T = 0.
3. kT (x)
n T = (T Texternal ) for x S.
In the first case the temperature is specified on the boundary. In the
second case the temperature flux is specified on the boundary. The
third case is a radiation condition, which is a generalization of the first
two cases. The limiting value of give
1 = T Texternal #1
1 = n
T 0 #2 with insulated boundary.
We now rewrite the general boundary condition (3) as
[
n T + (S)]T (x, t) = h(x, t)
for x S
(10.2)
where (S) = /kT (S) and h(S, t) = (/kT (S))Texternal . In the limit
1, T is given. In this case we recover boundary condition #1.
The radiation is essentially perfect, which says that the temperature
of the surface is equal to the temperature of the environment, which
corresponds to . In the other limit, for 1, n
T is given.
Thus we recover boundary condition # 2 which corresponds to 0.
By comparing the general boundary conditions for the heat equation
with the general N -dimensional elastic boundary condition,
[
n + (x)]u(x, t) = h(x, t)
we identify u(x, t) T (x, t) and (x) (x).
146
pr:heateq1
10.2
The Standard form of the Heat Eq.
10.2.1
Correspondence with the Wave Equation
We can make the conservation of energy equation 10.1 look more familiar by writing it in our standard differential equation form
!
T = q(x,
t)
L0 + cp
t
eq10DE
for x G
(10.3)
where the linear operator is

L0 = (kT (x)).
The correspondence with the wave equation is as follows:
Wave Equation
(x)
(x)
(x)f (x, t)
V (x)
7 Mar p3
Heat Equation
kT (x)
(x)cp
(x)q(x,
t)
no potential
For the initial condition, we only need T (x, 0) to fully specify the solution for all time.
10.2.2
Greens Function Problem
We know that because equation 10.3 is linear, it is sufficient to consider only the Greens function problem (which is related to the above
problem by pq(x,
t) = (x x0 )(t t0 ) and h(S, t) = 0):
!
L0 + cp
G(x, t; x0 , t0 ) = (x x0 )(t t0 ),
t
[
n + (S)]G(x, t; x0 , t0 ) = 0
G(x, t; x0 , t0 ) = 0
for x S,
for t < t0 .
10.2. THE STANDARD FORM OF THE HEAT EQ.
147
We lose symmetry in time since only the first time derivative appears.
We evaluate the retarded Greens functions by applying the standard
Fourier transform technique from chapter 6:
0
G(x, t; x , t ) =
Z
L
d i(tt0 )
e
G(x, x0 ; ).
2
is analytic in the Im > 0 plane.

We know by G = 0 for t < t0 that G
Thus we take L to be a line parallel to the real -axis in the upper half
plane. The Fourier Transform of the Greens function is the solution of
the problem
x0 ; ) = (x x0 ),
(L0 cp i)G(x,
x0 ; ) = 0
(
n + (x))G(x,
for x S,
which is obtained by Fourier transforming the above Greens function

problem.
10.2.3
Laplace Transform
pr:LapTrans1
We note that this problem is identical to the forced oscillation Greens

function problem with the substitutions cp and kT . Thus
we identify
x0 ; ) = G(x, x0 ; = i).
G(x,
The single time derivative causes the eigenvalues to be = i. To
evaluate this problem we thus make the substitution s = i. This
substitution results in the Laplace Transformation. Under this trans- 7 Mar p4
formation the Greens function in transform space is related by
x0 ; = is) = G(x, x0 ; = s).
G(x,
is now analytic in the right hand side plane: Re (s) > 0. This variable
G
substitution is depicted in figure 10.1. The transformed contour is
labeled L0 . The Laplace transform of the Greens function satisfies fig10a
the relation
G(x, t; x0 , t0 ) =
i Z
x0 ; = s0 )es(tt0 )
dsG(x,
2 L0
148
?L0
s
Figure 10.1: Rotation of contour in complex plane.
or, by changing the direction of the path, we have
G(x, t; x0 , t0 ) =
1 Z
x0 ; = s0 )es(tt0 ) .
dsG(x,
2i L0
(10.4)
In the following we will denote L0 as L.

The inversion formula
G()
=
d ei G(x, x0 , = t t0 )
is also rotated to become
G(s)
=
d es G(x, x0 ; ).
G(s)
is analytic for all Re (s) > 0. Note that the retarded condition
allows us start the lower limit at = 0 rather than = .
10.2.4
Eigen Function Expansions
7 Mar p5
We now solve the Greens function by writing it as a bilinear sum of

eigenfunctions:
X un (x)un (x0 )
G(x, x0 ; ) =
.
(10.5)
n
n
The eigenfunctions un (x) solve the problem
L0 un (x) = n cp un (x)
for x R
10.2. THE STANDARD FORM OF THE HEAT EQ.
149
Im s
s-plane

+
C
C
C
2
1

q
q

2 1
B 3
B
B
J
J
J
Q
sQ
Q
QP
PP
Re s

q
Figure 10.2: Contour closed in left half s-plane.

where L0 = (kT (x)) with the elastic boundary condition
(
n + (s))un = 0
for x S.
Because of the identification

x0 ; s) = G(x, x0 ; = s)
G(x,
we can substitute 10.5 into the transform integral 10.4 to get
ds X un (x)un (x0 ) s(tt0 )
e
n + s
L 2i n
X
1 I
ds s(tt0 )
=
un (x)un (x)
e
.
2i L n + s
n
G(x, t; x0 , t0 ) =
This vanishes for t < t0 . Close the contour in the left half s-plane
for t t0 > 0, as shown in figure 10.2.
This integral consists of fig10a1
contributions from the residues of the poles at n , where n = 1, 2, . . ..
So
1 I
ds s(tt0 )
0
e
= en (tt ) .
2i Cn n + s
Thus
X
0
G(x, t; x0 , t) =
un (x)un (x0 )en (tt ) .
(10.6)
n
150
We now consider the two limiting cases for t.

Suppose that t t0 . Then 10.6 becomes
tt0
un (x)un (x0 ) =
7 Mar p6
Thus we see that another interpretation of G is as the solution of an

initial value problem with the initial temperature
T (x, 0) =
why is this?
(x x0 )
.
cp
(x x0 )
cp
and no forcing term.

Now suppose we have the other case, t t0 1. We know n > 0
for all n since L0 is positive definite (physically, entropy requires k > 0
so that heat flows from hot to cold). Thus the dominant term is the
one for the lowest eigenvalue:
0
G u1 (x)u1 (x0 )e1 (tt )
(t t0 ) 1.
In particular, this formula is valid when (t t0 ) > 1/2 . We may thus

interpret 1/n = n as the lifetime of these states. After (t t0 ) N ,
all contributions to G from eigen values with n N are exponentially
small.
This is the physical meaning for the eigen values. The reason that
the lowest eigen function contribution is the only one that contributes
for t t0 1 is because for higher N there are more nodes in the
eigenfunction, so it has a larger spatial second derivative. This means
(using the heat equation) that the time derivative of temperature is
large, so the temperature is able to equilize quickly. This smoothing
or diffusing process is due to the term with a first derivative in time,
which gives the non-reversible nature of the problem.
10.3
Explicit One Dimensional Calculation
9 Mar p1
We now consider the heat equation in one dimension.
10.3. EXPLICIT ONE DIMENSIONAL CALCULATION
10.3.1
151
Application of Transform Method
pr:fsp1
Recall that the 1-dimensional Greens function for the free space wave
equation is defined by
9 Mar p2
(L0 )G = (x x0 )
for < x < .
We found that the solution for this wave equation is
1 ei |xx |/c
.
G(x, x ; ) =
c
2
0
Transferring from the wave equation to the heat

as discussed
q equation
above, we substitute kT , , c = / where =
KT /cp is the thermal diffusivity, and i s which means Im > pr:kappa1

0 becomes Re s > 0. The substitutions yield
kT KT and
s
1
cp e
2 s
x0 ; s) =
G(x,
or
|xx0 |
c cp ?
1 s/|xx0 |
0
cp G(x, x ; s) = e
.
2 s
We see that plays the role of a velocity. Now invert the transform 9 Mar p3
to obtain the free space Greens function for the heat equation:
cp G(x, t; x0 , t0 ) =
Z
L
ds s(tt0 )
x0 ; s)
e
cp G(x,
2i
Z
L
10.3.2
ds es(tt ) s/|xx |
.
2i
2 s
Solution of the Transform Integral
Our result has a branch on
s = |s|ei
s. We parameterize the s-plane:

for < <
This gives us Re s = |s|1/2 cos(/2) > 0. We choose the contour of

integration based on t.
9 Mar p5
152

Im s
s-plane

+

L1

L2
L3

-

s = i |s|
-
q
s = i |s|
Re s

B
B
B
L4
J
LJ5J
Q
sQ
Q
QP
PP
Figure 10.3: A contour with Branch cut.

For t < t0 we have the condition G = 0. Thus we close the contour
in the right half plane so that
h
i s
exp s(t t0 ) s|x x|/ 0
since both terms are increasingly negative. Since the contour encloses
no poles, we recover G = 0 as required.
For tt0 > 0, close contour in the left half plane. See figure 10.3. We
know by Cauchys theorem that the integral around the closed contour
L + L1 + L2 + L3 + L4 + L5 vanishes. We perform the usual Branch
cut evaluation, by treating the different segments separately. For L3 it
is convenient to use the parameterization s = ei for < < as
shown in figure 10.3. In this case the integral becomes
0
1 1 Z
d| |[1 + O((t t0 )) + O(1/2 |x x0 |/ )]0.

2 2
In this equation we assert that it is permissible to take the limit 0
before the other quantities are taken arbitrarily large.
q
For the contour L2 above the branch cut we have s = i |s|,

q
and for the contour L4 below the branch cut we have s = i |s|
153
Combining the integrals for these two cases gives

lim
ds es(tt ) 2 cos
2 2
|s|/|x x0 |
q
|s|
For L1 and L5 the integral vanishes. By letting s = Rei where

< < we have
h
i
0

|xx
| R2 cos 1
exp [s|x x0 |/]
exp

0.

s
R
Our final result is
q
1 Z ds s(tt0 )
e
cos s/(x x0 ).
cp G(x, t; x , t ) =
s
2 0
0
Substituting s = u2 gives
cp G(x, t; x0 , t0 ) =
2udu
u
2
0
cos |x x0 |eu (tt )
2u
or
1
cp G(x, t; x0 , t0 ) = I(t t0 , |x x0 |/ )
where (since the integrand is even)

1 Z du u2 t+iuy
e
.
I(t, y) =
2
This can be made into a simple Gaussian by completing the square:
I(t, y) = ey
2 /(4t)
du (u iy )2
2t
e
.
2
By shifting u u + iy/2t, the result is
9 Mar p6
ey /(4t)
I(t, y) =
.
4t
The free space Green function in 1-dimension is thus
cp G(x, t; x0 , t0 ) = q
pr:gaus1
4|t t0 |
(xx0 )2
4(tt0 )
(10.7)
154
10.3.3
The Physics of the Fundamental Solution
This solution corresponds to a pure initial value problem where, if x0 =

t0 = 0, we have
2
ex /4t
cp G(x, t) =
.
4t
At the initial time we have
t0
cp G(x x0 ) = (x).
9 Mar p7
2
1. For x
> 4t, the amplitude is very small. Since G is
small for
x 4t, diffusion proceeds at rate proportional to t, not t
as in wave equation. The average propagation is proportional to
t1/2 . This is indicative of a statistical process (Random walk). It
is non-dynamical in that it does not come from Newtons laws.
Rather it comes from the dissipativeconduction nature of thermodynamics.
2. For any t > 0 we have a non-zero effect for all space. This corresponds to propagation with infinite velocity. Again, this indicates
the non-dynamical nature of the problem. This is quite different
from the case of wave propagation, where an event at the origin
does not affect the position x until time x/c.
3. Another non-dynamical aspect of this problem is that it smoothes
the singularity in the initial distribution, whereas the wave equation propagates all singularities in the initial distribution forward
in time.
11 Mar p1
11 Mar p2
4. is a fundamental parameter whose role for the heat equation

is analogous to the role of c for the wave equation. It determines the rate of diffusion. (= kT /cp ) has the dimensions of
(distance)2 /time, whereas c has the dimensions of distance/time.
10.3.4
Solution of the General IVP
We now use the Greens function to solve the initial value problem:
2
d
kT 2 + cp
T (x, t) = 0
x
dt
!
for < x, x0 ,
155
T (x, 0) = T0 (x)
T 0
for |x|
The method of the solution is to use superposition and 10.8:

T (x, t) =
10.3.5
dx0 T0 (x0 )cp G(x0 , 0; x, t)
1 Z 0 (xx0 )2 /(4t)
dx e
T0 (x0 )
4t
(10.8)
Special Cases
Initial -function
Suppose T0 (x) = (x x0 ). Then we have T (x, t) = cp G(x00 , 0; x, t).
Thus we see that G is the solution to the IVP with the -function as
the initial condition and no forcing term.
Initial Gaussian Function
pr:gaus2
We now consider the special case of an initial Gaussian temperature 11 Mar p3

2
distribution. Let T0 (x)= (a/)1/2 eax . The width of the initial distribution is (x)0 = 1/ a. Plugging this form of T0 (x) into 10.8 gives
Z
02
1
0 2
T (x, t) =
dx0 e(xx ) /(4t)ax
4ta
x2
1
1
T (x, t) = q
e (1/a)+4t
(1/a) + 4t
1
=
1
2
e(x/x)
x

where x = (x0 )2 + 4t. The packet is spreading as (x)2 =

4t + (x0 )2 . Again, x t1/2 like a random walk, (again nondynamical). Suppose t (x0 )/4. This is the simplest quantity
with dimensions of time,qso is the characteristic time of the system.
We rewrite x = (x0 ) 1 + t/ . Thus for t ,
x
t/ (x0 ).
156
11 Mar p4
= (x0 )2 /4 is a fundamental unit of time in the problem. Since ask Baker

the region is infinite, there does not exist any characteristic distance about omitted
for the problem.
11 Mar p5
10.4
Summary
1. Conservation of energy for heat conduction is given by the equation

T
cp
= (kT T ) + q,
t
where is the mass density, cp is the specific heat, T is the temperature, kT is the thermal conductivity, and q is the rate of energy
production per unit mass by sources inside the region.
2. The general boundary condition for heat conduction is
[
n T + (S)]T (x, t) = h(x, t)
for x S.
3. The heat equation is

!
L0 + cp
T = q(x,
t)
t
for x G,
where the linear operator is

L0 = (kT (x)).
4. The Greens function equation for the heat conduction problem
is
!
G(x, t; x0 , t0 ) = (x x0 )(t t0 ).
L0 + cp
t
5. The solution of the heat equation for the initial value problem in
one dimension is
1 Z 0 (xx0 )2 /(4t)
T (x, t) =
dx e
T0 (x0 ),
4t
which is a weighted integration over point sources which individually diffuse with a gaussian shape.
10.5. REFERENCES
10.5
157
References
A similar treatment (though more thorough) is given in [Stakgold67b,

p194ff]. See also [Fetter80, p406ff].
The definitive reference on heat conduction is [Carslaw86].
158
Chapter 11
Spherical Symmetry
Chapter Goals:
Derive the form of the linear operator in spherical
coordinates.
Show that the angular part of the linear operator
L is hermitian.
28 Mar p1 (17)
m
Write the eigenvalue equations for Yl .

Write the partial wave expansion for the Greens
function.
Find the Greens function for the free space problem.
Our object of study is the Greens function for the problem
[L0 (x)]G(x, x0 ; ) = (x x0 )
(11.1)
with the regular boundary condition (RBC)

[
n + K(S)]G(x, x0 ; ) = 0
for x0 in a region R and x in the regions boundary S. The term x is a
field point, and x0 is a source point. The unit vector n
is the outward
normal of the surface S. The operator L0 is defined by the equation
L0 = ( (x)) + V (x).
159
eq11.1
160
CHAPTER 11. SPHERICAL SYMMETRY

z
(r, , )

7

r

y
l

l

l

l
l

Figure 11.1: Spherical Coordinates.

We have solved this problem for the one and two dimensional cases in
which there was a certain degree of symmetry.
11.1
Spherical Coordinates
28 Mar p2
pr:spher1
We now treat the problem in three dimensions. For this we use spherical
coordinates (since we will later assume angular independence). A point
in spherical coordinates is denoted (r, , ), where the range of each
variable is
0 r < ,
0 < ,
0 < 2.
We use the following transformation of coordinate systems:
z = r cos ,
x = r sin cos ,
y = r sin sin .
fig11.1
This relationship is illustrated in figure 11.1 For an arbitrary volume
11.1. SPHERICAL COORDINATES
161
element we have
d3 x = (dr)(rd)(r sin d)
= r2 ddr
where is the solid angle, and an infinitesimal of solid angle is d = pr:Omega1
sin dd.
We further define the delta function
f (r, , ) = f (x)
=
=
d3 x0 f (x0 ) (x x0 )
dr0 r0 sin 0 d0 d0 f (r0 , 0 , 0 ) (x x0 ).
From this we can extract the form of the -function for spherical coordinates:
1
(r r0 )( 0 )( 0 )
sin
(r r0 )
=
( 0 )
2
r
(x x0 ) =
r2
where the solid angle -function is

( 0 ) =
( 0 )( 0 )
.
sin
28 Mar p3
pr:delta2
We want to rewrite equation 11.1 in spherical coordinates. First we

define the gradient
pr:grad2
= r

+
+
.
r r r sin
See [Boas] for derivations of identities involving . The divergence is

A=
1 2
1
1
(r Ar ) +
(sin A ) +
A .
2
r r
r sin
r sin
When we apply this to the case

A = (x).
87 notes have
Gauss
law,
p18
162
the result is
!
1

( ) = 2
r2
+
sin
r r
r
r sin
r
!
1
+
r sin r sin
eq11.2
28 Mar p4
where
L
pr:Lthph1
(11.2)
where = (x) = (r, , ).

Now we can write L0 . We assume that , , and V are spherically
symmetric, i.e., they are only a function of r: (x) = (r), (x) = (r),
V (x) = V (r). In this case the linear operator is
1
L0 = 2
r2 (r)
r r
r
eq11.2b
=
sin
sin
(r)
L + V (r)
r2
(11.3)
1 2
,
sin2 2
which is the centrifugal term from equation 11.2. In the next few sections we will study the properties of L0 .
11.2
Discussion of L
Note that L is a hermitian operator on the surface of the sphere, as

shown by the following argument. In an earlier chapter we derived the
Greens Identity
Z
eq11.3
28 Mar p5
d3 xS (x)L0 u(x) =
d3 x (u (x)L0 S(x))
(11.4)
where u and S satisfy RBC. We use this fact to show the hermiticity
of L . Consider the functions
S(x) = S(r)S(, )
and
u(x) = u(r)u(, )
where u and S satisfy RBC. Such functions are a subset of the functions
which satisfy equation 11.4. Choose u(, ) and S(, ) to be periodic
in the azimuthal angle :
u(, ) = u(, + 2),
S(, ) = S(, + 2).
11.3. SPHERICAL EIGENFUNCTIONS
163
Now substitute d3 x = r2 drd and L0 (as defined in equation 11.3) into

equation 11.4. The term
1
2
r2 (r)
r r
r
+ V (r)
in L0 is hermitian so it cancels out in 11.4. All that is left is

Z
Z
(r)
r drS (r) 2 u(r) dS (, )L u(, ) =
r
Z
Z
(r)
r2 drS (r) 2 u(r) d (u (, )L S(, ))
r
This can be rewritten as

Z
28 Mar p6
(r)
r drS (r) 2 u(r)d
r
Z
Z
S (, )L u(, )
d (u (, )L S(, ))
= 0.
The bracket must then be zero. So

Z
d(S (, )L u(, ) =
d (u (, )L S(, )) .
(11.5)
This is the same as equation 11.4 with d3 x d and L0 L . Thus eq11.4

L is hermitian. If the region did not include the whole sphere, we
just integrate the region of physical interest and apply the appropriate
boundary conditions. Equation 11.5 can also be obtained directly from 28 Mar p7
the form of L by applying integration by parts on L twice, but
using L0 = L0 is much more elegant.
2
Note that the operators
2 and L commute:
"
2
, L = 0.
2
#
Thus we can reduce equation 11.1 to a one dimensional case and expand
the Greens function G in terms of a single set of eigenfunctions which
are valid for both 2 /2 and L . We know that L0 and L are
hermitian operators, and thus the eigenfunctions form a complete set.
For this reason this method is valid.
164
11.3
Spherical Eigenfunctions
28 Mar p8
We want to find a common set of eigenfunctions valid for both L and

2 /2 . Note that
pr:sphHarm1
im
2 eim
2e
=
m
2 2
2
m = 0, 1, 2, . . . .
So (2)1/2 eim are normalized eigen functions of 2 /2 . We define

the functions Ylm (, ) as the set of solutions to the equation
L Ylm (, ) = l(l + 1)Ylm (, )
eq11.5
for eigen values l(l + 1) and periodic boundary conditions, and the
equation
eq11.6
pr:orthon3
(11.6)
2 m
Y (, ) = m2 Ylm (, )
2 l
m = 0, 1, 2, . . . .
(11.7)
We can immediately write down the orthogonality condition (due to

the hermiticity of the operator L ):
is this true?
dYlm (, )Ylm
0 (, ) = l,l0 m,m0 .
(If there are degeneracies, we may use GramSchmidt techniques to

arrive at this result). We can choose the normalization coefficient to
be one. There is also a completeness relation which will be given later.
11.3.1
Reduced Eigenvalue Equation
28 Mar p9
pr:ulm1
eq11.6b
We now separate the eigenfunction into the product of a -part and a

-part:
eim
Ylm (, ) = um
(11.8)
l (cos )
2
(which explicitly solves equation 11.7, the differential equation involving ) so that we may write equation 11.6 as
"
m2
m
sin
+ 2 um
l (cos ) = l(l + 1)ul (cos ).
sin
sin
!
165
All we have left to do is solve this eigenvalue equation. The original

region was the surface of the sphere because the solid angle represents
area on the surface. We make a change of variables:
x = cos .
The derivative operator becomes
d
dx d
d
=
= sin
d
d dx
dx
so
1 d
d
=
.
sin d
dx
The eigen value equation for u becomes
"
d
d
m2
m
(1 x2 )
+
um
l (x) = l(l + 1)ul (x)
2
dx
dx
1x
!
(11.9)
defined on the interval 1 < x < 1. Thus x = 1 corresponds to = 0,

and x = 1 corresponds to = . Note that , which represents
the effective tension, is proportional to 1 x2 , so both end points are
singular points. On account of this we get both regular and irregular
solutions. A solution occurs only if the eigen value l takes on a special
value. Requiring regularity at x = 1 implies l is an integer. Note also
that equation 11.9 represents an infinite number of one dimensional
eigenvalue problems (indexed by m), which makes sense because we
started with a partial differential equation eigenvalue problem.
The way to solve the equation near a singular point is to look for
solutions of the form xp [power series], as in the solutions to Bessels
equation.
11.3.2
Determination of um
l (x)
We now determine the function um

l (x) which is regular at x = 1.
Suppose that it is of the form
2
um
l (x) = (1 x ) (power series).
We want to determine the power term . First we compute
eq11.7
28 Mar p10
Verify this
change this
30 Mar p1
30 Mar p2
(11.10)
eq11.8
166

(1 x2 )
d
(1 x2 ) = (2x)(1 x2 ) ,
dx
and then
"
d
d
(1 x2 ) (1 x2 )
dx
dx
= 2 (1 x2 )1 (2x)2
+ (2x)(1 x2 ) + . . . .
For the case x 1 we can drop all but the leading term:
"
d
d
(1 x2 ) (1 x2 ) 4x2 2 (1 x2 )1
dx
dx
eq11.9
x 1.
(11.11)
Plugging equations 11.10 and 11.11 into equation 11.9 gives

(1 x2 )1 A[4 2 + m2 ] (l + 1)lA(1 x2 ) ,
x 1.
where A is the leading constant from the power series. Note however
that (1 x2 ) approaches zero faster that (1 x2 )1 as x 1. So we
get m2 = 4 2 , or
m
= .
2
We thus look for a solution of the form
2 m/2
um
Cm (x),
l (x) = (1 x )
eq11.9a
30 Mar p3
(11.12)
where Cm (x) is a power series in x with implicit l dependence. We

expect regular and irregular solutions for Cm (x). We plug this equation
into equation 11.9 to get an equation for Cm (x). The result is
00
0
(1 x2 )Cm
+ 2x(m + 1)Cm
(l m)(l + m + 1)Cm (x) = 0. (11.13)
eq11.9b
We still have the boundary condition that um

l (x) is finite. In the case
that m = 0 we have
(1 x2 )C000 + 2xC00 l(l + 1)C0 = 0.
eq11.10
pr:LegEq1
(11.14)
This is called Legendres equation. We want to find the solution of

this equation which is regular at x = 1. We define Pl (x) to be such
167
a solution. The irregular solution at x = 1, called Q(x), is of interest

if the region R in our problem excludes x = 1 (which corresponds to
cos = 1 or = 0). Note that we consider l to be an arbitrary complex
number. (From the general theory, however, we know that the eigen
values are real.) By convention we normalize: Pl (1) = 1. We know
C0 (x) = Pl (x), because we defined C0 (x) to be regular at x = 1. But
l (x) to be the regular
x = 1 is also a singular point. We define R
solution and Il (x) to be the irregular solution at x = 1 We can write
l (x) + B(l)Il (x).
C0 (x) = Pl (x) = A(l)R
(11.15)
But if we further require that Pl must be finite (regular) at x = 1, we

then have B(l) = 0 for l = 0, 1, 2, . . ..
30 Mar p4
We now take Legendres equation, 11.14, with C0 (x) replaced by
Pl (x) (the regular solution), and differentiate it m times using Leibnitz
formula
pr:LeibForm1

m
X
dm
m di f dmi g
(f (x)g(x)) =
.
dxm
i dxi dxmi
i=0
This yields
d2
(1 x2 ) 2
dx
dm
d
Pl (x) + 2(m + 1)x
m
dx
dx
m
d
(l m)(l + m + 1) m Pl (x) = 0.
dx
!
dm
Pl (x)
dxm
(11.16)
Thus (dm /dxm )Pl (x) is also a solution of equation 11.13. We thus see eq11.11
that
dm
Cm (x) = m Pl (x),
(11.17)
dx
where still needs to be determined. Once we find out how to chose l eq11.11b
so Pl (x) is 0 at l, Pl0 for l0 6= l will also be zero. So we see that once we
determine constraints on l such that the Pl (x) which solves the m = 0
equation is zero at zero at x = 1, we can generate a solution for the
case m 6= 0.
We now calculate a recurrence relation for Pl (x). Set x = 1 in pr:recrel1
equation 11.16. This gives
dm+1
2(m+1)
Pl (x)
dxm+1
"
dm
= (lm)(l+m+1)
Pl (x)
dxm
"
x=1
. (11.18)
x=1
168
This tells us the (m + 1)th derivative of Pl (x) in terms of the mth eq11.12
derivative. We can differentiate l times if l is an integer. Take l to be
an integer. The case m = l yields
"
dl+1
Pl (x)
dxl+1
= 0.
(11.19)
x=1
So all derivatives are zero for m > l at x = 1. This means that P (x)
is an lth order polynomial, since all of its Taylor coefficients vanish for
m > l. Since Pl is regular at x = 1 and is a polynomial of degree l,
it must be regular at x = 1 also. If l were not an integer, we would
obtain a series which diverges at x = 1. Thus we conclude that l must
be an integer. Note that even for the solution which is not regular at
x = 1, for which l is not an integer, equation 11.18 is still valid for
calculating the series.
For a general m, we substitute equation 11.17 into equation 11.12
to get
m
2 m/2 d
(x)
=
(1
x
)
um
Pl (x).
(11.20)
l
dxm
This equation holds for m = 0, 1, 2, . . . . Furthermore this equation
dm
solves equation 11.13. Note that because dx
m Pl (x) is regular at x = 1
30 Mar p6
and x = 1, um
(x)
is
also.
l
We now compute the derivative. Using equation 11.18, for m = 0
we get
d
2 Pl (x)|x=1 = l(l + 1)Pl (1) = l(l + 1).
dx
By repeating this process for m = 1, 2, . . . and using induction, we find
yet to be veri- that the following polynomial satisfies equation 11.18
ask Baker isnt

this valid only
at x = 1
30 Mar p5
fied
Pl (x) =
eq11.13
pr:rodform1
1 dl 2
(x 1)l .
2l l! dxl
(11.21)
This is called Rodrigues formula for the Legrendre function.

We define the associated Legendre polynomial
Plm (x) = (1)m (1 x2 )m/2
dm
Pl (x)
dxm
(1 x2 )m/2 dl+m 2
=
(x 1)l ,
l
l+m
2 l!
dx
m0
m 0.
(11.22)

eq11.13b
169
For m > l, Plm (x) = 0. So the allowed range of m is l m l. Thus

the value of m affects what the lowest eigenvalue, l(l + 1), can be.
11.3.3
Orthogonality and Completeness of um

l (x)
We want to choose um
l (x) to be normalized. We define the normalized pr:normal4
eigen functions as the set of eigen functions which satisfies the condition
(with = 1)
Z
1
m
dx (um
l (x)) ul0 (x) = 1.
(11.23)
1
We need to evaluate 1
dx|Plm (x)|2 . Using integration by parts and eq11.13c
equation 11.22 we get (a short exercise)
dx|Plm (x)|2
dx (Plm (x)) Plm (x) =
2 (l + m)!
, (11.24)
2l + 1 (l m)!
so the normalized eigenfunctions are

v
u
u
m
ul (x) = t
30 Mar p7
2 (l + m)! m
P (x).
2l + 1 (l m)! l
(11.25)
The condition for orthonormality is

m
hum
l (x), ul0 (x)i
dx
eq11.14
(um
l (x))
pr:orthon4
um
l0 (x)
= ll0 ,
(11.26)
The corresponding completeness relation is (as usual, with = 1)
m 0
0
um
l (x)ul (x ) = (x x ).
(11.27)
l=m
The problem we wanted to solve was equation 11.6, so we substitute

back in x = cos into the completeness relation, which gives
m
0
0
um
l (cos )ul (cos ) = (cos cos )
l=m
= (( 0 )( sin ))
( 0 )
=
.
sin
170
In the second equality we used a Taylor expansion for near 0 , which

yields
cos cos 0 = ( 0 ) sin .
In the third equality we used the -function property (ax) = |a|1 (x).
The completeness condition for um
l (cos ) is thus
m
0
um
l (cos )ul (cos ) =
l=m
30 Mar p8
Similarly, the orthogonality condition becomes (since

R
0 sin d)
Z
11.4
pr:spherH1
1 Apr p1a
( 0 )
.
sin
m
d sin um
l0 (cos )ul (cos ) = ll0 .
(11.28)
R1
d(cos ) =
(11.29)
Spherical Harmonics
We want to determine the properties of the functions Ylm , such as

completeness and orthogonality, and to determine their explicit form.
We postulated that the solution of equations 11.6 and 11.7 has the form
(c.f., equation 11.8)
eim
Ylm (, ) = um
l (cos )
2
for integer l = m, m + 1, m + 2, . . . and m 0, where we have found
(equation 11.25)
v
u
u (l m)! 2l + 1
m
ul (cos ) = t
(sin )m
(lm )!
d
d cos
!m
Pl (cos ).
We define the Ylm (, ), for m > 0, as

Ylm (, ) (1)m Ylm (, )
im
me
= (1)
um
l (cos ).
2
im
pr:consho1
The term (1)m is a phase convention and e2 is an eigen function.

This is often called the Condon-Shortley phase convention.
11.4. SPHERICAL HARMONICS
171
Othonormality and Completeness of Ylm
11.4.1
We saw that the functions um

l satisfy the following completeness condition:
X
( 0 )
m
0
um
(cos
)u
(cos
)
=
for all m
l
l
sin
l=m
where m is fixed and positive. We know that
0
eim eim
= ( 0 ).
2
2
m=
Multiply
1
(
sin
(11.30)
0 ) into equation 11.30, so that
eq11.15
1 GApr 1b
( )( )
=
sin
=
im
m=
|m|
|m|
ul (cos )ul
l|m|
im0
e
(cos )
Ylm (, )Ylm (, 0 ),
m= l|m|
since
eim |m|
Ylm = (1)m ul (cos ) for m < 0,
2
and
Yl
im
me
= (1)
|m|
ul (cos ) for m < 0.
Thus we have the completeness relation

( 0 ) =
X
l
X
Ylm (, )Ylm (0 , 0 ).
(11.31)
l=0 m=l
We also note that L has (2l + 1)fold degenerate eigenvalues l(l + 1) eq11.16
in
L Ylm (, ) = l(l + 1)Ylm (, ).
Thus m is like a degeneracy index in this equation.
Next we look at the orthogonality of the spherical harmonics. The pr:orthon5
172
orthogonality relation becomes

Z
dYl
m0
Z 1
d
m0
(, )Yl (, ) =
mm0
d cos um
l (cos )ul0 (cos )
2
0
1
= ll0 mm0 ,
Z
where d = dd sin on the right hand side, Because the us are

orthogonal and the eim s are orthogonal, the right hand side is zero
when l 6= l0 or m 6= m0 .
11.5
GFs for Spherical Symmetry
1 Apr 2a
We now want to solve the Greens function problem for spherical symmetry.
11.5.1
GF Differential Equation
The first step is to convert the differential equation into spherical coordinates. The equation we are considering is
[L0 (x)]G(x, x0 ; ) = (x x0 ).
(11.32)
By substituting the L0 for spherically symmetric problems, which we

I dont know found in equation 11.3, we have
eq11.16a
how to fix this.
"
d
1 d
r2 (r)
2
r dr
dr
(r)
+ 2 L + V (r) Glm (x, x0 ; 0 )
r
(r r0 )
( 0 )
=
2
r
X
l
(r r0 ) X
=
Ylm (, )Ylm (0 , 0 ).
r2
l=0 m=l
eq11.17
pr:ExpThm2
(11.33)
where the second equality follows from the completeness relation, equation 11.31. Thus we try the solution form
G(x, x0 ; ) =
X
l
X
l=0 m=l
Ylm (, )Glm (r, r0 ; 0 )Ylm (, ).
(11.34)
11.5. GFS FOR SPHERICAL SYMMETRY

eq11.17b
173
Note that the symmetry of , and 0 , 0 in this solution form means

that Greens reciprocity principle is satisfied, as required. Substituting
this into equation 11.33 and using equation 11.6 results in L being
replaced by the eigenvalue of Ylm , which is l(l + 1). Superposition says
that we can look at just one term in the series. Since the linear operator
no longer involves , , we may divide out the Ylm s from both sides to
get the following radial equation
(r)
(r r0 )
+ 2 l(l + 1) + V (r) Glm (r, r0 ; 0 ) =
.
r
r2
(11.35)
The linear operator for this equation has no m dependence. That is, m eq11.18
is just a degeneracy index for the 2l + 1 different solutions of the L
equation for fixed l. Thus we can rewrite our radial Greens function
Glm as Gl and define the radial operator as
"
1 d
d
2
r2 (r)
r dr
dr
(l)
L0
d
d
=
r2 (r)
dr
dr
"
+r
(r)
l(l + 1) + V (r) .
r2
(11.36)
We have reduced the three dimensional Greens function to the standard eq11.18b
single dimensional case with effective tension r2 (r), effective potential
energy V (r), and a centripetal kinetic energy term ( (r)/r2 )l(l + 1).
11.5.2
Boundary Conditions
pr:bc5
What about the boundary conditions? If the boundary conditions are 1 Apr 2b
not spherically symmetric, we need to take account of the angles, i.e.,
[
n + (S)]G(x x0 , ) = 0,
for x on S and x0 in R.
Consider a spherical region as shown in figure 11.2. For spherically fig11b
symmetric boundary conditions, we can set int (S) = ka and ext (S) =
kb . Thus
#
"
+ ka G = 0 for r = a,
r
"
+ kb G = 0 for r = b.
r
174

' $
S2 -

R
S1
a
b
b

b &
%
Figure 11.2: The general boundary for spherical symmetry.

If we insert G from equation 11.34 into these conditions, we find the
following conditions on how Gl behaves:
"
+ ka Gl = 0 for r = a,
r
"
+ kb Gl = 0 for r = b,
r
for all l. These equations, together with equation 11.35, uniquely determine Gl .
With these definitions we can examine three interesting cases:
(1) the internal problem
(2) the external problem
(3) the all space problem
a ,
b 0, and
a 0 and b .
pr:spcProb1
These cases correspond to bound state, scattering, and free space problems respectively.
4 Apr p1
11.5.3
GF for the Exterior Problem
We will now look at how to determine the radial part of the Greens
function for the exterior problem. essential idea is that we have taken
a single partial differential equation and broken it into several ordinary
differential equations. For the spherical exterior problem the region R
of interest is the region outside a sphere of radius a, and the boundary S
is the surface of the sphere. The physical parameters are all spherically
11.5. GFS FOR SPHERICAL SYMMETRY
175
symmetric: (r), (r), V (r), and (S) = ka . Our boundary condition

is
"
#
+ ka G(x, x0 ; ) = 0,
r
where r = a for all , (that is, |x0 | > |x| = a). This implies
"
4 Apr p2,3
+ ka Gl (r, r0 ; ) = 0 for r0 > r = a.

r
The other boundary condition is that Gl is bounded as r .

We now want to solve Gl (r, r0 ; ). Recall that we have seen two
ways of expressing the Greens function in terms of solutions of the
homogeneous equation. One way is to write the Greens functions as
a product of the solution satisfying the upper boundary condition and
the solution satisfying the lower boundary condition, and then divide
by the Wronskian to ensure continuity. Thus we write
Gl (r, r0 ; ) =
1 ul1 (r< , )ul2 (r> , )

,
r2 (r)
W (ul1 , ul2 )
(11.37)
where ul1 and ul2 satisfy the equations
eq11.18c
[Ll0 (r)r2 ]ul1 (r, ) = 0,

[Ll0 (r)r2 ]ul2 (r, ) = 0,
and the boundary conditions
"
+ ka ul1 = 0 for r = a,
r
ul2 (r, ) < when r .
The other way of expressing the Greens function is to look at how

it behaves near its poles (or branch cut) and consider it as a sum of
residues. This analysis was performed in chapter 4 where we obtained
the following bilinear sum of eigenfunctions:
4 April p4
0
Gl (r, r , ) =
(l) 0
X u(l)
n (r)un (r )
n
(l)
(11.38)
176
Note that what is meant here is really a generic sum which can mean eq11.19
either a sum or an integral depending on the spectrum of eigenvalues.
For the external problem we are considering, the spectrum is a pure
continuum and sums over n should be replaced by integrals over n .
The u(l)
n (r) solve the corresponding eigen value problem
(l)
(l) 2
(l)
L0 u(l)
n (r) = n r (r)un (r)
with the boundary conditions that

"
+ ka u(l)
n = 0 for r = a,
r
l
and u(l)
n is finite as r . The interior problem, with un finite as
r 0, has a discrete spectrum.
The normalization of the u(l)
n (r) is given by the completeness relation
(l) 0
u(l)
n (r)un (r ) =
4 Apr p5
We insert equation 11.38 into 11.34 to get

G(x, x0 ; ) =
X u(l,m)
(x)u(l,m)
(x0 )
n
n
n
eq11.20
(r r0 )
.
r2 (r)
(l)
(11.39)
where
u(l,m)
(x) = Ylm (, )u(l)
n
n (r).
and (l)
n is the position of the nth pole of Gl . So the eigenvalues n are
(l)
the n determined from the r-space eigenvalue problem. The corresponding eigenfunctions u(l,m)
(x) satisfy
n
(l)
2
(l,m)
(x) = (l)
(x).
L0 u(l,m)
n
n r (r)un
The completeness relation for u(l,m)

(x) is found by substituting equation
n
11.39 into 11.32 and performing the same analysis as in chapter 4. The
result is
X
(x x0 )
(l)
(l) 0
un (r)un (r ) =
.
(x)
n
11.6. EXAMPLE: CONSTANT PARAMETERS
11.6
177
Example: Constant Parameters
4 Apr p6
Old HW#4
We now look at a problem from the homework. We apply the above

analysis to the case where V = 0 and and are constant. Our
operator for L becomes (c.f., equation 11.36)
(l)
L0
"
d
d
=
r2
dr
dr
+ l(l + 1) .
The equation for the Greens function becomes (after dividing by ):

"
d
d
r2
dr
dr
#
2 2
+ l(l + 1) k r
Gl (r, r0 ; ) =
1
(r r0 ),
where k 2 = / = /c2 .
11.6.1
Exterior Problem
We again have the boundary conditions

"
+ ka Gl (r, r0 ; ) = 0,
r
(11.40)
where r = a, r0 > a, and Gl bounded as r . As usual, we assume (eq11.21d

the solution form
(l)
Gl (r, r0 , ) =
(l)
1 u1 (r< , )u2 (r> , )

.
(l)
(l)
2
r (r)
W (u1 , u2
(11.41)
We solve for u1 and u2 :

"
d
d
r2
dr
dr
eq11.21a
!
#
2 2
+ l(l + 1) k r
4 Apr p7
(l)
u1,2 = 0,
where u1 and u2 satisfy the boundary

conditions as r = a and r
q
respectively and k /c = /c2 . This is the spherical Bessel equation from the third assignment of last quarter. We found the solution
R(r)
u(r) = ,
r
178
where R(r) satisfies the regular Bessel equation

"
d
d
r
dr
dr
(l + 12 )2
+
k 2 r R(r) = 0.
r
#
The solutions for this equation are:

R Jl+ 1 , Nl+ 1 .
2
pr:sphBes
By definition
J 1 (x),
2x l+ 2
r
nl (x) =
N 1 (x),
2x l+ 2
r
(1)
hl (x) =
H 1 (x),
2x l+ 2
where jl (kr) is a spherical Bessel function, nl (kr) is a spherical Neu(1)
mann function, and hl (kr) is a spherical Hankel function. So we can
write
u1 = Ajl (kr< ) + Bnl (kr< ),
r
jl (x) =
u2 = h1l (kr> ),
Is this correct? Note that the u2 solution is valid because it is bounded for large r:
i
(1)
lim hl (x) = eix (i)l .
x
11.6.2
Free Space Problem
4 Apr p8
We now take the special case where a = 0. The boundary condition

becomes the regularity condition at r = 0, which kills the nl (kr< )
solution.
(1)
(2)
The solutions ul and ul are then
(1)
ul
(2)
ul
4 Apr p9
= jl (kr),
(1)
= hl (kr).
(1)
Let be an arbitrary complex number. Note that since hl (x) =
11.6. EXAMPLE: CONSTANT PARAMETERS
179
jl (x) + inl (x), we have

(1)
W (jl (x), hl (x)) = iW (jl (x), nl (x)) =
i
.
x2
The last equality follows immediately if we evaluate the Wronskian for

large r using
jl (x) cos(x (l + 1)/2)
x ,
nl (x) sin(x (l + 1)/2)
x ,
and recall that W is a constant (for the general theory, c.f., problem
1, set 3) where is r2 for this problem. In particular we have
(1)
W (jl , hl ) =
i
.
(kr)2
We then get from equation 11.41

(1)
Gl
1 jl (kr< )hl (kr> )

= 2
i
r
k (kr)
2
=
ik
(1)
jl (kr< )hl (kr> ).
(11.42)
(eq11.21c
We have thus found the solution of

"
(x x0 )
,
2 G =
c
(11.43)
which is the fundamental three-space Greens function. We found

ik X m
G(x, x , ) =
Y (, )jl (kr< )hl (kr> )Ylm ,
lm l
0
(eq11.22
4 Apr p10
(11.44)
which is a simple combination of equation 11.34 and 11.42. In the first eq11.22a
homework assignment we solve this by a different method to find an
explicit form for equation 11.43. Here we show something related. In
equation 11.43 we have G = G(|x x0 |) since V = 0 and and
180
constant. This gives translational and rotational invariance, which corresponds to isotropy and homogeneity of space. We solve by choosing
x = 0 so that
jl (kr< ) = j0 (kr) + 00 s
as r0 0. Only the l = 0 term survives, since as r0 0 we have
l=0
jl (0) = 1,
l 6= 0
jl (0) = 0.
Thus we have
G(x, x0 ; ) =
ik 0 2
|Y | h0 (kr).
0
Since l = 0 impliesR m = 0, we get Y00 = const. = 1/4 since it satisfies

the normalization d|Y |2 = 1. We also know that
i
(1)
h0 (x) = eix .
x
This gives us
G(x x0 ; ) =
1 ikr
e .
4r
We may thus conclude that

0
eik|xx |
.
G(|x x |) =
4|x x0 |
0
(11.45)
eq11.25
stuff omitted
11.7
Summary
1. The form of the linear operator in spherical coordinates is

1
L0 = 2
r2 (r)
r r
r
where
L
=
sin
sin
(r)
L + V (r),
r2
1 2
.
sin2 2
4 Apr p11
11.8. REFERENCES
181
2. L is hermitian.
3. The eigenvalue equations for Ylm are
L Ylm (, ) = l(l + 1)Ylm (, ),
2 m
Y (, ) = m2 Ylm (, )
2 l
m = 0, 1, 2, . . . .
4. The partial wave expansion for the Greens function is

G(x, x0 ; ) =
X
l
X
Ylm (, )Glm (r, r0 ; 0 )Ylm (, ).
l=0 m=l
5. The Greens function for the free spce problem is

0
eik|xx |
.
G(|x x |) =
4|x x0 |
0
11.8
References
The preferred special functions reference for physicists seems to be

[Jackson75]. Another good source is [Arfken85].
This material is developed by example in [Fetter81].
182
Chapter 12
Steady State Scattering
Chapter Goals:
Find the free space Greens function outside a circle
of radius a due to point source.
Find the free space Greens function in one, two,
and three dimensions.
Describe scattering from a cylinder.
12.1
Spherical Waves
6 Apr p1
We now look at the important problem of steady state scattering.

Consider a point source at x0 with sinusoidal time dependence
pr:sssc1
f (x0 , t) = (x x0 )eit ,
whose radiated wave encounters an obstacle, as shown in figure 12.1
fig11.3
We saw in chapter 1 that the steady state response for free space
with a point source at x0 satisfies
"
2
L0 2 u0 (x, t) = (x x0 )eit ,
t
#
and was solved in terms of the Greens function,

u0 (x, ) = G0 (x, x0 , = 2 + i)eit
183
(12.1)
184
CHAPTER 12. STEADY STATE SCATTERING

x

A

A

AA
@
@

@

"
!
&
%
$

s
x0
$
s
'
#

!
%
Figure 12.1: Waves scattering from an obstacle.

0
eik|xx |iwt
=
4 |x x0 |
eq11.25b
pr:uoxo1
where the Greens function G0 satisfies the equation 11.43 and the
second equality follows from 11.45. The equation for G0 can be written
[2 k 2 ]G0 (x, x0 ; ) =
eq11.25c
6 Apr p3
(12.2)
1
(x x0 )
(12.3)
with the definition k = /c2 = /c (Remember that = 2 + i).

We combine these observations to get
u0 (x, x0 ; ) =
1
0
ei(t(xx )/c) .
0
4 |x x |
If there is an obstacle (i.e., interaction), then we have a new steady

state response
u(x, x0 ; ) = G(x, x0 ; = 2 + i)eit ,
where
[L0 ]G(x, x0 ; ) = (x x0 )
RBC.
(12.4)
This is the steady state solution for all time. We used u0 and G0 for
the free space problem, and u and G for the case with a boundary.
Note that equation 12.4 reduces to 12.2 if there is no interaction. The
scattered part of the wave is
GS eit = (G G0 )eit .
pr:GS1
6 Apr p4
12.2. PLANE WAVES
12.2
185
Plane Waves
We now look at the special case of an incident plane wave instead of

an incident spherical wave. This case is more common. Note that once
we solve the point source problem, we can also solve the plane wave
problem, since plane waves may be decomposed into spherical waves.
An incident plane wave has the form
0 = ei(wtkx)
where k = (/c)
n. This is a solution of the homogeneous wave equation pr:Phi0
"
1 2
2 + 2 2 0 = 0.
c t
#
So 0 is the solution to the equation without scattering.

Let be the wave when an obstacle is present,
"
1 2
+ 2 2 = 0
c t
#
which solves the homogeneous wave equation and the regular boundary
condition at the surface of the obstacle.
To obtain this plane wave problem, we let x0 go to . We now
describe this process. To obtain the situation of a plane wave approaching the origin from
z , we let x0 = r
z , as r0 . We want to find
out what effect this limit has on the plane wave solution we obtained
in equation 11.45,
0
eik|xx |
G0 =
.
4 |x x0 |
We define the angles and as shown in figure 12.2. From the figure pr:gamma1
we see that
fig12a
|x x0 | = r0 + r cos = r0 r cos ,
|x0 | .
We further recall that the dot product of unit vectors is equal to the
cosine of the separation angle,
x x0
cos =
= cos .
rr0
186

x
1

x x
r

x0
0
0
r
origin r cos
Figure 12.2: Definition of and ..
- z
So the solution is
0
eik|xx |
4 |x x0 |
1 ik(rr cos )
=
e
4 r0
1 ikr0 ikx(x0 )
=
e e
4 r0
0
eikr ikx
e
=
4 r0
G0 (x, x0 ; ) =
x0 ). Note that we have used the first two terms of the

where k = wc (
approximation in the exponent, but only the first term in the denominator. Thus we see that in this limit the spherical wave u0 in free space
due to a point source is
lim u0 = G0 eit
|x0 |
eikr
0
4r0
where
0 = ei(tkx) .
12.3
Relation to Potential Theory
pr:PotThy1
8 Apr p1
Consider the problem of finding the steady state response due to a

point source with frequency located at x0 outside a circular region of
radius a. The steady state response must satisfy the regular boundary
condition
"
#
G + a G = 0 for r = a.
(12.5)
r
12.3. RELATION TO POTENTIAL THEORY
187
In particular we want to find this free space Greens function outside a eq11A1
circle of radius a, where V = 0 and and are constant.
The Greens function G satisfies the inhomogeneous wave equation
[2 k 2 ]G(x, x0 ) =
(x x0 )
where k 2 = / = /c2 . The solution was found to be (from problem

1 of the final exam of last quarter)
G=
1 X
0
(1)
(1)
eim( ) [Jm (kr< ) + Xm Hm
(kr< )]Hm
(kr> ). (12.6)
4 m=
with
eq11A3
0
[kJm
(ka) Ka Jm (ka)]
.
(12.7) pr:Xm1
(1)
(1)
[kHm (ka) Ka Hm (ka)]
(1)
Note
that
H
(kr
)
comes
from
taking
Im
> 0. If we consider (Eq.BE)
>
m
(2)
Im < 0, we would have Hm
instead. All the physics is in the
functions Xm . Note that from this solution we can obtain the solution 8 Apr p2
to the free space problem (having no boundary circle). Our boundary
condition is then then G must be regular at |x| = 0:
Xm =
|x| = 0.
G regular,
(12.8)
(That is, equation 12.5 becomes 12.8.) How do we get this full space eq11A2
solution? From our solution, let a go to zero. So Xm in equation 12.6
goes the zero as a goes to zero, and by definition G G0 . The free
space Greens function is then
G0 =
X
i
0
(1)
eim( ) Jm (kr< )Hm
(kr> ).
4 m=
(12.9)
This is the 2-dimensional analog of what we did in three dimensions. eq11A4

We use this to derive the plane wave expression in 3-dimensions.
We may now obtain an alternative expression for the free space
Greens function in two dimensions by shifting the origin. In particular,
we place the origin at x0 . This gives us r0 = 0, for which
(
Jm (kr
(1)
)Hm
(kr)|r0 =0
(1)
H0 (kr) m = 0
0
else.
188
Thus equation 12.9 reduces to

G0 (r) =
i
(1)
H (kr),
4 0
which can also be written as

G0 (|x x0 |) =
eq11A5
i
(1)
H (k|x0 x|).
4 0
(12.10)
This is the expression for the two dimensional free space Greens function. In the process of obtaining it, we have proven the Hankel function
addition formula
(1)
H0 (k|x0 x|) =
(1)
eim( ) Jm (kr< )Hm
(kr> ).
m=
We now have the free space Greens functions for one, two, and
three dimensions:
i ik|xx0 |
e
,
2k
i
(1)
0
H0 (k|x0 x|),
G2D
0 (|x x |) =
4
0
eik|xx |
3D
0
G0 (|x x |) =
.
4 |x x0 |
0
G1D
0 (|x x |) =
eq11A6
pr:fsp2
8 Apr 5
(12.11)
We can interpret these free space Greens functions physically as follows. The one dimensional Greens function is the response due to a
plane source, for which waves go off in both directions. The two dimensional Greens function is the cylindrical wave from a line source. The
three dimensional Greens functions is the spherical wave from a point
source. Note that if we let k 0 in each case, we have
0
eik|xx | = 1 + ik|x x0 |,
and thus we recover the correct potential respectively for a sheet of
charge, a line charge, and a point charge.
12.4. SCATTERING FROM A CYLINDER
12.4
189
Scattering from a Cylinder
We consider again the Greens function for scattering from a cylinder,

equation 12.6
G=
1 X
0
(1)
(1)
(kr< )]Hm
(kr> ). (12.12)
4 m=
(1)
What is the physical meaning of [Jm (kr< ) + Xm Hm
(kr< )] in this equa- eq11A6b
tion? This is gives the field due to a point source exterior to the cylinder:
u = G(x, x0 , = 2 + i)eit
= G
eit + (G G0 )eit
| 0
{z
u0
{z
us
(12.13)
Note that Xm contains the physics of the boundary condition,

"
eq11A6c
+ a G = 0 for r = a.
r
From equations 12.12 and 12.13 we identify the scattered part of the
solution, us , as
us =
eit X
0
(1)
(1)
eim( ) Xm Hm
(kr)Hm
(kr0 ).
4 m=0
(12.14)
(1)
So we have expanded the total scattered wave in terms of Hm
, where eq11A7
Xm gives the mth amplitude. Why are the r> and r< in equation 12.11
but not in equation 12.14? Because there is a singular point at x = x0
in equation 12.11, but us will never have a singularity at x = x0 .
(Eq.j)
Now consider the more general case of spherical symmetry: V (r),
(r), (r). If these parameters are constant at large distances,
V (r) = 0,
(r) = = constant,
(r) = = constant,
190
then at large distances G must have the form of equation 12.11,

0
eik|xx |
G0 (|x x |) =
.
4 |x x0 |
0
We shall see that this formula is basic solution form of quantum mechanical scattering.
12.5
Summary
1. The free space Greens function outside a circle of radius a due

to point source is
G=
1 X
0
(1)
(1)
(kr< )]Hm
(kr> ).
4 m=
with
Xm =
0
[kJm
(ka) Ka Jm (ka)]
(1)
(1)
[kHm (ka) Ka Hm (ka)]
2. The free space Greens function in one, two, and three dimensions
is
i ik|xx0 |
e
,
2k
i
(1)
0
G2D
H0 (k|x0 x|),
0 (|x x |) =
4
0
eik|xx |
0
G3D
(|x
x
|)
=
.
0
4 |x x0 |
0
G1D
0 (|x x |) =
3. For the problem of scattering from a cylinder, the total response

u is easily decomposed into an incident part a scattered part us ,
where us contains the coefficient Xl .
12.6
References
See any old nuclear of high energy physics text, such as [Perkins87].
Chapter 13
Kirchhoff s Formula
As a further application of Greens functions to steady state problems,
let us derive Kirchhoffs formula for diffraction through an aperture. pr:Kirch1
Suppose we have a point source of sound waves of frequency at some
point x0 in the left half plane and at x = 0 we have a plane with a hole.
We want to find the diffracted wave at a point x in the right-half plane. fig19a
If we look at the screen directly, we see the aperture is the yz-plane
at x = 0 with a hole of shape 0 as shown in the figure. The solution
G(x, x0 ; ) satisfies the equations
[2 + k 2 ]G(x, x0 ; ) = (x x0 ),

G
= 0.

x x=00
x6
We will reformulate this problem as an integral equation. This

integral equation will have as a kernel the solution in the absence of
the hole due to a point source at x0 in the R.H.P. This kernal is the
free space Greens function, G0 (x, x0 ; ), which satisfies the equations
[2 + k 2 ]G0 (x, x0 ; ) = (x x0 ),

G0

x
= 0.
x=0
all y,z
This we solve by the method of images. The boundary condition is pr:MethIm2

191
192
CHAPTER 13. KIRCHHOFFS FORMULA
R
x
x0
Figure 13.1: A screen with a hole in it.

satisfied by adding an image source at x0 , as shown in figure 13.2 Thus
the Greens function for this boundary value problem is
0
1 eik|xx | eik|xx |
G(x, x ) =
+
4 |x x0 |
|x x0 |
"
Now by taking L0 = (2 + k 2 ) we may apply Greens second identity

Z
xR
(S L0 u uL0 S ) =
dS n
[uS S u]
xS
with u = G(x, x0 ) and S = G0 (x, x0 ) where R is the region x > 0 and

S is the yz-plane. This gives us
L0 u = 0
for x > 0,

u

= 0,
x x=0
x6
and
L0 S = (x x0 ),
S

= 0.
x x=0

193
x
*

x0 x0
v
v
x0
v
x0
x=0
Figure 13.2: The source and image source.
These identities allow us to rewrite Greens second identity as
dxG(x, x0 )(x x ) =
dy dzG0 (x, x ) G(x, x0 )

x
0
x=0
and therefore
0
G(x , x0 ) =
G(x, x0 )
.
dy dzG0 (x, x )

x
x=0
0
(13.1)
Thus the knowledge of the disturbance, i.e., the normal component of eq19a
the velocity at the aperture, determines the disturbance at an arbitrary
G at the
point x in the right half plane. We have then only to know x
aperture to know G everywhere.
Furthermore, if x0 approaches the aperture, equation 13.1 becomes
an integral equation for G for which we can develop approximation
methods.
G(x0 , x0 ) =
Z
x 0
dy dzG0 (x, x0 )
G(x, x0 ).
x
(13.2)
The configurations for the different Gs are shown in figure 13.3. Note eq19b
0
G0 (x, x )|x0 =0
1 eik|xx |
=
,
2 |x x0 |
and equation 13.2 becomes

0

1 Z
eik|xx |

G(x , x0 ) =
dy dz
G(x,
x
)
.
0
0

0
2 x
|x x | x
x=0
0
fig19b
194
CHAPTER 13. KIRCHHOFFS FORMULA
x0
x0
G(x0 , x0 )
G0 (x, x0 )
G(x, x0 )
Figure 13.3: Configurations for the Gs.

Now suppose that the size a of the aperture is much larger than
the wavelength = 2/k of the disturbance which determines the
distance scale. In this case we expect that the wave in the aperture
does not differ much from the undisturbed wave except for within a
few wavelengths near the aperture. Thus for ka 1 we can write

0
0

1 Z
eik|xx |
eik|xx |

dy dz
G(x, x0 )

,
G(x, x0 )

2 x0
|x x0 | x
x 4|x x0 | x=0
x=0
where we have used the substitution

0
eik|xx |
G(x, x0 )
=

.

4|x x0 |
x
x
x=0
x=0

fig19c
pr:Huyg1
This equation then gives us an explicit expression for G(x0 , x0 ) in terms

of propagation from the source at x to the field point x0 of the velocity
0
eik|xx |
disturbance at x of the velocity distribution x
produced by free
4|xx0 |
propagation to x from the point x0 of the disturbance. This yields Huygens principle and other results of physical optics (Babenets principle,
etc.).
13.1
References
See [Fetter80, pp327332] for a discussion of these results.
Chapter 14
Quantum Mechanics
Chapter Goals:
State the Greens function equation for the inhomogeneous Schrodinger equation.
State the Greens function for a bound-state spectra in terms of eigen wave functions.
State the correspondence between classical wave
theory and quantum particle theory.
The Schrodinger equation is
"
H i
h
(x, t) = 0
t
where the Hamiltonian H is given by

H=
h
2 2
+ V (x).
2m
This is identical to our original equation, with the substitutions =

h
2 /2m, L0 = H, and in the steady state case = E. For the free space
problem, we require that the wave function be a regular function.
The expression |(x, t)|2 is the probability given by the probability
amplitude (x, t). For the time dependent Schrodinger equation, we
have the same form as the heat equation, with cp i
h. In making the
195
8 Apr p8
pr:QM1
196
CHAPTER 14. QUANTUM MECHANICS

V (x)
E>0
E
E
EE
D
D
D
E<0

C
CC
B
B

A

QQ
Figure 14.1: An attractive potential.
pr:sss3
transition from classical mechanics to quantum mechanics, we use H =

p2 /2m + V (x) with the substitution p (
h/i); this correspondence
for momentum means that the better we know the position, and thus
the more sharply the wave function falls off, the worse we know the
subsequent position. This is the essence of the uncertainty principle.
We now look at the steady state form. Steady state solutions will
be of the form
(x, t) = eit (x),
where (x) satisfies the equation
[H h
] (x) = 0.
pr:enLev1
8 Apr p9
eq13.1
fig13.1
The allowed energy levels for h

are the eigen values E of H:
H = E.
(14.1)
So the allowed frequencies are = E/

h, for energy eigenvalues E. The
energy spectrum can be either discrete or continuous. Consider the
potential shown in figure 14.1. For E = En < V (0), the energy levels
are discrete and there are a finite number of such levels; for E > V (0),
the energy spectrum is continuous: any energy above V (0) is allowed.
A plot of the complex energy plane for this potential is shown in figure
14.2. The important features are that the discrete energies appear as
poles on the negative real axis, and the continuous energies appear as a
14.1. QUANTUM MECHANICAL SCATTERING

E-plane
u
u uu
197
Im E
Re E
Figure 14.2: The complex energy plane.

branch cut on the positive real axis. Note that where as in this problem
there are a finite number of discrete levels, for the coulomb potential
there are instead an infinite number of discrete levels. If, on the other
hand, we had a repulsive potential, then there would be no discrete
spectrum.
The Greens function solves the Schrodinger equation with an inhomogeneous -function term:
8 Apr p10
(H E)G(x, x0 ; E) = (x x0 )
(14.2)
where E is a complex variable. The boundary condition of the Greens eq13.2

function for the free space problem is that it be a regular solution. Unlike previously considered problems, in the quantum mechanical problems, the effect of a boundary, (e.g., the surface of a hard sphere) is
enforced by an appropriate choice of the potential (e.g., V = for
r > a). Once we have obtained the Greens function, we can look at its
energy spectrum to obtain the s and En s, using the formula obtained
in chapter 4:
X n (x)n (x0 )
.
G(x, x; E) =
En E
n
This formula relates the solution of equation 14.1 to the solution of
equation 14.2.
fig13.2
11 Apr p1
14.1
Quantum Mechanical Scattering
11 Apr p2
11 Apr p3
We now look at the continuum case. This corresponds to the prob- pr:QMS1
lem of scattering. We use the Greens function to solve the problem of
198
quantum mechanical scattering. In the process of doing this, we will

see that the quantum mechanics is mathematically equivalent to the
classical mechanics of waves. Both situations involve scattering. The
solution u for the classical wave problem is interpreted as a velocity
potential, whereas the solution for the quantum mechanical problem
is interpreted as the probability amplitude. For the classical wave problem, u2 is interpreted as intensity, while for quantum mechanics, ||2 is
interpreted as probability density. Thus, although the mathematics for
these problems is similar, the general difference is in the interpretation.
The case of quantum mechanical scattering is similar to classical
scattering, so we consider classical scattering first. We use the Greens
function in classical wave theory,
[L0 ]G(x, x; ) = (x x0 ),
with = 2 + i for causality, to obtain the steady state response due
to a point source,
u = eit G(x, x; = 2 + i).
This steady state solution solves the time dependent classical wave
equation,
#
"
2
(14.3)
L0 + 2 u(x, t) = (x x0 ).
t
eq13.3
We may decompose the solution u into two parts,

u = u0 + uscat ,
where
u0 = eit G0 =
ei(tkR)
4 R
where R = |x x0 |, and
uscat = eit [G G0 ].
pr:uscat1
Note that u0 is the steady state solution for a point source at x = x0 ,

solution for a point source at x, G0 is the solution for the free case,
and us is the solution for outgoing scattered waves. Note also that the
outgoing scattered waves have no singularity at x = x0 .
14.2. PLANE WAVE APPROXIMATION
14.2
199
Plane Wave Approximation

pr:PlWv1
If one solves the problem of the scattering of the spherical wave from 11 Apr p4
a point source, that is, for the Greens function problem, then we also
have the solution for scattering from a plane wave, merely by letting
|x0 | . We now define (x, t) as the solution of equation 14.3 for
the special case in which x0
z,
"
2
L0 + 2 (x, t) = 0.
t
#
(14.4)
eq13.4
For the steady state solution
11 Apr p5
(x, t) = eit (x, ),

we get the equation
[L0 2 ](x, ) = 0.
How is this equation solved for positive ? There isnt a unique solution
for this, just like there wasnt a unique solution for the Greens function.
We have already found a solution to this equation by considering the
Greens function with the source point going to infinity in the above
method, but it does not satisfy the boundary condition appropriate for 11 Apr p6
scattering.
In order to determine the unique scattering solution, we must introduce a new boundary condition appropriate for scattering. This is
necessary because although our previous equation
[L0 n ]un = 0
(14.5)
with RBC gave the eigenfunctions, it does not give unique physical eq13abc
solutions for the case of scattering.
To find the un s in equation 14.5 we would extract them from the
Greens function using, for the discrete case,
0
G(x, x ; ) =
X un un
n
200
or for the continuum case,

1
[G(x, x0 ; = n + i) G(x, x0 ; = n i)]
2i
1
=
Im [G(x, x0 ; = n + i)].
un =
But these un s are not solutions corresponding to the scattering boundary condition, since they contain both incoming and outgoing waves.
14.3
Quantum Mechanics
11 Apr p7
We now apply what we have said for particles to the case of quantum
mechanics. In this case the steady state solutions are of the form
= ei(E/h)t 0 .
We want the total wave to be a superposition of an incident plane wave
and a scattered wave.
= eikxit + s
11 Apr p9
where = E/
h. Note that eikx corresponds to an incident plane wave,
and s corresponds to outgoing waves. To get this form, we use the
Greens function for the free space problem (no potential)
G0 =
eikR
m eikR
=
,
4R
2
h2 R
where
k2 =
E
2mE
p2
= 2
=
=
.
h
/2m
h
2
h
2
Take the limit |x0 | , the free space Greens function becomes
ei(E/h)t G0 (x, x0 ; E) = 0
11 Apr p9
where 0 = eikxi(E/h)t .
m eikr
,
2
h2 r
14.4. REVIEW
14.4
13 Apr p1
201
Review
We have been considering the steady state response problem

"
2
L0 + 2 u(x, t) = (x x0 )eit .
t
#
The steady state response for outgoing waves (i.e., that which satisfies
the boundary condition for scattering) is
u(x, t) = eit G(x, x0 , = 2 + i)
(14.6)
where G solves
eq14.g
[L0 ]G(x, x0 ; ) = (x x0 ).
We want to get the response (x, t) for scattering from a plane
wave. We only need to let |x0 | go to infinity:
0
eikr
lim
u(x
,
t)
=
(x, t).
|x0 |
4r0
0
This gives scattering from a plane wave. is the solution of

2
](x, t) = 0
t2
which satisfies the boundary condition of scattering. For the case of
steady state response we can write
[L0 +
(x, t) = eit (x, )

where (x, ) solves the equation
[L0 2 ](x, ) = 0.
This equation satisfies the boundary condition of scattering:
(x, ) = eikx + s (x),
where s (x) has only outgoing waves. Note that
2
2
= 2 and k = k
n,
c
where = limr (r) and = limr (r).
k2 =
13 Apr p2
202
14.5
Spherical Symmetry Degeneracy
We now compare the mathematics of the plane wave solution with

that of the eigen function u. The eigenvalue equation can be written
[L0 2 ]u (x, 2 ) = 0.
In this equation u is a positive frequency eigen function with degeneracy number and eigen value 2 . The eigen functions can be obtained
directly from the Greens function by using
X
1
Im G(x, x0 ; = 2 + i) =
u (x, 2 )u (x0 , 2 ).
For the case of spherical symmetry we have

u (x, 2 ) = ulm (x, 2 )
= Ylm (, )ul (r, 2 )
13 Apr p3
where l = 0, 1, . . . and m = l, . . . , 0, . . . , l. The eigenvalues 2 are

continuous: 0 < 2 < . Because of the degeneracy, a solution of the
differential equation may be any linear combination of the degenerate
eigen functions:
X
(x, ) =
c u ,
where the c s are arbitrary coefficients. This relates the eigen function
to the plane wave scattering solution. In the next chapter we will see
how the c s are to be chosen.
14.6
Comparison of Classical and Quantum
pr:ClasMech1
Mathematically we have seen that classical mechanics and quantum

mechanics are similar. Here we summarize the correspondences between
their interpretations.
classical wave theory
u = wave amplitude
quantum particle theory

= probability amplitude
14.6. COMPARISON OF CLASSICAL AND QUANTUM

u2 = energy density
n = natural frequencies
Normal modes:
un (x)ein t = un (x, t)
L0 un = n2 un
L0 is positive definite:
n2 > 0
203
||2 = probability density

En = energy eigenvalues
Stationary states
n (x)ei(En /h)t = n (x, t)
Hn (x) = En n (x)
H + = H:
En R
The scattering problem is like the eigen value problem but we look
at the region of continuous spectrum. For scattering, we require that
En > 0. In this continuum case, look at the eigen values from:
(H E) (x, E) = 0,
13 Apr p4
(14.7)
where labels wave functions with degenerate eigenvalues. Rather eq13.25

than a wave we have a beam of particles characterized by some energy
E. The substitution from classical mechanics to quantum mechanics is
as follows:
2 E,
h
2
,
2m
k2
2
E
2mE
p
=
=
2
2
h
h
/2m
h
This last equation is the De Broglie relation.

pr:DeBr1
Now we want to look at the solution for quantum mechanical scat- 13 Apr p5
tering using Greens functions. Suppose we have a beam of particles
coming in. This incident free wave has the form eikxiEt/h which solve
the free space hamiltonian
H0 =
h
2 2
.
2m
We want the solution to equation 14.7 which corresponds to scattering.

That is, we want the solution for
(H E)(x, E) = 0
which is of the form
(x, E) = ekx + s
where s has only outgoing waves.
204
To solve this we look at the Greens function.

(H E)G(x, x0 ; E) = (x x0 ) for ImE > 0,
with the appropriate boundary conditions. We make the substitution
(x, t) = ei(E/h)t G(x, x0 ; E + i).
13 Apr p6
This corresponds not to a beam of particles but rather to a source of

particles.
Stuff omitted
14.7
Summary
1. The Greens function equation for the inhomogeneous Schrodinger

equation is
(H E)G(x, x0 ; E) = (x x0 ),
where
H=
h
2 2
+ V (x).
2m
2. The Greens function for a bound-state spectra in terms of eigen

wave functions is
G(x, x; E) =
X n (x)n (x0 )
n
En E
3. There is a close connection between classical and quantum mechanics which is discussed in section 14.6.
14.8
References
See your favorite quantum mechanics text.
Chapter 15
Scattering in 3-Dim
Chapter Goals:
State the asymptotic form of the response function
due to scattering from a localized potential.
Derive the scattering amplitude for a far-field observer due to an incident plane wave.
Derive the far-field form of the scattering amplitude.
15 Apr p1
Define the differential cross section and write it in

terms of the scattering amplitude.
Derive and interpret the optical theorem.
Derive the total cross section for scattering from a
hard sphere in the high energy limit.
Describe the scattering of sound waves from an oscillating sphere.
We have seen that the steady state case reduces the Greens function
problem to the equation
[L0 ]G(x, x0 ; ) = (x x0 )
205
206
CHAPTER 15. SCATTERING IN 3-DIM
with RBC, where the linear operator is given by

L0 = (x) + V (x).
In the spherically symmetric case we have V (r), (r), and (r). In
chapter 11 we saw that the Greens function can be written as an expansion in terms of spherical harmonics,
G(x, x0 ; ) =
Ylm (, )Gl (r, r0 ; )Ylm (0 , 0 ).
lm
Only
true.
In the last chapter we saw how to solve for scattering from a point
source and scattering from a plane wave. We did this for a particular
case in the problem set. This all had nothing to due with spherical
partly symmetry.
Now consider the case of spherical symmetry. From chapter 3 we
know that the radial Greens function can be written
ul1 (r< , )ul2 (r> , )
0
Gl (r, r ; ) = 2
.
(15.1)
r (r)W (ul1 , ul2 )
eq14.0
The us solve the same radial eigenvalue equation

"
eq14.1
15 Apr p2
1 d
d
2
r2 (r)
r dr
dr
(r)l(l + 1)
+
+ V (r) (r) ul1,2 = 0, (15.2)
r2
but different boundary conditions. The eigenfunction u1 satisfies the

boundary condition
l
u ul1 = 0 for r = a
r 1
and as a 0 we replace this with the boundary condition ul1 (r) finite
at r = 0. The eigenfunction u2 satisfies the boundary condition u2
finite as r .
By comparing equation 15.2 with previous one-dimensional equations we have encountered, we identify the second and third terms as
an effective potential,
(r)l(l + 1)
+ V (r).
r2
In quantum mechanics we have
Veff =
pr:Veff1
Veff
h
2 l(l + 1)
=
+ V (r).
2mr2
15.1. ANGULAR MOMENTUM
15.1
207
Angular Momentum
The above spherical harmonic expansion for the Greens function was
obtained by solving the corresponding eigenfunction equation for the
angular part,
2 l(l + 1)Ylm .
h
2 L Ylm = h
The differential operator L can be related to angular momentum by
recalling that the square of the angular momentum operator satisfies
the equation
L2op Ylm = h
2 l(l + 1)Ylm .
Thus we identify h
2 L as L2op , the square of the angular momentum
operator.
h
2 L L2op .
In the central potential problem of classical mechanics it was found
that
L2
Veff =
+ V (r),
2mr2
where
pr:bfL1
L=xp
and
L2 = L L.
In quantum mechanics the momentum operator is p = (
h/i), so that
L = x p Lop =
x
i
and thus
!
L2op
x
x
i
i
= h
2 L .
This gives the relation between angular momentum in classical mechanics and quantum mechanics.
208

'
#

A

A

AA
@
@

@

"
!
&
%
$

s
x0
$
s

!
%
Figure 15.1: The schematic representation of a scattering experiment.
15.2
pr:ExpScat1
fig14a
15 Apr p3
Far-Field Limit
We now take the far field limit, in which r , meaning the field
is measured far from the obstacle. This situation is accurate for experimental scattering measurements and is shown in figure 15.1. We
assume that in this r limit, we have (r) , (r) , and
rV (r) 0. If instead the potential went as V (r) = /r, e.g., the
Coulomb potential, then our analysis
would change somewhat. We will
q
also use the wave number k = / , where = 2 + i classically,
and = E + i for the quantum case. In classical mechanics we then
have k = /c and in quantum mechanics we have k = p/
h.
Our incident wave is from a point source, but by taking the sourceto-target separation r0 big, we have a plane wave approximation. After
making these approximations, equation 15.2 becomes
"
eq14.2
1 d
d
2
r2
r dr
dr
l(l + 1)
+
k 2 ul1,2 = 0.
r2
(15.3)
If we neglect the term l(l + 1)/r2 , compared to k 2 we would need kr

l, which we dont want. Instead we keep this term in order to keep
conventional solutions. In fact, it will prove easier to keep it, even
though it may vanish faster than V (r) as r , and since we also
have to consider l large, we dont want to kill it. In this limit the
Greens function is proportional to a product of the us,
lim Gl (r, r0 ; ) = Aul1 (r0 , )ul2 (r, ).
(15.4)
15.2. FAR-FIELD LIMIT
209
We assume the the point source is not in the region where things are eq14.3
really happening, but far away. In this case V (r0 ) = 0, (r0 ) = , and
(r0 ) = , for large r0 . Thus we are looking at the far field solution
where the point source is outside the region of interaction.
15 Apr p4
We already know the explicit asymptotic solution to the radial equation:
(1)
ul2 (r> ) hl (kr> )

ul1 (r< )
(1)
jl (kr< )
for kr> 1,
+
(1)
Xl hl (kr< )
(15.5)
for kr< 1.
(15.6)
Xl contains all the physics, which arises due to the boundary condition. eq14.5
In general, Xl must be evaluated numerically. For specific cases such as
in the problem set, V (r) = 0 so equation 15.3 is valid everywhere, and
thus we may obtain Xl explicitly. For our present situation we have
assumed the far field approximation and an interaction-free source, for
which the asymptotic form of the Greens function may be written in
terms of equation 15.5 and 15.6 as
(1)
(1)
lim Gl (r, r0 ; ) = A[jl (kr0 ) + Xl hl (kr0 )]hl (kr).
We have not yet specified r > r0 , only that r 1 and r0 1.

To obtain the scattered wave at large r, we look at G G0 . In
particular we will evaluate Gl Gl0 . So we look at Gl0 (r, r0 ; ) for
r r0 . Recall that Gl0 has the form
15 Apr p5
Gl0 (r, r0 ; ) =
ik
(1)
jl (kr< )hl (kr> ).
This is for the free problem; it solves all the way to the origin. Now
take the difference.
!
ik
(1)
(1)
(1)
Gl Gl0 = A
jl (kr< )hl (kr> ) + AXl hl (kr< )hl (kr> ).
This equation assumes only that we are out of the range of of interac(1)
tion. The term hl (kr> ) gives the discontinuity on dG/dr at r = r0 .
We now assert that A must equal ik/ because the scattering wave What does this
mean?
Gl Gl0 has to be nonsingular. Now look at the case r > r0 :
15 Apr p6
pr:scatWv1
210

Gl Gl0 =
ik
(1)
(1)
Xl hl (kr0 )hl (kr).
All that is left to see is what the scattered wave looks like. We take kr
to be large, as in the problem set.
To solve this equation not using Greens function, we first look for
a solution of the homogeneous problem which at large distances gives
scattered plus incident waves.
=
cm um (r, ).
m=
15 Apr p7
At large distances we have

eixt + outgoing waves.
18 Apr p1
18 Apr p2
15.3
Relation to the General Propagation

Problem
We could instead consider the general problem of propagation, but at

this time we are just considering the case of scattering, for which the
source lies in a homogeneous region where V (r) = 0 and and are
constant. The propagation problem is more general because it allows
the source to be anywhere.
15.4
pr:scGF1
Simplification of Scattering Problem
For the scattering problem, we are considering a beam of particles from

a distant (r0 1) point source in a homogeneous medium incident on
a target, which scatter and are detected by detectors far away (r 1).
This latter condition is called the far field condition. In this case we
have seen that the problem can be simplified, and that we may explicitly
calculate the scattered Greens function GlS (r, r0 ; ):
GlS = Gl (r, r0 , ) Gl0 (r, r0 , ) =
ik
Xl h1l (kr)h1l (kr0 ),
c
15.5. SCATTERING AMPLITUDE
211
with k = / = /c, where c is the speed. The value of Xl

depends on k and is obtained from the behavior of u1 at large r = |x|.
We already know that ul1 (r) must be of the form
18 Apr p3
lim ul1 (r) = jl (kr) + Xl (k)h1l (kr),
since this is the asymptotic form of the solution of the differential equation. Thus scattering reduces to this form. All we need is Xl , which
is obtained from the behavior of ul1 (r). In particular, we dont need to
know anything about ul2 (r) if we are only interested in the scattering
problem, because at large distances it cancels out.
The large distance behavior of the function which satisfies the boundary condition at small distances is what determines the scattering solution. In this case r and r0 are both large enough that we are in
essentially a homogeneous region.
15.5
What does this

mean
Physically? Isnt U1l
for distances
less
than
rsource ? Dont
we need r <
r0 ?
Scattering Amplitude
pr:scAmp1
Consider the special problem where V = 0, = const., and = const.,

with the boundary condition
u1
+ ku1 = 0 for r = a.
r
In the problem set we found Xl by satisfying this condition. The result
was
[kj 0 (ka) jl (ka)]
Xl = 1l0
.
[khl (ka) h1l (ka)]
This equation is valid for r > a.
18 Apr p4
Given Xl we can calculate the difference, Gl Gl0 so we can calculate
GlS . Thus we can determine the scattered wave, which we now do.
We calculate the scattered piece by recalling the expansion in terms
of spherical harmonics,
GS = G G0 =
Ylm (, )[Gl (r, r0 ; ) Gl0 (r, r0 ; )]Ylm (0 , 0 ).
l,m
(15.7)
We substitute into this the radial part of the the scattered Greens eq14.6
212

x
1

x x
r

x0
0
0
- z
r
origin r cos
Figure 15.2: The geometry defining and .
function,
0
G(r, r ; ) G0 (r, r ; ) =
X
ik
l=0
eq14.7
(1)
(1)
Xl (k)hl (kr)hl (kr0 ),
(15.8)
and the spherical harmonics addition formula
18 Apr p5
l
X
pr:addForm1
Ylm (, )Ylm (0 , 0 ) =
m=l
2l + 1
Pl (cos )
4
(1)l
=
(2l + 1)Pl (cos )
4
eq14.8
fig14b
(15.9)
where cos = x x0 . The geometry is shown in figure 15.2. The result

of plugging equations 15.8 and 15.9 into equation 15.7 is
GS = G G0 =
X
ik
l=0
15.6
l
1
1
0 (1)
Xl (k)hl (kr)hl (kr )
Pl (cos )(2l
+ 1).
Kinematics of Scattered Waves
We take the limit kr to get the far field behavior. In the asymptotic limit, the spherical Hankel function becomes
i
x
h(1) (x) (i)l eix .
x
Thus in the far field limit the scattered Greens function becomes
G G0
18 Apr p6
ik eikr (i) X
Xk h1l (kr0 )(i)l (2l + 1)Pl (cos ).
kr 4 l=0
This in the case for a detector very far away. We can write also this as
15.7. PLANE WAVE SCATTERING

G G0 =
213
eikr
f (, r0 , k),
r
where
1 X
f(, r0 , k) =
(2l + 1)(i)l Pl (cos )Xl (k)h1l (kr0 ).
4 l=0
(15.10)
This is an independent proof that the scattered Greens function, G eq14.9

G0 , is precisely an outgoing wave with amplitude f.
The scattered part of the solution for the steady state problem is
given by
us = eit (G G0 ).
The energy scattered per unit time per unit solid angle will be proportional to the energy per unit area, which is the energy flux u2 . This in
turn is proportional to the scattering amplitude f. Thus
dE
|f |2 .
dtd
We know the radial differential ds = r2 dr of the volume dV = dsd
for a spherical shell, so that we get
18 Apr p7
dE
dE ds
=
dtd
dtds dr
ds
dE
dE
= r12 and dr
= r2 , so that dtd
is
Note that dimensionally we have dtds
dimensionless. Thus it is the 1/r term in the scattered spherical wave
which assures conservation of energy.
pr:ConsE1
15.7
Plane Wave Scattering
We now look at scattering from a plane wave. Let r0 = |x0 | go to

infinity. This gives us
|kr0 |
0
(1)
0 r
l+1 e
hl (kr ) (i)
.
kr0
In this limit equation 15.10 becomes

0
e|kr |
0
f (, r , k)
f (, k)
4 r0
214
where
eq14fth
pr:ftrk1

0
ekr
iX
f
(,
k)
=
(2l + 1)Pl (cos )Xl .

r0
k l=0
18 Apr p8
(15.11)
f (, k) is called the scattering amplitude for a field observer from an

incident plane wave. We can now compute the total wave for the far
field limit with incident plane wave. It is
u = eit G = eit [G G0 + G0 ]
!
!
ikr0
ikr
e
e
= eit
eikx +
f .
4 r0
r
In this equation the term eikx corresponds to a plane wave and the
ikr
term e r f corresponds to an outgoing scattered wave. So
|f 2 |
dE
dtd
This is a problem in the problem set.
15.8
Special Cases
20 Apr p1
20 Apr p3
So far we have considered the case in which all the physics occurs within
some region of space, outside of which we have essentially free space.
We thus require that in the area exterior to the region, V0 = 0, =
constant, and = constant. The source emits waves at x0 , and we want
to find the wave amplitude at x. Note that for the Coulomb potential,
we have no free space, but we may instead establish a distance after
which we may ignore the potential.
15.8.1
Homogeneous Source; Inhomogeneous Observer
In this case x0 is in a region where V (r) 0, and and are constant.

We define u0 to be the steady state solution to the point source problem
without a scatterer present, i.e., u0 is the free space solution.
u0 = eit G0 ,
15.8. SPECIAL CASES
215
where
G0 =
eik|xx |
.
4 |x x0 |
Further we define the scattered solution

us u u0 .
To find us we use equation 15.1 to get the spherical wave expansion
G0l =
ik
(1)
jl (kr< )hl (kr> ).
Thus
us = eit (G G0 ).
For the case of a homogeneous source and an inhomogeneous observer
r> = r0 , r< = r = 0. We take
(l)
(1)
u2 (r0 ) = hl (kr0 ).
Remember that r0 is outside the region of scattering, so ul2 solves the 20 Apr p4
free space equation, 15.3,
"
1 d
d
2
r2
r dr
dr
l(l + 1)
+
k 2 u2 = 0,
r2
(15.12)
where k 2 = / with the condition ul2 (r) finite as r . The general eq14.10
solution to equation 15.12 is
(l)
(1)
u2 (r0 ) = hl (kr0 ).
We still need to solve the full problem for u1 with the total effective
potential V (r) 6= 0.
15.8.2
Homogeneous Observer; Inhomogeneous Source
In this case the source point is in the interior region. We want to find 20 Apr p5
u for x inside the medium, but we cannot use the us method as we did
in case 1. The reason why it is not reasonable to separate u0 and uS in
this case is because the source is still inside the scattering region.
216
We replace r> r and r< r0 so that

(l)
(1)
u2 (r) hl (kr)
(l)
and u1 (r) satisfies the full potential problem So once again we only
(l)
need to solve for u1 (r). The physics looks the same in case 1 and
case 2, and the solutions in these two cases are reciprocal. This is a
manifestation of Greens reciprocity principle. The case of a field inside
due to a source outside looks like the case of a field outside due to a
source inside.
15.8.3
Homogeneous Source; Homogeneous Observer
For this case both points are in exterior region. By explicitly taking
|x| > |x0 | we make this a special case of the previous case. Thus we
have r> r and r< r0 . Now both u1 and u2 satisfy the reduced
ordinary differential equation
"
eq14rad
d
1 d
r2
2
r dr
dr
l(l + 1)
+
k 2 u1,2 = 0,
r2
(15.13)
where u1 satisfies the lower boundary condition and u2 satisfies the

upper boundary condition. As we have seen, the asymptotic solutions
to this equation are
(l)
(1)
u1 (r0 ) jl (kr0 ) + Xl hl (kr0 ).

(l)
u2 (r)
eq14.11,12
20 Apr p6
(15.14)
(1)
hl (kr).
(15.15)
To obtain Xl we must solve

"
1 d
d
2
r2 (r)
r dr
dr
l(l + 1)
+
+ V (r) (r) u1 = 0,
r2
and then take r 1. We can then get Xl (k) simply by comparing

equations 15.14 and 15.15. The scattered wave is then
us = eit (G G0 ),
where
ik it
(1)
(1)
e
Xl hl (kr)hl (kr0 ).
We see that the field at x is due to source waves u0 and scattered waves
uS .
Gl G0l
15.8. SPECIAL CASES
217
Homogeneous Source and Observer, Far Field

For this case the source and the field point are out of the region of
interaction. We take r > r0 and r0 .
For these values of r and r0 we have V = 0 and and constant.
In this case
0
eik|xx |it
it
it
e
Ge
G0 =
(15.16)
4 |x x0 |
G0 =
X
(2l + 1)
l=0
where
G0l
(Eq.f )
(1)l Pl (cos )G0l
ik
(1)
= jl (kr< )hl (kr> )
(15.17)
(Eq.g)
(15.18)
(Eq.h)
it
u=e
it
G=e
G0 + us
(15.19)
where
(Eq.i)
us = eit (G G0 )
ik it X
(2l + 1)(1)l
(1)
(1)
=
e
Xl
Pl (cos )hl (kr)hl (kr0 )
4
l=0
22 Apr p3
where for large r,

(1)
u1 jl (kr) + Xl hl (kr)
(15.20)
This is the large r behavior of the solution satisfying the small r bound- (Eq.k)
ary condition.
15.8.4
Both Points in Interior Region
We put x very far away, next to a detector. The assumption that x lies
in the vicinity of a detector implies kr 1. This allows us to make
the following simplification from case 2:
(1)
hl (kr) (i)l
Thus we can rewrite u1 . We have
(i) ikr
e .
kr
(15.21)
(Eq.m)
20 Apr p7
218

us = eit (G G0 )
(15.22)
and the simplification
(Eq.n)
(1)
Gl0 jl (kr0 )hl (kr).
(15.23)
(Eq.0)
22 Apr p1
15.8.5
Summary
Here is a summary of the cases we have looked at

case 4
cases 1, 2
case 3
22 Apr p4
need to know u1 , u2 everywhere

need to know u1 everywhere
need to know u1 at large r only
We now look at two more special cases.
15.8.6
Far Field Observation
Make a large r expansion (r ):

(1)
hl (kr) (i)l (
i ikr
)e
kr
(15.24)
(EQ.l)
0
ei(tk|xx |) ei(tkr)
u=
+
f (, r0 , k).
4 |x x0 |
r
(Eq.m)
The term
found
ei(tkr)
f (, r0 , k)
r
is explicitly just the outgoing wave. We
1 X
(1)
(2l + 1)(i)l Pl (cos )Xl hl (kr0 )
f(, r0 , k) =
4 l=0
(Eq.n)
22 Apr p5
(15.25)
(15.26)
The term f(, r0 , k) is called the scattering amplitude for a point source
at r0 . The flux of energy is proportional to f2 .
15.9. THE PHYSICAL SIGNIFICANCE OF XL
15.8.7
219
Distant Source: r0
Let the distance of thee source go to infinity. Define

k = k(
x0 )
(Eq.o)
(15.27)
and in f, let r0 . This gives us

0
eikr
ei(tkr)
i(tkx)
e
+
f (, k)
u
4r0
r
"
(15.28)
We can then get
(Eq.p)
i X
f=
(2l + 1)Pl (cos )Xl
k l=0
22 Apr p6
(15.29)
This equation is seen in quantum mechanics. f is called the scattering (Eq.q)

amplitude at angle , and does not depend on due to symmetry. The
basic idea is that plane waves come in, and a scattered wave goes out.
The wave number k comes from the incident plane wave.
22 Apr p7
15.9
The Physical significance of Xl
Recall that Xl is determined by the large distance behavior of the solu- pr:Xl1
tion which satisfies the short distance boundary condition. Xl is defined
by
(1)
(1)
ul (kr) jl (kr) + Xl (k)hl (kr).
(15.30)
This equation holds for large r with V = 0 and , constant. By using eq14.20
the identity

1 (1)
(2)
jl (kr) =
hl (kr) + hl (kr) ,
2
we can rewrite equation 15.30 as
(1)
ul (kr)
i
1 h (2)
(1)
hl (kr) + (1 + 2Xl )hl (kr) .
2
We now define l (k) by
(15.31)
eq14.21
1 + 2Xl = e2il (k) ,
220
We will prove that l (k) is real. This definition allows us to rewrite

equation 15.31 as
(1)
ul (kr)
i
1 h (2)
(1)
hl (kr) + e2il (k) hl (kr) ,
2
or
eq14.22
h
i
1
(1)
(2)
(1)
ul (kr) = eil eil hl + eil hl .
(15.32)
2
The solution ul1 satisfies a real differential equation. The boundary
condition at r 0 gives real coefficients. Thus ul1 is real up to an
overall constant factor. This implies l real. Another way of seeing this
(1)
(2)
is to note that by the definition of hl and hl we have
h
(2)
(1)
hl (kr) = hl (kr) .
Ask Baker
22 Apr p8
Thus the bracketed expression in equation 15.32 is an element plus

its complex conjugate, which is therefore real. If ul1 (kr) R, then
l (kr) R.
We now look at the second term in equation 15.32 for far fields,
r
(1)
eil hl (kr) eil (k)
i
(i)l eikr .
kr
Note that
(i)l = eil/2 .
This gives
(1)
eil hl (kr) =
i i(krl/2+l )
e
kr
So
1
sin(kr l/2 + l (k))
r .
(15.33)
kr
Thus l (k) is the phase shift of the lth partial wave at wave number k.
In the case that V = 0 we have
ul1 (kr)
eq14.23
ul1 (kr) ul1,0 (kr).

If there is no potential, then we have
Xl (k) 0,
15.9. THE PHYSICAL SIGNIFICANCE OF XL
(l)
221
(l)
u1 (r) u1,0 (r)

Rn(0)
Rn
Figure 15.3: Phase shift due to potential.

22 Apr p9
and by using the asymptotic expansion of j, we see that equation 15.30

becomes
!
1
l
l
u1,0 (r)
sin kr
.
(15.34)
kr
2
Thus the phase shift l (k) is zero if the potential is zero.
eq14.24
Consider the values of r for which the waves u1 and u1,0 are zero in 25 Apr p1
the far field limit. For equation 15.33 and equation 15.34 respectively, 25 Apr p2
the zeros occur when
25 Apr p3
l
kRn
+ l = n,
2
and
l
= n.
2
By taking the difference of these equations we have
kRn0
k(Rn Rn0 ) = l (k).
(15.35)
Thus l (k) gives the large distance difference of phase between solutions (eq14.25
with interaction and without interaction. This situation is shown in figure 15.3. For the case shown in the figure, we have Rn0 > Rn , which fig14c
means l > 0. Note that turning on the interaction pulls in the scattered wave. Thus we identify two situations. l > 0 corresponds to an
attractive potential, which pulls in the wave, while l < 0 corresponds
to a repulsive potential, which pushes out the wave.
222
25 Apr p4
We now verify this behavior by looking at the differential equation

for the quantum mechanical case. We now turn to the quantum mechanical case. In this case we set (r) = h
2 /2m and k 2 (x) = 2mE
in
h2
the equation
"
1 d
d
2
r2
r dr
dr
V l (r) (r) l
+ eff
u1 (r) = 0.
(r)
(r)
#
So for the radial equation with no interaction potential we have / =

2mE/
h2 , while for the radial equation with an interaction potential we
have / = 2m(E V )/
h2 . Thus the effect of the interaction is to
change the wave number from
k02 =
2mE
,
h
2
to an effective wave number

k 2 (r) =
2m(E V )
.
h
2
(15.36)
eq14.26
25 Apr p5
fig14d
Suppose we have an attractive potential, V (r) > 0. Then from

equation 15.36 we see k 2 (r) > k02 , which means momentum is increasing.
Also, since k 2 (r) > 0, increasing k 2 increases the curvature of u, which
means the wavelength (r) decreases and the kinetic energy increases.
Thus the case k 2 (r) > l02 corresponds to an attractive potential pulling
in a wave, which means l (k) > 0. The phase shift l (k) > 0 is a
measure of how much the wave is pulled in. Note that this situation is
essentially that of a wave equation for a wave moving through a region
of variable index of refraction.
Now consider a repulsive potential with l = 0, as shown in figure
15.4. We have V (r) = E for r = r0 , and V (r) > E for r > r0 . In
this latter case equation 15.36 indicates that k 2 (r) > 0, which means
the wave will be attenuated. Thus, as the wave penetrates the barrier,
there will be exponential decay rather than propagation.
15.9.1
Calculating l (k)
From Griffies,
l
28 April, p2b It is possible to calculate l (k) directly from u1 without calculating
Xl (k) as an intermediate step. To do this, let r and then compare
15.10. SCATTERING FROM A SPHERE
223
V (r)
E
r0
Figure 15.4: A repulsive potential.

this ul1 with the general asymptotic form from equation 15.33, u
sin(kr l/2 + l (k))/r. A different method for calculating l (k) is
presented in a later section.
15.10
Scattering from a Sphere

pr:ScSph1
We now look at the example of scattering from a sphere, which was

already solved in the homework.
We have the boundary conditions
V = 0 at r = a
l
u + u1 = 0 at r = a.
r 1
We found in problem set 2, that Xl for this problem is
Xl =
[kjl0 (ka) jl (ka)]

(1)0
(1)
[khl (ka) hl (ka)]
25 Apr p7
(15.37)
by solving the radial equation.

Stuff missing
Now look at the long wavelength limit, which is also the low energy
limit. In this case ka 1 where k = 2/. We know asymptotically
that
jl (ka) (ka)l ,
224
and
(1)
hl (ka)
1
.
(ka)l+1
Thus we have
ka1
2l+1
Xl (k) (ka)
since
(ka)2l+1 =
25 Apr p8
l a
l + a
1
(ka)l
.
(ka)l1
Again, k = 2mE/
h2 . We now look at the phase shift for low energy
scattering. We use the fact
Xl (k) (ka)2l+1
to write
1 + 2Xl (k) = ei2(k)
= 1 + 2il + .
Thus we have
l (k) (ka)2l+1 .
25 Apr p9
15.10.1
A Related Problem
We now turn to a related problem. Take an arbitrary potential, for

example
V = V0 er/a .
fig14e
In this case the shape of Veff is similar, except that is has a potential
barrier for low values of r. V and Veff for this example are shown in
figure 15.5. The centrifugal barrier increases as l increases, that is, it
gets steeper. Thus, as l increases, the scattering phase shift gets smaller
and smaller since the centrifugal barrier gets steeper.
Recall that a represents the range of the potential and 2l + 1 represents the effect of a potential barrier. We assert that in the long
15.11. CALCULATION OF PHASE FOR A HARD SPHERE
V (r)
225
Veff (r)
r
r
a
r0
Figure 15.5: The potential V and Veff for a particular example.

wave length limit, that is, low energy scattering, the phases shift goes
generally as
for ka 1.
l (k) (ka)2l+1
This is a great simplification for low energy scattering. It means that
as long as ka 1, we need only consider the first few l in the infinite
series for the scattering amplitude f (). In particular, the dominant
contribution will usually come from the l = 0 term. For the case l = 0,
the radial equation is easier to solve, and X0 (k) is easier to obtain.
Thus the partial wave expansion is very useful in the long wavelength,
or low energy, limit. This limit is the opposite of the geometrical or
physical optics limit.
The low energy limit is useful, for example, in the study of the
nuclear force, where the range of the potential is a 1013 cm, which
gives ka 1. Note that in the geometrical optics limit, ka 1, it is
also possible to sum the series accurately. The summation is difficult in
the middle region, ka 1. In this case many terms of the series must
be retained.
27 Apr p1
27 Apr p2
15.11
27 Apr p3
Calculation of Phase for a Hard Sphere
We use the special case from above. Take (very high elastic
constant, very rigid media, a hard sphere). In this case u 0 when
226
r = a. Thus we get from equation 15.37

X0 (k) =
sinkaka
ika
ieka
ieika eika
2ieika
1
= [1 e2ika ].
2
=
So
e2i0 = 1 [1 e2ika ] = e2ika ,
and thus
0 (k) = ka.
eq14do
(15.38)
In terms of quantum mechanics, this is like having
27 Apr p4
V (r) = for r < a,

V (r) = 0 for r > a.
Outside we get the asymptotic solution form given in equation 15.33.
For l = 0 and substituting equation 15.38, this becomes
(0)
u1 =
fig14f
1
sin(kr ka).
r
By substituting this into equation 15.13 it is easy to verify that this

is an exact solution for ul=0
1 . This is exactly what we would expect:
a free space spherical wave which satisfies the boundary condition at
r = a. The wave is pushed out by an amount ka. We thus see that
l (k) is determined by the boundary condition. This situation is shown
in figure 15.6.
15.12
Experimental Measurement
pr:ExpMeas1
27 Apr p5
We now look at the experimental consequences. Assume that we have

solved for u1 and know Xl (k) and thus know l (k). By writing the
scattering amplitude from equation 15.11 in terms of the phase shift
l (k), we have
15.12. EXPERIMENTAL MEASUREMENT
227
u1l=0 (r)
V (r)
a
Figure 15.6: An infinite potential wall.
f () =
1X
e2il 1
(2l + 1)Pl (cos )
.
k l=0
2i
To get l for the solution for u1 , we look at large r. Note that

e2il 1
eil [eil eil ]
=
2i
2i
il
= e sin l .
So
1X
f () =
(2l + 1)eil sin l Pl (cos ).
k l=0
(15.39)
eq14.55
15.12.1
Cross Section
pr:CrSec1
This scattering amplitude is the quantity from which we determine the

energy or probability of the scattered wave. However, the scattering
amplitude is not a directly measurable experimental quantity.
Recall our original configuration of a source, an obstacle, and a
detector. The detector measures the number of particles intercepted
per unit time, dN/dt. (It may also distinguish energy of the intercepted pr:N2
particle.) This number will be proportional to the solid angle covered
by the detector and the incident flux of particles. If we denote the
proportionality factor as (, ), then this relationship says that the
rate at which particles are scattered into an element of solid angle is is
dN/dt = jinc d = jinc (d/d)d. Note that an element of solid angle pr:jinc1
is related to an element of area by r2 d = dA. The scattered current
through area dA is then dN/dt = jinc ((, )/d)dA/r2 . From this we
228
eq14cs1
identify the scattered current density

(, ) 1
jscat = jinc
r.
(15.40)
d r2
Now the quantum mechanical current density j is defined in terms of
the wave function:
"
#
h
j(r) = Re
,
im
where in the far-field limit the boundary condition of scattering tells us
that the wave function goes as
r
eikr
+
f (, ) .
r
!
ikz
scat N e
The wave function has an incident plane wave part and a scattered
spherical wave part. The current density for the incident wave is then
h
k
jinc = |N |2 z = jinc z,
m
and the current density for the scattered wave is
h
k |f (, )|2
|f (, )|2
3
r
+
O(r
)
j
r.
(15.41)
inc
m
r2
r2
By comparing equations 15.40 and 15.41, we identify the differential
cross section as
d
|f (, k)|2 .
(15.42)
d
jscat = |N |2
eq14cs2
eq14.57
This relationship between cross section and scattering amplitude

agrees with dimensional analysis. Note that the only dimensionful
quantity appear in equation 15.39 for f is k:
dim(f ()) = dim(k 1 ) = dim(length).
On the other hand, the dimension of the differential cross section is
dim[d/d] = dim[l2 /1] and dim[|f (, k)|2 ] = dim[l2 ].
Thus equation 15.42 is dimensionally valid. Note also that, because the
differential cross section is an area per solid angle, it must be real and
positive, which also agrees with |f |2 . The total cross section is
Z
Z
d
= d|f (, k)|2 .
(k) = d
d
15.12. EXPERIMENTAL MEASUREMENT
15.12.2
229
Notes on Cross Section
By using equation 15.39 we can calculate the differential cross section:
d
1 X
= 2
(2l + 1)(2l0 + 1)eil sin l eil0 sin l0 Pl (cos )Pl0 (cos )
d
k l,l0 =0
(15.43)
In this equation we get interference terms (cross terms). These interfer- eq14.60
ence terms prevent us from being able to think of the differential cross 27 Apr p6
section as a sum of contributions from each partial wave individually.
If we are measuring just , we can integrate equation 15.43 to get
d
(15.44)
d
Z
1 X
=
d 2
(2l + 1)(2l0 + 1)eil sin l eil0 sin l0 Pl (cos )Pl0 (cos )
k ll0 =0
We can simplify this by using the orthogonality of the Legendre poly- eq14.61
nomials:
Z
4
dPl (cos )Pl0 (cos) = ll0
.
2l + 1
In equation 15.44 the terms eil cancel. So we now have
(k)
=
d|f (, k)|2
4 X
(2l + 1) sin2 l .
k 2 l=0
From this we can conclude that

=
(15.45)
eq14.64
l ,
l=0
where
4
(2l + 1) sin2 l .
(15.46)
k2
Note that l is the contribution of the total cross section of scattering eq14.65
from the 2l + 1 partial waves which have angular momentum l. There
are no interference effects, which is because of spherical symmetry.
27 Apr p7
l =
230
Another way to think about this point is that the measuring apparatus has introduced an asymmetry in the field, and the we have interference effects in d/d. On the other hand, in the whole measurement
of , there is still spherical symmetry, and thus no interference effects.
A measurement of (k) is much more crude than a measurement of
d/d.
Because sine is bounded by one, the total cross section of the partial
waves are also bounded:
4
l lmax = 2 (2l + 1),
k
or, by using = 2/k,
lmax
eq14.67
pr:GeoLim1
= 4
2
!2
(2l + 1).
(15.47)
Note that the reality of l (k) puts a maximum value on the contribution
l (k) of the lth partial wave on the total cross section.
15.12.3
Geometrical Limit
In the geometrical limit we have ka 1, which is the long wavelength

limit, a. Recall that in this limit
l (ka)2l+1 .
Thus the dominant contribution to the cross section will come from
4
4
0 = 2 sin2 0 = 2 (ka)2 = 4a2 .
(15.48)
k
k
From equation 15.47 we have
0max
27 Apr p8
4
2
!2
and from equation 15.48 we have

2
a
.
Comparing these gives us 0 /0max 1 for a/ 1, which is the

fraction of the incident beam seen by an observer.
0 = 4
15.13. OPTICAL THEOREM
15.13
231
Optical Theorem
pr:OptThm1
We now take the imaginary part of equation 15.46:

Imf () =
1X
(2l + 1) sin2 l Pl (cos ).
k l=0
In the case that = 0 we get
1X
(2l + 1) sin2 l .
k l=0
Imf (0) =
By comparing this with equation 15.45 we obtain

Imf (0) =
k
.
4
This is called the optical theorem. The meaning of this is that the
imaginary part of the energy taken out of the forward beam goes into
scattering. This principle is called unitarity or conservation of momentum. The quantity Im f ()|=0 represents the radiation of the intensity
in the incident beam due to interference with the forward scattered
beam. This is just conservation of energy: energy removed from the
incident beam goes into the scattered wave.
29 Apr p1
15.14
Conservation of Probability Interpretation:

29 Apr p2
15.14.1
Hard Sphere
For the case of a hard sphere of radius a we found that

0 = ka.
In the case of (ka) 1, only the lower terms of equation 15.39 matter.
Exact scattering amplitude from a hard sphere k = 0? So for a sphere
of radius a, we have
l (ka)2l+1 .
k/4
as
a proportionality factor
pr:HardSph1
232

a2
a2
strong forward peak

Figure 15.7: Scattering with a strong forward peak.
In the case that k 0, we get (noting that eil 0):
f () =
29 Apr p3
i il
i
i
e sin 0 (k)P0 (cos ) = eil (ka) = (ka) = ia,
k
k
k
as k 0.
Thus
d
= a2 ,
as k 0.
d
Note that the hard sphere differential cross section is spherically symmetric at low energy (that is, when ka 1). In this case the total
cross section is
Z
d
= 4a2 .
d
For the geometrical optics limit, ka 1, corresponding to short
wavelength and high energy, we would expect a2 since the sphere
looks like a circle, but instead we get
2(a2 ).
fig14g
The factor of two comes from contributions from all partial waves and
has a strong forward peak. The situation has the geometry shown in
figure 15.7. The figure is composed of a spherically symmetric part and
a forward peak, which each contribute a2 to the total cross section .
29 Apr p4
15.15
Radiation of Sound Waves
pr:soundWv1
We consider a non-viscous medium characterized by a sound velocity

v. In this medium is a hard sphere oscillating about the origin along
15.15. RADIATION OF SOUND WAVES
233
the z-axis. The motion of the center of the sphere is given by

xc = aeit z,
where 1, and the velocity of the center of the sphere is then given
by
vc = iaeit z.
Note that the normal component of the velocity at the surface of the
sphere is
n
vsphere = iaeit cos .
(15.49)
where we have used n
z = cos with measures from the z-axis. The eq14.88
minus sign appears because we choose n to point into the sphere. For
the velocity of the fluid outside of the hard sphere we have
vfluid = ,
where is the velocity potential. Thus near the surface of the sphere
we have, up to first order in ,
n
v|r=a

.
=
r r=a

We want to find the velocity potential where the velocity potential

satisfies the equation
"
1 2
+ 2 2 (x, t) = 0,
c t
#
r > a,
with the hard sphere boundary condition that the fluid and the sphere
move at the same radial velocity near the surface of the sphere,
n
vsphere = n
vfluid .
The velocity of the fluid is then given by (using equation 15.49)
|r=a = iaeit cos .

r
(15.50)
eq14.93
234
15.15.1
Steady State Solution

pr:sss4
The steady state solution is of the form

(x, t) = eit (x, ),
where (x, ) satisfies
[2 + l2 ](x, ),
29 Apr p5
r > a,
with the outgoing wave boundary condition (from equation 15.50)
= ia cos(),
r
r = a.
Our boundary condition is of the form

= g(, ),
r r=a

where for our specific case

g(, ) = ia cos .
eq14.98
(15.51)
We want to solve the steady state equation subject to the boundary

condition. A more general form of the boundary condition is
+ = g(, ).
r
eq14.99
(15.52)
We write
[2 k 2 ]G(x, x0 ; ) = (x0 x)/c2 ,
where k 2 = /c2 . This is the standard form of the Greens function in
the case that = c2 , L0 = 2 , and |x0 ||x| > 0. The solution of this
equation, which we found previously, is
(1)
c2 G(x, x0 ; ) = ik
29 Apr p6
zl (kr< )hl (kr> )
.
(1)
(1) 0
hl (ka) khl (ka)
(15.53)
The general solution is given by a superposition of the Greens function
solution for source points on the surface of the sphere,
lm
(eq14.101
Ylm (, )Ylm
(0 , 0 )

(x) = c2
eq14.102
Z
x0 S
235
G(x, x0 ; = 2 + i)g(, )a2 d0 .
(15.54)
By setting = 2 + i, we have automatically incorporated the outgoing wave condition. Physically, g(, )a2 d0 is the strength of the
disturbance.
We use equation 15.53 with r> = 0 and r< = a. Thus zl is
(1) 0
(1)
zl (kr) = [hl (ka) khl (ka)]jl (kr)

(1)
[jl0 (ka) jl0 (ka)]hl (kr)

(1)
= kW (jl (ka)hl (ka)).

Recall that we have evaluated this Wronskian before, and plugging in
the result gives
i
i
zl (ka) = k
= 2
(15.55)
2
(ka)
ka
By combining equations 15.55 and 15.53 into equation 15.54, we obtain (eq14.106
29 Apr p7
(1)
X Ylm (, )hl (kr) Z

i
(x) = a2 ( 2 )(ik)
d0 Ylm
(0 , 0 )g(0 , 0 )
(1)
(1)
ka
lm hl (ka) khl (ka)
(1)
Ylm (, )hl (kr)
lm
hl (ka) khl (ka)
(1)
(1)
glm .
(15.56)
This is called the multipole expansion, where we also have defined

glm
d0 Ylm
(0 , 0 )g(0 , 0 ).
eq14.108
(15.57)
glm is the (l, m)th multipole moment of g(, ).
eq14.109
2 May p2
15.15.2
Far Field Behavior
pr:farFld1
At distances far from the origin (r ) the spherical Hankel functions

can be approximated by
(1)
hl (k, r) =
i
(i)l eilr ,
kr
r 1.
236
In this limit the velocity potential can be written

(x) =
eikr
f (, ),
r
r ,
where the amplitude factor f is given by

f (, ) =
iX
gl,m Ylm (, )(i)l
.
(1)0
k l,m h(1)
l (k, a) khl (k, a)
This amplitude may be further decomposed into components of particular l and m:

X
f=
fl,m Ylm (, ),
l,m
where
fl,m =
(i)l+1
gl,m
.
0
(1)
k hl (k, a) kh(1)
l (k, a)
The interpretation of the ls is

l=1
l=2
l=3
2 May p3
dipole radiation
m = 0, 1
quadrapole radiation m = 0, 1, 2
octopole radiation
m = 0, 1, 2, 3
For the case l = 1, the 3 possible ms correspond to different polarizations.
15.15.3
Special Case
We now return to the specific case of the general boundary condition,

equation 15.52, which applies to a hard sphere executing small oscillations. In this case the hard surface implies = 0 and the oscillatory
motion implies that g is given by equation 15.51, which can be rewritten
in terms of the spherical harmonic Y10 :
s
g(, ) = ia
3 0
Y ().
4 l
237
By plugging this into equation 15.57, we have obtained the (l, m) components of g(, ),
gl,m =
=
dYlm (, )g(, )
s
d(ia)Ylm (, )
s
= ia
3 0
Y (, )
4 1
3
m0 l1 .
4
This shows that the oscillating sphere only excites the Yl0 mode.
1 iat
f =
0
k kh(1)
l (ka)
ia cos
=
.
(1)0
k 2 hl (ka)
3 0
Y (, )
4 l
Thus we have pure dipole radiation for this type of oscillation. This
final equation gives the radiation and shows the dependence on k.
omitted
15.15.4
stuff
2 May p4
Energy Flux
Consider a sound wave with velocity

v = (x, t),
where in the far field limit the velocity potential is
(x, t)
eikr
f (, )eit ,
r
x .
We now obtain the rate dE/dt at which energy flows through a

surface. This is given by the energy flux through the surface,
dE Z
ds jE ,
dt
where jE is the energy flux vector. For sound waves the energy flux
vector can be expressed as a product of velocity and pressure,
jE = vp.
(15.58)
qm
238
This can be intuited as follows. The first law of thermodynamics says eq14jvp
that for an ideal fluid undergoing a reversible isentropic process, the FW p299
change in internal energy dE matches the work done on the element,
pdV . The total energy flowingout of through the surface S is
!
Z
dEs Z
dE Z
dr
ds jE =
= ds
= pdV = ds p
.
dt
dt
dt
S
S
S
S
2 May p5
By comparing integrands we obtain equation 15.58, as desired. Note

that jE = vp has the correct dimensions for flux that is, velocity
times pressure gives the correct dimensions for energy.
The velocity and pressure are defined in terms of the velocity potential and density,
x = ,
p=
.
t
We now look at the real parts of the velocity and the pressure for the
steady state solution,
1
Re v = veit + v eit ,
2
1
Re p = (peit + p eit ).
2
ask
Baker The flux is then
about this.
j = Re vR Re pR
1
=
(vp + v p) + e2it vp + e2it v p .
4
The time averaged flux is then
1
1
hji = (vp + v p) = Re (xp ),
4
2
where we have used
he2it i + he+2it i = 0.
239
The angled brackets represents the average over time. Note that x and
p are still complex, but with their time dependence factored out. To
obtain p we use
p(t) = eit p = eit ,
from which we obtain
p = .
Thus the time averaged flux is
1
hjE i = Re()(+ie) .
2
(15.59)
The radial derivative of is
eq14.149
r =
= ik.
r
Thus in this case the time averaged energy rate is

*
Therefore
dE
dt
= r2 d
r hjE i = r2 d
dE
dt
1
= k|f |2 ,
2
|f |2
k.
r2
r 1.
(15.60)
eq14.152
Plane Wave Approximation

pr:PlWv2
Now suppose that instead of a spherical wave, we have a plane wave,

(x, t) = eikxit .
In this special case the velocity and pressure are given by
v = = ik
r 1,
p = (i).
Using equation 15.59, the energy flux is
1
1
j = k = k.
2
2
240
The power radiated through the area element dA is then

dEA
=
dt
and we have
ds j =
dA
2 May p8
k
dA,
2
1 dEA
k
Incident flux =
.
dA dt
2
(15.61)
(eq14.158
15.15.5
Scattering From Plane Waves
The far field response to scattering from an incident plane wave is

(x) = eikx + f
eikr
.
r
Note that in the limit r , the scattered wave S is f eikr /r. So

dEs
1
= k|f |2 .
dt d
2
2 May p5
By definition, the differential cross section is given by the amount of

energy per unit solid angle per unit time divided by the incident energy
flux,
dEs
d
dt d
d
Incident flux
1
k|f |2
= 2 1
= |f |2 .
k
2
In the second equality we have used equation 15.60 and 15.61. This
duplicates our earlier result, equation 15.42.
If we are just interested in the radiated wave and not the incident
flux, the angular distribution of power is
dP
dE
1
=
= k|f |2 .
d
dtd
2
Now expand f in terms of spherical harmonics,
f=
X
l,m
Yl,m fl,m .
241
The radiated differential power can then be written
X
X
dP
1
1
0
= k|f |2 = k Ylm fl,m
Yl m
fl0 ,m0 ,
0
d
2
2
0
0
l,m
l ,m
where we have interference terms. The total power is

P =
2 May p6
X
dP
=
|fl,m |2 .
d
l,m
In this case there is no interference. This is the analogue for sound

wave of the differential cross section we studied earlier. For the case of
a sphere
fl,0 6= 0.
4 May p1
15.15.6
Spherical Symmetry
We now consider the situation where the properties of the medium

surrounding the fluid exhibit spherical symmetry. In this case the scattering amplitude can be expanded in terms of spherical harmonics,
f (, ) =
fl,m Ylm (, ).
l,m
This is called the multipole expansion. The term fl,m corresponds to

the mode of angular momentum radiation. Spherical symmetry here
means that the dynamic terms are spherically symmetric: (r), (r),
and V (r). However, any initial condition or disturbance, such as g,
may have asymmetry. We now look at the external distance problem.
gl,m =
dYlm (, )g(, ).
For the general boundary condition the scattering amplitude is related

to g by
(i)l+1
gl,m
.
fl,m =
0
(1)
k hl (ka) kh(1)
(ka)
l
242
For our case of small oscillations of a hard sphere, we have = 0 and 4 May p2
s
gl,m = l,1 m,0
3
ia.
4
In this case the scattering amplitude becomes

f (, ) =
Flm Ylm () =
l,m
ia
(1)
k 2 hl=1 (ka)
cos =
iac cos
(1)0
khl (ka)
Thus the differential power radiated is

dP
1 2 ac cos
=
k

k h(1)0 (ka)
d
2
1
1 a2 2 c2
2
=

2 cos .
0
(1)

2 h (ka)
1
Notice that this cos2 dependence is opposite that of dipole radiation,

which goes like sin2 . The total power radiated is in general given by
P =
4 May p3
X
dP
1
= ck 2
|fl,m |2 .
d
2
l,m
Note that there are no interference terms. It is simply a sum of power

from each partial wave.
15.16
Summary
1. The asymptotic form of the response function is

lim ul1 (r) = jl (kr) + Xl (k)h1l (kr).
2. The scattering amplitude for a far-field observer due to an incident plane wave is
f (, k) =
iX
(2l + 1)Pl (cos )Xl .
k l=0
15.17. REFERENCES
243
3. The phase shift l (k) is defined by the relation

1 + 2Xl = e2il (k) ,
which results in a scattered wave solution of the form
ul1 (kr)
1
sin(kr l/2 + l (k))
kr
r ,
where l (k) appears as a simple shift in the phase of the sine wave.
4. The scattering amplitude is given by
f () =
1X
(2l + 1)eil sin l Pl (cos ).
k l=0
5. The differential cross section represents the effective area of the

scatterer for those particle which are deflected into the solid angle
d, and can be written in terms of the scattering amplitude as
d
|f (, k)|2 .
d
6. The optical theorem is
Imf ()|=0 =
k
.
4
It relates forward wave to the scattered wave.

7. The total cross section for scattering from a hard sphere in the
high energy limit is
2(a2 ).
15.17
References
See any old nuclear or high energy physics text, such as [Perkins87].
244
Chapter 16
Heat Conduction in 3D
Chapter Goals:
State the general response to the time-dependent
inhomogeneous heat equation.
Describe the physical significance of the boundary
condition.
Derive the temperature exterior to a fixed temperature circle.
16.1
General Boundary Value Problem
We saw in an earlier chapter that the heat equation is

"
T (x, t) = q(x, t)
L0 + cp (x)
t
for x in R, with the linear operator

L0 = T (x).
For the time dependent problem need both an initial condition and a
boundary condition to determine a unique solution. The initial condition is
T (x, t) = T0 (x) for t = 0.
245
246
CHAPTER 16. HEAT CONDUCTION IN 3D
For our boundary condition we take the radiation condition,

T n
T = [Text (s, t) T (x, t)]
for x on s.
Recall the the radiation condition came from the equilibrium condition
for radiation conduction balance. As an example of this sort of problem,
consider the boundary to be the surface of the earth. In the evening
time the temperature of the surface is determined by radiation. This is
a faster method of transfer than heat conduction. The above radiation
condition says that there exists a radiation conduction balance. Note
that when we consider convection, we must keep the velocity dependent
term x and the problem becomes non-linear. In this context v is
the motion of the medium due to convection.
The solution in terms of the Greens function is given by the prinpr:GenSolHeat1 ciple of superposition
T (x, t) =
dt
Z
dt
Z
R
eq15.0
4 May p4
pr:Brom1
ds0 G(x, t; x0 , t0 )Text (s0 , t0 )
G(x, t; x0 , 0)(x0 )cp (x0 )T0 (x0 ).
The integral containing (x0 )q(x

0 , t0 ) represents contributions due to
volume sources; the integral containing Text (s0 , t0 ) represents contributions due to surface sources; and the integral containing (x0 )cp (x0 )T0 (x0 )
represents contributions due to the initial conditions. The integrations
over time and space can be done in either order, which ever is easiest.
The Greens function is given by
ds s(tt0 )
e
G(x, x0 ; = s),
(16.1)
2i
L
where L is the upward directed line along any constant Re s > 0. This
choice of contour is necessary since L0 is positive definite, which means
that all the singularities of G(x, x0 ; = s) lie on the negative real
s axis. This integral, which gives the inverse Laplace transform, is
sometimes called the Bromwich integral. The Laplace space Greens
function satisfies the differential equation
G(x, t; x0 , t0 ) =
eq15.1
Z
xS
dx0 G(x, t; x0 , t0 )(x0 )q(x

0 , t0 )
[L0 cp ]G(x, x0 ; ) = (x x0 )
x, x0 R,
16.2. TIME DEPENDENT PROBLEM
247

[T n
+ ]G(x, x0 , ) = 0
x R, x0 S.
If the dynamical variables cp (x), (x), and T (x) are spherically symmetric, then the Greens function can be written as bilinear product of
spherical harmonics,
G(x, x0 ; ) =
Ylm (, )Gl (r, r0 ; )Ylm (0 , 0 ).
By plugging this into equation 16.1, we obtain

G(x, t, x0 , t0 ) =
Ylm (, )Gl (r, t; r0 , t0 )Ylm (0 , 0 )
X
l,m
where
Z
G(r, t; r0 , t0 ) =
16.2
ds s(tt0 )
e
Gl (r, r0 ; = s).
2i
7 May p1
Time Dependent Problem
We now consider the case in which the temperature is initially zero, and
the volume and surface sources undergo harmonic time dependence:
T0 (x, t) = 0
it
q(x,
t) = q(x)e
Text (s0 , t) = Text (s0 )eit .

We want to find T (x, t) for t > 0. Note that if T0 (x) 6= 0 instead,
then in the following analysis we would also evaluate the third integral
in equation 16.1. For the conditions stated above, the temperature 7 May p2
response is
T (x, t) =
dt0
dx0 G(x, t; x0 , t0 )(x0 )q(x

0 )eit
Z
0
dt0
Z
xS
ds0 G(x, t; x0 , t0 )Text (s0 )eit .
248
We are looking for the complete time response of the temperature rather
than the steady state response. The time integration is of the form
Z
dtG(x, t; x0 , t0 )eit
Z t
ds st
0
0
e G(x, x0 ; = s) est it dt0
L 2i
0
Z
ds st
1 e(s+i)t
=
e G(x, x0 ; = s)
s + i
L 2i
Z
0
ds G(x, x ; = s) st
=
[e eit ].(16.2)
s + i
L 2i
eq15.10
The contour of integration, L, is any upward-directed line parallel to

the imaginary axis in the left half plane. We got the first equality by
substituting in equation 16.1 and interchanging the s and t integrations.
The second equality we got by noting
Z
est it dt =
e(s+i)t dt0

1
1 e(s+i)t .
s + i
If we allow T0 (x) 6= 0, then in evaluating the third integral of equation

16.1 we would also need to calculate the free space Greens function, as
was done in chapter 10.
G(x, t; x0 , 0) =
ds st
e G(x, x0 ; = s)
2i
0 2
e(xx ) /4t
=
.
4t
eq15.15
7 May p3
(16.3)
This applies to the special case of radiation in the infinite one-dimensional

plane.
16.3
Evaluation of the Integrals
Recall that the Greens function can also be written as a bilinear expansion of the eigenfunctions. The general form of solution for equation
16.3 is
( P
en t u (x)u (x0 )
G(x, t; x , 0) = R n 0 0nt 1 n
0
d e
interior
0
0
ImG(x,
x
,
+
i)
exterior.
(16.4)
16.3. EVALUATION OF THE INTEGRALS

eq15.16
7 May p4
249
In the case when there is explicit time dependence, it may prove useful
to integrate over t first, and then integrate over s. The expressions in
equation 16.4 are particularly useful for large times. In this limit only
a small range of n must be used in the evaluation. In contrast, for
short times, an expression like equation 16.3 is more useful.
If the functions Text , T0 , q and are spherically symmetric, then we
only need the spherically symmetric part of the Greens function, G0 .
This was done in the second problem set. In contrast, for the problem
presently being considered, the boundary conditions are arbitrary, but
the sources are oscillating in time.
Ask
Baker
about
rotating.
Now we will simplify the integral expression in equation 16.2.
To evaluate equation 16.2, we will use the fact from chapter 10 that 7 May p5
G(x, x0 ; s) has the form
e s
G(x, x , s) .
s
0
Note that the second term in equation 16.2 is

Z
ds G(x, x0 ; = s) it
e
= 0,
2i
s + i
because the integrand decays in the right-hand plane as e s . Thus

the fact that we have oscillating sources merely amounts to a change
in denominator,
Z
Z
ds
ds G(x, x0 ; = s) st
0
st osc.
G(x, x ; = s)e
e .
2i
2i
s + i
(16.5)
eq15osc
We thus need to evaluate the first term in equation 16.2. We close

the contour in the left-hand s-plane, omitting the branch along the
negative real axis, as shown in figure 16.1. By Cauchys theorem, the fig15a
closed contour gives zero:
e = 0.
s + i
L 2i
The integrand vanishes exponentially along L1 , L5 . Over the small
circle around the origin we have
Z
e = eit G(x, x0 , = i).
2i
s + i
250

Im s
s-plane

+

L1

L2
-

s = i |s|
Re s
L3
-

s = i |s|

B
L4
B
B
J
LJ5J
Q
sQ
Q
QP
PP
Figure 16.1: Closed contour around branch cut.

7 May p6
This looks like a steady state piece. We can use equation 16.4 to write
7 May p7
Z
7 May p8
m t un (x)un x )
discrete,
=
ds G(x, x0 , = s) st
n=0 e
(n )
0
e = R
1
0
0
t
Im G(x,x , i)
0e
2i
s + i
continuum.
0 d
i0
"
ds G(x, x0 , = s)est
1 Z0
ds0
=
G(x, x0 ; = s0 + i)
2i
s + i
2i s + i
#
Z
ds0
0
0
+
G(x, x ; = s i) .
s0 + i
0
Change variables
0 = s0 ,
to obtain
Z
9 May p1
Z
ds G(x, x0 , = s) st
d0
1
e =
[G(x, x0 , + i) G(x, x0 , 0 i)]
0
2i
s + i
i 2i
0
Z
0 t 1
Im G(x, x0 , 0 i)
0e
d
=
.
i 0
0
16.4. PHYSICS OF THE HEAT PROBLEM
16.4
251
Physics of the Heat Problem
We have been looking at how to evaluate the general solution of the

heat equation,
T (x, t) =
dt
+
+
dx0 G(x, t, x0 , t)(x0 )q(x

0 , t0 )
0
t
dt0
Z0
x in R
x in R
ds0 G(x, t, x0 , t0 )Text (s0 , t0 )
G(x, t, x0 , 0)(x0 ), cp (x0 )T0 (x0 ).
We now look at the physics. We require the solution to satisfy the

initial condition
T (x, t) = T0 (x) = 0
for t = 0,
and the general regular boundary condition

[T (x)
n + ]T (x, t) = Text (x, t)
x S.
This boundary condition represents the balance between conduction

and radiation.
16.4.1
The Parameter
pr:Theta1
We can rewrite the regular boundary condition as

n
T |x
on S
[Text (s, t) T (x, t)]x on S

Th
= [Text (s, t) T (x, t)]x on S ,
=
where = /Th . The expression on the left had side is the conduction
in the body, while the expression on the right hand side is the radiation
into the body. Thus, this equation is a statement of energy balance.
The dynamic characteristic parameter in this equation is , which has
the dimensions of inverse distance:
=
.
Th
distance
We now consider large and small values of .
9 May p2
252
Case 1: radiation important

In this region we have
1.
th
For this case we have radiation at large s and conduction is small, which
means
T (x, t) Text (s, t)
for x S.
=
Case 2: heat flux occurs

This applies to the case
=
1.
th
Thus we take
lim Tint F (s, t),
Tint
0
where F (s, t) is some particular heat flux at position s and time t. The
boundary condition then becomes a fixed flux condition,
th n
T(x, t)|x
on s
= F (s, t).
9 May p3
HW comments
omitted
16.5
Example: Sphere
The region is the exterior region to a sphere with

Text (, ; t) = Text (t).
So we can write
G(x, t, x0 , t0 ) =
9 May p4
1
G0 (r, t, r0 , t0 ).
4
So we just need to evaluate equation 16.1. We will get the typical

functions of the theory of the heat equation.
We take the temperature Text on the surface of the sphere to be
uniform in space and constant in time:
Tt (, , t) = Text .
16.5. EXAMPLE: SPHERE
253
In this case plugging equation 16.3 into 16.1 yields

Z
T (r, t) =
dt
dxG(x, t; x0 , 0)Text
0 2
e(xx ) /4t
Text
4t
!
a
ra
= Text erfc
,
r
4t
Z
where we define
and
dt
dx
ask Baker
2 Zx
2
erf x =
dzez ,
0
2 Z0
2
dzez .
erfc x = 1 erf x =
x
(16.6)
For short times, x is large, and for small times x is small and (Eq.14) eq15.32
is easy to evaluate. For large x we use integration by parts,
Mysterious
2 Z
21
zdzez
erfc x =
x
z

1 z2 1 Z
1 z2
1
2
dz e
= e
2
z
2
z2
x
!
Z
x2
1
dz z2
2 e
e
=
2x
2 x z2
2
2ex
=
equation omitted
1
1
3 + .
2x 4x

Thus we have a rapidly converging expansion for large x. For x 1, 9 May p5

2
we can directly place the Taylor series of ez inside the integral.
16.5.1
Long Times
We have standard diffusion phenomena. As t , the solution goes

t
T (r, t) ar Text . This is the steady state solution. It satisfies the conditions
2 T (x, t) = 0,
|x| > a,
T (x, t) = Text ,
|x| = a.
254
If
(r a)2
1
4t
then we satisfy the steady state condition, and we can define by
x2 =
(r a)2
(r a)2
= 1 and so =
.
4
4t
The variable is the characteristic time which determines the rate of
diffusion. So for t , the temperature T is if the form of the steady
state solution.
16.5.2
Interior Case
Having considered the region exterior to the sphere, we now consider

the problem for the interior of the sphere. In particular, we take the
surface source Text to have harmonic time dependence and arbitrary
spatial independence:
Text (t, s) = eit Text (s).
9 May p6
We further assume that there are no volume source and that the internal
temperature is initially zero:
q(x,
t) = 0, T (x, t = 0) = 0.
In this case equation 16.2 reduces to
T (x, t) =
ds0
x on s
or
T (x, t) =
Z
x on s
Z
0
dt0 G(x, t; x0 , t0 )eit Text (t, s),
ds Text (s )
it0
dt G(x, t, x , t )e
This equation was computed previously for an external region. The

solution was
T (x, t) =
Z
x0
ds Text (s ) eit G(x, x0 ; = i)

+
X
m
0
n t un (x)un (x )
i n
16.6. SUMMARY
255
This holds for the discrete case, which occurs when the region is the
interior of a sphere. For the continuous case, which is valid for the
external problem, we have
T (x, t) =
Z
x0
ds0 Text (s0 ) eit G(x, x0 ; = i)

1 Z 0 0 t Im G(x, x0 ; = 0 + i)
+
d e
.
0
i 0
The first term in the bracketed expression is the steady state part of 9 May p7
the response. The second term is the transient part of the response.
These transient terms do not always vanish, as is the case in the fixed
flux problem, in which there is a zero eigenvalue.
Many
16.6
HW comments
omitted
11 May p1
Summary
11 May p2
1. The general response to the time-dependent inhomogeneous heat

11 May p3
equation is
Z
T (x, t) =
dt
+
+
11 May p4
dx G(x, t; x , t )(x )q(x

,t )
Z0
R
dt0
Z
xS
ds0 G(x, t; x0 , t0 )Text (s0 , t0 )
G(x, t; x0 , 0)(x0 )cp (x0 )T0 (x0 ).
2. The boundary condition for the heat equation can be written

n
T |x
on S
= [Text (s, t) T (x, t)]x
on S
where = /Th . If 1, then radiation is dominant, otherwise if 1, then heat flux is dominant.
3. The temperature exterior to a fixed temperature circle is
!
a
ra
T (r, t) = Text erfc
,
r
4t
where
2 Z0
2
erfc x = 1 erf x =
dzez .
x
11 May p5
256
16.7
References
See the references of chapter 10.
Chapter 17
The Wave Equation
Chapter Goals:
State the free space Greens function in ndimensions.
Describe the connection between the even and
odd-dimensional Greens functions.
17.1
introduction
The Retarded Greens function for the wave equation satisfies

"
2
+ 2 G(x, t; x0 , t0 ) = (x x0 )(t t0 )
t
#
with the retarded boundary condition that GR = 0 for t < t0 . The

solution to this equation is
GR (x, t; x0 , t0 ) =
Z
L
d i(tt0 )
e
G(x, x0 ; = 2 ),
2
(17.1)
where the integration path L is any line in the upper half plane parallel eq16ft1
to the real axis and R = |x x0 | and where G(x, x0 ; ) satisfies
[ 2 ]G(x, x0 ; ) = (x x0 ).
(17.2)
We denote the solution of equation 17.2 in n-dimensions as Gn . We eq16B

257
258
CHAPTER 17. THE WAVE EQUATION
then have
ie
G1 (x, x ; ) =
/c2 R
2 /c2
i (1)
G2 (x, x0 ; ) =
H (k, R)
4 0
2
ei /c R
0
G3 (x, x ; ) =
,
4 R
(17.3)
(17.4)
(17.5)
where k = /c2 . It is readily verified that these three equations can

be written in the more general form
i
Gn (R; ) =
4
k
2R
! n 1
2
(1)
H n 1 (k, R).
2
The Fourier transform, equation 17.1, for the 3-dimensional case can be
reduced to the Fourier transform for the one dimensional case, which
we have already solved. The trick to do this is to rewrite the integral
as a derivative with respect to the constant parameter R, and then pull
the differential outside the integral.
d i(tt0 )
e
G3 (x, x0 ; = 2 )
L 2
Z
d i(tt0 ) e(i/c)R
e
=
4 R
L 2

Z
d i(tt0 )
1
2i ei c R
e
=
2R R 2 c
L 2
GR (x, t, x0 , t0 ) =
1 Z d ei c R i(tt0 )
=
ie
2R R L 2 2 c
1
=
G1 (x, t; x0 , t0 )
2R R

1
c
0
=
(c(t t ) R)
2R R 2
where the -function satisfies d(x/dx = (x). Note that
f (ax) =
1
(x).
|a|
17.2. DIMENSIONALITY
11 May p6
259
Thus we can write

1
c
G3 (x, t; x , t ) =
(c(t t0 ) R)
2R R 2
1
=
(c(t t0 ) R/c)
4R
1
=
(t t0 R/c).
4R
0
Our result is then

G3 (R, t t0 ) =
1
(t t0 R/c)
G1 (R, t t0 ) =
.
R R
4R
(17.6)
eq16.4
13 May p1
17.2
Dimensionality
17.2.1
Odd Dimensions
pr:oddDim1
(1)
H n 1 (k, R)
2
is a trigonometric function for any odd integer.

Note that
Thus for n odd, we get
13 May p2
1
Gn (R, t t0 ) =
2R R
and
1
Gn (R, ) =
2R R
Thus
i
Gn (R, ) =
2R
for n odd. We also have

! n1
G1 (R, t t0 )
! n1
2
G1 (R, ).
n1
2
Gn (R, t t ) =
2R R
2R R
G1 (R, )
! n3
2
! n3
2
G3 (R, t t0 )
(17.7)
(t t0 Rc )
.
4 R
(17.8)
260
17.2.2
13 May p3
Even Dimensions
pr:evDim1
Recall that the steady state Greens function for 2-dimensions is

i (1)
G2 (R, ) =
H (k, R).
4 0
If we insert this into equation 17.1 we obtain the retarded Greens
function,
c (c(t t0 ) R)
q
.
G2 (R, t t0 ) =
2 c2 (t t0 )2 R2
17.3
Physics
There are two ways to define electrostatics. The first is by Guasss law
and the second is by its solution, Coulombs law. The same relationship
is true here.
17.3.1
Odd Dimensions
We consider the n = 3-dimensional case,

(t t0 R/c)
G3 (R, t t ) =
.
4R
At time t the disturbance is zero everywhere except at the radius R =
c(t t0 ) from x0 . We only see a disturbance on the spherical shell.
0
17.3.2
13 May p3
Even Dimensions
In two dimensions the disturbance is felt at locations other than the

surface of the expanding spherical shell. In two dimensions we have
c [c(t t0 ) R]
q
G2 =
=
2 c2 (t t0 )2 R2
= 0 R > c(t t0 ),
6= 0 R < c(t t0 ).
(17.9)
The case G = 0 for R > c(t t0 ) makes sense since the disturbance has
not yet had time to reach the observer. We also have
G2 =
c [c(t t0 ) R]
q
2 c2 (t t0 )2 R2
as R c(t t0 ).
17.3. PHYSICS
261
G2
c(t t0 ) R
Figure 17.1: Radial part of the 2-dimensional Greens function.
Thus the maximum disturbance occurs at R c(t t0 ). Finally, G 6= 0
for R < c(t t0 ). Thus we have propagation at speed c, as well as all
smaller velocities. This is called a wake. The disturbance is shown in
figure 17.1. We have not yet given a motivation for why G2 6= 0 for fig16a
R > c(t t0 ). This will be done in the next section, where we will also
give an alternative derivation of this result.
17.3.3
Connection between GFs in 2 & 3-dim
We now calculate the Greens function in 2-dimensions using the Greens

function in 3-dimensions. This will help us to understand the difference
between even and odd dimensions. Consider the general inhomogeneous
wave equation in three dimensions,
"
2
23 + 2 u(x, t) = f (x, t).
t
#
(17.10)
From our general theory we know that the solution of this equation can eq16.1
be written in terms of the Greens function as
u(x, t) =
Z
0
dt0
dx0 G3 (x, t; x0 , t0 )f (x0 , t0 ).
We now consider a particular source,
(17.11)
eq16.2
f (x0 , t0 ) = (x0 )(y 0 )(t t0 ).

This corresponds to a line source along the z-axis acting at time t = t0 .
What equation does u satisfy for this case? The solution will be 13 May p6
completely independent of z: u(x, t) = u(x, y, t) = u(x2 +y 2 , t) = u(, t)
262
where = x2 + y 2 , and the second equality follows from rotational

invariance. For this case equation 17.10 becomes
"
22

1
+
u(x, y, t) = (x)(y)(t t0 ).
2
t
Thus
u(x, y, t) = G2 (, t t0 ),
eq16.6
13 May p7
where G2 was given in equation 17.9. We should be able to get the same
result by plugging the expression for G3 , equation 17.6, in equation
17.11. Thus we have
Z t
Z
1
(t t0 R/c)(x0 )(y 0 )(t0 t0 )
u(x, t) =
dt0 dx0 dy 0 dz 0
4R
0
(17.12)
0
Now let u(x, t) = u(x, y, 0, t) = G2 (, t t ) on the left hand side of
equation 17.12 and partially evaluate the right hand side to get
G2 (, t t0 ) =
eq16.7
fig16b
16 May p1
Z
0
dt0
dz 0
1 (t t0 R/c)
2
.
4
+ z 02
(17.13)
The disturbance at time t at the field point will be due to contributions

at z = 0 from = c(t t0 ). We also have disturbances at farther
distances which were emanated at an earlier time. This is shown in
figure 17.2.
Note that only the terms at z 0 contribute, where z 02 +2 = c2 (tt0 )2 .
So we define
q
z 0 = z = c2 (t t0 )2 2
We now consider the value of G2 using equation 17.13 for three
different regions.
G2 = 0 if > c(t t0 ) for a signal emitted at z. This is true since
a signal emitted at any z will not have time to arrive at since
in travels at velocity c.
16 May p2
If = c(t t0 ), then the signal emitted from the point z = 0 at

time t0 arrives at at time t. Thus z = 0.
Finally if < c(t t0 ), then the signals emitted at time t = t0
from the points z = z arrive at time t.
This is the origin of the wake.
17.4. EVALUATION OF G2
263
field
point HH
j
H
2 + z 02
z0
z=0
z+ z
AA
K
A
A
A
line source
on z-axis
Figure 17.2: A line source in 3-dimensions.
17.4
Evaluation of G2
We make a change of variables in equation 17.9,

R0 =
2 + z 02
and thus
dR0 =
so
zdz 0
,
R0
dR0
dz 0
.
=
R0
R02 2
The Greens function G 2 is then

G2 =
Z
dR0 /c
1
02
(2)
(t t0 R0 /c)
4
R 2
so the answer is
G2 =
c (c(t t0 ) )
q
.
2 c2 (t t0 )2 2
We would get the same result if we took the inverse Fourier transform 16 May p3
(1)
of H0 . For heat equation, the character of the Greens function is
independent of dimension; it is always Gaussian.
264
17.5
Summary
1. The free space Greens function in n-dimensions is

i
Gn (R; ) =
4
k
2R
! n 1
2
(1)
H n 1 (k, R).
2
2. The connection between the fact that the 3-dimensional Greens

function response propagates on the surface of a sphere and the
fact the the 2-dimensional Greens function response propagates
inside of a cylinder is illustrated.
17.6
References
See [Fetter80] and [Stakgold67]. This chapter is mostly just an exploration of how the number of dimensions affects the solution form.
Chapter 18
The Method of Steepest
Descent
pr:StDesc1
Chapter Goals:
Find
the solution to the integral I()
R
f (z)
dze
g(z).
C
Find the asymptotic form of the Gamma function.

Find the asymptotic behavior of the Hankel function.
Suppose that we integrate over a contour C such as that shown in figure
18.1:
Z
I() =
dzef (z) g(z).
(18.1)
C
We want to find an expression for I() for large . Without loss of eq17.a
generality, we take to be real and positive. This simply reflects the pr:Iint1
choice of what we call f (z). The first step will be to take the indefinite
integral. The second step will will then be to deform the contour C
into a contour C0 such that
df
=0
dz z=z0

where z0 lies on the contour C0 .

In order to perform these operations we will first digress to a review
265
266
CHAPTER 18. THE METHOD OF STEEPEST DESCENT
z0
C0
Figure 18.1: Contour C & deformation C0 with point z0 .
fig17.1
of the methods of complex analysis which are needed to compute this

integral. Then we shall explicitly solve the integral.
18.1
pr:anal2
Review of Complex Variables
Let z = x + iy and f (x, y) = u(x, y) + iv(x, y) where f (z) is analytic

on the region which we are considering. In general a function f of
the complex variable z is analytic (or holomorphic) at a point z0 if
its derivative exists not only at z0 but also at each point z in some
neighborhood of z0 , and a function f is said to be analytic in a region
R if it is analytic at each point in R. In this case we have:
d
du
dv
df
= (u + iv) =
+i .
dz
dz
dz
dz
eq17.1
eq17.2
pr:CReq1
Since the function is analytic, its derivative is independent of the path

of approach.
If we differentiate with respect to an infinitesimal change dz = dx,
we get
df
du
dv
=
+i ,
(18.2)
dz
dx
dx
and if we differentiate with respect to an infinitesimal change dz = idy,
we get
df
du
dv
du dv
=
+i
= i + .
(18.3)
dz
d(iy)
d(iy)
dy dy
By comparing equations 18.2 and 18.3 and separating the resulting
equation into real and imaginary parts we get the Cauchy-Riemann
equations:
18.1. REVIEW OF COMPLEX VARIABLES
267
Im z
v constant
u constant
Re z
Figure 18.2: Gradients of u and v.
dv
dv
du
du
=
and
= .
dx
dy
dx
dy
These facts allow us to make the following four observations about
differentiation on the complex plane:
Observation 1. The gradient of a complex valued function is depicted in figure 18.2 for an integral curve of an analytic function.
fig17.2
The product of gradients is given by the equation
u v =
du dv du dv
+
= 0.
dx dx dy dy
The last equality follows from the Cauchy-Riemann equations. This

means that the lines for which u is constant are perpendicular (i.e.,
orthogonal) to the lines for which v is constant.
Observation 2. For the second derivatives we have the following
relations:
!
d2 u
d du
d dv
=
=
(18.4)
dx2
dx dx
dx dy
and
eq17.3
!
!
d2 u
d du
d
dv
=
=
.
(18.5)
dy 2
dy dy
dy
dx
The differentials commute, so by combining equations 18.4 and 18.5 eq17.4
we get
d2 u d2 u
+
= 0,
dx2 dy 2
268

u = Re f (z)
x = Re z
y = Im y
Figure 18.3: f (z) near a saddle-point.
and similarly
pr:LapEq1
d2 v d2 v
+
= 0.
dx2 dy 2
This means that analytic functions satisfy Laplaces equation.
Observation 3. From Observation 2 we find that:
If
eq17.5
d2 u
d2 u
>
0
then
< 0.
dx2
dy 2
(18.6)
Thus we cannot have a maximum or a minimum of both u and v occur

anywhere in the complex plane. The point z0 = x0 + iy0 for which
du
du
|z0 = 0 and
| =0
dx
dy z0
fig17.3
is called a saddle point. The CauchyRiemann equations and equation

18.6 imply that if df /dz = 0 at z0 , then z0 is a saddle point of both
u(x, y) and v(x, y). This is illustrated in figure 18.3.
Observation 4. For an analytic function f = u+iv and a differential
dl we have
df = dl f
= dl u + idl v.
Note that |df /dz| is independent of the direction of dl due to analyticity.
Suppose that we chose dl to be perpendicular to u. In this case
dl v = 0, so
df = dl u
for dlku.
18.2. SPECIFICATION OF STEEPEST DESCENT
269
As the magnitude of f changes, the change dl u is purely real, since

u(x, y) is real. Thus the real part of f has maximum change in the
direction where dl v = 0, since |df /dz| is independent of direction.
Therefore dlv = 0 gives the path of either steepest descent or steepest
ascent. The information given so far is insufficient to determine which.
18.2
Specification of Steepest Descent
We want to evaluate the integral from equation 18.1,

I() =
dzef (z) g(z),
for large. We take to be real and positive. In the previous section

we wrote f (z) = u(z)+iv(z). Thus we want to know Re (f (z)) in order
to determine the leading order behavior of I() for 1.
To solve for I() we deform C C0 such that most of the contribution of the integral when 1 comes from a small region on C0 . Thus
we need to make an optimal choice of contour. We want df /dz = 0 at
some point z = z0 on the deformed contour C0 . We parameterize C0
with the line
z( ) = x( ) + iy( ).
We want the region of the curve where u( = Re (f ( )) to be as localized as possible. Thus we want the contour to run in the direction
where u( ) has maximal change. As we saw at the end of the previous section, this occurs when v(z( )) = v( ) remains constant. So our
deformed contour C0 has the property that
v( ) = a constant on C0 .
(18.7)
This will uniquely determine the contour.

eq17.6
Note that we assume there is only one point where df /dz = 0. If
there were more than one such point, then we would merely repeat this
process at the new point and add its contribution.
Equation 18.7 is equivalent to the condition
Im[f (z( )) f (z0 )] = 0.
270
The path for which this condition is satisfied is also the one for which
Re[f (z) f (z0 )] = u(z) u(z0 )
changes most rapidly.
We want to evaluate I() for large. Recalling the condition for a
local maximum or minimum that df /dz = 0, we note that it is useful
to rewrite the integral defined in equation 18.1 as
I() =
dzef (z) g(z) = ef (z0 )
dze(f (z)f (z0 )) g(z).
C0
Note that since f (z) is an analytic function, the integral over C0 is equal
to the integral over C. The main contribution is at the maximum of the
difference f (z) f (z0 ). We want to find the curve with the maximum
change, which has a local maximum at z0 , which means we want the
quantity f (z) f (z0 ) to be negative. Thus we want the curve along
which
Re[f (z) f (z0 )] = u(z) u(z0 )
changes most rapidly and is negative. This is called the curve of steepest
descent. This condition specifies which of the two curves specified by
Im[f (z( )) f (z0 )] = 0 we choose: we choose the path of steepest
descent.
18.3
Inverting a Series
pr:invSer1
We choose the parameterization

f (z) f (z0 ) 2
so we get
z( = 0) = z0 .
Note that is real since v( ) = 0 along the curve and f (z) < f (u0 ).
We need to invert the integral. Expand f (z) f (z0 ) in a power
series about z0 :
f 000 (z0 )
f 00 (z0 )
2
(z z0 ) +
(z z0 )3 + . . . = 2 . (18.8)
f (z) f (z0 ) =
2!
3!
18.3. INVERTING A SERIES

eq17.7
271
We also get
z z0 =
an n = a1 + a2 2 + a1 3 + . . . .
(18.9)
n=1
Note that there is no constant term in this series. This is because = 0 eq17.8
implies z z0 = 0. Thus, if we had an n = 0 term, we couldnt satisfy
this stipulation.
Plug equation 18.9 into 18.8:
f 00 (z0 ) X
2 =
an n
2!
n=1
!2
f 000 (z0 ) X
+
an n
3!
n=1
!3
To calculate a1 , forget the terms (f 000 (z0 )/3!)(z z0 )3 on. The calculation of a2 includes this term and the calculation of a3 includes the
following term. Thus
2 =
f 00 (z0 ) 2 2
a1 + O( 3 ).
2!
Now let
f 00 (z0 )
Re+i .
2!
Plugging this into equation 18.10 and canceling 2 yields
(18.10)
eq17.9
1 = a21 Rei
so
a21 =
where 1 = ei . So
eii
R
1
a1 = ei( 2 2 ) .
(18.11)
R
The calculation of the a0i s is the only messy part involved in finding eq17.10
subsequent terms of the inverted series. For our purposes, it is sufficient
to have calculated a1 . The in equation 18.11 gives us two curves for
the first term:
z z0 a1 = ei()/2 .
R
272
We now assume that we have calculated the whole series, and use
the series to rewrite I(). We now take our integral
f (z0 )
I() = e
dze(f (z)f (z0 )) g(z),
C0
and make a variable substitution

dz =
to obtain
f (z0 )
I() = e
dz
d
d
d e
dz
g(z( )),
d
where + and are on the curve C0 on opposite sides of = 0. We

expand the z( ) in the function g(z( )) as
z = a1 + a2 2 + . . .
and thus
eq17.11
X
dz
g(z( )) =
cn n ,
d
n=0
(18.12)
where the cn can be determined from the an and g(z( )). Thus we can
write
X Z +
2
f (z0 )
I() = e
d e cn n .
n
Thus, with no approximations being made so far, we can assert

I() = ef (z0 )
cn
d e n .
Now let and + . Our integral becomes

wf (z0 )
I(w) = e
cn
n=0
d ew n .
This is an elementary integral. We know

Z
w 2
d e
,
w
18.4. EXAMPLE 1: EXPANSION OF FUNCTION
273
r
d Z
d
w 2
d e
=
d e
=
=
,
3/2
d
d w
2
(this is called differentiating with respect to a parameter), and similarly
2 w 2
w 2
d 2m e
d
d
!n Z
d ew =
1 3 5 (2m 1)
.
2m (2m+1)/2
Since odd n gives zero by symmetry, we have
wf (z0 )
I(w) = e
cn
n=0,2,4,...
d ew n .
All this gives

wf (z0 )
I(w) = e
" r
c0
X
1 3 5 (2m 1)
c2
+
.
+
c2m
2m w(2m+1)/2
w
2 w3/2
m=2
The term c0 / corresponds to Sterlings formula and the term c22 w3/2
is the first correction to Sterlings formula. The only computation remaining is the dz/d in equation 18.12.
18.4
Example 1: Expansion of function

pr:Gamma1
We want to evaluate the integral

I(w) =
et tw dt.
18.4.1
Transforming the Integral
We want to get this equation into the standard form. We make an

elementary transformation to get it into the form
Z
dzewf (z) g(z).
We substitute t = zw to get
I(w) =
dzewz (zw)w
= ww+1
Z
0
dzew[log zz] .
(18.13)
(18.14)
274
For g = 1 we have
f (z) = log z z
and we have
df
1
= 1 = 0 at z = 1.
dz
z
This is a saddle point. We chose to define on the interval
< <
so that
z = rei ,
log z = log r + i.
This is analytic everywhere except the negative real axis, which we

dont need.
18.4.2
The Curve of Steepest Descent
Since we know the saddle point, we can write

f (z) f (z0 ) = log z z + 1.
So we just need to calculate
0 = Im[f (z) f (z0 )]
= r sin .
(18.15)
(18.16)
We expect the lines of steepest assent and descent passing through z0

to be perpendicular to each other. The two solutions of this equation
correspond to these curves. The solution = 0 gives a line on the
positive real axis. The other solution is
sin
2
1+
for 1.
6
r =
(18.17)
(18.18)
We havent yet formally shown which one is the line of steepest ascent
and descent. This is determined by looking at the behavior of f (z) on
each line.
18.4. EXAMPLE 1: EXPANSION OF FUNCTION
275
By looking at log z z + 1 we see that f (z) f (z0 ) can be written

f (z) f (z0 ) = log z (z 1)
= 2
X
=
cn (z 1)n .
(18.19)
We have to invert this in order to get the asymptotic expansion of

the gamma function. We expand equation 18.19 in a power series, after
noting that log z = log[(z 1) + 1]:
1
1
1
2 = (z 1)2 + (z 1)3 (z 1)4 + . . . .
2
3
4
We plug in
z( ) 1 = A + B 2 + C 3 + O( 4 )
1
1
1
2 = (A + B 2 + C 3 )2 + (A + B 2 )3 (A )4 + O( 5 )
2
4
! 3
!
2
2
1 2 2
A
B
A4
3
2
= A A B

+ AC A B +
4 + O( 5 ).
2
3
2
4
Comparing coefficients on the left and right hand side, we get
2 :
3 :
4 :
A2 = 2
A(B A2 /3) = 0
B2
A4
+ AC A2 B +
= 0.
2
4
This method is called inverting the power series. We find
A = 2,
C = 2/8.
The positive roots were chosen for convenience. Now we calculate what
the cn s are from
X
dz
g(z( )) =
cn n .
d
n=0
276
Since g(z) = 1 and

dz
= A + 2B + 3C 2 + . . . ,
d
we know that
C0 =
C2 =
2,
2
.
6
Finally, we plug in these values:

Z
0
et tw dt = I(w)
#
c2
= e
c0
+
+ ...
w
2 w3/2
s
2
2
= ew
+
+ . . . ,
w
12w3/2
wf (z0 )
" r
which agrees with Abramowitz & Stegun, formula 6.1.37.
18.5
Example 2: Asymptotic Hankel Function
pr:Hankel1
We want to find the asymptotic form of the Hankel function, starting

with the integral representation
H(1) (z)
fig17.4
1 Z +i z sinh ww
=
e
dw
i
The contour of integration is the figure 18.4. The high index and argument behavior of the Hankel function
H(1) (z)
are important in high energy scattering. The index is related to the
effect of an angular momentum barrier, and z to an energy barrier. In
this equation is an arbitrary complex number and z is an arbitrary
complex number in a certain strip of the plane.
18.5. EXAMPLE 2: ASYMPTOTIC HANKEL FUNCTION
277
Im w
Re w
Figure 18.4: Defining Contour for the Hankel function.
We now relabel the Hankel function as Hp(1) (z), where
p
= cos 0 .
z
Note that p/z is real and 0 < p/z < 1, which implies 0 < 0 < /2. So
Hp (z) =
1 Z
dzezf (w) g(w)
i C
where g(w) = 1, with

f (w) = sinh w pw/z = sinh w w cos 0 .
To examine asymptotic values |z| 1 with 0 fixed, we want to deform
the contour so that it goes through a saddle point. Using the usual
method, we have
df (w)
= cosh w cos 0 = 0.
dw
We define
w0 = i0
so that
cosh w0 cosh i0 = cos 0 .
Thus
f (w = i) = sinh i0 i0 cos 0 = i[sin 0 0 cos 0 ]
so that
f (w) f (w0 ) = sinh w w cos 0 i[sin 0 0 cos 0 ].
278

Im w
Ascent
/4
Descent
Re w
Figure 18.5: Deformed contour for the Hankel function.
We want to find out what the curves are. Note that

d2 f (w)
= sinh w
= i sin 0

2
dw w=i0
w=i0

so we can write
1
2 = f (w) f (i0 ) = i(sin 0 )(w i0 )2 + . . . .
2
To invert this series, we write
w i0 = A + B 2 + C 3 + . . . .
For now, we are just interested in the leading order term. So
1
2 = A2 2 i(sin 0 )
2
which implies
2i
.
sin 0
Recall that we are looking for the tangent of C0 at w0 . Thus we have
A2 =
A = ei/4
so
w i0 =
fig17.5
2
,
sin 0
2 i/4
e .
sin 0
The deformed curve C0 has the form shown in figure 18.5. The picture
neglects to take into account higher order terms. We choose the plus
18.5. EXAMPLE 2: ASYMPTOTIC HANKEL FUNCTION
279
sign to get the direction correct. Note that the curve of steepest ascent
is obtained by a rotation of /2 of the tangent, not the whole curve.
In the power series
g(z( ))
X
dz
=
cn n
d
n=0
where g(z( )) = 1, we have

dw
= A + higher order terms,
d
so
s
i/4
c0 = A = e
Note that
sin 0 =
2
.
sin 0
p2
1q 2
=
z p2 .
z2
z
Usually we consider z p, so that

1q 2
z p2 = 1.
z
The equation for C0 comes from
Im [f (w) f (w0 )] = 0
where w = u + iv. So in the equation
Im [f (w) f (w0 )] = cosh u sin v v cos 0 (sin 0 0 cos 0 ) = 0.
Thus,
u +
implies
cosh u +
so
v = 0, ,
implies
cosh u
so
v = 0, ,
This gives the line of steepest assent and descent. The orientation of
the curves of ascent and descent are shown in figure 18.6
fig17.6
280

Im w
Descent
Re w
Figure 18.6: Hankel function contours.
18.6
Summary
1. The asymptotic solution of the integral

I() =
dzef (z) g(z)
is
wf (z0 )
" r
I(w) = e
c0
X
1 3 5 (2m 1)
c2
c2m
.
+
+
2m w(2m+1)/2
w
2 w3/2
m=2
2. The asymptotic expansion for the Gamma function is

Z
0
2
2
et tw dt = ew
+
+ . . . .
w
12w3/2
s
3. The asymptotic behavior of the Hankel function is discussed in

section 17.4.
18.7
References
See [Dennery], as well as [Arfken85].
Chapter 19
High Energy Scattering
Chapter Goals:
Derive the fundamental integral equation of scattering.
25 May p1
Derive the Born approximation.

Derive the integral equation for the transition operator.
The study of scattering involves the same equation (Schrodingers) as
before, but subject to specific boundary conditions. We want solutions
for the Schrodinger equation,
i
h
(x, t) = H(x, t),

t
(19.1)
where the Hamiltonian is

H=
eq18.1
h
2 2
+ V (x) = H0 + V.
2m

(x, t) = ei(E/h)t E (x),
(19.2)
using the association E = h

. In particular we want E > 0 solutions, eq18.3
since the solutions for E < 0 are bound states. By substituting equation pr:bound1
281
282
CHAPTER 19. HIGH ENERGY SCATTERING
19.2 into 19.1 we find that E satisfies

(E H)E (x) = 0.
(19.3)
eq18.4
The boundary condition of scattering requires that the wave function

be of the form
E (x) = eiki x + S (x)
(19.4)
eq18.5
where the incident wave number ki is

s
ki =
2mE
ez .
h
2
We interpret equation 19.4 as meaning that the total wave function is

the sum of an incident plane wave eiki x with wavelength = 2/ki ,
and a wave function due to scattering. This solution is illustrated by
the following picture:
E (x) =
25 May p2
- +
eikx
S (x)
At distances far from the scatterer (r 1), the scattered wave
function becomes1
s (x) =
eq18.7

z

eikr
f (ki , kf ; E)
r
for r 1
(19.5)
where the final wave number kf is

s
2mE
pf
=x
kf =
.
h
h
2
The unit vector x simply indicates some arbitrary direction of interest.
Equation 19.5 is the correct equation for the scattered wave function.
The angular function f (ki , kf ; E) is called the form factor and contains
the physical information of the interaction.
1
Again, see most any quantum mechanics text.
19.1. FUNDAMENTAL INTEGRAL EQUATION OF SCATTERING283

For the case of spherical symmetry
f (kf , ki ; E) = f (kf ki ; E).
where kf ki = kf ki cos .
We now would like to formulate the scattering problem for an arbitrary interaction. Thus we look at the relation of the above formulation
to Greens functions. The form of equation 19.3 appropriate for Greens
functions is
(E H)G(x, x0 ; E) = (x x0 ).
Note the minus sign (used by convention) on the left hand side of this G 5/25/88
equation. We solved equation 19.3 by writing (for the asymptotic limit
|x0s | )
0
m eikr
m eikr ikx
G
(x)
=
[e
+ s (x)]
E
2
h2 r0
2
h2 r0
where E (x) satisfies
(E H)E (x) = 0.
The Greens function holds asymptotically since (xx0 ) 0 as |x0 |
. This is the solution of the Schrodinger equation which has the This
fixin
boundary condition of scattering.
needs
pr:bcos1
19.1
Fundamental Integral Equation of Scattering

G 5/25/88
The equation for a general Greens function is

(E H)G(x, x0 ; E) = (x x0 ).
25 May p3
(19.6)
Since H = H0 + V , where H0 =
h2 2 /2m, the free space Greens eq18.13
function satisfies
(E H0 )G0 (x, x0 ; E) = (x x0 ).
As we have seen, the solution to this equation is
See also Jackson, p.224.
284

G0 =
m eikR
,
2
h2 R
(19.7)
where R = |x x0 |. We now convert equation 19.6 into an integral eq18.15

equation. The general Greens function equation can be written
(E H0 )G = (x x0 ) + V (x)G.
eq18.16
(19.8)
We can now use Greens second identity. We define an operator

L0 E H0 .
25 May p4
The operator L0 is hermitian, since both E and H0 are. Recall that

Greens second identity is
Z
(S L0 u) =
(uL0 S),
where we now choose

S = G(x, x0 ; E),
u = G0 (x, x00 ; E),
with the L0 from above. We now have (using equation 19.8)
Z
dxG(x, x0 ; E)(x0 x00 ) =

Z
dxG0 (x, x00 ; E)[(x x0 ) + V (x)G(x, x00 ; E)],
so
G(x00 , x0 ) = G0 (x0 , x00 ) +
dxG0 (x, x00 )V (x)G(x, x0 ).
We can use the fact that G is symmetric (see equation 19.7) to write
00
00
G(x , x ; E) = G0 (x , x ; E) +
dxG0 (x00 , x; E)V (x)G(x, x0 ; E).
So, for x00 x and x x1 , we have

G(x, x0 ; E) = G0 (x, x0 ; E) +
dx1 G0 (x, x1 ; E)V (x1 )G(x1 , x0 ; E).

(19.9)
19.2. FORMAL SCATTERING THEORY

eq18.23
pr:FundInt1
285
This is called the fundamental integral equation of scattering. This

integral is completely equivalent to equation 19.6. We now describe a
way to write equation 19.9 diagrammatically. We establish the following
correspondences.
G(x, x ; E) =
x
G0 (x, x0 ; E) =

G
x0
x0

G
x0
dx1 G0 (x, x1 ; E)V (x1 )G(x1 , x ; E) =

z

x
x1
Thus a line indicates a free Greens function. A dot indicates a potential, and a G in a circle represents the Greens function in the presence
of the potential. The point x1 represents the position of the last interaction. Thus equation 19.9 can be written
25 May p5

=
G
0
x
x
+
0
x
x
z

x1

G
x0
The arrowheads indicate the line of causality. This helps us to

remember the ordering of x0 , x1 and x.
19.2
Formal Scattering Theory
Now we want to derive this equation again more formally. We will

use the operator formalism, which we now introduce. The free Greens pr:OpForm1
function equation is
"
h
2 2
E+
G0 (x, x0 ; E) = (x x0 ),
2m
#
286
where the right hand side is just the identity matrix,

hx|1|x0 i = (x x0 ),
and (1)ij = ij . We also write
h
2 02
= (x x0 ),
G0 (x, x ; E) E +
2m
"
where the operator 2 operates to the right. From these equations we

can write symbolically
[E H0 ]G0 = 1
(19.10)
eq18.25
and
G0 = 1[E H0 ].
eq18.26
(19.11)
This uses the symmetry of G0 operating on x or x0 . Thus our manipulations are essentially based on hermiticity. Because G is also symmetric,
we may also write
[E H]G = 1,
and
G[E H] = 1.
We now want to rederive equation 19.9. We write
25 May p6
[E H]G = 1
and
[E H0 V ]G = 1.
We now multiply on the left by G0 to get
G0 [E H0 V ]G = G0 1 = G0 .
With the aid of equation 19.11 this becomes
G G0 V G = G0 ,
or
G = G0 + G0 V G,
(19.12)
19.2. FORMAL SCATTERING THEORY

eq18.31
287
where the term G0 V G symbolizes matrix multiplication, which is thus

as integral. This is equivalent to equation 19.9, which is what we wanted
to derive.
Note that
hx|G|x0 i = G(x, x0 )
and
hx|V |x0 i = V (x)(x x0 ).
19.2.1
A short digression on operators
If an integral of the form

C(x1 , x2 ) =
dx0 A(x1 , x0 )Bx0 , x2 )
were written as a discrete sum, we would let x1 i, x2 j, and

x0 k. We could then express it as
Cij =
Aik Bkj .
But now A, B, and C are just matrices, so we can express C as a

matrix product C = AB. This can also be viewed as an operator
equation. Quantum mechanically, this can be represented as a product
of expectation values, either for a discrete spectrum,
hi|C|ji =
hi|A|kihk|B|ji,
or for a continuous spectrum,

hx1 |C|x2 i =
dx0 hx1 |A|x0 ihx0 |B|x2 i.
We now show that the form of the fundamental integral of scattering

expressed in equation 19.12 is equivalent to that in equation 19.9. If
we reexpress equation 19.12 in terms of expectation values, we have
hx|G|x0 i = hx|G0 |x0 i + hx|G0 V G|x0 i.
288
By comparing with equation 19.10, we see that the last term can be
written as
0
hx|G0 V G|x i =
=
=
dx1 dx2 hx|G0 |x1 ihx1 |V |x2 ihx2 |G|x0 i
dx1 dx2 hx|G0 |x1 iV (x1 )(x1 x2 )hx2 |G|x0 i
dx1 hx|G0 |x1 iV (x1 )hx1 |G|x0 i.
So by identifying
hx|G|x0 i = G(x, x0 )
hx|G0 |x0 i = G0 (x, x0 )
we have our final result, identical to equation 19.9,
0
G(x, x ) = G0 (x, x ) +
27 May p1
dx1 G0 (x, x1)V (x1 )G(x1 , x0 )
which we obtained using equation 19.12.
19.3
Summary of Operator Method
We started with
H = H0 + V
and the algebraic formulas
(E H)G = 1,
(E0 H)G0 = 1.
We then found that G satisfies the integral equation
G + G0 + G0 V G.
By noting that G(E H) = 1, we also got
G = G0 + GV G0 .
27 May p2
19.3. SUMMARY OF OPERATOR METHOD
289
Derivation of G = (E H)1
19.3.1
The trick is to multiply by G0 . Thus

G(E H0 V ) = 1 G0 ,
and so
G GV G0 = G0 .
Operators are nothing more than matrices. By inverting the equation, we get
1
.
G=
EH
So G is the inverse operator of [E H]. In this context it is useful to
define G in terms of its matrix elements:
G(x, x0 ; E) hx|G(E)|x0 i
where Im E = 0. This arithmetic summarizes the arithmetic of Greens
Second Identity. We also found that the Greens function solves the 27 May p3
following integral equation
G(x, x0 ; E) = G0 (x, x0 ; E) +
dxG0 (x, x1 ; E)V (x1 )G(x1 , x0 ; E).
We were able to express this graphically as well. The other equation

gives
0
G(x, x ; E) = G0 (x, x ; E) +
19.3.2
dxG(x, x1 ; E)V (x1 )G0 (x1 , x0 ; E).
Born Approximation
Suppose the V (x) is small. Then in the first approximation G G0 .

We originally used this to calculate the scattering amplitude f . We 27 May p4
now use perturbation methods to obtain a power series in V .
290
19.4
Physical Interest
We now want to look at E = E + i. We place the source point at

x0 r0 z, where r0 . We can then write the free space Greens
function as
0
m eikr ikx
0
G0 (x, x ; E) =
e
2
h2 r0
where
s
2mE
k=
.
h
2
In this limit the full Greens function becomes, for the fundamental
integral equation of scattering, equation 19.9,
0
eq18.13a
m ekr ikx Z
lim G(x, x ; E) =
e + dx1 G(x, x1 ; E)V (x1 )eikx .
2
0
xr 0
lz
r
2
h
r 0
(19.13)
Note that since
h
2 2
H=
2m
we have
[E H0 ]eikx = 0.
27 May p5
We define
(+)
k (x)
twsedblst
ikx
dx1 G(x, x1 ; E)V (x1 )eikx1 .
(19.14)
Then what we have shown is

0
m eikr (+)
lim
G(x,
x
;
E)
=
(x)
r0
2
h2 r0 l
0
+
where [E H]+
k = 0, and k satisfies outgoing wave boundary condi(+)
tion for scattering. We can get k to any order in perturbation since
we have an explicit expression for it and G.
Now consider the case in which r0 with
1. G = G0 + G0 V G
2. G = G0 + GV G0
19.4. PHYSICAL INTEREST
291
Case 2 implies that

0
m eikr (+)
G=
(x)
2
h2 r0 n
as r0 .
(19.15)
By inserting Eq. (19.15) into 1., we get

0
eq:twoseva
m eikr (+)
m eikr ikx Z
(+)
N (x) =
e
+ G0 (x, x1 ; E)V (x1 )k (x1 ) .
2
2
0
0
2
h r
2
h r

(+)
This implies that k (x) satisfies

(+)
k (x) = eikx +
27 May p6
(+)
G0 (x, x1 ; E)V (x1 )k (x1 ).
(19.16)
This is the fundamental integral expression for . Compare with Eq. twosevst
(19.14).
(+)
We now prove that k satisfies scattering equation with the scattering condition. We use the form of (+) in Eq. (19.14).
(+)
(E H)k
= (E H0 V )eikx
+
dx1 (E H0 V )x G(x, x1 ; E)V (x1 )eikx1

ikx
= V (x)e
dx1 (x x0 )V (x1 )eikx1
= V (x)eikx + V (x)eikx
= 0.
27 May p7
19.4.1
Satisfying the Scattering Condition
We use the form of equation (19.16) to prove that it does satisfy the
scattering condition. Let x = er r where r . We use the result
lime G0 (x, x1 ; E) =
x=r
r
r
m eikr ikf x1
e
2
h2 r
292
where kf = k
er . There is a minus in the exponent since we are taking
the limit as x approaches er rather than
er as in equation (19.15).
Thus the limiting case is
(+)
k (x)
ikx
=e

eikr
m Z
(+)
ikf x1
+
dx1 e
V (x1 )k (x1 ) .
r
2
h2
We further define
(+)
lime k (x) eikx +

x=r
r
r
27 May p8
eikr
f (k, kf ; E)
r
where f is the scattering amplitude to scatter a particle of incident wave

k to outgoing kf with energy E. This is the outgoing wave boundary
condition.
19.5
Physical Interpretation
(+)
We defined the wave function k (x) using equation 19.13 whose components have the following interpretation.
(+)
k (x)
ikx
= e
(+)
dx1 G0 (x, x1 ; E)V (x1 )k (x1 )
incident (x) + scattered (x)

where
eikr
f (k, kf ; E).
|x|=r
r
The physical interpretation of this is shown graphically as follows.
lim
(+)
k (x) =
s (x) =
- +

z

incident (x) scattered (x)
19.6
Probability Amplitude
27 May p9
pr:diffcrsec1
The differential cross section is given by
19.7. REVIEW
293
d
(k kf ) = |f (k, kf ; E)|2
d
where f is the scattering amplitude for scattering with initial momentum p = h

k and final amplitude pf = h
kf from a potential V (x). The
scattering amplitude f is sometimes written
f (k, kf ; E) =
19.7
m
(+)
hkf |V |k i.
2
h2
Review
1 Jun p1
We have obtained the integral equation for the Greens function,

0
G(x, x ; E) = G0 (x, x ; E) +
dxG0 (x, x1 ; E)V (x1 )G(x1 , x0 ; E).
For the case of a distant source we have seen

0
m eikr (+)
lim
G(x,
x
;
E)
=
(x),
x0
z
2
h2 r0 E
0
where
(+)
E (x)
ikx
=e
(+)
dx1 G0 (x, x0 ; E)V (x1 )E (x1 ).
The first term is a plane wave. The integral represents a distorted wave.
(+)
Note that E (x) automatically satisfies the outgoing wave boundary
condition. (This is the advantage of the integral equation approach
over the differential equation approach.) To verify this, we took the
limit x . We also obtained
(+)
E (x) = eikx +
dx1 G(x1 , x0 ; E)V (x1 )eikx1 .
We let E E + i to get a scattering solution,

0
eik|xx |
m
G0 (x, x ; E + i) =
,
0
|x x |
2
h2
0
where |k| =
(+)
2mE/
h2 . We also have shown that E (x) satisfies
(+)
[E H]E (x) = 0.
294

(+)
The wave function E (x) can also be written in the form

eikr
,
r
where we have obtained the following unique expression for f ,
m Z
(+)
f (k, k1 ; E) =
dxeikf x V (x)E (x),
2
2
h
where the integral represents a distorted wave. In particular, the term
eikf x is a free wave with the final momentum, V (x) is the interaction
(+)
potential, and E (x) is the distorted wave. So the integral expression
for f is the overlap of and V with the outgoing final wave. Note
that we have made no use of spherical symmetry. All x contribute, so
we still need short distance behavior even for far distance results. The
differential cross section can be written in terms of f as
(+)
x
n
E (x) eikx + f (k, k1 ; E)
d

= |f |2 .
d kkf
19.8
The Born Approximation
pr:BornAp1
(+)
We now study a particular approximation technique to evaluate E (x)

in
Z
(+)
(+)
ikx
E (x) = e
+ dx1 G0 (x, x0 ; E)V (x1 )E (x1 ).
We assume that the potential is weak so that the distortion, as represented by the integral, is small. The condition that the distortion is
small is
small distortion
(+)
|E (x) eikx | 1.
In this case the potential must be sufficiently small, such that

Z
(+)
dx1 G0 (x, x0 ; E)V (x1 )E (x1 ) 1.
We now introduce the short hand of representing this integral by VB ,

the Born parameter:
VB
(+)
dx1 G0 (x, x0 ; E)V (x1 )E (x1 ) 1.
19.8. THE BORN APPROXIMATION
295
(+)
For VB 1 we may let E (x1 ) be replaced by eikx in f (k, kf ; E). In

this case f becomes fBorn (k, kf ; E), defined by
m Z
f (k, kf ; E) =
dxeix(kxf ) V (x).
2
2
h
In this approximation the cross section becomes
d Born dB
= |fB |2 .
d
d
This is called the first Born approximation. This approximation is valid
in certain high energy physics domains.
We now introduce the matrix notation
Z
(+)
(+)
dxeixkf V (x)k (x) hxf |V |k i.
So in terms of this matrix element the differential cross section is

d
= |f (k, kf ; E)|2 ,
d
where the scattering amplitude is given by
m
(+)
f (k, kf ; E) =
2 hxf |V |k i.
2
h
We also define the wave number transfer q,
q = kf k = (pf pi )/
h.
Thus q is the same as (momentum transfer)/
h. This allows us to write
m
fB =
V (q),
2
h2
where the fourier transformed potential, V (q), is given by
V (q) =
dxeiqx V (x).
So in the first Born approximation, fB depends only on q.

Suppose that q 0. In this case the potential simplifies to
q0
V (q)
dxV (x).
So the first Born approximation just gives us the f dependence on

the average of the potential. Notice that the first Born approximation
looses the imaginary part of f (k, kf ; E) for fB in R, the real numbers.
296

1

>

q

kf

-
-
k
Figure 19.1: Geometry of the scattered wave vectors.
19.8.1
fig18a
Geometry
The relationship between k, kf , q and is shown in figure 19.1. For

the special case of elastic scattering we have
h2
kf2 = k 2 = 2mE/
(elastic scattering).
In this case q 2 is given by

q2 =
=
=
=
(kf k) (kf k)
2k 2 2k 2 cos
2k 2 (1 cos )
4k 2 sin2 (/2).
Thus we have q = 2 sin(/2). We thus know that q will be small for

either k 0 (the low energy limit) or sin(/2) 0 (forward scattering).
19.8.2
Spherically Symmetric Case
In this case the potential V (x) is replaced by V (r). We choose the

z-axis along q and use spherical coordinates. The fourier transform of
the potential then becomes
V (q) =
r2 drdd(cos )eiqr cos V (r)
= 2
Z
0
r drV (r)
d(cos )eiqr cos
4 Z
rdrV (r) sin qr.
q 0
This is a 1-dimensional fourier sine transform.
19.8. THE BORN APPROXIMATION
19.8.3
297
Coulomb Case
We now choose a specific V (r) so that we can do the integral. We

choose the shielded Coulomb potential,
V (r) =
V0 r/a
e
.
r
In the problem set set we use instead of V0 . The parameter characterizes the charge. The fourier sine transform of this potential is
Z
(q) = 4
dr sin rqer/a ,
q 0
and the differential cross section is then

d

= |fB |2
d Born

2

m

=
2 V (q)
2
h
!2
2

1
2 ma
= 4a
.
q 2 a2 + 1
h
2
This is the shielded Coulomb scattering differential cros section in the
first Born approximation, where q = 2k sin(/2). Notice that as
this reduces to Rutherford scattering, which is a lucky accident.
We now look at characteristics of the differential cross section we
have obtained. Most of the cross section contribution comes from
qa = 2k sin(/2) 1. Now if ka 1, then we must require 1,
which means that we can use the small angle approximation. In this
case out dominant cross section condition becomes qa 2ka(/2) 1,
or 1/ka. This gives a quantitative estimation of how strongly forward peaked the scattered wave is. The condition ka 1 corresponds
to the small , or high energy, limit. In this case the wavelength is
much smaller that the particle, which means that most of the scattering will be in the forward direction. We can see how good the first Born
approximation is by evaluating the Born parameter in this limit. We
find
ma2 1
VB = V0 2
1.
k ka
298
In this equation V0 is the strength of the potential and a is the range of

the potential. Notice that ka 1 can make VB 1 even if V0 is large.
Thus we have a dimensionless measure of the strength of the potential.
19.9
Scattering Approximation
We now want to look at the perturbation expansion for the differential

cross section,

d

= |f (k, kf )|2 ,
d kkf
where the scattering amplitude f (k, kf ) is
m Z
(+)
dxeikf x V (x)E (x),
f (k, k1 ; E) =
2
2
h
eq18b1
where the incident wave function is

(+)
E (x)
eq18b2
(19.17)
ikx
=e
dx1 G(x1 , x0 ; E)V (x1 )eikx1 .
(19.18)
(+)
E (x) satisfies the outgoing wave condition. By combining equation

19.17 into 19.18 be obtain
Z
m
dxekf x V (x)ekx
f (k, k1 ; E) =
2
hZ2

dxdx0 ekf x V (x)G(x, x0 ; E)V (x0 )ekx .
The first integral represents a single interaction, while the second integral represents two or more interactions. By introducing the transition
operator, we can simplify the expression for the scattering amplitude,
m Z
f (k, k1 ; E) =
dxdx0 ekf x G(x, x0 ; E)ekx .
2
h2
pr:transOp1
We now define the transition operator T . In function notation it is

T (x, x0 ; E) V (x)(x x0 ) + V (x)G(x, x0 ; E)V (x0 ).
In operator notation, we can rewrite this equation as
T = V + V GV.
19.10. PERTURBATION EXPANSION
299
Thus in matrix notation, our old equation for f ,

f (k, kf ; E) =
m
(+)
2 hxf |V |k i,
2
h
is replaced by
m
hxf |T |ki,
2
h2
Thus we now have two equivalent forms for expressing f .
In the first Born approximation we approximate T = V . T plays the
role in the exact theory what V plays in the first Born approximation.
f (k, k0 ) =
19.10
Perturbation Expansion
pr:pertExp1
We now look at how the transition operator T can be used in diagramatic perturbation theory. We make the following correspondences
between terms in the formulas and the graphical counterparts (these
are the Feynman rules):
pr:FeynRul1
An incoming line:
k
HH
j
HHx represents eikx .
An outgoing line:
kf
x

*

A vertex point:
represents eikf x .
u represents V (x).
A free propagator:
x1
-
A circled G:
x0
x2
represents G0 (x2 , x1 ).
x
- Gm- represents G(x, x0 ).
Thus we can write the transition operator matrix element as

k 0
kf
Hjk x k
x
j
f + HH

HH
HHxu *
*
hxf |T |ki = H
u
m
u

G
The first diagram represents the first Born approximation, which corresponds to a single scatterer. The second diagram represents two or
300
more scatterer, where the propagation occurs via any number of interactions through G.
The integral equation for the full Greens function,
G(x, x0 ; E) = G0 (x, x0 ; E) +
dx1 G(x, x1 ; E)V (x1 )G(x1 , x0 ; E),
has the following symbolic representation:

x0
- Gm- =
x0
-
x
x1
x
+ - uGm-
where x0 is the source point, x is the field point, and x1 is one of the
interaction points.
19.10.1
Perturbation Expansion
In matrix language the integral equation for the full Greens function
is
G = G0 + GV G0 ,
which implies
G = G0 (1 V G0 )1 .
Thus the following geometric series gives the solution to the integral
equation,
G = G0 (1 + V G0 + (V G0 )(V G0 ) + (V G0 )(V G0 )(V G0 ) + ).
In symbolic notation, this expansion corresponds to
x0
- Gm-
x0
-
x0 x
x x0 x1 x2 x
+ - u1+ - u- u-
x0 x
x
x
x
+ - u1- u2- u3+ .
We could also write the series expansion in integral notation. In this
case the third order in V term, (V G0 )(V G0 )(V G0 ), is (writing right to
left)
Z
dx1 dx2 dx3 G0 (x, x3 )V (x3 )G0 (x3 , x2 )V (x2 )G0 (x2 , x1 )V (x1 )G0 (x1 , x0 ),
19.10. PERTURBATION EXPANSION
301
where, for example,

G0 (x3 , x2 ) =
m eik|x2 x3 |
.
2
h2 |x2 x3 |
Think of these terms as multiply scattered terms.

Now we can use this series to get a perturbation expansion for the
scattering amplitude f , that is, for the matrix element hxf |T |ki. In
symbolic language it is
is this correct?
hxf |T |ki =
=
k kf
k 0
HH
Hj
j

HHx
HHxu *
u
+ H
Gm-
kf
x

*
u

kf
k 0
kf
k kf
k 0
Hj
HH
Hj
j

HHxu - xu1- x
HHx
HHxu - x

*
u*
+ H
u*
u
+ H
kf
k 0
HH

HHxu - xu1- xu2- x

u*
+ .
+ j
To convert this to integral language we note that, for example, the

fourth Born approximation term is
k 0 x
kf
HH
x2 x
j
1

H x
*
Hu -
u-
u-
u

In integral notation this is expressed as

Z
dxdx0 dx1 dx2 eikf x V (x)G0 (x, x2 )V (x2 )G0 (x2 , x1 )V (x1 )G0 (x1 , x0 )ekx .
We must integrate over all space since each of the interaction points
may occur at any place.
19.10.2
Use of the T -Matrix
An alternative approach is to eliminate all direct reference to G without perturbation theory. We then obtain an integral equation for the
transition matrix. By using
G = G0 (1 V G0 )1 ,
we have
V G = V G0 (1 V G0 )1 ,
302
so we can write the transition matrix as

T =
=
=
=
V + V G0 (1 V G0 )1 V
[1 + V G0 (1 V G0 )1 ]V
[(1 V G0 ) + V G0 ](1 V G0 )1 V
(1 V G0 )1 V.
This provides us with a new solution for T :

T = (1 V G0 )1 V.
We can write this as an integral equation, which would have the operator form
(1 V G0 )T = V,
or
T = V + V G0 T.
This gives us another Lippman/Schwinger equation. Notice that T =
(1 V G0 )1 V may be expanded in a power series in V just as was the
previous expression for G.
19.11
Summary
1. The fundamental integral equation of scattering is

G(x, x0 ; E) = G0 (x, x0 ; E) +
19.12
dx1 G0 (x, x1 ; E)V (x1 )G(x1 , x0 ; E).
References
See [Neyfeh, p360ff] for perturbation theory.
Appendix A
Symbols Used
hS, ui the brackets denote an inner product, 13.
as a superscript, represents complex conjugation, 13.
nabla, the differential operator in an arbitrary number of dimensions, 7.
A1 , A2 constants used in determining the Greens function, 28.
a the horizontal displacement between mass points on a string; an arbitrary position on the string, 2, the left endpoint of a string,
6.
a1 , a2 constants used in discussion of superposition, 23.
B1 , B2 constants used in determining the Greens function, 28.
b the right endpoint of a string 6.
b(x) width of a water channel, 108.
C a constant used in determining the Greens function, 29.
c left endpoint used in the discussion of the -function, 24; constant
characterizing velocity, 39, 45.
D a constant used in determining the Greens function, 29.
303
304
APPENDIX A. SYMBOLS USED
d the differential operator; right endpoint used in the discussion of the

-function, 24.
p change in momentum, 87.
ui the transverse distance between adjacent points (ui ui1 ) on a
discrete string, 4.
x the longitudinal distance between adjacent points on a discrete
string, 4.
(x x0 ) the delta function, 24, 129, 161.
mn the Kroneker delta function, 36.
E energy, 74, 143.
e = 2.71 .
a small distance along the string, 27.
F (x, t) the external force on a continuous string, 1.
Fcd the force over the interval [c, d], used in the discussion of the function, 24.
Fielastic the elastic force on the ith mass point of a discrete string, 3.
Fiext the external force on the ith mass point of a discrete string, 3.
Fiyi the transverse force at the ith mass point on a string due to tension,
3.
Ftot the total force on the ith mass point of the string.
f (x) is the external force density divided by the mass density at position x, 4.
f (x0 ) a finite term used in discussion of asymptotic Greens function,
42.
f1 , f2 force terms used in discussion of superposition, 23.
305
f (, k) the scattering amplitude for a field observer from an incident
plane wave, 214.
f(, r0 , k) scattering amplitude, 214.
G(x, xk ; 2 ) the Greens function for the Helmholtz equation, 26.
GA the advances Greens function, 87.
GS the scattered part of the steady state Greens function, 184.
GR the retarded Greens function, 86.
Gm (r, r0 ; ) reduced Greens function, 132.
the Fourier transform of the Greens function, 88, the Laplace transG
form of the Greens function, 147.
gn (x, x0 ) asymptotic coefficient for Greens function near an eigen value,
41.
angular difference between x and x0 used in scattering discussion,
185.
H the Hamiltonian, 195
(2)
(1)
(x) the first and second Hankel functions, 79.
(x), Hm
Hm
h(x) equilibrium height of a surface wave, 108.

hl (x) the spherical Hankel function, 178.
ha (t) the effective force exerted by the string: Fa /a , 6.
hS (t) same as ha (t), generalized for both endpoints, 8.
h
the reduced Planks constant, 74.
I() a general integral used in discussion of method of steepest descent,
265.
Im another Bessel function, 80.
306
i the index of mass points on a string, 2.

i unit vector in the x-direction, 128.
J Jacobian function, 128.
j(r) the quantum mechanical current density, 227.
jinc the incident flux, 227.
jl (x) the spherical Bessel function, 178.
jn heat current, 144.
j unit vector in the y-direction, 128.
Km another Bessel function, 80.
k the wave number, 38.
k 2 a short hand for V / used in infinite string problem, 63.
ki the spring constant at the ith mass point, 3.
the thermal diffusity, 151.
a the effective spring constant exerted by the string at endpoint a:
ka /a , 6.
L0 linear operator, 5.
L centrifugal linear operator, 162.
L the angular momentum vector, 207.
l dimension of length, 3.
l the direction along the string in the positive x direction, 7.
an arbitrary complex number representing the square of the frequency continued into the complex plane, 27; wavelength of surface waves, 108.
307
n nth eigen value for the normal mode problem, 37.
(m )
(m)
the nth eigenvalue of the reduced operator L0
n
, 133.
m dimension of mass, 3.
mi the mass of the particle at point i on the discrete string, 2.
m eigenvalues for circular eigenfunctions, 131.
N the number of mass particles on the discrete string, 2; the number
of particles intercepted in a scattering experiment, 227.
nl (x) the spherical Neumann function, 178.
n
the outward normal, 7.
solid angle, 161.
angular frequency, 9.
n the natural frequency of the nth normal mode, 32.
p momentum, 74, 207.
solution of the Klein Gordon equation, 75; total response due to a
plane wave scattering on an obstacle, 185.
0 incident plane wave used in scattering discussion, 185.
angular coordinate, 128, 160.
n (xi , t) the normal modes, 38.
quantum mechanical wave function, 195
R(r) function used to obtain Bessels equation, 178.
Re take real value of whatever term imediately follows.
r radial coordinate, 128.
308
S the surface (i.e., endpoints) of a one dimensional string, 7; an

arbitrary function used in the derivation of the Greens identities,
13.
S(x) cross sectional area of a surface wave, 108.
the cross section, 227.
(x) the mass density of the string at position x 4.
T (x, x0 ; E) transition operator, 298.
t time, dimension 3, variable, 3.
i the tension on the segment between the (i1)th and ith mass points
on a string, 2.
parameter in RBC for the heat equation, 251.
the angle of the string between mass points on a discrete string, 3;
angle in parameterization of complex plane, 63.
u(x, t) transverse displacement of string, 5; displacement of a surface
wave from equilibrium height, 108.
u0 (x) an arbitrary function used in the derivation of the Greens identities, 13.
u0 (x) value of the transverse amplitude at t = 0, 8.
u0 (x, ) steady state in free space due to a point source, 184.
u1 (x) value of the derivative of the transverse amplitude at t = 0, 8.
u1 , u2 functions used in discussion of superposition, 23.
u1 solution of the homogeneous fixed string problem, 46.
ui the vertical displacement of the ith mass particle on a string, 2.
(m )
u(m)
n (r) the nth eigenfunction of L0
, 133.
um
l (x) the normalized -part of the spherical harmonic, 164.
309
uscat the scattered part of the steady state response, 198.
u1 modified solution of the homogeneous fixed string problem, 47.
V (x) the coefficient of elasticity of the string at position x, 1.
Veff the effective potential, 206.
W (u1 , u2 ) the Wronskian, 30.
Xl coefficient of the scattered part of the wave relative to the incident
part, 187, 219.
x continuous position variable, 4.
x< the lower of the position point and source point, 30.
x> the higher of the position point and source point, 30.
x0 the location of the -function disturbance, 24.
xi discrete position variable, 4.
xk the location of the -function disturbance, 26.
Ylm (, ) the spherical harmonics, 164.
z(x) hight of a surface wave, 108.
310
Bibliography
[1] Arfken, George B. Mathematical Methods for Physicists. Academic
Press, 1985.
[2] Barton, Gabriel, Elements of Greens Functions and Propagation:
potentials, diffusion, and waves. Oxford, 1989, 1991.
[3] *Boas, Mary, Mathematical Methods in the Physical Sciences.
[4] Carslaw, Horatio Scott & Jaeger, J. C. Conduction of heat in solids.
Oxford, 1959, 1986.
[5] Dennery, Phillippi and Andre Krzywicki, Mathematics for Physicists.
[6] Fetter, Alexander L. & Walecka, John Dirk. Theoretical Mechanics
of Particles and Continua, Chapters 913. McGraw-Hill, 1980.
[7] *Griffiths, David, Indroduction to Electrodynamics, Prentice-Hall,
1981.
[8] *Halliday, David and Robert Resnick, Physics, John Wiley, 1978.
[9] *Jackson, David, Classical Electrodynamics, John Wiley, 1975.
[10] Morse, Philip M. & Feshbach, Herman. Methods of Theoretical
Physics. McGraw-Hill, 1953.
[11] Neyfeh, Perturbation Methods.
[12] Stakgold, Ivar. Boundary Value Problems of Mathematical Physics.
Macmillan, 1967.
311
312
BIBLIOGRAPHY
[13] Stakgold, Ivar. Greens Functions and Boundary Value Problems.

John Wiley, 1979.
[14] *Symon, Keith R., Mechanics, Addison-Wesley, 1971.
Index
addition formula 212
advanced Greens function 87
all-spce problem 117, 174
analytic 46, 266
angular momentum 207
associated Legendre polynomial
168
asymptotic limit 49
Babenets principle 194
Bessels equation 79
Born approximation 294
bound states 281
boundary conditions 5, 111, 116,
145, 173; of scattereing
283
boundary value problem 1
branch cut 45, 60
Bromwich integral 246
Cartesian coordinates 128
Cauchys theorem 54
Cauchy-Riemann equations 266
causality 87
characteristic range 83
classical mechanics (vs. quantum
mechanics) 202, 207
closed string 6, discrete 37
coefficient of elasticity 4
completeness relation 51, 57, 76,
131, 169
Condon-Shortley phase convention 170

conservation of energy 144, 213
continuity condition 28
Coulomb potential 208
cross section 227
cutoff frequency 38, 66
De Broglie relation 203
degeneracy 39
delta function 24, 129
differential cross section 292
differential equation 3
diffraction 191
Dirichlet boundary conditions 8
discrete spectrum 49
dispersion relation 38
divergence 129, 161
effective force 7
effective spring constant 7
eigen function 32
eigenfunction expansion 131
eigen value problem 28, 68, 121,
133, 134, 140
eigen vector 32
elastic boundary conditions 6, 116
limiting cases, 7
elastic force 3
elastic media 8
elastic membrane 109
313
314
energy 74, 143
energy levels 196
even dimensions, Greens function in, 260
equations of motion 2
expansion theorem 57, 172
experimental scattering 208, 226
exterior problem 117, 122, 174
external force 3
far-field limit 208, 235
Feynman rules 299
forced oscillation problem 31, 73,
118
forced vibration 3
Fourier coefficient 58
Fourier integral 78, see also expansion theorem
Fourier Inversion Theorem see inverse Fourier transform
Fourier-Bessel transform 83
Fourier transform 88
Fredholm equation 40
free oscillation problem 32
free space problem 151, 188
free vibration 3
fundamental integral equation of
scattering 285
Gamma function 273
Gaussian 153, 155
gradient 129, 161
general response problem 103, 117,
119
general solution, heat equation,
246
Generalized Fourier Integral 59
geometrical limit of scattering 230
Greens first identity 14, 119
INDEX
Greens function for the Helmholtz
equation, 26
Greens reciprocity principle 30,
Greens second identity 15, 119
Hamiltonian 195
Hankel function 79; asymptotic
form, 276
hard sphere, scattering from a,
231
heat conduction 143
heat current 144
heat equation 146
Helmholtz equation 9, 26
Hermitian analyticity 43
Hermitian operator 17, 119
holomorphic see analytic
homogeneous equation 28, 45
Huygens principle 194
impulsive force 86
infinite string 62
initial conditions 8
initial value problem 92, 119
inner product 13
inverting a series 270
interior problem 116, 122, 174
inverse Fourier transform 91
Jacobian 128
Kirchhoffs formula 191
Klein Gordon equation 74
Lagrangian 110
Laplace transform 147
Laplaces equation 268
Legendres equation 166
Legendre polynomial 168
Leibnitz formula 167
linear operator 5, 24
linearly independent 31
INDEX
mass density 4
membrane problem 138
method of images 122, 191
momentum operator 207
natural frequency 26, 32, 37, 42
natural modes 32, 37
Neumann boundary conditions
8
Newtons Second Law 3
normal modes 32, 37, 117, 134
normalization 44, 58, 135, 169
odd dimensions, Greens function
in, 259
open string 6
operator formalism 285
optical theorem 231
orthogonal 36
orthonormality 36, 58, 164, 169,
171
oscillating point source see forced
oscillation problem
outward normal 7
partial expansion 131
partial differential equation 5
periodic boundary conditions 6,
111, 116
perturbation expansion 299
plane wave 199, 213, 239
polar coordinates 128
poles 44
positive definite operator 20
potential energy 20
potential theory 186
principle of superposition 24, 131
quantum mechanical scattering
197
quantum mechanics 195
315
radiation 81
Rayleigh quotient 40
recurrence relation 167
reduced linear operator 132
regular boundary conditions 8
residues 44
retarded Greens function 86, 120,
136
Rodrigues formula 168
scattered Greens function 210
scattering Amplitude 211
scattering from a sphere 223
scattering wave 209
Schrodinger equation 39, 195
self-adjoint operator 52, 119
singular boundary conditions 8
shallow water condition 108
singularity 54
sound waves, radiation of, 232
specific heat 143
spectral theory 42
spherical coordinates 160
spherical harmonics 170
steady state scattering 183
steady state solution 9, 135, 196,
234
steepest descent, method of, 265
string 1
superposition see principle of
surface waves 108
temperature 143
tension 2
transition operator 298
transverse vibrations 2
travelling wave 38
wave propagation 66
wedge problem 136
316
Wronskian 30
INDEX

(Baker M., Sutlief S.) Green's Functions in Physics

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Baker M., Sutlief S.) Green's Functions in Physics

Uploaded by

Copyright:

Available Formats

Greens Functions in Physics

2.4.2 A Note on Potential Energy . . . .

4 Properties of Eigen States

4.6.2 Summary, Interp. & Asymptotics . .

5 Steady State Problems

Semi-Infinite String with Fixed End . . . . .

12 Steady State Scattering

12.4 Scattering from a Cylinder . . . . . . . . . . . . . . . . . 189

17 The Wave Equation

Method of Steepest Descent

19 High Energy Scattering

The pointed string . . . . . . . . . . . . . . . . . . . . . 27

The closed string with discrete

6.1 The contour L in the -plane. . . . . . . . . . .

Water waves moving in channels. . . . . . . . . . . . . . 108

9.1 The region R as a circle with radius a. . . . . . . . . . . 130

The schematic representation of a scattering experiment.

16.1 Closed contour around branch cut. . . . . . . . . . . . . 250

Contour C & deformation C0 with point z0 .

19.1 Geometry of the scattered wave vectors. . . . . . . . . . 296

A Note About Copying:

CHAPTER 1. THE VIBRATING STRING

Figure 1.1: A string with mass points attached to springs.

We consider a massless string with equidistant mass points attached. In

Forces on the String

1.1. THE STRING

where dim(ki ) = (m/t2 ). This situation can be visualized by imagining pr:Fel1

Equations of Motion for a Massless String

CHAPTER 1. THE VIBRATING STRING

Equations of Motion for a Massive String

In the limit we take a 0, N , and define their product to be

The limiting case allows us to redefine the terms of the equations of

V (xi ) = coefficient of elasticity of the media;

1.2. THE LINEAR OPERATOR FORM

This allows us to write 1.3 as

This is a partial differential equation. We will look at this problem in eq1diff

The Linear Operator Form

We define the linear operator L0 by the equation

We can now write equation (1.9) as

This is an inhomogeneous equation with an external force term. Note eq1waveone

To obtain a unique solution for the differential equation, we must place

CHAPTER 1. THE VIBRATING STRING

Figure 1.2: A closed string, where a and b are connected.

Case 1: A Closed String

which is the condition that the ends meet, and

Case 2: An Open String

For an elastically bound open string we have the boundary condition

The homogeneous terms of this equation are a u

1.3. BOUNDARY CONDITIONS

The boundary condition at b can be similarly defined:

For a more compact notation, consider points a and b to be elements

for x on S, for all t.

It is also worthwhile to consider the limiting cases for an elastically

CHAPTER 1. THE VIBRATING STRING

The boundary condition a 0 corresponds to an elastic media, and pr:ElMed1

then the boundary conditions are called regular boundary conditions.

Thus regular boundary conditions correspond to the case in which there

The complete description of the problem also requires information about

for a < x < b

u(x, t)|t=0 = u1 (x)

1.4. SPECIAL CASES

In the case that a , we require that u(x) have a finite limit as

Oscillatory External Force