QFT Lecture Notes
Uploaded by
RajiQFT Lecture Notes
Uploaded by
RajiQuantum Field Theory
Lecture Notes Summer Term 2022
Nicolai Lang*
Institute for Theoretical Physics III
University of Stuttgart
Updated August 3, 2022
* [email protected]
Contents
Preliminaries 4
1 Elements of Classical Field Theory 8
1.1 Lagrangian and Hamiltonian Formalism . . . . . . . . . . . . . . . . . . . . . . 8
1.2 Symmetries and Conservation Laws . . . . . . . . . . . . . . . . . . . . . . . . 11
2 The Klein-Gordon Field 19
2.1 Canonical Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 The Klein-Gordon Field in Space-Time . . . . . . . . . . . . . . . . . . . . . . 22
3 The Dirac Field 29
3.1 The Dirac Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Free-Particle Solutions of the Dirac Equation . . . . . . . . . . . . . . . . . . . 33
3.3 Dirac Field Bilinears . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.4 Quantization of the Dirac Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.5 Discrete Symmetries of the Dirac Theory . . . . . . . . . . . . . . . . . . . . . 44
4 Interacting Fields and Feynman Diagrams 49
4.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2 Perturbation Expansion of Correlation Functions . . . . . . . . . . . . . . . . . 51
4.3 Wick’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.4 Feynman Diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.5 Cross Sections and the S -Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.6 Computing S-Matrix Elements from Feynman Diagrams . . . . . . . . . . . . . 70
4.7 Feynman Rules for Quantum Electrodynamics . . . . . . . . . . . . . . . . . . . 78
5 Elementary Processes of Quantum Electrodynamics 89
5.1 Cross section of e C e ! C scattering . . . . . . . . . . . . . . . . . . . . 89
5.2 Summary of QED calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6 Radiative Corrections of QED 94
6.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
6.2 Soft Bremsstrahlung . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
6.3 The Electron Vertex Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.3.1 Formal Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.3.2 The Landé g-factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.3.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
6.3.4 The Infrared Divergence . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.3.5 Summation and Interpretation of Infrared Divergences . . . . . . . . . . 113
6.4 Field-Strength Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.4.1 Structure of Two-Point Correlators in Interacting Theories . . . . . . . . 121
6.4.2 Application to QED: The Electron Self-Energy . . . . . . . . . . . . . . 125
6.5 Electric Charge Renormalization . . . . . . . . . . . . . . . . . . . . . . . . . . 130
7 Systematics of Renormalization 141
7.1 Counting UV-Divergences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2 Renormalized Perturbation Theory . . . . . . . . . . . . . . . . . . . . . . . . . 150
8 Functional Methods 158
8.1 Path Integrals in Quantum Mechanics . . . . . . . . . . . . . . . . . . . . . . . 158
8.2 Path Integrals for scalar fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
8.3 Application: Quantization of the Electromagnetic Field . . . . . . . . . . . . . . 166
9 Non-Abelian Gauge Theories 170
9.1 The Geometry of Gauge Invariance . . . . . . . . . . . . . . . . . . . . . . . . 171
9.2 The Yang-Mills Lagrangian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
10 Excursions 179
10.1 The Higgs Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
10.1.1 Abelian Example: The Standard Approach . . . . . . . . . . . . . . . . 179
10.1.2 Bonus: A Gauge-Invariant Approach . . . . . . . . . . . . . . . . . . . . 184
10.2 The Standard Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
10.2.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
10.2.2 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
10.2.3 The Glashow-Weinberg-Salam Theory . . . . . . . . . . . . . . . . . . . 189
10.2.4 Quantum Chromodynamics . . . . . . . . . . . . . . . . . . . . . . . . 198
10.2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
Bibliography 203
LE C T U R E 0
Preliminaries
Requirements for this course
For this course, we assume that students are familiar with the following concepts:
• Non-relativistic quantum mechanics and second quantization
• The Lagrangian and Hamiltonian formalism of classical mechanics
• Special theory of relativity and tensor calculus
• Complex analysis (contour integrals, residue theorem, ...)
Literature recommendations
• Weinberg: The Quantum Theory of Fields (Volume 1) [1]
ISBN 978-0-521-67053-1
Standard reference, very rigorous & mathematical, #formulas/#text = high
• Itzykson & Zuber: Quantum Field Theory [2]
ISBN 978-0-486-44568-7
Standard reference, #formulas/#text = high
• Peskin & Schroeder: An Introduction to Quantum Field Theory [3]
ISBN 978-0-201-50397-5
Standard reference for courses on QFT, #formulas/#text = medium
• Zee: Quantum Field Theory in a Nutshell [4]
ISBN 978-0-691-14034-6
Compact and pedagogical introduction to the field, #formulas/#text = low
For a first introduction to QFT, Peskin & Schroeder is a good choice (which we will use in this
course). Then, if you are hooked and want to understand QFT in depth (in particular its
mathematical foundations) read Weinberg afterwards.
Goals of this course
The goal of this course is to gain a thorough understanding of relativistic quantum field theory,
the concepts of Feynman diagrams, renormalization for quantum electrodynamics, and to
extend this knowledge to non-abelian gauge theories. In particular (★ optional):
• Relativistic quantum mechanics (Klein-Gordon and Dirac field)
• Quantization of free fields
• Perturbative analysis of interacting fields
• Feynman rules and diagrams
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 4
LE C T U R E 0
• Elementary processes and first corrections of quantum electrodynamics
• Path integral formalism
• Renormalization
• Non-abelian gauge fields ★
• The Standard Model ★
This course follows and partially covers Part I (field quantization, perturbation theory, Feynman
rules) and Part II (path integrals, renormalization) of “An Introduction to Quantum Field
Theory” by Peskin & Schroeder. If there is time, we close with a brief perspective on Part III
(non-abelian gauge theories, standard model).
Notes on this document
• This document is not an extension of the material covered in the lectures but the script
that I use to prepare them.
• Please have a look at Peskin & Schroeder for more comprehensive coverage; the
corresponding pages are noted in the headers (→ PS:xx–yy).
• The content of this script is color-coded as follows:
– Text in black is written to the blackboard.
– Notes in red should be mentioned in the lecture to prevent misconceptions.
– Notes in blue can be mentioned/noted in the lecture if there is enough time.
– Notes in green are hints for the lecturer.
• One page of the script corresponds roughly to one covered panel of the blackboard.
• Enumerated lists are used for more or less rigorous chains of thought:
1| This leads to …
2| this. By the way:
i| This leads to …
ii | this leads to …
iii | this.
3| Let’s proceed …
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 5
LE C T U R E 0
Key
The following abbreviations and glyphs are used in this document:
cf confer (“compare”)
dof degree(s) of freedom
eg exempli gratia (“for example”)
etc et cetera (“and so forth”)
et al et alii (“and others”)
ie id est (“that is”)
viz videlicet (“namely”)
vs versus (“against”)
wlog without loss of generality
wrt with respect to
^ “consider”
! “therefore”
$ non-obvious equality that may require lengthy, but straightforward calculations
D non-trivial equality that cannot be derived without additional input
ı
! “it is easy to show”
! “it is not easy to show”
) logical implication
^ logical conjunction
_ logical disjunction
repeated expression
anonymous reference
w/o “without”
w/ “with”
→ internal forward reference (“see below/later”)
← internal backward reference (“see above/before”)
↑ external reference to advanced concepts (“have a look at an advanced textbook on…”)
↓ external reference to basic concepts (“remember your basic course on…”)
→ reference to previous or upcoming exercises
★ optional choice/item
⁂ implicit or explicit definition of a new technical term (“so called …”)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 6
LE C T U R E 1 → P S : x i –x v i
→ Topics of Lecture 1
1. Ontology of quantum field theory in high-energy physics and condensed matter physics
2. Basics of classical field theory (Hamiltonian, Lagrangian)
3. Symmetries and conservation laws
→ Topics of Problemset 1
1. Functional derivatives
2. Lorentz covariance
3. Maxwell equations
Before we start our journey, here a few general remarks:
• Quantum field theory (QFT) is concerned with the quantization of fields that live on
smooth manifolds (e.g. Euclidean space-time, Minkowski space-time).
• The most prominent example (which you probably already encountered in one form or
the other) is the quantum theory of the electromagnetic field, which also initiated the field at
the beginning of the 20th century.
• As Maxwell theory has special relativity “built in,” its QFT must be relativistic as well
(i.e., the Lorentz group must be a global symmetry). Relativistic QFTs are standard in
high-energy physics and will be the focus of this course.
• QFTs are riddled with infinities in their expressions, which makes it hard to define them
rigorously as mathematical objects (this is still true for most of them).
• In the mid of the 20th century, the technique of renormalization was developed to
systematically deal with these infinities and extract physical predictions. This was a
crucial step to make quantum electrodynamics (and the standard model afterwards) a
useful and accepted QFT.
• At the same time, the development of Feynman diagrams as a systematic approach to
perturbation theory paved the way to successful applications of QFTs, in particular
quantum electrodynamics.
• In the second half of the 20th century, the toolbox of QFT was imported from high-
energy physics into condensed matter physics for effective, large scale & low energy
descriptions of many-body systems (such as magnets and superconductors); in particular,
it proved useful for the description of phase transitions (↑ Conformal Field Theories).
These QFTs are typically not relativistic as the Lorentz group is not a symmetry of
condensed matter systems.
• In the context of condensed matter physics, the method of renormalization is less
opaque and has a physical interpretation. These insights led to a better understanding of
renormalization in high-energy physics as well.
• While the methods of QFT in high-energy physics and condensed matter physics are
very similar, their ontology is very different:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 7
LE C T U R E 1 → PS:15–19
– In high-energy physics, fields are elementary and particles are emergent
– In condensed matter physics, particles are elementary and fields are emergent
! In this course, fields are the fundamental entities of the world; particles are emergent,
effective models for localized excitations of these fields.
1 Elements of Classical Field Theory
1.1 Lagrangian and Hamiltonian Formalism
Recap: Classical mechanics of “points”
With “points” we mean a discrete set of degrees of freedom.
1| Degrees of freedom qi labeled by i D 1; : : : ; N
2| Lagrangian L.fqi g; fqP i g; t / D T V
We write q for fqi g D fq1 ; : : : ; qN g.
T is the kinetic, V the potential energy.
3 | Action SŒq D dt L.q.t /; q.t
R
P /; t / 2 R
This is a functional of trajectories q D q.t /.
4| Hamilton’s principle of least action:
ıS Œq Š
Z
Š
D0 , ıS D dt ıL D 0 (1.1)
ıq
ı denotes functional derivatives/variations (→ Problemset 1).
5| Euler-Lagrange equations (i D 1; : : : ; N ):
@L d @L
D0 (1.2)
@qi dt @qP i
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 8
LE C T U R E 1 → PS:15–19
Analogous: Lagrangian Field Theory
Now we consider a continuous set of degrees of freedom:
1| ^ One or more fields .x/ on spacetime x 2 R1;3 =R4 with derivatives @ .x/
where @0 D @ t and @iD1;2;3 D @x;y;´
(R1;3 : Minkowski space, R4 : Euclidean space; in the following, we focus on R1;3 )
2| Lagrangian density L. ; @ ; x/
Most general form: L.f
R 3k g; f@ k g; fx g/. No explicit x -dependence in the following!
! Lagrangian L D d x L. ; @ /
(We omit the “density” in the following.)
3| Action:
Z Z Z
S ΠD dt L D dt d 3x L. ; @ / D d 4x L. ; @ / (1.3)
SŒ is a functional of “field trajectories” in R1;3 .
4| Action principle:
Z
Š
0 D ıS Œ D d 4x ıL (1.4a)
@L @L
Z
4
D d x ı C ı.@ / (1.4b)
@ @.@ /
Add zero and use ı.@ / D @ .ı /
@L @L @L
Z
4
D d x ı @ ı C @ ı (1.4c)
@ @.@ / @.@ /
Gauss theorem
@L @L @L
Z Z
D dA ı C d 4x @ ı (1.4d)
@ @.@ / „ƒ‚… @ @.@ /
D0 „ ƒ‚ …
D0
Note that is fixed on the boundary @ and therefore ı D 0.
The second term vanishes because the integral must vanish for arbitrary variations ı .
5| Euler-Lagrange equations (one for each field ):
@L @L
@ D0 (1.5)
@ @.@ /
Note the Einstein summation over repeated indices.
This expression is manifestly Lorentz invariant if L is a Lorentz scalar.
Recap: Hamiltonian Mechanics
Lagrangian Legendre transformation Hamiltonian
! (1.6)
P t/
L.q; q; Conjugate momentum H.q; p; t / D p qP P t/
L.q; q;
@L
p @qP
, qP D q.p/
P
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 9
LE C T U R E 1 → PS:15–19
Analogous: Hamiltonian Field Theory
1| ^ x D xi ¶ i discrete spatial coordinates:
@L @L @
Z
D pi ¶ p.x/ D D d 3y L. .y/; .y//
P (1.7a)
@qP i P
@ .x/ P
@ .x/
X @ @L
d 3y P
L. .y/; .y// D d 3x (1.7b)
@ P
.x/ @ P
.x/
y „ ƒ‚ˇ … „ƒ‚…
ıx;y @L ˇ .x/
P ˇyDx
@
(In L. .y/; .y//
P we omit the time dependece!)
! Momentum density conjugate to is D @LP
@
2| Hamiltonian:
.x/ d 3x P d 3x
P
x L. .x/; .x//
X ‚…„ƒ ‚…„ƒ
H D P
p.x/ .x/ L (1.8)
x
Therefore
Z
d 3x .x/ .x/
P P
˚
H D L. ; / (1.9)
„ ƒ‚ …
Hamiltonian density H . ;/
Note that P D ./.
P
→ Example 1.1: Free scalar field
1| Real field W R3 R ! R with .x;
E t / 7! .x;
E t / D .x/
2| Lagrangian (density): L D 12 .@ t /2 12 .r /2 12 m2 2 D 12 .@ /2 12 m2 2
It is .@ /2 @ @ D .@ t /2 .@x /2 .@y /2 .@´ /2 with signature
g D diag .1; 1; 1; 1/. Note that then @ @ D @2t r 2 .
3| Interpretation:
In L, m is refered to as mass. This is not the inertial mass of the pendula but the
stiffness of the harmonic potential!
Continuum of spring-coupled pendula for m D 0 , 1D rubber band
4| Equation of motion (“field equation”):
m2 @ .@ / D 0 , .@ @ C m2 / D 0 (1.10)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 10
LE C T U R E 1 → PS:15–19
This is the classical (!) ⁂ Klein-Gordon equation.
5 | Conjugate momentum field: D @L P
D P
@
6| Hamiltonian (density):
1 P2 1 1
C .r /2 C m2 2
H D P (1.11a)
2 2 2
1 2 1 2 1 2 2
D C .r / C m (1.11b)
2 2 2
The Hamiltonian is H D d x H . ; /.
R 3
1.2 Symmetries and Conservation Laws
What follows is based on Sénéchal “Conformal Field Theory” (pp. 36–42,45–46) [5].
1| ^ Transformations of coordinates and fields:
x 7! x 0 D x 0 .x/ and .x/ 7! 0 .x 0 / D F . .x// (1.12)
Two effects: coordinates and fields transformed
These are active transformations that change physics!
→ Example 1.2: Rotation of a vector field E
i| ^ 3-component field E D . 1 ; 2 ; 3 / and R 2 SO.3/ rotation:
ii | xE 0 D RxE and E 0 .x 0 / D R .x/
E E
D R .R 1x0/
This defines a ⁂ vector field.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 11
LE C T U R E 1 → PS:15–19
2| Change of the action under !
7 0:
Z
S S ΠD d dx L. 0 .x/; @ 0 .x//
0 0
(1.13a)
Rename integration variables x ! x 0
Z
D d dx 0 L. 0 .x 0 /; @0 0 .x 0 // (1.13b)
Definition
Z
D d dx 0 L.F . .x//; @0 F . .x/// (1.13c)
Substitution
ˇ 0ˇ
ˇ @x ˇ
ˇ L F . .x//; @x @ F . .x//
Z
D d dx ˇˇ (1.13d)
@x ˇ @x 0
Skip first step, use colors for primes.
→ Example 1.3: Translations
1| x 0 WD x C a and 0 .x 0 / WD .x/ D .x 0 a/
This defines a ⁂ scalar field.
@x
2| F D 1 trivial, 0 .x 0 / D F . .x// D .x.x 0 //, and @x 0
D ı
3| Action:
Z Z
0 d 0 0
SΠD d x L. .x/; @ .x// D d dx L. .x/; @ .x// D S Π(1.14)
The action is translation invariant: S D S 0 !
This follows generally from the missing x-dependence of L for scalar fields.
→ Example 1.4: Scale transformations
1| x 0 WD x and 0 .x 0 / WD .x/ D . 1 x 0 /
is the ⁂ scaling dimension of the field
ˇ ˇ
2 | F . / D and @x @x 1 ı and ˇ @x 0 ˇ D d
0 D ˇ @x ˇ
3| Action:
Z
0 d
S ΠD d dx L.
.x/; 1
@ .x// (1.15a)
^ Massless scalar field: S ΠD 12 d dx .@ /2
R
Z
d 2 2
D d dx L. .x/; @ .x// D d 2 2 S Π(1.15b)
d
! S 0 D S iff D 2 1
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 12
LE C T U R E 1 → PS:15–19
This is an example of a ↑ Conformal Field Theory (CFT).
→ Example 1.5: Phase rotation
1| x 0 WD x and 0 .x 0 / WD e i .x/
! There are symmetries that only transform the fields but not the coordinates.
ˇ 0ˇ
@x
2 | F . / D e and @x 0 D ı and ˇ @x
i D1
ˇ ˇ
@x ˇ
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 13
LE C T U R E 2 → PS:15–19
→ Topics of Lecture 2
1. Infinitesimal transformations and continuous symmetries
2. Noether’s theorem and conserved quantities
3. Application to the energy-momentum tensor
Infinitesimal Transformations
We are interested in continuous symmetries (↓ Lie groups).
1| ^ Infinitesimal transformations (IT):
ıx ıF
x 0 D x C wa .x/ and 0 .x 0 / D .x/ C wa .x/ (1.16)
ıwa ıwa
Here, wa denotes infinitesimal parameters of the transformation (sum over a implied!).
They may vary from point to point: wa D wa .x/ (see below).
2| Generator of IT:
ıw .x/ WD 0 .x/ .x/ iwa Ga .x/ (1.17)
With (omit first line and refer to previous equation)
ıF
0 .x 0 / D .x/ C wa .x/ (1.18a)
ıwa
ıx ıF 0
D .x 0 / wa @ .x 0 / C wa .x / C O.w 2 / (1.18b)
ıwa ıwa
it follows (replace x 0 by x; these are just labels!)
ıx ıF
iGa D @ (1.19)
ıwa ıwa
This function describes the infinitesimal change of the field at the same point.
→ Example 1.6: Translations
ıx ıx
1| x 0 WD x C w x C w ıw with ıw
D ı
ıF
2 | ıw D 0 (For a scalar or a vector field.)
3| @
iG D ı 0 and therefore
G D i @ P (1.20)
! The “momentum operator” generates translations.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14
LE C T U R E 2 → PS:15–19
→ Example 1.7: Scale Transformations
G D ix @ D ! Generates “dilations” in spacetime.
ıF
This simple form is valid for a scalar field with scaling dimension D 0 so that ı
D 0.
→ Example 1.8: Spatial Rotations
G D i.x @ x @ / C S for ; D 1; 2; 3
The first term generates coordinate rotations (↓ orbital angular momentum operator).
S are spin matrices that generate field transformations (for non-scalar fields).
Question: What generates G if either D 0 or D 0? Answer: Boosts.
Noether’s Theorem
1| ^ Transformation Eq. (1.16) which is a
Symmetry of the action W, SΠD S Π0 (1.21)
for wa independent of x (⁂ rigid transformation).
2| Assume that Eq. (1.16) is not rigid: wa D wa .x/
We assume that wa is sufficiently smooth so that @ wa is infinitesimal as well, i.e.,
O.wa / D O.@ wa /.
0
ˇ 0ˇ
ıx ıx
3 | Jacobian: @x C@ ˇ @x ˇ
@x D ı w a ıwa
! ˇ @x ˇ D 1 C @ w a ıwa
Use det.1 C A/ D 1 C TrŒA C O.A2 /.
@x ıx
4| Inverse Jacobian matrix: @x 0 D ı
@ w a ıwa
This is true in linear order of wa and @ wa .
5| Use Eq. (1.13d):
ıx
Z
0 d
S D d x 1 C @ wa (1.22)
ıwa
ıx
ıF ıF
L C wa ; ı @ w a @ C @ wa
ıwa ıwa ıwa
@wa
6| Expand in 1st order of wa and @x
7| ^ ıS S 0 S ! Only terms / @w a
@x
remain
Because the transformation is a symmetry of the action by assumption, i.e., for
wa D const (a rigid transformation) it is S 0 D S !
This is equivalent to the definition of a symmetry (of the action).
8| For generic, non-rigid transformation we find
Z
ıS D d dx ja @ wa (1.23)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 15
LE C T U R E 2 → PS:15–19
with the current
ıx
@L @L ıF
ja $ @ ı L (1.24)
@.@ / ıwa @.@ / ıwa
ıx ıF
associated to the IT ıw a
and ıwa
.
This is only true for transformations that are symmetries of the action!
9 | Integration by parts ! ıS D d dx wa @ ja
R
Here we assume that the variations wa .x/ vanish on the boundaries (possibly at infinity).
10 | ^ that obeys the equations of motion ! ıS D 0 for arbitrary variations 0 D C ı
In particular, for arbitrary non-rigid transformations wa .x/!
It follows Noether’s (first) theorem:
@ ja D 0 8x;a (1.25)
This is a conservation law with conserved current ja .
11 | Conserved charge:
Z
Qa WD d d 1x ja0 (1.26)
Space
Indeed:
dQa
Z Z Z
Noether Gauss
D d d 1x @0 ja0 D d d 1x @k jak D dk jak D 0
dt Space Space Surface
(1.27)
Here we assume that jak 0 on the boundaries—typically at spatial infinity, i.e., the
universe is closed. k D 1; 2; 3 denotes the spatial coordinates.
→ Note 1.1
The current Eq. (1.24) is called canonical current as there is an ambiguity:
jQa WD ja C @ Ba with Ba D Ba arbitrary ) @ jQa D 0 (1.28)
→ Note 1.2
Symmetric Lagrangian ) Symmetric action ) Symmetric EOMs (1.29)
„ ƒ‚ …
! Conserved currents
Continuous symmetries of the EOMs do not imply conserved currents!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 16
LE C T U R E 2 → PS:15–19
Application: The Energy-Momentum-Tensor (EMT)
Special relativity:
Global spacetime symmetries (Lorentz transformations + Translations = Poincaré group)
General relativity:
Local spacetime symmetries (↑ Diffeomorphisms ! Gauge symmetries)
ıx ıF
1| ^ Infinitesimal spacetime translations: x 0 D x C " ! ı"
D ı ,ı"
D0
2| ^ Translation-invariant action: S 0 D S
This includes translations in time!
3| Conserved currents:
ıx
@L @L
T D @ ı L
D @ ı L (1.30)
@.@ / ı"
„ƒ‚… @.@ /
ı
@L
T D g T D @ g L (Energy-Momentum Tensor)
@.@ /
(1.31)
with @ T D 0 and four conserved charges
Z
P D d 3x T 0 (1.32)
Note that these quantities are only conserved for classical solutions of the EOMs.
4| Energy ( D 0) (skip first step):
Z
@L P
Z Z
0 3 00 3
P D d xT D d x L D d 3x H . ; / D H (1.33)
@ P
! The Hamiltonian is the component of a 4-vector and not Lorentz invariant!
By contrast, the Lagrangian is Lorentz invariant (for relativistic field theories).
5| Kinetic momentum ( D i ):
@L
Z Z Z
P i D d 3x T 0i D d 3x . @i / D d 3x @i (1.34)
@ P
is the canonical momentum.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 17
LE C T U R E 2 → PS:15–19
→ Note 1.3
In general T ¤ T for the canonical EMT. But:
TQ WD T C @ K with K D K (1.35)
Choose K such that TQ D TQ (↑ ⁂ Belinfante EMT )
→ Example 1.9: Electromagnetism (EM) in vacuum
Details → Problemset 1
1| Four-component gauge field: A D . ; A1 ; A2 ; A3 /
2| EM field tensor: F D @ A @ A
Contains E- and B-field components.
1
3| Lagrangian: Lem .A; @A/ D 4 F F
4 | Action: Sem D d 4x Lem
R
5| Euler-Lagrange equations: @ F D 0 (inhomogeneous Maxwell equations)
6| Sem is Lorentz invariant and translation invariant (= Poincaré invariant)
! EMT = conserved currents
Why is this obvious?
@Lem
7| Canonical EMT: Tem D @.@ A /
@ A g Lem
8| Symmetric EMT using K WD F A :
1
TQem $ g F F F F (1.36)
4
• TQ 00 D 12 .E 2 C B 2 / (↓ Energy density)
• TQ 0i D .EE B/
E i (↓ Pointing vector)
• TQ ij D ij (↑ Maxwell stress tensor)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 18
LE C T U R E 3 → PS:19–26
→ Topics of Lecture 3
1. Canonical quantization of the Klein-Gordon field
2. Heisenberg picture: Time evolution of the quantized Klein-Gordon field
→ Topics of Problemset 2
1. The classical complex Klein-Gordon field
2. The quantized complex Klein-Gordon field
2 The Klein-Gordon Field
2.1 Canonical Quantization
1| Theory:
i| Real field .x/ (→ Problemset 2 for the complex analog)
ii | Lagrangian: L D 12 .@ /2 1 2 2
2m (⁂ free scalar field)
iii | EOM: .@2 C m2 / D 0 (Klein-Gordon equation)
iv | Hamiltonian: H D 12 2 C 12 .r /2 C 12 m2 2
2| Canonical quantization:
E D i ı .3/ .xE
E .y/
.x/; E
y/
E .y/
.x/; E D0 (2.1)
E .y/
.x/; E D0
with D , D (“real” field operators) and xE 2 R3 .
This is completely analog to the canonical quantization of “points” known from
undergraduate courses on quantum mechanics if Kronecker deltas are replaced by delta
distributions: Œqi ; pi D i ıij ! Œ .x/; .y/ D i ı.x y/.
For now, we are in the Schrödinger picture where the fields do not depend on time!
3| Goals:
• Representation of field operators ↑ Eq. (2.1)
• Spectrum of Hamiltonian
• Time evolution of field operators .x/
E 7! .x/
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 19
LE C T U R E 3 → PS:19–26
4| Motivation:
(The following facts are used to come up with a tentative representation for the field
operators; that it is correct can be later checked rigorously.)
i| Fourier transform of KG equation in space:
d 3p i pExE Q
Z
E t/ D
.x; e .p; E t/ (2.2)
.2/3
Then
2
E 2 C m2 / .
Q p;
@ t C .jpj E t/ D 0 (2.3)
! Decoupled harmonic oscillators with frequency !pE D jpj
p
E 2 C m2
and constraint Q .p; Q p;
E t / D . E t / (since D ).
ii | ^ Hamiltonian HSHO D 1 Q 2 C 12 ! 2 Q 2 and introduce
2 q
Q D p1 .a C a / and Q D i !2 .a a / with Œa; a D 1
2!
! HSHO D !.a a C 12 / (diagonal!)
5| This motivates the Field operators
d 3p 1
Z
i pEx
E i pEx
E
E WD
.x/ a e C a e
.2/3 2!pE pE pE
p
d 3p 1
Z
D a C a e i pExE
.2/3 2!pE pE pE
p
„ ƒ‚ … (2.4)
Q p/
. E
d 3p
r
!pE
Z
E WD
.x/ . i/ apE a e i pExE
.2/3 2 pE
„ ƒ‚ …
.
Q p/
E
(Use colors to skip second line.)
The pE is necessary to make the fields “real”: D and D .
with momentum modes
h i
apE ; aqE D .2/3 ı .3/ .pE qE/ (2.5)
ı
! Eq. (2.5)^ Eq. (2.4))Eq. (2.1).
6| Hamiltonian:
d 3p 1h
Z i
H $ ! a a C a ; a (2.6)
.2/3 pE pE pE 2 „ pEƒ‚ pE…
/ ı.0/D1
Ignore the infinite term since only relative energies are physical!
This infinity accounts for the zero-point energies of all harmonic oscillator modes.
Dropping this infinity is called normal ordering (→ later).
7| Eigenstates & Spectrum:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 20
LE C T U R E 3 → PS:19–26
ı
i| ! ŒH; apE D !pE apE
ii | Vacuum j0i ! Eigenstates apE aqE : : : j0i (span complete Hilbert space)
iii | Energy: EpE D !pE D C jpjE 2 C m2 (relativistic dispersion, positive energies!)
p
iv | (Kinetic) momentum:
d 3p i
Z Z
i 3
P D E
d x .x/. E $
@i / .x/ p apE apE (2.7)
.2/3
This is now an operator!
v| Statistics: apE aqE j0i D aqE apE j0i
! Excitations apE commute and carry additive energy & momentum
! Bosonic particles (in momentum space)
8| Normalization:
i| ^ ƒ D R0 L3 .ˇ/R 2 SOC .1; 3/ ! p 0 D .EpE0 ; pE0 / D ƒp with p D .EpE ; p/ E
Recall that all Lorentz transformations can be generated from spatial rotations and
a boost L3 .ˇ/ in ´-direction!
dp 0
0
@pE E 0
ii | Jacobian in space: det $ dp33 $ EpE
@pE E
p
EpE0
! ı .3/ .pE qE/ D EpE ı .3/ .pE0 qE0 /
! ı .3/ .pE qE/ is not Lorentz invariant but EpE ı .3/ .pE qE/ is!
Use colors to shorten this!
3D volumes are not invariant under boosts due to Lorentz contraction!
iii | Single-particle eigenstates:
q
jpi
E WD 2EpE apE j0i ) E q i D .2/3 2EpE ı .3/ .pE
hpjE qE/ (2.8)
„ ƒ‚ …
Lorentz invariant
This follows directly from the commutation relations.
The 2 is just convention.
9| Lorentz transformations ƒ 2 SOC .1; 3/:
We need a unitary representation of the Lorentz group SOC .1; 3/ on the Hilbert space!
s
EƒpE
U.ƒ/jpiE WD jƒpi E , U.ƒ/apE U 1 .ƒ/ D a (2.9)
EpE ƒpE
It is .ƒp/
E i ƒi p (i.e., the spatial projection).
Note that the “boost part” of ƒ is hidden in the normalization of the state!
10 | Interpretation of .x/:
E
d 3p 1
Z
i pEx
E
E
.x/j0i D e jpi
E (2.10)
.2/3 2EpE
For non-relativistic jpj
E m ) EpE const
! State jxi
E of particle at position xE
E creates particle at position xE
! .x/
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 21
LE C T U R E 3 → PS:19–26
• This interpretation is also consistent with the “position-space representation”
h0j .x/j
E piE $ e i pExE .
• The factor 2E1 suppresses large momenta and “smears out” the position of the
E
p
particle on length scales of its Compton wavelength c D 1=m (→ space-like
two-point correlation function h0j .x/ .
E y/j0i
E E ).
e mjxE yj
→ Note 2.1
R d 3p
• Projector on single-particle sector: 11 D .2/ E 2E1 hpj
3 jpi E
E
p
R d 3p f .p/
• ^ f .p/ Lorentz invariant ! .2/ 3 2E is Lorentz invariant
E
p
2.2 The Klein-Gordon Field in Space-Time
So far: Schrödinger picture
Now: Heisenberg picture
1| Heisenberg operators: .x/ D .x;
E t / D e iH t .x/e
E iH t (similar for .x/)
2| Heisenberg equation: i @ t O D ŒO; H for O D ; yields
1 2 1 1 2 2
Z
3 2
i@ t .x/ D .x/; d y E t / C .r .y;
.y; E t // C m .y;
E t/
2 2 2
Z
D d 3y i ı .3/ .xE y/.
E y; E t/
D i .x/ (2.11a)
2 2
i@ t .x/ $ i. r C m / .x/ (2.11b)
) .@2t r 2 C m2 / .x/ D 0 (Klein-Gordon equation) (2.12)
3| Time-evolution of modes:
e iH t apE e iH t
D apE e iEpE t
(2.13a)
e iH t apE e iH t
D apE e CiEpE t (2.13b)
Use colors to skip last row.
This can be shown informally by using HpE D EpE apE apE and counting excitations (i.e., on
the number basis).
4| Field operators:
d 3p 1
Z ˇ
ipx
.x/ D apE e C apE e ipx ˇ (2.14)
ˇ
.2/3 2EpE
p
p 0 DEpE
.x/ D @ t .x/ (2.15)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 22
LE C T U R E 3 → PS:19–26
Here, px D p x D EpE t E note that p 0 D EpE .
pEx;
In the following, apE and apE always denote the time-independent Schrödinger operators!
→ Note 2.2
1| Hamiltonian generates time translations:
E t/ D e iH t .x;
.x; E 0/ e iH t
„ ƒ‚ …
.x/
E
2| Total momentum operator generates space translations:
E E i PE xE
E D e i P xE .0/e
.x/
3| Four-momentum operator generates space-time translations:
.x/ D e iP x .0/e iP x
Here, P D .H; PE / where PE is defined in Eq. (2.7).
→ Note 2.3
Note that p 0 D EpE is always positive:
• e ipx $ positive-frequency solution of KG equation $ destruction operator apE
• e Cipx $ negative-frequency solution of KG equation $ creation operator apE
As single-particle wavefunctions, solutions with positive/negative frequency correspond to
solutions with positive/negative energy. Note that there are only excitations with positive
energy in the quantized field theory!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 23
LE C T U R E 4 → P S : 27 – 3 1
→ Topics of Lecture 4
1. Causality
2. Green’s functions of the Klein-Gordon theory
3. The Feynman propagator
Causality
^ Amplitude for a particle to propagate from y to x:
d 3p 1
Z
ip.x y/
D.x y/ h0j .x/ .y/j0i $ e (2.16)
.2/3 2EpE
This expression is Lorentz invariant, i.e., D.ƒ.x y// D D.x y/ for all ƒ 2 SOC .1; 3/
[more generally, for all orthochronous Lorentz transformations ƒ 2 OC .1; 3/].
This is not true for non-orthochronous Lorentz transformations which flip the sign of D.x y/
since EpE D p 0 7! p 0 D EpE !
1| ^ Time-like distance: x 0 y 0 D t and xE yE D 0
Z 1
4 p2 p
i p 2 Cm2 t
D.x y/ D dp e (2.17a)
.2/3 0
p
2 p 2 C m2
1
Z 1 p
D dE E 2 m2 e iE t (2.17b)
4 2 m
t !1
¤ 0 (actually not convergent) (2.17c)
t !1 imt
e (this is very hand-wavy) (2.17d)
! Does not vanish ! Propagation possible
2| ^ Space-like distance: x 0 y 0 D 0 and xE yE D rE
d 3p 1 i pE
Z
D.x y/ D e Er (2.18a)
.2/3 2EpE
Z 1
2 p 2 e ipr e ipr
D dp (2.18b)
.2/3 0 2EpE ipr
Z 1 ipr
i pe
D dp p (2.18c)
2
2.2/ r 1 p 2 C m2
Use Cauchy’s integral theorem with the following path:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 24
LE C T U R E 4 → P S : 27 – 3 1
cot
Oc— 5
• Show that the curved sections vanish for R ! 1 and " ! 0, respectively!
Showing that B; F ! 0 for R ! 1 is actually tricky and requires some kind of
regularization (that P&S are silent about) to exponentially suppress the oscillating
terms close to the real axis. One way to to fix this, is to focus on the asymptotics
r ! 1 (which is our goal here). The oscillating terms can then be exponentially
suppressed in the limit r ! 1 so that the contributions from the arcs become
negligible. Strictly speaking, the non-convergent integral Eq. (2.18c) should be
defined by such an appropriately chosen limit.
• It is C D E since the minus from the opposite direction and the branch cut cancel.
Then
D.x y/ D C E D 2C (2.19)
Z 1
i p e ipr
D dp (2.20)
.2/2 r im
p
p 2 C m2
Z 1
D ip 1 e r
D d (2.21)
4 2 r m
p
2 m2
r!1 mr
e (2.22)
! Vanishes exponentially (but non-zero!) ! Problem?
The integral Eq. (2.21) can be evaluated in terms of modified Bessel functions of the second
kind, the asymptotics of which is known and yields the given exponential decay. Note
that simply upper bounding the integrand by e r e mr leaves a diverging integral
behind so that one ends up with the useless upper bound e mr 1.
3| ^ Measurements A and B: can affect each other iff ŒA; B ¤ 0
Simplest choice: A D .x/ and B D .y/
Causality is preserved if all observables commute at space-like separations!
Since D @ t , it is sufficient for Π.x/; .y/ to vanish for .x y/2 < 0.
d 3p 1 d 3q 1
Z Z
Π.x/; .y/ D (2.23a)
.2/3 2EpE .2/3 2EqE
p p
h i
apE e ipx C apE e ipx ; aqE e iqy C aqE e iqy
d 3p 1 ip.x y/
Z
ip.x y/
D e e (2.23b)
.2/3 2EpE
D D.x y/ D.y x/ (2.23c)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 25
LE C T U R E 4 → P S : 27 – 3 1
Let .x y/2 < 0 space-like ! 9ƒ 2 SOC .1; 3/ W ƒ .x y/ D .x y/:
The proper orthochronous Lorentz group SOC .1; 3/ is a connected subgroup of the Lorentz
group O.1; 3/, the elements of which connect continuously to the identity.
N "Time
Continuous transformations (rotations in space and boosts) allow for .x y/ 7! .x y/
only if .x y/2 < 0. For time-like distances, this requires discontinuous transformations
(time-reversal).
Then
Œ .x/; .y/ D D.x y/ D.ƒ .y x//
.x y/2 < 0
D D.x y/ D.x y/ 0 (Causality) (2.24)
For time-like separation, .x y/2 > 0, there is no such continuous transformation and
the argument breaks down.
The first line follows from the Lorentz invariant integral measure in Note 2.1 and the
definition of the propagator in Eq. (2.16). Remember that D is only invariant under
orthochronous Lorentz transformations but picks up a minus sign under time inversion!
The Propagator
1| Since Π.x/; .y/ / 1 (the commutator is a c-number), we can write (x 0 > y 0 for now)
“c-number” historically denotes scalar multiples of the identity, i.e. classical/commut-
ing/complex “numbers”.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 26
LE C T U R E 4 → P S : 27 – 3 1
d 3p 1
Z
ip.x y/
h0j Π.x/; .y/ j0i D e e ip.x y/
(2.25a)
.2/3 2EpE
Substitute pE ! pE to obtain the second term:
( ip.x y/
e ip.x y/ jp0 D
)
Z
d 3p e jp0 DEpE EpE
D C (2.25b)
.2/3 2EpE 2EpE
Residue theorem with clockwise orientation (therefore the 1):
d 3p dp 0 1
Z Z
x 0 >y 0
D 2 2
e ip.x y/ (2.25c)
.2/3 2 i p m
„ ƒ‚ …
.p 0 EpE /.p 0 CEpE /
d 4p i
Z
ip.x y/
D e (2.25d)
.2/4 p 2 m2
with contours
The arc vanishes in the lower/upper-half pane for x 0 > y 0 and x 0 < y 0 , respectively.
Therefore
DR .x y/ .x 0 y 0 /h0j Π.x/; .y/ j0i D Eq. (2.25d) C (2.26)
2| Interpretation:
.@2 C m2 /DR .x y/ $ i ı .4/ .x y/ (2.27)
! Retarded Green’s function of Klein-Gordon operator
“R” for “retarded” since it vanishes for x 0 < y 0 .
We could have found Eq. (2.25d) directly from Eq. (2.27) by Fourier transformation.
3| Alternative contour :
©
O5
N N >
7 ”
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 27
LE C T U R E 4 → P S : 27 – 3 1
• x 0 > y 0 : close contour below
• x 0 < y 0 : close contour above
d 4p i
Z
ip.x y/
DF .x y/ Eq. (2.25d) C D 4 2
e (2.28)
.2/ p m2 C i "
(Feynman propagator)
The infinitesimal i " shifts the poles to p 0 ˙.EpE i "=2EpE / D ˙.EpE i / and
yields an equivalent prescription of the Feynman propagator without the need to specify
a contour. Note that "=2EpE are both infinitesimals.
We find (using Eq. (2.25d) and Eq. (2.16))
(
D.x y/ for x 0 > y 0
DF .x y/ D (2.29)
D.y x/ for x 0 < y 0
D .x 0 y 0 /h0j .x/ .y/j0i (2.30)
0 0
C.y x /h0j .y/ .x/j0i
h0jT .x/ .y/j0i (2.31)
with the time-ordering (meta-)operator T
T orders products of operators by time with the latest to the left.
It is a meta-operator as it operates on descriptions of operators. Note that this is different
from super-operators (such as the Lindbladian) which operate on operators.
The Feynman propagator is a Green’s function of the KG equation (with different
boundary conditions than the retarded/advanced Green’s functions).
Later: Feynman propagator & Interactions ! Feynman rules
However, so far we only studied the free KG field (! linear field equation). Without in-
teractions, however, there is no scattering so that there are no characteristic observations
possible.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 28
LE C T U R E 5 → PS:35–49
→ Topics of Lecture 5
1. (Review of the Poincaré group)
2. Construction of the Dirac equation
3. Free-particle solutions of the Dirac equation
→ Topics of Problemset 3
1. Fock states and coherent states
2. Free-particle solutions of the Dirac equation
3 The Dirac Field
3.1 The Dirac Equation
So far: Simplest relativistic field equation ! Klein-Gordon equation
Now: Second simplest relativistic field equation ! Dirac equation
1| Observation: Lorentz symmetry of the KG equation:
We view Lorentz transformations as active transformations, mapping solutions to different
solutions! This is equivalent to the passive viewpoint where the coordinate system is
transformed instead.
i| ^ Coordinate transformation: x 0 D ƒx & Field transformation: 0 .x 0 / D .x/
ii | ^ with .@2 C m2 / .x/ D 0 for all x
iii | ! 0 .x/ D .ƒ 1 x/ is a new solution:
Use the chain rule in the first step twice:
.g @ @ C m2 / 0 .x/ D Œg .ƒ 1
/ @ .ƒ 1
/ @ C m2 .ƒ 1
x/ (3.1a)
Use invariance of the metric
D .g @ @ C m2 / .ƒ 1
x/ (3.1b)
solution
D .@2 C m2 / .ƒ 1
x/ D 0 (3.1c)
2| Observation: ^ Vector fields under rotations: E 0 .x/ D R .R
E 1 x/
! In general, a field .x/ 2 C n can transform under n-dimensional Lorentz transforma-
tions as
a0 .x/ D Mab .ƒ/ b .ƒ 1
x/ a D 1; : : : ; n (3.2)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 29
LE C T U R E 5 → PS:35–49
where
Š
M.ƒ0 /M.ƒ/ .ƒ 1
ƒ0 1
x/ D M.ƒ0 ƒ/ ..ƒ0 ƒ/ 1
x/ (3.3)
is a n-dimensional representation of the Lorentz group SOC .1; 3/
3| We want a first-order relativistic field equation:
.@ @ C const/ D 0 ) .i @ C const/ D 0 (3.4)
The i anticipates wave-like solutions for real .
4| Then (combine 1 & 2)
i| ^ Coordinate transformation: x 0 D ƒx & Field transformation: 0 .x 0 / D
M.ƒ/ .x/
ii | ^ with .i @ C const/ .x/ D 0 for all x
iii | When is 0 .x/ D M.ƒ/ .ƒ 1 x/ is a new solution?
Š
.i @ C const/ 0 .x/ D Œi .ƒ 1
/ @ C const M.ƒ/ .ƒ 1
x/ D 0 (3.5)
Multiply with M 1 .ƒ/:
1 Š
, Œi M .ƒ/ M.ƒ/.ƒ 1
/ @ C const .ƒ 1
x/ D 0 (3.6)
„ ƒ‚ …
Š
D
! must be n n-matrices with
1
M .ƒ/ M.ƒ/ D ƒ
(3.7)
The -matrices “translate” the “spinor”-representation M.ƒ/ into the “vector”-
representation ƒ and vice versa.
5| How to find and M.ƒ/? SOC .1; 3/ is a Lie group:
i ˛ˇ ! 1 i
ƒ D exp !˛ˇ J 1 !˛ˇ J ˛ˇ (3.8a)
2 2
i ˛ˇ ! 1 i
M.ƒ/ D exp !˛ˇ S 1 !˛ˇ S ˛ˇ (3.8b)
2 2
!˛ˇ antisymmetric tensor ! 3 rotations (angles) + 3 boosts (rapidities)
˛ ıˇ
It is .J ˛ˇ / D i.ı
ˇ
ı˛ ı /.
The 4 4 matrices J ˛ˇ generate the vector-representation ƒ, . 12 ; 12 /, the n n-matrices
S ˛ˇ the spinor-representation M.ƒ/, . 12 ; 0/ ˚ .0; 12 /. The generators are antisymmetric
in the spacetime indices.
• Infinitesimal form of Eq. (3.7):
h i
; S ˛ˇ $ .J ˛ˇ /
$ i.g ˛ ˇ
g ˇ ˛
/ (3.9)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 30
LE C T U R E 5 → PS:35–49
• J ˛ˇ ! Lie-algebra of Lorentz group (J D S; J)
J ;J $ i.g J g J g J C g J / (3.10)
The Lie algebra defines the structure of the Lie group by integration and is therefore
the same for all representations.
6| Solution: Dirac’s trick: ^ such that
f ; g D 2g 1nn ⁂ Dirac algebra (3.11)
This is the Clifford algebra C`1;3 .C/.
Then
i
S
; (3.12)
4
satisfies the Lorentz algebra Eq. (3.10) and Eq. (3.9)
7| Representations:
• At least 4-dimensional
(think of the as Majorana modes and construct ladder operators ! 2 modes)
• All 4-dimensional representations are unitarily equivalent
(actually, they constitute the unique irrep of the Dirac algebra which is 4-
dimensional)
• We use the Weyl representation (sometimes called chiral representation):
i
0 0 1 i 0
D and D (3.13)
1 0 i 0
• Henceforth: ƒ 1 M.ƒ/
2
Two “copies” of a spin- 12 projective representation.
8| Setting const D m, we find:
.i @ m/‰ D 0 ⁂ Dirac equation (3.14)
‰.x/ is a bispinor-field with values in C 4 D C 2 ˚ C 2 .
9| The components of the Dirac spinor field satisfy the KG equation:
0D. i @ m/.i @ m/‰ $ .@2 C m2 /‰ (3.15)
The Dirac differential operator is the “square root” of the Klein-Gordon differential
operator.
10 | Dirac adjoint:
Goal: Lagrangian (which must be a Lorentz scalar).
! How to form Lorentz scalars from spinors?
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 31
LE C T U R E 5 → PS:35–49
i| First try: ‰ ‰
‰ 0 ‰ 0 D ‰ ƒ 1 ƒ 1 ‰ ¤ ‰ ‰ (3.16)
2 2
„ ƒ‚ …
¤1
ƒ 1 is not unitary because S is not hermitian for boosts ( D 0 and D 1; 2; 3).
2
This is a consequence of the non-compactness of the Lorentz group due to boosts!
ii | Define
‰ D ‰ 0
⁂ Dirac adjoint (3.17)
ı 0
! ‰ ‰ 0 D ‰ƒ 1 1 ƒ 1 ‰ D ‰‰ )Lorentz scalar
2 2
Use Eq. (3.12) and Eq. (3.8b) and the Dirac algebra to show this!
11 | Lagrangian:
LDirac D ‰.i @ m/‰ (3.18)
ı
! Euler-Lagrange equations yield Dirac equation.
→ Note 3.1
0
• Let .1; E /T and .1; E /T and D
0
! Dirac equation:
m i @ L
D0 (3.19)
i @ m R
• L and R are called left- and right-handed ⁂ Weyl spinors
• They do not mix under Lorentz transformations
They form the . 12 ; 0/ and .0; 12 / projective irreps of the Lorentz group. Note that
the reducibility of the . 12 ; 0/ ˚ .0; 12 / bispinor representation is manifest in the Weyl
basis:
i i
0i i h 0 ii 0
S D ; D (Boosts, antihermitian) (3.20a)
4 2 0 i
1 ij k k 0
ij i h i ji
S D ; D " (Rotations, hermitian) (3.20b)
4 2 0 k
• For m D 0, the Dirac equation decouples into the ⁂ Weyl equations:
i @ L D0 and i @ R D0 (3.21)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 32
LE C T U R E 5 → PS:35–49
Solutions R and L are eigenstates of the helicity operator h D pO 2E with h D C 12
called right-handed and h D 12 left-handed. Here, pO D p=EE is the normalized
3-momentum for a massless particle.
3.2 Free-Particle Solutions of the Dirac Equation
Here we consider the Dirac equation as a wave equation for a single particle, or, equivalently,
the classical field equation of a complex bispinor field; what follows is therefore “first-
quantized” quantum mechanics. We do this because we need the eigenfunctions of the Dirac
differential operator to construct the field operators when we quantize the Dirac field (“second
quantization”).
Detailed calculations: → Problemset 3.
1| Eq. (3.14) ).@2 C m2 /‰ D 0 (Klein-Gordon equation), therefore
‰ ˙ .x/ D ˙
.p/e ipx with p 2 D m2 and p 0 > 0 (3.22)
Here ˙ .p/ 2 C 4 is a complex-valued four-component bispinor.
We set p 0 > 0 for both positive (C) and negative ( ) frequency solutions and change
the sign of p in the exponent (to simplify the discussion below).
2| Eq. (3.22) in Eq. (3.14) yields
˙
˙ m ˙p L
.˙ p m/ .p/ D ˙ D0 (3.23)
˙p m R
3| Note (→ Problemset 3):
• .p/.p / D p 2 D m2
• Eigenvalues of p and p: p 0 ˙ jpj
E ! for p 0 > 0 and m > 0 positive spectrum
p
In particular, p and p are invertible and the positive squareroots p and
p are Hermitian.
p
p
˙
4| ^ L D p ˙ with arbitrary, normalized [. ˙ / ˙ D 1] spinor ˙ 2 C 2 :
p
Eq. (3.23) ) m p ˙ ˙ p ˙
R D0 (3.24)
p p
p pDm m
˙
D ˙ p ˙ D ˙ p ˙
p
, R (3.25)
p
The second equation in Eq. (3.23) yields the same solution.
5| Solutions:
Conventional notation: C 7! , 7! and C 7! u, 7! v
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 33
LE C T U R E 5 → PS:35–49
1 0
Basis states: s with 1 D and D
2 (same for s )
0 1
p s
C p ipx
‰ .x/ D p e (positive frequency solutions) (3.26)
p s
„ ƒ‚ …
us .p/
p
p s
‰ .x/ D e Cipx (negative frequency solutions) (3.27)
p s
p
„ ƒ‚ …
v s .p/
with p 2 D m2 , p 0 > 0 and s D 1; 2
! Four linearly independent solutions for each 3-momentum pE (˙ and s D 1; 2).
6| Some relations (→ Problemset 3):
• Orthonormality:
Let us .us / 0 and v s .v s / 0, then
ur us D 2m ı rs and .ur / us D 2EpE ı rs
vr vs D 2m ı rs and .v r / v s D 2EpE ı rs
(3.28)
v r us D ur v s D 0
ur .p/v
E s . p/
E D v r . p/u
E s .p/
E D0
Note that uu is Lorentz invariant whereas u u / EpE is not!
Note that .ur / v s ¤ 0 and .v r / us ¤ 0!
For massless particles, the normalization condition is given by .ur / us D 2EpE ı rs .
• Spin sums:
Let p
= p (⁂ Feynman slash notation), then
X
us .p/us .p/ D p = C m1
s
X (3.29)
v s .p/v s .p/ D p
= m1
s
Useful if one wants to sum over spin-polarizations of fermions (→ later).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 34
LE C T U R E 6 → PS:49–62
→ Topics of Lecture 6
1. Dirac field bilinears
2. Fermi statistics and the quantization of the Dirac field
3.3 Dirac Field Bilinears
1| Definition:
5 0 1 2 3 i Weyl basis 1 0
i D " D (3.30)
4Š 0 1
with
5 5 5 2 5
˚
. / D ; . / D 1; ; D0 (3.31)
The last relation implies Π5 ; S D 0, i.e., the Dirac bispinor representation must be
reducible according to Schur’s lemma: . 12 ; 0/ ˚ .0; 12 /
2| The following bilinears ‰ ‰ transform under the Lorentz group as …
D 1 scalar 1
vector 4
2i Π; Di Πtensor 6 (3.32)
5 pseudo-vector 4
5 pseudo-scalar 1
The notation Π: : : denotes the completely antisymmetrized product.
Any 44 matrix can be decomposed into these 16 matrices with definite transformation
properties under Lorentz transformations.
The prefix pseudo- marks quantities that transform under continuous Lorentz transforma-
tions ƒ 2 SOC .1; 3/ as usual but pick up an additional sign under parity transformations.
This is similar to the cross product a b in three dimensions which produces a pseudo-
vector from the two vectors a and b with respect to the Euclidean group (= isometries of
Euclidean space). E.g., angular momentum L E D rE pE is not a vector but a pseudo-vector.
For example,
0
.j /0 D ‰ ‰ 0 D ‰ƒ 1 1
ƒ 1 ‰ D ƒ ‰
‰ D ƒ j (3.33)
2 2
transforms as a Lorentz 4-vector.
ı
! j is the conserved Noether current corresponding to the continuous symmetry
‰ ! e i˛ ‰ of the Dirac Lagrangian.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 35
LE C T U R E 6 → PS:49–62
3.4 Quantization of the Dirac Field
1| Lagrangian: L D ‰.i @
m/‰
2| Canonical momentum: …a D @L Pa
D i‰a
@‰
3 | Hamiltonian: H D d 3x ‰ Œ i ˛r
E C mˇ ‰ with ˛E D 0E and ˇ D 0
R
„ ƒ‚ …
DHD
HD is the Dirac Hamiltonian of single-particle quantum mechanics.
! Expand ‰ in eigenmodes of HD to diagonalize H
4| Eigenmodes: HD us .p/e
E i pExE D EpE and HD v s .p/e
E i pExE D EpE
This can be seen from Œi 0 @0 C i E r m‰ D 0 (← last lecture).
5| Mode expansion:
XZ d 3p 1 h i
E D
‰.x/ apsE us .p/e
E i pExE C bpsE v s .p/e
E i pEx
E
(3.34)
.2/3 2EpE
p
s
apsE and bpsE are operator-valued expansion coefficients. We do not yet fix their algebra!
6| Use
s
XZ d 3p EpE h s s i
E D
HD ‰.x/ E i pExE
apE u .p/e bpsE v s .p/e
E i pEx
E
(3.35)
s
.2/3 2
then (using the orthonormality relations Eq. (3.28))
Z X Z d 3p s s
3
H D d x ‰ HD ‰ $ 3
EpE .apE apsE bpE bpsE / (3.36)
s
.2/
You do not need reordering of operators to show this. The algebra is still undefined!
First try: Commutator
7| Canonical quantization with equal-time commutators:
h i
E D i ıab ı .3/ xE yE E D ıab ı .3/ xE
E …b .y/
‰a .x/; , E ‰b .y/
‰a .x/; yE
E ‰b .y/
‰a .x/; E D0
(3.37)
ı
8| ! Mode algebra
h i h i
s s
aprE ; aqE D bprE ; bqE D .2/3 ı rs ı .3/ pE
qE (3.38)
h i
s./
aprE ; bqE D0 (3.39)
Show (using the mode expansion Eq. (3.34) and the spin sums Eq. (3.29)) that this is
equivalent to the commutators of the fields.
Beware: Eq. (3.89) of P&S is mathematically ill-defined since ‰‰ is a matrix but ‰ ‰
is not (it’s just sloppy math that doesn’t belong in a textbook for students). Do it right,
i.e., componentwise: Œ‰a .x/; E D ıab ı .3/ xE yE .
E ‰b .y/
! Irreducible Representation = Bosonic Fock space
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 36
LE C T U R E 6 → PS:49–62
n n!1
s
9| Problem: bpE j0i has energy nEpE ! 1
! No stable vacuum state (the spectrum of H is unbounded below)
10 | Fix (?): b $ b (Use colors to modify the previous derivation.)
s
i| E D : : : ŒapsE C bpE : : :
‰.x/
s s
ii | H D : : : .apE apsE bpsE bpE /
s
ŒbprE ; bqE D .2/3 ı rs ı .3/ pE
iii | qE
s s
iv | H D : : : .apE apsE bpE bpsE / C const
s s s
v| ŒH; bpE D EpE bpE ! bpE creates a particle with positive energy! ! H 0
It seems that we solved the problem: The spectrum of the Hamiltonian is now
bounded from below.
vi | But:
h i
s s
kbpE j0ik2 D h0j bpsE ; bpE j0i D .2/3 ı .3/ .0/ < 0 (3.40)
! Negative norm states (i.e., the constructed representation is not a Hilbert space)
11 | Conclusion: Eq. (3.37) implies
• either an instability of the vacuum
• or a loss of unitarity
! No consistent quantization possible!
Second try: Anticommutator
7| Canonical quantization with equal-time anticommutators:
n o
E D ıab ı .3/ xE and
˚
E ‰b .y/
‰a .x/; yE E ‰b .y/
‰a .x/; E D0 (3.41)
Note that these are equal-time anticommutators!
ı
8| ! Mode algebra
n o n o n o
s s s./
aprE ; aqE D bprE ; bqE D .2/3 ı rs ı .3/ pE and aprE ; bqE
qE D0
(3.42)
The proof is similar to the bosonic case above.
! Irreducible Representation = Fermionic Fock space
s
9| Problem: bpE j0i has energy EpE & infinite sum over momenta
! Still no stable vacuum state
(The spectrum of H is still unbounded below due to the sum over momenta.)
10 | Fix (?): b $ b (we saw above that it changes the sign of the excitation energies)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 37
LE C T U R E 6 → PS:49–62
i| Hamiltonian:
XZ d 3p s s
H D 3
EpE .apE apsE bpsE bpE / (3.43)
s
.2/
XZ d 3p s s
D E .a as C bpE bpsE / 1 (3.44)
s
.2/3 pE pE pE
We will drop the infinite constant henceforth. (Cross the 1.)
ii | The mode algebra Eq. (3.42) is invariant under b $ b !
! Unitarity is preserved and Hamiltonian is bounded from below
! With anticommutation relations, quantization is consistently possible!
11 | Heisenberg picture:
Now that we have a representation where the Hamiltonian generates a unitary time
evolution, we can switch to the Heisenberg picture:
With
e iH t apsE e iH t
$ apsE e iEpE t
and e iH t bpsE e iH t
$ bpsE e iEpE t
(3.45)
and ‰.x/ D e iH t ‰.x/e
E iH t we find
XZd 3p 1 h
s s ipx s s ipx
i
‰.x/ D a pE
u .p/e C b v .p/e (3.46)
.2/3 2EpE pE
p
s
d 3p 1
Z h i
s s
X
ipx s s ipx
‰.x/ D a u .p/e C b pE
v .p/e (3.47)
.2/3 2EpE pE
p
s
These are operator-valued spinor fields, i.e., functions (more precisely: distributions) on
Minkowski spacetime that assign to an event x a tuple (“spinor”) of operators that act
on the fermionic Fock space where the states of the quantized theory live.
Continuous symmetries & Conserved charges
• Time translation ! Hamiltonian (see above)
• Spatial translations ! Momentum operator
d 3p
Z XZ
3 s s
PE $ d x ‰ . i r/‰ $ 3
pE .apE apsE C bpE bpsE / (3.48)
s
.2/
• Rotations ! Angular momentum operator JE
1
Z
3
JE D d x ‰ xE . i r/ C † E ‰ with E D E 0
† (3.49)
2 0 E
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 38
LE C T U R E 6 → PS:49–62
• Global phase rotations e i˛ ‰
ı
! Conserved current j D ‰ ‰
! Conserved charge:
d 3p
Z XZ
3 s s
QD d x‰ ‰ $ .a as C b s pE b pE / (3.50)
s
.2/3 pE pE
XZ d 3p s s
D .a as bpE bpsE / C 1 (3.51)
s
.2/3 pE pE
In QED we will couple the fermions to the EM field; then, Q is the total EM charge of
the fermion field.
Recall that in single-particle quantum mechanics the global phase rotation symmetry gives
rise to a positive density and a current that can be interpreted as probability current;
the conserved charge corresponds then to the total probability to find the single particle
somewhere. This interpretation clearly does no longer apply as Q can become negative.
The operators of conserved charges generate symmetry transformations of the Hamiltonian.
Excitations = Particles
s
apE j0i W Fermion with energy EpE ,
momentum p, E
1
spin J D 2 (polarization s),
and charge Q D C1
(3.52)
s
bpE j0i W Antifermion with energy EpE ,
momentum p, E
1
spin J D 2 (polarization opposite to s),
and charge Q D 1
In QED, the fermions will be electrons and the antifermions positrons.
→ Note 3.2
• The two states for s D 1; 2 suggest a spin- 12 representation
• To show this, the action of JE (see Eq. (3.49)) on one-particle states must be studied
• One finds for particles at rest:
s 1 s s 1 s
J´ a E j0i D ˙ a E j0i and J´ b E j0i D b E j0i (3.53)
0 2 0 0 2 0
1 0
with sD1 D and sD2 D .
0 1
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 39
LE C T U R E 6 → PS:49–62
Lorentz transformations
s
1| ^ Lorentz transformation ƒ 2 SOC .1; 3/ on single particle state jp;
p
E si 2EpE apE j0i:
jp;
E si 7! U.ƒ/jp;
E si (3.54)
U.ƒ/: representation of SOC .1; 3/ on Fock space
For generic rotations/boosts, this mixes the two spin components!
2| ^ Special case: quantization axis parallel to boost and/or rotation axis
! Spin polarizations do not mix:
s
s 1
EƒpE s
U.ƒ/ apE U .ƒ/ D a (3.55)
EpE ƒpE
Note that spins mix under generic Lorentz transformations: ap1E $ aq2E .
3| Consider this special case, then:
q ; ri D 2EpE .2/3 ı .3/ pE qE ı rs D hp;
E sjU .ƒ/U.ƒ/jE
hp;
E sjE q ; ri (3.56)
„ ƒ‚ …
Lorentz invariant
! U.ƒ/ is unitary
4| Now we have 3 representations:
ƒ acts on 4-vectors in R1;3 D=4 not unitary
ƒ 1 acts on bispinors in C 2 ˚ C 2 D=4 not unitary (3.57)
2
U.ƒ/ acts on states in fermionic Fock space D=1 unitary
ı
5| Action by conjugation on field operators !
1
U.ƒ/‰.x/U .ƒ/ D ƒ 1 1 ‰.ƒx/ (3.58)
2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 40
LE C T U R E 7 → PS:62–71
→ Topics of Lecture 7
1. The spin-statistics theorem
2. The Dirac propagator
3. Causality
4. Discrete symmetries of the Dirac theory
→ Topics of Problemset 4
1. The relativistic hydrogen atom
2. Parity transformation of Dirac spinors
Spin-statistics theorem
• Observation:
Klein-Gordon field : Spin 0 (scalar) ! commutator ! bosonic excitations
1
Dirac field ‰: Spin 2 (spinor) ! anticommutator ! fermionic excitations
(3.59)
This is no coincidence but hints at a more fundamental connection:
• Spin-statistics theorem:
Lorentz invariance
9
>
Causality Integer spin $ Bosons
>
=
) (3.60)
Positive energies >
> Half-integer spin $ Fermions
Positive norms
;
This means, whenever you quantize a relativistic field that transforms under a (pro-
jective) half-integer spin representation, the Poisson bracket must be replaced by
anticommutators. Otherwise unitarity is lost or the vacuum becomes unstable.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 41
LE C T U R E 7 → PS:62–71
• “Proof by picture”:
Rigorous proofs are elaborate and quite technical.
↑ http://math.ucr.edu/home/baez/spin_stat.html
Dirac Propagator
All that follows is very similar to our discussion of the Klein-Gordon propagator.
For details, we refer the student to the corresponding notes.
1| Propagation amplitudes (use colors to skip this calculation):
d 3p 1
Z X
ip.x y/
h0j‰a .x/‰ b .y/j0i D e usa .p/usb .p/ (3.61)
.2/3 2EpE s
„ ƒ‚ …
.pCm/
= ab
D .i @=x C m/ab D.x y/ (3.62)
d 4p i.p= C m/ab
Z
x 0 >y 0 ip.x y/
D e (3.63)
4
.2/ p 2 m2 C i "
d 3p 1
Z X
ip.y x/
h0j‰ b .y/‰a .x/j0i D e vas .p/v sb .p/ (3.64)
.2/3 2EpE s
„ ƒ‚ …
.p
= m/ab
D .i @=x C m/ab D.y x/ (3.65)
d 4p i.p= C m/ab
Z
x 0 <y 0 ip.x y/
D e (3.66)
4
.2/ p 2 m2 C i "
Remember:
©
O5
N N >
7 ”
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 42
LE C T U R E 7 → PS:62–71
• x 0 > y 0 : close contour below
• x 0 < y 0 : close contour above
2| Therefore we define the Feynman propagator of the Dirac field:
d 4p i.p= C m/ab
Z
SFab .x y/ D 4 2
e ip.x y/
(3.67)
.2/ p m2 C i "
(
h0j‰a .x/‰ b .y/j0i for x 0 > y 0
D (3.68)
h0j‰ b .y/‰a .x/j0i for x 0 < y 0
h0jT ‰a .x/‰ b .y/j0i (3.69)
Note: For t1 > t2 it is T ‰.t2 /‰.t1 / ‰.t1 /‰.t2 / for fermionic fields!
The Feynman propagator SF .x y/ of the Dirac field is a 4 4 matrix.
3| Similarly, one can derive the Retarded Green’s function:
ab
.x y/ .x 0 y 0 / h0j ‰a .x/; ‰ b .y/ j0i $ .i @=x C m/ab DR .x
˚
SR y/ (3.70)
Here, DR .x y/ is the retarded Green’s function of the Klein-Gordon field;
@=x denotes derivatives with respect to the variables x for D 0; 1; 2; 3 and generates
the p= in the integral.
Causality
P Qeven N ./ ./ ./
1| O
Measurable operators: O.x/ D .‰i .x/ _ @‰i .x/ _ @2 ‰i .x/ : : : /
i D1
Example: j D ‰ ‰ (check that this is Hermitian!) (but not ‰a C ‰a !)
Restricting observables to field polynomials of even degree ensures that space-like
separated observables commute if space-like separated fields anticommute (which is the
best we can hope for given our quantization conditions).
The answer to the question “Why restrict observables to even degree expressions in the fields?”
is therefore: Because these are the only observables that do not violate causality in a
theory built from fermionic fields. (There are also more rigorous arguments for this:
↑ Superselection.)
2| Causality for fermionic fields , f‰a .x/; ‰ b .y/g D 0 for .x y/2 < 0
All other anticommutators vanish trivially. Note that here x D .t; x/
E and y D .t 0 ; y/,
E
i.e., we consider the anticommutator at different times.
We find (using results from above)
˚
‰a .x/; ‰ b .y/ $ .i @=x C m/ab ŒD.x y/ D.y x/ (3.71)
2
.x y/ < 0
D .i @=x C m/ab ŒD.x y/ D.x y/ D 0 (3.72)
The argument is the same as for the Klein-Gordon field.
R d 3p 1
Recall: D.x y/ D .2/ 3 2E e
ip.x y/
E
p
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 43
LE C T U R E 7 → PS:62–71
3.5 Discrete Symmetries of the Dirac Theory
Review of the Lorentz group
• Lorentz group O.1; 3/ = Lie group with four disconnected components
• Continuous Lorentz transformations = Proper orthochronous Lorentz group SOC .1; 3/
• Four components connected by discrete transformations:
Parity P W .t; x/
E 7! .t; x/
E (3.73)
Time reversal T W .t; x/
E 7! . t; x/
E (3.74)
Si
cnouovS ) qu 1% 17
.
vo Kue
PVPpr
Lo. ıt ‚= me
LG) N)
Sr) amt sur
+: det +1 (proper )
> - 2 detA
T
L(n Le \
Vo proper 20
53€ U \
?
er
Parity
Details: → Problemset 4
1| Unitary representation on Fock space:
U.P / apsE U 1
.P / D a as pE and U.P / bpsE U 1
.P / D b b s pE (3.75)
„ƒ‚… „ƒ‚…
C1 1
E D rE pE
Note that we do not want spin to change under P because angular momentum L
also does not pick up a sign under inversion (it is a pseudo-vector).
Note that often U.P / is simply written P .
2| Equivalent to
1 0
E
U.P /‰.t; x/U .P / D ‰. t; xE / (3.76)
„ƒ‚… „ƒ‚…
P1 Px
2
The 0 -matrix exchanges the left- and right-handed Weyl sectors of the bispinor; this
makes sense as a parity transformation of space should switch chirality.
3| Dirac field bilinears (examples):
1
U.P /‰‰U .P / $ C‰‰.t; x/
E ! scalar (3.77)
5 1 5
U.P /‰ ‰U .P / $ ‰ E
‰.t; x/ ! pseudo-scalar (3.78)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 44
LE C T U R E 7 → PS:62–71
Time Reversal
1| Time reversal should …
• U.T /‰.t; x/U
E 1 .T / D T 1 ‰. t; x/,
E
2
• U.T /apsE U 1 .T / Da ‹
pE
,
E D rE pE 7!
• flip spins (motivated by L E
L),
• be a symmetry of the Dirac theory: ŒU.T /; H D 0,
• obey U 1 .T / D U .T /.
This is required for any symmetry to preserve overlaps: ↑ Wigner’s theorem.
Note that often U.T / is simply written T .
2| Problem:
E D e iH t ‰.x/e
‰.t; x/ E iH t
(3.79)
1
) E
U.T /‰.t; x/U .T / D e iH t U.T /‰.x/U
E 1
.T /e iH t
(3.80)
iH t
) E
T 1 ‰. t; x/j0i De E
T 1 ‰.x/j0i (3.81)
2 2
iH t
) T1 e E
‰.x/j0i D e iH t T 1 ‰.x/j0i
E (3.82)
2 2
2iH t
) e„ ƒ‚ E
T 1 ‰.x/j0i D T 1 ‰.x/j0i
E (3.83)
… 2 2
time-dependent!
Here we used that ŒU.T /; H D 0 and H j0i D 0.
! Not possible (for invertible T 1 and arbitrary times t)!
2
3| Solution: U.T / must be antiunitary/antilinear:
U.T /c D c U.T / for c 2 C (3.84)
Highlight the differences with colors in the derivation above:
E D e iH t ‰.x/e
‰.t; x/ E iH t
(3.85)
1 iH t 1 iH t
) E
U.T /‰.t; x/U .T / D e E
U.T /‰.x/U .T /e (3.86)
iH t
) E
T 1 ‰. t; x/j0i De E
T 1 ‰.x/j0i (3.87)
2 2
iH t iH t
) T1 e E
‰.x/j0i De E
T 1 ‰.x/j0i (3.88)
2 2
) 1 E
T 1 ‰.x/j0i D T 1 ‰.x/j0i
E (3.89)
„ƒ‚… 2 2
time-independent!
4| Transformation of spin:
i| Spinors: ^ Spin basis s (s D 1; 2) along arbitrary axis nE :
! !
i sin
cos e
1 D i 2 and 2 D 2 (3.90)
e sin 2 cos 2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 45
LE C T U R E 7 → PS:62–71
That is, 1 D j"i and 2 D j#i.
“Time-reversed” (=flipped) spinors:
( )
1 2
2 s
s i . / ) D (3.91)
2 1
Indeed, if nE E D C, we have
nE E . i 2 / D i 2 . nE E / D i 2 . / D . i 2 / (3.92)
where we used E 2 D 2 . E /.
Note that T D i 2 K (where K denotes complex conjugation) is the conventional
representation of time-reversal symmetry for spinful fermions that you might know
from condensed matter physics (e.g., to classify symmetry-protected topological
phases).
ii | Bispinors:
p p
p s s
s s p
u .p/ and v .p/
p s
p p
p s
p p ! (3.93)
p s s
p
s
u .p/ p and v s .p/
p s
p
p s
Use colors to skip the second row.
Note that here us is not the Dirac adjoint us !
Recall that the basis s in the definition of v s .p/ was arbitrary.
iii | Define the modes:
8 9 8 9 (
<a 1 = ap2E <b 1 = 2
( ) )
pE
b
and pE
pE
(3.94)
:a 2 ; ap1E :b 2 ; bp1E
pE pE
Skip the second part.
Note that this is analog to Eq. (3.91)!
iv | Let pQ .p 0 ; p/
E and show
1 3
us .p/
Q $ Œus .p/
(3.95)
1 3
v s .p/
Q $ Œv s .p/
Note: s D s used
p in v 2
s
p
Use Eq. (3.93) and pQ D 2 p to show this!
5| Definition:
Antilinearity Eq. (3.84) > 1
9
> E
U.T /‰.t; x/U .T /
=
s 1 s 1 3
U.T / apE U .T / a pE ) $. E
/ ‰. t; x/ (3.96)
> „ ƒ‚ …
U.T / b s U 1 .T / b s ; T1
>
pE pE 2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 46
LE C T U R E 7 → PS:62–71
Use Eq. (3.95) and Eq. (3.94) and ap2E u2 .p/ D ap1E u1 .p/ etc. to show this!
Note that in Weyl representation
iy
0
T1 D (3.97)
2 0 iy
i.e., time-reversal acts on spins but does not mix chiralities (as parity did). This makes
sense, because (for massless particles) chirality = helicity and helicity is the projection of
spin on momentum: SE p. E Since both spin (angular momentum) SE and linear momentum
pE change sign under time-reversal, helicity does not.
6| Dirac field bilinears (example: j D ‰ ‰):
(
Cj . t; x/
E for D 0
U.T /j .t; x/U
E 1
.T / $ (3.98)
j . t; x/E for D 1; 2; 3
! As expected for density ( D 0) and 3-current ( D 1; 2; 3)
Charge Conjugation
1| Discrete, non-spacetime symmetry that exchanges particle and antiparticle:
U.C /apsE U 1
.C / D bpsE and U.C /bpsE U 1
.C / D apsE (3.99)
Note that there is no representation on Minkowski space as this is an “internal”
symmetry.
Often U.C / is simply written C .
2| Use Eq. (3.93) to show:
us .p/ $ i 2
.v s .p// and v s .p/ $ i 2
.us .p// (3.100)
3| Then
U.C /‰.x/U 1 .C / (3.101)
X Z d 3p 1 h s
i
2
D 3
p i .v s .p// bpsE e ipx
i 2
.us .p// apE e ipx (3.102)
s
.2/ 2E p
E
2
$ i .‰ /T D i.‰ 0 2 T
/ (3.103)
4| Therefore:
1 0 2 T
U.C /‰U .C / D i.‰ / .D i 2 ‰/ (3.104)
„ƒ‚…
C 1
2
1 0 2 T
and U.C /‰U .C / $ i. ‰/ (3.105)
• Note that C essentially exchanges ‰ $ ‰ but is not antiunitary!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 47
LE C T U R E 7 → PS:62–71
• To show this, recall that 0 and 2 are symmetric matrices.
.
• It is C 1 D C 1 , C 12 D 1 and C 1 C
1 D
2 2 2 2 2
• Note that the expression in parantheses is only true for the transformation of
classical (i.e. “first quantized”) Dirac fields and can be used to show the symmetry
of the classical Dirac equation. However, if you take the to conjugate complex
numbers and Hilbert space operators, ‰ .‰ /T , it is valid for the quantized
field as well.
5| Dirac field bilinears (examples):
1
U.C /‰‰U .C / $ ‰‰ (Scalar) (3.106)
1
U.C /‰ ‰U .C / $ ‰ ‰ (Vector) (3.107)
→ Note 3.3
"
• Any relativistic QFT must be invariant under SOC .1; 3/ (D LC )
• The (classical) Dirac equation .i @
m/‰ D 0 is fC; P; T g-invariant
• The (quantized) Dirac theory is fC; P; T g-invariant:
ŒH; U.X / D 0 for X D P; T; C
• Weak interactions (of the standard model) violate C and P but preserve CP and T
(↑ Wu experiment)
• Rare processes (decay of neutral kaons) violate CP and T but preserve CP T
• CP T seems to be a perfect symmetry of nature
• CP T theorem:
SOC .1; 3/ invariance
9
>
Causality
>
=
) CP T symmetry (3.108)
Locality >
>
Stable vacuum
;
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 48
LE C T U R E 8 → PS:77–90
→ Topics of Lecture 8
1. Interacting fields
2. Perturbation expansion of correlation functions
3. Wick’s theorem
4 Interacting Fields and Feynman Diagrams
4.1 Preliminaries
• Up to now: No interactions, no scattering, Fourier modes are eigenmodes
• Now: Include non-linear terms in the Hamiltonian/Lagrangian that couple Fourier
modes
• Causality ! Interactions = Products of fields at same spacetime point
• In the following:
Z Z
3
Hint D d x Hint . .x// D d 3x Lint . .x// (4.1)
Lint is only a function of ! Hint D Lint .
• Examples:
1. 4 -theory:
1 1 2 2 4
L 4 D .@ /2 m (4.2)
2 2 4Š
: dimensionless coupling constant
Why do we choose 4 and not 3 ? Energy unbounded from below for 3 !
The 4 -interaction arises in the standard model (Higgs field) and also in statistical
mechanics.
! Equation of motion is no longer linear:
3
.@2 C m2 / D (4.3)
3Š
! Cannot be solved by Fourier modes!
2. Yukawa theory:
1 1 2 2
LYukawa D ‰.i @= m/‰ C .@ /2 m g‰‰ (4.4)
„ ƒ‚ … „ 2 ƒ‚ 2 „ ƒ‚ …
Dirac Interaction
…
Klein-Gordon
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 49
LE C T U R E 8 → PS:77–90
g: dimensionless coupling constant
Yuakawa theory = QED for a scalar field instead of a vector field A .
In the standard model, Yukawa couplings describe the coupling of the Higgs field
to quarks and leptons (→ later).
3. QED (Quantum Electrodynamics):
1
LQED D ‰.i @= m/‰ .F /2 e‰ ‰A (4.5)
„ ƒ‚ … 4 „ ƒ‚ …
Dirac Interaction
„ ƒ‚ …
Maxwell
1
=
D ‰.i D m/‰ .F /2 (4.6)
4
e D jej < 0: Electron charge
D @ C i eA .x/: Gauge covariant derivative
1
QED has a U.1/ gauge symmetry: A0 D A e @ ˛.x/ and ‰ 0 .x/ D e i˛.x/ ‰.x/.
! Equations of motion:
=
.i D m/‰.x/ D 0 and @ F D ej .j D ‰
‰/ (4.7)
Quantizing the EM field is subtle due to gauge invariance. We will demonstrate one
possibility at the end of this course using path integrals.
→ Note 4.1: Minimal coupling
The coupling via @ 7! D in Eq. (4.6) is called minimal coupling. It is minimal
in the sense that only the zeroth moment of the charge distribution described
by ‰ is coupled to the electromagnetic field (remember that e‰ ‰ is the
charge density current). It can also be derived as necessary addition to the
Dirac Lagrangian to feature a local U.1/ gauge symmetry (→ Section 9.1); this
is another sense in which it is minimal.
A term beyond minimal coupling that one could add is Pauli coupling
e
LPauli D ‰ ‰ F (4.8)
8m
which couples the electromagnetic field directly to the first moment (essentially
the spin current) and has therefore direct influence on the g-factor of the
electron. Here, is a free coupling constant and D i=2 Π; has
been defined in Eq. (3.32). Deriving the Dirac equation from the minimally
coupled Dirac Lagrangian extended by Eq. (4.8) yields for the electron g-factor
g D 2 C , i.e., the g-factor can be freely tuned by the coupling constant [6].
The fascinating thing about nature is that only minimal coupling is needed
to describe our observations (this is true for all forces described by the
Standard Model); couplings to higher moments emerge automatically and
can be computed ab initio; → Eq. (6.29). (We will do this in Sections 6.3.2
and 6.3.3 where we start from minimal coupling and compute corrections to
the g-factor that arise perturbatively.) By the way, the term Eq. (4.8) would
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 50
LE C T U R E 8 → PS:77–90
render our theory non-renormalizable and therefore “useless” (→ Chapter 7
and ↑ p. 517 of Ref. [1]).
• The list of possible interaction terms is finite due to constraints like gauge invariance and
renormalizability (which we will discuss in the second half of the semester).
• The standard model includes all of the allowed interactions. The three examples above
cover nearly half of them!
• No known exactly solvable interacting QFTs in D > 1 C 1!
Examples of exactly solvable interacting QFTs in D D 1 C 1 are ↑ conformal field theories
which have an extensive set of symmetry generators.
! Perturbation theory
(we hope/assume that the coupling constants are small enough!)
4.2 Perturbation Expansion of Correlation Functions
Details: → Problemset 5
1| Goal: Two-point Green’s function hjT .x/ .y/ji of 4 -theory
ji: Ground state of interacting theory
j0i: Ground state of free theory (free=non-interacting)
2| Remember: Without interactions, this is the Feynman propagator:
d 4p i e ip.x y/
Z
h0jT .x/ .y/j0i D DF .x y/D (4.9)
.2/4 p 2 m2 C i "
3| Now:
Z
H 4 D H0 C d 3x 4 .x/
E (4.10)
„ƒ‚… 4Š
KG Hamiltonian „ ƒ‚ …
Hint W Interaction = Perturbation
! Expand hjT .x/ .y/ji in powers of
4| Todo:
free field I .x/
.x/
Express in terms of (4.11)
ji free vacuum j0i
Note that both .x/ D e iH t .x/e
E iH t and ji depend on the interaction.
5| ^ Reference time t0 , then
d 3p 1
Z
E D
.t0 ; x/ a e i pExE C apE e i pEx
E
(4.12)
.2/3 2EpE pE
p
This follows, because the equal-time commutation relations are still valid.
The modes apE now implicitly and non-trivially depend on the reference time t0 ! This
dependence only drops out for the free theory where the Fourier modes are stationary
eigenmodes.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 51
LE C T U R E 8 → PS:77–90
6| Definitions:
E e iH.t
.t; x/ t0 /
E
.t0 ; x/e iH.t t0 /
Heisenberg picture (4.13)
iH0 .t t0 / iH0 .t t0 /
E e
I .t; x/ E
.t0 ; x/e Interaction picture (4.14)
Then
d 3p 1
Z
ipx
E $
I .t; x/ a e C apE e ipx (4.15)
.2/3 2EpE pE
p
This is analogous to the free field!
and
E D U .t; t0 / I .t; x/U.t;
.t; x/ E t0 / with U.t; t0 / D e iH0 .t t0 /
e iH.t t0 /
(4.16)
Our goal is to express in terms of I since we know its time evolution!
7| The time-evolution operator is determined by U.t0 ; t0 / D 1 and the differential equation
i @ t U.t; t0 / $ HI .t /U.t; t0 / (4.17)
with
4
Z
iH0 .t t0 / iH0 .t t0 /
HI .t / D e Hint e D d 3x E
.t; x/ (4.18)
4Š I
8| The solution of Eq. (4.17) is given by the Dyson series:
Z t
U.t; t0 / D 1 C . i/ dt1 HI .t1 / (4.19)
t0
. i /2 t
Z
C dt1 dt2 T fHI .t1 /HI .t2 /g C : : :
2Š t0
Z t
T exp i ds HI .s/ (4.20)
t0
The Dyson series yields an expansion for .t; x/E in terms of I .t; x/
E in powers of .
This is the definition of the time-ordered exponential.
9| Properties: (Proofs: → Problemset 5)
iH.t t 0 / iH0 .t 0 t0 /
U.t; t 0 / D e iH0 .t t0 /
e e
1
U .t; t 0 / D U .t; t 0 / (4.21)
U.t1 ; t2 /U.t2 ; t3 / D U.t1 ; t3 /
Here, t t 0 and t1 t2 t3 ; the definition for t 0 ¤ t0 is given by Eq. (4.20).
10 | Ground state ji?
1 ! hj0i ¤ 0 (this is not a rigorous but a reasonable assumption)
X
iH T iEn T
e j0i D e jnihnj0i (4.22)
n
X
iE0 T iEn T
De jihj0i C e jnihnj0i (4.23)
n¤0
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 52
LE C T U R E 8 → PS:77–90
Then (since En > E0 for n ¤ 0)
1
iE0 T iH T
ji D lim e hj0i e j0i (4.24)
T !1.1 i"/
1
iE0 .t0 CT /
$ lim e hj0i U.t0 ; T /j0i (4.25)
T !1.1 i"/
(Details: → Problemset 5)
Similar:
1
iE0 .T t0 /
hj D lim h0jU.T; t0 / e h0ji (4.26)
T !1.1 i"/
11 | Two-point correlator: (let x 0 > y 0 > t0 )
hj .x/ .y/ji D (4.26) (4.16) (4.16) (4.25) (4.27)
(4.21)
D lim NT 1 h0jU.T; x 0 / I .x/U.x 0 ; y 0 / I .y/U.y 0 ; T /j0i
T !1.1 i"/
(4.28)
with (use hji D 1)
(4.26)(4.25) (4.21)
NT D h0jU.T; t0 /U.t0 ; T /j0i D h0jU.T; T /j0i (4.29)
For y 0 > x 0 we can do the same calculation for hj .y/ .x/ji by replacing x $ y.
(Details: → Problemset 5)
This leads to the final result:
x 0 ? y 0 arbitrary !
(4.20)
hjT .x/ .y/ji D (4.30)
n h R io
T
h0jT I .x/ I .y/ exp i T dt HI .t / j0i
lim n h R io (4.31)
T
T !1.1 i"/ h0jT exp i T dt HI .t / j0i
The right-hand side of this expression only includes known entities!
This derivation goes through for arbitrary n-point correlators.
4.3 Wick’s Theorem
Eq. (4.18) and Eq. (4.31) ! (expand the time-ordered exponential in orders of )
X
hjT .x/ .y/ji D : : : h0jT I .x1 / I .x2 / : : : I .xn /j0i : : : (4.32)
„ ƒ‚ …
How to evaluate this efficiently?
Solution: Wick’s theorem!
We could just use the mode expansion of the fields and calculate the n-point correlators the
brute force way. But Wick’s theorem provides a more systematic approach.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 53
LE C T U R E 8 → PS:77–90
1| Define
d 3p 1 d 3p 1
Z Z
ipx
I .x/ D a e C a e Cipx (4.33)
3 2EpE pE 3 2EpE pE
p p
.2/ .2/
„ ƒ‚ … „ ƒ‚ …
IC .x/ I .x/
Useful because IC j0i D 0 and h0j I D 0.
2| Observation for x 0 > y 0 and n D 2:
T I .x/ I .y/ D IC .x/ IC .y/ C IC .x/ I .y/ C I .x/ IC .y/ C I .x/ I .y/ (4.34)
D IC .x/ IC .y/ C I .y/ IC .x/ C I .x/ IC .y/ C I .x/ I .y/
C IC .x/; I .y/
(4.35)
For y 0 > x 0 we find:
T I .x/ I .y/ D IC .y/ IC .x/ C IC .y/ I .x/ C I .y/ IC .x/ C I .y/ I .x/ (4.36)
D IC .y/ IC .x/ C I .x/ IC .y/ C I .y/ IC .x/ C I .y/ I .x/
C IC .y/; I .x/
(4.37)
Use colors to skip this step!
3| This motivates the Definitions:
(We drop the I henceforth as contractions always operate on interaction picture fields).
Contraction: (4.38)
ΠC .x/; .y/ for x0 y0
>
.x/ .y/ $ DF .x y/1 (4.39)
ΠC .y/; .x/ for y > x 0
0
Normal order: (4.40)
./
Wa1 : : : an./ W .creation operators/ .annihilation operators/ (4.41)
Example for normal order: W C .x/ .y/W D .y/ C .x/
Recall that Π.x/; .y/ D ΠC .x/; .y/ C Π.x/; C .y/ D D.x y/ D.y x/.
Like time ordering T , normal ordering WW is a meta operator that acts on symbolic
strings (= descriptions of operators = the free algebra of ai and ai ). In particular,
normal ordering is not well-defined on the CCR algebra: a a D Waa W ¤ Wa a C 1W D
Wa aW C W1W D a a C 1, ↑ https://physics.stackexchange.com/a/368084/45257.
The vacuum expectation value of normal-ordered products vanishes!
!
T .x/ .y/ D W .x/ .y/ C .x/ .y/W (4.42)
) h0jT .x/ .y/j0i D DF .x y/ (4.43)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 54
LE C T U R E 8 → PS:77–90
4| The generalization of this statement is called Wick’s theorem:
T f .x1 / : : : .xn /g D W .x1 / : : : .xn / C all possible contractionsW (4.44)
where WA i B j C W DF .xi xj / WABC W (4.45)
Proof: → Problemset 5
Wick’s theorem is not specific to QFT but a quite generic, combinatorical state-
ment, ↑ https://physics.stackexchange.com/a/24180/45257. For instance, in
probability theory, it is well known that the expectation values of arbitrary products
of Gaussian random variables are completely determined by two-point correlators,
↑ https://en.wikipedia.org/wiki/Isserlis%27_theorem.
5| Corollary:
h0jT f .x1 / : : : .xn /gj0i D all full contractions (4.46)
Wick’s theorem in this form is only valid for expectation values w.r.t. the non-interacting
vacuum j0i of non-interacting fields (recall that we omit here the subscript I , i.e.,
D I ).
6| Example ( i .xi /):
T f 1 2 3 4 g DW 1 2 3 4 C 1 2 3 4 C 1 2 3 4 C 1 2 3 4 (4.47)
C 1 2 3 4 C 1 2 3 4 C 1 2 3 4
C 1 2 3 4 C 1 2 3 4 C 1 2 3 4 W
Therefore
h0jT f 1 2 3 4 gj0i D h0j 1 2 3 4 C 1 2 3 4 C 1 2 3 4 j0i (4.48)
D DF .x1 x2 /DF .x3 x4 / (4.49)
CDF .x1 x3 /DF .x2 x4 /
CDF .x1 x4 /DF .x2 x3 /
1 2 1 2 1 2
D C C (4.50)
3 4 3 4 3 4
We associate each spacetime point xi with a vertex and each propagator connecting
two points with an edge. These are Feynman diagrams, here for the trivial example of
free fields. We interpret edges as particles propagating from one point to another; the
propagation amplitude is then the superposition of all possible ways for two particles to
propagate between four points.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 55
LE C T U R E 9 → PS:90–95
→ Topics of Lecture 9
1. Feynman rules and diagrams of 4 -theory in position space
2. Feynman rules in momentum space
→ Topics of Problemset 5
1. Perturbation expansion of correlation functions
2. Wick’s theorem
4.4 Feynman Diagrams
Details: → Problemset 6
1| ^ Numerator of Eq. (4.31) (on the right-hand side I 7! for simplicity)
Z
hjT .x/ .y/ji / h0jT .x/ .y/ C .x/ .y/ i dt HI .t / C : : : j0i
(4.51)
We focus now on 4 -theory and develop the formalism for this specific theory.
→ Problemset 6 for an analogous treatment of the complex Klein-Gordon field.
2| 0 -term: h0jT .x/ .y/j0i D DF .x y/ D x y
3| 1 -term:
8 9
ˆ
ˆ >
>
ˆ >
. i /
< Z =
4
h0jT .x/ .y/ d ´ .´/ .´/ .´/ .´/ j0i (4.52)
ˆ
ˆ 4Š >
>
ˆ
: R
„ ƒ‚
R
… >
;
dt d 3x
Wick’s theorem
. i /
Z
D 3 DF .x y/ d 4´ DF .´ ´/DF .´ ´/ (4.53)
4Š
. i /
Z
C12 d 4´ DF .x ´/DF .y ´/DF .´ ´/
4Š
Dx y C (4.54)
x y
! Interpretation:
edges D ⁂ propagators
8 9
$ DF R = Analytic
Feynman
<
internal points D ⁂ vertices 4
$ . i / d ´ expres-
diagram :
external points D spacetime points $ x; y; : : : sion
;
„ ƒ‚ …
Feynman diagram ¶ Process of particle creation & propagation & annihilation
(4.55)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 56
LE C T U R E 9 → PS:90–95
Internal points are vertices with four emanating edges that are associated to an integration.
External points are vertices that are endpoints of a single edge and associated to boundary
conditions (i.e. given spacetime points x; y; : : : of the correlation function).
4| Prefactors:
• Feynman diagram = sum of all identical terms (= prefactor)
• ^ O.n /
1
! factor nŠ and n integrals/vertices
1
! nŠ possibilities to interchange vertices cancels nŠ
1
! ignore the nŠ
• 4 contractions at each vertex
! 4Š possibilities to interchange contractions
1 1
! 4Š of interactionRcancels 4Š (this is the reason for the 4Š in the first place)
! associate . i / d 4´ with each vertex
• Symmetries of diagrams reduce the number of different contractions
! divide expression by the symmetry factor S
• Examples:
0 1
!
S D2 and S @ AD222D8 (4.56)
B C
x y
Imagine the diagram is made from strings pinned at external points and placed
flat on the table. Strings emanating from a vertex are marked with a colored flag.
Count the configurations that look the same when one forgets about the flags but
are different when the flags are taken into account.
Therefore:
1
Z
x y D DF .x y/. i / d 4´ DF .´ ´/DF .´ ´/ (4.57)
8
1
Z
D . i / d 4´ DF .x ´/DF .y ´/DF .´ ´/ (4.58)
x y 2
5| Therefore
n R o X Feynman diagrams with two
i dt HI .t/
h0jT .x/ .y/e j0i D (4.59)
external points x and y
with the position/real-space Feynman rules for the 4 -theory
1: For each propagator, x y D DF .x y/
2: For each vertex, D . i / d 4´
R
(4.60)
3: For each external point, x D1
1
4: Divide by the symmetry factor, S :::
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 57
LE C T U R E 9 → PS:90–95
The integration over spacetime coordinates ´ at each internal vertex accounts for the
superposition principle: We sum over all spacetime positions where the absorption/emis-
sion of particles – represented by vertices – can occur.
6| Often calculations are simpler in momentum space:
d 4p i
Z
ip.x y/
DF .x y/ D e (4.61)
.2/4 p 2 m2 C i "
Assign arbitrary orientation to edges (since DF .x y/ D DF .y x/) and perform
vertex integrals:
p2
Z
p1
p3 D . i / d 4´ : : : D . i /.2/4 ı.p1 C p2 p3 p4 / (4.62)
p4
Details: → Problemset 6
! Momentum conservation at vertices
Note that it is actually
Z Z T Z
4 0
d ´ ::: D lim d´ d 3´ : : : (4.63)
T !1.1 i"/ T
! Momentum-space Feynman rules:
1: For each propagator, p D i
p 2 m2 Ci"
p2
p1
2: For each vertex, p3 D . i /.2/4
ı.p1 Cp2 p3 p4 /
p4 (4.64)
3: For each external point, x D e ipx
p
Q R d 4pi
4: Integrate momenta, i .2/4
:::
1
5: Divide by sym. factor, S :::
Equivalence between momentum- and position space Feynman rules: → Problemset 6
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 58
LE C T U R E 10 → PS:96–99
→ Topics of Lecture 10
1. Resummation of disconnected diagrams
2. Vacuum energy
3. Cross sections
7| Problem: Disconnected pieces of diagrams diverge!
Example:
1
Z
D . i / d 4´ DF .0/DF .0/ (4.65)
8 „ ƒ‚ …
const
/ .2T / .volume of space/ D .2/4 ı.0/ (4.66)
Interpretation: This “detached” process can happen anytime and anywhere in an
infinitely large (for T ! 1.1 i "/) spacetime volume—and we have to sum up all
these amplitudes!
8| Exponentiation of disconnected diagrams:
This is a preliminary step to cancel the divergencies with the denominator, → below.
i| Typical diagram:
7sieceS
7
x and y are always connected because the sum of all degrees of all vertices of a
connected graph is always even (=twice the number of edges). Note that the only
(graph) vertices with odd degree in a 4 -Feynman diagram are the external points.
ii | Let
<Set of all disconnected =
8 9
V D fV1 ; V2 ; : : : g Feynman diagrams (4.67)
without external points
: ;
Set of all connected
8 9
ˆ >
Feynman diagrams
ˆ
< >
=
xy
F (4.68)
ˆ
ˆwith external points x > >
and y
: ;
8 9
ˆ
< >
=
xy
! Feynman diagram F D F
„ƒ‚… ; V1 ; : : : ; V1 ; V2 ; : : : ; V2 ; V3 ; : : :
:Connected part „ ƒ‚ … „ ƒ‚
ˆ … >
Multiplicity n1 n
;
2
(4.69)
Abuse of notation: Vi denotes also the value of the corresponding diagram.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 59
LE C T U R E 10 → PS:96–99
iii | Amplitude of F :
Y 1
F D F xy .Vi /ni (4.70)
ni Š
i „ƒ‚…
Si
Si : Symmetry factor for exchanging the ni copies of Vi
iv | Then
n R o
i dt HI .t/
h0jT .x/ .y/e j0i (4.71)
" #
X X Y 1
D F xy .Vi /ni (4.72)
ni Š
F xy 2F xy n1 ;n2 ;::: i
" # " #
X X Y 1
xy
D F .Vi /ni (4.73)
n1 ;n2 ;::: i
n i Š
F xy 2F xy
" # " #
X YX 1
D F xy .Vi /ni (4.74)
n i Š
F xy 2F xy i ni
" # " #
X X
D F xy exp Vi (4.75)
F xy 2F xy i
In words:
.sum of all diagrams/ D (4.76)
.sum of all connected pieces/ expŒsum of all disconnected pieces (4.77)
!
n R o
i dt HI .t/
h0jT .x/ .y/e j0i D †.F xy / e †.V / (4.78)
with †.X / x2X x
P
9| Denominator of Eq. (4.31):
n R o
i dt HI .t/
h0jT e j0i D e †.V / (4.79)
The argument runs along the same lines as for the numerator.
10 | Two-point correlator:
hjT .x/ .y/ji D †.F xy / (4.80)
Sum of all connected diagrams
D (4.81)
with two external points
11 | Generalization to n-point correlators:
hjT .x1 / : : : .xn /ji D †.F x1 :::xn / (4.82)
Sum of all connected diagrams
D (4.83)
with n external points
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 60
LE C T U R E 10 → PS:96–99
In 4 -theory, correlators with n odd vanish identically as the set F x1 :::xn of allowed
connected diagrams is empty. This follows also from Wick’s theorem where full
contractions are only possible with an even number of fields.
→ Note 4.2
• Connected diagrams are connected to external points and not necessarily connected
graphs:
• Disconnected diagrams = “Vacuum bubbles”
• Interpretation of vacuum bubbles:
With Eq. (4.26) and Eq. (4.25)
( " #)
Z T
lim h0jT I .x/ I .y/ exp i dt HI .t / j0i (4.84)
T !1.1 i"/ T
0 1
D hjT .x/ .y/ji lim @jh0jij2 e iE0 .2T / A (4.85)
„ ƒ‚ … T !1.1 i"/ „ ƒ‚ …
D†.F xy / /expŒ†.V /
With Vi D VQi .2T V / where V D Volume of space:
E0 X
Di VQj (independent of T ) (4.86)
V
j
! Total vacuum energy E0 / V (good!)
! Vacuum bubbles determine vacuum energy density
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 61
LE C T U R E 10 → PS:99–108
4.5 Cross Sections and the S -Matrix
• So far: Time-ordered correlators hjT 1 : : : n ji ! Cannot be directly measured
• Now: Cross sections (can be measured with scattering experiments in particle accelerators)
• Recipe: Feynman diagrams ! S-matrix ! Cross section
The Cross Section
1| ^ Scattering experiment:
Collide two beams of particles with well-defined momenta and observe the outcome:
F
..
.*
... u . .. .
m m m m m .
- — {m et
. I;
° ‚!
.
fl. “
2| Cross section:
# of scattering events (with outcome X )
.X/ ) ΠD L2 D Area (4.87)
A `A B `B A
! Encodes the likelihood of scattering event X
! Intrinsic property of the colliding particles
In particular, the cross section is independent of the parameters of the experiment (like
beam size, particle density etc.).
3| Real experiments: Densities not homogeneous across beam: X 7! X .x; y/
If interaction range and wavepacket size are much smaller than the beam diameter, the
densities can be taken as (locally) constant and the following derivations apply. The only
difference is:
# of scattering events (with outcome X ) (4.88)
Z
D .X/ `A `B dx dy A .x; y/B .x; y/ (4.89)
Beam cross section
homogeneous beam: X D const (4.90)
.X/ NA NB
D (4.91)
A
NX : # of particles of type X in the interaction volume `X A
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 62
LE C T U R E 10 → PS:99–108
4| Typically there are many outcomes X possible, e.g.
8
ˆ
ˆ
ˆ eCe
<C ( : muon)
ˆ
eCe ! (4.92)
ˆ
ˆ
ˆ
C
( : photon)
ˆ
:: : :
The possible outcomes depend on the field content of the theory and the interactions
that couple them.
5| Differential cross section:
^ Scattering outcome X of n final particles with momenta .pE1 ; : : : ; pEn / 2 Vp
Vp R3n : final-state 3-momentum subspace
d
Z
X jVp D d 3p1 : : : d 3pn (4.93)
d p : : : d 3pn
3
Vp
„ 1 ƒ‚ …
Differential cross section
! Constrained by 4-momentum conservation: i pi D const
P
(This follows from spacetime translation symmetry; there are 4 independent constraints.)
Special case: n D 2
! 6 dof (degrees of freedom) .pE1 ; pE2 / and 4 constraints
! 2 dof ! Scattering direction . ; / in center-of-mass frame:
d d
! (4.94)
d 3p1 d 3p2 d
Differential of the solid angle: d D sin. / d d
Here we skip another measurable quantity: the decay rate
# of decays per unit time (into state X )
.X/ (4.95)
# of particles A present
In scattering experiments, the decay of unstable intermediate particles modifies the scattering
cross section according to the Breit-Wigner formula (a Lorentzian distribution)
1
/ (4.96)
.E 2 m2 /2 C m2 2
with m the rest mass of the unstable intermediate particle and E the center-of-mass energy of
the collision (this is called a resonance).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 63
LE C T U R E 11 → PS:99–108
→ Topics of Lecture 11
1. The S - and the T -matrix
2. Compute cross sections from S -matrix
→ Topics of Problemset 6
1. Feynman diagrams for 4 -theory
2. Feynman rules for the interacting complex Klein-Gordon field
The S-Matrix
Goal: Compute cross sections
Recipe: Start with initial states ! evolve in time ! compute overlap with final states
Note: Henceforth we consider the scattering of two particles A and B
1| ^ One-particle wavepacket
d 3k 1 d 3k
Z Z
j i D E ki
.k/j E with E 2 D 1 D h j i
j .k/j (4.97)
3 .2/3
q
.2/ 2E kE
q
E one-particle state of interacting theory (jki
jki: E0D 2EkE a E j0i for free theory)
k
Just as the vacuum j0i 7! ji is “dressed” by the vacuum fluctuations due to the
E 0 7! jki.
interactions, the single particle states are as well: jki E
2| We want the probability
P D jout h 1 : : : n j A B iin j2 ¶ (4.98)
„ ƒ‚ …
formal expression (definition: → below)
j A B iin W ⁂ in-state at T ! 1 of two separated wavepackets
(4.99)
j 1 : : : n iout W ⁂ out-state at T ! C1 of n separated wavepackets
3| Fourier transform in-states (wlog):
EE
d 3kA d 3kB A .kEA / B .kEB /e i b kB E E
Z Z
E in D
j A B .b/i jkA kB iin (4.100)
.2/3 .2/3
q
.2EkE /.2EkE /
A B
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 64
LE C T U R E 11 → PS:99–108
E ⁂ Impact parameter
b:
Here we assume that the wave packets in the in-states are far apart such that interactions
between the particles can be neglected. Therefore the state of two particles is simply
given by the tensor product of two single-particle states.
E are collinear; shifts by bE
By convention, all wavefunctions constructed from B .k/
perpendicular to the axis of incidence are then realized by e i bEkE :
4| Simplification:
j 1 : : : n iout ! jpE1 : : : pEn iout (4.101)
This simplification can be justified by the characteristics of particle detectors which
predominantly measure the energy (and therefore the momentum) of scattered (on-shell)
particles and cannot resolve positions on the scale of de Broglie wavelengths.
With Eq. (4.100), we are interested in
out hp
E E
E1 : : : pEn jkA kB iin ¶ (4.102)
5| S-matrix (⁂ Scattering matrix):
out hp
E E
E1 : : : pEn jkA kB iin WD lim CT hp
E E
E1 : : : pEn jkA kB i T (4.103)
T !1
D lim t0 hp
E1 : : : pEn je iH.2T /
jkEA kEB i t0 (4.104)
T !1
t0 hp
E E
E1 : : : pEn jSjkA kB i t0 (4.105)
Example: S D 1 for free theory (momentum modes are eigenstates of H D H0 !)
The states above are all Heisenberg states. However, if we label Heisenberg states by
eigenvalues of operators, we have to specify a time because the operators themselves
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 65
LE C T U R E 11 → PS:99–108
evolve in time. E.g., the state jki t denotes the eigenstate of the momentum operator
P .t/ at time t with eigenvalue k,
P .t /jki t D kjki t : (4.106)
Heisenberg states that are labeled by a fixed quantum number (= eigenvalue of an
operator) therefore depend on time! The translation of states in time for a fixed
eigenvalue is given by
jki t D e iH.t t0 /
jki t0 : (4.107)
Note the different sign in the exponential compared to time-evolution operators of
Schrödinger states!
In Eq. (4.103), the Heisenberg states are labeled by the single-particle momenta of
incoming and outgoing particles at different times. To translate them to a common
reference time t0 , we use
jkEA kEB i T D e iH. T t0 /
jkEA kEB i t0 (4.108)
jpE1 : : : pEn iCT D e iH.CT t0 /
jpE1 : : : pEn i t0 (4.109)
which yields Eq. (4.104).
6| T -matrix:
S 1 C iT (4.110)
„ƒ‚… „ƒ‚…
particles miss each other non-trivial scattering
7| Translation-invariant Hamiltonian H ! 4-momentum conservation !
hpE1 : : : pEn jiT jkEA kEB i .2/4 ı .4/ kA C kB
P
f pf (4.111)
„ ƒ‚ …
kinematics
dynamics
‚ …„ ƒ
i M.kA kB 7! fpf g/
„ ƒ‚ …
⁂ Invariant matrix element
Note: Here all 4-momenta are on-shell, i.e., p 0 D EpE D pE2 C m2 .
p
Invariant matrix element ¶ Scattering amplitude of one-particle quantum mechanics
Two questions:
• M D‹ (→ later)
‹
• D .M/ (→ now)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 66
LE C T U R E 11 → PS:99–108
^ Probability to scatter in infinitesimal momentum volume d Vp D f d 3pf :
Q
8|
Q d 3pf 1 E in j2
dP .ABbE 7! 1 : : : n/ D f .2/3 2EpE jout hpE1 : : : pEn j A B .b/i (4.112)
f
„ ƒ‚ …
for normalization
¶ (4.113)
The normalization defines a Lorentz invariant measure, as discussed before. The 1=2EpEf
is necessary for normalization
q because of our convention to normalize single-particle
states with a prefactor of 2EkE . Note that the integral over dP should be unity if there
is only one decay channel.
Here we assume that amplitudes for different momenta do not interfere as the particle
ˇ2 R
detector measures momentum distributions: ˇ V dp hpj iˇ V dp jhpj ij2 .
ˇR
9| ^ Single target particle A and many incident particles BbE :
Z
d.# scattering events/ D d 2b nB dP .ABbE 7! 1 : : : n/ (4.114)
nB : Area density of B-particles
E l0 )
By assumption, nB const on the interaction length scale l0 (i.e., dP 0 for jbj
!
d.# scattering events/
Z
d D D D d 2b dP .ABbE 7! 1 : : : n/ (4.115)
nB NA nB 1
„ƒ‚… „ƒ‚…
B `B A `A A
insert everything
Z !
Q d 3pf 1 i .kEi /
d 3ki q d 3qi q i .E
qi /
d 2b i DA;B
Q R R
D f .2/3 2E (4.116)
Ef
p .2/3 2E E .2/ 3
2EqE
k i i
E E
e„i b.Eqƒ‚
B kB /
out hfpEf gjfkE gi
j in out hfp
Ef gjfE
qj giin
… „ ƒ‚ …
? ?
.2/2 ı .2/ kB qB
„ ƒ‚ …
i M.fkj g 7! fpf g/ i M .fqj g 7! fpf g/
P .2/4 ı .4/ P q P
pf
.2/4 ı .4/
P
kj pf j
For the matrix elements, we ignored the identity 1 in the S -matrix as we are only
interested in non-trivial scattering events given by the T -matrix.
! Evaluate the six qi -integrals:
Only the two qi´ -integrals are non-trivial; note that we assume wlog that bE ? eE´ :
y
? x ? ?
(this follows from ı .2/ kB
? ?
)
i | qB D .qB ; qB /-integrals )qB D kB qB
? x y ? ?
ii | qA D .qA ; qA /-integrals )qA D kA
(this follows from (i) in combination with the remaining two ı-functions)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 67
LE C T U R E 11 → PS:99–108
´ ´
iii | ^ qA qB -integrals (here we focus only on the ı-functions and omit the fields):
0 1
Z X X
´ ´ ´ ´
dqA dqB ı qA C qB pf´ ı @ EA C EB Ef A
B C
„ ƒ‚ …
Depend on qi´ !
0 1ˇ
ˇ
Z ˇ
Bq q X Cˇˇ
C
´ B 2 2 2
D dqA ı B qEA C mA C qEB C m2B Ef Cˇ (4.117)
@„ ƒ‚ …Aˇˇ
´
g.qA / ´
pf´ qA
´
ˇ P
qB D
ˇ
1 ˇ
D 0 ´ ˇ (4.118)
ˇ
jg .qA /j ˇ ´
g.qA /D0
1 1
$ˇ ´ ´ ˇ (4.119)
ˇ qA qB ˇ jvA vB j
ˇ EA EB ˇ
´
where i), ii), and qB D pf´ qA ´
are implied;
P
´ ´
qA is a solution of g.qA / D 0 , EA C EB D Ef .
P
@E.E
q/ qE
vX is the velocity of particle X in the lab frame; recall: vEgroup D @E
q
D E.Eq/
for
relativistic particles.
´
Note that from the two four-dimensional delta distributions it follows that kA C
´ ´ ´ E E
k D q Cq and E.kA /CE.kB / D E.E qA /CE.E qB /. Together with q D k ?
?
B A B i i
these constraints are solved by ki´ D qi´ and therefore kEi D qEi ; this will be used in
the next step to simplify the expression further. In particular, qEi is a function of
kEi and therefore still integrated over; this extends to the expression in Eq. (4.119)
which implicitly depends on kEi !
These calculations are sloppy and lack mathematical rigour. Can this be improved?
10 | ^ i .kEi / peaked around pEi for i D A; B !
(pull all continuous functions of kEi out of the integrals)
jM.pA pB 7! fpf g/j2
Q d 3pf 1
d D f .2/3 2EpE
f 2EpEA 2EpEB jvA vB j
d 3kA d 3kB
Z Z X
j A .kEA /j2 j B .kEB /j2 .2/4 ı .4/ kA C kB pf
.2/3 .2/3
(4.120)
Note that we cannot simply demand plane waves for the incoming wave packets, because
we assumed that the incoming particles were well-separated and non-interacting! The
best we can do is to demand i .kEi / to be peaked around pEi while still describing an
elongated but localized wave packet.
11 | Particle detectors project onto momentum eigenstates with finite resolution
! Cannot resolve momentum spread of initial wavepackets
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 68
LE C T U R E 11 → PS:99–108
! kA C kB pA C pB
1 Q d 3pf 1
d D f .2/3 2EpE (4.121)
2EpEA 2EpEB jvA vB j f
X
jM.pA pB 7! fpf g/j2 .2/4 ı .4/ pA C pB pf
Here we used the normalization of the initial wavepackets.
Note that this result is independent on the shape of the initial wavepackets!
For j i .kEi /j2 ı .3/ kEi pEi this approximation becomes exact.
Note that d is not Lorentz invariant because the prefactor
R
1
(4.122)
2EpEA 2EpEB jvA vB j
is not (it transforms non-trivially under boosts perpendicular to the axis of incidence (E
e´ )
because of Lorentz contraction).
However, the remaining terms are Lorentz invariant (Li): (1) the measure is Li as
shown before, (2) the invariant matrix element is Li because T commutes with the
unitary representation of Lorentz transformations on the asymptotic Hilbert space [for
a proof, ↑ pp. 116–121 of Weinberg’s The Quantum Theory of Fields (Vol 1) [1]] (note
that this requires additional assumptions since the Hamiltonian does not commute
with the generators of boosts), and (3) the ı-distribution is Li since the equation
pA C pB D pf is Lorentz covariant (i.e., valid in all inertial systems).
P
Special Cases
Details: → Problemset 7
12 | ^ Two final particles (p1 and p2 ) in center-of-mass frame:
(pEA C pEB D 0 , pE1 D pE2 )
d
1 jpE1 j
$ jM.pA pB 7! p1 p2 /j2 (4.123)
d cm 2EpEA 2EpEB jvA vB j .2/2 4Ecm
.pA C pB /2 D ŒEpEA C EpEB cm : center-of-mass energy (Lorentz invariant!)
p
Ecm D
To end up with Eq. (4.123) one has to perform 4 momentum integrals to reduce the 6 dof
that are restricted by 4-momentum conservation to the remaining 2 dof encoded in the
angular dependency d .
13 | If, in addition, mA D mB D m1 D m2 :
d jM.pA pB 7! p1 p2 /j2
$ (4.124)
d cm 64 2 Ecm
2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 69
LE C T U R E 12 → PS:108–115
→ Topics of Lecture 12
1. Computing S -matrix elements perturbatively from Feynman diagarams
2. Fully connected and amputated Feynman diagrams
4.6 Computing S -Matrix Elements from Feynman Diagrams
The main result of this section will be motivated but not rigorously derived. For the proof, a
technical result known as ⁂ LSZ reduction formula is needed. For details, ↑ Chapter 7.2 in
Peskin & Schroeder.
Motivation
1| We want
iH.2T /
hpE1 : : : pEn jS jpEA pEB i D lim t0 h p
E1 : : : pEn je jpEA pEB i t0 (4.125)
T !1
We omit the common reference time t0 of the Heisenberg states in the following.
2| Problem:
Eigenstates of H0
p p
jpEA pEB i0 D 2EpEA 2EpEB apE apE j0i
A B
jpEA pEB i D ‹ E Eigenstate of H D H0 C Hint
jpi:
(4.126)
Interactions “deform” not only the vacuum j0i 7! ji but also the single-particle states
jpi E in a highly non-trivial way.
E 0 7! jpi
3| Remember: For the vacuum we found
iE0 T 1 iH T
ji D lim .e hj0i/ e j0i (4.127)
T !1.1 i"/
4| Assume it holds similarly
iH T
jpEA pEB i D lim e jpEA pEB i0 (4.128)
T !1.1 i"/ „ƒ‚…
Prefactors & Overlaps
This construction is not easy and we deliberately omit prefactors and overlaps!
Remember that in the case of vacuum expectation values, these prefactors canceled; here
the same happens in the end.
5| If this holds, we could write
iH.2T /
hpE1 : : : pEn jS jpEA pEB i / lim 0 hp
E1 : : : pEn je jpEA pEB i0 (4.129)
T !1.1 i"/
" #
Z T
/ lim E1 : : : pEn jT exp
0 hp i dt HI .t / jpEA pEB i0
T !1.1 i"/ T
(4.130)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 70
LE C T U R E 12 → PS:108–115
h i
In the first line, we used that e iH T .1 i"/ e iH 2T e iH T .1 i"/ De iH Œ2T .1 i"/ .
In the last line, we used that
" Z #
T
def
U.T; T / D T exp i dt HI .t / $ e iH0 .T t0 /
e iH.2T /
e iH0 . T t0 /
(4.131)
T
and that jpEA pEB i0 and 0 hpE1 : : : pEn j are eigenstates of H0 .
6| Correct result:
hpE1 : : : pEn jiT jpEA pEB i D
( " # )
Z T
lim 0 hp
E1 : : : pEn jT exp i dt HI .t / jpEA pEB i0 (4.132)
T !1.1 i"/ T
fc&a
fc&a = “fully connected and amputated”
= restriction on Feynman diagrams that contribute to this amplitude (→ below)
Interpretation & Application
Here: ^ 4 -theory
Details: → Problemset 7
1| 0 -order: (assume pEA ¤ pEB )
q
hpE E
p j E
p
0 1 2 A B 0 E
p i D 2EpE1 2EpE2 2EpEA 2EpEB h0japE1 apE2 apE apE j0i (4.133)
A B
( .3/ .3/ )
ı pEA pE1 ı E
pB E
p2
$ 2EpEA 2EpEB .2/6 (4.134)
Cı .3/ pEA pE2 ı .3/ pEB pE1
1 2 1 2
D C (4.135)
A B A B
! State does not change (Bosons!)
! Contributes to 1 in S D 1 C iT (! not part of fc&a diagrams)
Note that there is only one particle type in 4 theory, so all particles have naturally the
same mass.
2| 1 -order:
i|
Z
4 4
0 hpE1 pE2 j i d x T f I .x/g jpEA pEB i0 (4.136)
4Š
Wick’s theorem
Z
4 4
D0 hpE1 pE2 j i d x W I .x/ C contractionsW jpEA pEB i0 (4.137)
4Š
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 71
LE C T U R E 12 → PS:108–115
ii | Careful: Not only full contractions survive because the states contain particles:
d 3k 1
Z q
IC .x/jpi
E 0D a e ikx
2EpE apE j0i $ e ipx
j0i (4.138)
.2/3 2E kE
q
EEk
d 3k 1 q
Z
0 hpj
E I .x/D h0j 2EpE apE a E e Cikx $ h0je Cipx (4.139)
.2/3 2E
q
k
EEk
Recall that not fully contracted, normal-ordered products contain IC fields on the
right and I fields on the left.
iii | Definition:
E e
I .x/jpi ipx j0i ¶ p
(4.140)
E I .x/ h0je Cipx
hpj ¶ p
E q i 2EpE .2/3 ı .3/ pE
hpjE qE ¶ q p
We omit the subscript 0 for states whenever it is implied by the context to lighten
the notation.
Feynman diagrams for S-matrix elements contain external lines (labeled by mo-
menta) instead of external points (labeled by positions) as compared to the diagrams
for correlation functions.
iv | Then
Sum of all full contractions of
n o
0 hp
E1 : : :jT f a : : : gjpEA : : :i0 D
fields and external-state momenta
(4.141)
This is a generalization of Wick’s theorem for states with external momenta.
Example:
0 hp
E1 pE2 jpEA pEB i0 D hpE1 pE2 jpEA pEB i C hpE1 pE2 jpEA pEB i (4.142)
D (4.135) (4.143)
v| Application to Eq. (4.137):
Z
i d 4x 0 hpE1 pE2 jT f I4 .x/gjpEA pEB i0 D : : : (4.144)
4Š
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 72
LE C T U R E 12 → PS:108–115
! Terms with and (red) do not contribute to T
! Only fully connected diagrams contribute to T
fully connected = all external lines are connected to each other
The integral in Eq. (4.137) yields a momentum-conserving ı-distribution at the
vertices.
vi | Therefore
p1 p2
hpE1 pE2 jiT jpEA pEB i (4.145)
pA pB
Z
D .4Š/ i d 4x e i.pA CpB p1 p2 /x
(4.146)
4Š
D i .2/4 ı .4/ .pA C pB p1 p2 / (4.147)
def
D i M.2/4 ı .4/ .pA C pB p1 p2 / (4.148)
! M.pA pB 7! p1 p2 / D C O.2 /
The factor 4Š comes from the 4Š possibilities to contract the four external momenta
with the four fields (above we show only one of these contractions exemplarily).
! (→ Problemset 7)
2
total D (4.149)
2
32Ecm
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 73
LE C T U R E 12 → PS:108–115
By measuring total in a particle collider, one can determine the coupling constant .
Note the factor 1=2 as the two final particles are indistinguishable (that is, the final
states hpE1 pE2 j and hpE2 pE1 j are physically equivalent and must not be counted twice)!
3| Higher-order contributions:
hpE1 pE2 jiT jpEA pEB i D (4.150)
8 9
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
< >
=
(4.151)
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
: >
;
fc&a
Which of these diagrams qualify as “fc&a”?
(a) Not fully connected (← above) 7
(b) 1 -order contribution (← above) 3
(c) Higher-order contributions 3
(d) Diagrams with vacuum bubbles
! Bubbles exponentiate & drop out (as before) 7
(e) Fully connected diagrams with “appendices to external legs” 7/3?
p1 p2
k 1
Z
d 4p 0 i
Z
d 4k i
$ (4.152)
2 .2/ p4 02 m2 .2/ k4 2 m2
p0
pA . i /.2/4 ı .4/ pA C p 0 p1 p2
pB
. i /.2/4 ı .4/ pB p 0
1 1 1
2 D D D1 (4.153)
pB m2 Ep2E 2
pEB m 2 0
B
The two momentum integrals come from the two propagators after integrating out
the vertex positions. The prefactor 1=2 is the symmetry factor of the loop.
Note that the external momenta are on-shell, p 2 D m2 , whereas the momentum
integrals of internal momenta go over off-shell momenta, p 2 ¤ m2 , as well.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 74
LE C T U R E 12 → PS:108–115
! Eq. (4.132) makes only sense without these diagrams!
Interpretation:
! Not related to scattering! ! 7
! “Amputate” legs for calculation of M
4| Amputation of diagrams:
Starting from the tip of each external leg, cut at the last point at which the diagram can
be cut by removing a single propagator, such that this operation separates the leg from
the rest of the diagram.
Example:
5| !
i M .2/4 ı .4/ pA C pB
P
(4.132) D pf (4.154)
< Sum of all fully connected, amputated Feyn-=
8 9
D man diagrams with pA , pB incoming and (4.155)
fpf g outgoing
: ;
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 75
LE C T U R E 12 → PS:108–115
6| ! Position-space Feynman rules for scattering amplitudes in 4 -theory:
1: For each edge, x y D DF .x y/
2: For each vertex, D . i / d 4´
R
(4.156)
3: For each external line, p De ip´
1
4: Divide by the symmetry factor, S :::
Only (3) is modified as compared to Feynman rules for correlation functions.
^ Momentum-space representation of DF & vertex integration
! ı-distributions at vertices & momentum integrals over internal momenta
7| ! Momentum-space Feynman rules for scattering amplitudes in 4 -theory:
1: For each edge, p D i
p 2 m2 Ci"
p2
p1
2: For each vertex, p3 D . i /.2/4
ı.p1 Cp2 p3 p4 /
p4
(4.157)
3: For each external line, p D1
d 4pi
4: Integrate int. momenta,
Q R
i .2/4
:::
1
5: Divide by sym. factor, S :::
Only (3) is modified as compared to Feynman rules for correlation functions.
8| Because of the many ı-distributions, the expressions obtained from the momentum-space
Feynman rules can be simplified considerably. On pp. 114–115 of P&S this is mentioned
and, after canceling the global momentum conservation, a set of Feynman rules where
only integrals over “undetermined loop momenta” are left is given. This prescription is
rather obscure as they do not define what and how many of these “loop momenta” there
are.
So let us think about this more carefully:
i| Consider a fully connected, amputated Feynman diagram with Ne external mo-
menta, Ni internal momenta, and Nv vertices.
ii | We can interpret the Feynman diagram as a connected graph (in the sense of graph
theory) with E D Ni C Ne edges and V D Nv C Ne vertices (these are now
“graph theory vertices”, i.e., external legs terminate at vertices).
iii | By variable substitutions, the Nv ı-distributions can be rewritten as follows:
P
ı.: : : / ı.: : : / D ı.pA C pB pf / ı.: : : / ı.: : : / (4.158)
„ ƒ‚ … „ ƒ‚ …
Nv Nv 1
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 76
LE C T U R E 12 → PS:108–115
(The argument of the global ı-distribution is just the sum of all Nv arguments
of the original ı-distributions at vertices.) Note that the global momentum
conservation cannot be used to remove a momentum integral; but it can be
cancelled with the same expression in Eq. (4.111) so that the remaining expression
equals i M.pA pB 7! fpf g/.
iv | This remaining expression has Ni momentum integrals but only Nv 1 ı-
distributions, so that
#.Loop integrals/ D Ni Nv C 1 (4.159)
integrals remain after integrating over all ı-distributions.
v| To see why these are integrals over “loop momenta”, we have to put on our graph
theory goggles again: For a given (connected) graph, the set of all closed circuits
(= loops = Eulerian subgraphs) forms a binary vector space (adding two loops is
done modulo-2 on the edges), the so called cycle space C. It is well-known that the
dimension of this space (= the number of basis-loops) is given by
dim C D E V C 1 D Ni Nv C 1 : (4.160)
This suggests, that for each basis-loop of a given Feynman diagram, there is one
undetermined “loop momentum” to integrate over.
Add more details?
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 77
LE C T U R E 13 → PS:115–116,123–124
→ Topics of Lecture 13
1. The fermion sector of quantum electrodynamics (QED)
2. Wick’s theorem for fermions
3. The photon sector of QED and the photon propagator
4. Feynman rules for QED
→ Topics of Problemset 7
1. Cross section of two scattering particles
2. Important relations for gamma matrices
4.7 Feynman Rules for Quantum Electrodynamics
Setting the Stage
Here we leave 4 -theory and switch to fermionic fields.
We will use and generalize the results on interactions derived for 4 -theory for this new theory
without detailed derivations (as these are very technical).
1| Fields:
Fermions: ‰.x/ (bispinor field)
(4.161)
Photons: A .x/ (vector field)
2| Lagrangian:
LQED D LDirac C LMaxwell C Lint (4.162)
1
D ‰.i @= m/‰ F F e‰ ‰ A (4.163)
4 „ ƒ‚ …
j
1
=
D ‰.i D m/‰ F F (4.164)
4
m: Mass of fermions
e: Charge of fermions (= coupling constant)
D : covariant derivative: D D @ C i eA
The replacement @ 7! D is called minimal coupling and constitutes a general recipe
for coupling gauge fields to matter fields in a gauge-invariant way (← Note 4.1 and
→ Section 9.1).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 78
LE C T U R E 13 → PS:115–116,123–124
3| Hamiltonian:
HQED D HDirac C HMaxwell C Hint (4.165)
Z
with Hint D e d 3x ‰ ‰ A (4.166)
4| Equations of motion:
=
.i D m/‰ D 0 (gauge-covariant Dirac equation) (4.167)
@ F Dj (inhomogeneous Maxwell equations) (4.168)
→ Note 4.3
LQED is invariant under U.1/ gauge transformations,
‰ 0 .x/ D e ie˛.x/ ‰.x/ (4.169)
A0 .x/ D A .x/ @ ˛.x/ (4.170)
for arbitrary ˛ W R1;3 ! R.
This is the simplest example of an (abelian) gauge theory of the Yang-Mills form.
→ Note 4.4
The QED-sector of the standard model includes several copies of the fermion field that
all couple to the same photon field,
2 3
1
LSM
X6
qf ‰f ‰f A 7 F F ;
7
6‰f .i @= (4.171)
QED D mf /‰f
4 „ ƒ‚ … 5 4
f
jf
with mass mf and charge qf of fermion type
f 2 f Leptons; Quarks g D fe; ; ; e ; ; ; u; d; c; s; t; bg : (4.172)
Here we restrict our discussion to a single fermion type f (think of electrons/positrons).
The situation in the standard model is actually a lot more complicated than suggested
by LSM
QED due to gauge symmetry constraints that forbid mass terms (a situation that is
compensated by the Higgs mechanism and electroweak symmetry breaking; → later).
Notes on the Fermion/Dirac Sector
We have already quantized the free Dirac field LDirac and diagonalized the non-interacting
Hamiltonian HDirac !
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 79
LE C T U R E 13 → PS:115–116,123–124
Remember: Feynman propagator:
d 4p i.p= C m/ab
Z
SFab .x y/ D 4 2
e ip.x y/
(4.173)
.2/ p m2 C i "
(
h0j‰a .x/‰ b .y/j0i for x 0 > y 0
D (4.174)
h0j‰ b .y/‰a .x/j0i for x 0 < y 0
h0jT ‰a .x/‰ b .y/j0i (4.175)
To deal with Hint perturbatively, we need Wick’s theorem for fermions:
The proofs for all that follows are very similar to the bosonic case (except for the signs).
1| Time ordering: Eq. (4.175) suggests for 2 f‰; ‰g
Tf 1 : : : N g . 1/# 1::: N for x10 > > xN
0
(4.176)
: Permutation of f1; 2; : : : N g
. 1/# : Signum of with # number of operator interchanges
Note that here we suppress spinor indices!
s s
2| Normal order: Define for x 2 fapsE ; bpsE ; apE ; bpE g
Wx1 : : : xN W . 1/# .creation operators/ .annihilation operators/ (4.177)
#: Number of operator interchanges
3| Contraction: Define
a .x/ b .y/ Tf a .x/ b .y/g W a .x/ b .y/W (4.178)
Here, a and b are spinor indices!
This definition of the contraction is analogous to the bosonic case.
and show
( )
f‰aC .x/; ‰ b .y/g for x 0 > y 0
‰a .x/‰ b .y/ $ C $ SFab .x y/ (4.179)
f‰ b .y/; ‰a .x/g for x 0 < y 0
‰a .x/‰b .y/ $ 0 (4.180)
‰ a .x/‰ b .y/ $ 0 (4.181)
r
The last two contractions vanish since fapsE ; bqE g D 0.
4| Contraction & Normal order:
WA a .x/ B b .y/ C W . 1/# a .x/ b .y/ WABC W (4.182)
#: Number of operator interchanges (i.e., a .x/ with A and b .y/ with AB)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 80
LE C T U R E 13 → PS:115–116,123–124
5| Wick’s theorem: For 2 f‰; ‰g and a; b; : : : spinor indices
Tf a .x1 / b .x2 / : : : g DW a .x1 / b .x2 / C all possible contractionsW
(4.183)
Due to the adjusted definitions of time- and normal order, Wick’s theorem takes the same
form as for bosonic fields!
Notes on the Photon/Maxwell Sector
1| Observation: A has four degrees of freedom but there are only two photon polarizations!
2| Problem: Gauge invariance
! Unphysical degrees of freedom
! Fix gauge to quantize only physical degrees of freedom
3| Different solutions:
• Coulomb gauge r AE D 0 (not Lorentz invariant) (↓ Advanced quantum mechanics)
• Lorenz gauge @ A D 0 (Lorentz invariant)
(Gupta-Bleuler formalism, → Itzykson & Zuber, Quantum Field Theory, pp. 127–134)
• Faddeev-Popov procedure (→ later)
4| Motivation:
i| ^ Lorenz gauge: @ A D 0 ! EOMs for LMaxwell : @2 A D 0
Each component of A .x/ satisfies the Klein-Gordon equation for m D 0.
Recall: @ F D @2 A @ @ A D 0
Note that the Lorenz gauge does not fix the gauge freedom completely.
ii | Expand field in classical solutions:
3
d 3p 1 Xh r r
Z i
ipx r r
A .x/ D apE .p/ e C apE .p/ e ipx
.2/3 2EpE
p
rD0
(4.184)
with p 2 D 0 , p 0 D EpE D jpj E (dispersion of the massless KG equation)
: polarization 4-vectors (Lorenz gauge ! p
r r D p r D 0).
5| Results:
i| Impose constraints on external (physical) photons:
0
.p/ D and pE E.p/ D 0 (transverse polarization) (4.185)
E.p/
This reduces the number of degrees of freedom from 4 to 2!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 81
LE C T U R E 13 → PS:115–116,123–124
! Two r; s D 1; 2 independent bosonic modes for each momentum p:
E
h i h i h i
s r s
aprE ; aqE D .2/3 ırs ı .3/ pE and aprE ; aqsE D 0 D apE ; aqE
qE
(4.186)
ii | Propagator (Feynman gauge):
d 4q ig
Z
iq.x y/
h0jT fA .x/A .y/gj0i D e (4.187)
.2/4 q 2 C i "
We will derive the photon propagator using path integrals at the end of this course.
As each component of A satisfies the KG equation, the propagator should be
similar to the massless KG propagator DF .x y/. The two-point correlator is
a second-rank tensor that should be invariant under Lorentz transformations (as
the theory is relativistically invariant with a unitary representation of the Lorentz
group on the Hilbert space), which is realized only by g (see Eq. (6.60) below).
The sign makes the space-like components D D 1; 2; 3 positive and ensures
positive norm for states of the form Ai .x/j0i. In turn, states with A0 .x/j0i have
negative norm – but it can be shown that these states are never produced in physical
processes.
Feynman Rules
1. Expectations:
a) Two fields (‰a and A ) ! Two propagators ! Two line-types:
Fermions (with spinor indices a and b): a b (4.188)
Photons (with 4-vector indices and ): (4.189)
The arrow for fermions denotes the (negative) charge flow, not the momentum.
Since for fermion fields, particles are distinct from antiparticles, the arrow cannot
be neglected: It originates at a field ‰ that creates a particle (annihilates an
anti-particle) and terminates at a field ‰ that annihilates a particle (creates an
antiparticle).
! Two particle types: (anti-)fermions & photons
! Two types of external states:
Fermion/Antifermion: jp;
E sia=b (s: Spin; a: Fermion; b: Antifermion)
(4.190)
Photon: jp;
E ri (r: Polarization) (4.191)
For each in state (ket), there is a corresponding out state (bra).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 82
LE C T U R E 13 → PS:115–116,123–124
b) Interaction with three fields (Hint ‰ b ba
‰a A )
a
! Vertices of degree 3:
b
2. Momentum-space Feynman rules (for scattering amplitudes):
Note: In many textbooks, the colored indices are omitted.
The proofs are very technical but conceptually they parallel 4 -theory.
Note that there are three types of (graph) vertices:
• : internal vertex, corresponds to an interaction
• j : external vertex, corresponds to an in- or outgoing state
• : virtual cut of the diagram where Lorentz- or spinor-indices are summed
Examples & Applications: → next lectures
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 83
LE C T U R E 13 → PS:115–116,123–124
Propagators
p i.pCm/
=
Fermions: D ba
p 2 m2 Ci"
¶ ‰b .x/‰ a .y/
a b
p (simplified)
a b
ig
Photons: D q 2 Ci"
¶ A .x/A .y/
q
Vertices
a
¶ . i e/ d 4´
R
D ie ba ba
b
External legs
a js
Fermions: D usa .p/ ¶ ‰a jp;
E sia
p
sj a
D usa .p/ ¶ hp;
E sja ‰ a
p
a js
Antifermions: D v sa .p/ ¶ ‰ a jp;
E sib
p
sj a
D vas .p/ ¶ hp;
E sjb ‰a
p
jr
Photons: r .q/
D ¶ A jE
q ; ri
q
rj r .q/
D ¶ hE
q ; rjA
q
Evaluation
1. Impose momentum conservation at each vertex.
2. Integrate over all undetermined momenta.
3. Compute the overall sign of the diagram.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 84
LE C T U R E 14 → PS:125,131–138
→ Topics of Lecture 14
1. First application of QED: The Coulomb potential
2. Cross section of electron-positron scattering
3. Recipe for computing scattering cross sections in QED
First application: The Coulomb Potential
Before we start with the computation of relativistic QED predictions in the next chapter, let us
draw our first Feynman diagram and evaluate it in the non-relativistic limit to make contact
with known results.
1| ^ Møller scattering:
Electron (e ) C Electron (e ) ! Electron (e ) C Electron (e ) (4.193)
i| Contribution to the tree-level amplitude (sufficient for distinguishable fermions):
“Tree-level” refers to Feynman diagrams without loops; these correspond to
lowest/leading-order contributions to the scattering amplitude and do not contain
integrations over undetermined momenta.
For simplicity, we omit the spin labels s:
iM.e .p/e .k/ ! e .p 0 /e .k 0 // (4.194)
D (4.195)
0 ig
D ud .p /. i e /u .p/
dc c
ub .k 0 /. i e ba /ua .k/ (4.196)
q2
ig
D u.p 0 /. i e /u.p/ u.k 0 /. i e /u.k/ (4.197)
q2
Momentum conservation & integration over undetermined momenta !
with p p 0 D q D k 0 k
: sign of the diagram (→ below)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 85
LE C T U R E 14 → PS:125,131–138
• Note that the order of matrix-vector chains always follows the arrows of a
directed fermion path through the diagram; the different fermion paths are
connected by photon lines. The terms that correspond to different fermion
paths are commuting numbers indexed by as many spacetime indices as there
are vertices along the path.
• So far we did not encounter internal fermion lines that correspond to Feynman
propagators!
• Typically we omit the spinor indices and imply matrix-vector products.
• As electrons are indistinguishable, there is another tree-level diagram where the
outgoing states are exchanged. This diagram has to be added with the correct
sign to obtain the true tree-level scattering amplitude.
ii | Nonrelativistic limit: jpj
E 2 m2 ! Keep only lowest-order terms in p
(We will discuss a full-relativistic calculation in the next chapter in detail.)
p
p
p 1 1
! u.p/ D p m and 0 2
(4.198)
p .p p / jpE pE0 j2
Therefore
(
0 2mp0 p D0
u.p / u.p/ (4.199)
0 D 1; 2; 3
and
i e2
iM .2mp0 p /.2mk 0 k / (4.200)
jpE pE0 j2
iii | Compare with nonrelativistic scattering theory (↓ Born approximation):
hp 0 jiT jpi D i VO .E
q /.2/ı.EpE0 q D pE0
EpE / (E E
p) (4.201)
„ ƒ‚ …
Di M
VO .E
q /: Fourier transform of the scattering potential
(This is the first-order Born approximation which can be derived from the
Lippmann-Schwinger equation. Note that because of the static potential V , only
energy—but not momentum—is conserved.)
!
e2 e2 ˛
VO .E
q/ D 2 ) V .Er / $ D (4.202)
jE
qj 4jEr j r
˛ D e 2 =4 1=137: fine-structure constant (in natural units c D „ D "0 D 1)
The terms .2mp0 p / etc. are due to the QFT normalization conditions and must
be ignored for a sensible comparison with nonrelativistic scattering theory.
For the Fourier transform of the Coulomb potential in three dimensions, a
regularization is necessary. To this end, one Fourier transforms the Yukawa
mr
potential V .r/ D e4 r instead and sets the screening mass m to zero in the end;
for this integration, the residue theorem is needed.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 86
LE C T U R E 14 → PS:125,131–138
iv | Sign of the diagram: (here we suppress both spinor and spacetime indices)
E0 ; kE0 j ‰ ‰ A‰ ‰ A jp;
aa hp
E aa
E ki (4.203)
Dh0j akE0 apE0 ‰ ‰ A ‰ ‰ A apE a E j0i (4.204)
k
! 1+1+2=4 interchanges ! D C1 (4.205)
v| ! Repulsive Coulomb potential:
e2
Ve e .r/ D C (4.206)
4 r
! Equal charges repel each other (As it should be!)
2| ^ Bhabha scattering:
Electron (e ) C Positron (e C ) ! Electron (e ) C Positron (e C ) (4.207)
i| Contribution to the tree-level amplitude:
i M.e .p/e C .k/ ! e .p 0 /e C .k 0 // (4.208)
D (4.209)
0 ig
D ud .p /. i e /u .p/
dc c
v a .k/. i e ab /vb .k 0 / (4.210)
q2
0 ig
D u.p /. i e /u.p/ v.k/. i e /v.k 0 / (4.211)
q2
with p p0 D q D k0 k (Skip the spinor indices and reuse the diagram above.)
There is another tree-level contribution where an electron and a positron annihilate
to a virtual photon which then decays into an electron-positron pair. The sum of
both diagrams yields the tree-level scattering amplitude.
ii | Nonrelativistic limit ! Same result as Eq. (4.200) (with k $ k 0 ), but what is ?
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 87
LE C T U R E 14 → PS:125,131–138
iii | Sign of the diagram:
E0 ; kE0 j ‰ ‰ A‰ ‰ A jp;
ab hp
E ab
E ki (4.212)
Dh0j bkE0 apE0 ‰ ‰ A ‰ ‰ A apE b E j0i (4.213)
k
! 2+1+2=5 interchanges ! D 1 (4.214)
iv | ! Attractive Coulomb potential:
e2
VeC e .r/ D (4.215)
4 r
! Opposite charges attract each other (As it should be!)
These examples demonstrated four things:
• How to translate Feynman diagrams into analytical expressions.
• How to determine the sign of Feynman diagrams with fermions.
• The predictions of QED seem to be reasonable!
• Signs of diagrams are important!
The sign of amplitudes can also be determined from the diagrams directly by identification of
certain features of the diagram (like fermion loops). However, as we will rarely need this, we
skip the derivation/discussion of these rules and resort to counting fermion field interchanges.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 88
LE C T U R E 14 → PS:125,131–138
5 Elementary Processes of Quantum
Electrodynamics
In this short chapter, we use the machinery developed in the last few chapters to study
predictions of QED. Here we focus on ⁂ tree-level amplitudes. Diagrams with loops will be the
focus of the next chapter (⁂ radiative corrections).
5.1 Cross section of e C e ! C scattering
1| ^ Reaction
Electron (e ) C Positron (e C ) ! Muon ( ) C Antimuon (C ) (5.1)
This process is the simplest non-trivial QED process and used to calibrate e C e colliders.
2| Note: Both electrons and muons are spin- 12 fermions with equal charge qe D qm D e D
jej but different mass me mm :
We use m to label muons since is already used for spacetime indices.
2 3
1
Le;m
X 6
qf ‰f ‰f A 7 F F
7
6‰f .i @= (5.2)
QED D mf /‰f
4 „ ƒ‚ … 5 4
f De;m
jf
So there is one Fermion field for electrons/positrons ‰e and one Fermion field for
muons/antimuons ‰m . Mathematically, they only differ in the mass parameter mf that
enters the propagator. Note that the two Fermion fields can never couple directly but
only indirectly via the photon (gauge) field A !
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 89
LE C T U R E 14 → PS:125,131–138
3| Tree-level amplitude:
i M.e .p/e C .p 0 / ! .k/C .k 0 // (5.3)
D (5.4)
0 ig 0
D .v se /d .p 0 /. i qe dc /.use /c .p/ .urm /b .k/. i qm ba
r
/.vm /a .k 0 / (5.5)
„ ƒ‚ … q2 „ ƒ‚ …
Electron sector (e) Muon sector (m)
0 ig 0
D v se .p 0 /. i qe
/use .p/ urm .k/. i qm r
/vm .k 0 / (5.6)
q2
ie 2
v.p 0 / 0
D u.p/ u.k/ v.k / (5.7)
q2
with p C p 0 D q D k C k 0
• Typically we omit the spinor indices and imply matrix-vector products.
• In the following, we also suppress the spin superscripts and the fermion flavour
subscripts.
4| We want d / jMj2 ! need M . Use .v u/ $ .u v/:
e4
jMj2 D v.p 0 /
v.p 0 / u.k/ 0
/v.k 0 / u.k/
u.p/u.p/ v.k (5.8)
4
q „ ƒ‚ …
5| Typical collider setup:
• e C - and e -beam unpolarized ! Average over spin polarizations of in-states
• Muon detector cannot resolve spin ! Sum over spin polarizations of out-states
!
1 XX
d / jM.s; s 0 ! r; r 0 /j2 (5.9)
4 0 0
s;s r;r
6| Use spin sums Eq. (3.29) and spinor indices to evaluate :
X 0 s0
v sa .p 0 / ab usb .p/usc .p/ cd vd .p 0 / $ Tr .p=0 me /
= C me /
.p (5.10)
s;s 0
Details: → Problemset 8
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 90
LE C T U R E 14 → PS:125,131–138
7| !
1 X e4
jMj2 D 4 Tr .p=0
Tr .k= C mm / .k=
0
me / = C me /
.p mm /
4 0 0 4q
s;s ;r;r
(5.11)
Any squared and spin-summed QED amplitude with external fermions can be converted
into a trace of products of -matrices!
8| Trace technology: (due to Feynman, for derivations → Problemset 7)
Trace identities: (5.12)
Tr Œodd # of ’s D 0 (5.13)
Tr D 4g
(5.14)
Tr D 4 g g
g g Cg g (5.15)
Tr 5 D 0
(5.16)
Tr 5 D 0
(5.17)
Tr 5 D 4i "
(5.18)
Tr : : : D Tr : : :
(5.19)
Contraction identities: (5.20)
D4 (5.21)
D 2 (5.22)
D 4g (5.23)
D 2 (5.24)
These identities are useful for many QED calculations!
9| !
Tr .p=0 me / .p
$ 4 p 0 p C p 0 p g .pp 0 C m2e /
= C me / (5.25)
Tr .k= C mm / .k=0 mm / $ 4 k k0 C k k 0
g .kk 0 C m2m /
(5.26)
10 | Since me =mm 1=200, we set me D 0 henceforth:
(→ Problemset 8 for the general result with me ¤ 0)
1 X 8e 4
jMj2 $ 4 .pk/.p 0 k 0 / C .pk 0 /.p 0 k/ C m2m .pp 0 /
(5.27)
4 0 0 q
s;s ;r;r
11 | ^ Center-of-mass frame: pE C pE0 D 0 D kE C kE0
• wlog p D .E; E Ó /, p 0 D .E; E Ó / (since me D 0)
E D E 2 m2 (since E D Ee .p/ D Ee .p 0 / D Em .k/ D Em .k 0 /)
• jkj
p
m
• kE Ó D jkj
E cos
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 91
LE C T U R E 14 → PS:125,131–138
" M
et
c-
e- P' (E QQ E)
=
¥i
-
,
's
a.
This leads to the following expressions for the 4-momentum inner products:
q 2 D .p C p 0 /2 D 4E 2 (5.28a)
pp 0 D 2E 2 (5.28b)
pk D p 0 k 0 D E 2 E cos
Ejkj (5.28c)
0 0 E cos
pk D p k D E C Ejkj 2
(5.28d)
!
m2m m2m
21 X 2 4 2
jMj jMj $ e 1C 2 C 1 cos (5.29)
4 0 0 E E2
s;s ;r;r
12 | Differential scattering cross section from Eq. (4.123):
d
1 E
jkj
D jMj2 (5.30)
d cm 2EpE 2EpE0 jvp vp0 j .2/2 4Ecm
s
˛ 2 m2m
m2m
m2m
2
D 2
1 1C 2 C 1 cos (5.31)
4Ecm E2 E E2
It is Ecm D 2E and jvp vp0 j D jp 3 =EpE p 03 =EpE0 j D 2.
13 | Total cross section:
s
4 ˛ 2 m2m m2m
total $ 1 1C (5.32)
2
3Ecm E2 2E 2
„ ƒ‚ …
14 | Discussion:
• For Ecm < 2mm no pair-production is possible.
• Prediction of QED: non-trivial energy dependence of M
Experimental results verify this additional dependence!
(↑ P&S Fig. 5.2 on p. 138 or Ref. [7])
Recall that the energy-dependence of the prefactor was derived on very general
grounds and is not QED-specific!
• Measuring total as a function of Ecm yields the muon mass mm .
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 92
LE C T U R E 14 → PS:125,131–138
5.2 Summary of QED calculations
1| Draw relevant Feynman diagrams.
2| Use Feynman rules to calculate M.
3 | Calculate jMj2 D spins jMj (use spin-sum relations).
2
P
4| Evaluate traces (use trace technology).
5| Fix a frame of reference and express all 4-momenta in terms of kinematic variables
(energies, angles …).
6| Plug in jMj2 in Eq. (4.121) and integrate over phase-space variables that are not
measured.
Following this procedure, one can evaluate cross sections for many other QED processes (like
Compton scattering) and compare them with measurements from particle colliders (→ P&S
pp. 139–169). We will not dwell on these often very technical calculations but proceed with a
more interesting question: What happens if we go beyond tree-level diagrams?
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 93
LE C T U R E 15 → PS:175–186
→ Topics of Lecture 15
1. Overview of radiative corrections in QED
2. Soft bremsstrahlung
3. Formal structure of the electron vertex function
→ Topics of Problemset 8
1. Rutherford scattering
2. Scattering cross-section for electron-positron scattering in QED
6 Radiative Corrections of QED
6.1 Overview
1| Process: For simplicity (→ below), ^ e scattering of a very heavy particle, e.g.,
lim f Electron (e ) C Muon ( ) ! Electron (e ) C Muon ( ) g (6.1)
m !1
2| Tree-level:
The computation runs along the same lines as for e e ! e e scattering.
In the following, however, we do not need the tree-level result.
-
-
e
µ
¥
•¥Fn
'
A 13
new.
←
A P2
M
-
Alternatively, crossing symmetry relates the process to e C e ! C and allows us to
reuse the results we obtained for finite electron mass (→ Problemset 8) with suitable
substitutions.
3| Radiative corrections =
Higher-order contributions to tree-level amplitudes from diagrams with …
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 94
LE C T U R E 15 → PS:175–186
• loops:
The 6 additional one-loop diagrams involving the heavy particle can be neglected as
these include propagators of the heavy particle that vanish for m ! 1.
Physically, the heavy particle does not accelerate much upon absorption/emission
of a photon but behaves like a “static wall”.
(a) Vertex correction: UV-divergence & IR-divergence
(most interesting, → below)
UV-divergence: divergence for k ! 1 in integral of loop momentum
IR-divergence: divergence for k ! 0 in integral of loop momentum
(The vertex correction yields the anomalous magnetic moment of the electron.)
(b) External leg corrections: UV-divergence & IR-divergence
(not amputated, → later)
(c) Vacuum polarization: UV-divergence
(complicated evaluation, → later)
• extra final-state photons (⁂ Bremsstrahlung):
! IR-divergence for k ! 0
In this limit, photons cannot be measured by detectors, so we should add these
diagrams to the scattering amplitude.
4| Spoilers:
• UV-divergences: cancel in observable quantitites
• IR-divergences: cancel with the divergences of the bremsstrahlung diagrams
That is, radiative corrections are only consistent if both types of corrections (loops
and bremsstrahlung) are included.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 95
LE C T U R E 15 → PS:175–186
6.2 Soft Bremsstrahlung
1| ⁂ Bremsstrahlung = Electromagnetic radiation emitted by decelerated, charged particles
⁂ Soft = Low-energy photons (k 0)
2| Can be classically derived from Maxwell’s equations (↑ P&S pp. 177–182)
3| ^ Corresponding QFT process:
i i
M0 : (unkown) interaction amplitude
This is a 4 4-matrix with spinor indices and (potentially) multiple 4-vector indices.
!
= k= C m/
i.p
8 9
0 >
ˆ
ˆ M 0 .p ; p k/ >
.p k/2 m2 C i "
< =
0
i M D i e .k/u.p / u.p/ (6.2)
i.p=0 C k= C m/
:C 0
ˆ >
ˆ M0 .p C k; p/ >
;
.p 0 C k/2 m2 C i "
4| Simplifications:
• Use p 2 D m2 and k 2 D 0:
.p k/2 m2 D 2pk (6.3)
0 2 2 0
.p C k/ m D 2p k (6.4)
E jpE0 pj
• Soft photons: jkj E
! M0 .p ; p k/ M0 .p 0 C k; p/ M0 .p 0 ; p/ (cross ks in amplitudes)
0
!p = k= p = etc. (cross k= s in numerators of propagators)
• Dirac algebra !
= C m/
.p u.p/ $ 2p
u.p/ (6.5)
u.p 0 / =0
.p C m/ $ u.p 0 /2p 0
(6.6)
Here we use the Dirac algebra and the spin-completeness relations that imply
= m/u.p/ D 0.
.p
5| Then
0
p
p
i M D u.p 0 /M0 .p 0 ; p/u.p/ e (6.7)
„ ƒ‚ … p0k pk
elastic scattering „ ƒ‚ …
bremsstrahlung
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 96
LE C T U R E 15 → PS:175–186
6| Scattering cross section (cf. Eq. (4.121) for two incoming particles):
d 3k X e 2 ˇˇ p 0 r p r ˇˇ2
Z ˇ ˇ
0 0
d .p ! p C / D d .p ! p / (6.8)
.2/3 r 2jkj
E ˇ p0k pk ˇ
„ ƒ‚ …
d Pk .p!p 0 /
d Pk .p ! p 0 /: differential probability to emit a photon into d 3 k under the condition
that the electron scatters from p to p 0 .
We integrate over the photon momentum and sum over its polarizations because we are
only interested in the probability that an additional photon is emitted.
7| Evaluation:
1
d k X ˇˇ p 0 r p r ˇˇ2
ˇ ˇ
˛ 1
Z Z Z
d Pk D dk (6.9)
0 k 4 r ˇ p 0 kQ p kQ
ˇ
„ ƒ‚ …
I.p;p 0 /
˛
D I.p; p 0 / Πlog.1/ log.0/ (6.10)
„ ƒ‚ … „ƒ‚…
Problem 1 Problem 2
with kQ D k=jkj
E D .1; k/
O
8| Approximations:
i| Problem 1: Soft-photon approximation breaks down at k jE
q j D jpE pE0 j
! Introduce upper cutoff at jE
qj
ii | Problem 2: Probability of radiating a very soft photon is infinite!
! IR-divergences of perturbative QED
(Note that in the limit k ! 0 our soft-photon approximation is exact!)
Solution: Regularization with finite photon mass > 0:
1 1 1
D 7! p (6.11)
k EkE C k2
2
This is a purely mathematical ad-hoc solution to control the IR-divergence. Later
we will see that in physical observables the unphysical parameter drops out.
and therefore
Z jEq j p !
1 2 C jEq j2 C jE
qj
dk p D log (6.12)
0 2 C k 2
2
jE
qj jE
qj 1 jE
qj
log 2 log D log (6.13)
2 2
(asymptotically for ! 0)
iii | Relativistic limit (Ep;p0 m):
2
0 q
I.p; p / 2 log with q2 D .p 0 p/2 0 (6.14)
m2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 97
LE C T U R E 15 → PS:175–186
Proof: ↑ P&S pp. 180–182, starting at Eq. (6.12)
Recall that for two time-like momentum vectors p and p 0 , p 2 D p 02 D m2 , their
difference q D p 0 p is space-like, q 2 0 (use Cauchy-Schwarz inequality to
show this). Therefore there always exists a coordinate system with q 0 D 0, or,
equivalently, p 0 D E D p 00 . In this system, it is q 2 D jE
q j2 .
9| Result:
q2
2
0 ˛ 0 q
d .p ! p C / d .p ! p / log 2
log (6.15)
m2
„ ƒ‚ …
⁂ Sudakov double logarithm
for ! 0 (regularization) and Ep;p0 m or q 2 D jE
q j2 ! 1 (relativistic limit)
10 | Two problems:
• Dependence of unphysical photon mass
Should drop out from physical predictions!
• Logarithmic divergence for q 2 ! 1 (! cannot be interpreted as probability)
We will see that the correct interpretation is that of the emitted number of photons.
6.3 The Electron Vertex Function
6.3.1 Formal Structure
1| Scattering amplitude:
iM.e .p/ .k/ ! e .p 0 / .k 0 // (6.16)
D (6.17)
D (6.18)
1
D i e 2 ue .p 0 / .p 0 ; p/ue .p/ 2 um .k 0 /
um .k/ (6.19)
q
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 98
LE C T U R E 15 → PS:175–186
Note that we consider only amputated diagrams (1) without loops connecting to the heavy
particle and (2) ignore the vacuum polarization diagrams as these describe corrections to
the photon propagator and are not related to the interaction between fermions and gauge
field.
Below we will explicitly evaluate the first loop correction (yellow).
2| General form:
.p 0 ; p/ D f .p ; p 0 ;
; m; e; C/ (6.20)
5 is forbidden since QED does not violate parity symmetry (recall that . 5 /2 D 1 and
5 produces a pseudo vector and 5 a pseudo scalar)!
3| Restrictions:
All equations that follow are required to hold if sandwiched between bispinors u and u!
i| Lorentz covariance: transforms like !
D A
C BQ p C CQ p 0 (6.21)
0 0
DA C B .p C p / C C .p p / (6.22)
must be a linear function of the available Lorentz vectors.
ii | Recall p
= u.p/ D m u.p/ and u.p 0 /p=0 D u.p 0 / m !
X D X.p ; p 0 ; m; e; C/ 1 for X D A; B; C (6.23)
Use the spin-sum identities Eq. (3.29) to show this.
iii | q 2 D .p 0 p/2 D 2.m2 pp 0 / only non-trivial scalar !
X D X.q 2 ; m; e; C/ for X D A; B; C (6.24)
Recall that p 2 D p 02 D m2 are constants.
iv | Ward identity for U.1/ gauge-symmetry of QED Lagrangian:
q .p 0 ; p/ D 0 (6.25)
↑ P&S pp. 238–244 for a proof and pp. 159–161 for a motivation
This is the quantum version of the classically conserved current @ j .x/ D 0 in
Fourier space.
Ward identities = QFT analog of Noether’s theorem
!
0 D q D A q CB q .p 0 C p / CC q 2 (6.26)
„ƒ‚… „ ƒ‚ …
D0 D0
!C D0
The first term vanishes only if sandwiched between bispinors,
u.p 0 /.p=0 = /u.p/ D .m
p m/u.p 0 /u.p/ D 0 ; (6.27)
and the second vanishes identically since p 2 D p 02 D m2 .
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 99
LE C T U R E 15 → PS:175–186
4| Gordon identity:
p 0 C p i q
0
u.p / u.p/ $ u.p 0 /
u.p/ u.p /0
u.p/ (6.28)
2m 2m
Absorb the first term in A.
Recall that D 2i Π; produces a second-rank Lorentz tensor.
5| Therefore
i q
.p 0 ; p/ D
F1 .q 2 / C F2 .q 2 / D
C O.˛/ (6.29)
„ ƒ‚ … 2m „ ƒ‚ …
1CO.˛/ 0CO.˛/
Fi .q 2 /: ⁂ Form factors
Note that we can use the Gordon identity wlog because the vertex amplitude is always
sandwiched between bispinors u and u.
Expanding in orders of ˛ D e 2 =4, the lowest order term must give back the tree-level
vertex D . Therefore F1 D 1 C O.˛/ but F2 D 0 C O.˛/.
Note that the form of the vertex function was derived on very general grounds and holds
for any fermion coupling to the electromagnetic field. Measuring scattering cross sections
can be used to experimentally determine the two form factors—even in situations where
ab initio computations are hard (→ Problemset 9).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 100
LE C T U R E 16 → PS:186–196
→ Topics of Lecture 16
1. The Landé g-factor of the electron
2. Evaluation of the vertex correction
6.3.2 The Landé g-factor
Observation: F1 and F2 encode the electric and magnetic response of the electron completely.
Goal: Express electric charge and magnetic moment as function of form factors.
1| Setting: ^ Classical, external field Acl
.x/: (→ Problemset 8)
Z
Hint D e d 3x ‰.x/
‰.x/ Acl
.x/ (6.30)
i M .2/ı.p 00 p0/ D (6.31)
D i eu.p 0 / .p 0 ; p/u.p/ Acl
.q D p
0
p/ (6.32)
Note that Acl
.x/ is a parameter and not an operator; in particular, it has no dynamics!
In general, a static potential breaks translational invariance and therefore 3-momentum is
no longer conserved. However, as it is static, it does not break time translation invariance,
so that energy is still conserved, i.e. p 00 D p 0 . This is like a hard wall in mechanics that
can absorb momentum but not energy.
2| Electric charge:
i| ^ Acl
.x/ D . .x/;
E 0/E ) Acl 0
.q/ D ..2/ı.q / .E
E
q /; 0/
ii | iM D i eu.p 0 / 0 .p 0 ; p/u.p/ .E
q/
iii | E slowly varying ! .E
^ .x/ q / concentrated at qE D 0 ! take limit qE ! 0:
E 2 m2
jpj
iM i eF1 .0/ u.p 0 / 0
u.p/ .E
q/ q / 2m 0
i eF1 .0/ .E (6.33)
Recall Eq. (4.199) for the non-relativistic limit of bispinors.
iv | ! Born approximation with potential
E D eF1 .0/ .x/
V .x/ E (6.34)
Recall Eq. (4.201) for the Born approximation.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 101
LE C T U R E 16 → PS:186–196
Š .0/
v| Charge e D eF1 .0/ and F1 D1!
.n/
F1 .0/ D 0 for n1 (6.35)
P1 .n/ n
It is F1 D nD0 F1 ˛ with ˛ the fine-structure constant.
3| Magnetic moment:
i| ^ Acl E E
.x/ D .0; A.x// ) Acl 0 E q //
.q/ D .0; .2/ı.q /A.E
ii | Then
i q
iMD i eu.p /0 2
F1 .q / C F2 .q / u.p/ Acl
2
.E
q/ (6.36)
2m
" #
i i q
D Ci eu.p 0 / i 2
F1 .q / C 2
F2 .q / u.p/ Aicl .E
q/ (6.37)
2m
„ ƒ‚ …
Vanishes for q D 0 and jpj
E 2 m2 , see Eq. (4.199)
Note that qE D 0 , q D 0.
iii | ^ F1 -term and expand bispinors in linear order of p,
E pE0 :
0
0 i ı 0 p E E i ipEE
u.p / u.p/ 2m C (6.38)
2m 2m
p i C p 0i
0 0 i ij k j k
$ 2m C 2m " q (6.39)
„ 2mƒ‚ … „
2m
ƒ‚ …
(A) (B)
In the second step, we used i j D ı ij C i "ij k k .
Only (B) is spin-dependent and affects the magnetic moment!
Term (A) describes the kinetic energy pEAE C AEpE of a charged particle in a magnetic
field in nonrelativistic quantum mechanics.
iv | ^ F2 -term and expand bispinors in lowest order of p, E pE0 :
i q ı i ij k j k
u.p 0 / i u.p/ 2m 0 " q (6.40)
2m 2m
p
Use u.p/ m [← Eq. (4.199)], Œ i ; j D 2i "ij k k and qi D qi .
v| In summary:
" #
0 i i i q
2
i MD i eu.p / F1 .q / C F2 .q / u.p/ Aicl .E
2
q/ (6.41)
2m
h
q!0 0 1 k i
ie ŒF1 .0/ C F2 .0/ i "ij k q j Aicl .E
q / .2m/ (6.42)
2m „ ƒ‚ …
DBclk .E
q/
with BEcl D r AEcl (B k D "ij k @i Aj ) B k .E
q / D i "ij k q i Aj .E
q /)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 02
LE C T U R E 16 → PS:186–196
vi | ! Born approximation with potential
E D
V .x/ E BEcl .x/
hi E (6.43)
yields the magnetic moment
e k
hi
E D ŒF1 .0/ C F2 .0/ 0 g B hSE i (6.44)
m 2
e
with Bohr magneton B D 2m and Landé factor
g D 2 ŒF1 .0/ C F2 .0/ D 2 C 2F2 .0/ (6.45)
.1/
D 2
„ƒ‚… C 2˛F2 .0/ C O.˛ 2 / (6.46)
Dirac equation
„ ƒ‚ …
Anomalous magnetic moment
.1/
Here we use F1 .0/ D 1 in all orders of ˛ and that F2 D ˛ F2 C O.˛ 2 /.
.1/
This result motivates our subsequent evaluation of the first loop correction F2 !
6.3.3 Evaluation
The techniques that we use below can be applied to the evaluation of all loop diagrams in QED.
1| Scattering amplitude:
u.p 0 /Œ˛ .1/ .p 0 ; p/ u.p/ (6.47)
D (6.48)
d 4k ig i.k=0 C m/ i.k= C m/
Z
D 4 2
u.p 0 /. i e
/
. ie
/u.p/
.2/ qQ C i " k 02 m2 C i " k2 m2 C i "
(6.49)
Contraction identities: D 2 etc.
0
0 2
d 4k u.p / k= k= C m 2m.k C k 0 / u.p/
Z
2
$2ie (6.50)
.2/4 .qQ 2 C i "/.k 02 m2 C i "/.k 2 m2 C i "/
In the following, the regularizations i " will be crucial to make the expressions well-
defined!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 103
LE C T U R E 16 → PS:186–196
2| Feynman Parameters:
Goal: Introduce new integration variables to combine the three factors in the denominator
so that we can solve the integral by completing the square.
n Z
! !
1
1 Y X .n 1/Š
D dxi ı xi 1 (6.51)
A1 : : : An Œx1 A1 C C xn An n
i D1 0 i
xi : Feynman parameters
(Proof: → Problemset 9)
3| Application to denominator of Eq. (6.50):
1
1 2
Z
D dx dy d´ ı.x C y C ´ 1/
.qQ 2 C i "/.k 02 m C i "/.k 2
2 m2 C i "/ 0 D3
(6.52)
with (using x C y C ´ D 1 and qQ D p k and k 0 D k C q)
D $ k 2 C 2k.yq ´p/ C yq 2 C ´p 2 .x C y/m2 C i " (6.53)
Complete the square: l k C yq ´p
D l2 C i" (6.54)
where xyq 2 C .1 ´/2 m2 > 0 (“effective mass squared”) since q 2 < 0 (always
spacelike)
4| Express the numerator of Eq. (6.50) in terms of l (k D l yq C ´p ):
u.p 0 / k= k=0 C m2 2m.k C k 0 / u.p/
(6.55)
This step is only valid under the integral d 4l (→ notes below)!
R
< 1 l 2 C Œ y q C ´p Œ.1 y/q C ´p =
8 9
ı = = = =
u.p 0 / 2 u.p/ (6.56)
Cm2 2mŒ.1 2y/q C 2´p
: ;
For this step, you have to use the Dirac algebra (→ notes below).
8 9
1 2 2 2 2
ˆ
ˆ l C .1 x/.1 y/q C .1 2´ ´ /m > >
2
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ „ ƒ‚ … >
>
A
ˆ
ˆ >
>
< =
0 0
$ u.p / C.p C p/ Œm´.´ 1/ u.p/
ˆ
ˆ „ ƒ‚ … >
>
ˆ
ˆ
ˆ B >
>
>
ˆ >
ˆ
ˆ
ˆ
ˆ Cq Œm.´ 2/.x y/ >
>
>
>
: „ ƒ‚ … ;
C
(6.57)
This structure was expected, recall Eq. (6.22).
R d 4l l
• The first step follows with .2/ 4 D.l 2 / D 0 due to symmetry and
d 4l l l d 4l g l 2
Z Z
L D : (6.58)
.2/4 D.l 2 / .2/4 4 D.l 2 /
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 04
LE C T U R E 16 → PS:186–196
This identity can be shown as follows: First, notice that under a Lorentz transfor-
mation ƒ 2 SOC .1; 3/
d 4l l 0 l 0 d 4l 0 l 0 l 0
Z Z
0
L D D D L (6.59)
.2/4 D.l 2 / .2/4 D.l 02 /
(in the second step we used that det.ƒ/ D 1 and l 2 is a scalar) and therefore
L D g C.l 2 / (6.60)
which follows from Schur’s lemma [8] and the observation that the only scalar
available is l 2 . Finally, C.l 2 / can be determined by contracting with g :
d 4l l2
Z
g
) g L D D 4C.l 2 / (6.61)
.2/4 D.l 2 /
• For the second step, use
– p
= D 2p p
=
– p
= u.p/ D mu.p/ and u.p 0 /p=0 D mu.p 0 / and therefore u.p 0 /q=u.p/ D 0
– xCyC´D1
5| C is antisymmetric and D is symmetric under x $ y ! drop C
R1 R1
Formally: 0 dx 0 dy C.x; y/=D.x; y/ D 0
This result was expected from the Ward identity!
6| Gordon identity Eq. (6.28) !
u.p 0 /Œ˛ .1/ .p 0 ; p/ u.p/ (6.62)
Z 1 Z 1 Z 1
d 4l 2
Z
D 2i e 2 dx dy d´ ı.x C y C ´ 1/ (6.63)
.2/4 0 0 0 D.l 2 /3
8 9
1 2 2 2 2 >
l C .1 x/.1 y/q C .1 4´ C ´ /m >
ˆ
ˆ
2
< =
0
u.p / q
u.p/
ˆ
:C i 2 >
2m ´.1 ´/
ˆ >
;
2m
Note that the Gordon identity contributes also to the -term, thus the modifications in
the last term proportional to m2 .
7| Momentum integral:
i| Problem:
l 2 D .l 0 /2 lE2 cannot be integrated in four-dimensional spherical coordinates.
Solution:
Wick rotation = Evaluation of a contour integral (blue) along a rotated contour
(green) that encircles the same poles (red):
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 05
LE C T U R E 16 → PS:186–196
Note that this requires the integrand to vanish faster than 1=jl 0 j so that the
contribution of the half-circle vanishes for R ! 1.
Parametrization of the new contour:
l 0 i lE
0
and lE lEE with lE 2 R 4 (6.64)
) l D 2 0 2
.lE / lEE
2
D 2
lE (6.65)
Here, l 2 is the squared “norm” of a four-dimensional vector in the Minkowski
metric and lE2
in the norm squared in the Euclidean metric.
ii | Then (m > 2) (we are interested in m D 3)
d 4l 1 i 1 1
Z Z
lim 4 2 m
D m 4
d 4lE 2 (6.66)
"!0 .2/ .l C i "/ . 1/ .2/ .lE C /m
Z 1 3
i. 1/m lE
Z
D d 4 d l E 2
(6.67)
.2/4 0 .lE C /m
„ ƒ‚ … „ ƒ‚ …
D2 2 D 1
2.m 1/.m 2/m 2
i. 1/m 1
D (6.68)
.4/2 .m 1/.m 2/m 2
d 4 D 2 2 is the surface area of the unit sphere in four dimensions.
R
and similarly (m > 3)
d 4l l2 i. 1/m 1 2
Z
lim $
"!0 .2/4 .l 2 C i "/m .4/2 .m 1/.m 2/.m 3/m 3
(6.69)
Problem: For m D 3 the integral diverges!
This is a UV-divergence since it occurs for lE
2
lE2 kE2 ! 1.
Note that in this case also the contribution of the half-circle does not vanish and
the Wick rotation is not justified.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 106
LE C T U R E 16 → PS:186–196
iii | Fix: Pauli-Villars regularization:
ig ig ig
7! 2 (6.70)
qQ 2 C i" qQ C i " qQ 2 ƒ2 C i "
for large ƒ (= additional, heavy photon with mass ƒ)
For ƒ ! 1 we obtain the original expression. The regularization essentially
introduces a UV-cutoff at momenta k & ƒ where the difference is suppressed.
Hope: ƒ does not appear in physical predictions
ı
! Only change:
ƒ D xyq 2 C .1 ´/2 m2 C´ƒ2 (6.71)
iv | Therefore (m D 3):
• Eq. (6.68) 7! Eq. (6.68) O.ƒ1 / D Eq. (6.68) O.ƒ 2 / Eq: (6.68)
Drop contribution to the convergent integral since ƒ 2 ! 0 for ƒ ! 1.
• Eq. (6.69) 7!
d 4l l2 l2
Z
lim (6.72)
"!0 .2/4 .l 2 C i "/3 .l 2 ƒ C i "/3
Wick rotation
" #
Z 1 5 5
i 2lE 2lE
D d lE 2 2
(6.73)
.4/2 0 .lE C /3 .lE C ƒ /3
2
i ƒ ƒ!1 i ´ƒ
$ 2
log ! 2
log (6.74)
.4/ .4/
Details: → Problemset 10
8| Result: (with ƒ ´ƒ2 for ƒ ! 1)
u.p 0 /Œ˛ .1/ .p 0 ; p/ u.p/ (6.75)
Z 1 Z 1 Z 1
˛
$ dx dy d´ ı.x C y C ´ 1/
2 0 0 0
2
.1 x/.1 y/q 2 4´ C ´2 /m2
8 9
´ƒ .1
ˆ
ˆ log C C >
>
ˆ
ˆ >
>
ˆ
ˆ >
ˆ
ˆ „ ƒ‚ …>
>
>
.1/
ˆ >
< ¶F1 .q 2 / =
u.p 0 / u.p/
i q 2m2 ´.1 ´/
ˆ >
C
ˆ
ˆ >
>
2m
ˆ
ˆ >
>
ˆ
ˆ >
>
ˆ
ˆ „ ƒ‚ … >
>
.1/
: ;
¶F2 .q 2 /
(6.76)
The “¶” signifies that the integrals over Feynman parameters and the prefactor belong
to the form factors.
9| Discussion of F1 :
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 07
LE C T U R E 16 → PS:186–196
.1/ .1/
i| Problem 1: It should be F1 .0/ D 0, but here F1 .0/ ¤ 0!
Fix 1:
.1/ .1/ .1/
F1 .q 2 / 7! F1 .q 2 / F1 .0/ (6.77)
We cannot justify this substitution at this point; a rigorous derivation requires the
LSZ reduction formula and is rooted in field strength renormalization (→ later).
The origin of this term can be traced back to our omission of the external leg loop
corrections.
ii | Problem 2: In addition, there is a IR-divergence for qQ 2 ! 0
^ q 2 D 0 for simplicity:
1 1 1
1 4´ C ´2
Z Z Z
dx dy d´ ı.x C y C ´ 1/ (6.78)
0 0 0 .1 ´/2
1 Z 1 ´
2 C .1 ´/.3 ´/
Z
D d´ dy (6.79)
0 0 .1 ´/2
1
2
Z
D d´ Cfinite terms (6.80)
1 ´
„0 ƒ‚ …
1
Fix 2: Add a small photon mass > 0 !
7! D xyq 2 C .1 ´/2 m2 C´2 (6.81)
In this regularization we recover the original result for ! 0.
We will discuss this IR-divergence later.
iii | Fix 1 + Fix 2 !
.1/
F1 .q 2 / D 1 C ˛F1 .q 2 / C O.˛ 2 / (6.82)
with
1 1 1
1
Z Z Z
.1/
F1 .q 2 / D d´ ı.x C y C ´ 1/
dx dy (6.83)
2 0 0 0
´/2 m2 .1
2 3
log
6
6 m2 .1 ´/2 q 2 xy 7
7
6 m2 .1 4´ C ´2 / C q 2 .1 x/.1 y/ 7
6 7
6C
m2 .1 ´/2 q 2 xy C ´2
7
6 7
6 7
4 m2 .1 4´ C ´2 / 5
m2 .1 ´/2 C ´2
• Here we already set D 0 in the logarithm where it is not needed to control
the IR-divergence.
• Note how the unphysical Pauli-Villars regulator ƒ dropped out because of the
subtraction.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 108
LE C T U R E 16 → PS:186–196
10 | Discussion of F2 :
No divergences in F2 ! Yay!
.1/
F2 .q 2 / D ˛F2 .q 2 / C O.˛ 2 / (6.84)
with
1 1 1
1
Z Z Z
.1/
F2 .q 2 / D dx dy d´ ı.x C y C ´ 1/ (6.85)
2 0 0 0
2
2m ´.1 ´/
m2 .1 ´/2 q 2 xy
11 | Landé g-factor:
Z 1 Z 1 Z 1
2 ˛ 2´
F2 .q D 0/D dx dy d´ ı.x C y C ´ 1/ C O.˛ 2 / (6.86)
2 0 0 0 1 ´
Z 1 Z 1 ´
˛ 2´
D d´ dy C O.˛ 2 / (6.87)
2 0 0 1 ´
˛ 2
D C O.˛ / (6.88)
2
Therefore the anomalous magnetic dipole moment of the electron is
g 2 ˛
ae D 0:0011614 (6.89)
2 2
exp
ae 0:0011597 (6.90)
Note that ˛ 2 0:5 10 4 so that the deviation can be explained by higher-order
corrections.
→ Note 6.1
• Our first-order result was obtained by Schwinger in 1948 [9].
˛
The first-order correction 2 is engraved on Schwinger’s tombstone:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 109
LE C T U R E 16 → PS:186–196
• Modern values:
aeSM D 0:001 159 652 182 031.15/.15/.720/ (6.91)
exp
ae D 0:001 159 652 180 73.28/ (6.92)
! Agree to 11 significant digits
This is the most accurate prediction of physics to date!
The experimental value is from Ref. [10] and the theoretical value is from Ref. [11]
(erratum). The theoretical result is based on numerical evaluations of contributions
up to order ˛ 5 . Analytical results are known up to order ˛ 3 [12]. Note that the
theoretical value also includes small contributions beyond QED, namely from the
electroweak and hadronic sector of the standard model. The main contribution
comes from higher-order QED diagrams, though.
• Our first-order result applies also to the muon since the mass cancels:
.1/ ˛
a D D ae.1/ (6.93)
2
However, in higher-order there seem to be discripancies between the standard
model predictions (as for the electron, this goes beyond QED) and measurements:
exp SM 11
a a D 261.63/.48/ 10 (6.94)
For details, ↑ http://pdg.lbl.gov/2019/reviews/rpp2018-rev-g-2-muon-
anom-mag-moment.pdf and references therein.
This deviation may hint at new physics beyond the standard model, for example
contributions from supersymmetric partners.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 110
LE C T U R E 17 → PS:199–208
→ Topics of Lecture 17
1. Infrared divergences
2. Resummation and interpretation of infrared divergences
→ Topics of Problemset 9
1. Rosenbluth formula
2. Feynman parameters
6.3.4 The Infrared Divergence
1| Goal: Understand asymptotics of jF1 .q 2 /j ! 1 for ! 0
2| Show in → Problemset 10:
2 !0 ˛ 2 A
F1 .q / D (6.82) 1 fIR .q / log C O.˛ 2 / (6.95)
2 2
where A 2 f q 2 ; m2 g (both are asymptotically equivalent but, depending on the
additional limit q 2 ! 0=1, one or the other must be chosen) and
1
m2 q 2 =2
Z
fIR .q 2 / D d 1 0 (6.96)
0 m2 q 2 .1 /
Note that q 2 0 and .1 / 1=4 for 0 1.
3| ^ Cross section for electron scattering off a static potential:
2 3
d .pE ! pE0 / d
6 ˛ A 7
61 fIR .q 2 / log CO.˛ 2 7
/ (6.97)
6
d d 2
7
„ ƒ‚ …0
4 5
„ ƒ‚ …
Tree-level result Problem: ! 1 for ! 0
Recall that d / jMj2 Œ 2 ŒF1 .q 2 /2 . Just like e, F1 is a prefactor to so that
e 7! e F1 .q 2 / is enough to obtain the contribution of F1 to the scattering cross section.
The factor 1=2 vanishes because the expression must be squared for the cross section.
The contribution of F2 does not affect the asymptotic behaviour as it is finite for ! 0.
Problem: The negative, diverging O.˛/ contribution to the scattering cross section is
clearly unphysical!
4| ^ Limit q 2 ! 1:
1
q 2 =2 q2
Z
2 ı
fIR .q / d 2 2
log (6.98)
0 q .1 / C m m2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 111
LE C T U R E 17 → PS:199–208
We drop the constant 1 and the mass m2 in the numerator for the asymptotic behaviour.
!
2 2
!0 ˛ q q
F1 . q 2 ! 1/ 1 log 2
log CO.˛ 2 / (6.99)
2 m 2
„ ƒ‚ …
Sudakov double logarithm
Here we have to use A D q 2 and not A D m2 since q 2 ! 1.
5| Comparison with bremsstrahlung Eq. (6.15) for q 2 ! 1:
d .pE ! pE0 / !0 d
2 2
˛ q q 2
1 log log C O.˛ / (6.100)
d d 0 m2 2
d.pE ! pE0 C / !0 d
2 2
˛ q q 2
C log log C O.˛ / (6.101)
d d 0 m2 2
! Both are divergent but their sum is finite and independent of !
6| Suggested solution:
Photon detectors cannot detect photons below a lower threshold Emin !
d d .pE ! pE0 / d .pE ! pE0 C .k < Emin //
D C (6.102)
d measured d d
To show: The cancellation does not only occur for q 2 ! 1 but for arbitrary q.
7| For general q:
˛ A
2 3
2
1 f IR .q / log
6„
6 2 7
7
6 ƒ‚ … 7
d
d
6 Elastic scattering 7
!0
(6.103)
6 ! 7
2
d d 0 ˛ E
6 7
measured 0 min 2 7
log
6
6 C I.p; p / CO.˛ /7
6
4 2 2 7
5
„ ƒ‚ …
Bremsstrahlung
with I.p; p 0 / defined in Eq. (6.9) as
d k X ˇˇ p 0 r p r ˇˇ2
Z ˇ ˇ
0
I.p; p / D (6.104)
4 r ˇ p 0 kQ p kQ
ˇ
Recall that after introducing the small photon mass , we found for the Bremsstrahlung
cross section with Eq. (6.9) and Eq. (6.13)
2
0 0 ˛ 0 jE
qj
d .p ! p C / D d .p ! p / I.p; p / log (6.105)
2 2
where we introduced the upper cutoff jE q j because there the soft-photon approximation
breaks down and invalidates the result.
Here we replace this upper bound by the physically motivated cutoff Emin < jEq j and find
!
2
0 0 ˛ 0 Emin
d .p ! p C / D d .p ! p / I.p; p / log (6.106)
2 2
(Use colors to skip the second equation.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 112
LE C T U R E 17 → PS:199–208
8| Show (using a Feynman parameter)
I.p; p 0 / D 2fIR .q 2 / for all p; p 0 (6.107)
Proof: ↑ P&S p. 201, starting at Eq. (6.69); see also P&S pp. 180–181 Eqs. (6.12)-(6.15)
9| Then
d
(6.108)
d measured
" ! #
d
!0 ˛ A
1 fIR .q 2 / log C O.˛ / 2
(6.109)
d 0 2
Emin
2 3
!
d q2 q2
6 7
q 2 m2 ˛ 2 7
log log
6 7
61 CO.˛ / (6.110)
d 6
04 m2 2
Emin 7
5
„ ƒ‚ …
Correction by Sudakov double logarithm
! Independent of but dependent on experimental conditions (Emin ) (which is fine)
We did not evaluate the exact dependence on q (since A 2 f q 2 ; m2 g) but for q 2 m2
(or q 2 ! 1) the result is correct.
This is an example how an unphysical regularization parameter does not affect measurable
results.
6.3.5 Summation and Interpretation of Infrared Divergences
1| Problems:
• Did not show the cancellation of the IR divergences for higher orders
• Cross section Eq. (6.110) becomes negative (and therefore unphysical) for Emin ! 0
The solution of the second problem will follow from the solution of the first one.
The following discussion is only a sketch and not mathematically rigorous as it skips
several technical details that are beyond the scope of this course (and P&S).
2| Notation:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 113
LE C T U R E 17 → PS:199–208
• Real photons (with arrow) are on-shell, transversely polarized and are connected
with only one end to the Feynman diagram
• Virtual photons (without arrow) can be off-shell, longitudinally polarized and are
connected with both ends to the Feynman diagram
• The momentum of soft photons (red) is upper bounded:
E < Emin (real)
k 2 < E 2 (virtual) and jkj
E min
• The momentum of hard photons (blue) is lower bounded:
E > Emin (real)
k 2 > E 2 (virtual) and jkj
E min
The subscript E denotes norms in the Euclidean norm after Wick rotation.
Virtual photons are not physical and can never be measured. Real photons can only be
measured if they are hard. Soft, real photons cannot be measured due to finite detector
sensitivity.
3| Origin of IR divergences:
! No IR divergence
! IR divergence (yellow)
Soft (real or virtual) photons on the legs of scattering vertices with hard photons lead to
IR divergences via singular (i.e. on-shell) fermion propagators.
4| ^ Generic process:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 14
LE C T U R E 17 → PS:199–208
Todo: Sum all such diagrams!
5| ^ Outgoing leg:
Here we do not care whether the soft photons are real or virtual, and, if they are virtual,
whether they connect to each other (and form a leg correction) or to the incoming leg
(forming a vertex correction).
i| Feynman rules !
i.p=0 C k= 1 C m/ i.p=0 C k= 1 C k= 2 C m/
u.p 0 /. i e 1
/ . i e 2
/ (6.111)
2p 0 k1 C O.k 2 / 2p 0 .k1 C k2 / C O.k 2 /
0
i.p= C k= 1 C C k= n C m/
. i e n / 0 .i Mhard / : : : (6.112)
2p .k1 C C kn / C O.k 2 /
Note that ki2 D 0 is only true for real (on-shell) photons. Since we do not specify at
this point, whether we interpret the soft photons ki as real or virtual, we cannot,
strictly speaking, set ki2 D 0 in the denominators (the terms ki kj may even be
non-zero for real photons). However, in the soft-photon approximation, we drop
the O.k 2 / terms anyway and their presence is irrelevant in the end.
ii | Soft-photon approximation (ki ! 0)
• Drop non-singular terms k= i in the numerators
• Drop O.k 2 / terms in the denominators
• Use repeatedly .p
=0 C m/ D . p=0 C m/ C 2p 0 (Dirac algebra)
• Use repeatedly u.p 0 /. p=0 C m/ D 0 (recall the spin sums Eq. (3.29))
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 115
LE C T U R E 17 → PS:199–208
ı
!
p 01 p 02 p 0n
0
u.p / e 0 e 0 e 0 (6.113)
p k1 p .k1 C k2 / p .k1 C C kn /
„ ƒ‚ …
D
iii | Sum over all orderings of k1 ; : : : ; kn :
X
.ki 7! k.i/ / D ‹ (6.114)
2Sn
Sn denotes the permutation group of n elements and is a particular permutation.
If we connect two photons ki and kj to form a virtual photon (see below), the
permutation ki $ kj would not be necessary so that we overcount the weight of
the diagram by a factor of 2 (recall that virtual photon lines are unoriented). To
compensate for that, we will multiply by 12 when we form a photon loop (→ below).
iv | Use
n
X 1 1 1 Y 1
$
p k.1/ p .k.1/ C k.2/ / p .k.1/ C C k.n/ / p ki
2Sn i D1
(6.115)
Proof by induction over n.
v| Then
Y p 0i
0
u.p / e 0 (6.116)
p ki
i
The “” hints at the soft-photon approximation.
6| ^ Incoming leg:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 116
LE C T U R E 17 → PS:199–208
! The same arguments yield
Y p i
e u.p/ (6.117)
p ki
i
Only difference: One additional minus sign per photon since ki 7! ki .
7| ^ Sum over n soft photons attached either to the incoming or the outgoing leg:
Eq. (6.116) & Eq. (6.117) !
Y p 0i p i
0
u.p /i Mhard u.p/ e (6.118)
p 0 ki p ki
i
This is a process that involves only bremsstrahlung and no vertex loops.
8| Virtual photon between vertex i and j :
i| Set kj D ki k
ig
ii | Multiply by photon propagator k 2 Ci"
iii | Integrate over k
1
iv | Multiply by 2 to account for the symmetry ki $ kj (← note above)
!
0
e2 d 4k p 0 p p
i p
Z
X (6.119)
2 4 2
.2/ k C i " p0 k pk p0 k pk
This prescription allows us to convert two real, soft photons into a virtual soft photon
which is a loop correction of either the vertex or one of the two legs.
To evaluate this integral by contour integration, as before, a regularization by introducing
a small photon mass > 0 is needed to control the IR divergence.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 117
LE C T U R E 17 → PS:199–208
9| Evaluation of X: ^ Special case with one virtual photon:
(6.120)
u.p 0 /i Mhard u.p/ X (6.121)
q2
Š ˛
D u.p 0 /i Mhard u.p/ fIR .q 2 / log (6.122)
2 2
„ ƒ‚ …
.1/
Known IR limit of F1 .q 2 /, see Eq. (6.95)
.1/ .1/
Note that the known IR limit of F1 .q 2 / followed after ad-hoc subtraction of F1 .0/.
This is related to the fact that in X we also sum over the leg corrections which we ignored
in our original discussion of the form factors. The details are quite technical and beyond
the scope of this course.
!
2
˛ 2 q
XD fIR .q / log (6.123)
2 2
This result can also be obtained by direct evaluation of the integral Eq. (6.119).
10 | ^ Sum of arbitrary many soft, virtual photons:
Eq. (6.118) & Eq. (6.119) !
1
0
X Xm
u.p /i Mhard u.p/ (6.124)
mŠ
mD0
D u.p 0 /i Mhard u.p/ exp.X/ (6.125)
1
The factors mŠ compensate for overcounting since the order of virtual photons does not
matter and gives rise to equivalent diagrams.
As for the factor of 12 , this is a consequence of our “symmetrization” above. For instance,
symmetrization over n D 4 photons includes the (distinct) summands
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 118
LE C T U R E 17 → PS:199–208
which become identical after connecting pairs to virtual photons:
1
Here, m D 2 and 2Š would cancel the factor of 2.
11 | ^ Emission of a real photon ki D k:
i| Multiply by polarization vector Œ
r .k/ (external outgoing photon)
ii | Square the amplitude
iii | Phase-space integration of photon momentum kE (with upper cutoff jkj
E < Emin )
iv | Sum photon polarizations r D 1; 2
!
Emin
d 3k 1 X 2 ˇˇ p 0 r p r ˇˇ2
Z ˇ ˇ
e ˇ 0 Y (6.126)
.2/3 2k r p k pk ˇ
Recall the discussion of bremsstrahlung, i.e. Eq. (6.9) and Eq. (6.13)
˛ 0 Emin
I.p; p / log (6.127)
!
2
(6.107) ˛ 2 Emin
D fIR .q / log (6.128)
2
Note that the complex conjugate vanishes because of the absolute value after squaring.
We ignore here the amplitude of the hard process.
12 | ^ Cross section for emission of arbitrary number of soft photons:
1 1
X d d X 1 n
.pE 7! pE0 C n / D .pE 7! pE0 / Y (6.129)
d d nŠ
nD0 „ ƒ‚ … nD0
/ ju.p 0 /iMhard u.p/j2
d
D .pE 7! pE0 / exp.Y / (6.130)
d
1
The prefactor nŠ is needed since the outgoing photons are indistinguishable bosons, i.e.,
whether a photon originates from vertex i or any other outgoing vertex does not change
the physical state. Since we treated the vertices as distinct when symmetrizing, we have
1
to compensate for that by nŠ .
13 | ! Measured cross section for process
e .p/ E < Emin /
E ! e .pE0 / C .Any number of photons with jkj (6.131)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 119
LE C T U R E 17 → PS:199–208
to all orders in ˛ is
d (6.125) & (6.130) d
exp.2X C Y / (6.132)
d measured d 0
" !#
(6.123) & (6.128) d 2
˛ q
D exp fIR .q 2 / log (6.133)
d 0 2
Emin
„ ƒ‚ …
0 and 1 (for q 2 > Emin )
" !#
d 2 q2
q 2 m2 ˛ q
exp log log (6.134)
d 0 m2 2
Emin
„ ƒ‚ …
Sudakov form factor
The exponent 2X follows because we have to square the amplitude Eq. (6.125).
This cross section describes the combination of an arbitrary number of soft virtual
photons with an arbitrary number of soft real photons.
→ Note 6.2
• Sudakov form factor = Probability of a hard process to not emit hard photons
• As the result is independent of , it demonstrates the cancellation of IR
divergences in all orders of ˛.
• We can recover our previous result Eq. (6.110) by expanding the exponential.
However, for Emin ! 0 the exponent becomes large (and negative) so that
this expansion is no longer valid. This explains our earlier, unphysical result
of purportedly negative scattering cross sections. That is, by lowering our
detector sensitivty, higher-order corrections become more and more important
to explain the observed cross sections.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 120
LE C T U R E 18 → PS:211–222
→ Topics of Lecture 18
1. General structure of two-point correlators in interacting theories
2. Källén-Lehmann spectral representation
3. Field-strength renormalization and the electron self-energy
6.4 Field-Strength Renormalization
So far, we glossed over radiative corrections to external legs twice:
• When evaluating S-matrix elements perturbatively, we considered only amputated
Feynman diagrams. We identified the diverging contributions from loops attached to legs
as modification of the propagation of particles in an interacting theory (which are not
part of the scattering process itself ).
• When discussing the scattering of an electron from a heavy target, we ignored the two
diagrams with leg corrections, postponing their treatment to the future.
The future has come:
6.4.1 Structure of Two-Point Correlators in Interacting Theories
• Before we discuss the modification of the electron propagator due to radiative corrections
in QED, we first study the general structure of two-point correlators in interacting field
theories.
• Note that the results of this discussion are exact and not built on perturbation theory.
For now: ^ 4 -theory (later: QED)
1| Goal: Study structure of hjT .x/ .y/ji in an interacting theory
2| Interpretation for free theory:
h0jT .x/ .y/j0i = Amplitude of particle to propagate from y to x (for x 0 > y 0 )
! Effect of interactions?
3| Mathematical preliminaries:
i| ^ Hilbert space of interacting theory Hint
ii | Basis of Hint :
ŒH; PE D 0 ! jpE i eigenstates with energy EpE ./ and momentum pE
Here, H is the interacting Hamiltonian and the states jpE i can contain an arbitrary
number of excitations. Note that we will refer to the vacuum state still as ji.
iii | ^ Boost ƒpE 2 SOC .1; 3/ such that
EpE ./ m
q
ƒpE 1 D E with EpE ./ E 2 C m2
jpj (6.135)
pE 0
Here we assume that a rest frame exists, i.e., that the state has a mass gap.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 121
LE C T U R E 18 → PS:211–222
! 8 jpE i 9 ƒpE 9 j0 i W j0 i D U.ƒpE 1 /jpE i with
H j0 i D m j0 i and PE j0 i D 0 (6.136)
H jpE i D EpE ./jpE i and PE jpE i D pj
E pE i (6.137)
Here U.ƒ/ denotes the unitary (and therefore infinitely dimensional) representation
of the Lorentz group on the Hilbert space Hint of the interacting theory.
Note that P D .H; PE / transforms like a 4-vector,
U .ƒ/P U.ƒ/ D ƒ
P
(6.138)
E j0 i follows
so that from P j0 i D .m ; 0/
P jpE i D P U.ƒpE /j0 i (6.139)
D U.ƒpE /U .ƒpE /P U.ƒpE /j0 i (6.140)
D U.ƒpE /.ƒpE / P j0 i (6.141)
D .ƒpE / .m ; 0/ U.ƒpE /j0 i
E (6.142)
D .EpE ./; p/
E jpE i (6.143)
The bottom line is that every eigenstate jpE i can be obtained from a state j0 i with
vanishing 3-momentum by a boost ƒpE : jpE i D U.ƒpE /j0 i.
iv | Typical spectrum of P D .H; PE / of an interacting theory with mass gap:
Every state j0 i with vanishing momentum and “mass” (=rest energy) m
is associated to a hyperboloid (the “mass shell”) of states pE with energies
q
EpE ./ D jpj E 2 C m2 that are generated by boosts.
Note that the two-particle states occupy a continuum of hyperboloids because
the energy of two particles can take any value 2m E < 1 for vanishing total
momentum pE D 0.
Depending on the interactions, bound states of two particles can exist where the
energy 2m of the free particles is reduced by the binding energy. In this course, we
do not discuss bound states of interacting QFTs.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 122
LE C T U R E 18 → PS:211–222
v| Identity on Hint :
d 3p 1
Z
.1/1 particle D jpih
E pjE (6.144)
.2/3 2EpE
X Z d 3p 1
Generalization ! 1 D jihj C j ih j (6.145)
.2/3 2EpE ./ pE pE
Here, we choose the same normalization as for one-particle states.
The sum runs over all zero-momentum excited states j0 i.
4| Insert identity between fields !
X Z d 3p 1
hj .x/ .y/ji D hj .x/jpE ihpE j .y/ji C const (6.146)
.2/3 2EpE ./
Here, we drop the constant term coming from hj .x/ji (which vanishes often anyway
due to the symmetries of the vacuum).
5| With
hj .x/jpE i D hje iP x .0/e iP x
jpE i (6.147)
ipx
D hj .0/jpE i e jp0 DEpE ./ (6.148)
D hjU.ƒpE / U .ƒpE / .0/U.ƒpE / j0 i e ipx
jp0 DEpE ./ (6.149)
„ ƒ‚ … „ ƒ‚ …
hj .ƒpE 1 0/D .0/
ipx
D hj .0/j0 i e jp0 DEpE ./ (6.150)
Here we use the Poincaré invariance of the (interacting) vacuum, U.ƒ/ji D ji and
e iP x ji D ji, and the scalar nature of the field, U.ƒ/ .x/U .ƒ/ D .ƒx/.
6| …we find
d 3p 1
X Z
2 ip.x y/
hj .x/ .y/ji D jhj .0/j0 ij 3
e jp0 DEpE ./ (6.151)
.2/ 2EpE ./
Introduce p 0 -integration [recall Eq. (2.16) ff.]
d 4p i
Z
x 0 >y 0 X 2 ip.x y/
D jhj .0/j0 ij e (6.152)
.2/ p4 2 m2 C i "
„ ƒ‚ …
DF .x yIm2 /
x 0 <y 0
D hj .y/ .x/ji (6.153)
7| ! Källén-Lehmann spectral representation of the two-point correlator:
1
dM 2
Z
hjT .x/ .y/ji D .M 2 /DF .x yI M 2 / (6.154)
0 2
with spectral density
X
.M 2 / D 2 ı.M 2 m2 /jhj .0/j0 ij2 (6.155)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 123
LE C T U R E 18 → PS:211–222
Interpretation: The two-point correlator of an interacting QFT is the sum of propagators
of all possible states with mass m that can be created by a single field operator from the
interacting vacuum: jhj .0/j0 ij2 ¤ 0.
Note that this result is exact!
8| Typical spectral density of an interacting theory with mass gap:
.M 2 / D 2ı.M 2 m2 / Z C f multi-particle states for M 2 & .2m/2 g (6.156)
with (We assume here that the theory has only one massive particle D 1.)
Field-strength renormalization Z D jhj .0/j0 D 10 ij2 (6.157)
Physical mass m D m1 (6.158)
(given by H j10 i D m1 j10 i) (6.159)
Bare mass m0 (6.160)
(given by H D : : : 1=2 m20 2 : : : ) (6.161)
• Free theory: Z D jh0j .0/jpE D 0i0 j2 D 1 and m D m0
• Interacting theory: Z ¤ 1 and m ¤ m0
• Only m is observable
• Field-strength renormalization = Probability jhj .0/j10 ij2 that the field .0/
creates the interacting single particle state j10 i from the interacting vacuum ji.
9| Fourier transform of the two-point correlator:
Z 1
dM 2 i.M 2 /
Z
4 ipx
d x e hjT .x/ .0/ji D (6.162)
0 2 p 2 M 2 C i "
Z 1
(6.156) i Z dM 2 i.M 2 /
D C
p 2 m2 C i " .2m/2 2 p
2 M 2 C i"
(6.163)
free i 1
D (6.164)
p2 m20 C i "
! Typical analytical structure in the complex p 2 -plane:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 124
LE C T U R E 18 → PS:211–222
6.4.2 Application to QED: The Electron Self-Energy
Goal: Use perturbation theory to
1. verify the non-perturbative results from above for QED and
2. compute m and Z in first order of ˛.
Details: → Problemset 10.
1| 4 -theory 7! QED
= C m/
iZ2 .p
Z
d 4x e ipx hjT ‰.x/‰.0/ji D 2 C ::: (6.165)
p m2 C i "
The name “Z2 ” for the field-strength renormalization is conventional.
2| On the other side:
Z
d 4x e ipx hjT ‰.x/‰.0/ji (6.166)
Feynman rules for correlation functions
D (6.167)
3| ˛ 0 -order:
= C m0 /
i.p
(a) D (6.168)
p2 m20 C i "
4| ˛ 1 -order:
= C m0 /
i.p
(b) D (6.169)
p2m20 C i "
" #
d 4k i.k= C m0 / i
Z
. i e/2
.2/4 k 2 m20 C i " .p k/2 C i "
„ ƒ‚ …
i†2 .p/
= C m0 /
i.p
p2 m20 C i "
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 125
LE C T U R E 18 → PS:211–222
! IR- and UV-divergences
(Use regularization with ; ƒ and the techniques developed for the vertex corrections.)
Evaluation: → Problemset 10 !
" #
1
˛ xƒ2
Z
ƒ!1
†2 .p/ dx .2m0 = / log
xp (6.170)
2 0 .1 x/m20 C x2 x.1 x/p 2
ı
! This expression has a branch cut (in the complex p 2 -plane) emanating from
p 2 D .m0 C /2 , i.e., at the threshold of a two-particle state consisting of an electron
of mass m0 and a photon of (artificial) mass . There is, however, no simple pole at
p 2 D m2 .
5| Summation to all orders in ˛:
This is needed to recover the isolate one-particle pole at p 2 D m2 .
i| Definitions:
One-particle irreducible (1PI) diagram D Bridgeless one-particle diagram
(6.171)
A bridgeless graph cannot be separated into two pieces by deleting a single edge.
Examples:
Let furthermore
i †.p/ WD f Sum of all 1PI diagrams g (6.172)
D i †2 .p/ C O.˛ 2 / (6.173)
†.p/ does not include the propagators of the two external legs, recall Eq. (6.169).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 126
LE C T U R E 18 → PS:211–222
ii | Then
Z
d 4x e ipx hjT ‰.x/‰.0/ji (6.174)
D f Sum of all one-particle diagrams g (6.175)
D (6.176)
= C m0 /
i.p = C m0 /
i.p = C m0 /
i.p
D C Œ i †.p/ C ::: (6.177)
p2 m20 p2 m20 p2 m20
2
2
Use p= D p and †.p/; p= D 0 and write †.p
= / instead of †.p/
1 n
†.p=/
i X
D (6.178)
p
= m0 p
= m0 nD0
Geometric series (for matrices)
i 1
D (6.179)
= m0 1 †.p/
p =
p
= m0
i
D (6.180)
p
= m0 †.p
=/
Here we omit the infinitesimals " for the sake of simplicity.
It is Π.p/; p
= D 0 since, similar to our discussion of the general structure of
the vertex function previously, the matrix †.p/ must be a Lorentz scalar,
ƒ 1 †.p/ƒ 1 1 D †.ƒp/, and therefore can only be constructed from contracted
2 2
pairs of -matrices and the four-vector p, i.e.,
†.p/ D f . p / C g.p p / C c.
/ D f .p
= / C g.p
2
= / C c D †.p
=/
(6.181)
where f and g are arbitrary (analytic) functions and c is a constant (recall
D 4); note that ƒ 1 p
= ƒ 1 1 D p=0 . This also demonstrates that † can
2 2
equivalently be interpreted as a function of p
=.
6| Laurent series:
i Š iZ2
D C ::: (6.182)
p
= m0 †.p
=/ p
= m
! Expect simple pole for p
= D 1 m D m:
m m0 D †.p
= D m/ (6.183)
This is an implicit equation for the physical mass m.
! Expand denominator around this root:
d† ˇˇ
ˇ
m/2
p
= m0 = / D .p
†.p = m/ 1 C O .p
= (6.184)
dp
= ˇpDm
=
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 27
LE C T U R E 18 → PS:211–222
! 1
d† ˇˇ
ˇ
) Z2 D 1 (6.185)
dp
= ˇpDm
=
7| Results in leading order O.˛/:
i| Physical mass:
ım D m m0 D †.p
= D m/ (6.186)
2
D †2 .p
= D m/ C O.˛ / (6.187)
2
D †2 .p
= D m0 / C O.˛ / (6.188)
Use Eq. (6.170), → Problemset 10
!
ƒ!1 3˛ ƒ2 ƒ!1
m0 log 2
! 1 (6.189)
4 m0
In the last line we expanded †2 around m0 in ˛ and kept only the lowest order.
! Mass shift is UV-divergent!
ii | Field-strength renormalization:
Use 1 1 x D 1 C x C O.x 2 /.
ıZ2 D Z2 1 (6.190)
d† ˇˇ
ˇ
D C O.˛ 2 / (6.191)
dp= pDm
ˇ
=
d†2 ˇˇ
ˇ
D C O.˛ 2 / (6.192)
dp
= ˇpDm=
8 9
xƒ2
x log
ˆ
ˆ >
>
Z 1 ˆ
2 2 2
>
˛ < .1 x/ m C x =
$ dx (6.193)
2 0 ˆ x.1 x/m2 >
:C2.2 x/
ˆ
ˆ >
>
2 2
.1 x/ m C x 2
;
Note that the lowest order of † D †2 C O.˛ 2 / is linear in ˛ since we excluded
the external propagators from the definition of †.
! Field-strength renormalization is also UV-divergent!
→ Note 6.3
• The diverging mass of the electron is classically expected as it includes the energy
of its electrostatic field in the vicinity of the electron. This energy diverges for a
charged sphere with vanishing radius re as r1e ƒ ! 1.
• Our results on QED processes all involved the bare mass m0 . To compare them
with experiments, we should express m0 in terms of the observed mass m, which,
however, is problematic because their difference diverges! This conceptual impasse
motivates the introduction of a renormalized perturbation theory for QED where the
physical mass m instead of m0 shows up in the Feynman propagator (→ later).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 128
LE C T U R E 18 → PS:211–222
• It is easy to show that
.1/
ıZ2 D F1 .0/ (6.194)
.1/
where F1 .0/ was the term that we subtracted from the form factor of the vertex
correction to ensure that F1 .0/ D 1, recall Eq. (6.77). An application of the LSZ
reduction formula yields a correction to the form factor, namely
.1/ .1/ .1/
F1 .q 2 / D 1 C F1 .q 2 / C ıZ2 D 1 C F1 .q 2 / F1 .0/ (6.195)
which justifies our subtraction.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 129
LE C T U R E 19 → PS:244–257
→ Topics of Lecture 19
1. Renormalization of the electric charge
2. Running of the fine-structure constant
3. Landau pole
→ Topics of Problemset 10
1. Infrared divergence of the electron vertex function
2. The electron self-energy
6.5 Electric Charge Renormalization
Remember the radiative corrections:
(a) Vertex correction
! Form factors and anomalous magnetic moment, IR-div. and UV-div., had to subtract
.1/
F1 .0/ ad hoc
(b) Soft bremsstrahlung
! IR-div. cancelled with IR-div. of vertex correction
(c) Electron self-energy
.1/
! Field-strength and mass renormalization, explained subtraction of F1 .0/ and thereby
removed UV-div. of vertex correction
Here ^ Vacuum polarization diagram (d) ! Photon self-energy
This is analogous to the electron self-energy (c) which modified the electron propagation due to
virtual photons. Here, the photon propagator will be modified due to the presence of virtual
electron-positron pairs. This will lead to momentum dependent modifications of the strength
of the electromagnetic field.
1| One-loop correction:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 130
LE C T U R E 19 → PS:244–257
As before, we exclude the photon propagators of the legs from all expressions.
(6.196)
d 4k =C
i.k m/bd i.k
= C q= C m/ca
Z
2
D . 1/ . i e/ ab k 2 dc (6.197)
„ƒ‚… .2/4 m2 .k C q/2 m2
Fermion loop
d 4k = i.k= C q= C m/
i.k C m/
Z
2
D . 1/. i e/ Tr (6.198)
.2/4 k 2 m2 .k C q/2 m2
i …2 .q/ (6.199)
The sign of the fermion loop follows from the contraction ‰‰‰‰ D ‰‰‰‰ where
we used that ‰‰ D ‰‰ for fermionic fields (recall that ‰.x/‰.y/ D SF .x y/ is
the Feynman propagator).
2| ^ Sum of all 1-particle irreducible diagrams:
i … .q/ D i …2 .q/ C O.˛ 2 /
(6.200)
What follows is analogous to our discussion of the generic structure of the vertex
correction :
i| Only tensors available: g and q q ! … .q/ D A.q 2 / g C B.q 2 / q q
ii | Ward identity (recall Eq. (6.25) and references below for the vertex correction ):
q … .q/ D 0 ! B D qA2 ! … .q/ D .q 2 g q q / qA2
iii | ! … .q/ has no pole for q 2 D 0
Motivation: Poles at q 2 D 0 arise from massless intermediate states with propagator
ig
q 2 Ci"
– but these do not occur in 1-particle irreducible diagrams. A rigorous proof
of this statement is possible but non-trivial.
2/
! ….q 2 / A.q q2
regular at q 2 D 0 !
… .q/ D .q 2 g q q / ….q 2 / (6.201)
Note that q 2 D 0 does not imply q D 0 and thus q q can be chosen finite for
q 2 ! 0; consequently, ….q 2 / must be regular in this limit if … .q 2 / is.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 131
LE C T U R E 19 → PS:244–257
3| ^ Sum of all diagrams with two photon legs:
(6.202)
ig ig 2 2
ig
D C i.q g q q /….q / C : : : (6.203)
q2 q2 q2
Define ı q q =q 2 and use g g D ı
ig ig ig 2 2
D 2
C 2
….q 2 / C … .q / C : : :
q q q2
(6.204)
Use D
1
ig ig X n 2
D C … .q / (6.205)
q2 q2
nD1
„ ƒ‚ …
1
D 1
1 ….q 2 /
Geometric series
i q q i q q
D 2 g C (6.206)
q Œ1 ….q 2 / q2 q2 q2
4| ^ … .q/ contracted with a fermion vertex to form an S -matrix element:
Ward identity !
ig
¶ (6.207)
q 2 Œ1 ….q 2 /
The “¶” signifies that this equation is only true for computations in S -matrix elements.
For a proof ↑ P&S pp. 238–244, in particular Eq. (7.66).
Note that this propagator has a pole at q 2 D 0 (to all orders in ˛) because ….q 2 / is
regular so that the photon remains exactly massless. Formally, this is a consequence of
the Ward identity.
5| Charge renormalization:
i| Define
1
Z3 (6.208)
1 ….0/
This is a finite number since ….q 2 / is regular at q 2 D 0 (and we assume j…j < 1
as otherwise the resumation with the geometric series is not justified).
Then, for q 2 ! 0 (i.e. “almost-on-shell photons”)
e 2 g Z3 e 2 g
D ::: ::: ! D ::: : : : (6.209)
q2 q2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 32
LE C T U R E 19 → PS:244–257
ii | ! Charge renormalization:
Bare charge (given by Hint D e0 ‰
e0 ‰A ) (6.210)
p
Physical charge e Z3 e0 (6.211)
e2 e02
Fine-structure constant D ˛ Z3 ˛0 D Z3 (6.212)
4 4
Note that the bare charge e0 was previously
p called e. The charge measured in
experiments is the physical mass e D Z3 e0 , hence the new notation.
That is, for soft photons, q 2 ! 0, we can incorporate the effects of the vacuum
polarization diagrams by simply replacing the bare charge e0 by the renormalized,
physical charge e in computations of S-matrix elements.
Note that in lowest order ˛ 0 , it is Z3 D 1 so that e D e0 and ˛ D ˛0 . In
general, Z3 D 1 C O.˛0 / so that ˛ D ˛0 C O.˛02 /. In particular, we can write
O.˛ 2 / D O.˛02 /.
iii | In addition, for q 2 ¤ 0 and ….q 2 / D …2 .q 2 / C O.˛ 2 /, each virtual photon line
comes with (the charges come from the interaction vertices)
ig e02 ig e 2 Œ1 ….0/
D (6.213)
q2 1 ….q 2 / q2 1 ….q 2 /
Use .1 x/ 1 D 1 C x C O.x 2 / (6.214)
ig e 2 Œ1
…2 .0/
D 2
C O.˛ 2 / (6.215)
q 1 …2 .q 2 /
Use .1 x/ D .1 C x/ 1 C O.x 2 / (6.216)
ig e2
D C O.˛ 2 / (6.217)
q2 Œ1 …2 .q 2 / Œ1 C …2 .0/
ig e2
D C O.˛ 2 / (6.218)
q2 1 Œ…2 .q 2 / …2 .0/
! q 2 -dependent charge/fine-structure constant:
e02 =4 ˛
˛eff .q 2 / D C O.˛ 2 / (6.219)
1 ….q 2 / 1 Œ…2 .q 2 / …2 .0/
That is, for arbitrary momenta, the effect of replacing the tree-level propagator with
the full propagator is a q 2 -dependent electric charge, or, equivalently, fine-structure
constant.
6| Computation of …2 :
i| From Eq. (6.199):
i˛ i˛0
We use m and e instead of m0 and e0 since = m
k
D = m0
k
C O.˛02 / and we are
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 133
LE C T U R E 19 → PS:244–257
only interested in linear order corrections (note that …2 is already of order ˛0 ).
d 4k = i.k= C q= C m/
i.k C m/
Z
i…2 .q/ D . i e/2 Tr
(6.220)
.2/4 k2 m2 .k C q/2 m2
Trace identities
d 4k k .k C q/ C k .k C q/ g .k .k C q/ m2 /
Z
2
D 4e
.2/4 .k 2 m2 /..k C q/2 m2 /
(6.221)
0 0
Feynman parameters, Substitution l k C xq, Wick rotation l i lE
Z 1
d 4lE
Z
2
$ 4i e dx
0 .2/4
2 2 2
g lE C g lE 2x.1 x/q q C g .m2 C x.1 x/q 2 /
d 2
.lE C /2
(6.222)
where m2 x.1 x/q 2
In the last line, the term d2 g lE2
in the numerator follows for spacetime
dimension d D 4 from Eq. (6.59) ff. (also all terms linear in lE have been dropped).
We keep the spacetime dimension d explicitly as we need it for the dimensional
regularization below.
ii | Strong UV-divergence: ^ UV-cutoff jlE j < ƒ, then
ƒ!1
i …2 .q/ e 2 ƒ2 g !1 (6.223)
This follows simply by power counting.
Note that this result also violates the Ward identity q …2 D 0 as there is no
corresponding term / q q ; this violation produces a (infinite) photon mass!
! To make sense of this result (and restore the Ward identity), a regularization is
needed!
iii | Dimensional regularization: (Details: → Problemset 11)
1. Lower the spacetime dimension d 2 N until the UV-divergence vanishes
2. Generalize all expressions to d 2 R (below even d 2 C is fine)
3. Take the limit d % 4 in observable quantities
We could also use Pauli-Villars regularization (with the same results), which,
however, is in this case much more complicated than dimensional regularization.
For d 2 C we find (Proof: → Problemset 11)
d
.n d2 /
n
d dlE
1 1 1
Z 2
2
D (6.224)
.2/d .lE C /n .4/d=2 .n/
d
d .n d2 1/
2 n 1
d dlE lE
1 1
Z 2
2
D (6.225)
.2/d .lE C /n .4/d=2 2 .n/
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 134
LE C T U R E 19 → PS:244–257
For d 2 N these are identities that must be proven, for d 2 C n N, these are
definitions of the left-hand side.
^ n D 2:
d
.´/ has poles at ´ D 0; 1; 2; : : : ! .2 2 / has isolated poles at d D 4; 6; : : :
^d D4 " and use
d 2
2 D ."=2/ D C O."/ (6.226)
2 "
: Euler-Mascheroni constant
Note: g g D d , so that in invariant integrals over spacetime the substitution
1 2
l l ¶ l g (6.227)
d
is valid [generalization of Eq. (6.59)].
iv | Evaluate Eq. (6.222) with Eq. (6.224) & Eq. (6.225) & Eq. (6.227)
(and use ´.´/ D .1 C ´/) !
i …2 .q/ $ .q 2 g q q / i …2 .q 2 / (6.228)
x.1 x/.2 d2 /
Z 1
8e 2
with …2 .q 2 / D dx (6.229)
.4/d=2 0 Œm2 x.1 x/q 2
2 d=2
„ ƒ‚ …
D
Note that …2 .q/ has the expected form Eq. (6.201) (and therefore satisfies the
Ward identity).
v| Use Eq. (6.226) to expand in ":
1
2˛ 2
Z
2
…2 .q / $ dx x.1 x/ log ./ C log .4/ C O."/ (6.230)
0 "
To show this, use 2Cd=2 D 1 2" log ./ C O."2 / and .4/ d=2 D .4/ 2 .1 C
"
2 log .4// C O." /, and keep only constant and diverging terms.
2
7| O.˛/ charge renormalization:
e2 e02 ….0/
D Z3 1D (6.231)
e02 1 ….0/
D …2 .0/ C O.˛ 2 / (6.232)
"!0 2˛ "&0
! 1 (6.233)
3"
! If the observed charge is finite, 1 < e < 0, the bare charge diverges: e0 D 1.
Note that e0 is not observable so that this is not a falsifiable prediction of QED!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 135
LE C T U R E 19 → PS:244–257
8| O.˛/ q 2 -dependence of ˛eff .q 2 / depends on
(This is an experimentally observable prediction!)
2˛ 1 m2
Z
O 2 2
…2 .q / …2 .q / …2 .0/ D dx x.1 x/ log
0 m2 x.1 x/q 2
(6.234)
Note that the UV-divergence for " ! 0 drops out!
O 2 .q 2 /:
9 | Analysis & Interpretation of …
i| Note:
s-channel: ) s .p C p 0 /2 D q 2 >0 (6.235)
t-channel: ) t .p k/2 D q 2 <0 (6.236)
u-channel: ) u .p k 0 /2 D q 2 <0 (6.237)
The eponymous variables s, t , and u are known as Mandelstam variables.
(To show the inequalities, use the Cauchy-Schwarz inequality. Because of p 2 D
p 02 D k 2 D k 02 D m2 > 0 all momenta are time-like vectors pointing into the
future light-cone, i.e., for each momentum there is a rest frame with p 0 D m > 0.)
• t- and u-channel: … O 2 .q 2 / is analytic in the left half of the (complex) q 2 -plane.
• s-channel: … O 2 .q 2 / has a branch cut on the real axis for m2 x.1 x/q 2 < 0,
i.e., starting at m2 D q 2 =4 , q 2 D .2m/2 where a real (on-shell) electron-
positron pair can be produced.
ii | ^ Effective potential in nonrelativistic limit [recall Eq. (4.199) ff.]:
d 3q i qExE e2
Z
E D
V .x/ e (6.238)
.2/3 q j2 Œ1 …
jE O 2 . jEq j2 /
d 3q i qExE e2
Z
D e Œ1 C … O 2 . jE q j2 / C O.˛ 2 / (6.239)
.2/3 q j2
jE
d 3q i qExE
2 h
e ˛
Z
jq 2 j m2
i
2
D e 1 C jE
q j C O.˛ 3 / (6.240)
.2/3 q j2
jE 15 m2
ı ˛ 4˛ 2 .3/
ı xE (6.241)
jxj
E 15m2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 136
LE C T U R E 19 → PS:244–257
Recall that q 2 D .p
p / 2jpE pE j and jq j m in the
0 2 0 2 2 2 nonrelativistic
1
limit. Use log 1 x D x C O.x / to expand the logarithm in q =m2 and use that
2
R1 2
0 dx x .1 x/2 D 1=30.
! Electromagnetic force becomes much stronger at small distances
That is, QED tells us that the Coulomb potential of charged point particles is a
low-energy/large-distance approximation!
iii | Experimental verification:
Energy shift of s-orbitals in the hydrogen atom (contributes to the Lamb shift):
4˛ 2 .3/ 4˛ 2
Z
3 2 lD0
E d x j .x/j E 2
ı xE D 2
j .0/j2 < 0 (6.242)
15m 15m
˛ .3/
Note that the Darwin term HDarwin D 2m xE has a similar form but follows
2ı
already from the (first quantized) Dirac equation, i.e., at tree-level. In contrast,
the above correction is of loop-order ˛ 2 and contributes to the Lamb shift of
1058 MHz (but only 2 % = -27 MHz; the dominant part of 1011 MHz is due to
the self-energy of the electron, an additional part of 68 MHz comes from the
anomalous magnetic moment [13]), a famous prediction of QED that cannot be
derived from the Dirac equation and explains the observed splitting of the hydrogen
orbitals 2S1=2 and 2P1=2 with total angular momentum j D 1=2 (Dirac theory
result: → Problemset 4). Note that the theoretical prediction of 1058 MHz contains
higher-order contributions, ↑ Ref. [13] p. 270.
iv | More generally, one finds the Uehling potential:
˛ e 2mr
˛
V .r/ D 1C p C ::: (6.243)
r 4 .mr/3=2
A contour integration is needed to derive this, ↑ P&S p. 254.
Note that the range of the correction is given by the electron’s Compton wavelength
c D h=mc D 2=m. Since the length scale of variations of atomic orbitals is given
by the Bohr radius, a0 D c =.2 ˛/ 22 c , the nonrelativistic approximation
from above is sufficient for atomic physics.
v| Interpretation: Vacuum polarization:
The vacuum behaves as a dielectric medium where electric dipoles of size 1=m,
formed by electron-positron pairs, screen the bare charge e0 . The energy scale
q at which we observe the electron determines the size of the sphere r D 1=q
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 37
LE C T U R E 19 → PS:244–257
that contributes to the observed charge e; for r & 1=m the screening due to
electron-positron pairs kicks in, for r . 1=m the screening becomes weaker and
the observed charge approaches the bare charge. Note that in this picture, the
e e C -pairs that traverse the surface of the sphere are responsible for reducing the
infinite bare charge to the finite physical charge.
vi | ^ Relativistic limit q 2 m2 :
1
m2 m2
2˛
Z
O 2 .q 2 / D
… dx x.1 x/ log log .x.1 x// C O
0 q2 q2
(6.244)
q2 m2
˛ 5
$ log CO (6.245)
3 m2 3 q2
x
R1
Use log xCa D log xa C O.x/ and 0 dx x.1 x/ log .x.1
x// D 5=18.
! “Running” of ˛eff with the length scale r D q1 ! 0
q 2 m2 ˛
˛eff .q 2 / (6.246)
˛ q2
1 3 log Am2
with A D exp .5=3/:
Note that Eq. (6.246) is only valid for relativistic jq 2 j m2 and not for jq 2 j !
1
0 , r ! 1; thus the approach of ˛eff to ˛ 137 cannot be extracted from
this expression! However, Eq. (6.246) demonstrates the growth of the coupling
strength at small length scales r ! 0 , jq 2 j ! 1 where the screening due to the
polarized vacuum becomes weaker (→ note below).
This modification is crucial for explaining scattering cross sections at high energies,
↑ P&S p. 256 Fig. 7.7.
So when one refers to the fine-structure constant, the “constant” can only refer to
constancy in time and the “the” refers to the energy/length/momentum scales that
1
we typically observe, i.e., where ˛eff 137 .
! Renormalization (→ next lectures)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 138
LE C T U R E 19 → PS:244–257
→ Note 6.4
Eq. (6.246) describes the coupling strength at very high energies q 2 m2 . Strangely,
the expression predicts a divergence of ˛eff at a large but finite momentum q 2 ƒL2
:
!
2
˛ ƒL Š 3
1 log 2
D 0 ) ƒL me 2˛ 10286 eV ; (6.247)
3 Am
1
where we used the renormalized mass m 0:5 MeV of the electron, ˛ 137 , and
dropped the insignificant constant A.
This divergence is known as ⁂ Landau pole [14].
• Its occurrence is often interpreted as onset of a non-perturbative regime of QED
and/or as the necessity for a UV-completion of QED (embedded in some “grand
unified theory”). Note that the existence – let alone the exact position – of the
Landau pole cannot be inferred from perturbative computations because the very
premise of perturbation theory is the smallness of the coupling constant (which is
1
justified on our energy scales where ˛eff 137 but clearly breaks down long before
the purported Landau pole is reached).
• Thus, to study whether the Landau pole bears physical meaning or is a mere
artifact of perturbation theory, one must resort to non-perturbative methods like
↑ ⁂ Lattice QED, e.g. [15]; these studies suggest that the Landau pole is “real”
in the sense that pure QED seems to be ↑ ⁂ quantum trivial (i.e., ˛ renormalizes
to zero if the momentum cutoff is taken to infinity), but is rendered inaccessible
by spontaneous symmetry breaking. However, the QED we use is not “pure”
but embedded into the standard model which is expected to be embedded into a
UV-complete grand unified theory. Thus “our” QED is more likely an effective,
incomplete theory and the Landau pole may be a mere artifact of perturbation
theory that occurs when we extrapolate this theory into a regime beyond its validity.
• There is another perspective on this: As Freeman Dyson argued [16], the pertur-
bative expansion of any quantity (e.g., the g-factor) in QED in the fine structure
constant ˛ / e 2 does not converge (more precisely: its convergence radius is zero).
This is so because setting ˛ < 0 (equivalently: e 7! i e) leads to an unstable vacuum
where electrons and positrons repel each other. Such a vacuum would quickly fill
with spatially separated condensates of electrons and positrons without lower bound
on its energy. If the convergence radius of any perturbative expansion in ˛ were
non-zero, it would necessarily contain a range of negative ˛ (which it cannot because
of the instability). The existence of the Landau pole in the perturbation series can be
seen as sign of this non-convergence. (Which doesn’t mean that the first N orders
in the series do not improve on the predictions of QED – they do! It only means
that after some (very high) order N the predictions will get worse because QED is
an effective QFT that lacks a proper UV completion. A common [16, 17] estimate
of the order at which the series starts to become unreliable is N 1=˛ 137; it
is based on non-perturbative terms of the form e c=˛ becoming comparable to the
perturbative contributions: e c=˛ ˛ N ) N 1=˛. Clearly Feynman diagrams
with O.137/ loops are far beyond anything of practical relevance.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 39
LE C T U R E 19 → PS:244–257
• The problem of the Landau pole is of no practical importance for actual predictions
of QED because the energy scale at which it occurs are beyond reach:
ELHC 1013 eV EPlanck 1028 eV ƒL 10286 eV : (6.248)
• Note that the occurrence of a Landau pole is not specific to QED but a rather
generic feature of field theories that are not asymptotically free (cf QCD).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 0
LE C T U R E 20 → P S : 3 1 5 – 32 3
→ Topics of Lecture 20
1. Systematic approach to renormalization
2. Counting UV-divergencies
3. The mass dimension
4. A short note on the Einstein-Hilbert action
7 Systematics of Renormalization
Remember:
• IR-divergences:
– Due to massless particles (photons)
(The amplitudes for k ! 0 real/virtual photons diverge.)
– Regulate with small photon mass ()
– Divergences from soft virtual photons (vertex correction) and soft bremsstrahlung
cancel
! Not a fundamental problem (we do not have to reinterpret/change the theory)
• UV-divergences:
– Due to unbounded high momenta of particles (= unbounded small length scales) in
all three radiative corrections:
– Regulate with Pauli-Villars (ƒ) or dimensional regularization (")
– Cancelled in several observable quantities
(The UV-divergence of field-strength renormalization cancelled with the UV-
divergence of the vertex correction. The momentum-dependence of the effective
electric charge did not depend on the UV-regulator ".)
– Diverging differences between physical and bare quantities
This is clearly a conceptual problem as the physical quantities are obviously finite
(as given by experiments), thus the bare quantities (so far seen as fixed parameters
of the microscopic Lagrangian) must then be cutoff dependent and diverge for
ƒ ! 1.
! Fundamental problem (of most interacting QFTs)
! UV-divergences are considered more severe problems than IR-divergences.
! Study UV-divergences systematically
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 1
LE C T U R E 20 → P S : 3 1 5 – 32 3
7.1 Counting UV-Divergences
1| Goal: Classify UV-divergences in QED
Which Feynman diagrams diverge and how many diverging quantitites are hidden in the
amplitudes of QED?
2| Definitions:
Ne D # external electron lines (7.1)
N D # external photon lines (7.2)
Pe
Y 1
Pe D # electron propagators ! (7.3)
k= i m
i
P
Y 1
P D # photon propagators ! (7.4)
i
ki2
V D # vertices (7.5)
L Z
Y d 4ki
L D # independent loops ! (7.6)
.2/4
i
This is valid for (diagrams of ) scattering amplitudes; for (diagrams of ) vertex functions,
count the propagators to external points as external lines and not as propagators (they are
multiplicative and therefore irrelevant for the UV-behaviour of the diagram).
3| Superficial degree of divergence:
DQED .3L C L/ .Pe C 2P / (7.7)
„ ƒ‚ … „ ƒ‚ …
Numerator Denominator
Note that a 4-dimensional integral diverges as ƒ4 ; e.g., in spherical coordinates, 3L
comes from the Jacobian and L from the integrations. DQED quantifies the divergence of
the integral, not the integrand.
Intuition:
8
<> 0
ˆ W Divergence with ƒD
DQED D 0 W Divergence with log ƒ (7.8)
No divergence
ˆ
<0 W
:
Example:
log ƒ and D D41 .2 C 2 1/ D 0 (7.9)
However: Not always correct!
Reasons:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 2
LE C T U R E 20 → P S : 3 1 5 – 32 3
• Divergence may be weaker (or absent) if symmetries make divergent terms cancel:
log ƒ although DQED D 4 1 .2 C 2 0/ D 2
(7.10)
(Recall the restrictions on the general structure of this diagram due to the Ward
identity in Eq. (6.201), → discussion below.)
• Divergence may be worse if diagram contains divergent subdiagrams (yellow):
log ƒ although DQED D 4 1 .2 C 2 2/ D 2
(7.11)
• Tree-level diagrams with no propagators have D D 0 but no divergence:
1 although DQED D 4 0 .0 C 2 0/ D 0 (7.12)
4| Use (standard graph theory identities, ← discussion of Fenyman rules)
L D Pe C P V C 1 D Cycle space dimension (7.13)
„ ƒ‚ …
# Edges
1
V D 2P C N D .2Pe C Ne / (7.14)
2
To show the two expressions in the second row, use that every internal photon line
corresponds to two vertices and every external photon line to one vertex; the same is true
for fermion lines except that every vertex is shared by two lines so that the final count
must be divided by 2.
to show
3
DQED $ 4 N Ne (7.15)
2
! Independent of number of vertices!
5| Aside: Furry’s theorem:
Feynman diagrams with an odd number of photons
(7.16)
as their only external lines vanish identically.
Proof: Follows from charge conjugation symmetry (C ) of QED
(Use C ji D ji and Cj C D j with j D ‰ ‰, ↑ P&S p. 318.)
This tells us that a single, real photons can never be produced or absorbed by the
interacting vacuum of QED.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 3
LE C T U R E 20 → P S : 3 1 5 – 32 3
6| Enumerate diagrams with DQED 0:
We consider only amputated and one-particle irreducible diagrams as all other diagrams
are products of these. Grey blobs denote the sums of all such diagrams with specified
external lines. We consider the amplitudes as functions of their external momenta and
express them as power series with unkown coefficient (which may or may not diverge
with the UV-cutoff ƒ). In the following, “” denotes asymptotic scaling up to regular
terms. Recall that Ne D 0; 2; 4; : : : in QED.
i| Ne D 0:
a| N D 0 (DQED D 4):
badly divergent (7.17)
Unobservable vacuum energy shift ! Ignore diagram
b| N D 1 (DQED D 3):
Furry
D 0 (7.18)
c| N D 2 (DQED D 2): [Recall our first-order result in Eq. (6.230).]
D .g q 2 q q /….q 2 / (7.19)
const
.g q 2 qq / (7.20)
"
.g q 2 q q / const log ƒ (7.21)
„ ƒ‚ …
a0 .ƒ/
The divergence comes from ….q 2 / and is logarithmic. Recall that we used
dimensional regularization for our first-order calculation Eq. (6.230), so that the
divergence log ƒ with a Pauli-Villars regulator ƒ is not obvious. a0 .ƒ/ is a
q-independent quantity diverging with ƒ.
This is one of the examples where symmetry (via the Ward identity) makes
the divergence less severe than superficially predicted by D: the zeroth and
first-order coefficients of the q-expansion must vanish due to symmetry; the
divergence left comes from the quadratic term and is only logarithmic.
d| N D 3 (DQED D 1):
Furry
D 0 (7.22)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 4
LE C T U R E 20 → P S : 3 1 5 – 32 3
e| N D 4 (DQED D 0):
Ward
1 (7.23)
This follows from symmetry arguments (Ward identity) that make potentially
diverging terms vanish identically, ↑ P&S p. 320.
Note: This diagram describes light-by-light scattering (Halpern scattering) in
QED. The lowest-order amplitude is very weak (of order ˛ 2 =m4e ); therefore,
we do not experience this in everyday life and the linearity of classical
electrodynamics is a good approximation. Nevertheless, it has consequences:
In astronomy, observable -rays are restricted to energies below 80 TeV; above
this threshold, the photons scatter at the ubiquitous microwave background
and the universe becomes opaque. For direct experimental observations at
LHC, see the recent paper [18].
ii | Ne D 2:
a| N D 0 (DQED D 1): [Recall our first-order result in Eq. (6.170).]
const log ƒ Cp
= const log ƒ (7.24)
„ ƒ‚ … „ ƒ‚ …
a1 .ƒ/ a2 .ƒ/
It can be shown that this scaling is true in all orders, ↑ P&S p. 319.
b| N D 1 (DQED D 0): [Recall our first-order result in Eq. (6.76).]
ie log ƒ (7.25)
„ ƒ‚ …
a3 .ƒ/
It can be shown that this scaling is true in all orders, ↑ P&S p. 319.
! Diagrams only diverge if they contain Eq. (7.21), Eq. (7.24) or Eq. (7.25) as
subdiagrams.
! QED contains only four UV-divergent numbers: a0 ; a1 ; a2 ; a3 .
7| Idea: Absorb finite number of diverging quantitites in finite number of diverging but
unobservable Lagrangian parameters ! Renormalization (→ below)
“Hiding” the divergences in unobservable parameters makes all other observable quantities
(like scattering amplitudes and physical parameters) cutoff-independent and UV-finite.
8| Generalization: ^ QED in d spacetime dimensions !
DQED D dL Pe 2P (7.26)
d 4 d 2 d 1
$dC V N Ne (7.27)
2 2 2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 145
LE C T U R E 20 → P S : 3 1 5 – 32 3
To show this, note that the identities Eq. (7.13) and Eq. (7.14) are still valid.
! Observations:
• For d < 4, diagrams of higher order (V ! 1) are always superficially convergent
(independent of potentially diverging subdiagrams)
• For d D 4, DQED is independent of the order V
(the divergence of diagrams can be traced back to a finite number of diverging
amplitudes/subdiagrams)
• For d > 4, diagrams of higher order (V ! 1) are always superficially divergent
This means, that the “reductionistic approach” only works in d D 4 dimensions where
the divergence of all diagrams can be traced back to a finite number of diverging
subdiagrams.
9| This is also valid for other QFTs and motivates three classes of theories:
• Super-Renormalizable theory:
Only a finite number of Feynman diagrams (not amplitudes = sums of diagrams!)
superficially diverge.
Example: QED in d D 2 C 1
• Renormalizable theory:
Only a finite number of amplitudes superficially diverge.
! Divergences at all orders in perturbation theory.
Example: QED in d D 3 C 1
• Non-Renormalizable theory:
All amplitudes diverge at sufficiently high order in perturbation theory.
Example: QED in d D 4 C 1
→ Note 7.1
• There are examples in which the divergences are not as bad as superficially predicted
due to symmetries that cancel diverging amplitudes.
• The diverging amplitudes of superficially renormalizable theories can always be
absorbed into a finite number of unobservable Lagrangian parameters (→ below).
Alternative approach
1| ^ n -theory (for simplicity)
1 1 2 2 n
L n D .@ /2 m with n 2 N (7.28)
2 2 nŠ
2| Definitions:
N D # external lines (7.29)
P D # propagators (7.30)
V D # vertices (7.31)
L D # independent loops (7.32)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 6
LE C T U R E 20 → P S : 3 1 5 – 32 3
3| Superficial degree of divergence:
D n dL 2P (7.33)
d 2 d 2
$dC n d V N (7.34)
2 2
Use the graph identities L D P V C 1 and nV D N C 2P to show this.
! For n D 4 in d D 4 independent of V ! Renormalizable
4| Alternative approach via dimensional analysis:
h 2
i| Recall: „ D c D 1 and c D mc D m
! Dimension of length: Œc D M 1 (M : dimension of mass)
ii | Dimension of action: ŒS D 1 (since „ D 1)
iii | S D d dx L and Œd dx D M d !
R
Dimension of Lagrangian density: ŒL D M d
As all dimensions can be expressed in M , we say that “L has (mass) dimension d ”.
iv | From Eq. (7.28) follows (use Œ@ D M ):
d 2
ΠD M 2 (7.35)
Œm D M (consistent!) (7.36)
ΠD M d n.d 2/=2
(7.37)
v| ^ Amplitude M of single diagram with N external lines
! Could arise (on tree-level) from interaction N ! ΠD M d N .d 2/=2
! ŒM D Œ D M d N .d 2/=2 (recall M D C O.2 / from Eq. (4.147))
vi | ^ Diagram with V vertices ! M V ƒD for the UV-cutoff ƒ ! 1
(This is an implicit definition of the superficial degree of divergence D.)
! (use Œƒ D M )
ŒV ŒƒD D ŒM D M d N .d 2/=2
(7.38)
d 2
V logM ΠC D D d N (7.39)
2
!
d 2
D n D d logM ΠV N D Eq. (7.34) (7.40)
„ ƒ‚ … 2
d n.d 2/=2
5| Therefore we find the equivalent characterization:
• Super-Renormalizable theory:
Coupling constant has positive mass dimension: logM Π> 0.
• Renormalizable theory:
Coupling constant is dimensionless: logM ΠD 0.
Example: QED with Œe D 1 is superficially renormalizable.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 7
LE C T U R E 20 → P S : 3 1 5 – 32 3
• Non-Renormalizable theory:
Coupling constant has negative mass dimension: logM Π< 0.
This argument remains valid for other QFTs as well, in particular QED.
Aside: Why quantum gravity is special
1| Fields: Components of the metric tensor g .x/
Note that in general relativity, the metric is position dependent, that is, a field.
2| Einstein-Hilbert action of pure gravity:
1
Z
d 4x
p
SEH D j det g.x/j ŒR.g.x// 2ƒc (7.41)
16G
R D g R : Ricci scalar with Ricci tensor R
ƒc : Cosmological constant
G: Gravitational constant = Coupling constant of gravity
If matter is present, this action is extended by the covariant action of the matter fields
(e.g., LQED ) which then generates a non-vanishing energy-momentum tensor in the
Einstein equations below.
ı
! Equations of motion = Einstein’s field equations (in vacuum):
1
R g R C ƒc g D 0 (7.42)
2
3| Recall:
R g R g R g @
g @ .g @ g / (7.43)
3 2
) ŒR D Œg Œ@ (7.44)
are the Christoffel symbols of the second kind, R
is the Riemann curvature tensor.
and
ds 2 D g dx dx (7.45)
) L2 D Œds 2 D ŒgŒdx2 D ŒgL2 (7.46)
) Œg D 1 (7.47)
ds 2 is the squared length element of the Riemannian spacetime.
such that
ŒR D Œ@2 D L 2
D M2 (7.48)
Š
4| From Eq. (7.41) it follows ŒG 1 Œdx4 ŒR D ŒG 1M 4M 2 D ŒS D 1, i.e.,
logM ŒG D 2<0 (7.49)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 8
LE C T U R E 20 → P S : 3 1 5 – 32 3
! Einstein gravity is superficially non-renormalizable!
„c 1
Recall that G D 2 D 2 with the Planck mass mP , consistent with our result.
mP mP
Superficial non-renormalizability does not prove non-renormalizability as there may still
be non-trivial cancellations that make the theory UV-finite.
If you wonder why the negative mass dimension of G is a problem although the action is
1
proportional to G (which has positive mass dimension): To apply our reasoning (which is
completely based on perturbation theory!) we must bring the action first into a form of
a free (Gaussian) theory that is perturbed by non-quadratic terms proportional to some
(small) coupling constant [19]; after a rescaling of the fields, this constant is essentially G
and only then corresponds to the of 4 -theory or the e of QED.
→ Note 7.2
• At one-loop level, pure Einstein gravity (no matter fields) is – quite unexpectedly! –
UV-finite [20].
• However, when matter is coupled to gravity, the one-loop diagrams are UV-divergent,
see Ref. [20] for the example of a scalar field and references in Ref. [21].
• At two-loop level, pure Einstein gravity is proven to be UV-divergent [21]. That is,
no unexpected cancellations occur.
• Therefore it is widely believed (though, to my knowledge, not proven) that no
unexpected cancellations occur beyond two-loop order; therefore, Einstein gravity is
perturbatively not renormalizable.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 14 9
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
→ Topics of Lecture 21
1. Bare perturbation theory
2. Renormalizted perturbation theory
3. Application to QED
7.2 Renormalized Perturbation Theory
Goal: Compute finite predictions from given physical parameters m and e for ƒ ! 1
Recipe:
Historically, this was the first widely accepted “fix” for the UV-problems of QFTs.
(i) Compute UV-divergent amplitude with UV-regulator ƒ to some order in ˛0 :
M D M.m0 ; e0 I ƒ/ C O.˛0 / (7.50)
(ii) Compute physical mass, physical charge and field-strength renormalization:
m D m.m0 ; e0 I ƒ/ C O; e D e.m0 ; e0 I ƒ/ C O; Z D Z.m0 ; e0 I ƒ/ C O (7.51)
The order O of these computations should be consistent with the order of M. The
field-strength renormalization Z is only needed for the computation of S-matrix elements
(where we sum only over amputated and fully connected diagrams), but not for correlation
functions (where we sum over all connected diagrams). This follows from the LSZ
reduction formula (which we did not discuss, ↑ P&S pp. 222–230, in particular Eq. (7.45)
on p. 229).
(iii) Renormalization:
Eliminate m0 and e0 in favour of m and e (which are fixed and given by experiments):
e0 D e0 .m; eI ƒ/; m0 D m0 .m; eI ƒ/ (7.52)
We did this previously when discussing the charge renormalization where we replaced
m0 and e0 by m and e in lowest order.
(iv) Then
M.m; e/ lim M.m0 .m; eI ƒ/; e0 .m; eI ƒ/I ƒ/ (7.53)
ƒ!1
is finite and indepenent of ƒ in all orders of ˛.
This is a remarkable, non-trivial observation! Note that this requires the bare parameters
to be cutoff dependent and divergent for ƒ ! 1, i.e., we change the “microscopic”
theory parametrically with ƒ. This interpretation is justified by the numerous extremely
precise predictions of QED like the anomalous magnetic moment (where we – somewhat
naively – used the physical value ˛ and not the bare value ˛0 to evaluate the numerical
correction for ge .).
! Bare perturbation theory (since the Feynman rules involve bare parameters)
! Works for all renormalizable QFTs (but can be cumbersome)
! Alternative (but equivalent!) formalism: Renormalized perturbation theory (→ today)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 150
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
1| ^ 4 -theory in d D 3 C 1 dimensions (→ QED below):
1 1 2 2 0 4
L 4 D .@ /2 m (7.54)
2 2 0 4Š
2| With D 4 D 4 N [Eq. (7.34)] and N D 0; 2; 4; : : : one finds the divergent
amplitudes:
Note that all vertices of 4 -theory have degree 4 so that only an even number of external
legs is possible.
D 4 D 4 unobservable vacuum energy shift (7.55)
D 4 D 2 ƒ2 C p 2 log ƒ (7.56)
D 4 D 0 log ƒ (7.57)
! 3 divergent quantities
! Absorb in 3 unobservable parameters: bare mass m0 , bare coupling 0 , fields
3| Recall:
iZ
Z
d 4x e ipx hjT .x/ .0/ji D C ::: (7.58)
p2 m2
The dots denote terms regular at p 2 D m2 .
Absorb unobservable Z in rescaled fields:
1
r p (7.59)
Z
Then
i
Z
d 4x e ipx hjT r .x/ r .0/ji D C ::: (7.60)
p2 m2
Note that this expression is no longer affected by Z ! 1 for ƒ ! 1 since we rescale
the field strength of r accordingly.
4| Lagrangian in new fields:
1 1 2 0 2 4
L 4 D Z.@ r /2 m Z 2 Z r (7.61)
2 2 0 r 4Š
5| Split terms into observable parameters and unobservable ones:
Physical parameters (fixed)
‚ …„ ƒ
1 1 2 2 4
L 4 D .@ r /2 m r (7.62)
2 2 4Š r
ı ı ı
Z ‚ …„m ƒ ‚ …„ ƒ
1 ‚ …„ ƒ 1 2 1
C .Z 1/.@ r /2 .m Z m2 / r2 .0 Z 2 / r4
2
„ 2 0 ƒ‚ 4Š …
Counterterms (cutoff-dependent)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 151
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
! ıZ , ım , and ı absorb unobservable, diverging shifts of bare and physical quantities
So far, we only redefined quantities and shuffled them around! No magic here.
6| Experimental input ! Renormalization conditions:
Š i
D C ::: (7.63)
p2 m2
2 3
6 7 Š
6
4
7
5 D i (7.64)
fc&a
E/
pi D.m;0
We need to force the theory to match the observed, physical parameters m and to
extrapolate from these and make non-trivial predictions.
Motivation:
• Eq. (7.63) includes two conditions: it fixes the pole of the propagator at the physical
mass m and the residue (and thereby the field strength) at 1. This enforces the
scaled fields r from above.
• Recall that in bare perturbation theory for the amplitude iM.p1 p2 7! p3 p4 / D
i0 C O.20 / as shown in Eq. (4.147). This motivates Eq. (7.64) which then is an
operational definition of the physical parameter as the measured amplitude for the
depicted scattering process at zero momentum. Note that the choice of momenta
(playing the role of experimental settings) is arbitrary. Changing these would
change the interpretation and the numerical value of , but not the predictions of
the theory.
7| Perturbation theory of Eq. (7.62) !
Feynman rules for renormalized perturbation theory
of 4 -theory in momentum space for S -matrix elements:
i
1: Edges: p D p 2 m2 Ci"
2: Vertices: D i
˝ D i ı
(7.65)
˝ D i.p 2 ıZ ım /
3: External lines: D1
4: Impose momentum conservation at all vertices
5: Integrate over all undetermined momenta
6: Divide by the symmetry factor
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 152
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
The propagator and the first vertex are the same as before, only that now the physical
mass and the physical coupling enter the perturbation series. Note that the counterterms
give rise to two additional vertices.
To understand the term for the two-leg vertex, retrace our derivation of Feynman rules
in Section 4.4 and recall that in momentum space the derivatives translate to p 2 (use
integration by parts).
8| Procedure for computing amplitudes:
(i) Sum all relevant diagrams built from the Feynman rules above.
(ii) If loop integrals diverge, introduce a regulator.
(iii) The results depend on the (yet undetermined) parameters fı g, the fixed physical
parameters m and e, and the regulator (ƒ or ").
(iv) Choose (“renormalize”) the parameters fı g such that the renormalization condi-
tions Eq. (7.63) and Eq. (7.64) are satisfied.
(v) With these fı g, the amplitude is finite, independent of the regulator, and depends
only on the physical parameters.
9| Bare perturbation theory (← beginning of this lecture) and renormalized perturbation theory
are equivalent and yield the same results.
Which one to choose depends on personal preference and the application.
10 | Example for renormalized perturbation theory in one-loop order:
i| ^ Amplitude
M.p1 p2 7! p3 p4 / D (7.66)
(7.67)
$ i C . i /2 ŒiV .s/ C iV .t / C iV .u/ i ı (7.68)
with Mandelstam variables s D .p1 C p2 /2 , t D .p3 p1 /2 , and u D .p4 p1 /2 .
Note that we include all one-loop diagrams with two physical vertices / but
only the tree-level diagram with one counterterm / ı . This is consistent because
ı D O.2 / as we will see below.
To construct the three one-loop diagrams, enumerate all possibilities to connect
two external momenta pi with i D 1; 2; 3; 4 at a one vertex.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 153
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
ii | Evaluate loop integral with dimensional regularization:
. i /2 iV .s/ D (7.69)
p 2 D .p1 C p2 /2 D s
. i /2 d 4k i i
Z
D (7.70)
2 .2/ k4 2 m .k C p/2 m2
2
Feynman parameter, Substitution, Wick rotation,
Dimensional regularization
2
8 9
C log.4/
Z 1
"!0 i < =
. i /2 dx " ; (7.71)
32 2 0
log m2 x.1 x/p 2
:
iii | Enforce renormalization condition Eq. (7.64) to determine ı :
Š
i MjsD4m2 ;tDuD0 D i (7.72)
solved by
2 V .4m2 / C 2V .0/
ı WD (7.73)
< 6 3 C 3 log.4/
8 9
2 Z 1
"!0
=
dx " (7.74)
32 2 0
log m2 x.1 x/4m2 2 log m 2
: ;
Here, the amplitude we want to calculate is the same that we need for the
renormalization condition. This is a special case! Note that ı depends only on ,
m and "; it is quadratic in the physical coupling which explains our perturbative
expansion above.
iv | Amplitude:
iM.p1 p2 7! p3 p4 / D i i 2 F .fpi gI m/ (7.75)
F : finite function of the momenta fpi g, parametrized by the physical mass m.
Important: The regulator " drops out!
Note that F D 0 for s D 4m2 and t D u D 0, as demanded by the renormalization
condition. The prediction of the theory is therefore not the amplitude for zero
momentum [pi D .m; 0/] E but the non-trivial dependency on fpi g for non-zero
momenta!
v| Enforce renormalization condition Eq. (7.63) to determine ıZ and ım :
a| Define
iM 2 .p 2 / WD (7.76)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 154
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
b| It follows along the same lines as for the electron self-energy
(7.77)
(7.78)
i
D (7.79)
p2 m2 M 2 .p 2 /
Š i 1
D C ::: (7.80)
p2 m2
c| Eq. (7.63) is equivalent to
dM 2 .p 2 / ˇˇ
ˇ
2 2 Š Š
and
ˇ
M .p /ˇp2 Dm2 D 0 D0 (7.81)
dp 2 ˇp2 Dm2
The first relation fixes the pole at p 2 D m2 , the second relation fixes the
Š
residue of this simple pole at 1, i.e., dpd 2 p 2 m2 M 2 .p 2 / p2 Dm2 D 1 1 .
d| In one-loop order:
iM 2 .p 2 / (7.82)
1 d 4k i
Z
D . i / C i.p 2 ıZ ım / (7.83)
2 .2/4 k 2 m2
Wick rotation, Dimensional regularization
i 1 .1 d=2/
$ d=2
C i.p 2 ıZ ım / (7.84)
2 .4/ .m2 /1 d=2
! Eq. (7.81) solved by
1 .1 d=2/
ıZ WD 0 and ım WD d=2
(7.85)
2 .4/ .m2 /1 d=2
Note that ıZ D 0 in one-loop order is a special case of 4 -theory since the
first term does not depend on p 2 . Note that ım is a diverging (for d ! 4)
function of the physical parameters (m and ) and the UV-regulator (d ).
Application to QED
We briefly summarize the analogous results for the renormalized perturbation theory of QED.
1| Original Lagrangian:
1
LQED D .F /2 C ‰.i @= m0 /‰ e0 ‰
‰A (7.86)
4
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 155
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
2| Interacting propagators:
iZ2 iZ3 g
D C ::: and D C : : : (7.87)
p
= m q2
3| Absorb Z2 and Z3 ! Renormalized fields:
1 1
‰r WD p ‰ and A
r WD p A (7.88)
Z2 Z3
4| Insert renormalized fields into Lagrangian:
1 1=2
LQED D Z3 .Fr /2 C Z2 ‰ r .i @= m0 /‰r e0 Z 2 Z 3 ‰r
‰r .Ar / (7.89)
4
1=2
5| Define Z1 WD Z2 Z3 ee0 with physical charge e
The physical charge e is defined by measurements at large distances, i.e., for q ! 0
(→ below).
6| Split Lagrangian into terms with physical parameters and counterterms:
1 2
LQED D .F / C ‰ r .i @= m/‰r e ‰ r ‰r .Ar / (7.90)
4 r
1
ı3 .Fr /2 C ‰ r .i ı2 @= ım /‰r eı1 ‰ r ‰r .Ar /
„ 4 ƒ‚ …
4 counterterms
with
ıi WD Zi 1 for i D 1; 2; 3 and ım WD Z2 m0 m (7.91)
7| Feynman rules (we omit external lines etc.):
i
Edges: p D p
= mCi"
ig
D q 2 Ci"
q
Vertices: D ie
(7.92)
D ı
˝ ie 1
˝ D i.g q 2 q q /ı3
˝ D i.p
= ı2 ım /
There are three additional counterterm vertices.
The counterterm for the photon two-leg vertex follows, similarly to the two-leg vertex of
4 -theory, with integration by parts.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 156
LE C T U R E 21 → P S : 32 3 – 32 8 , 3 3 0 – 3 32
8| 4 Counterterm coefficients ! 4 Renormalization conditions:
1. Fix electron mass to m:
" #
Š
D = D m/
i †.p D0
pDm
=
2. Fix residue of electron propagator to 1 (choose ‰r ):
" #
d†.p/
ˇ
d = ˇ Š
dp=
D i dp
=
ˇ D0
pDm
=
pDm
=
3. Fix residue of photon propagator to 1 (choose Ar ):
2 3 (7.93)
Š
D i ….q 2 D 0/
6 7
6
4
7
5 D0
g q 2 qq
q 2 D0
4. Fix electron charge to e:
2 3
6 7
Š
D i e .q D 0/ D ie
6 7
6 7
4 5
fc&a
qD0
These are redefinitions of †, … and in terms of the renormalized Feynman rules above.
The definition of involves now the physical charge e.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 157
LE C T U R E 22 → P S : 27 5 – 2 8 4
→ Topics of Lecture 22
1. The path integral as an alternative quantization method
2. Path integrals for scalar fields
→ Topics of Problemset 11
1. Dimensional regularization
2. Thomas-Fermi screening
8 Functional Methods
• So far:
Hamiltonian ! Canonical quantization ! Feynman rules (8.1)
The Hamiltonian is not Lorentz invariant (generates translations in time direction)!
• Alternative [22]:
Lagrangian ! Path integral ! Feynman rules (8.2)
The Lagrangian is Lorentz invariant (for a relativistic field theory)!
• Two descriptions of the same physics
• Application: Derivation of the photon propagator (easier with path integrals)
8.1 Path Integrals in Quantum Mechanics
p2
1| ^ Nonrelativistic particle in 1D: H D 2m C V .x/
i
2| ↓ Time evolution operator: U.xa ; xb I T / D hxb je „ H T jxa i
Known from canonical quantization in the Hamiltonian formalism.
3| Path integral (PI) formalism ! Alternative expression for U :
‚Functional
…„ ƒ Z xb
i F Œx.t /
X
U.xa ; xb I T / D e„ ƒ‚ … D Dx.t / e iF Œx.t/ (8.3)
x
All paths x.t/ with Pure phase „a ƒ‚ …
x.0/ D xa , x.T / D xb Functional integral
„ ƒ‚ …
Superposition principle
Paths are weighted with pure phases ! Interference (all paths are equivalent)
Functional integral = Integration over space of functions
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 158
LE C T U R E 22 → P S : 27 5 – 2 8 4
4| Conditions on F :
i| Describes the system
ii | Functional of path x.t /
iii | Classical path xcl .t / dominates (for „ ! 0):
X
U.xa ; xb I T / e iF Œx.t/ (8.4)
Paths close to xcl .t/
e + ec
Therefore
ˇ
ıF ˇˇ S 1
Z
Š
D0 ) F D D dt L.x.t // (8.5)
ıx xDxcl
ˇ „ „
This is the correspondence principle in action! The function S.xa ; xb I T / S Œxcl
with the classical trajectory xcl .t / that starts at xa at t D 0 and reaches xb
at t D T is known as ⁂ Hamilton’s principal function; it is the solution of
the ↑ ⁂ Hamilton-Jacobi equation. The correspondence principle then reads
Š i
U.xa ; xb I T / e „ S.xa ;xb IT / for „ ! 0 and embeds classical mechanics as well-
defined limit into quantum mechanics (in particular, it explains determinism as
emergent phenomenon of an underlying non-deterministic theory).
5| Propagation amplitude (Propagator):
Z x.T /Dxb
i ‹ i
U.xa ; xb I T / D Dx.t / e „ SŒx.t/ D hxb je „HT jxa i (8.6)
x.0/Dxa
We show the equivalence to canonical quantization for a free particle below.
So far, the PI is just a sketchy idea and not a well-defined mathematical concept!
6| Definition of PI via time slices:
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 159
LE C T U R E 22 → P S : 27 5 – 2 8 4
N 1Z
1 dx1 dxN 1 Y dxk
Z Z Z
1
Dx.t / WD lim D lim (8.7)
N !1 C" C" C" N !1 C" C"
kD1
T
with " D N and C" a constant (→ below)
C" determines the measure of the functional integral.
→ Example 8.1: Particle in potential V .x/
m 2
1| Lagrangian: L D P
2x V .x/
2| Action:
T N
X1 m .xkC1 xk /2
xkC1 C xk
Z
SD dt L "V (8.8)
0 2 " 2
kD0
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 160
LE C T U R E 22 → P S : 27 5 – 2 8 4
3| Recursion:
1
dx 0 i m.xb x 0 /2 xb C x 0
i
Z
U.xa ; xb I T / D exp "V (8.9a)
1 C" „ 2" „ 2
U.xa ; x 0 I T "/
Use V Œxb C x 0 =2 D V .xb / C O."/ under the Gaussian integral.
Z 1
dx 0 i m.xb x 0 /2
i
D exp 1 "V .xb / C : : :
1 C" „ 2" „
" #
@ .x 0 x /2 @2
b
1 C .x 0 xb / C C :::
@xb 2 @xb2
U.xa ; xb I T "/
(8.9b)
Compute Gaussian integrals with regularization.
Note that terms with odd powers of .x 0 xb / vanish!
r !
1 2„"
$
C" im
„ ƒ‚ …
Š
D1
To see this, consider both sides of the equation for " ! 0.
" #
i i „ @2 2
1 "V .xb / C " C O." / U.xa ; xb I T "/
„ 2m @xb2
4| ! PI measure:
r
2„"
C" D (8.10)
im
This is not generic but depends on L!
5| Use U.xa ; xb I T "/ D U.xa ; xb I T / " @T U C O."2 / and compare terms of
order ":
@ „2 @2
i„ @T U D 2m @x 2 C V .xb / U D H U
b
(8.11)
(Schrödinger equation)
Behold: We derived the Schrödinger equation and the quantized form of the
Hamiltonian from first principles (namely, the concept of weighting paths with
phases proportional to their classical action)!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 161
LE C T U R E 22 → P S : 27 5 – 2 8 4
6| Initial condition: set N D 1 (No integral!) !
1 i m 2
U.xa ; xb I "/ D exp .xb xa / C O."/ (8.12a)
C" „ 2"
r
im i m .xb xa /2
e „ 2" (8.12b)
2„"
"!0
! ı.xa xb / D U.xa ; xb I 0/ D hxb jxa i (8.12c)
With Eq. (8.11) we only showed that the path integral propagator obeys the same
differential equation as the time-evolution operator of canonical quantization; to
prove their equality as functions, we need to check their equality at some common
reference time, e.g., T D 0.
7| The last two steps conclude the proof of the second equality in Eq. (8.6) for
p2
H D 2m C V .x/.
Generalization
Now we reverse the reasoning:
We start with canonical quantization and derive the path integral.
Details: → Problemset 12
1| ^ Coordinates qi , conjugate momenta pi , Hamiltonian H.E E
q ; p/
2| ↓ Canonical quantization: Œqi ; pj D i „ıij ! U.E
qa ; qEb I T / D hE
qb je i HO T jE
qa i
We set „ D 1 to simplify equations.
i HO T i HO " i HO "
3| Time slicing: e D e„ e
ƒ‚ …
N
Insert N 1 identities 1k D qk j (k D 1; : : : ; N
R
4| d qEk jE
qk ihE 1) !
i HO "
hE
qkC1 je jE
qk i D hE
qkC1 j1 i HO " C O."2 /jE
qk i (8.13)
5| For HO D HO 1 .E
q / C HO 2 .p/
E (Proof: → Problemset 12):
d pEk qEkC1 C qEk
Z
O ; pEk exp i pEk .E
hE
qkC1 jH jE qk i $ H qkC1 qEk / (8.14)
2 2
Note that HO is an operator whereas H is a function!
This is more subtle for generic Hamiltonians H D H.p; q/ with terms like q 2 p 2 where
ordering is important, → Problemset 12.
The expression Eq. (8.14) maps a function H.E E on phase-space to an operator HO
q ; p/
on the Hilbert space (given in position representation). This map is known as ⁂ Weyl
transform or ⁂ Weyl quantization; ↑ p. 264 of Ref. [23]. The operator HO is given in
⁂ Weyl order, resolving the ordering ambiguities when transitioning from functions to
operators, → Problemset 12.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 162
LE C T U R E 22 → P S : 27 5 – 2 8 4
6| ! Hamiltonian phase-space path integral: (We restore „)
qa ; qEb I T / $
U.E (8.15)
2 3
q ;qEP /
¶L.E
qEb 6i Z T
Z ‚
…„ ƒ
7
E / exp 6
D qE.t / D p.t
6 P
dt pE qE H.E E 7
q ; p/
7
(8.16)
qEa 4„ 0 5
„ ƒ‚ … „ ƒ‚ …
Ek d p
dq Ek q ;p¶SŒE
SŒE E q
lim
Q R
k 2„
N !1
• The functional integral measure is called canonical measure and does not depend on
the system.
• In most cases (when H depends quadratically on p),
E the functional integration over
pE can be evaluated. Then one ends up with the simpler form Eq. (8.6) that sums
only over trajectories. The integration over momentum trajectories p.t
E / yields the
PI measure C" (→ Problemset 12).
• In Eq. (8.16), qE and pE do not satisfy the Hamiltonian EOMs (! Heisenberg
uncertainty principle); thus replacing the expression in the exponent by the
Lagrangian L.E q ; qEP / (resp. action S ŒE
q ) is not justified at this level (this is why
we use “¶” and not “D”) because we would have to use the classical relations
pE D @LP to do so. That is, the exponent should be seen as a functional S ŒE E of
q ; p
@qE
two independent variables qE and p.
E Classical solutions then indeed relate the two in
Š
the conventional way: ıS
D qEP @H D 0.ı pE @pE
• The Hamiltonian PI over phase space Eq. (8.16) is more general than the Lagrangian
PI over trajectories Eq. (8.6) [24].
8.2 Path Integrals for scalar fields
Identification: qi $ .x/
→ Example 8.2: Real scalar field
" Z #
b
i T 4
1 2 1
Z
i HO T
h b je j a i D D D exp d x P .r /2 V . /
a „ 0 2 2
(8.17a)
Evaluate -integration
" Z #
i T 4
Z
D D exp d x L. ; @ / (8.17b)
„ 0
• Lagrangian: L. ; @ / D 12 .@ /2 V . /
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 163
LE C T U R E 22 → P S : 27 5 – 2 8 4
• Boundaries: .x; E and .x;
E 0/ a .x/ E T / b .x/
E
• All symmetries of L are manifest in the PI formalism
This is not true for the Hamiltonian formalism which singles out a time direction!
• Abandon the Hamiltonian formalism and use Eq. (8.17b) to define the time evolution
• Goal: Derive correlation functions & Feynman rules directly from PIs
Here we only discuss correlation functions, for Feynman rules ↑ P&S pp. 284–289.
Correlation functions
1| We would like to evaluate the two-point correlator with path integrals:
Z .CT /D b
‹ R CT
d 4x L. /
hjT H .x1 / H .x2 / ji ! D .x1 / .x2 / e i T (8.18)
„ ƒ‚ … . T /D a „ ƒ‚ …
Operators Numbers
H are interacting Heisenberg field operators.
2| Split functional integral:
Z Z Z Z .x20 ;x/D
E 2 .x/
E
D .x/ D E
D 1 .x/ E
D 2 .x/ D .x/ (8.19)
.x10 ;x/D
E 1 .x/
E
3 0<LS
3| !
Z Z
Eq. (8.18) D E
D 1 .x/ E 1 .xE1 / 2 .xE2 /
D 2 .x/
„ ƒ‚ …
(i)
iH.T x20 / iH.x20 x10 / iH.x10 CT /
h b je j 2 i h 2 je j 1 i h 1 je j a i
„ ƒ‚ …„ ƒ‚ …„ ƒ‚ …
(ii) (iii) (iv)
(8.20)
Here we use only the definition of the propagator!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 164
LE C T U R E 22 → P S : 27 5 – 2 8 4
Use D 1 .x/E j 1 ih 1 j D 1 and S .xE1 /j 1 i D 1 .xE1 /j 1 i:
R
4|
Compare: qjqi
O D qjqi; S labels a Schrödinger field operator: S .x/
E D H .x;
E t D 0/.
(i) (ii) (iii) (iv) (8.21a)
x20 >x10 iH.T x20 / iH.x20 x10 / iH.x10 CT /
D h b je S .xE2 /e S .xE1 /e j a i (8.21b)
iH T iH T
D h b j e„ ƒ‚ (8.21c)
… T f H .x1 / H .x2 /g e„ ƒ‚ … j a i
!/jihj !/jihj
T !1.1 i"/
! C hjT H .x1 / H .x2 /ji (8.21d)
5| Result („ D 1):
hjT H .x1 / H .x2 /ji (8.22)
h R i
CT
D .x1 / .x2 / exp i T d 4x L. /
R
D lim h R i (8.23)
CT
T !1.1 i"/ D exp i T d 4x L. /
R
The denominator ensures independence of the boundaries at T ! ˙1, a and b .
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 165
LE C T U R E 23 → PS:294–298
→ Topics of Lecture 23
1. Quantization of the electromagnetic field
2. Fadeev-Popov gauge fixing
8.3 Application: Quantization of the Electromagnetic Field
ig
Goal: Apply PI formalism to derive the photon propagator k 2 Ci"
1| Action:
1
Z
S ŒA D d 4x .F /2 (8.24a)
4
jx j!1
Partial integration with A ! 0; use F D @ A @ A
1
Z
d 4x A .x/ @2 g @ @ A .x/
$ (8.24b)
2
Fourier transform
1 d 4k Q
Z
2 Q
D A .k/ k g C k k A . k/ (8.24c)
2 .2/4 „ ƒ‚ …
Set AQ .k/ D k ˛.k/ ! D 0 ! S ŒA D 0 ! DA e 0 D 1
R
2|
That’s bad!
3| Problem: Gauge invariance A ! A C 1e @ ˛
Integration over continuity of gauge-equivalent configurations A 0 , A / @ ˛
leads to divergence!
4| Solution: Count each physical configuration once (↑ Faddeev & Popov procedure [25])
Š
i| Gauge fixing: G.A/ D 0 (e.g. Lorenz gauge: G.A/ D @ A )
ii | Let A˛ WD A C 1e @ ˛, then
ıG.A˛ /
Z
˛
1 D D˛ ı.G.A // det (8.25)
ı˛
→ Note 8.1
To understand this, consider the n-dimensional ı-distribution:
" Z # "Y Z #
Y @ E
g
dai ı .n/ .g.E
E a// det D dgi ı .n/ .g/
E D1 (8.26)
@Ea
i i
@gE
Here, ı .n/ .g.E
E a// D ı.g1 .E a// and det
a// ı.gn .E @E
a
is the Jacobian of the
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 166
LE C T U R E 23 → PS:294–298
vector-valued map gE D g.E
E a/.
˛
iii | Assume that ıG.A
ı˛
/
is independent of A and ˛ (true for the Lorenz gauge!)
This cannot be satisfied for non-abelian gauge theories ! ↑ ⁂ Ghost fields
iv |
ıG.A˛ /
Z Z Z
iSŒA
DA e D det D˛ DA e iSŒA ı.G.A˛ // (8.27a)
ı˛
1
Substitute AQ D A˛ D A C @˛ ! D AQ D DA
e
Q
Use gauge invariance: S ŒA D S ŒA
ıG.A /˛ Z Z
Q
D det D˛ D AQ e iSŒA ı.G.A//
Q (8.27b)
ı˛
„ ƒ‚ … „ ƒ‚ …
D1 Only physically distinct configurations
The infinite integral over ˛ can be interpreted as the “volume” of the gauge orbit.
ıG.A˛ /
v | Choose G.A/ D @ A !.x/ ! det D det 1e @2
ı˛
Here !.x/ is an arbitrary scalar function (→ next step). @2 is a linear operator
on a function space; since the latter is infinite dimensional, think of @2 as an
“infinite-dimensional matrix”.
Z Z
1 2
Eq. (8.27b) D det @ D˛ DA e iSŒA ı.@ A !.x// (8.28)
e
(We renamed AQ as A.)
vi | True for any ! ! True for normalized linear combinations:
D1
‚ Z …„ ƒ
2
i d 4x !
R
Eq. (8.28) D N./ D! e 2 (8.29)
„ƒ‚…
Normalization „ ƒ‚ …
Linear combination
Z Z
1 2
det @ D˛ DA e iSŒA ı.@ A !.x//
e
So far, 2 R is arbitrary.
Z
1 2
D N./ det @ D˛ (8.30)
e
Z 2
4 .@ A /
Z
iSŒA
DA e exp i d x
2
„ ƒ‚ …
New term (breaks gauge symmetry)
Note that breaking gauge invariance in the new effective Lagrangian does not alter
expectation values of physical (and therefore gauge-invariant) operators. Different
Lagrangians can describe the same physics!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 67
LE C T U R E 23 → PS:294–298
vii | O gauge invariant operator: O.AO˛ / D O.A/,
^ O.A/ O then
O
hjT O.A/ji (8.31)
n R h io
CT 1
DA O.A/ exp i T d 4x L 2 .@ A /2
R
D lim n R h io (8.32)
CT 1
T !1.1 i"/ DA exp i T d 4x L 2
R
.@ A /2
This follows along the same lines as Eq. (8.23). The gauge-invariance of the
Q
operator is needed in step Eq. (8.27b) where we substitute A by A.
Important: The unkown and diverging prefactors have canceled!
5| New action (same calculation as in step 1):
1 1
Z
Q 4 2 2
S ŒA D d x .F / .@ A / (8.33a)
4 2
jx j!1
Partial integration with A ! 0; use F D @ A @ A
1
Z
$ d 4x A .x/Œ@2 g .1 1 /@ @ A .x/ (8.33b)
2
Fourier transform
1 d 4k Q
Z
D A .k/Πk 2 g C .1 1 /k k AQ . k/ (8.33c)
2 .2/4 „ƒ‚…
New!
Skip first and second step.
! Argument of Step 2 no longer applies!
6| Propagator:
DF .x y/ D hjT A .x/A .y/ji (8.34)
! hjAQ .k/AQ .q/ji D 0 for k ¤ q (due to translation invariance)
Therefore
DQ F .q/ D hjAQ .q/AQ . q/ji (8.35a)
Use AQ . q/ D .AQ .q// since A is real.
Add Ci " for regularization to the action.
M .k/ (symmetric)
R d 4k ‚ …„ ƒ
2 1 Q
DA AQ .q/AQ . q/ expf 2i .2/ Q
R
4 A .k/ Πk g C .1 /k k A . k/g
D n R 4 o
DA exp 2i .2/ d k Q
1 /k k AQ . k/
R
2 C .1
4 A .k/Πk g
(8.35b)
Y
PI measure: DA D d.Re AQ .k// d.Im AQ .k// .
I k;k 0 >0
Diagonalize M , complete the square, and evaluate Gaussian integrals.
Details → Problemset 12.
1
$ i.M .q// (8.35c)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 168
LE C T U R E 23 → PS:294–298
Finally
qq
i
DQ F .q/ $ g .1 / 2 (8.36)
q2 C i " q
Check that this is the inverse of M .q/!
7| Gauges:
• Set D 1:
ig
DQ F .q/ D (Feynman gauge) (8.37)
q2 C i "
This form is Lorentz invariant since g is.
• Set D 0:
qq
i
DQ F .q/ D g (Landau gauge) (8.38)
q2 C i " q2
→ Note 8.2
• Correlators of gauge invariant operators are independent of .
• For ! 1 we have k 2 g C k k D M .k/.
Since . k 2 g C k k /k D 0, the inverse M 1 .k/ does not exist!
• k 2 lim D g k k =k 2 D T is a projector on transversal fields:
!1 M
T k D 0 and T T D T
! The (original) divergence is due to longitudinal gauge fields: AQ .k/ D k ˛.k/.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 169
LE C T U R E 24 → PS:481–491
→ Topics of Lecture 24
1. Non-abelian gauge theories
2. The Yang-Mills Lagrangian
→ Topics of Problemset 12
1. Propagator in the path integral formalism
2. Path integral and Weyl order
9 Non-Abelian Gauge Theories
Motivation:
1| Fact I: So far the only interactions considered where 4 and ‰ ‰A .
Can we construct theories (of massless vector bosons) with more complicated interactions
like A4 or .@A/A2 ? In particular, is it possible to construct theories with force carriers (=
vector bosons) that are themselves charged (photons are not!)?
2| Fact II:
i| Massless particles carry helicity representations [ISO.2/] and not spin [SO.3/]
(⁂ Wigner’s Little groups ↑ p. 69ff. of Ref. [1]). This is why photons only have two
transversal polarizations despite being the excitations of a vector (= spin 1) field A .
ii | Unitarily transforming the single-particle helicity modes yields for a quantized field
in the vector representation ƒ (↑ p. 246ff. of Ref. [1])
1 1
U.ƒ/A .x/U .ƒ/ D .ƒ / A .ƒx/ C @ .x; ƒ/ : (9.1)
„ ƒ‚ …
Gauge!
Compare this to the transformation of a Dirac field [Eq. (3.58)] in the bispinor
representation ƒ 1 :
2
1
U.ƒ/‰.x/U .ƒ/ D ƒ 1 1 ‰.ƒx/ : (9.2)
2
iii | Conclusion: To preserve unitarity and Lorentz invariance, theories of massless vector
bosons (e.g., photons, gluons) must be gauge theories! (This is not true for massive
vector bosons, ↑ ⁂ Proca action.)
3| Fact III: Historically, the renormalizability of QED was believed to be rooted in the gauge
invariance of the Lagrangian; this was proven later by ’t Hooft [26, 27]. As we can
only make sense of renormalizable theories, this relation between gauge invariance and
renormalizability makes a strong case for studying more general gauge theories than QED.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 70
LE C T U R E 24 → PS:481–491
4| Fact IV: Quantization of the electromagnetic field is complicated by negative-norm states
of the time component A0 :
E D ig ı .3/ xE yE but g00 D gi i
E … .y/
A .x/; (9.3)
(↑ Gupta-Bleuler quantization of Maxwell theory).
Observation: In Maxwell theory, negative-norm states and longitudinal polarization
states decouple from the transversal polarization states. This is guaranteed by the gauge
symmetry (↑ Ward identity).
5| Idea: Generalize Maxwell theory (or QED, if matter is involved) to gauge theories with
other symmetry groups.
6| Spoiler: This type of theory turns out to describe all fundamental forces of nature (except
gravity); it is the foundation of the Standard Model!
We start with a thorough analysis of the gauge symmetry of QED. In a second step, we
generalize our findings to non-abelian gauge groups. This yields the famous ⁂ Yang-Mills
theories.
9.1 The Geometry of Gauge Invariance
1| ^ Local U.1/ symmetry G of Dirac field:
Q
‰.x/ D e i˛.x/ ‰.x/ (9.4)
for arbitrary ˛.x/ W R1;3 ! R
(In the following, a tilde always denotes symmetry-transformed quantities.)
2| Goal: Construct invariant Lagrangian
3| No problem without derivatives:
All terms invariant under global U.1/ transformations allowed (e.g. ‰.x/‰.x/)
4| ^ Directional derivative along n 2 R1;3 :
‰.x C "n/ ‰.x/
n @ ‰ WD lim (9.5)
"!0 "
‰.x C "n/ and ‰.x/ transform differently under G
! n @ ‰ has no simple transformation law under G
(! Not a useful building block for symmetric Lagrangians.)
To see why, calculate n @ ‰;Q the result is not just e i˛.x/ n @ ‰. This makes the
construction of invariant terms for a symmetric Lagrangian very complicated.
5| Thus we need a sensible way to compare fields at different points:
Postulate the existence of a “comparator” U W R1;3 R1;3 ! C with transformation
UQ .y; x/ D e i˛.y/ U.y; x/e i˛.x/
and U.y; y/ D 1 (9.6)
[we require U.y; x/ D e i .y;x/ ]
! ‰.y/ and U.y; x/‰.x/ have same transformation law and therefore can be meaning-
fully compared.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 171
LE C T U R E 24 → PS:481–491
Note that we do neither prove the existence of U nor provide its construction; we simply
take such a function for granted. For more details, ↑ fiber bundles in differential geometry.
In particular, the “comparator” relates to the concept of ↑ parallel transport between
fibers of principal bundles.
6| Covariant derivative:
‰.x C "n/ U.x C "n; x/‰.x/
n D ‰ WD lim (9.7)
"!0 "
7| Assume U.y; x/ continuous !
U.x C "n; x/ D 1 i e "n A .x/ C O."2 / (9.8)
e: arbitrary constant (rescales A )
A : new vector field = (gauge) connection
ı
8| Eq. (9.8) in Eq. (9.7) !
D ‰.x/ D @ ‰.x/ C i eA ‰.x/ (9.9)
ı
9| Eq. (9.8) in Eq. (9.6) !
1
AQ .x/ D A .x/ @ ˛.x/ (9.10)
e
ı
10 | !
DQ ‰.x/
Q D e i˛.x/ D ‰.x/ (9.11)
! D‰ transforms like ‰ (this makes it easy to construct invariant terms!)
! All terms invariant under global U.1/ transformations allowed if @ is replaced by D
[e.g. ‰.x/.i D/‰.x/]
=
11 | Conclusion:
Local symmetry ! Gauge field A needed for covariant derivatives
Note that we did not put in the gauge field by hand! It automatically emerges as a
necessary ingredient for terms that are locally symmetric and involve derivatives.
12 | Last but not least: Kinetic energy term for A ?
(= Locally invariant term that depends only on A and its derivatives.)
i| ^ Locally invariant loop (= local limit of a Wilson loop):
(1E and 2E are two arbirary orthogonal unit vectors.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 72
LE C T U R E 24 → PS:481–491
U .x/ WD E
U.x; x C "2/
E x C "1E C "2/
U.x C "2; E
U.x C "1E C "2;
E x C "1/
E
E x/
U.x C "1;
(9.12)
Q D U by construction [remember Eq. (9.6)]
!U
ii | Use
h " i
U.x C "n; x/ D exp i e "n A x C n C O."3 / (9.13)
2
To derive this form, recall U.y; x/ D e i .y;x/ and U.x; x/ D 1. Without changing
our definition of A in Eq. (9.8), we can restrict U to the form
U.y; x/ D exp i e "n A .x C "C / C O."3 /
(9.14)
where C is arbitrary. The additional constraint U .x; y/ D U.y; x/ then deter-
mines C D 12 n. Relaxing this assumption introduces additional vector fields (for
the higher orders) that render the theory more complicated than necessary.
!
U .x/ $ 1 i "2 e Œ@1 A2 .x/ @2 A1 .x/ CO."3 / (9.15)
„ ƒ‚ …
DWF12
!
F WD @ A @ A (Field-strength tensor) (9.16)
is locally gauge invariant by construction
Note that a similar construction (that is, parallel transport along a small closed
loop) gives rise to the notion of curvature of Riemannian manifolds. Consequently,
the Riemann curvature tensor R of general relativity plays a similar role than the
field-strength tensor F of Maxwell theory.
13 | Most general gauge (and Lorentz-) invariant Lagrangian in D D 3 C 1:
• Gauge invariant !
Constructed from ‰, D ‰, F , @ F etc. and globally U.1/-invariant
• Relativistic ! Lorentz scalar
• Renormalizable ! Terms of mass dimension at most 4
Otherwise the coupling constants of such terms have negative mass dimension
(because the Lagrangian density has mass dimension 4) and render the theory
non-renormalizable.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 73
LE C T U R E 24 → PS:481–491
!
1
=
L D ‰.i D/‰ m‰‰ .F /2 (9.17)
4
c1 "˛ˇ F˛ˇ F (9.18)
„ ƒ‚ …
breaks P and T
2
C c2 .‰‰/ C : : : (9.19)
„ ƒ‚ …
non-renormalizable
! Most general P =T -symmetric Lagrangian: Minimally coupled Maxwell-Dirac
! QED
Note that "˛ˇ is a pseudo tensor, in contrast to g ˛ g ˇ ; therefore the c1 -term is a
pseudo scalar; i.e., it is only invariant under SOC .1; 3/ but not O.1; 3/.
9.2 The Yang-Mills Lagrangian
Goal: Replace local symmetry group U.1/ by non-abelian Lie group G
Examples: SO.3/, SU.2/, SU.3/, …
Details: → Problemset 13
1| ^ Lie group G represented by n n unitary matrices V
Typically, we consider
the
fundamental (or defining) representation of matrix Lie groups,
j
e.g., V D exp i !j 2 for G D SU.2/ with n D 2 and j Pauli matrices.
2| Fields ‰ D .‰1 ; : : : ; ‰n /T are ⁂ n-plets of Dirac fields ‰i :
‰ W R1;3 ! C 4 ˝ C n ' C 4n and transform as
Q
‰.x/ D V .x/‰.x/ D Vij .x/‰j .x/ (9.20)
with V W R1;3 ! G arbitrary
Note that i; j are not spinor- but G-indices; each ‰j is a Dirac bispinor!
3| G Lie group ! Lie algebra g with N Hermitian generators t a (n n-matrices, a D
1; : : : ; N ) that obey
h i
t a ; t b D if abc t c Einstein notation! (9.21)
with ⁂ structure constants f abc 2 C.
The structure constants define the Lie algebra. One can always choose a basis ft a g such
that they are completely antisymmetric in the three indices. Note that a D 1; : : : ; N
is finite since G is assumed to be compact and the matrix representations of t a are
Hermitian because the matrix representations V are assumed to be unitary. ↑ P&S
pp. 495–502 for details.
!
V .x/ D exp i ˛ a .x/t a D 1 C i ˛ a .x/ t a C O.˛ 2 /
(9.22)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 74
LE C T U R E 24 → PS:481–491
4| The “comparator” is now a n n unitary matrix with transformation
UQ .y; x/ D V .y/U.y; x/V .x/ and U.y; y/ D 1 (9.23)
U.x C "n; x/ D 1 C ig "n Aa t a C O."2 / (9.24)
g: arbitrary constant (rescales Aa )
Aa : N vector fields (= gauge connections, one for each generator t a )
The “comparator” acts on ‰, i.e., on the representation of G given by V ; thus its
infinitesimal action must be generated by the representation of the corresponding Lie
algebra ft a g.
5| Eq. (9.7) ! Covariant derivative:
D $ @ igAa t a (9.25)
It is often written A Aa t a so that A is a Lie-algebra valued field (a n n- matrix).
6| Transformation of Aa :
i| Eq. (9.24) in Eq. (9.23) !
1 C ig "n AQa t a D V .x C "n/ 1 C ig "n Aa t a V .x/
(9.26)
ii | Use
h i
V .x C "n/V .x/ $ 1 C "n V .x/ @ V .x/ C O."2 / (9.27)
[Recall that 0 D @ 1 D @ .V V / D .@ V /V C V .@ V /.]
to show
i
A t D V .x/ A t C @ V .x/
Qa a a a
(9.28)
g
This transformation law is exact, i.e., true for any V . Note that @ acts only on V
and not on what comes after AQa t a !
iii | @ V .x/ is not easy to evaluate (non-commuting operators in the exponent!)!
^ Infinitesimal transformation V .x/ 1:
V .x/ D 1 C i ˛ a .x/t a C O.˛ 2 / (9.29)
a a 2
and @ V .x/ D i @ ˛ .x/t C O.˛ / (9.30)
With f cba D f abc !
ı 1
AQa Aa C @ ˛ a C f abc Ab ˛ c (9.31)
g „ ƒ‚ …
New!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 175
LE C T U R E 24 → PS:481–491
This transformation law is only true for infinitesimal transformations V 1 (hence
the “”). For an abelian Lie group [such as U.1/], it is f abc 0 and this
expression is exact.
7| Eq. (9.28) in Eq. (9.25) ! Transformation of D ‰:
DQ ‰
Q $ V D ‰ (9.32)
Use again .@ V /V D V .@ V / to show this.
! D ‰ transforms like ‰
! ‰D ‰ is gauge-invariant and ‰ D‰
= is both gauge- and Lorentz invariant.
8| Last but not least: Kinetic energy term for Aa ?
Here, we follow an alternative approach to find such terms (without using the infinitesimal
loop construction from above):
Q DQ ‰
i | Iteration of Eq. (9.32) implies D Q D V D D ‰
DQ ; DQ ‰ Q DV D ; D ‰ D V D ; D V ‰
Q
) (9.33)
DQ ; DQ
) DV D ; D V (9.34)
ii | On the other hand: Eq. (9.25) !
a a
ig F t WD D ; D (9.35)
„ƒ‚…
DWF
a
with F $ @ Aa @ Aa C gf abc Ab Ac (9.36)
a : N field-strength tensors
F
Note that F Fa t a is a n n-matrix, not a derivative!
iii | Eq. (9.34) !
a a
FQ D FQ t D VF V (9.37)
! F is no longer gauge invariant
Cf. Maxwell theory where V and t a are 1 1-matrices so that FQ D F .
iv | Simplest invariant term:
1 1 h i
Tr F 2 a a
Tr .F t /.F b t b /
LYM D (9.38)
2 2
Use Tr.t a t b / D 12 ı ab
1 a 2
D .F / (Yang-Mills theory) (9.39)
4
One can always choose a basis ft a g of g where Tr.t a t b / D 12 ı ab , ↑ P&S p. 498ff.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 76
LE C T U R E 24 → PS:481–491
→ Note 9.1
Important:
F 2 .@A/2 C f .@A/AA C f 2 AAAA (9.40)
„ ƒ‚ …
Interactions
! Interacting QFT for f ¤ 0 (= non-abelian)!
! Gauge bosons scatter off each other
Example:
Quantum Chromodynamics [G D SU.3/] (→ last lecture)
Gauge bosons = Gluons ! Pure gluon vertices in Feynman diagrams:
! Bound states of (typically 2 or 3) gluons: Glueballs (not yet directly
observed, but progress has been made lately [28])
That the mass of glueballs cannot be arbitrarily small is (part of ) one
of the Millennium Prize Problems of the Clay Mathematics Institute: the
“Yang–Mills Existence and Mass Gap” problem, ↑ https://www.claymath.
org/millennium-problems.
9| Couple Dirac fermions to Yang-Mills gauge field:
1 a 2
=
LYMCD D ‰ i D m ‰ .F / (9.41)
4
Two parameters:
m: Fermion mass
g: Coupling constant (hidden in D and F 2 )
This is the most general Lagrangian that is …
• gauge invariant
• Lorentz invariant
• renormalizable
• P - and T -symmetric
! Yang-Mills theories describe all fundamental forces of the standard model!
(Without the Dirac mass term in Eq. (9.41) though, and not necessarily respecting the P -
and T - symmetries; → last lecture.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 177
LE C T U R E 24 → PS:481–491
→ Note 9.2
Let us be precise about the symbols:
= D
D D D @ 1n igAa a
t (9.42)
where
a
t ˝ ta D ˛ˇ
a
tmn D. a
t /.˛;m/.ˇ;n/ (9.43)
so that
‰ W R1;3 ! C 4 ˝ C n ' C 4n (9.44)
carries a four-dimensional bispinor representation of the Lorentz group, ƒ 1 D
2
i i
exp 2 ! S
with S D 4 Π; , and the representation V of the gauge
group G, V D exp .i ˛ a t a /. Then it follows in particular
Q D .V ‰/
‰ 0
D ‰ V ˝ 0
D ‰ 0
V D ‰V (9.45)
so that ‰‰ is gauge invariant.
→ Note 9.3
The mass term A2 is not allowed as it is not gauge invariant!
Recall that 12 m 2 was responsible for the mass gap of 4 -theory and m‰‰ for the mass
of Dirac fermions.
! Gauge bosons of Yang-Mills theories are massless.
For QED and QCD, this is fine: The photon and gluons are massless.
Problem:
The weak interaction is short-ranged, i.e., its gauge bosons W ˙ and Z have mass!
Solution: Higgs mechanism (→ next lecture)
↑ Notes
N ICO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 178
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
→ Topics of Lecture 25
1. Mass generation with the Higgs mechanism
2. The Goldstone theorem
10 Excursions
10.1 The Higgs Mechanism
Motivation:
• Problem 1: Recall that we cannot add a mass term A2 to the Yang-Mills Lagrangian as it
would break gauge invariance (← note at the end of last lecture).
How do the W ˙ and Z bosons that mediate the short-ranged weak interaction obtain their
observed masses?
• Problem 2: Although we have shown that a Dirac mass term ‰‰ is allowed in general
Yang-Mills theories, in the particular case of the standard model, it is forbidden (→ next
lecture).
How do quarks and leptons gain their observed masses?
Solution to both problems: Higgs mechanism
(For simplicity, will consider only classical field theories and skip their quantization as the
crucial mechanisms are already present at this level. Here we will only find a solution to Problem
1, how the Higgs field couples to fermions and solves Problem 2 will become clearn when we
discuss the Standard Model.)
For the quantization of gauge theories with Higgs field, ↑ Chapter 21 of P&S (p. 731ff.)
10.1.1 Abelian Example: The Standard Approach
This approach follows loosely the essay ↑ http://philsci-archive.pitt.edu/9295/
1/Spontaneous_symmetry_breaking_in_the_Higgs_mechanism.pdf (see also references
therein); this is also roughly the approach of P&S, ↑ pp. 690–692.
Goal: ^ Abelian gauge theory to understand the Higgs mechanism
1| ^ Maxwell theory coupled to a complex scalar field:
1
LD .F /2 C jD j2 V . / (10.1)
4
with potential V . / D 2 j j2 C j j4 (10.2)
and D D @ C i eA
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 179
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
2| L is Invariant under the U.1/ gauge transformations:
1
Q
.x/ D e i˛.x/ .x/ and AQ .x/ D A .x/ @ ˛.x/ (10.3)
e
3| ^ V . / for > 0 in the complex plane 2 C:
• 2 > 0: Unique minimum with h i D 0
• 2 < 0 ! Mexican hat potential:
Degenerate minima with non-zero vacuum expectation value (VEV)
s
2
0 WD h i and v WD j 0 j D ¤0 (10.4)
2
! Ground states are not symmetric under global phase rotations
! Spontaneous symmetry breaking (SSB) of the global U(1) symmetry
4| Aside: The Goldstone theorem:
If a global, continuous symmetry is spontaneously
broken, there is one massless scalar (= Spin-0)
(10.5)
particle for each broken symmetry generator; these
particles are known as (Nambu-)Goldstone bosons.
“Proof by picture:”
Long wavelength deformations of the field with the broken symmetry generator (red
arrows) cost arbitrary low energy ! Gapless Goldstone mode
Examples:
• Breaking of translation and rotation invariance in crystals
! Transversal and longitudinal phonons
This is a subtle example. There are in total 6 generators that are broken: 3
translations Px ; Py ; P´ and 3 rotations Lx ; Ly ; L´ —but there are only 3 (not 6!)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 180
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
Goldstone modes, namely two transversal and one longitudinal phonon. The reason
is that the Euclidean group of translations and rotations is E.3/ D O.3/ Ë T.3/
with rotations O.3/ and translations T.3/ D R3 and not E.3/ D O.3/ T.3/
(“Ë” denotes the ↑ ⁂ semidirect product of groups); in particular, the generators of
rotations Li (= angular momentum operators) and translations Pi (= momentum
operators) do not commute. Thus for nonrelativistic field theories, the above
statement is only true if the different generators commute; ↑ Ref. [29] for details on
counting the Goldstone modes correctly in such theories.
• Breaking of rotation symmetry in a ferromagnet
! Magnons (= Spin waves)
But there is one notable exception:
In conventional superconductors the U.1/ symmetry (generated by particle number
conservation) is broken spontaneously (↑ Ginzburg-Landau theory) – but there is no
massless Goldstone boson! (Recall that the photon in superconductors is short-ranged
and therefore massive; it is also not a scalar.)
! How can the Goldstone theorem fail?
! Answer: Gauge symmetry & Higgs mechanism (→ below)
5| Assume that h i D 0 D v breaks the global U.1/ symmetry
! Expand in small fluctuations around h i:
.x/ D Œv C h.x/ e i'.x/ (10.6)
with two real fields:
h.x/ : Higgs field and '.x/ : Goldstone boson (10.7)
(The terms “field”, “mode” and “boson” are often used interchangeably.)
!
1 h ih i
LD .F /2 C .@ C i eA /.v C h/e i' .@ i eA /.v C h/e i'
(10.8)
4
2 .v C h/2 .v C h/4
1
$ .F /2 C e 2 v 2 A2 C .@ h/2 m2h h2 (10.9)
„ 4 ƒ‚ … „ ƒ‚ …
Massive gauge field (Yay!) Higgs field with mass m2h D 4v 2
C v 2 .@ '/2 C 2ev 2 .@ '/A C „ƒ‚…
:::
„ ƒ‚ … „ ƒ‚ …
Massless Goldstone mode Quadratic coupling Interactions
The interactions include terms cubic and quartic in the dynamical fields ', h and A .
Use Eq. (10.4) to write 2 D 2v 2 which simplifies the expression and explains the
resulting mass term of the Higgs field.
Note that this Lagrangian is still gauge invariant under the gauge transformation
1
'Q D ' C ˛ and AQ D A @ ˛ and hQ D h : (10.10)
e
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 181
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
6| Fix the gauge in the
Š
unitary gauge D , '0 (10.11)
with the gauge transformation ˛.x/ D '.x/:
i' 1
Q D e and AQ D A C @ '.x/ :
(10.12)
e
As the gauge is now fixed, the theory has no longer a gauge symmetry! Indeed, D
is violated by the transformation Q D e i˛.x/ . Note that the local gauge symmetry is
lost not because of SSB but because of explicit gauge fixing!
!
1 Q 2 Q 2 m2 hQ 2
LQ D .F / C e 2 v 2 AQ2 C .@ h/
4 „ ƒ‚ h …
Massive Higgs field
„ ƒ‚ …
Massive gauge field
C Interactions (10.13)
! Goldstone mode ' has disappeared!
Reason: ' is a pure gauge dof and therefore not physical!
Q '/
The gauge symmetry of the theory manifests in the relation L.AQ ; h; Q D L.A ; h; '/;
Q Q Q 'Q 0/.
here we fixed the gauge by demanding 'Q 0 and defining L.A ; h/ WD L.AQ ; h;
Q
L. Q describes no longer a gauge theory since L.
Q AQ ; h/ Q ¤ L.A
Q AQ ; h/ Q ; h/ for general
gauge transformations.
• This is what one means by
“The Goldstone boson is ’eaten’ by the Gauge boson to give it a mass.”
Personally, I do not like this metaphoric description as a didactic “auxiliary
structure” for a mathematically subtle mechanism because it explains nothing and
makes only sense if you already understood the math.
• This explains how the Goldstone theorem can fail for gauge theories. Conventional
superconductivity is therefore a non-relativistic example for the Higgs mechanism
where the Goldstone mode vanishes and instead the gauge boson (the photon, now
a quasiparticle excitation) obtains a mass m which leads to the Meissner effect;
the London penetration depth is then given by L / m 1 . Historically, this
observation in condensed matter physics motivated the application of the Higgs
mechanism to the problem of mass generation in high-energy physics.
7| Consistency check: Counting physical degrees of freedom:
#(dof ) before SSB D 2 (massless vector boson) C 2 (complex scalar field) D 4 (10.14)
#(dof ) after SSB D 3 (massive vector boson) C 1 (real scalar Higgs field) D 4 (10.15)
! We did not loose any dof but merely “mixed” them differently!
Note that a massless vector boson (like the photon) has only two transversal polarizations
(it has helicity h D ˙1). By contrast, a massive vector boson has an additional
longitudinal polarization (it has spin m D ˙1; 0).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 182
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
→ Note 10.1
• We have seen that the Goldstone theorem is not valid for gauge theories (since the
Goldston boson can become “pure gauge”).
• The Higgs mechanism also describes conventional superconductivity as spontaneous
U.1/ symmetry breaking in a charged superfluid (↑ Ginzburg-Landau theory).
In a superconductor, the photon (then a quasiparticle) acquires a mass and can no
longer propagate (↑ Meissner effect).
• There is also an intuitive picture how the Goldstone theorem fails in the presence of
a gauge field:
The proof of the Godstone theorem relies on the absence of long-range interactions (like
the Coulomb interaction). Only then, a massless Goldstone boson can be predicted.
However, coupling a (yet massless) gauge field to the (yet U.1/-symmetric) complex
scalar field adds exactly such long-range interactions between fluctuations of the
scalar. Due to these long-range interactions, the long-wavelength fluctuations of
the real mode of the scalar field “parallel” to the symmetry—that under normal
circumstances give rise to the massless Goldstone mode—develops a mass gap and
mixes with the gauge bosons. The result is a massive Spin-1 gauge boson, now a
collective “quasiparticle” excitation of the former gauge field and the Goldstone
mode of the scalar. The other real mode of the complex scalar that is “orthogonal”
to the symmetry gives rise to the Higgs boson.
• The Higgs mechanism is sometimes explained as “spontaneous breaking of a gauge
symmetry.” This is a misleading statement as gauge symmetries are redundancies of
our mathematical description; breaking a gauge symmetry should consequently not
lead to observable phenomena. As the mass generation due to the Higgs mechanism
is clearly observable, it cannot be rooted in the breaking of a gauge symmetry. In
addition, there is Elitzur’s theorem [30] that rigorously forbids SSB for local (gauge)
symmetries.
Indeed, there are equivalent descriptions of the Higgs mechanism that circumvent
the concept of “gauge symmetry breaking” altogether. Here a few references for
the interested student:
– A gauge-invariant treatment of the Higgs mechanism (for the weak interaction)
is given in Ref. [31].
– A gauge-invariant treatment of the U.1/ symmetry breaking in superconduc-
tors is discussed in Ref. [32].
– A few general remarks on the impossibility of spontaneously breaking gauge
symmetries can be found in ↑ http://web.physics.ucsb.edu/~d_else/
gauge_rant.pdf.
→ Gauge-invariant approach below
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 183
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
10.1.2 Bonus: A Gauge-Invariant Approach
This approach is based on Chapter 6.1 (p. 105ff.) of Ref. [33].
1| ^ Again Eq. (10.2):
1
LD .F /2 C jD j2 2 j j2 j j4 (10.16)
4
2| Let 2 < 0 (= symmetry-broken phase) ! Classical ground state (= vacuum):
0 .x/ D e i˛.x/ 0 (10.17)
q
2
with ˛.x/ arbitrary (wlog ˛.x/ 0) and j 0 j D 2
D v ¤ 0 (wlog 0 D v).
3| ^ Small fluctuations around 0 and introduce the new real fields h.x/, '.x/ and B .x/:
1
Œv C h.x/ e i'.x/ and B .x/ A .x/ C @ '.x/ (10.18)
e
Note that '.x/ is only well-defined if .x/ ¤ 0 everywhere and we can ignore the
ambiguity ' D ' C 2; this is true for small fluctuations around the vacuum 0 D v.
! Gauge transformations:
'Q D ' C ˛ ! pure gauge = only gauge dof (10.19)
hQ D h ! gauge invariant = only physical dof (10.20)
BQ D B ! gauge invariant = only physical dof (10.21)
! B is not a gauge field as it is gauge invariant!
Indeed:
1 1
BQ D AQ C @ 'Q D A C @ .'Q ˛/ D B (10.22)
e e
Compare this to
Q D e i˛ ! gauge dependent = physical and gauge dof (10.23)
1
AQ D A @ ˛ ! gauge dependent = physical and gauge dof (10.24)
e
4| Express Lagrangian in new fields:
Š
1 B DA
L$ .B /2 C e 2 v 2 B2 C .@ h/2 m2h h2 C : : : D (10.13) (10.25)
4
with B WD @ B @ B $ @ A @ A D F (10.26)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 184
LE C T U R E 25 → P S : 6 8 9 – 6 92 , 70 0 – 705
• Gauge dof ' drops out and is unconstrained by the Lagrangian
• L is manifestly gauge-invariant (Note that L D LQ only if we fix the unitary gauge
Š
B D A [this is a constraint on A ].)
• B is a massive vector boson
• h is a massive Higgs mode
5| Take-home-message:
The crucial ingredient of the Higgs mechanism is the non-zero vacuum expectation value of
the Higgs field 0 D v which can be explained by the spontaneous breaking of a global,
continuous symmetry. However:
There is no spontaneous breaking of local gauge symmetries in
the Higgs mechanism.
As a matter of fact, local gauge symmetries can never break
spontaneously (↑ Elitzur’s theorem [30]), they are a (10.27)
consequence of redundancies in our mathematical description.
In particular, they do not give rise to conserved charges
(↑ Noether’s second theorem [34, 35]).
→ Note 10.2
The Higgs mechanism can be straightforwardly generalized to non-abelian gauge symme-
tries.
→ Next lecture for the electroweak interaction with SU.2/ U.1/ gauge symmetry.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 185
LE C T U R E 26 → P S : 70 0 – 727
→ Topics of Lecture 26
1. Field content and gauge symmetries of the Standard Model
2. The Glashow-Weinberg-Salam theory
3. Quantum chromodynamics
4. Summary of all fields/particles in the Standard Model
→ Topics of Problemset 13
1. Non-abelian gauge theories and the Yang-Mills Lagrangian
10.2 The Standard Model
This section does not follow P&S but is a collage of various sources.
10.2.1 Preliminaries
1| Define the chiral projectors
1 5 Weyl 0 0 1 5 Weyl12 0
PR WD .14 C / D and PL WD .14 / D (10.28)
2 0 12 2 0 0
and the chiral fermion fields
‰R WD PR ‰ and ‰L WD PL ‰ : (10.29)
2| With ‰PR D ‰L and ‰PL D ‰ R show that
‰.i @= m/‰ $ ‰ R .i @=/‰R C ‰L .i @=/‰L m‰L ‰R m‰ R ‰L (10.30)
Use PR C PL D 14 , PR=L 2
D PR=L , PR=L D PR=L and f 5 ; g D 0 to show this.
Only the mass term mixes right- and left-handed fermions. We did not use this notation
so far, because there was no reason to (and the left-hand side is shorter).
3| The Dirac representation is reducible, a fact that is manifest in the Weyl basis, recall
Eq. (3.20). Alternatively, it is easy to check that
h i
PR=L ; ƒ 1 $ 0 (10.31)
2
so that the decomposition Eq. (10.30) is irreducible for Lorentz transformations.
! Terms like ‰ R .i @=/‰R and ‰L ‰R are Lorentz invariant on their own and do not mix
with their counterparts ‰L .i @=/‰L and ‰ R ‰L under continuous Lorentz transformations
ƒ 1 2 SOC .1; 3/.
2
! Under additional (gauge) symmetries, the left- and right-handed fields ‰L=R (then
multiplets) can transform under different representations of these new symmetry groups!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 186
LE C T U R E 26 → P S : 70 0 – 727
10.2.2 Overview
1| Field content:
• Fermions (= Spin- 12 ):
Generation n I II III
eL eR L R L R
Leptons
eL .eR / L .R / L . R / (10.32)
uL uR cL cR tL tR
Quarks
dL dR sL sR bL bR
– Here, each symbol xL=R denotes a four-component, chiral bispinor field which
describes both a fermion and its corresponding antifermion (recall the QED
Lagrangian). Note that the chirality is reversed for the antiparticles: eL
describes left-handed electrons and right-handed positrons.
– The right-handed neutrinos (in parantheses) have not been observed. In
the standard model, these fields are completely uncharged (mathematically
speaking, they transform under the trivial representation of all gauge groups);
thus these fields are typically omitted in the Lagrangian (however, they may be
required to explain the neutrino masses, → below).
– The three generations of fermions are not necessary for the symmetry
considerations that follow. We will simply sum over the generation index n. It
is unclear why there are three generations; however, so far there is no evidence
for a fourth generation. All stable baryonic matter in the observable universe is
made from first generation fermions as the other generations are much heavier
and decay quickly into first generation particles.
– While the total number of generations is not determined, the fact that each
generation contains three chiral leptons (e.g. eL ; eR ; eL ) and four chiral
quarks (e.g. uL ; dL ; uR ; dR ) is crucial to cancel the so called chiral anomaly
when quantizing the theory. The number of quarks and leptons is therefore
not independent, ↑ P&S pp. 705–707!
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 187
LE C T U R E 26 → P S : 70 0 – 727
• Vector bosons (= Spin-1):
Force Electroweak Strong
Gauge group SU.2/L U.1/Y SU.3/C
# Generators 3C1D4 8
(10.33)
Gauge fields Wi .i D 1; 2; 3/; B G
a .a D 1; : : : ; 8/
„ ƒ‚ …
Before Higgs SSB
Gauge bosons ; W C; W ; Z 8 Gluons
„ ƒ‚ …
After Higgs SSB
Warning: The gauge field B of the U.1/Y symmetry does not correspond to the
photon of QED (→ Higgs mechanism in the GWS theory below).
• Scalar bosons (= Spin-0):
3 SSB
2 Complex Higgs fields C 0 ! 1 Real Higgs field h (10.34)
„ ƒ‚ … „ ƒ‚ …
Before Higgs SSB After Higgs SSB
The three missing dof after SSB give the three vector bosons W ˙ and Z their mass
($ longitudinal component).
Why is nature like this? That doesn’t look very pretty!
Well, we don’t know! The most likely answer is that at very high energies (the
⁂ GUT scale 1025 eV) the picture becomes more symmetric with fewer distinct
fields. The mess we observe may be caused by spontaneous symmetry breaking at our
“low” energies. Finding a “prettier” construction is the quest for a GUT, a Grand
Unified Theory. However, one should always be careful when assessing the “beauty” (=
simplicity?) of theories. We are not entitled to live in a “simple and comprehensible”
universe!
2| Question: How to put this “chaos” into a consistent (= relativistic, renormalizable) QFT?
Answer:
LSM D LEWS C LQCD C LGF C LGhost (Standard model) (10.35)
The above Lagrangian contains all physical fields and interactions of the Standard
Model. However, for actual computations one has to add two “auxiliary” terms LGF
and LGhost to account for the complications that arise from the quantization of Yang-
Mills gauge theories [36]. LGF contains Gauge Fixing terms similar to .@ A /2 =2
(← Faddeev-Popov procedure, in particular Eq. (8.30)) to prevent the overcounting of
gauge-equivalent field configurations. LGhost contains unphysical ghost fields which are
needed to account for the determinant det .ıG.A˛ /=ı˛/ which, for non-abelian gauge
theories, is no longer independent of the gauge field (cf Eq. (8.27b) and the step before).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 188
LE C T U R E 26 → P S : 70 0 – 727
3| Two parts:
• Electroweak Standard Model LEWS
= Glashow-Weinberg-Salam (GWS) Theory
= Unification of weak & electromagnetic force
(+ mass generation through Higgs mechanism)
• Quantum Chromodynamics LQCD = Strong force
10.2.3 The Glashow-Weinberg-Salam Theory
GWS theory = Unification of the electromagnetic and weak interaction of the standard model;
explains the masses of W ˙ and Z bosons and all the fermions (including quarks) with the
Higgs mechanism.
Goal: Generalize the Higgs mechanism to the Standard model
1| Lagrangian:
LEWS D LFermion C LYang Mills C LHiggs C LYukawa (10.36)
We will discuss each term separately in the following.
2| Gauge symmetry (pre-Higgs, i.e., without SSB of the vacuum):
SU.2/L U.1/Y (10.37)
„ ƒ‚ … „ƒ‚…
Weak isospin Weak Hypercharge
• SU.2/L ! 3 generators T i , i D 1; 2; 3 with
h i
T i ; T j D i "ij k T k (10.38)
! Irreducible representations: (hats denote representation matrices)
– 1D: Trivial representation TO i D 0 (= Singlet representation)
i
– 2D: Pauli matrices TO i D 2 (= Doublet representation)
(In the following, TO i always denotes the doublet representation.)
! Eigenvalue of TO 3 = The weak isospin T 3
(T 3 D ˙ 12 for doublet and T 3 D 0 for singlet)
(For eigenvalues we do not write hats as these are not matrices but numbers.)
• U.1/Y ! 1 generator Y
h i
Y; T i D 0 (10.39)
(Since the gauge group is a direct product of SU.2/L and U.1/Y .)
Schur’s lemma ! YO D Number 1 D Hypercharge Y 1
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 189
LE C T U R E 26 → P S : 70 0 – 727
3| SU.2/L Representations:
We focus here on the first generation fermions. The values (= representations) of the
weak hypercharge cannot be infered at this point; we will discuss them after the Higgs
mechanism.
• Left-handed fields = Isospin doublets:
uL eL cL L tL L
‰L D ; ; ; ; ; (10.40)
dL eL sL L bL L
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
Gen. I Gen. II Gen. III
! Weak isospin: T 3 .eL / D C 12 and T 3 .eL / D 1
2 …
The notation used here is conventional but a bit confusing: With
eL
(10.41)
eL
we mean that the chiral bispinor field eL .x/ is of the form
1
eL .x/ D L .x/ ˝ 2 L2 .R1;3 / ˝ C 4 ˝ CL
2
(10.42)
0
with some left-chiral bispinor field L .x/. The last factor CL 2
is the spin- 12
representation space of SU.2/L and C is the representation space of the Dirac
4
bispinor. It is then
3 1 3 3 1 1
T .eL / D C W, TO eL .x/ D L .x/ ˝ TO D C eL .x/ : (10.43)
2 0 2
1 0
The basis vectors that span CL , say eL
2
and eL are refered to
0 1
as flavours (of first-generation leptons). Similarly, uL and dL are the flavours of
first-generation quarks.
• Right-handed fields = Isospin singlets:
R D uR ; dR ; eR ; cR ; sR ; R ; tR ; bR ; R (10.44)
„ ƒ‚ … „ ƒ‚ … „ ƒ‚ …
Gen. I Gen. II Gen. III
In the following, ‰ denotes a doublet and a singlet. If we write ‰L , we refer to a
doublet of left-handed components (as above).
! Weak isospin: T 3 .eR / D 0 …
– Note that we omit right-handed neutrinos eR ; : : : because such particles have
never been observed [37]. If they exist, one could add them to Eq. (10.44) just
as any other right-handed fermion (→ below).
– One may wonder why left- and right-handed fields are treated so differently.
The answer is observations: Experiments show that the weak interaction only
couples to left-handed fermions (and right-handed antifermions). To represent
this fact about nature mathematically, right-handed fermions must be isospin
singlets.
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 190
LE C T U R E 26 → P S : 70 0 – 727
• Higgs fields = Isospin doublet:
C
ˆD (10.45)
0
1 1
! Weak isospin: T 3 . C / D 2 and T 3 . 0 / D 2
– Both ˆ and its scalar components C and are often refered to as “Higgs
0
field”. Note that despite the vectorial notation, the fields C and 0 are com-
plex (Lorentz) scalars. That is, “scalar” refers to their trivial transformation
under Lorentz transformations. The Higgs field ˆ does not transform trivially
under SU.2/L gauge transformations, as its doublet structure reveals. In a
nutshell: The Higgs field is a Lorentz scalar (= Spin-0 irrep of SOC .1; 3/) but
a SU.2/L gauge doublet (= Spin- 12 irrep of SU.2/).
– The fact that the Higgs field is a isospin doublet is essential for the Higgs
mechanism (→ below); it is this property that allows for the generation of Dirac
mass terms for the fermions.
! Gauge transformations on fields:
Left-handed doublet: Q L D e i YOL ˛.x/ e i TO i ˇ i .x/ ‰L
‰ (10.46)
„ ƒ‚ …
VL .x/
Right-handed singlet: Q R D e i YOR ˛.x/ R (10.47)
Higgs doublet: Q̂ D e i YOH ˛.x/ e i TO i ˇ i .x/ ˆ (10.48)
where YOL D Y 12 , YOR D Y 1 and YOH D Y 12
Note that here also the hypercharge is an operator. As we consider a direct sum of
possibly unitary equivalent but different copies of irreps, Y can take different values on
these irreps.
Note: The weak hypercharge Y is a fixed number for each irrep, e.g., Y .uL / D Y .dL /,
but can differ for different irreps: Y .uL / ¤ Y .eL / (→ Higgs mechanism below)
4| Kinetic energy for fermions & Minimal coupling:
X X
LFermion D = L /‰L C
‰L .i D = R/ R
R .i D (10.49)
‰L R
The sums go here over the doublets in Eq. (10.40) and the singlets in Eq. (10.44).
with covariant derivatives
DL D @ igWi TO i ig 0 B YOL (10.50)
DR D @ ig 0 B YOR (10.51)
g = Wi : coupling constant / gauge field for weak isospin
g 0 = B : coupling constant / gauge field for weak hypercharge
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 191
LE C T U R E 26 → P S : 70 0 – 727
Note that ŒTO i ; YOL D 0 for all i so that the fields Wi and B do not mix under gauge
transformations and thus can have different coupling constants g and g 0 , respectively.
The Lagrangian Eq. (10.49) violates the symmetries C (swaps left-handed fermions with
left-handed antifermions) and P (swaps left-handed and right-handed fermions) as much
as possible since left(right)-handed (anti)fermions couple weakly but right(left)-handed
(anti)fermions do not. Note that CP swaps a left-handed fermion with a right-handed
antifermion so that Eq. (10.49) is CP -symmetric.
! Transformation of the gauge fields:
1 i
BQ D B C 0 @ ˛ and Q
W D VL W C @ VL (10.52)
g g
Recall Eq. (9.28) and Eq. (9.10). Here we use the shorthand notation W Wi TO i .
→ Example 10.1: Beta decay
To draw a connection to previous knowledge, focus on the two left-handed
first-generation terms in Eq. (10.49)
uL eL
LFermion D uL d L .i D = L/ = L/
C eL eL .i D C : : : (10.53)
dL eL
and
WC
g 0
DL D ig.W1 TO 1 C W2 TO 2 / C ::: D ip C ::: (10.54)
2 W 0
p
with W˙ WD 1= 2 W1 iW2 . Thus we find terms of the form
LFermion WC uL
dL C W eL
eL C h.c. C : : : (10.55)
Now think of a second-order pro-
cess (← Eq. (4.132)) that includes
both vertices that derive from
these terms (there would be a
contraction of two W˙ fields in-
volved that produces a gauge field
propagator) and add an up and
a down quark that do not partic-
ipate in the interaction for good
measure:
This is nothing but the ˇ -decay you already know from your physics course in
kindergarten. Note that all fermions/anti-fermions that connect to the vertices are
left-handed/right-handed (eL describes a left-handed electron and a right-handed
positron).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 1 92
LE C T U R E 26 → P S : 70 0 – 727
5| Dirac mass terms? Should be of the form
m.‰L R C R ‰L / ! Undefined! (10.56)
^ Elementary terms of the form xL yR with x; y Dirac spinors
! Not SU.2/L gauge invariant since
• xL is component of a SU.2/ doublet
• but yR transforms as a SU.2/ singlet
The argument here is the same as, e.g., for an expression like Ep 2 that is not a Lorentz
scalar since E D p 0 is component of a four vector.
! xL yR is not a SU.2/ singlet (i.e., not gauge invariant)
! We cannot add Dirac mass terms to the Lagrangian!
Solution: Yukawa coupling and Higgs mechanism (→ below)
6| Kinetic energy for gauge bosons ! Yang-Mills Lagrangian:
1 2 1
LYang Mills D .B / .W i /2 (10.57)
4 4
with B D @ B @ B (10.58)
i
W D @ Wi @ Wi Cg" ij k
Wj Wk (10.59)
„ ƒ‚ …
Interactions between
gauge bosons
Here, "ij k D f ij k are the structure constants of SU.2/, → Problemset 13.
7| Higgs field:
LHiggs D .DH ˆ/ .DH ˆ/ 2 ˆ ˆ .ˆ ˆ/2 (10.60)
with covariant derivative
DH D @ igWi TO i ig 0 B YOH (10.61)
Note that .ˆ ˆ/2 ¤ j C j4 C j 0 j4 ; the latter term is Lorentz- but not gauge invariant
so that only the former is an allowed interaction. The form of the Higgs potential is
then given by the condition of renormalizability. To make the vacuum stable, > 0 is
required.
8| Higgs mechanism Part I: Masses for the gauge bosons
i| Let 2 < 0 ! Non-zero VEV of Higgs field:
s
2
1 0
Wlog hˆi D ˆ0 D p with vD (10.62)
2 v
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 193
LE C T U R E 26 → P S : 70 0 – 727
ii | Define the electric charge (operator)
Q D T 3 C Y 2 su.2/L ˚ u.1/Y (10.63)
! Choose Y .ˆ/ D C 12 so that
1 1 O
O
Qˆ0 D C ˆ0 D 0 ) e i Q˛.x/ ˆ0 D ˆ0 (10.64)
2 2
This is why the lower Higgs field is called 0 : it is uncharged, Q. 0 / D 0. By
contrast, the upper field C has charge Q. C / D C1. Choosing the hypercharge
fixes the representation of the Higgs field under U.1/Y , just as demanding the
doublet structure fixed the representation under SU.2/L .
! Gauge symmetry U.1/Q generated by Q is unbroken:
3 SSB
SU.2/L U.1/Y ! U.1/Q (10.65)
„ ƒ‚ …
Unbroken gauge
group of QED
Three generators of the global symmetry group are spontaneously broken while
one generator (Q) remains unbroken. This is what we want, as we know that there
should be one massless gauge boson: the photon.
Conclusion: The generator of U.1/Y (the weak hypercharge Y ) and the generator
of U.1/Q (the electric charge Q) are not the same!
iii | ^ Fluctuations of ˆ around ˆ0 in the unitary gauge:
1 0
ˆ.x/ D p (10.66)
2 v C h.x/
h.x/: real scalar Higgs field
The excitations of this field are the famous Higgs bosons.
iv | ˆ.x/ in Eq. (10.60):
v2 ˚ 2 1 2
.DH ˆ/ .DH ˆ/ $ g .W / C .W2 /2 C . gW3 C g 0 B /2 C : : :
8
(10.67)
(We focus here on the terms that generate the gauge boson masses.)
v| Define the new fields
1
W˙ WD p W1 iW2
(10.68)
2
1
gW3 g 0 B
Z WD p (10.69)
2
g Cg 02
1
g 0 W3 C gB
A WD p (10.70)
2
g Cg 02
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 194
LE C T U R E 26 → P S : 70 0 – 727
Thep ratio of g and g defines the so called ⁂ Weinberg angle W : cos W D
0
g= g C g . This parameter is not predicted by the SM but one of the many
2 02
input parameters that have to be determined experimentally. It is also called weak
mixing angle as it describes the mixing of W3 and B that yields A .
!
gv 2 1 v 2 2
(10.67) $ WC W
C .g C g 02 / .Z /2 C : : : (10.71)
2 …
„ ƒ‚ 2 „2 ƒ‚ …
m2W m2Z
and (express the covariant derivative in the new fields)
gg 0
DH D @ .: : : / i p A QO (10.72)
g 2 C g 02
„ ƒ‚ …
Electron charge e
We conclude:
• A : massless, neutral (Q D 0) gauge field of QED
• W˙ : massive, charged (Q D ˙1) gauge bosons of weak interaction
• Z : massive, neutral (Q D 0) gauge boson of weak interaction
9| Interlude:
With Eq. (10.63) we can fix the hypercharge Y by the (observed) electric charge Q.
Examples:
3 1 1
Y .eL / D Q.eL / T .eL / D 1 D (10.73)
2 2
Y .eR / D Q.eR / T 3 .eR / D 1 0D 1 (10.74)
We will need these two hypercharges to understand the Yukawa coupling below.
10 | Higgs mechanism Part II: Masses for the fermions
i| How to form a gauge invariant term including left- and right-handed fermions?
Must be a SU.2/L singlet and hypercharge-neutral (Y D 0)!
! Couple left-handed fermion doublet, Higgs doublet, and right-handed fermion
singlet via a Yukawa term:
[Compare: ‰‰ (Yukawa) vs. A ‰ ‰ (Maxwell)]
eL
e .‰L ˆ/ eR C h.c. with ‰L D (10.75)
eL
e: coupling constant
Note that
1 1
Y .ˆ/ C Y .eR / Y .‰L / D 1 D0 (10.76)
2 2
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 195
LE C T U R E 26 → P S : 70 0 – 727
and
C
eR D C eL eR C 0 eL eR
.‰L ˆ/ eR D eL eL (10.77)
0 „ ƒ‚ …
„ ƒ‚ … Scalars Dirac inner products
SU.2/L singlet
so that Eq. (10.75) is both SU.2/L and U.1/Y invariant. The last expression reveals
the Yukawa-form of the interaction clearly.
p
Higgs mechanism: C 7! 0 and 0 7! v= 2 !
ev
(10.75) D p .eL eR C e R eL / C : : : (10.78)
2
p
with fermion mass me D e v= 2
The same works for the other charged leptons and the quarks but not the neutrinos
if their right-handed counterparts are excluded. Thus, in the Standard Model,
neutrinos are massless because of their missing right-handed partners!
ii | In general, we can couple different fermion generations:
This is possible since fermions of the same type (charged lepton l, neutrino
, up-type u and down-type d quark) but different generations have the same
hypercharge and isospin.
u m m
LYukawa D mn QL Ô unR d
mn QL ˆ dRn
l m n m n
mn LL ˆ lR mn LL Ô R C h.c. (10.79)
There are implicit sums over the fermion generations m and n. All other symbols
are fixed labels.
• m; n 2 fI,II,IIIg: fermion generations
• x 2 fu; d; l; g: fermion types
I II
Examples: lR D eR , lR D R , uIR D uR , uIIR D cR , …
• mn
x : coupling constants
Example: I,I
l
D e from above
• QLm
, LL
m
: left-handed quark- resp. lepton doublets of generation m
I II
Examples: QL D uL d L and LL D L L , …
• Ô i "ij ˆj : Higgs doublet with opposite hypercharge: Y . Ô / D 12
This representation is required to make the terms hypercharge-neutral [to
m
see this, use Y .unR / D 23 , Y .dRn / D 13 and Y .QL / D 16 ]. Note that Ô
transforms in the same isospin irrep as ˆ.
iii | The Yukawa couplings Eq. (10.79) …
• … generate mass terms for quarks and charged fermions
• … cannot generate mass terms for neutrinos if there are no right-handed
neutrinos (! Neutrinos are massless in the standard model.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 196
LE C T U R E 26 → P S : 70 0 – 727
→ Note 10.1
Adding right-handed neutrinos R n
to produce the Dirac mass term in
Eq. (10.79) can be used to explain the experimentally observed masses of
neutrinos [38, 39] and might even contribute to dark matter. Alternatively
(or in addition), right-handed neutrinos can obtain mass through a
↑ Majorana mass term (which comes with an additional free parameter mR
unrelated to the Higgs VEV v and the Yukawa coupling strength mn )
1
LMajorana D mR R .R /c C h.c. (10.80)
2
with the charge conjugation of a field (← Eq. (3.104))
c WD C 1 D i 2
: (10.81)
2
(Note that without charge conjugation it is trivially R R 0. For the
Majorana mass term, it does not matter whether R is part of a Dirac or
Majorana spinor as long as it is not charged.)
Right-handed neutrinos R are called ⁂ sterile neutrinos as they have
vanishing weak isospin T 3 D 0 and electric charge Q D 0 (and therefore
hypercharge Y D 0). Hence they do not interact through the three
forces described by the standard model; in particular, the Lagrangian
Eq. (10.80) is gauge invariant (this would not be true for any other fermion
with non-vanishing U.1/ charge!). Note that they do interact via the
Yukawa coupling in Eq. (10.79) with the Higgs field and the left-handed
lepton fields. So, if they exist, right-handed neutrinos could be detected
indirectly by studying their decay channels.
Both of these mass terms (Dirac from LYukawa and Majorana from
LMajorana ) are combined in the ↑ (Type I) Seesaw mechanism [40, 41] to
explain the extremely small neutrino masses (compared to all other
fermions) [42]. The masses of the mass eigenstates (which are then
Majorana fermions) are given by [41]
m2
m1 and m2 mR so that m1 m2 m2 (10.82)
mR
p
for mR m D v= 2. (Remember that mR is unconstrained,
↑ Ref. [41] for arguments why it could be very large. Note that
Eq. (10.80) breaks lepton number conservation and therefore should
happen at mass scales much larger than the electroweak scale.)
The very small m1 would correspond to the mass of our left-handed
neutrinos while the very large m2 would be the mass of the hypothetical
right-handed neutrino. The relation m1 m2 m2 with fixed Dirac mass
m gives rise to the name “Seesaw mechanism”: Large m2 corresponds
to small m1 and vice versa. Note that the Seesaw mechanism naturally
leads to m1 m where m is expected to be of the scale of the other
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 197
LE C T U R E 26 → P S : 70 0 – 727
leptons/quarks (which is way to heavy to be consistent with the extremely
light neutrinos).
Note: Massive neutrinos are “beyond Standard Model physics”. However,
since they can be described by straightforward extensions of the SM
(e.g. by the Seesaw mechanism) without modifying its gauge group, the
neutrino masses are sometimes treated as part of SM physics.
• … lead to generation-changing transitions of quarks.
(↑ CKM matrix and P&S pp. 721–724)
The generation mixing in Eq. (10.79) implies that the quark states that take part
in weak interactions (= interaction eigenstates) are not the eigenstates of the
mass operator (= mass eigenstates) that describe freely propagating particles.
Then one can show that a (mass eigenstate) s-quark that propagates freely can
decay into a (mass eigenstate) u-quark by coupling to a (virtual) W -boson. If
there are no right-handed neutrinos, such transitions are forbidden for leptons
(which so far matches experimental observations).
• … lead to generation-changing transitions of neutrinos if right-handed neutrinos
are added. (↑ PMNS matrix and ↑ Neutrino oscillations)
10.2.4 Quantum Chromodynamics
We discuss QCD here only superficially to connect with concepts that we learned previously.
1| Gauge symmetry:
SU.3/C (10.83)
„ ƒ‚ …
Color charge
! 8 generators K a , a D 1; : : : ; 8 with (in general, SU.N / has N 2 1 generators)
h i
K a ; K b D if abc K c (10.84)
Here we use the unconventional label K a to distinguish the generators from the SU.2/L
generators T i of the weak force.
! Irreducible representations:
• 1D: Trivial representation KO a D 0 (= Singlet representation)
• 3D: Defining representation (physicist parlance: fundamental representation):
KO a D a with 3 3 Hermitian Gell-Mann matrices a (= Triplet representation)
2
Gell-Mann matrices are the analog of Pauli matrices for SU.3/.
2| Field representations:
• Quarks = SU.3/C triplets
0 1
qr
q D @qg A for q 2 fu; d; c; s; t; bg (10.85)
qb
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 198
LE C T U R E 26 → P S : 70 0 – 727
with colors r (red), g (green), b (blue)
Note that each color field qc is a Dirac bispinor, i.e., we extended the number of
quark fields threefold!
The notion of “colors” is not gauge invariant: For instance, a “red” quark qr can be
transformed into a mixture of red, green, and blue quarks by a gauge tranformation
UC .x/:
0 1 0 1
qQ r qr
@qQ g A D UC @ 0 A (10.86)
qQ b 0
• Leptons & Higgs fields = SU.3/C singlets ! Ignore them in QCD
Since the leptons (e, e , …) do not interact via the strong force, they carry the
trivial (singlet-) representations of SU.3/C , i.e., their fields are not extended into
triplets and it is KO a D 0 for actions of SU.3/C transformations on their fields.
! Gauge transformation of fields:
Oa a
Quark triplet: qQ D e„i Kƒ‚
ˇ .x/
…q (10.87)
UC .x/
3| Lagrangian:
X 1 a 2
LQCD D = C /q
q.i D .G / (10.88)
q
4
Note the missing mass terms! As explained above, the masses of quarks are generated by
the Higgs mechanism and electroweak SSB.
with covariant derivative
a Oa
DC D @ igs G K (10.89)
gs : coupling constant of the strong force
Ga : 8 gauge fields ! 8 gauge bosons = 8 Gluons
← Eq. (9.28) for the transformation of G G aKO a under UC .x/.
The Gauge field strength is defined as usual:
a
G D @ Ga a
@ G C gs f abc G
b c
G (10.90)
→ Note 10.3
• No additional Higgs mechanism:
– Quarks masses are generated by electroweak SSB
– Gluons are massless
• Gluons carry color charges and can therefore interact with each other (← Note 9.1)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 199
LE C T U R E 26 → P S : 70 0 – 727
Mathematically, this means that gluons transform in a non-trivial representation
of SU.3/C (not the three-dimensional defining irrep of quarks but the so called
adjoint representation which is 8-dimensional for SU.3/). Gluons act then on colored
quarks and change their color. That is, if we write jci for the three color states of a
quark (c D r; g; b), gluon states can be generated from matrices of the form jcihc 0 j
(one says that gluons carry a color c and an anticolor c 0 ). However, this suggests
3 3 D 9 gluon states, but there are only 8! The hitch is that the linear combination
KO 0 jrihrj C jgihgj C jbihbj (10.91)
is forbidden (physically, this means that a gluon can never transform in the singlet
representation, i.e., a gluon cannot be coloreless). That KO 0 is not part of the
generating set of su.3/ can be seen easily since
O 0
ei K D 13 (10.92)
has determinant 1! That is, a coloreless gluon would imply a gauge group U.3/
rather than SU.3/. However, such a gluon would not be constrained by confinement,
and therefore contradicts current experimental evidence. Thus the gauge group of
QCD is SU.3/ with 8 gluons and not U.3/ with 9 (↑ p. 279ff. of Ref. [43]).
The Gell-Mann matrices are then 8 particular linear combinations of the 9 matrices
jcihc 0 j that are linearly independent of KO 0 , e.g., 1 D jrihgj C jgihrj.
gs2
• Renormalization: Let ˛s 4 , then
q 2 !1
˛seff .q 2 / !0 ! Asymptotic freedom (10.93)
q 2 !0
˛seff .q 2 / ! 1 ! Confinement (10.94)
Compare this with the running of ˛eff .q 2 / in QED, ← Eq. (6.246).
For experimental results ↑ P&S Fig. 17.23 on p. 595.
That is, quarks at very high energies (e.g. in hadrons) behave almost like free
particles. By contrast, at low energies, their interaction becomes so strong that free
particles that carry a color charge (i.e. are not a color singlet) do not exist (thus we
observe only mesons and baryons that are “white”).
(*) Note that a diverging coupling constant does not prove confinement (QED
also has a divergence, known as Landau pole, which is not related to confinement,
← Note 6.4). Up to know, the existence of confinement in QCD and the true
IR behavior of ˛seff remains unproven resp. unknown (confinement is supported
by numerical lattice QCD calculations though) since this regime is not accessible
by perturbation theory. The reason for the opposite running of the coupling
constant (compared to QED) is the existence of virtual gluon bubbles in the vacuum
polarization diagrams that lead to “antiscreening” (↑ p. 293ff of Ref. [43]).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 200
LE C T U R E 26 → P S : 70 0 – 727
10.2.5 Summary
• Gauge symmetry group of the standard model:
SU.2/L U.1/Y SU.3/C (10.95)
„ ƒ‚ … „ƒ‚… „ ƒ‚ …
Weak isospin Weak Hypercharge Color charge
„ ƒ‚ …
Electroweak SSB
! U.1/Q
„ ƒ‚ …
Electric charge
Our vacuum has lost the global SU.2/L U.1/Y symmetry since the Higgs field
developed a VEV (by the way, it is unclear why this happened). The “true Lagrangian
of the universe” still has this symmetry, only our low-energy vacuum “hides” this
symmetry from us. Thus spontaneous symmetry breaking is sometimes refered to as
spontaneous symmetry hiding.
• Fermions and their interactions:
In total there are
Œ2 Leptons C 2 Quarks 3 Colors 3 Generations D 24 Dirac bispinors (10.96)
each consisting of 4 complex fields ! 96 complex fields for fermions. (This count
includes right-handed neutrinos for good measure.)
• The standard model Lagrangian LSM contains 18 parameters (can be more if addi-
tional extensions to the SM are considered, e.g., neutrino masses ← Note 10.1) that
cannot be derived but must be measured by experiments:
– 9 Fermion masses: me , mu ,… (recall that neutrinos are massless in the SM)
– 1 Higgs mass mh 125 GeV
(This is the famous result from the observation at LHC in 2012 [44])
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 201
LE C T U R E 26 → P S : 70 0 – 727
– 1 Higgs field VEV v
– 3 Gauge field couplings: g, g 0 , gs
– 4 CKM matrix parameters: 12 ,… (describe the mixing of quark generations and
possible CP -violating terms, hidden in the Yukawa coupling matrices of Eq. (10.79))
(↑ P&S p. 721ff.)
In conclusion, the SM does not seem to be good candidate for a truly fundamental theory
(which should be a more efficient “compression” of the laws of nature). This is one of
the reasons to look for a GUT (↑ Grand Unified Theory = Unification of all three forces of
the SM) or even a TOE (↑ Theory Of Everything = Unification of all three forces of the
SM and gravity) that allows for the computation of some (or all) of these parameters ab
initio.
• Have a look at Ref. [45] if you want to know more about the representation theory of the
standard model (and its possible extensions to GUTs).
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 2 02
BIBLIOGRAPHY
Bibliography
[1] S. Weinberg, The Quantum Theory of Fields, vol. 1, Cambridge University Press, ISBN 9780521585552 (1996).
(Cited on pages 4, 51, 69, and 170.)
[2] C. Itzykson and J.-B. Zuber, Quantum field theory, Dover Publications, ISBN 0486445682 (2005). (Cited on
page 4.)
[3] M. E. Peskin and D. V. Schroeder, An introduction to quantum field theory, Addison-Wesley Pub. Co., ISBN
0201503972 (1995). (Cited on page 4.)
[4] A. Zee, Quantum Field Theory in a Nutshell, Princeton University Press, ISBN 9780691140346 (2010). (Cited
on page 4.)
[5] P. Francesco, P. Mathieu and D. Senechal, Conformal Field Theory, Springer, ISBN 9781461222569 (2012).
(Cited on page 11.)
[6] R. Turcati, C. A. B. Quintero, J. A. Helayël-Neto and E. Arias, Spin-polarized current, spin-transfer torque, and
spin Hall effect in presence of an electromagnetic non-minimal coupling, Annalen der Physik 532(3), 1900357 (2020),
doi:10.1002/andp.201900357. (Cited on page 50.)
[7] W. Bacino, T. Ferguson, L. Nodulman, W. Slater, H. K. Ticho, A. Diamant-Berger, M. Faessler, A. Hall,
G. Irwin, J. Kirkby, F. Merritt, S. Wojcicki et al., Measurement of the Threshold Behavior of C Production in
e C e Annihilation, Physical Review Letters 41(1), 13 (1978), doi:10.1103/physrevlett.41.13. (Cited on
page 92.)
[8] M. Arminjon, Lorentz-invariant second-order tensors and an irreducible set of matrices, Journal of Geometry and
Symmetry in Physics 50, 1 (2018), doi:10.7546/jgsp-50-2018-1-10. (Cited on page 105.)
[9] J. Schwinger, On Quantum-Electrodynamics and the magnetic moment of the electron, Physical Review 73(4), 416
(1948), doi:10.1103/physrev.73.416. (Cited on page 109.)
[10] D. Hanneke, S. F. Hoogerheide and G. Gabrielse, Cavity control of a single-electron quantum cyclotron: Measuring
the electron magnetic moment, Physical Review A 83(5), 052122 (2011), doi:10.1103/physreva.83.052122.
(Cited on page 110.)
[11] T. Aoyama, M. Hayakawa, T. Kinoshita and M. Nio, Tenth-order electron anomalous magnetic mo-
ment: Contribution of diagrams without closed lepton loops, Physical Review D 91(3), 033006 (2015),
doi:10.1103/physrevd.91.033006. (Cited on page 110.)
[12] S. Laporta and E. Remiddi, The analytical value of the electron (g - 2) at order ’3 in QED, Physics Letters B
379(1-4), 283 (1996), doi:10.1016/0370-2693(96)00439-x. (Cited on page 110.)
[13] K. Gottfried and V. F. Weisskopf, Concepts of Particle Physics, vol. 2, Oxford University Press, USA, ISBN
9780195033939 (1986). (Cited on page 137.)
[14] L. Landau, An asymptotic expression for the photon Green function in quantum electrodynamics, In Collected Papers
of L.D. Landau, pp. 616–620. Elsevier, doi:10.1016/b978-0-08-010586-4.50085-7 (1965). (Cited on
page 139.)
[15] M. Göckeler, R. Horsley, V. Linke, P. Rakow, G. Schierholz and H. Stüben, Is there a landau pole problem in QED?,
Physical Review Letters 80(19), 4119 (1998), doi:10.1103/physrevlett.80.4119. (Cited on page 139.)
[16] F. J. Dyson, Divergence of perturbation theory in quantum electrodynamics, Physical Review 85(4), 631 (1952),
doi:10.1103/physrev.85.631. (Cited on page 139.)
[17] M. Flory, R. C. Helling and C. Sluka, How I learned to stop worrying and love QFT,
doi:10.48550/arxiv.1201.2714 (2012). (Cited on page 139.)
[18] G. Aad, B. Abbott, D. Abbott, O. Abdinov, A. A. Abud, K. Abeling, D. Abhayasinghe, S. Abidi, O. AbouZeid,
N. Abraham, H. Abramowicz, H. Abreu et al., Observation of Light-by-Light Scattering in Ultrape-
ripheral Pb C Pb Collisions with the ATLAS Detector, Physical Review Letters 123(5), 052001 (2019),
doi:10.1103/physrevlett.123.052001. (Cited on page 145.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 203
BIBLIOGRAPHY
[19] A. Shomer, A pedagogical explanation for the non-renormalizability of gravity, doi:10.48550/ARXIV.0709.3555
(2007). (Cited on page 149.)
[20] G. ’t Hooft and M. Veltman, One-loop divergencies in the theory of gravitation, In Euclidean Quantum Gravity, pp.
3–28. World Scientific, doi:10.1142/9789814539395_0001 (1993). (Cited on page 149.)
[21] M. H. Goroff and A. Sagnotti, The ultraviolet behavior of Einstein gravity, Nuclear Physics B 266(3-4), 709
(1986), doi:10.1016/0550-3213(86)90193-8. (Cited on page 149.)
[22] R. P. Feynman, Space-time approach to non-relativistic quantum mechanics, In Feynman's Thesis — A New Approach
to Quantum Theory, pp. 71–109. World Scientific, doi:10.1142/9789812567635_0002 (2005). (Cited on
page 158.)
[23] B. C. Hall, Quantum Theory for Mathematicians, Springer, ISBN 9781461471158 (2013). (Cited on page 162.)
[24] R. P. Feynman, An operator calculus having applications in quantum electrodynamics, Physical Review 84(1), 108
(1951), doi:10.1103/physrev.84.108. (Cited on page 163.)
[25] L. Faddeev and V. Popov, Feynman diagrams for the Yang-Mills field, Physics Letters B 25(1), 29 (1967),
doi:10.1016/0370-2693(67)90067-6. (Cited on page 166.)
[26] G. ’t Hooft, Renormalization of massless Yang-Mills fields, Nuclear Physics B 33(1), 173 (1971),
doi:10.1016/0550-3213(71)90395-6. (Cited on page 170.)
[27] G. ’t Hooft, Renormalizable Lagrangians for massive Yang-Mills fields, Nuclear Physics B 35(1), 167 (1971),
doi:10.1016/0550-3213(71)90139-8. (Cited on page 170.)
[28] V. M. Abazov, B. Abbott, B. S. Acharya, M. Adams, T. Adams, J. P. Agnew, G. D. Alexeev, G. Alkhazov,
A. Alton, G. A. Alves, G. Antchev, A. Askew et al., Odderon exchange from elastic scattering differences between
pp and pp data at 1.96 TeV and from pp forward scattering measurements, Phys. Rev. Lett. 127, 062003 (2021),
doi:10.1103/PhysRevLett.127.062003. (Cited on page 177.)
[29] H. Watanabe and H. Murayama, Unified description of Nambu-Goldstone bosons without Lorentz invariance, Physical
Review Letters 108(25), 251602 (2012), doi:10.1103/physrevlett.108.251602. (Cited on page 181.)
[30] S. Elitzur, Impossibility of spontaneously breaking local symmetries, Physical Review D 12(12), 3978 (1975),
doi:10.1103/physrevd.12.3978. (Cited on pages 183 and 185.)
[31] W. Buchmüller, Z. Fodor and A. Hebecker, Gauge invariant treatment of the electroweak phase transition, Physics
Letters B 331(1-2), 131 (1994), doi:10.1016/0370-2693(94)90953-9. (Cited on page 183.)
[32] M. Greiter, Is electromagnetic gauge invariance spontaneously violated in superconductors?, Annals of Physics 319(1),
217 (2005), doi:10.1016/j.aop.2005.03.008. (Cited on page 183.)
[33] V. Rubakov, Classical theory of gauge fields, Princeton University Press, ISBN 0691059276 (2002). (Cited on
page 184.)
[34] E. Noether, Invariante Variationsprobleme, Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen,
Mathematisch-Physikalische Klasse 1918, 235 (1918). (Cited on page 185.)
[35] S. G. Avery and B. U. W. Schwab, Noether’s second theorem and Ward identities for gauge symmetries, Journal of
High Energy Physics 2016(2) (2016), doi:10.1007/jhep02(2016)031. (Cited on page 185.)
[36] J. C. Romao and J. P. Silva, A resource for signs and Feynman diagrams of the Standard Model (2012),
doi:10.48550/ARXIV.1209.6213. (Cited on page 188.)
[37] M. Goldhaber, L. Grodzins and A. W. Sunyar, Helicity of neutrinos, Physical Review 109(3), 1015 (1958),
doi:10.1103/physrev.109.1015. (Cited on page 190.)
[38] Y. Fukuda, T. Hayakawa, E. Ichihara, K. Inoue, K. Ishihara, H. Ishino, Y. Itow, T. Kajita, J. Kameda, S. Kasuga,
K. Kobayashi, Y. Kobayashi et al., Evidence for oscillation of atmospheric neutrinos, Physical Review Letters 81(8),
1562 (1998), doi:10.1103/physrevlett.81.1562. (Cited on page 197.)
[39] Q. R. Ahmad, R. C. Allen, T. C. Andersen, J. D.Anglin, J. C. Barton, E. W. Beier, M. Bercovitch, J. Bigu,
S. D. Biller, R. A. Black, I. Blevis, R. J. Boardman et al., Direct evidence for neutrino flavor transformation from
neutral-current interactions in the sudbury neutrino observatory, Physical Review Letters 89(1), 011301 (2002),
doi:10.1103/physrevlett.89.011301. (Cited on page 197.)
[40] P. Minkowski, !e” at a rate of one out of 109 muon decays?, Physics Letters B 67(4), 421 (1977),
doi:10.1016/0370-2693(77)90435-x. (Cited on page 197.)
[41] T. Yanagida, Horizontal symmetry and masses of neutrinos, Progress of Theoretical Physics 64(3), 1103 (1980),
doi:10.1143/ptp.64.1103. (Cited on page 197.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 2 04
BIBLIOGRAPHY
[42] M. Aker, A. Beglarian, J. Behrens, A. Berlev, U. Besserer, B. Bieringer, F. Block, S. Bobien, M. Böttcher,
B. Bornschein, L. Bornschein, T. Brunst et al., Direct neutrino-mass measurement with sub-electronvolt sensitivity,
Nature Physics 18(2), 160 (2022), doi:10.1038/s41567-021-01463-1. (Cited on page 197.)
[43] D. J. Griffiths, Introduction to elementary particles, Wiley, ISBN 0471603864 (1987). (Cited on page 200.)
[44] S. Chatrchyan, V. Khachatryan, A. Sirunyan, A. Tumasyan, W. Adam, E. Aguilo, T. Bergauer, M. Dragicevic,
J. Erö, C. Fabjan, M. Friedl, R. Frühwirth et al., Observation of a new boson at a mass of 125 GeV with the CMS
experiment at the LHC, Physics Letters B 716(1), 30 (2012), doi:10.1016/j.physletb.2012.08.021. (Cited
on page 201.)
[45] J. C. Baez and J. Huerta, The algebra of Grand Unified Theories (2009), doi:10.48550/arxiv.0904.1556.
(Cited on page 202.)
↑ Notes
N I CO L A I L A N G • I T P I I I • U N I V E R S I T Y O F ST U T TG A RT PAG E 2 05
You might also like
- An Introduction To Relativistic Processes and The Standard Model of Electroweak Interactions - Carlo M. Becchi, Giovanni Ridolfi (2006 Springer, 140p)No ratings yetAn Introduction To Relativistic Processes and The Standard Model of Electroweak Interactions - Carlo M. Becchi, Giovanni Ridolfi (2006 Springer, 140p)140 pages
- Einstein's Vierbein Field Theory of Curved Space: Jeffrey YepezNo ratings yetEinstein's Vierbein Field Theory of Curved Space: Jeffrey Yepez27 pages
- Rose RelativisticElectronTheory OuttakesNo ratings yetRose RelativisticElectronTheory Outtakes43 pages
- Dreiner H.K., Haber H.E., Martin S.P. Supersymmetry (CUP DrafNo ratings yetDreiner H.K., Haber H.E., Martin S.P. Supersymmetry (CUP Draf272 pages
- Frontiers in Modern Condensed Matter PhysicsNo ratings yetFrontiers in Modern Condensed Matter Physics19 pages
- Matrix Product States in Gauge TheoriesNo ratings yetMatrix Product States in Gauge Theories10 pages
- Lorentz and Poincaré Group RepresentationsNo ratings yetLorentz and Poincaré Group Representations11 pages
- QFT Solutions (Including Peskin & Schroeder 3.1 and 3.4)100% (1)QFT Solutions (Including Peskin & Schroeder 3.1 and 3.4)19 pages
- Julian Schwinger: Source Theory and The UCLA Years - From Magnetic Charge To The Casimir EffectNo ratings yetJulian Schwinger: Source Theory and The UCLA Years - From Magnetic Charge To The Casimir Effect15 pages
- Magnetism Radiation Relativity SchroederNo ratings yetMagnetism Radiation Relativity Schroeder39 pages
- Cosmology Lectures Part III Mathematical Trips - Daniel Baumann, CambridgeNo ratings yetCosmology Lectures Part III Mathematical Trips - Daniel Baumann, Cambridge130 pages
- Mathematical Methods of Physics Problems With Solutions Igor V KolokolovNo ratings yetMathematical Methods of Physics Problems With Solutions Igor V Kolokolov359 pages
- ElectromagneticFieldTheory ZahnSolutionsManual PDFNo ratings yetElectromagneticFieldTheory ZahnSolutionsManual PDF302 pages
- Indirect Exchange Magnetic Electrons : of Nuclear Moments by ConductionNo ratings yetIndirect Exchange Magnetic Electrons : of Nuclear Moments by Conduction4 pages
- Fundamentals of Statistical and Thermal Physics - Reif - Solutions Manual PDFNo ratings yetFundamentals of Statistical and Thermal Physics - Reif - Solutions Manual PDF126 pages
- Spontaneous Breakdown of Parity in A Class of Gauge TheoriesNo ratings yetSpontaneous Breakdown of Parity in A Class of Gauge Theories31 pages
- Greiner W., Reinhardt. Field QuantizationNo ratings yetGreiner W., Reinhardt. Field Quantization458 pages
- Advanced Concepts of Theoretical Physics: Uwe-Jens WieseNo ratings yetAdvanced Concepts of Theoretical Physics: Uwe-Jens Wiese145 pages
- Advanced Quantum Mechanics, Notes Based On Online Course Given by Leonard Susskind - Lecture 1No ratings yetAdvanced Quantum Mechanics, Notes Based On Online Course Given by Leonard Susskind - Lecture 17 pages
- Introduction To General Relativity Solutions 1-5No ratings yetIntroduction To General Relativity Solutions 1-55 pages
- Quantum Field Theory by Peskin - Chap15 SolutionNo ratings yetQuantum Field Theory by Peskin - Chap15 Solution7 pages
- Mathematical Theory of Electromagnetism: Piero BassaniniNo ratings yetMathematical Theory of Electromagnetism: Piero Bassanini384 pages
- Janis-Newman Algorithm: Generating Rotating and NUT Charged Black HolesNo ratings yetJanis-Newman Algorithm: Generating Rotating and NUT Charged Black Holes68 pages
- Introduction To Quantum Field Theory, Based On The Book by Peskin and Schroeder67% (3)Introduction To Quantum Field Theory, Based On The Book by Peskin and Schroeder155 pages
- Introduction To Quantum Field Theory: Marina Von Steinkirch State University of New York at Stony Brook March 3, 2011100% (2)Introduction To Quantum Field Theory: Marina Von Steinkirch State University of New York at Stony Brook March 3, 2011121 pages
- Introduction To Relativistic Quantum FieldsNo ratings yetIntroduction To Relativistic Quantum Fields260 pages
- Closed String Field Theory From PolyhedrNo ratings yetClosed String Field Theory From Polyhedr15 pages
- Problems in Quantum Field Theory With Fullyworked Solutions 1st Edition Gelis DownloadNo ratings yetProblems in Quantum Field Theory With Fullyworked Solutions 1st Edition Gelis Download79 pages
- Jean Zinn-Justin - Quantum Field Theory and Critical Phenomena-Oxford University Press (2021)100% (3)Jean Zinn-Justin - Quantum Field Theory and Critical Phenomena-Oxford University Press (2021)1,074 pages
- Combinatorial Physics: Combinatorics, Quantum Field Theory, and Quantum Gravity Models Adrian Tanasa PDF Download100% (4)Combinatorial Physics: Combinatorics, Quantum Field Theory, and Quantum Gravity Models Adrian Tanasa PDF Download50 pages
- Lecture 2: Perturbative Renormalization (Continued) : 2.1. Renormalizability of Quantum Eld TheoriesNo ratings yetLecture 2: Perturbative Renormalization (Continued) : 2.1. Renormalizability of Quantum Eld Theories12 pages
- Quantum Theory of Many-Body Systems Techniques and Applications (PDFDrive)No ratings yetQuantum Theory of Many-Body Systems Techniques and Applications (PDFDrive)291 pages
- Introduction To The AdS-CFT CorrespondenceNo ratings yetIntroduction To The AdS-CFT Correspondence122 pages
- Understanding Feynman Diagrams in PhysicsNo ratings yetUnderstanding Feynman Diagrams in Physics20 pages
- Arxiv Sunil Mukhi String Theory Review 1110.2569No ratings yetArxiv Sunil Mukhi String Theory Review 1110.256945 pages
- 8.325: Relativistic Quantum Field Theory III Problem Set 2No ratings yet8.325: Relativistic Quantum Field Theory III Problem Set 21 page
- Quantum Field Theory: Feynman Path Integrals and Diagrammatic Techniques in Condensed Matter 1st Edition Lukong Cornelius Fai DownloadNo ratings yetQuantum Field Theory: Feynman Path Integrals and Diagrammatic Techniques in Condensed Matter 1st Edition Lukong Cornelius Fai Download94 pages
- Feynman Rules for Quantum ElectrodynamicsNo ratings yetFeynman Rules for Quantum Electrodynamics35 pages
