Problems In Quantum Field Theory With

Fullyworked Solutions 1st Edition Gelis download

https://ebookbell.com/product/problems-in-quantum-field-theory-
with-fullyworked-solutions-1st-edition-gelis-34479932

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Problems In Quantum Mechanics And Field Theory With Mathematical

Modelling Aleksander V Chichurin Elena M Ovsiyuk Viktor M Redkov

https://ebookbell.com/product/problems-in-quantum-mechanics-and-field-
theory-with-mathematical-modelling-aleksander-v-chichurin-elena-m-
ovsiyuk-viktor-m-redkov-58948076

Fundamental Problems In Quantum Field Theory T Fujita

https://ebookbell.com/product/fundamental-problems-in-quantum-field-
theory-t-fujita-32734368

A Complete Solution To Problems In An Introduction To Quantum Field

Theory Zhongzhi Xianyu

https://ebookbell.com/product/a-complete-solution-to-problems-in-an-
introduction-to-quantum-field-theory-zhongzhi-xianyu-5639910

The Canonical Operator In Manyparticle Problems And Quantum Field

Theory V P Maslov

https://ebookbell.com/product/the-canonical-operator-in-manyparticle-
problems-and-quantum-field-theory-v-p-maslov-49194670
Problem Book In Quantum Field Theory Voja Radovanovi

https://ebookbell.com/product/problem-book-in-quantum-field-theory-
voja-radovanovi-46499378

Problem Book In Quantum Field Theory 2nd Edition Voja Radovanovi Auth

https://ebookbell.com/product/problem-book-in-quantum-field-
theory-2nd-edition-voja-radovanovi-auth-4208158

Problems In Quantum Mechanics I I Goldman V D Krivchenkov

https://ebookbell.com/product/problems-in-quantum-mechanics-i-i-
goldman-v-d-krivchenkov-51211056

Problems In Quantum Mechanics With Solutions 2nd Ed 2017 Demilio

https://ebookbell.com/product/problems-in-quantum-mechanics-with-
solutions-2nd-ed-2017-demilio-38432006

Problems In Quantum Mechanics Third Edition Dover Books On Physics 3rd

Edition D Ter Haar Editor

https://ebookbell.com/product/problems-in-quantum-mechanics-third-
edition-dover-books-on-physics-3rd-edition-d-ter-haar-editor-38496998
To Kanako, Nathan and Simon.
Reality is that which, when you stop believing in it, doesn’t go away.

P HILIP K. D ICK

How to build a universe that doesn’t fall apart two days later (1978)
Contents

List of Figures and Tables x

Preface xii

Acknowledgements xiv

Notation and Conventions xv

1 Quantum Field Theory Basics 1

Introduction 1
1 Relationship between the Equations of Motion of φ and φin 11
2 Little-Group Elements for Massless Particles 12
3 S-matrix in Terms of φin and φout 15
4 Constraints on the S-matrix from Causality 17
5 Landau Equations for Soft and Collinear Singularities 22
6 Multi-loop Integration in D Dimensions 28
7 Weinberg Convergence Theorem 31
8 Electron Anomalous Magnetic Moment 34
9 Ward–Takahashi Identities and Lorentz Invariance 38
10 Equivalence Principle and Lorentz Invariance 41
11 Lee–Nauenberg Theorem 42
12 Classical External Field Approximation 49
13 Subleading Soft Radiation in Scalar QED 53
14 Low–Burnett–Kroll Theorem 57
15 Coherent States in Quantum Field Theory 62
16 Running Couplings in a Two-Field Scalar Field Theory 64
viii CONTENTS

17 Solution of the Running Equation at Two Loops 68

2 Functional Methods 71
Introduction 71
18 Weyl Quantization 77
19 Combinatorial Interpretation of the Legendre Transform 79
20 Coherent State Path Integral 83
21 Coherent State Path Integral (Continued) 86
22 Spin Coherent States and the Berry Phase 90
23 Zeta Function Regularization of the Klein–Gordon Operator 101
24 Casimir Zero-Point Energy 103
25 The Feynman Diagrams behind the Casimir Force 106
26 Gross–Neveu Model 110
27 D’Hoker–Gagné Formula 114
28 Stochastic Quantization 118
29 Complex Langevin Equation 123

3 Non-Abelian Fields 130

Introduction 130
30 Basic su(n) Identities 137
31 More su(n) Identities 138
32 More su(n) Identities (Continued) 140
33 An su(2) Identity 142
34 Area Derivative of a Wilson Loop 144
35 Endpoint Derivative of a Wilson Line 145
36 Makeenko–Migdal Loop Equation 148
37 Classical Non-Abelian Particles and the Wong Equations 151
38 Wong Equations from the Worldline Formalism 154
39 Non-Abelian Stokes’s Theorem 157
40 QED in the ’t Hooft–Veltman Gauge 162
41 LSZ Reduction Formula and Gauge Invariance 167
42 Large-N Expansion in SU(N) Yang–Mills Theory 171
43 Soft Radiation and Antenna Effects in QCD 175
44 An Identity Obeyed by On-Shell QCD Amplitudes 183
45 Low–Burnett–Kroll Theorem in QCD 184
46 Globally Gauge Invariant Operators 188
47 Banks–Zaks Fixed Points 191
48 Instantons and Tunneling in Quantum Mechanics 193
49 Instantons as Transitions between Pure Gauges in Yang–Mills Theory 196
50 Vacua of Yang–Mills Theory 199
CONTENTS ix

4 Scattering Amplitudes 205

Introduction 205
51 Enumeration of Tree Graphs 212
52 Six-Point − − − + + + Amplitude 214
53 Two Spinor-Helicity Identities 216
54 Auxiliary Vector Dependence of the Polarization Vectors 218
55 Polarization Sum in the Spinor-Helicity Formalism 220
56 Photon Decoupling Identity 221
57 Scalar–Antiscalar–Gluon Amplitudes 223
58 Color Decomposition with a Quark–Antiquark Pair 226
59 Quark–Antiquark–Gluon Amplitudes 229
60 Quark–Antiquark–Gluon Amplitudes (Continued) 233
61 Gluon–Gluon Scattering in the Spinor-Helicity Framework 238
62 Soft Limit of Gluon Amplitudes 242
63 Collinear Limit of Gluon Amplitudes 247
64 Enumeration of One-Loop Graphs 252
65 Ultraviolet Finiteness of One-Loop All-Plus Amplitudes 258
66 Rationality of One-Loop All-Plus Amplitudes 259
67 One-Loop Four-Point + + + + Amplitude from Generalized Unitarity 261
68 One-Loop Five-Point + + + + + Amplitude from Generalized Unitarity 267

5 Lattice, Finite T , Strong Fields 276

Introduction 276
69 Banks–Casher Relation 283
70 Nielsen–Ninomiya Theorem 285
71 Wigner Transform of a Convolution Product 290
72 Center Symmetry and Deconfinement 293
73 Photon Hard Thermal Loop from Classical Particles 297
74 Gluon Hard Thermal Loop from Colored Classical Particles 301
75 Landau–Pomeranchuk–Migdal Effect 305
4
76 Microcanonical Equilibration in φ Scalar Theory 312
77 Combinatorics of Multi-particle Production 315
78 Generating Functional for Particle Production 320
79 Correlations in the Schwinger Mechanism 324
80 Schwinger Mechanism from Bogoliubov Transformations 331
81 Unruh Effect, Hawking Radiation 339
82 Anderson Localization 347
83 Scaling Theory of Anderson Localization 350

Index 353
Figures and Tables

Figures

1.1 Domains Ω1 and Ω2 , with Ω2 under the causal influence of Ω1 . 19

1.2 Domains Ω1,2 , such that Ω2 does not causally influence Ω1 . 19
1.3 Space-like separated domains and their transformation under boosts. 20
1.4 One-loop example of application of the Landau equations. 26
1.5 The β function that controls the running of the ratio ρ/λ. 68

2.1 Notation for the angles of the spherical triangle (n  n

n  0 ). 94
2.2 Geometrical interpretation of the Berry phase. 95
2.3 Derivation of the area of a spherical triangle. Triangulation of a pentagon by
three spherical triangles. 97
2.4 Triangulations of a sphere. 98
2.5 Definition of the angle θv at a vertex v of a triangulation. 99
2.6 Gauss mapping of a vertex of the triangulation of a surface. 100
2.7 Setup for the calculation of the Casimir energy. 104
2.8 Diagrams contributing to the free energy in the presence of two walls. 108
2.9 Moments obtained from the fixed point of the Fokker–Planck equation. 126

3.1 Addition of a small appendix γ  to a loop γ, at the base point x. 144

3.2 Closed path γ with a self-intersection at the point x. 150
3.3 Setup for the non-Abelian Stokes’s theorem. 158
3.4 Topological invariance of the Euler characteristic. 174
3.5 Polytopes with the Euler characteristic of a 2-sphere. 174
3.6 Euler characteristics of polytopes with faces removed. 175
LIST OF FIGURES AND TABLES xi

3.7 Deformation of a polytope with two holes into a torus. 175

3.8 Allowed regions for soft gluon emission from a singlet dipole. 179
3.9 Sketch of a β function with β0 < 0 and β1 > 0. 192
3.10 Double-well potential. Classical solution of the Euclidean time equation of
motion. 195

4.1 Contributions to the six-point color-ordered amplitude 1− 2− 3− 4+ 5+ 6+ . 215

4.2 Color-ordered amplitude with a quark–antiquark line. 227
4.3 Cuts of the one-loop four-point + + + + amplitude. 260
4.4 Color-ordered graphs contributing to the amplitude Ascalar
4 (1+ 2+ 3+ 4+ ). 267

5.1 Two-dimensional vector fields with zeroes of index +1 and −1. 288
5.2 Illustration of the Poincaré–Hopf theorem in two dimensions. 289
5.3 Special vector field used in the proof of the Poincaré–Hopf theorem. 290
5.4 Illustration of the Landau–Pomeranchuk–Migdal effect. Feynman graphs
contributing to the photon self-energy. 311
5.5 Phase-space evolution of an ensemble of classical fields. 315
5.6 Illustration of Rindler space. 341
5.7 β function for the conductance in d = 1, 2, 3 dimensions. 352

Tables

4.1 Number of graphs contributing to n-point tree amplitudes. 213

4.2 Couplings of a gluon to an on-shell quark–antiquark current. 231
4.3 Number of n-point one-loop graphs. 255
4.4 Number of n-gluon graphs with a ghost or quark loop. 256
4.5 Numbers of tree and one-loop n-gluon color-ordered graphs. 258
Preface

This project grew out as an extension to my previous Quantum Field Theory book (hereafter
referred to as book I), in the form of an additional set of solved problems. The starting
observation was that most textbooks (and mine is no exception) have a general inclination
towards the exposition of the concepts, rather than the more practical aspects. This is of course
quite understandable for such a vast subject, where one needs to absorb a quite large volume
of concepts before becoming operational. The intended goal of the present book is to help fill
the gap between theory and applications by providing a text almost exclusively geared towards
actual practice (but occasionally new concepts are also introduced).
The set of problems included in this book cover most of the subjects treated in book I,
and expose the reader to a broad variety of techniques. Occasionally, the same question is
addressed by various methods, in order to shed light on it from different perspectives. The
problems proposed in this set cover a broad range of difficulties and fall into several categories:

• More sophisticated and lengthier applications of the techniques and results of book I,
that are too long to be treated as examples in a textbook
• Problems that extend book I towards new topics and concepts that are too specialized to
fit reasonably in a textbook
• Classic results of quantum field theory (QFT) presented in a more modern fashion and
with uniform notation, with the goal of making them more accessible than from a reading
of the original literature
• More recent results, presented in a way that is accessible to the readers of book I; these
are meant to be a bridge between the material of a typical textbook and contemporary
research articles.

When deciding which problems to include in this collection, I made the deliberate choice to
keep almost exclusively problems that can be worked out analytically “by hand” in a reasonable
amount of time. Occasionally, straightforward but tedious calculations have been avoided by
the use of a computer algebra system (for these problems, PYTHON notebooks will be provided
as a separate online resource). This prejudice is of course an important limitation since these
PREFACE xiii

computer tools are a common aid in contemporary theoretical research, but my impression is
that the didactical virtue of working out simpler problems by hand is higher.
Although this book has an obvious lineage with book I (e.g., the two books share the
same notation), a significant effort has been made to ensure that it is self-contained and can
be used on its own. Each chapter starts with a reminder of the important concepts and tools
relevant for the problems of that chapter, but an important word of caution is in order here:
these introductions are meant to be a concise refresher for a reader who has already studied the
corresponding subjects, but they are not an appropriate source for learning a subject for the first
time. This book is also self-contained by the inclusion of detailed solutions to all the problems.
When necessary, a quick exposition of some relevant mathematical tools is included in the
solutions, to avoid a detour via a mathematical textbook (this is intended to give plausibility
to a given mathematical statement, not to provide a thorough and rigorous description of the
underlying mathematics).
The intended readership of this book is of course primarily students who are in the process
of learning quantum field theory, as well as their instructors. Roughly speaking, Chapters
1 and 2 cover the topics one would usually learn in a first QFT course. Chapter 3, on non-
Abelian gauge theories, tends to be a bit more advanced and is often treated in a second course.
Chapters 4 and 5 deal with more specialized subjects, respectively the newly developed tools
for calculating scattering amplitudes, and aspects of quantum field theory that are on the border
of many-body physics. Hopefully, more experienced readers will also find the book useful,
both for the discussion of these more advanced topics and for a more modern treatment of
some classic QFT results.
Notation and Conventions

We list here some notation and conventions that are used throughout this book:

• c = 1 : length = time, energy = momentum = mass

• h̄ = 1 : momentum = wavenumber, energy = frequency

• pμ ≡ (p0 , p) : 4-momentum; p : three-dimensional spatial components


• Ep ≡ p2 + m2 : positive on-shell energy of a particle of momentum p and mass m
• L : Lagrangian density (more rarely, the Lagrangian itself)
• H : Hamiltonian; S : action

• μ, ν, ρ, σ, . . . (more rarely α, β, γ, . . . ): Lorentz indices in Minkowski space

• i, j, k, l, . . . : Lorentz indices in Euclidean space
• a, b, c, d, . . . : group indices in the adjoint representation
• i, j, k, l, . . . : group indices in the fundamental representation

• μνρ··· : Levi–Civita symbol. In situations where it makes sense to raise or lower the
indices (e.g., for Lorentz indices in Minkowski space), the normalization convention is
with lowered indices, 012··· = +1

• The normalization of creation and annihilation operators is defined so that the canonical
commutation relation reads [ap , a†q ] = (2π)3 2Ep δ(p − q) (except in Problems 80 and
81, where it is convenient to omit the factor 2Ep ). Their dimension is (mass)−1 .
 CHAPTER 1

Quantum Field Theory Basics

Introduction

This chapter is devoted to basic aspects of quantum field theory, ranging from the foundations to
perturbation theory and renormalization, and is limited to the canonical formalism (functional
methods are treated in Chapter 2) and to the traditional workflow (Lagrangian → Feynman rules
→ time-ordered products of fields → scattering amplitudes) for the calculation of scattering
amplitudes (the spinor-helicity formalism and on-shell recursion are considered in Chapter
4). The problems of this chapter deal with questions in scalar field theory and quantum
electrodynamics, while non-Abelian gauge theories are discussed in Chapter 3.

Non-interacting Field Theory

A non-interacting field theory may be defined by a quadratic Lagrangian. In the simplest case
of a scalar field theory, it reads
   

L ≡ d3 x 12 ∂μ φ ∂μ φ) − 12 m2 φ2 . (1.1)

Such a Lagrangian defines a dynamical system with infinitely many degrees of freedom, corre-
sponding to the values taken by φ(x) at every point x of space. The momentum canonically
conjugate to φ(x) is given by
∂L
Π(x) ≡ = ∂0 φ(x), (1.2)
∂(∂0 φ(x))
which leads to the Hamiltonian
   
 2
H ≡ d x Π(x)∂ φ(x) − L = d3 x 12 Π2 + 12 ∇φ + 12 m2 φ2 .
3 0
(1.3)
2 1 QUANTUM FIELD THEORY BASICS

From the Hamiltonian or Lagrangian, one obtains the equation of motion of the field, which in
the present example reads
 
x + m2 φ(x) = 0, (1.4)

known as the Klein–Gordon equation. Generic real solutions of this linear equation are
superpositions of plane waves:
  
d3 k ∗ −ik·x
φ(x) = α e + α k eik·x
, (1.5)
(2π)3 2Ek k


where Ek ≡ p2 + m2 is the dispersion relation associated with the wave equation (1.4),
and αk is a function of momentum that depends on the boundary conditions imposed on the
solution.
Canonical quantization consists in promoting the coefficients αk , α∗k into annihilation and
creation operators ak , a†k that obey the following commutation relations:
 
ap , a†q ≡ (2π)3 2Ep δ(p − q). (1.6)
   
The normalization in Eqs. (1.5) and (1.6) is chosen so that H, a†p = Ep a†p and H, ap =
−Ep ap , which means that a†p increases the energy of the system by Ep while ap decreases it
by the same amount. As a consequence, this setup describes a collection of non-interacting
particles. The commutation relation (1.6) implies the following equal-time commutation
relation between the field operator and its conjugate momentum:
 
φ(x), Π(y) = i δ(x − y), (1.7)
x0 =y0

which one may view as the quantum version of the classical Poisson bracket between a
coordinate and its conjugate momentum.

Interacting Field Theory and Interaction Representation

Interactions are introduced via terms of degree higher than two in the Lagrangian:
   

L ≡ d3 x 12 ∂μ φ ∂μ φ) − 12 m2 φ2 −V(φ) . (1.8)
   
L0 , non-interacting theory interactions

(In order to have a causal theory, the potential V(φ) must be a local function of the field φ(x);
see Problem 4.) In the presence of interactions, the Klein–Gordon equation of motion becomes
 
x + m2 φ(x) + V  (φ(x)) = 0. (1.9)

Since the degree of V(φ) is higher than two, this equation is non-linear, which induces a mixing
between the Fourier modes of the field and prevents writing its solutions as superpositions of
plane waves.
By assuming that the interactions are turned off at large times, x0 → ±∞, we may
define free fields φin and φout that coincide with the interacting field φ of the Heisenberg
INTRODUCTION 3

representation, respectively when x0 → −∞ and x0 → +∞. For instance, φ and φin are
related by
φ(x) = U(−∞, x0 ) φin (x) U(x0 , −∞),
 t2
U(t2 , t1 ) ≡ T exp −i dx0 d3 x V(φin (x)) . (1.10)
t1

In this representation, the time dependence of the field φ(x) is split into a trivial one that comes
from the free field φin and the time evolution operator U that depends on the interactions.
Since they are free fields obeying Eq. (1.4), φin and φout can be written as superpositions of
plane waves, with coefficients ap,in , a†p,in and ap,out , a†p,out , respectively. These two sets of
creation and annihilation operators define two towers of Fock states, i.e., states with a definite
particle content at x0 = −∞ and x0 = +∞, respectively.

Lehmann–Symanzik–Zimmermann Reduction Formulas

Experimentally measurable quantities, such as cross-sections, may be related to correlation

functions of the field operator as follows. An intermediate step involves the transition ampli-
tudes between in and out states,
 
q1 · · · qn out p1 · · · pm in ≡ (2π)4 δ pi − qj T(q1···n |p1···m ), (1.11)
i j

in terms of which a cross-section in the center of momentum frame is given by


1 2
σ12→1···n center of = √ dΓn (p1,2 ) T(q1,··· ,n |p1,2 ) , (1.12)
momentum 4 s |p1 |
d3 qj  
with dΓn (p1,2 ) ≡ (2π)4 δ p1 + p2 − qj , s ≡ (p1 + p2 )2 .
(2π) 2Eqj
3
j j

In turn, the transition amplitudes from in to out states are expressed in terms of expectation
values of time-ordered products of field operators by the Lehmann–Symanzik–Zimmermann
(LSZ) reduction formulas:
 m
im+n
q1 · · · qn out p1 · · · pm in = m+n d4 xj e−ipi ·xi (xi + m2 )
Z 2 i=1
 n
× d4 yj eiqj ·yj (yj + m2 ) 0out T φ(x1 ) · · · φ(xm )φ(y1 ) · · · φ(yn ) 0in , (1.13)
j=1

where Z is the wavefunction renormalization factor.

Generating Functional, Feynman Propagator

The calculation of expectation values of time-ordered products of field operators is usually

organized by encapsulating them in a generating functional
δn Z[j]
0out T φ(x1 ) · · · φ(xn ) 0in = , (1.14)
iδj(x1 ) · · · iδj(xn ) j≡0
4 1 QUANTUM FIELD THEORY BASICS


with Z[j] ≡ 0out T exp i d4 x j(x)φ(x) 0in (1.15)
   
δ
= exp − i d x V
4
0in T exp i d4 x j(x)φin (x) 0in .
iδj(x)
 
Z0 [j], non-interacting theory
(1.16)
The last factor, the generating functional of the non-interacting theory, is a Gaussian in the
auxiliary source j:
  
1
Z0 [j] = exp − d xd y j(x)j(y)GF (x, y) ,
4 4 0
(1.17)
2

where G0F (x, y) is the free Feynman propagator, which can be expressed in various equivalent
ways:
G0F (x, y) = 0in T φin (x)φin (y) 0in (1.18)
  
d3 p −ip·(x−y) +ip·(x−y)
= θ(x 0
− y0
) e + θ(y0
− x 0
) e , (1.19)
(2π)3 2Ep
i
G0F (p) = 2 . (1.20)
p − m2 + i0+

Feynman Rules of Scalar Field Theory

The effect of interactions can be calculated order-by-order by expanding the first exponential in
Eq. (1.16). The successive terms of this expansion are obtained from a diagrammatic expansion,
where each diagram is converted into a formula by means of Feynman rules. Below we list
these rules in momentum space, for a scalar field theory:
1. Draw all the graphs with as many external lines as field operators in the correlation
function, and a number of vertices equal to the desired order. The vertices allowed in
these graphs must have valences equal to the degrees of the terms in V(φ). Graphs with
multiple connected components need not be considered in the calculation of scattering
amplitudes.
2. A 4-momentum k is assigned to each internal line of the graph, and the associated
Feynman rule is a free propagator G0F (k):
p
i
= .
p2 − m2 + i0+
No propagator should be assigned to the external lines of a graph when calculating a
scattering amplitude (because of the factors  + m2 in the reduction formulas).
 
3. For an interaction n! φ , each vertex of valence n brings a factor −iλ(2π)4 δ
λ n
i ki ,
where the ki are the momenta incoming into this vertex:

= −iλ.
INTRODUCTION 5

3. All the internal momenta that are not constrained by the delta functions at the vertices
should be integrated over with a measure d4 k/(2π)4 . In a connected graph with nI
internal lines and nV vertices, there are nL = nI − nV + 1 of them, which is also the
number of loops in the graph.

4. Each graph must be weighted by a symmetry factor, defined as the inverse of the
order of the discrete symmetry group of the graph (assuming interaction terms properly
symmetrized, as in V(φ) = φn /n!).

Dimensional Regularization

The momentum integrals that correspond to loops in Feynman diagrams may be divergent at
large momentum. Convergence may be assessed from the superficial degree of divergence
of a graph, ω(G) ≡ 4nL − 2nI for a graph with nL loops and nI internal lines in a scalar
field theory with quartic coupling in four spacetime dimensions: the graph G is convergent
if ω(G) < 0 and the superficial degree of divergence of all its subgraphs is negative as well.
In order to safely manipulate possibly divergent loop integrals, the first step is to introduce a
regularization, i.e., a modification of the Feynman rules such that all loop integrals become
well defined. Many regularization methods are possible: Pauli–Villars subtraction, lattice
discretization, momentum cutoff, dimensional regularization.
Dimensional regularization, based on the observation that loop integrals calculated in an
arbitrary number D of dimensions have an analytical continuation which is well defined at all
D’s except a discrete set of values, is particularly adapted to analytical calculations. With this
regularization scheme, some common (Euclidean) loop integrals are given by

 D  
dD kE 1 Δ 2 −n Γ n − D2
=   ,
(2π)D (k2E + Δ)n D Γ n
(4π) 2
 D D  
d kE kμE kνE gμν Δ 2 +1−n Γ n − 1 − D2
=   ,
(2π)D (k2E + Δ)n 2 D Γ n
(4π) 2
 D D  
d kE kμE kνE kρE kσE gμν gρσ + gμρ gνσ + gμσ gνρ Δ 2 +2−n Γ n − 2 − D
=   2
,
(2π)D (k2E + Δ)n 4 D Γ n
(4π) 2
 D
d kE kμE 1 · · · kEμ2n+1
= 0. (1.21)
(2π)D (k2E + Δ)n

The first of these equations is obtained by integration in D-dimensional spherical coordinates,

and the subsequent equations follow from Lorentz invariance.

Renormalization

The list of correlation functions that exhibit ultraviolet divergences can be obtained from
the superficial degree of divergence ω(G) (except in situations where a symmetry produces
cancellations that cannot be anticipated by power counting). For a scalar field theory with a
quartic coupling, one has ω(G) = 4 − nE + (D − 4)nL in D spacetime dimensions, where
6 1 QUANTUM FIELD THEORY BASICS

nE is the number of external points and nL the number of loops. The Weinberg convergence
theorem states that a graph is ultraviolet convergent if and only if the superficial degree of
divergence of the graph, and of any of its subgraphs, is negative.
In D = 4 spacetime dimensions, ω(G) is negative for all correlation functions with
nE > 4 points, implying that only a finite number of correlation functions have intrinsic
divergences. Moreover, these divergent correlation functions are the expectation values of the
operators already present in the Lagrangian, (∂μ φ)2 , φ2 , φ4 . The divergences that appear in
these functions can be subtracted order-by-order via a redefinition of their coefficients in the
Lagrangian, i.e., Z (this one is usually not explicit in the bare Lagrangian because it is set to
1), m2 and λ, respectively. Such a quantum field theory is called renormalizable.
In D > 4 dimensions, ω(G) increases with the number of loops at fixed nE . This
implies that any correlation function exhibits intrinsic ultraviolet divergences beyond a certain
loop order. Removing these divergences would require that one adds arbitrarily many new
terms in the Lagrangian, reducing considerably the predictive power of such a theory (but
it may nevertheless be of some use in an effective sense, at low loop orders). It is called
non-renormalizable.
When D < 4, the superficial degree of divergence of any correlation function eventually
becomes negative after a certain loop order. These theories have a finite number of ultraviolet
divergent Feynman graphs, whose calculation is sufficient to determined the renormalized
Lagrangian once and for all. These theories are called super-renormalizable.
For general interactions in arbitrary dimensions, the above criteria can be expressed in
terms of the mass dimension of the prefactor that accompanies the operator in the Lagrangian.
The corresponding operator is renormalizable if the mass dimension of its coupling constant is
zero, non-renormalizable if this dimension is negative, super-renormalizable if it is positive.

Renormalization Group

In a renormalized quantum field theory, one may still freely choose the renormalization scale
μ at which the conditions that define the parameters of the renormalized Lagrangian (masses,
couplings, etc.) are imposed. Physical results should not depend on this scale. The dependence
of various renormalized quantities with respect to μ is controlled by the Callan–Symanzik
equations, also known as renormalization group equations. For the renormalized n-point
correlation function Gn , this equation reads
 
μ∂μ + β∂λ + γm m∂m +nγ Gn = 0, (1.22)
 
≡ Dμ
1 ∂ ln(Z) ∂λ ∂ ln(m)
with γ ≡ , β≡ , γm ≡ (1.23)
2 ∂ ln(μ) ∂ ln(μ) ∂ ln(μ)
(γ is called an anomalous dimension, and β is the β function). Physical quantities are invariant
under the action of Dμ , i.e., under the simultaneous change of the scale μ and of the parameters
Z, λ, m as prescribed by the above differential equations (the solutions λ(μ) and m(μ) are
called the running coupling and running mass, respectively). The curves (Z(μ), λ(μ), m(μ))
in the parameter space of the renormalized theory, along which physical quantities are invariant,
define a vector field called the renormalization flow.
From the Callan–Symanzik equation satisfied by the propagator, (Dμ + 2γ)G2 = 0, one
obtains the corresponding flow equations for the pole mass mP (defined from the value of p2
INTRODUCTION 7

at the pole of the propagator) and for the residue Z at the pole:

Dμ mP = 0, (Dμ + 2γ) Z = 0. (1.24)

Thus, a n-point scattering amplitude An ∼ Z−n/2 Gn also satisfies Dμ An = 0. Amputated

correlation functions Γn ≡ (G2 )−n Gn obey (Dμ − nγ)Γn = 0.

Spin-1/2 Fields

The representation
 μ ν of the Lorentzμ algebra of lowest even dimension is defined byμ theν 
generators
Mμν ≡ i
γ , γ , where the γ are the Dirac 4 × 4 matrices, which obey γ , γ = 2 gμν .
1/2 4
 i 
Under a Lorentz transformation Λ ≡ exp − 2 ωμν Mμν , a Dirac spinor is a four-component
field that transforms as
 
ψ(x) → exp − 2i ωμν Mμν 1/2
ψ(Λ−1 x). (1.25)

In the absence of interactions, such a field obeys the – Lorentz invariant – Dirac equation,
 μ 
iγ ∂μ − m ψ = 0, (1.26)

which can be obtained as the equation of motion that results from the following Lagrangian:
 
L = ψ iγμ ∂μ − m ψ, with ψ ≡ ψ† γ0 . (1.27)

The canonical quantization of a free spinor (i.e., a solution of the Dirac equation (1.26))
consists in replacing its Fourier coefficients by creation and annihilation operators:
  
d3 p † +ip·x −ip·x
ψ(x) ≡ d v
sp s (p)e + b u
sp s (p)e . (1.28)
s=±
(2π)3 2Ep

Since ψ is not Hermitian, the two operators in this decomposition need not be mutual conjugates
(except in the special case of Majorana fermions). The spinors us , vs are a basis of free spinors
in momentum space defined by
 μ   μ 
γ pμ − m us (p) = 0, γ pμ + m vs (p) = 0, (1.29)
u†r (p)us (p) = 2Ep δrs , v†r (p)vs (p) = 2Ep δrs . (1.30)

For the Hamiltonian of this system to have a well-defined ground state, these creation and
annihilation operators must obey anti-commutation relations. The non-zero ones read

{dsp , d†s  p  } = {bsp , b†s  p  } = (2π)3 2Ep δss  δ(p − p  ), (1.31)

or, equivalently,

{ψα (x), ψ†β (y)} = δαβ δ(x − y). (1.32)

x0 =y0
8 1 QUANTUM FIELD THEORY BASICS

The Dirac Lagrangian has a U(1) symmetry, ψ → e−iα ψ, which by Noether’s theorem leads
to a conserved current Jμ ≡ ψγμ ψ and conserved charge
 
d3 p  † †

Q ≡ d3 x J0 = (2π) 2Ep
3 bsp bsp − dsp dsp . (1.33)
s=±

Spin-1 Fields

A vector field Aμ (x) is a four-component field that transforms as Aμ (x) → Λμ ν Aν (Λ−1 x)

under a Lorentz transformation Λ. The simplest such (massless) field is the electromagnetic
field, whose Lagrangian reads
L = − 14 Fμν Fμν , with Fμν ≡ ∂μ Aν − ∂ν Aμ . (1.34)
The corresponding equation of motion, ∂μ Fμν = 0, has several remarkable properties:
• Gauge invariance: for any function θ, Aμ − ∂μ θ is a solution if Aμ is a solution.
• The field A0 is not dynamical, but given by a constraint from the spatial components Ai .
• Only the transverse (i.e., transverse to the momentum ki in Fourier space) components
of Ai are constrained by the equation of motion.
The unphysical redundancy due to gauge invariance is removed by imposing a gauge condition
– e.g., ∂μ Aμ = 0 (Lorenz gauge), ∂i Ai = 0 (Coulomb gauge), A0 = 0 (temporal gauge) – leaving
only two independent dynamical solutions per Fourier mode. The quantization of the vector
field Aμ amounts to replacing the coefficients in its Fourier decomposition by creation and
annihilation operators:

d3 p  † μ∗ μ −ip·x

A (x) ≡
μ
a (p) eip·x
+ a λp λ (p) e , (1.35)
(2π)3 2|p| λp λ
λ=1,2
 
with the canonical commutation relation aλp , a†λ  p  = (2π)3 2|p|δλλ  δ(p − p  ) and where
the objects μλ (p) are polarization vectors that encode the Lorentz indices of a vector field of
polarization λ and momentum p. The polarization vectors may depend on the choice of gauge
condition, but always satisfy pμ μλ (p) = 0.

Quantum Electrodynamics

The conserved charge of the Dirac fermions can be interpreted as an electrical charge. In-
teractions between these fermions and photons are introduced by minimal coupling, i.e., by
requesting that the modified Dirac Lagrangian is invariant under spacetime-dependent U(1)
transformations, ψ(x) → e−ieθ(x) ψ(x). This is achieved by replacing the ordinary derivative
by a covariant derivative, Dμ ≡ ∂μ − ieAμ . Perturbation theory in QED has the following
Feynman rules:
p
/ + m)
i(p p i Cμν (p)
= μ ν =
p − m2 + i0+
2 p2 + i0+

= −i e γμ fermion
loop
= (minus sign)
μ
INTRODUCTION 9

The numerator Cμν in the photon propagator depends on the gauge fixing (for instance,
Cμν = −gμν in Feynman gauge).

Ward–Takahashi Identity

A crucial property of QED amplitudes with external photons is the Ward–Takahashi identity,
namely
pμ Γ μ··· (p, . . . ) = 0, (1.36)
where Γ μ··· (p, . . . ) is an amplitude amputated of its external propagators, containing a photon
of momentum p with Lorentz index μ. The dots represent the other external lines, either
photons or charged particles. The conditions of validity of this identity, which follows from
the conservation of the electrical current, are the following:

• All the external lines corresponding to charged particles must be on-shell, and contracted
in the appropriate spinors if they are fermions.
• The gauge fixing condition must be linear in the gauge potential, in order not to have
three- and four-photon vertices.

The Ward–Takahashi identity plays a crucial role in ensuring that QED scattering amplitudes
are gauge invariant, and that they fulfill the requirements of unitarity despite the presence of
non-physical photon polarization in certain gauges.

Unitarity, the Optical Theorem and Cutkosky’s Cutting Rules

The time evolution operator from x0 = −∞ to x0 = +∞ (also called the S-matrix) is unitary,
SS† = 1. Writing it as S ≡ 1 + iT to separate the interactions, this implies the optical theorem:
1 2
Im αin T αin = αin T βin .
2
states β

This relation implies that the total probability of scattering from the state α to any state β (with
at least one interaction) equals twice the imaginary part of the forward scattering amplitude
α → α. In perturbation theory, the imaginary part of a transition amplitude Γ can be obtained
by means of Cutkosky’s cutting rules:
1  
Im Γ = Γ γ,
2 cuts γ

where a cut is a fictitious line that divides the graph

  into two subgraphs, with at least one
external leg on each side of the cut. A cut graph Γ γ is calculated with the following rules:

• Left of the cut: use the propagator G0++ (p) = i

p2 −m2 +i0+
and the vertex −iλ,
−i
• Right of the cut: use the propagator G0−− (p) = p2 −m2 −i0+
and the vertex +iλ,

• The propagators traversing the cut should be G0+− (p) = 2π θ(−p0 )δ(p2 ).
10 1 QUANTUM FIELD THEORY BASICS

About the Problems of this Chapter

• Problem 1 establishes a crucial relationship between the field operators φ (Heisenberg

representation) and φin (interaction representation), namely that the former obeys the
interacting equation of motion if the latter obeys the free Klein–Gordon equation.
• In Problem 2, we derive an explicit form of the elements of the little group for massless
particles. This is then used in Problem 9 in order to show that, in a theory with massless
spin-1 bosons, the Lorentz invariance of scattering amplitudes implies a property that
may be viewed as a weak form of the Ward–Takahashi identity. This observation, due to
Weinberg, is extended to gravity in Problem 10.
• Problem 3 establishes some formal relationships between various expressions for the
time evolution operator and the S-matrix. Then, Problem 4 shows that the expression
for the S-matrix as the time-ordered exponential of a local interaction term is to a large
extent a consequence of causality.
• In Problem 5, we derive a set of conditions, known as the Landau equations, for a given
loop integral to have infrared or collinear singularities. An explicit multi-loop integration
is studied in Problem 6, which provides another point of view on these conditions.
• Problem 7 establishes Weinberg’s convergence theorem in the simple case of scalar
field theory, a crucial result in the discussion of renormalization since it clarifies the
role of the superficial degree of divergence in assessing whether a particular diagram is
ultraviolet divergent.
• The electron anomalous magnetic moment is calculated at one loop in Problem 8. This
is a classic QED calculation of great historical importance, which has now been pushed
to five loops and provides one of the most precise agreements between theory and
experiment in all of physics.
• Problem 11 derives the Lee–Nauenberg theorem, an important result about soft and
collinear singularities which states that such divergences are removed by summing
transition probabilities over degenerate states, thereby providing a link between the
finiteness of a quantity and its practical measurability.
• In Problem 12, we discuss the external classical field approximation, thanks to which a
heavy charged object may be replaced by its classical Coulomb field.
• Problems 13 and 14 are devoted to a derivation of the Low–Burnett–Kroll theorem,
a result that states that the emission probability of a soft photon is proportional to the
probability of the underlying hard process, at the first two orders in the energy of the
emitted photon.
• Coherent states are introduced in Problem 15 and their main properties established.
They will be discussed further in Problems 20, 21 and 22.
• Problems 16 and 17 study the running coupling in a scalar field theory with two fields,
and in a QCD-like theory at two-loop order.
1 RELATIONSHIP BETWEEN THE EQUATIONS OF MOTION OF φ AND φin 11

1. Relationship between the Equations of Motion of φ and φin Recall that the field
operators in the Heisenberg representation (φ) and in the interaction representation (φin ) are
related by
φ(x) = U(−∞, x0 ) φin (x) U(x0 , −∞). (1.37)
The goal of the problem is to show that, if V(φ) is the interaction potential, this implies the
following relationship between the left-hand sides of their respective equations of motion:
 
( + m2 )φ(x) + V  (φ(x)) = U(−∞, x0 ) ( + m2 )φin (x) U(x0 , −∞),
t
provided that U(t2 , t1 ) ≡ T exp −i t12 d4 x V(φin (x)). This is an important consistency check,
since it implies that φin is a free field while φ evolves as prescribed by the self-interaction
term in the Lagrangian.
1.a Apply a derivative ∂μto (1.37). Note that spatial derivatives
 do not act on the U’s. In par-
ticular, show that ∂0 U(−∞, x0 )φin (x)U(x0 , −∞) = U(−∞, x0 )Πin (x)U(x0 , −∞).
How did the terms coming from the time derivative of the U’s cancel?
1.b Apply a second time derivative to this result, to obtain
 
∂20 U(−∞, x0 )φin (x)U(x0 , −∞)
 
 
= U(−∞, x ) ∂0 φin (x) − i d3 y Πin (x), V(φin (x0 , y)) U(x0 , −∞).
0 2

1.c Calculate the commutator on the right-hand side (one may prove that if [A, B] is an
object that commutes with all other operators, then [A, f(B)] = f  (B) [A, B]).

1.a Let us start from

φ(x) = U(−∞, x0 )φin (x)U(x0 , −∞).
Since the evolution operators depend only on time, we have trivially
 
(−∇2 + m2 )φ(x) = U(−∞, x0 ) (−∇2 + m2 )φin (x) U(x0 , −∞),
and the main difficulty is to deal with the time derivatives. The first time derivative reads
   
∂0 φ(x) = ∂0 U(−∞, x0 ) φin (x)U(x0 , −∞) + U(−∞, x0 )φin (x) ∂0 U(x0 , −∞)
 
+ U(−∞, x0 ) ∂0 φin (x) U(x0 , −∞)
 
= i U(−∞, x0 ) φin (x), I(x0 ) U(x0 , −∞)
 
+ U(−∞, x0 ) ∂0 φin (x) U(x0 , −∞),

where we denote I(x0 ) ≡ − d3 x V(φin (x0 , x)). The first line contains an equal-time
commutator of φin with some functional of φin , which is zero, leaving only the non-vanishing
term of the second line.
12 1 QUANTUM FIELD THEORY BASICS

1.b A second differentiation with respect to time gives

 
∂20 φ(x) = i U(−∞, x0 ) ∂0 φin (x), I(x0 ) U(x0 , −∞)
 
+ U(−∞, x ) ∂0 φin (x) U(x0 , −∞).
0 2

1.c The commutator in the first line is an equal-time commutator between the canonical
momentum ∂0 φin and a functional of the field φin . In order to evaluate it, we need the following
result:

[A, f(B)] = [A, B] f  (B) , valid when [[A, B], B] = 0.

This can be shown by using the Taylor series of f(B), by first showing by recursion that
[A, Bn ] = n[A, B]Bn−1 . Then, we can write

   
i ∂0 φin (x), I(x ) = −i d3 y ∂0 φin (x), V(φin (x0 , y))
0


 
= −i d3 y ∂0 φin (x), φin (x0 , y) V  (φin (x0 , y))
 
−iδ(x−y)

= −V (φin (x)).

Now, using U(−∞, x0 )V  (φin (x))U(x0 , −∞) = V  (φ(x)), we get

 
( + m2 )φ(x) − L  (φ(x)) = U(−∞, x0 ) ( + m2 )φin (x) U(x0 , −∞).

Therefore, the left-hand sides of the equations of motion for the Heisenberg representation
(interacting) field φ and for the interaction representation (free) field φin are related by a
unitary transformation identical to the formula that relates the field operators themselves.

2. Little-Group Elements for Massless Particles The little group is the subgroup of the
Lorentz group that leaves a fixed reference vector qμ invariant. In this problem, we derive a
particularly convenient explicit form of the elements of the little group in the case where qμ is
the light-like vector qμ ≡ (ω, 0, 0, ω).

2.a First, show that an infinitesimal little-group transformation of this kind can be written as
follows:

R ≈ 1 − iθ J3 + iα1 B1 + iα2 B2 ,

with three generators J3 , B1 , B2 (the first one being the generator of rotations about the
third direction of space) that one should determine explicitly. Check also that they satisfy
the following commutation relations (after an appropriate normalization):
 3 i  1 2
J , B = i ij Bj , B , B = 0.
2 LITTLE-GROUP ELEMENTS FOR MASSLESS PARTICLES 13

2.b Show that any finite element R of the massless little group can be written as
⎛ ⎞ ⎛ ⎞
β2 +β2 β2 +β2
1 0 0 0 1 + 12 2 β1 β2 − 12 2
⎜ ⎟ ⎜ ⎟
⎜0 cos θ − sin θ 0⎟ ⎜ β1 1 0 −β1 ⎟
Rμ ν =⎜ ×⎜
⎝0 sin θ cos θ 0⎟
⎠ ⎝ β2 0 1 −β2 ⎟.
⎠
0 0 0 1 β21 +β22 β21 +β22
2 β1 β2 1 − 2

Hint: use the exact Baker–Campbell–Hausdorff formula:

  1
 
ln e eiX iY
= i X + i dt F et adY eadX Y,
0

  ln(z)
where adX (Y) ≡ −i X, Y and F(z) ≡ ,
z−1
in order to exponentiate the infinitesimal form.

2.a This question is not difficult to solve by “brute force,” i.e., by looking for the most
general little-group transformation

Rμ ν ≡ δ μ ν + ω μ ν

such that ωμ ν qν = 0 for all values of the index μ, with the additional constraint that ωμν =
0 is antisymmetric (so we have a legitimate infinitesimal Lorentz transformation). If we
parameterize
⎛ ⎞
0 a b c
ωμν ≡ ⎝−a 0
−b −d 0 f ,
d e⎠
−c −e −f 0

the condition that qν ≡ (ω, 0, 0, ω) is invariant is equivalent to

c = 0, a = e, b = f,

implying that there is a three-parameter family of ωμ ν ’s that fulfill all the requirements:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 0 0 0 0 1 0 0 0 0 1 0
ωμ ν = θ⎝00 10 −1 0⎠ ⎝1 0 0 −1⎠
0 0 + α1 0 0 0 0 + α2 1 0 0 −1
⎝0 0 0 0 ⎠
0 0 0 0 0 1 0 0 0 0 1 0
 12 μ  10 
13 μ
 20 μ
= −i θ M ν + i α1 
M  −M ν + i α2  M  − M23 ν .
J3 ≡ B1 ≡ B2
 
(The identification in the second line follows from ωμν = − 2i ωαβ Mαβ μν , valid for the
spin-1 representation of the Lorentz algebra, where (Mαβ )μν = i(δα μ δβ ν − δα ν δβ μ ).) Note
that the first term corresponds to rotations about the x3 axis, which trivially leaves invariant
any vector whose only non-zero spatial component is along the third direction.
14 1 QUANTUM FIELD THEORY BASICS

The announced commutation relations can be checked by an explicit evaluation of the

corresponding matrix products. Alternatively, they can also be obtained from
J3 = M12 , B1 = M10 − M13 , B2 = M20 − M23 ,
and by using the defining commutation relation of the Lorentz algebra,
 μν 
M , Mρσ = i(gνρ Mμσ − gμρ Mνσ ) − i(gνσ Mμρ − gμσ Mνρ ).

2.b Given the infinitesimal form of little-group transformations for massless particles de-
rived above, any finite little-group transformation R can be obtained by exponentiating the
inifinitesimal ones:
3 +α 1 +α 2)
R ≡ ei(−θJ 1B 2B
.

Note first that the factor on the left of the proposed formula is nothing but the rotation e−iθJ ,
3

i.e., a rotation by an angle θ about the x3 axis, which affects only the coordinates 1, 2:
⎛ ⎞
1 0 0 0
 −iθJ3 μ 0 cos θ − sin θ 0⎠.
e ν = ⎝0 sin θ cos θ 0
0 0 0 1
Thus, the form proposed in the statement of the problem suggests that R may also be written as
R = e−iθJ ei(β1 B
3 1 +β B2 )
2
.
Our ansatz for the form of the second factor is based on the case θ = 0 (in this case, we
should of course have βi = αi , but this is not necessarily true for θ = 0). Verifying that
the two expressions for R are equivalent could in principle be performed by calculating both
expressions and equating them in order to find the relationship between the coefficients α1,2
and β1,2 , but this is rather challenging.
A much more efficient method is to use the exact form of the Baker–Campbell–Hausdorff
formula, recalled in the statement of the problem. In the present case, the commutation
relations among J3 , B1 , B2 lead to
 
ad−θJ3 (βi Bi ) = −θ βi ij Bj , ead−θJ3 βi Bi = βi e−θ ij Bj , adβi Bi (Bj ) = 0,
and the Baker–Campbell–Hausdorff formula gives
 i
 θ 
ln e−iθJ eiβi B = −iθJ3 − iβi −θ
3
Bj .
e − 1 ij
Using the fact that ij jk = −δik , we have
e−θ − 1 = cos θ − 1 − sin θ,
 i  
ln e−iθJ eiβi B = −iθJ3 − i θβi (cosθ − 1 − sin θ)−1 ij Bj .
3

 
−αj

This formula shows that ei(−θJ +αi B ) = e−iθJ eiβi B with the following relationship between
3 i 3 i

the coefficients α1,2 and β1,2 :

 
βi = θ−1 sin θ + (1 − cos θ) ij αj .
(Note that βi = αi when θ → 0, as expected trivially in this limit.)
3 S-MATRIX IN TERMS OF φin AND φout 15

i i 1 2
In order to calculate the factor eiβi B , one should first note that eiβi B = eiβ1 B eiβ2 B
since B1 and B2 commute. Using the explicit representations of B1,2 , simple algebra shows
that
⎛ ⎞
1 0 0 −1
 1 2 μ  2 2 μ
(iB ) ν = (iB ) ν = ⎝0 0 0 0⎠
0 0 0 0 ,
1 0 0 −1
and
 μ  μ
(iB1 )n ν = (iB2 )n ν = 0 for n ≥ 3.
Therefore, we have
⎛ β2 β2
⎞
1+ 1 β1 0 − 21
 μ ⎜ β1 2
 μ 1 0 −β1 ⎟
=⎜ ⎟
1
eiβ1 B ν = 1 + iβ1 B1 + 12 β21 (iB1 )2 ν ⎝ 0 0 1 0 ⎠,
β21 β21
2 β1 0 1 − 2

⎛ β2 β2
⎞
1+ 2 0 β2 − 22
 μ ⎜ 0 2
 μ 1 0 0 ⎟
=⎜ ⎟
2
eiβ2 B ν = 1 + iβ2 B2 + 12 β22 (iB2 )2 ν ⎝ β2 0 1 −β2 ⎠ ,
β22 β22
2 0 β2 1 − 2

and finally
⎛ β2 +β2 β2 +β2
⎞
1 + 12 2 β1 β2 − 12 2
 μ ⎜ β1 1 0 −β1 ⎟
=⎜ ⎟
i
eiβi B ν ⎝ β2 0 1 −β2 ⎠ ,
β21 +β22 β21 +β22
2 β1 β2 1 − 2

which establishes the announced result.

3. S-matrix in Terms of φin and φout Given an interaction Lagrangian LI , the field
operators in the Heisenberg representation (φ) and in the interaction representation (φin ) are
related by means of a time evolution operator
  t2  
Uin (t2 , t1 ) ≡ T exp i d4 x LI φin (x)
t1

expressed in terms of the free field φin (we have added a subscript in to this evolution operator,
in order to recall that it is defined in terms of φin , since we are about to introduce its counterpart
defined in terms of φout ). Likewise, we define a similar evolution operator in terms of φout , the
field operator in the interaction picture that coincides with the Heisenberg picture at x0 = +∞:
  t2  
Uout (t2 , t1 ) ≡ T exp i d4 x LI φout (x) .
t1
16 1 QUANTUM FIELD THEORY BASICS

3.a Show that Uin (+∞, −∞) = Uout (+∞, −∞). In other words, the S-matrix (i.e., the
time evolution operator over the entire time range) does not depend on whether it is
defined in terms φin or φout .

3.b Are Uin and Uout identical in general? Find the relationship between the two.

3.c Show that the S-matrix is also given by S = Uin (x0 , −∞)Uout (+∞, x0 ), for any inter-
mediate time x0 . Note that, on the surface, this expression does not seem to be properly
time-ordered. Why is it nevertheless a correct formula?

3.a Recall the relationship between the interacting field φ and the free field φin of the
interaction representation:

φ(x) = Uin (−∞, x0 ) φin (x) Uin (x0 , −∞),

  t2  
Uin (t2 , t1 ) ≡ T exp i d4 x LI φin (x) . (1.38)
t1

(Since we shall shortly write the analogous relationship with φout , it is important to have a
subscript in on the evolution operator to avoid confusion, since it depends on φin .) By taking
the limit x0 → +∞ in this equation, and using the fact that in this limit the interacting field
becomes identical to the free field φout , we obtain a first relationship between φin and φout :

φout (x) = Uin (−∞, +∞) φin (x) Uin (+∞, −∞).
x0 →+∞

Strictly speaking, this limiting procedure gives a relationship between the two fields only for
large x0 . Then, we use the fact that two fields obeying the same equation of motion (here,
φin and φout both obey the Klein–Gordon equation) and identical in some region of time are
equal at all times (this argument relies on the uniqueness of the solutions of the Klein–Gordon
equation, if their value and that of their first time derivative are prescribed at some time).
Therefore, the above equation is in fact valid at all times.
Note that the right-hand side of this equation depends only on φin , but in a completely
non-linear and non-local fashion because of the evolution operators. Another noteworthy aspect
of this equation is that, despite the fact that both φin and φout are free fields, the relationship
between the two involves the interactions.
The easiest way to invert the relationship between φin and φout is to write the analogue of
(1.38) for the free field φout :

φ(x) = Uout (+∞, x0 ) φout (x) Uout (x0 , +∞),

  t2  
Uout (t2 , t1 ) ≡ T exp i d4 x LI φout (x) . (1.39)
t1

Taking the limit x0 → −∞ in this equation leads to a second form of the formula that relates
4 CONSTRAINTS ON THE S-MATRIX FROM CAUSALITY 17

φin and φout :

φin (x) = Uout (+∞, −∞) φout (x) Uout (−∞, +∞),

or, equivalently,

φout (x) = Uout (−∞, +∞) φin (x) Uout (+∞, −∞).

In order for the two relations we have obtained to be consistent, we must have

Uin (+∞, −∞) = Uout (+∞, −∞).

Therefore, the evolution operators over the entire time range are identical, regardless of whether
they are constructed with the fields φin or φout . For this combination of time arguments, we
may drop the subscripts in/out on the evolution operators.

3.b But it is important to realize that this property is not true for arbitrary time intervals. By
requesting that (1.38) and (1.39) give the same interacting field φ, we must have

Uin (−∞, x0 ) φin (x) Uin (x0 , −∞)

= Uout (+∞, x0 ) φout (x) Uout (x0 , +∞)
= Uout (+∞, x0 )U(−∞, +∞) φin (x) U(+∞, −∞)Uout (x0 , +∞),

implying that in general we have

Uin (x0 , −∞) = U(+∞, −∞)Uout (x0 , +∞). (1.40)

Writing Uin (x0 , −∞) = Uin (x0 , y0 )Uin (y0 , −∞) and using the same identity with x0 replaced
by y0 , we obtain

Uout (x0 , y0 ) = U(−∞, +∞)Uin (x0 , y0 )U(+∞, −∞).

(This relation could also have been obtained from the definition of Uout , by performing its
Taylor expansion in powers of φout , replacing every occurrence of φout by its expression in
terms of φin , and at the end repackaging the series to obtain a Uin .)

3.c By multiplying (1.40) on the right by the inverse of Uout (x0 , +∞), we obtain another
formula for the full evolution operator,

U(+∞, −∞) = Uin (x0 , −∞)Uout (+∞, x0 ),

which is rather counterintuitive since the order of the operators on the right-hand side may
(wrongly) suggest that it is inconsistent with the time ordering. The resolution of this paradox is
that U(+∞, −∞) is time-ordered when expressed entirely in terms of φin or entirely in terms
of φout ; but the right-hand side of the above formula mixes φin and φout , and the relationship
between φin and φout is non-local in time, which obscures the actual time ordering of the
operators.
18 1 QUANTUM FIELD THEORY BASICS

4. Constraints on the S-matrix from Causality The goal of this problem is to derive
general constraints on the S-matrix from causality. To that end, let us assume that the coupling
constant λ that controls the interactions is a function of spacetime, λ(x). With this modification,
the S-matrix becomes a functional S[λ].
4.a Consider two regions of spacetime, Ω1 and Ω2 , such that Ω2 lies in the future light-cone
of Ω1 , and denote by λ1,2 (x) the coupling function restricted to these domains (we
assume it is zero outside of Ω1 ∪ Ω2 ). Show that S[λ1 + λ2 ] = S[λ2 ]S[λ1 ].
4.b Generalize this result to the more general situation where Ω1 and Ω2 are simply sepa-
rated by a locally space-like surface.
4.c Consider now the casewhere the separation
 between any pair of points of Ω1 and Ω2 is
space-like. Show that S[λ1 ], S[λ2 ] = 0.
4.d If two coupling functions λ and λ  coincide for x0 ≤ y0 , show that S[λ  ]S† [λ] does not
depend on the behavior of the coupling at times ≤ y0 . By considering an infinitesimal
variation of the coupling function, show that
 
δ δS[λ] †
S [λ] = 0 if y is not in the future light-cone of x.
δλ(y) δλ(x)

4.e Solve the constraints of causality and unitarity to obtain the form of S[λ] up to O(λ2 ).

4.a Let us consider two regions Ω1 and Ω2 of spacetime, as shown in Figure 1.1. Thus, for
any pair of points x1 ∈ Ω1 , x2 ∈ Ω2 , we have x02 > x01 and (x1 − x2 )2 > 0. We assume that
interactions exist only in Ω1 ∪ Ω2 and are zero elsewhere, and we denote by λ1,2 the coupling
functions in these two domains. The coupling function over the entire spacetime is thus
λ(x) = λ1 (x) + λ2 (x),
and the full S-matrix is S[λ1 + λ2 ]. Recall that the S-matrix is the operator that connects the in
states at x0 = −∞ and the out states at x0 = +∞:
αout = αin Sin [λ1 + λ2 ].
For the time being, it is safer to add a subscript in on the S-matrix in order to indicate that it is
expressed in terms of the field operator φin . Given the relative configuration of the domains
Ω1 and Ω2 , we could also construct another version of the interaction representation, where
the fields coincide with the Heisenberg representation ones at some intermediate time located
between Ω1 and Ω2 . Let us call intermediate this representation, and φinter the corresponding
free field operator. We have
αinter = αin Sin [λ1 ], αout = αinter Sinter [λ2 ].
Thus, we obtain
Sin [λ1 + λ2 ] = Sin [λ1 ] Sinter [λ2 ].
Note that this equation is subject to the same paradox regarding the ordering of the operators
as in the last equation in Problem 3: the seemingly unnatural ordering between Sin [λ1 ] and
4 CONSTRAINTS ON THE S-MATRIX FROM CAUSALITY 19

out

Ω2

intermediate time

Ω1
in

Figure 1.1 Domains Ω1 and Ω2 , with Ω2 under the causal influence of Ω1 .

out

int
e
sp rme
ac
e-l diate
ike
su
rfa
Ω2
ce

Ω1
in

Figure 1.2 Domains Ω1,2 , such that Ω2 does not causally influence Ω1 .

Sinter [λ2 ] is due to the fact that the latter is implicitly expressed in terms of the field φinter while
the former depends on the field φin . Using exactly the same manipulations as in Problem 3,
we can rewrite this expression solely in terms of φin , which gives

Sinter [λ2 ] = Sout [λ2 ],

Sin [λ1 + λ2 ] = Sin [λ1 ]Sout [λ2 ],
Sin [λ1 + λ2 ] = Sin [λ2 ] Sin [λ1 ].

From now on, we implicitly assume that all S-matrix operators are expressed in terms of φin ,
and we suppress the subscript in.

4.b In words, the previous setup could be described by saying that Ω2 is under the influence
of Ω1 . A much more general situation would be to simply request that Ω2 does not influence
Ω1 . This is achieved by dividing spacetime with a hyper-surface Σ located between the
domains Ω1 and Ω2 , provided that this surface is locally space-like (i.e., no signal can travel
from the surface towards the domain below it). This setup is shown in Figure 1.2. This ordering
of the domains Ω1 and Ω2 is sufficient to reproduce the preceding arguments, leading again to
S[λ1 + λ2 ] = S[λ2 ] S[λ1 ].

4.c Consider now the situation where the interval between any point in Ω1 and any point
in Ω2 is space-like, as illustrated in Figure 1.3 (left panel). Since it is possible to find an
20 1 QUANTUM FIELD THEORY BASICS

out out out

in e-l
sp

te i k
ac

rm e s
ed ur
ia fac
te e

Ω1 Ω2
Ω1 Ω2
Ω2 Ω1

in in in

Figure 1.3 Left: domains Ω1 and Ω2 with mutual space-like separations. Center and right:
displacements of the domains under a Lorentz boost.

appropriate space-like surface separating the two domains, the previous result applies. But
now there is an ambiguity regarding which domain should be considered as being “before” the
other one. In fact, what is special about this kinematical configuration is that the time ordering
between the two domains can be altered by applying a Lorentz boost, as illustrated by the
second (β < 0) and third panels (β > 0) of Figure 1.3. In this case, because of this lack of
absolute time ordering between the two domains, the following two equations are both true:

S[λ1 + λ2 ] = S[λ1 ] S[λ2 ] and S[λ1 + λ2 ] = S[λ2 ] S[λ1 ],

 
which implies that S[λ1 ], S[λ2 ] = 0.

4.d Consider now two distinct coupling functions in the domain Ω2 , namely λ2 and λ2 . Thus,
we have two realizations of the coupling function:

λ ≡ λ1 + λ2 , λ  ≡ λ1 + λ2 .

For these two coupling functions, the S-matrix is given by

S[λ] = S[λ2 ] S[λ1 ], S[λ  ] = S[λ2 ] S[λ1 ],

and we therefore have

S[λ  ] S† [λ] = S[λ2 ] S[λ1 ] S† [λ1 ] S† [λ2 ] = S[λ2 ] S† [λ2 ].

This combination is independent of the function λ1 , i.e., independent of the behavior of the
coupling function in the portion of spacetime that cannot receive any causal influence from Ω2 .
Let us now assume that the difference δλ ≡ λ2 − λ2 is infinitesimal. To first order in δλ, we
4 CONSTRAINTS ON THE S-MATRIX FROM CAUSALITY 21

have

 δS[λ]
S[λ ] = S[λ] + d4 x δλ(x) + O(δλ2 ),
δλ(x)

and

 † δS[λ] †
S[λ ] S [λ] = 1 + d4 x δλ(x) S [λ] + O(δλ2 ).
δλ(x)

Since this should be independent of λ1 for any variation δλ, we must have

δS[λ] †
S [λ] is independent of λ1 if Ω1 is not under the causal influence of x.
δλ(x)

This condition can also be phrased as

 
δ δS[λ] †
S [λ] = 0 if y is not in the future light-cone of x. (1.41)
δλ(y) δλ(x)

Note that the same identity would be true in a theory where the fields are coupled to some
external source Jext , if we replace λ by Jext , again thanks to causality.

4.e In order to see the consequences of this constraint on the S-matrix, let us write a formal
Taylor expansion of the functional S[λ]:
 
1
S[λ] = 1 + d4 x S1 (x) λ(x) + d4 xd4 y S2 (x, y) λ(x)λ(y) + · · · .
2

In this expansion, the objects S(x1 , . . . , xn ) are operator-valued symmetric functions of their
arguments.
  addition to the constraint (1.41), the S-matrix must be unitary, and also satisfy
In
S[λ1 ], S[λ2 ] = 0 when the supports of λ1 and λ2 have purely space-like separations. The
last constraint implies that the coefficients Sn are multi-local operators (i.e., Sn (x1 , . . . , xn )
depends only on the field operator and its derivatives at the points x1 , . . . , xn ) constructed with
the field operator and its derivatives (non-locality would lead to violations of this commutation
relation). In the first two orders, the unitarity of S implies that

S1 (x) + S†1 (x) = 0, S2 (x, y) + S†2 (x, y) + S†1 (x)S1 (y) + S†1 (y)S1 (x) = 0.

(In deriving the second equation, we must be careful to symmetrize the coefficient that
multiplies λ(x)λ(y) in the integrand of the second-order term.) These equations can be
rewritten as

S†1 (x) = −S1 (x), S2 (x, y) + S†2 (x, y) = S1 (x)S1 (y) + S1 (y)S1 (x).

Note that unitarity can only constrain the Hermitian part of the coefficients Sn , and does not
say anything about their anti-Hermitian part. To put constrains on the latter, we need to make
22 1 QUANTUM FIELD THEORY BASICS

use of (1.41). It is straightforward to check that

 
δ δS[λ] †
S [λ] = S2 (x, y) + S1 (x)S†1 (y) + O(λ).
δλ(y) δλ(x)

Therefore, at lowest order, (1.41) tells us that

S2 (x, y) = −S1 (x)S†1 (y) = S1 (x)S1 (y) if y is not in the future light-cone of x.

Using the fact that S2 (x, y) should be symmetric, we also have

S2 (x, y) = S2 (y, x) = S1 (y)S1 (x) if x is not in the future light-cone of y.

(Note that these two conditions on x, y are both satisfied if their separation is space-like. For
such a space-like separation, we could thus use either of the two formulas. This does not
lead to any contradiction, provided that S1 (x) is a local function of the field operator and its
derivatives.) Therefore, the answer valid for any x, y can be written as

S2 (x, y) = T (S1 (x), S1 (y)).

(One may check a posteriori that the unitarity constraint is satisfied.) Using the same constraints,
one could show by induction that the coefficient of order n, Sn (x1 , . . . , xn ), is the time-ordered
product of n factors S1 . Therefore, we see that unitarity and causality provide an almost closed
form for the S-matrix,
 
S[λ] = T exp d4 x S1 (x)λ(x) ,

in which the only remaining unknown is the first coefficient S1 (x). The latter can be related to
the interaction Lagrangian by considering a λ(x) which is non-zero in an infinitesimal region
of spacetime. Therefore, regardless of the microscopic details of a given theory – which control
the first coefficient S1 –, the general structure of the S-matrix is governed to a large extent by
the constraints provided by unitarity and causality.

5. Landau Equations for Soft and Collinear Singularities The goal of this problem is to
study the singularities that may occur in a Feynman integral due to vanishing denominators
(not to be confused with ultraviolet divergences, due to an integrand that decreases too slowly
at large momentum). Consider a Feynman integral with L loops and m denominators:
 L
dD j N({pi }, {j })
I({pi }) ≡ .
(2π)D (q21 − m21 + i0+ ) · · · (q2m − m2m + i0+ )
j=1

In this integral, the pi are the external momenta, the j the loop momenta, and the qr the
momenta of the propagators in the loops, i.e., linear combinations of loop momenta and
external momenta of the form qr ≡ Lj=1 rj j + Δr (where the coefficients rj take values in
{−1, 0, +1} and the Δr depend only on the external momenta).
5 LANDAU EQUATIONS FOR SOFT AND COLLINEAR SINGULARITIES 23

5.a Use Feynman parameters xr to combine the m denominators into a single one, Dm .

5.b By considering the following elementary examples,

 +1  +1  +1
dx dx dx
, , ,
−1 x + i0+ 0 x + i0+ −1 (x + i0+ )(x − i0+ )

explain why singularities occur only when a pole is located on the boundary of the
integration manifold, or when multiple poles “pinch” the integration manifold.
5.c Study the variations of D as a function of the loop momenta j , and explain why the
zeroes of D must also be extrema of D in order to produce singularities.
5.d Now study the dependence with respect to the Feynman parameters and show that, for
each r, one must have either xr = 0 or q2r = m2r in order to have a singularity.
5.e Conclude that the conditions for a singularity are given by the Landau equations:
m
For each propagator r: xr (q2r − m2r ) = 0; For each loop j: μ
rj xr qr = 0.
r=1

5.f Determine the singularities of the one-loop q2

k'

Feynman diagram shown on the right (assume k+k'

q1=

k2 = k 2 = 0, (k + k  )2 > 0 and all internal q3
k
particles are massless)

5.a Consider a Feynman integral with L loops and m denominators. Generically, such an
integral may be written as
 L
dD j N({pi }, {j })
I({pi }) ≡ ,
(2π) (q1 − m1 + i0+ ) · · · (q2m − m2m + i0+ )
D 2 2
j=1

where the pi are the momenta external to the loops and the j are L independent loop momenta.
The momenta qr are the momenta carried by the various propagators along the loops. They are
all of the form
L
qr ≡ rj j + Δr ,
j=1

where the coefficients rj take values in {−1, 0, +1} (a propagator may belong to a loop or not,
and may be oriented in the same way as the loop momentum or in the opposite direction) and
where the Δr depend only on the external momenta (they are thus constants from the point
of view of evaluating the loop integrals). N({pi }, {j }) is a numerator that comprises all the
momentum dependence that may arise, e.g., from three-gluon vertices in QCD or from the
Dirac traces if there are fermion loops. This factor plays no role in analyzing the singularities of
24 1 QUANTUM FIELD THEORY BASICS

the integral, except in those rare situations where a singularity due to a vanishing denominator
may be canceled by an accidental concomitant vanishing of the numerator.
The first step is to combine the m denominators into a single one thanks to Feynman
parameterization:
 L  m
dD j   N({pi }, {j })
I({pi }) = Γ (m) dxr δ 1− xr  m , (1.42)
(2π)D r D({qr }, {xr })+i0+
j=1 xr ≥0 r=1
m
D({qr }, {xr }) ≡ xr (q2r − m2r ).
r=1

5.b In this problem, we assume that the ultraviolet divergences have been properly disposed of
by some regularization, and we are chiefly interested in the possibility of additional singularities
that may arise from a vanishing denominator. Obviously, that the equation
D({qr }, {xr }) = 0
has solutions in the integration domain is a necessary condition for having such singularities.
But a zero of the denominator of the integrand does not always lead to a singularity in the
integral. In order to see this, let us consider the following toy examples:
 +1  +1  
dx 1
+
= dx P − iπ δ(x) = −iπ,
−1 x + i0 −1 x

 +1
dx
= ∞,
0 x + i0+
 +1  +1
dx dx
= lim = ∞.
−1 (x + i0+ )(x − i0+ ) ε→0+ −1 x2 + ε2
In the first example, the integral is finite despite the fact that denominator vanishes at x = 0,
because the integration contour is slightly shifted from the singularity by the presence of the
i0+ . An infinite result would be obtained when it is impossible to shift the contour to avoid
the pole. In the second example, we cannot avoid the singularity because the pole occurs on
the boundary (here, at an endpoint) of the integration domain. In the third example, there are
two poles in the interior of the integration range, but these poles pinch the integration contour,
which prevents moving the contour to avoid the poles and also leads to an infinite result.

5.c Although the conditions of occurrence of a genuine singularity are the same, the situ-
ation we have to analyze is arguably more complex than these toy examples because of the
multivariate nature of the denominator. For a multi-dimension integral, the above condition is
that the poles of the integrand cannot be avoided by deforming the integration manifold. We
can make the following observations:

• Note first that the fact that D is raised to the m-th power in (1.42) is irrelevant for this
discussion: every zero of D leads to a pole of order m of the integrand, but this is
equivalent to having m poles all on the same side of the integration domain, so this
cannot produce a pinch.
5 LANDAU EQUATIONS FOR SOFT AND COLLINEAR SINGULARITIES 25

R
• For the L loop momenta j , the integration domain is DL . We may add to this domain
a point at infinity (assuming an ultraviolet regularization, the integrand goes to zero in
all directions at infinity), which leads to an integration domain topologically equivalent
to a DL-dimensional sphere SDL . This domain is boundaryless and therefore the only
possibility of singularities when integrating over the loop momenta is to have a pinch.
• For the Feynman parameters xr , the integration domain is a (m − 1)-dimensional
simplex,
 
(x1 , . . . , xm ) xr ≥ 0, xr = 1 ,
r

i.e., a line segment for m = 2, a triangle for m = 3, a tetrahedron for m = 4, etc.

Clearly, this domain has a boundary and therefore endpoint singularities may occur.

Let us now study the behavior of the denominator D. This function is quadratic in the loop
momenta, and linear in the Feynman parameters. Note also that, after a Wick rotation of the
loop momenta, D is a negative definite quadratic form in the Euclidean loop momenta, at fixed
{xr } (this is obvious from the fact that the second-degree part of D is a sum of squares weighted
by the negative coefficients −xr ). Therefore, when varying the loop momenta at fixed {xr }, the
denominator D has a maximum. We can distinguish the following cases:

• If max{j } (D) < 0, the denominator is always non-zero and there is no singularity.

• If max{j } (D) > 0, the denominator can vanish, but the zeroes are simple zeroes that
cannot pinch the integration manifold for the variables {j }. The integral is also finite in
this case.
• The dangerous situation is when max{j } (D) = 0, because the location of the maximum
is then a double zero (as in the third of the toy examples considered earlier) in all the μj
variables.

Therefore, we are seeking zeroes of D that are also extrema of its dependence with respect to
all the loop momenta:
∂D
D = 0, = 0.
∂μj

Note that these conditions are still not sufficient for a genuine singularity since we have only
discussed what happens at fixed {xr }, and we have not yet analyzed whether it may be avoided
by a deformation of the integration domain for the xr ’s. This discussion can be divided into
two cases:

• First, note that if q2r = m2r , a small variation of xr will change the value of D and move
the denominator away from zero. For xr in the interior of its allowed range, this means
that there is no actual singularity. The only exception is at xr = 0, since this is on the
boundary of the integration range.
• In contrast, when q2r = m2r , the denominator D is independent of xr , and a zero of D
persists at all values of xr .
26 1 QUANTUM FIELD THEORY BASICS

k'
q2
k+k'
q1=
q3
k

Figure 1.4 One-loop example of application of the Landau equations.

5.d We now have all the information to give the conditions of occurrence of a singularity in
the Feynman integral I({pi }):
∂D
D = 0, = 0; for each r, either xr = 0 or q2r = m2r .
∂μj

(Note that the last condition can be phrased as xr (q2r − m2r ) = 0 for each r, which makes
the first one, D = 0, redundant.) These conditions are known as the Landau equations. By
explicitly evaluating the derivative of the denominator with respect to the loop momenta, these
conditions can be written as
m
for each propagator r: xr (q2r − m2r ) = 0; for each loop j: μ
rj xr qr = 0.
r=1

5.e Consider now the example of Figure 1.4. The corresponding Landau equations read

x1 2 = x2 ( + k  )2 = x3 ( − k)2 = 0, x1 μ + x2 (μ + k μ ) + x3 (μ − kμ ) = 0.
A first solution is obtained for
2 = ( − k)2 = 0, x2 = 0, x1 μ + x3 (μ − kμ ) = 0,

which (since we have assumed that k2 = 0) is equivalent to

x3
2 = k2 =  · k = 0, μ = kμ .
x 1 + x3
This type of singularity is called a collinear singularity, since it occurs when the loop momen-
tum is aligned with one of the external momenta. A similar type of singularity is obtained at
x3 = 0, i.e.,
2 x2 
2 = k =  · k  = 0, μ = − k μ.
x1 + x 2

Another way to fulfill the Landau equations is to have μ kμ , k μ . In this limit, the first

three conditions are automatically satisfied, and the last one becomes x2 k μ = x3 kμ . Since kμ

and k μ are a priori not collinear, this implies that x2 = x3 = 0. This last type of singularity
is called a soft or infrared singularity, since it occurs when all the components of the loop
momentum go to zero.
5 LANDAU EQUATIONS FOR SOFT AND COLLINEAR SINGULARITIES 27

Coleman–Norton Interpretation: A more intuitive physical interpretation of the Landau

equations was found by Coleman and Norton (see Coleman, S. and Norton, R. E. (1965),
Nuovo Cimento 38: 438). They propose to interpret a Feynman graph literally, as if the
graph represents a process in spacetime, with the vertices being the locations where some
instantaneous interactions happen. They further propose that the spacetime separation dxμr
between two vertices is parallel to the momentum qμr , and proportional to the Feynman
parameter xr :

dxμr = xr qμr .

(The constant of proportionality could be different from 1, but should be the same for all
propagators. Its value is not important, as one may freely rescale the entire diagram without
affecting the Landau equations.) Based on this identification, one may make the following
observations:

• If a singularity happens at xr = 0, the spacetime separation is zero, and we may shrink

the corresponding propagator to a point. The resulting graph is called a reduced graph.
A similar reduction is possible if qμr = 0, in the case of a soft divergence.
• If xr = 0, the Landau equations tell us that we should have q2r = m2r instead. Thus, in
this interpretation, the propagators of a reduced graph represent the on-shell propagation
of a particle between two interactions. In this case, we may also write

dx0r qμr dxμr qμr

xr = , dxμr = dx0r , i.e., = = vμr ,
q0r q0r dx0r q0r

where vμr is the 4-velocity of an on-shell particle of momentum qμr .

• The last of the Landau equations becomes
μ
rj dxr = 0, for every loop j.
r

This equation is consistent with dxμr being a separation in spacetime, since adding these
separations along a closed loop should obviously give zero. This also implies that the
spacetime separation between two vertices does not depend on which path we follow on
the graph to connect them, as it should for this interpretation to make sense.

As an illustration, let us show the reduced diagrams for the three singularities we have found
for the triangle one-loop graph studied above:
k'
k'
+k' k'

 
−k k k
k

In more complicated cases, where solving the Landau equations may be difficult, the Coleman–
Norton interpretation and the associated reduced diagrams can be used as guidance for identi-
fying the possible solutions. This is based on the fact that there is a one-to-one correspondence
28 1 QUANTUM FIELD THEORY BASICS

between the solutions of the Landau equations and the spacetime diagrams proposed by Cole-
man and Norton. For instance, in the case of the above example, we could use it as a way of
checking that there are no other solutions. Indeed, there could in principle be two additional
reduced diagrams:

k'
k'
k'

−k k
k

However, neither of them is kinematically allowed; the left graph would have two non-collinear
particles propagating along the same spacetime interval, and the right graph would have a
physical free particle making a closed loop.

6. Multi-loop Integration in D Dimensions Start from equation (1.42) in Problem 5, in

the special case where the numerator is trivial, i.e., N ≡ 1. Note that the denominator D is a
quadratic form in the components of the loop momenta, which we may arrange as follows:

D ≡ Ltμ ALμ + Ltμ Bμ + Btμ Lμ + C, Lμ ≡ (μ1 , μ2 , . . . , μL ).

6.a Perform explicitly the integration over the L loop momenta in D dimensions in order to
obtain

DL  m
  (L+1)D
iL Γ (m − 2 ) δ 1− r xr det A)m− 2
I({pi }) = dxr ,
DL  m− DL
(−4π) 2 xr ≥0 r=1 C det A − Btμ CtA B μ
+i0+ 2
 
≡Δ

where CA is the matrix of co-factors of A.

⎛ ⎞
A11 . . . A1L Bμ1
⎜ ⎟
⎜A21 . . . A2L Bμ2 ⎟
⎜ . .. ⎟
6.b Show that Δ = det ⎜ . ⎟
..
⎜ .. . ⎟ and is of degree two in each variable xr .
⎜ μ⎟
⎝AL1 . . . ALL BL ⎠
B1μ . . . BLμ C
6 MULTI-LOOP INTEGRATION IN D DIMENSIONS 29

6.a Using the notation of Problem 5, qr ≡ j rj j +Δr , the denominator D can be rewritten
as

D= xr (q2r − m2r )
r
   
= xr rj rk μj kμ + 2 xr μ
rj Δr jμ + xr (Δ2r − m2r )
j,k
r  j
r  r 
≡ Ajk ≡ Bμ
j
≡C

= Ltμ A Lμ + Ltμ Bμ + Btμ Lμ + C.

In this expression, note that:

• A is a real symmetric L × L matrix, whose coefficients are linear in the Feynman

parameters xr and are independent of the external momenta. This matrix is positive
since zj Ajk zk = r xr ( j rj zj )2 , and positive definite if all xr > 0 (depending on the
details of the Feynman graph under consideration, it can still be positive definite if some
– but not all – of the xr are zero).

• Bμ is a column vector with L components, linear in the Feynman parameters and linear
in the external momenta.

• C is linear in the Feynman parameters, and also contains the squared masses and Lorentz
invariant scalar products of external momenta.

Let us first rearrange the quadratic form D as follows:

D = Ltμ A Lμ + Ltμ Bμ + Btμ Lμ + C

= (Lμ + A−1 Bμ )t 
A (L A−1 Bμ ) + C − Btμ A−1 Bμ
 + 
μ

Ωt DΩ
Kμ
= (ΩKμ ) D (ΩK ) + C − Bμ A−1 Bμ
t μ t

= Rμ Rμ + C − Btμ A−1 Bμ , with Rμj ≡ Djj (ΩK)μj .

The second line is the standard manipulation that eliminates the linear terms from the quadratic
form (the Jacobian is 1, since this is just a translation). In the third line, we used the fact that
the symmetric matrix Ajk (of size L × L) is diagonalizable by an orthogonal transformation
Ω (the Jacobian is also 1). In the last line, we have rescaled the various loop momenta in
order to absorb the diagonal elements of D. In the final form, we may view Rμ Rμ as the norm
of a unique vector with DL components (D being the dimension of spacetime). The overall
30 1 QUANTUM FIELD THEORY BASICS

Jacobian of this sequence of transformations is given by

 −D/2
dDL L = det A × dDL R.

Therefore, the integration over the loop momenta reads

 
dDL L 1 1 dDL R 1
=  m
(2π)DL D m (det A)D/2 (2π)DL
Rμ R + C − Btμ A−1 Bμ
μ

iL dDL RE (−1)m
=  
(det A)D/2 (2π)DL R2E + Btμ A−1 Bμ − C m
 t −1 μ  DL −m
(−1)m iL Γ (m − DL2 )
Bμ A B − C 2
= .
DL Γ (m) (det A)D/2
(4π) 2

In the second line, we have applied a Wick rotation to the temporal components of the L loop
momenta, and the result of the last line is a standard integration in DL dimensions (see the
first of Eqs. (1.21)). Therefore, the expression for the L-loop integral is
 m
(−1)m iL Γ (m − 2 )
DL  
I({pi }) = DL
dx r δ 1 − x r
(4π) 2 xr ≥0 r=1  t −1 r μ  DL −m
Bμ A B − C − i0+ 2
× .
(det A)D/2
 −1
(We have reinstated the i0+ prescription of Feynman propagators.) Using A−1 = det A CtA ,
where CA is the matrix of co-factors of A, we may rewrite this as

DL  m  (L+1)D
iL Γ (m −
2 )
  det A)m− 2
I({pi }) = dxr δ 1− xr .
DL  m− DL
(−4π) 2 xr ≥0 r=1 r C det A−Btμ CtA Bμ +i0+ 2
 
≡Δ

Written in this form, the fraction in the integrand is a rational function of the Feynman
parameters, with Δ ≡ C det A − Btμ CtA Bμ a homogeneous polynomial of degree L + 1 and
det A a polynomial of degree L.

6.b This expression for the loop integral can shed extra light on the discussion of singularities
in Problem 5, which led to the Landau equations. First, note that the determinant of A is zero
only in accidental situations, since in general D is a negative definite quadratic form of the
Euclidean loop momenta. Thus, the singularities in the above expression come from zeroes
in Δ. This quantity is precisely the maximum of the denominator D, viewed as a function of
the Euclidean loop momenta μj . Therefore, we recover the fact that the singularities of I({pi })
may only occur when D has a vanishing maximum.
Recall now that C det A − Btμ CtA Bμ = C det A − Btμ (det A) A−1 Bμ . Consider now the
linear equation Ajk Xμk = Bμj (where μ is treated as a fixed parameter). Its solution may be
7 WEINBERG CONVERGENCE THEOREM 31

written as
 
(det A) (A−1 )jk Bμk = (det A) Xμj = det A[1,j−1] Bμ A[j+1,L] ,
where the notation A[1,j−1] Bμ A[j+1,L] stands for the L × L matrix obtained with columns 1 to
j − 1 of A, the column vector Bμ , and columns j + 1 to L of A. Therefore, we have
 
Δ = C det A − Btjμ det A[1,j−1] Bμ A[j+1,L]
 
= C det A − Btjμ (−1)L−j det A[1,j−1] A[j+1,L] Bμ
⎛ ⎞
A11 . . . A1L Bμ1
⎜ μ⎟
⎜A21 . . . A2L B2 ⎟
⎜ . .. ⎟
= det ⎜⎜ ..
..
. . ⎟
⎟. (1.43)
⎜ μ⎟
⎝AL1 . . . ALL BL ⎠
B1μ . . . BLμ C
In the second line, we just need to count the number of column permutations necessary to
bring the column Bμ to the rightmost position. The last equality can be checked by expanding
the determinant on the right-hand side according to the minors of the last line. Observe now
that the function of x1 , x2 , . . . , xm obtained after integrating over the loop momenta does
not depend on how we labeled the loop momenta (possibly up to a permutation of the {xr }).
Therefore, let us assume that, for the internal propagator r = 1, we have made the choice
qμ1 = μ1 , i.e., 1j = δ1j , Δμ1 = 0.
With this choice, we have
Ajk = x1 δ1j δ1k + xr rj rk , Bμj = xr μ
rj Δr ,
r≥2 r≥2

C= −m21 x1 + xr (Δ2r − m2r ).

r≥2

In particular, the only coefficients that depend (linearly) on x1 are A11 and C. By expanding
the determinant in (1.43), we see that Δ is a quadratic function of x1 (with the other xr fixed).
Since our choice of setting q1 = 1 was arbitrary, we conclude that Δ is of degree two in every
variable xr (and overall homogeneous of degree L + 1).

7. Weinberg Convergence Theorem The goal of this problem is to establish a criterion

for ultraviolet convergence for loop integrals in scalar field theories (this restriction simplifies
the problem a bit, since the integrands have a trivial numerator equal to one). This result,
known as Weinberg’s convergence theorem, states that a loop integral is ultraviolet convergent
if the superficial degree of divergence of the loop integral, and of any of its restrictions to
hyperplanes obtained by setting linear combinations of the loop momenta to constants, is
negative. To establish this, we consider general Euclidean integrals of the form

  d4 1 · · · d4 m
I C, q, m ≡ n  ,
j=1 (Cji i + qj ) + mj
2 2

where Cji is an n × m constant matrix.

32 1 QUANTUM FIELD THEORY BASICS

7.a Why can one consider instead the simpler integrals


d4 1 · · · d4 m  
J(C) ≡ n , D ≡ (μ1 , . . . , μm ) 1 ≤ (C)2 ,
j=1 (Cji i )
2
D

that do not depend on the masses or on the external momenta?

7.b Show that I(C, q, m) is absolutely convergent if 4m − 2n < 0 and if the restriction of
this integral to any hyperplane of co-dimension 4 is also absolutely convergent.
Hint: define kj ≡ Cji i and write the integral J(C) as a sum of terms Jj (C) in which the
squared norm k2j is the smallest among all the k2 ’s, then rescale all the k=j by |kj |. Show
that the integral Jj (C) factorizes into a one-dimensional integral whose convergence is
determined by power counting, and an integral over a subspace of co-dimension 4 of the
same type as the original integral.

7.c Show that an equivalent convergence criterion is that the superficial degree of divergence
of I(C, q, m), and of restrictions of I to any hyperplane defined by setting some linear
combinations of the momenta to constants, is negative.

7.a Since we are interested only in the ultraviolet convergence of this integral, the masses
mj and the shifts qj do not play any role at large momenta (even though they matter for the
convergence in the infrared and for the precise value of the integral). As far as ultraviolet
convergence is concerned, we may as well set these parameters to zero, but we need to cut
out a small region around j = 0 in order to avoid infrared problems. Thus, we may consider
instead

d4 1 · · · d4 m  
J(C) ≡ n , D ≡ (μ1 , . . . , μm ) 1 ≤ (C)2 , (1.44)
j=1 (Cji i )
2
D

which has exactly the same ultraviolet behavior as the original integral. (For technical reasons
that will become clear later, it turns out to be a bit simpler to remove a ball of radius unity in
the space of the variables kj ≡ Cji j rather than i .)

7.b Note that the first of the two conditions listed in the statement of the problem for the
ultraviolet convergence of this integral, namely 4m − 2n < 0, is nothing but the demand
that the superficial degree of divergence of this integral be negative. This is a necessary
condition because the integrand cannot decrease faster than ξ4m−2n when all the i ’s are
rescaled according to i → ξi . The reason why this condition alone is insufficient to ensure
the convergence is that, depending on the matrix Cji , there could be directions in 4m along R
which the decrease of the integrand is slower, for instance if one or more of the i ’s do not
appear in one of the kj ’s. As we shall see, the second condition ensures that this problematic
situation does not occur.
8 WEINBERG CONVERGENCE THEOREM 33

Then, we may write the integral J(C) as a sum of integrals Jj (C) in which the squared
norm k2j is the smallest among all the k2l ’s:
n 
d4 1 · · · d4 m
J(C) = Jj (C), Jj (C) ≡ n 2 ,
j=1 D j l=1 kl
 
Dj ≡ (μ1 , . . . , μm ) 1 ≤ k2j ≤ · · · .

The next step is to perform a linear change of the integration variables, t ≡ A, such that

det A = 1, t1 = kj .

In terms of the tμi , the integral Jj (C) may be rewritten as

 
d4 t2 · · · d4 tm  
Jj (C) = 4
d t1 n −1
, U ≡ (tμ2 , . . . , tμn ) t21 ≤ (CA−1 t)2 .
l=1 (Cli Aik tk )
2
t21 ≥1 U

Then, we rescale t2 , . . . , tn by the Euclidean norm of t1 ,

  
tμi ≡ |t1 | iμ |t1 | ≡ t21 ,

which leads to the following form of the integral Jj (C):

 
d4   · · · d4  
Jj (C) = d t1 |t1 |
4 4m−2n−4
×  2   m2 , (1.45)
t21 ≥1 U l=j (Clk k )
 

with Clk ≡ Cli A−1
ik , U  ≡ (2μ , . . . , nμ ) 1 ≤ (C    )2 .

In this form, it is clear that:

• The first integral on the right-hand side is absolutely convergent if 4m − 2n < 0.

• The second integral is of the form (1.44) with one less integration variable. Since the two
integrals on the right-hand side are independent, it must also be convergent for Jj (C) to
be finite.

This proves the announced convergence criterion.

7.c If we apply this criterion recursively, we get integrals of lower and lower dimensionality,
until the last step where the analogue of (1.45) contains only the first factor (this corresponds
to a one-loop integral). When we reach this point, we can conclude about the convergence
from power counting only. Thus, the convergence of the original integral is ensured if its
degree of divergence 4m − 2n, and the degree of divergence of all the sub-integrals obtained
by restricting the integration domain to hyperplanes of lower dimension, are negative. In this
form, the convergence criterion is known as Weinberg’s convergence theorem (note that here
we have studied a less general situation than in the original theorem, which also considers
the possibility of a polynomial of the integration variables in the numerator of the integrand).
Let us add a final remark: the criterion derived in this problem is a criterion for absolute
convergence. When it is not satisfied, it could still happen that the integral is nevertheless
34 1 QUANTUM FIELD THEORY BASICS

(weakly) convergent because of cancellations among various parts of the integration domain.
And of course, there could also be cancellations among the contributions of various Feynman
graphs. This is typically what happens in gauge theories, where the gauge symmetry can
induce cancellations among graphs that form a gauge invariant set (a single graph is usually
not gauge invariant).
Source: Hahn, Y. and Zimmermann, W. (1968), Commun Math Phys 10: 330.

8. Electron Anomalous Magnetic Moment Consider the amputated renormalized

photon–electron–positron vertex function Γrμ (k, p, q) (with the convention that all momenta
are incoming).

8.a Using the Dirac equation, prove the following relationship:

i μν  
(p − q ) u(−q)u(p) = 2m u(−q) γ + M (p + q)ν u(p),
μ μ μ
m

known as Gordon’s identity (with Mμν ≡ 4i [γμ , γν ]).

8.b Show that its contribution to electron scattering off an external field Aμ can be parame-
terized as follows:
 Mμν kν 
Aμ (k)u(−q)Γrμ (k, p, q)u(p) = er Aμ (k)u(−q) F1 (k2 )γμ +i F2 (k2 ) u(p).
mr

Hint: the most general form of Γrμ is

Γrμ = Cμ1 1 + Cμα μ μα μαβ

2 γα + C3 γ5 + C4 γα γ5 + C5 Mαβ .

Then, use Lorentz invariance and the Ward–Takahashi identity to show that Cμ1 ∝
(pμ − qμ ), and then the Gordon identity to bring this term to the announced form. The
same reasoning can be used to bring the remaining four terms in Γrμ to the announced
form.

8.c Approximate this formula for a constant magnetic field. In particular, for an electron at
rest in a homogeneous magnetic field B in the x3 direction, show that

er B  
Aμ (k)u(−q)Γrμ (k, p, q)u(p) becomes 1 + F2 (0) u(0)M12 u(0).
mr

μ
8.d Calculate the relevant parts of Γr at one loop in order to show that F2 (0) = 2π
α
. (This
result led to one of the first experimental verifications of quantum electrodynamics.)
8 ELECTRON ANOMALOUS MAGNETIC MOMENT 35

8.a Let us rewrite the Poincaré algebra generator Mμν as follows:

i  i  i 
Mμν = γμ γν − γν γμ = γμ γν − gμν = gμν − γν γμ .
4 2 2
Then, consider
i  
u(−q)Mμν (pν + qν )u(p) = u(−q) γμ γν − gμν pν u(p)
2
i  
+ u(−q) gμν − γν γμ qν u(p)
2
i(pμ − qμ )
= im u(−q)γμ u(p) − u(−q)u(p).
2
This relation, known as Gordon’s identity, can be rewritten as
 i 
(pμ − qμ ) u(−q)u(p) = 2m u(−q) γμ + Mμν (p + q)ν u(p). (1.46)
m

8.b The object u(−q)Γrμ u(p) is a radiative correction to an electromagnetic current. Since
the incoming and outgoing electrons are on-shell, it must satisfy the following Ward–Takahashi
identity:
kμ u(−q)Γrμ (k, p, q)u(p) = 0.
Moreover, this identity must be satisfied even if the photon is off-shell, k2 = 0. Obviously, the
right-hand side of the announced formula obeys this identity, after one uses the Dirac equation
and the antisymmetry of Mμν .
Let us now give a glimpse of how one would prove the converse, namely that this form is
the only possible one. The starting point is to note thatΓrμ is a 4 × 4 matrix carrying
 a pair of
Dirac indices, which we may decompose on the basis 1, γα , γ5 , γα γ5 , Mαβ :
Γrμ = Cμ1 1 + Cμα μ μα μαβ
2 γα + C3 γ5 + C4 γα γ5 + C5 Mαβ .
Since these terms are linearly independent, they must fulfill the Ward–Takahashi identity
independently. From charge conjugation and parity symmetry, the term in γ5 should be zero,
Cμ3 ≡ 0. The Lorentz indices of the remaining coefficients must be carried by the vectors p, q
(or k = −p − q), the metric tensor, and possibly the Levi–Civita symbol in the case of γα γ5 ,
with prefactors that depend only on Lorentz invariant quantities. In fact, all Lorentz invariant
quantities can be expressed in terms of the electron mass and the photon virtuality k2 , since
2
p · q = k2 − m2 , p · k = q · k = − k2 .
Consider for instance the coefficient Cμ1 of the identity. It may be written as
Cμ1 = C1a (k2 ) pμ + C1b (k2 ) kμ .
The Ward–Takahashi identity implies
 
0 = C1a k · p + C1b k2 = k2 C1b − 12 C1a .
Therefore, this coefficient must have the following form:
 
Cμ1 = C1b (k2 ) 2pμ + kμ = C1b (k2 ) (pμ − qμ ).
Thanks to the Gordon identity (1.46), this term of Γrμ indeed has the general form quoted in the
statement of the problem. This turns out to be true for all the terms in this decomposition, as one
36 1 QUANTUM FIELD THEORY BASICS

may check by first writing the most general Lorentz structure allowed by the Ward–Takahashi
identity and then by using the Dirac equation to simplify the result after insertion between
u(−q) · · · u(p).

8.c The coefficients F1 (k2 ) and F2 (k2 ), called the form factors of the electron, describe the
properties of the cloud of photons and virtual pairs that surround the electron, as one varies
the virtuality k2 of the photon that probes the electron (in a sense, k2 plays the role of a the
resolution scale at which the photon probes this cloud). In the limit k2 → 0, the photon probes
this cloud on very large distance scales. Given its similarity with the bare vertex, the term
proportional to γμ encodes the electrical charge seen by the photon. On very large distance
scales, this must be the usual charge of the electron as we know it from atomic physics, which
means that F1 (0) = 1. Let us now discuss the meaning of F2 (0). Consider a constant external
magnetic field B in the x3 direction, corresponding to

A2 (x) = Bx1 , A2 (k) = iB∂k1 , F12 (x) = −F12 (x) = B,

where Fμν is the field strength. When an electron at rest is embedded in this field, the dressed
coupling of the photon to the current reads (in the limit k → 0)
 M12 
er B u(−p) iγ2 ∂p1 + F2 (0) u(p) .
mr p=0

The derivative ∂p1 acting on a spinor of small momentum acts like the boost K1 in the direction
x1 . More precisely, we have

iK1 γ1 γ0
u(−p)iγ2 ∂p1 u(p) = u(−p)iγ2 u(p) = −u(−p)iγ2 u(p)
p=0 mr p=0 2mr p=0
1
= u(−p)M12 u(p) .
mr p=0

(The Dirac equation for zero momentum spinors reduces to γ0 u(p) = u(p).) Therefore, the
above coupling becomes
er B  
1 + F2 (0) u(−p)M12 u(p) .
mr p=0

The 1 in 1 + F2 (0) encodes the bare coupling of the magnetic field to the electron, and F2 (0) is
therefore a correction to this coupling.

8.d At one loop, the QED Feynman rules lead to the following expression for the vertex
function contracted between the spinors of the incoming and outgoing fermions:
 D
d  u(−q)γσ (/−q / +m)γμ (/+p/ +m)γσ u(p)
u(−q)Γ (k, p, q)u(p) = −i e
μ 3
(2π) D  (( − q) − m )(( + p)2 − m2 )
2 2 2
 D
d  Nμ
≡ −i e3 .
(2π)D D
For the time being, we use dimensional regularization to make all intermediate expressions
finite, but we shall see shortly that the form factor F2 is ultraviolet finite. The denominators of
8 ELECTRON ANOMALOUS MAGNETIC MOMENT 37

the three propagators can be combined into a single one by introducing Feynman parameters:
1
1 δ(x1 + x2 + x3 − 1)
=2 dx1 dx2 dx3 .
ABC 0 (x1 A + x2 B + x3 C)3

In the situation of interest here, the resulting denominator reads

x1 (( + p)2 − m2 ) + x2 (( − q)2 − m2 ) + x3 2

= L2 + x1 (1 − x1 )p2 + x2 (1 − x2 )q2 − (x1 + x2 )m2 + x1 x2 p · q
= L2 −(x1 + x2 )2 m2 + x1 x2 k2 ,
 
≡ −Δ

where we have defined L ≡  + x1 p − x2 q and where in the final line we have assumed the
electrons to be on-shell (p2 = q2 = m2 , k = −(p + q)).
The next step is to express the numerator Nμ in terms of the new integration variable L:

Nμ = u(−q)γσL / γμ L
/γσ u(p)
+ terms linear in L
+ u(−q)γσ (m − x1p / − (1 − x2 )q
/ )γμ (m + (1 − x1 )p
/ + x2 q
/ )γσ u(p). (1.47)

The terms linear in L can be dropped since the denominator is even in L. The combination of
Dirac matrices in the first line can be rewritten as

(2 − D)2 2 μ
γσ L / γσ =
/ γμ L L γ .
D
Since it is quadratic in L, this terms leads to a logarithmic ultraviolet divergence, but the
proportionality to γμ indicates that it contributes only to the form factor F1 (the ultraviolet
divergence in F1 leads to a renormalization of the electron electrical charge). Since our goal
is to evaluate F2 , we can disregard this term from now on. The term on the last line of (1.47)
does not depend on L and therefore gives an ultraviolet finite integral over L, which implies
that we can perform the Dirac algebra for this term in D = 4. Using

γμ γν γμ = −2 γν , γμ γν γρ γμ = 4 gνρ , γμ γν γρ γσ γμ = −2 γσ γρ γν

and the Dirac equations obeyed by the two spinors, a straightforward but somewhat tedious
calculation leads to

u(−q)γσ (m − x1p
/ − (1 − x2 )q
/ )γμ (m + (1 − x1 )p / + x2 q
/ )γσ u(p)
 
 pμ −qμ μ

= 4m u(−q)u(p) 2 (x 3 − x 2
3 ) + k
2 (x 1 + x 2
1 − x 2 − x 2
2 )
   
+ 2 u(−q)γμ u(p) m2 (x23 + 2x3 − 1) − k2 (x3 + x1 x2 ) .

The term of the last line, in u(−q)γμ u(p), contributes only to F1 and can be ignored. In the
first line on the right-hand side, the term in kμ is odd under the exchange of x1 and x2 , while
the denominator is even. Its integral over the Feynman parameters is therefore zero. Thus, we
38 1 QUANTUM FIELD THEORY BASICS

need only consider further the term in pμ − qμ , which can be rearranged thanks to the Gordon
identity (1.46):

u(−q)γσ (m − x1p / − (1 − x2 )q/ )γμ (m + (1 − x1 )p

/ + x2q
/ )γσ u(p)
i 
= u(−q)Mμν u(p) kν (−4m2 )(x3 − x23 )
m
+ terms in u(−q)γμ u(p)
+ terms that integrate to zero.

By performing the integral over L (using a Wick rotation to have a Euclidean integration
momentum),
  
dD L 1 ΔD/2−3 Γ 3 − D2 i
= −i = − ,
(2π) (L − Δ)
D 2 3 (4π)D/2 Γ (3) D=4 32π2 Δ

and then by comparing the term in Mμν with the expression for the vertex function in terms of
the electron form factors, we obtain
1
e2 x3 − x23
F2 (k ) = 2
2
dx1 dx2 dx3 δ(x1 + x2 + x3 − 1) .
4π 0 (1 − x3 )2 − x1 x2 (k2 /m2 )

Noting that
1
dx1 dx2 δ(x1 + x2 + x3 − 1) = 1 − x3 ,
0

we finally get

e2 α
F2 (0) = 2
= .
8π 2π
This is the celebrated result obtained by Schwinger in 1948. As of 2019, the QED calculation
of this quantity has been pushed to order α5 , allowing a comparison with experimental
measurements of the electron anomalous magnetic moment with an unprecedented accuracy.

9. Ward–Takahashi Identities and Lorentz Invariance The goal of this problem is to

discuss the relationship between the Lorentz invariance of scattering cross-sections involving
photons and the Ward–Takahashi identities obeyed by amplitudes with external photons. We
will see that the Ward–Takahashi identities imply the Lorentz invariance, but also that the
converse is true to some extent: namely a theory of massless spin-1 particles must obey at least
some weak form of Ward–Takahashi identities in order to be Lorentz invariant. To that end,
we consider an S-matrix element with an external photon, written as μλ (p)Mμ (p, . . . ).

9.a What is the transformation law of Mμ (p, . . . ) under a Lorentz transformation Λ?

Discovering Diverse Content Through
Random Scribd Documents
The Project Gutenberg eBook of The Round
Towers of Ireland; or, The History of the
Tuath-De-Danaans
This ebook is for the use of anyone anywhere in the United States
and most other parts of the world at no cost and with almost no
restrictions whatsoever. You may copy it, give it away or re-use it
under the terms of the Project Gutenberg License included with this
ebook or online at www.gutenberg.org. If you are not located in the
United States, you will have to check the laws of the country where
you are located before using this eBook.

Title: The Round Towers of Ireland; or, The History of the Tuath-De-
Danaans

Author: Henry O'Brien

Release date: April 15, 2013 [eBook #42538]

Most recently updated: October 23, 2024

Language: English

Credits: Produced by The Online Distributed Proofreading Team at

http://www.pgdp.net (This file was produced from images
generously made available by The Internet Archive.)

*** START OF THE PROJECT GUTENBERG EBOOK THE ROUND

TOWERS OF IRELAND; OR, THE HISTORY OF THE TUATH-DE-
DANAANS ***
THE ROUND TOWERS OF
IRELAND

Author of “The Round Towers of Ireland.”

 THE ROUND TOWERS
OF IRELAND
OR

THE HISTORY OF THE TUATH-DE-

DANAANS

BY
HENRY O’BRIEN

A NEW EDITION
WITH INTRODUCTION, SYNOPSIS, INDEX, ETC.

London: W. THACKER & CO., 2 Creed Lane, E.C.

Calcutta: THACKER, SPINK & CO.
1898
[All Rights Reserved]

750 Copies only of this Edition have been

printed for Sale and the Type distributed,
of which this is No. 324.
 CONTENTS
PAGE

Introduction vii
Synopsis xxxvii
Dedication (First Edition) lxi
Preface (First Edition) lxiii
Dedication (Second Edition) xciii
List of Illustrations xcv
Author’s Text (Second Edition) 1
List of the Principal Irish Towers and Crosses 525
Index 529
 INTRODUCTION
“When all is dark, who would object to a
ray of light, merely because of the faulty
or flickering medium by which it is
transmitted? And if those round towers
have been hitherto a dark puzzle and a
mystery, must we scare away O’Brien,
because he approaches with a rude and
unpolished but serviceable lantern?”—
Fraser’s Magazine for August 1835.
Henry O’Brien, the most daring and ingenious explorer of that
recondite mystery, the origin and purpose of Irish Round Towers,
was born in 1808. On both his father’s and his mother’s side he
came of good descent,[1] being connected with two of the oldest and
most influential families in the west of Ireland. At the time of his
birth that branch of “the O’Briens” to which he belonged were
settled in Kerry, where his father resided in a wild, mountainous
district, known as Iveragh, forming a portion of the Marquis of
Lansdowne’s Irish estates. That his family were in affluent
circumstances is improbable, for up to the age of twelve the boy’s
education seems to have been neglected in a way very uncommon
with Irish people who are well off. “Though I could then tolerably
well express myself in English,” he says,[2] referring to this portion of
his life, “the train of my reflections always ran in Irish. From infancy
I spoke that tongue; it was to me vernacular. I thought in Irish, I
understood in Irish, and I composed in Irish”; and again, “I was
twelve years of age before ever I saw a Testament in any language.”
From this unusual neglect, coupled with the fact of his becoming a
private tutor soon after he had settled in London, and an obscure
reference to certain “difficulties” at the outset of his career as an
author, we are probably justified in assuming that money was a
rather scarce commodity in the paternal home. There is, however,
reason to suppose that when he had reached the age of twelve, or
thereabouts, his education was taken in hand, though how, or by
whom, does not appear. Evidence of his having been sent to school
and placed under systematic and qualified instruction is not
forthcoming. In fact, circumstances go to negative that supposition.
His acquaintance with Greek and Latin authors seems to have been
more extensive than accurate, and his quotations from them are
marked by solecisms which any properly taught schoolboy would
avoid, but in which the self-educated are prone to indulge. It is true
that (at p. 481) he describes in terms of unqualified praise a “tutor”
with whom he commenced the study of the Greek Testament; but
there is internal evidence in the same passage that such praise was
not wholly deserved, and that the tutor in question was certainly not
the person referred to in Father Prout’s statement that O’Brien had
been “brought up at the feet of the Rev. Charles Boyton.”[3] Mr.
Boyton was at the time a highly distinguished Fellow of Trinity
College, Dublin, who, in addition to holding the position of Greek
Lecturer at that University, was the most eminent mathematical
“coach” of his day; and the only connection likely to have existed
between him and young O’Brien was that of college-tutor and
undergraduate in statu pupillari. The probability is, therefore, that
any instruction which the boy received at this early period of his life
was of a very elementary character, and that his education was
mainly conducted by himself, a probability which is certainly not
discounted by the wide and promiscuous character of his reading.
From the outset of his introduction to letters he is known to have
been an omnivorous reader of all books that came in his way, nor
was his mode of studying classical authors that by which the
scholastic proficiency essential to aspirants for success at college
examinations is usually attained. O’Brien did not resemble the
ordinary boy-student, to whom Roman or Greek classics represent
merely a given quantity of “text” possessing certain peculiarities of
diction or allusion which have to be nicely dissected, analysed, and
mastered, but who regards the subject-matter of each work as being
of very minor importance. On the contrary, he manifestly read them
as authors, or rather authorities upon the subjects with which they
respectively dealt, paying, so far as we can perceive, little or no
attention to the diction or distinctive literary character of their
writings. The result was what might be expected. If, whilst an
undergraduate of Dublin University, it be true that he was regarded
by many of his fellow-students as a prodigy of learning, their seniors
appear to have been less enthusiastic about his scholarship, for we
have not been able to discover his name in the college archives.[4]
Still, from the fact of his having obtained, after he took his degree in
1831, the appointment of private tutor to the sons of the then
Master of the Rolls,[5] it is possible that he may have distinguished
himself previously.
What seems absolutely certain is, that during his stay at the
University he must have availed himself to the full of opportunities
presented by the library for which Trinity College is famous. Here, no
doubt, he laid the foundation of that Oriental learning in which he
was second to no Irishman of his day, and probably to few
Englishmen. It is hardly too much to say that in the early part of the
century Orientalism was comparatively untrodden ground. Sir
William Jones had indeed, many years before, thoroughly explored
this field of knowledge, but the results of his splendid labours had
not as yet been properly assimilated by the general mass of readers,
or supplemented to any remarkable extent by other workers in the
same field. Hence the scope of European knowledge of the East was
by no means so extensive then as now; and an enthusiastic student
thereof, which O’Brien undoubtedly was, had it in his power to
acquire an almost complete mastery of the subject, so far as it was
then known. It was one peculiarly fitted to his ardent, dreamy, and
speculative nature. He read, he pondered, he divined, he foresaw.
Dark places in the history of his own country began to grow clear in
the light of this Eastern dawn. Hitherto, like so many of his
compatriots, he had found no way of accounting for the
extraordinary contrast between the distinctive superiority of “the
Ireland that was” and the relative obscurity of “the Ireland that is.”
To what, he must apparently have asked himself, was the fact to be
attributed, that a people who in days of old were admittedly pre-
eminent in learning and civilisation, should have afterwards lost all
claim to such distinction; or how was it that, in a land covered with
the ruins of structures evincing the ripest skill and most fanciful
artistic device, architecture should have sunk to a level that was
almost barbarous? Why was it that this decadence did not take place
gradually, as one would expect, but was plainly the result of a
sudden check that stopped the erection of such edifices at once and
for ever? Why were the materials, structure, and conformation of the
edifices in question so different from those of other ancient buildings
found in their immediate neighbourhood? Why had their sculptured
ornamentation reference to what was unconnected with, nay even
opposed to, the teachings of that religious faith to which its
execution was attributed; and why did the peasantry, inheriting the
tradition of bygone ages, not recognise them as identified with that
religion? Questions like these are very stimulating to inquisitive
young souls, which usually become fired with an ambition to solve
them; and as O’Brien pored over Sir William Jones and The Asiatic
Researches—not to mention his beloved, though decried, Herodotus
—it was only natural that he should draw certain conclusions from
the undoubted affinity that exists between the languages, folk-lore,
customs, superstitions, and modes of thought of his own country
and those of the Orient. Similar conclusions had forced themselves
upon older people who did not possess a tithe of his Eastern lore.
Moore, that versatile Anacreontic, in his ill-fitting disguise of an
Edinburgh Reviewer, avowed “That there exist strong traces of an
Oriental origin in the language, character, and movements of the
Irish people, no fair inquirer into the subject will be inclined to
deny;” and it is further instanced by the same reviewer how the
famous traveller, Bishop Pococke, on visiting Ireland after his return
from the East, was much struck with “the amazing conformity” he
observed between the Irish and the Egyptians.[6] From early
childhood the questions to which we have referred seem to have
been present to O’Brien—even from the time when he gazed upon
the stunted ruin of Bally-Carbery Round Tower, not far from his
father’s house, and had been told by awestruck peasants that the
real name of that desolate and unsightly object was Cathoir Ghall, or
“The Temple of Delight” (p. 48). Since then he had seen other and
complete round towers; had noticed that all were of the same
peculiar shape, and possibly had detected for himself, or learned
from other sources, the existence of that phallic analogy upon which
he so strongly insists. He must have read in Sir William Jones and
elsewhere how, in Eastern lands, the idea which lay beneath this
same analogy formed the basis of a widespread religious faith, and
was expressed in structures devoted to public worship. His next step
was, almost inevitably, one of conjecture. If, as the voice of national
tradition asserted, the round towers are “temples,” and if certain
analogous associations are connected with them, might they not
have been temples of a kindred religious belief? Having settled this
to his own satisfaction, the speculation would naturally rise—How
came that particular form of belief to prevail in Ireland? Was it native
to the soil; or if not, by whom was it introduced, and when? His
book being mainly an answer to these questions, we need not
continue to follow the various stages by which conjecture may have
passed into theory, and theory into conviction. With men of O’Brien’s
temperament the hypothetical interval is rarely of long duration.
Before he had assumed the toga virilis of a full-fledged graduate, he
probably felt confident that in an Eastern origin lay the true solution
of the mystery of the round towers; and the more he studied the
subject, the stronger grew his belief. Being an ambitious man, too,
he had no intention to forego the honour which he was persuaded
must accrue to the discoverer of this key to a problem that had
baffled so many generations of inquirers, and longed for an
opportunity to display his acquisition.
That opportunity soon came. In December 1830, the Royal Irish
Academy offered the prize of a gold medal and fifty pounds to “the
author of an approved essay on the Round Towers, in which it is
expected that the characteristic architectural peculiarities belonging
to all those ancient buildings now existing shall be noticed, and the
uncertainty in which their origin and uses are involved be
satisfactorily removed.” Unfortunately, the advertisement of this offer
escaped O’Brien’s notice, and he did not join in the competition
which it evoked. But on the 21st February 1832 the advertisement
was repeated, and this time it caught his attention. It declared that
none of the essays which had been sent in “satisfied the conditions
of the question,” and extended the period of competition for another
three months (i.e. until 1st June 1832), in the alleged hope “of
receiving other essays on said subject,” and also for allowing the
authors of the essays already sent in “to enlarge and improve them.”
Considering the task that was set, new competitors were thus placed
at a singular disadvantage—being expected to do in three months
what the others had been unable to accomplish in two years. With
all due respect to the Royal Irish Academy, it is difficult to believe
that its members can have fully realised the nature of their own
conditions. There still exist some scores of round towers in a more
or less perfect state; and they are scattered all over Ireland, being
situated for the most part in remote and not easily accessible places.
The work of visiting and inspecting these—which was, surely, a
necessary preliminary to describing “the characteristic architectural
peculiarities belonging to all”—would require much time, after which
candidates must apply themselves to the by no means trifling task of
dispelling “the uncertainty in which their origin and use are
involved,” and all within three short months.[7] O’Brien was not,
however, to be deterred by considerations of time or space when
confronted with such a chance of winning deathless fame. Besides,
he was, in one respect at any rate, well equipped for the enterprise,
having already made up his mind as to the “origin and uses” of the
Round Towers. That he had examined them all is not to be
supposed, nor is it at all likely that at his age he could have
possessed sufficient technical knowledge of architecture, in its
historical and scientific aspects, to profit much by their inspection.
Still, he was probably acquainted with whatever had been written on
that branch of the subject, and had actually made an examination of
some towers, which would give him a fair general idea of the whole.
Moreover, he had a formidable quantity of Eastern learning to fall
back upon, in which latter respect he would have enjoyed an
immense advantage over all other possible competitors, if his judges
had only been qualified to appreciate that learning as it deserved. Be
his equipment for the enterprise what it might, the enthusiastic
young Irishman saw no rocks ahead, felt no mistrust, and rushed
into the fray. “I grappled with the question,” he assures us, “with all
the ardour of my nature; and, heaven and earth, night and day, in
difficulties and in sorrow, I laboured until I finished my ‘essay’
against the appointed hour, when—a brain blow to their (sc. the
Academy’s) expectation—I sent it in—fully satisfied, from the
consciousness of its imperturbable axioms, that all the powers of
error and wickedness combined could not withhold from it the
suffrage of the advertised medal.”[8] The meaning of this passionate
reference to malign influences in the background will appear later
on; as yet, he had no cause for misgiving on the subject of fair play,
and his overweening self-confidence precluded any anticipation of
failure. Bad omens seem to have attended his venture from the very
outset. The Academy had requested that each essay should be
inscribed with some motto; and it would appear that the motto
appended to O’Brien’s was “Φωνη εν τη ερεμω” (sic[9])—a sorry
introduction to the notice of learned Academicians.
The heartburnings of suspense, with which most young authors are
familiar, soon began. Four days after his essay had been sent in, the
Academy issued a third advertisement, requiring all the essays to be
taken back, and extending the period of preparation by an additional
month, “so as to admit of the receiving of other essays on said
subject, and for allowing the authors of essays already given in to
improve and enlarge them.” O’Brien afterwards saw fit to attribute
this fresh delay to a cause very different from that alleged; but just
then, being persuaded that his triumph was merely postponed, he
reconciled himself as best he could to the infliction, and calmly
waited for apotheosis. Six months more passed by—wearily enough,
we may be sure; and then, one direful morning, just at the close of
1832, came news that the premiums had been adjudged as follows:
—“£50 and the gold medal to George Petrie, and £20 to Henry
O’Brien, Esq.”
It may be stated here that an additional premium of £100, which
had been placed by Lord Cloncurry at the disposal of the Academy,
was also awarded in its entirety to Mr. Petrie, and that the essay sent
in by that gentleman was, by order of the Academy, printed in their
Transactions. It further appears that O’Brien’s essay was at first
accepted for publication in the Transactions, but afterwards rejected
on the ground of having been made too lengthy by the insertion of
additional matter, though in its most enlarged form it never attained
to the dimensions of Mr. Petrie’s work, and, presumably, must have
been smaller in its original than in its present shape. The true reason
for its exclusion from the Transactions (as will, we think, appear
from what follows) was that the Academy took offence at the way in
which O’Brien received their decision. Nor was such resentment to
be wondered at. So confidently had our author reckoned upon an
overwhelming triumph for the revelation which, as we have seen, he
believed to be not only unprecedented, but given to the world with
flawless perfection of statement, that the award seems to have
almost maddened him. Belonging to a race which has never been
remarkable for the silent endurance of wrongs, he lost no time in
giving expression to his feelings of disappointment. At first came
distant mutterings of the storm that was brewing. “On hearing of the
decision,” he informs us, “I wrote off to the secretary, tendering, in
indignant irony, my thanks for their adjudication, taking care,
however, to tell them that I had expected an issue more flattering to
my hopes.” This dignified attitude having apparently failed to imbue
the Academy with a desire to remedy his grievance, he flung off the
mask of satire, and rushed into downright, unmistakable
personalities of a kind rarely addressed to august and learned
associations. He declared that, from information which had come to
his knowledge, he was prepared to prove “that the Royal Irish
Academy, at the very moment in which they published their second
invitation (i.e. that by which the time for receiving essays was
extended to 1st June 1832), had actually determined to award the
gold medal and premium to one of their own Council.”[10] He then
went on to denounce the successful essay as “a farrago of
anachronisms and historical falsehoods.” He prophesied that when
both essays were published, and the public given an opportunity of
seeing “the truth,” in the shape of his own essay, there would be a
general acclamation of “This alone is right.” He warned the Academy
that, “though separated from them by a roaring sea” (he was living
in London at the time), his eye was on their plans, and he
demanded from them an opportunity for making his ascription of the
Round Towers “a mathematical demonstration by all the varieties
and modes of proof”; and further, that upon such demonstration
they should at once award him the gold medal and premium, “or, if
that could not be recalled, an equivalent gold medal and premium”—
not that, as he is careful to assure them, this offer was to be
construed as an admission that his original essay was not “all-
sufficient, all-conclusive, all-illustrative, and all-convincing.” As was
only to be expected, the reply sent to this challenge ran to the effect
that, “whatever might be the merits of any additional matter
supplied to them after the day appointed by advertisement, the
Academy could not make any alteration or revocation of their
award.” Then came the rejoinder,—“I do not want them either to
‘alter’ or ‘revoke’ their award; but simply to vote me ‘an equivalent
gold medal and premium’ for my combined essay, or, if they prefer,
the new portion of it. Should this be refused, I will put my cause into
the hands of the great God who has enlightened me, and make Him
the Umpire between me and the Academy.”[11] One is not surprised
to learn that “no answer was received to this communication,”
which, as already pointed out, may have afforded one of the reasons
why the Academy declined to publish the essay in their Transactions.
We may sympathise with O’Brien’s disappointment, and even go
further in deprecation of the attitude assumed by the Academy; but
it is impossible to deny that his conduct showed a want of dignity
and common sense, excusable only on the ground of youth.
As regards the Academy’s decision, assuming that the competition
was conducted fairly,—and, a priori, everything seemed in favour of
that assumption,—it is not easy to see how it could well have been
other than it was. With all possible admiration for O’Brien’s talents
and learning, candour obliges us to own that his essay—taken
merely as a literary performance—was inferior to that of his rival.
Apart from the question as to whether his theory was the true one,
and that of Dr. Petrie the reverse, the Academy were in a manner
bound by regard for their own dignity, and by the literary standard
then prevailing, to withhold the meed of their unqualified approval
from a composition which violated in so many respects the
established precedents of literary “form,” not to mention the canons
of good taste. Besides, O’Brien was, in archæological matters, so far
in advance of his generation, that a body of elderly gentlemen, who
simply represented the standard of knowledge prevalent at the time,
might well be excused for declining to follow him. They had, in fact,
to decide between the respective merits of two essays,—one of
which was well put together, conforming, at least in appearance, to
the stipulated conditions, expressing the most approved views,
bearing the marks of careful and systematic investigation and of
superior technical knowledge, also of literary skill much above the
average; the other, daring, novel, incoherent, propounding views
which were not only unfamiliar, but even shocking, to grave and
reverend seignors, rambling in method, deficient in proof, and
slipshod in language. Was it not, then, almost inevitable that they
should have preferred the former? But if one has to pronounce upon
the way in which the competition was started, carried on, and finally
decided, we are by no means sure that O’Brien had not some reason
to complain. First of all, with regard to his charge of the Academy
having awarded the prize to a member of their own Council, the
evidence to support it is primâ facie strong. Upon turning to vol. xvi.
of the Transactions of the Royal Irish Academy, we find the names of
“The Committee of Antiquities of the Council” for the year 1830 (that
in which the competition was first invited) given as follows:—“Isaac
D’Olier, LL.D.; Thomas Herbert Orpen, M.D.; Hugh Ferguson, M.D.;
Sir William Betham; John D’Alton, Esq.; George Petrie, Esq.; and the
Rev. Cæsar Otway.” In the next volume of the Transactions,
extending to 1837, the above list is given without any alteration,
except that Mr. D’Alton’s name is omitted, that of the Dean of St.
Patrick’s being substituted. From this the inference seems only
natural that “George Petrie, Esq.,” was a member of the Council
(being likewise, as we find, “antiquarian artist to the Academy”) at
the time when the idea of offering a prize for an essay on the Round
Towers was first started; that he continued to be a member while
the competition was in progress, and was actually one when the said
prize was adjudicated. Next, as to the charge that the Academy had
predetermined to award the prize to a member of its own Council,
we have the very compromising letter of the Rev. Mr. Otway (himself
a member of the Council) to the editor of the Dublin Penny Journal,
which is cited in the Preface to the first edition of this work,[12]
coupled with those repeated postponements of the date for sending
in essays, which O’Brien assures us were inexplicable on any other
ground than that of giving Mr. Petrie time to finish his essay. We are
far from contending that the reasons adduced in support of both
these charges should weigh against the high repute which the Royal
Irish Academy has always enjoyed from the time of its foundation;
still, it is impossible to deny that, in the absence of all satisfactory
explanation,—at least so far as we have been able to discover any,—
they wear a rather ugly look.
O’Brien was resolved that, as the Academy would not publish his
essay, he must do so himself; but in the meantime he had been
engaged upon a translation of Dr. Villanueva’s Ibernia Phœnicia,
which appeared in 1833. Personal liking for the author must have
been his motive for undertaking this task, as his own views do not
always harmonise with those of the Spanish savant; and certain
letters which are quoted in the “Translator’s Preface” show that the
two were very intimate. Having made this concession to friendship,
he busied himself with the production of an enlarged and amended
version of his essay. The first edition of this was published, early in
1834, by Whittaker & Co. of London, and J. Cumming of Dublin. It
seems to have met with a ready sale, for a second edition appeared
during the same year, bearing the imprint of Parbury & Allen,
London, and J. Cumming, Dublin. Both editions are in octavo, and to
outward appearance uniform, but differ in some respects. On the
title-page of the first it is described as the “Prize Essay of the Royal
Irish Academy, enlarged”—a description omitted in the second.
Further, the title itself is given as “The Round Towers of Ireland (or
the Mysteries of Freemasonry, of Sabaism, and of Budhism, for the
first time unveiled)”; but the words within brackets are absent from
the title-page of the second. A few corrections, too, appear in the
latter edition; but, upon the whole, it is not much more carefully
edited than the first—the curious omission of chapters vii. and xxxii.
being common to both. What is known in the book-trade as “The
Long Preface,” together with an amusingly comprehensive
“Dedication,” is omitted from the second edition, a much more
commonplace dedication to the Marquis of Lansdowne (described, of
course, as “The Mæcenas of his age”) being substituted for the
latter. As the second, and last, edition is that which had the author’s
latest revisions, it has been thought advisable to reproduce it in the
present issue. No interference with its text has been attempted—
typography and pagination being alike preserved. Nor has anything
in the shape of comment been inserted. A few supplementary
additions to the original work will probably not be considered out of
place. Together with this Introduction, they comprise a “Synopsis,” of
which the object is to assist readers in following the track of the
main argument—not always an easy task in the face of the author’s
numerous divagations, annotated lists of the principal Round Towers
and crosses, and an Index to the body of the work.
The reception accorded to the book by those whose verdict was
most important to its success, was decidedly hostile, and—what
must have been especially galling to a man like O’Brien—took the
shape of ridicule. Though it cannot be said that he had given no
occasion for the latter, it is equally apparent that much of it was
owing to ignorance; for there is not to be found among all the
censorious judgments of those “irresponsible reviewers” a single
attempt at sterling criticism. They attacked his style, and they
laughed his theory out of court, but they never resorted to anything
that deserved to be called refutation; and showed plainly by the
character of their strictures that they were quite in the dark with
respect to the nature of the evidence which he adduced in support
of his statements. It was profanely said of the late Professor Jowett,
that whatever he did not happen to know was held by him not to be
knowledge; and such was the view which his critics seem to have
taken of O’Brien’s dependence upon Eastern authorities, with which
they themselves were unfamiliar. As occasionally happens in Irish
affairs, a countryman of his own led the attack. In one of the
weakest articles that ever appeared in the Edinburgh Review,[13]
Moore, the poet, accused O’Brien of plagiarism and other misdeeds.
Considering the extent of Moore’s acquaintance with Oriental
literature, and the character of his mind, it is perhaps not surprising
that he mistakes the whole drift of O’Brien’s argument, fails to
perceive the force of those analogies upon which the latter chiefly
relied, and, in fact, only succeeds in proving his own incapacity as a
critic. But it is less conceivable that he should seek to overwhelm a
young aspirant for literary honours, who was of his own nationality,
and with whom he was on terms of at least nominal friendship, with
unfounded charges and clumsy ridicule. The secret of this otherwise
unaccountable severity is disclosed to us by “Father Prout,” in his
article on “The Rogueries of Tom Moore.” From it we learn that
Moore had endeavoured unsuccessfully to secure the co-operation of
O’Brien in his forthcoming History of Ireland, and that, upon the
negotiation falling through, a “coolness” ensued between the two. As
“Father Prout” had the whole correspondence laid before him, the
story does not rest upon O’Brien’s own version of what took place.
But, be it reliable or not, there is no denying that the poet went out
of his way—and out of his depth, too—in the effort to crush a young
author, who might fairly be supposed to have some claim upon his
sympathy. The scent which Moore thus struck was followed up by
the whole critical pack. The Gentleman’s Magazine, for instance,[14]
without attempting anything like serious criticism, quizzed O’Brien
unmercifully. He committed the fatal indiscretion of sending a
lengthy, but for him most temperate, reply, in which he is fain to cite
the Freemason’s Quarterly Review as his solitary backer. The
Gentleman’s Magazine reserved its answer until he was no more;
when, in an obituary notice (November 1835), it flung back this
retort: “Fondly imagining that he was the author of most profound
discoveries, and as it were the discoverer of a new historical creed,
Mr. O’Brien was always in a state of the highest excitement; and
when his lucubrations were treated with ridicule instead of serious
refutation, he was acutely irritated”—which last observation
somehow reminds one of that fastidious man-o’-war’s man, who,
whether the bo’sun “hit him high or hit him low,” took no pleasure in
being flogged. In fact, there was no real scholarly criticism of the
book from any quarter, though its eccentricities of style and
treatment received due attention. Superficially regarded, indeed, it
bristled with salient points for attack, and of these the gentlemen of
the press naturally availed themselves. They described it as “wild
and extravagant”—and no one could say them nay; but they failed to
point out, probably because they failed to see, that under this same
wildness and extravagance lay profound knowledge of a most
unusual kind, powerful if somewhat erratic reasoning, and the only
theory as to the genesis of ancient Irish proficiency in the arts of
civilisation which is consistent with the traditions, customs,
superstitions, folk-lore, and antiquities of the country.
O’Brien had now settled in London, where such time as could be
spared from his tutorial duties was spent in the study of his favourite
literature. It appears that he had at least two works then in
contemplation—one a Dissertation on the Pyramids, partly written,
and the other a Celtic Dictionary—which latter project excited the
ribaldry, altogether unfounded,[15] of certain critics. His health, never
strong, was now such as to cause some apprehension to his friends;
still he was able to share the pleasures which London life affords. He
went into the fashionable world—which, by the way, does not appear
to have taken him quite seriously, while acknowledging his talents
and erudition. The Marquis of Lansdowne’s house was open to him;
and mainly, no doubt, through the influence of that kindly nobleman,
he was even presented at Court. The military career, for which, as he
informs us (p. 130), he had a predilection second only to “his love
for truth and the rectification of his country’s honour,” was no longer
an object of ambition; and he may be regarded as having resigned
himself contentedly to the peaceful avocations of a man of letters.
Bad health, aggravated by his studious habits, seems indeed to have
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com