P. 1


|Views: 102|Likes:
Published by jimling123

More info:

Published by: jimling123 on Jan 23, 2011
Copyright:Attribution Non-commercial


Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less






  • A. Relativistic Quantum Mechanics
  • B. Conventions and Notation
  • 1. Units
  • 2. Relativistic Notation
  • 3. Fourier Transforms
  • 4. The Dirac Delta “Function”
  • C. A Na¨ıve Relativistic Theory
  • A. Multi-particle Basis States
  • 1. Fock Space
  • 2. Review of the Simple Harmonic Oscillator
  • 3. An Operator Formalism for Fock Space
  • 4. Relativistically Normalized States
  • B. Canonical Quantization
  • 1. Classical Particle Mechanics
  • 2. Quantum Particle Mechanics
  • 3. Classical Field Theory
  • 4. Quantum Field Theory
  • C. Causality
  • A. Classical Mechanics
  • B. Symmetries in Field Theory
  • 1. Space-Time Translations and the Energy-Momentum Tensor
  • 2. Lorentz Transformations
  • C. Internal Symmetries
  • 1. U(1) Invariance and Antiparticles
  • 2. Non-Abelian Internal Symmetries
  • D. Discrete Symmetries: C, P and T
  • 1. Charge Conjugation, C
  • 2. Parity, P
  • 3. Time Reversal, T
  • A. Particle Creation by a Classical Source
  • B. More on Green Functions
  • C. Mesons Coupled to a Dynamical Source
  • D. The Interaction Picture
  • E. Dyson’s Formula
  • F. Wick’s Theorem
  • G. S matrix elements from Wick’s Theorem
  • H. Diagrammatic Perturbation Theory
  • I. More Scattering Processes
  • J. Potentials and Resonances
  • A. Decays
  • B. Cross Sections
  • C. D for Two Body Final States
  • A. Feynman Diagrams with External Lines off the Mass Shell
  • 1. Answer One: Part of a Larger Diagram
  • 2. Answer Two: The Fourier Transform of a Green Function
  • 3. Answer Three: The VEV of a String of Heisenberg Fields
  • B. Green Functions and Feynman Diagrams
  • C. The LSZ Reduction Formula
  • 1. Proof of the LSZ Reduction Formula
  • A. Transformation Properties
  • B. The Weyl Lagrangian
  • C. The Dirac Equation
  • 1. Plane Wave Solutions to the Dirac Equation
  • D. γ Matrices
  • 1. Bilinear Forms
  • 2. Chirality and γ5
  • E. Summary of Results for the Dirac Equation
  • 2. Space-Time Symmetries
  • 3. Dirac Adjoint, γ Matrices
  • 4. Bilinear Forms
  • 5. Plane Wave Solutions
  • A. Canonical Commutation Relations
  • B. Canonical Anticommutation Relations
  • C. Fermi-Dirac Statistics
  • D. Perturbation Theory for Spinors
  • 1. The Fermion Propagator
  • 2. Feynman Rules
  • E. Spin Sums and Cross Sections
  • A. The Classical Theory
  • B. The Quantum Theory
  • C. The Massless Theory
  • 1. Minimal Coupling
  • 2. Gauge Transformations
  • D. The Limit µ→0
  • 1. Decoupling of the Helicity 0 Mode
  • E. QED
  • F. Renormalizability of Gauge Theories

PHY2403F Lecture Notes

Michael Luke
(Dated: Fall, 2007)

These notes are perpetually under construction. Please let me know of any typos or errors. I claim little originality; these notes are in large part an abridged and revised version of Sidney Coleman’s field theory lectures from Harvard.

I. Introduction A. Relativistic Quantum Mechanics B. Conventions and Notation 1. Units 2. Relativistic Notation 3. Fourier Transforms 4. The Dirac Delta “Function” C. A Na¨ Relativistic Theory ıve II. Constructing Quantum Field Theory A. Multi-particle Basis States 1. Fock Space 2. Review of the Simple Harmonic Oscillator 3. An Operator Formalism for Fock Space 4. Relativistically Normalized States B. Canonical Quantization 1. Classical Particle Mechanics 2. Quantum Particle Mechanics 3. Classical Field Theory 4. Quantum Field Theory C. Causality III. Symmetries and Conservation Laws A. Classical Mechanics B. Symmetries in Field Theory 1. Space-Time Translations and the Energy-Momentum Tensor 2. Lorentz Transformations C. Internal Symmetries 1. U (1) Invariance and Antiparticles 2. Non-Abelian Internal Symmetries D. Discrete Symmetries: C, P and T 1. Charge Conjugation, C 2. Parity, P 3. Time Reversal, T IV. Example: Non-Relativistic Quantum Mechanics (“Second Quantization”) V. Interacting Fields A. Particle Creation by a Classical Source B. More on Green Functions 5 5 9 9 10 14 15 15 20 20 20 22 23 23 25 26 27 29 32 37 40 40 42 43 44 48 49 53 55 55 56 58 60 65 65 69

C. Mesons Coupled to a Dynamical Source D. The Interaction Picture E. Dyson’s Formula F. Wick’s Theorem G. S matrix elements from Wick’s Theorem H. Diagrammatic Perturbation Theory I. More Scattering Processes J. Potentials and Resonances VI. Example (continued): Perturbation Theory for nonrelativistic quantum mechanics VII. Decay Widths, Cross Sections and Phase Space A. Decays B. Cross Sections C. D for Two Body Final States VIII. More on Scattering Theory A. Feynman Diagrams with External Lines off the Mass Shell 1. Answer One: Part of a Larger Diagram 2. Answer Two: The Fourier Transform of a Green Function 3. Answer Three: The VEV of a String of Heisenberg Fields B. Green Functions and Feynman Diagrams C. The LSZ Reduction Formula 1. Proof of the LSZ Reduction Formula IX. Spin 1/2 Fields A. Transformation Properties B. The Weyl Lagrangian C. The Dirac Equation 1. Plane Wave Solutions to the Dirac Equation D. γ Matrices 1. Bilinear Forms 2. Chirality and γ5 E. Summary of Results for the Dirac Equation 1. Dirac Lagrangian, Dirac Equation, Dirac Matrices 2. Space-Time Symmetries 3. Dirac Adjoint, γ Matrices 4. Bilinear Forms 5. Plane Wave Solutions X. Quantizing the Dirac Lagrangian A. Canonical Commutation Relations 70 71 73 77 80 83 88 92 95 98 101 102 103 106 106 107 108 110 111 115 117 125 125 131 135 137 139 142 144 146 146 146 147 148 149 151 151

Perturbation Theory for Spinors 1. The Fermion Propagator 2. Vector Fields and Quantum Electrodynamics A. Gauge Transformations D. The Limit µ → 0 1. Feynman Rules E. Canonical Anticommutation Relations C. The Quantum Theory C. QED F. Decoupling of the Helicity 0 Mode E. Fermi-Dirac Statistics D. Spin Sums and Cross Sections XI. The Massless Theory 1. Minimal Coupling 2. How Not to Quantize the Dirac Lagrangian B.4 or. The Classical Theory B. Renormalizability of Gauge Theories 151 153 155 157 159 161 166 168 168 173 176 179 182 184 186 187 189 .

For example. Why does the addition of Lorentz invariance complicate quantum mechanics? The answer is very simple: in relativistic systems. I. In a quantum system. The proton and antiproton beams travel perpendicular to the page. this has profound implications. colliding at the origin of the tracks.5 jet #2 jet #3 jet #1 jet #4 e+ FIG. additional symmetries simplify physical problems. the radius of the curvature of the path of a particle provides a means to determine its mass. At higher energies where relativity is important things gets more complicated. Relativistic Quantum Mechanics Usually. and therefore identify it. because if E ∼ mc2 there is enough energy to pop additional particles out of the vacuum (we will discuss how this works at length in the course). before and after the scattering process. I.1 The results of a proton-antiproton collision at the Tevatron. The tracks are curved because the detector is placed in a magnetic field. There is only one particle. Each of the curved tracks indicates a charged particle in the final state. E mc2 where relativity is unimportant. the number of particles is not conserved. INTRODUCTION A. in non-relativistic quantum mechanics (NRQM) rotational invariance greatly simplifies scattering problems. in p-p (proton-proton) scattering with a centre of mass energy E > mπ c2 (where mπ ∼ 140 MeV is the mass of the neutral pion) the process p + p → p + p + π0 . scattering a particle in potential. NRQM provides a perfectly adequate description. Consider. For example. The incident particle is in some initial state. At low energies. for example. and one can fairly simply calculate the amplitude for it to scatter into any final state.

particle anti-particle pairs can pop out of the vacuum. On the other hand. the up and down quarks which make up the proton have masses of order 10 MeV (λc 20 fm) and are confined to a region the size of a proton. I. the problems with NRQM run much deeper.511 MeV × 197 MeV fm ∼ 4 × 10−11 cm. E > 2mp c2 . but rather the state of lowest energy . what started out as a simple two-body scattering process has turned into a many-body problem. or about 1 fm. The most energetic accelerator today is the Tevatron at Fermilab. However.6 is possible. so typical collisions produce a huge slew of particles (see Fig. There is therefore no sense in which it is possible to localize a particle in a region smaller than its Compton wavelength. So there is no problem localizing an electron on atomic scales. we can localize the particle in an arbitrarily small region. or about 103 mp c2 . Consider the familiar problem of a particle in a box. and it is necessary to calculate the amplitude to produce a variety of many-body final states. L < ¯ /µc (where h/µc ≡ λc . Therefore. is 1/0. the ¯ ∼h uncertainty in the energy of the system is large enough for particle creation to occur . or about 10−3 Bohr radii. one. Clearly we will have to construct a many-particle quantum theory to describe such a process. one can produce an additional proton-antiproton pair: p+p→p+p+p+p and so on. and relativistic effects will be huge. which collides protons and antiprotons with energies greater than 1 TeV. this does not introduce any problems. h In the relativistic regime. In atomic physics. and the relativistic corrections due to multi-particle states are small.1). as long as we accept an arbitrarily large uncertainty in its momentum. or even zero . In the nonrelativistic description. The smaller the distance scale you look at it.which in an interacting quantum theory is not the zero-particle state. At higher energies. But relativity tells us that this description must break down if the box gets too small. this translates to an uncertainty of order hc/L in the particle’s energy. ¯ For L small enough.511 MeV/c2 ).is complicated. Even the vacuum state . the more complex its structure. The Compton wavelength of an electron (mass µ = 0. Consider a particle of mass µ trapped in a container with reflecting walls of side L. Thus. where NRQM works very well. Clearly the internal structure of the proton is much more complex than a simple three quark system. there is no such thing in relativistic quantum mechanics as the two. outside Chicago. The uncertainty in the particle’s momentum is therefore of order ¯ /L. making the number of particles in the container uncertain! The physical state of the system is a quantum-mechanical superposition of states with different particle number. as a brief contemplation of the uncertainty principle indicates. the Compton wavelength of the particle).

because observables separated by spacelike separations will be able to . which is that of causality. In both relativistic and nonrelativistic quantum mechanics observables correspond to Hermitian operators. even incomplete (usually perturbative) solutions will give us a great deal of understanding and predictive power. having different observables at each space-time point) we will run into trouble with causality.e. λc = h/µ. in a relativistic theory we have to be more careful. As a general conclusion. is totally intractable analytically. the number of particles in the box is therefore indeterminate. and so on. The first casualty of relativistic QM is the position operator. Nevertheless. However. and it will not arise in the formalism which we will develop. I. single particle quantum theory. Even the nature of the vacuum state in the real world. the uncertainty in the energy of the system allows particle production to occur. observables are not attached to spacetime points . a horribly complex sea of quark-antiquark pairs.2 A particle of mass µ cannot be localized in a region smaller than its Compton wavelength. as we shall see in this course. one is always dealing with the infinite body problem. body problem! In principle. however. So we will have to set up a formalism to handle many-particle systems. Furthermore. relativistic. Thus. gluons. it should be clear from this discussion that our old friend the position operator X from NRQM does not make sense in a relativistic theory: the {| x } basis of NRQM simply does not exist.7 L L (a) L> h/µc > (b) L< <h/µc FIG. except in very simple toy models (typically in one spatial dimension). There is a second. it is impossible to solve any relativistic quantum system exactly. In NRQM. you cannot have a consistent. since particles cannot be localized to arbitrarily small regions. the momentum operator. because making a measurement forces the system into an eigenstate of the corresponding operator. electron-positron pairs as well as more exotic beasts like Higgs condensates and gravitons.one simply talks about the position operator. ¯ At smaller scales. intimately related problem which arises in a relativistic quantum theory. Unless we are careful about only defining observables locally (i.

So imagine that Observer One has an electron and measures the x−component of its spin. In this case their measurements can interfere with one another. since there are reference frames in which Observer One’s second measurement preceded Observer Two’s measurement (recall that the time-ordering of spacelike separated events depends on the frame of reference). Therefore. so observables at point 1 must commute with all observables at point 2. One and Two. the next time Observer One measures σx it has a 50% chance of being in the opposite spin state. forcing it into an eigenstate of the spin operator σx . so the time ordering is frame dependent. One could be here and Observer Two in the Andromeda galaxy. Classical physics got away from action at a distance by introducing electromagnetic and gravitational fields.). and the dynamics of the fields are purely local . and don’t refer to particular space-time points. This of course violates causality.8 interfere with one another. They have communicated at faster than the speed of light. Now suppose that these two observers both decide to measure non-commuting observables.. measurements made at the two points cannot interfere. Consider applying the NRQM approach to observables to a situation with two observers. A Lorentz boost will move the observer O2 along the hyperboloid (∆t)2 − |∆x|2 = constant. If Observer Two measures a non-commuting observable such as σy . at space-time points x1 and x2 .. as well . which are separated by a spacelike interval. and they are not in causal contact. and so she can immediately tell that Observer Two has made a measurement. Observer t t´ O2 O1 boost O1 O2 FIG.3 Observers O1 and O2 are separated by a spacelike interval. The fields are defined at all spacetime points.the dynamics of the field at a point xµ are determined entirely by the physical quantities (the various fields and their derivatives. and leads to all sorts of paradoxes (maybe Observer Two then changes his mind and doesn’t make the measurement . I. The problem with NRQM in this context is that it has action at a distance built in: observables are universal.

or more commonly MeV. 1. From the fact that velocity (L/T ) and action (M L2 /T ) are dimensionless we find that length and time have units of eV−1 . in these units all dimensionful quantities may be expressed in terms of a single unit. we can get away from action at a distance by promoting all of our operators to quantum fields: operator-valued functions of space-time whose dynamics is purely local. (I. h h Indeed. O2 (x2 )] = 0 for (x1 − x2 )2 < 0. or.) This makes life much simpler. relativistic quantum mechanics is usually known as “Quantum Field Theory. In the old units it is α= e2 1 = . which is a fundamental dimensionless number characterizing the strength of the electromagnetic interaction to a single charged particle. For example. Consider the fine structure constant. which we usually take to be mass. therefore. equivalently. Spacelike separated measurements cannot interfere with one another. Conventions and Notation Before delving into QFT. In relativistic quantum mechanics. energy (since E = mc2 becomes E = m). or between frequency ω and energy E = ¯ ω. we must have [O1 (x1 ). in which h ¯ = c = 1 (we do this by choosing units such that one unit of velocity is c and one unit of action is h ¯ . we will set a few conventions for the notation we will be using in this course. When we refer to the dimension of a quantity in this course we mean the mass dimension: if X has dimensions of (mass)d . Units We will choose the “natural” system of units to simplify formulas and calculations. if O1 (x1 ) and O2 (x2 ) are observables which are defined at the space-time points x1 and x2 . GeV (=109 eV) or TeV (=1012 ) eV.1) B.” The requirement that causality be respected then simply translates into a requirement that spacelike separated observables commute: as our example demonstrates. In particle physics we usually take the choice of energy unit to be the electron-Volt (eV). Hence. we no longer have to distinguish h between wavenumber k and momentum P = ¯ k. we write [X] = d. by setting ¯ = 1.04 . 4π¯ c h 137.9 as the charge density) at that point.

Just to remind you of the distinction. 2) subscripts are labels. 2) basis.4) .7 MeV 134 MeV 938. 4π 137. h ¯ = 6.97 × 10−11 MeV cm (I. h It is easy to convert a physical quantity back to conventional units by using the following.3) where 1 fm (femtometer. not coordinates).2) By multiplying or dividing by these factors you can convert factors of MeV into sec or cm.58 × 10−22 MeV sec h ¯ c = 1. Relativistic Notation When dealing with non-orthogonal coordinates. not indices: e1 and e2 are vectors.10 In the new units it is α= e2 1 = . consider the set of two-dimensional non-orthogonal coordinates on the plane shown in Fig.6 MeV 5. Some particle masses in natural units are: particle e− (electron) µ− (muon) π 0 (pion) p (proton) n (neutron) B (B meson) W + (W boson) Z 0 (Z boson mass 511 keV 105.4). Now consider the coordinates of a point x in the (1. it is of crucial importance to distinguish between contravariant coordinates xµ and covariant coordinates xµ .04 Thus the charge e has units of (¯ c)1/2 in the old units. but it is dimensionless in the new units. (I.279 GeV 80. A useful conversion is h ¯ c = 197 MeV fm (I. or “fermi”)= 10−13 cm is a typical nuclear scale. In terms of the unit vectors e1 ˆ and e2 (where the (1.2 GeV 91.3 MeV 939. ˆ ˆ ˆ we can write x = x1 e1 + x2 e2 ˆ ˆ (I.17 GeV 2.

5) which are also shown on the figure. we have x · y = x · (y 1 e1 + y 2 e2 ) ˆ ˆ = y 1 x · e1 + y 2 x · e2 ˆ ˆ = y 1 x1 + y 2 x2 = y1 x1 + y2 x2 (I. However. away from Euclidean space (in particular. From the definitions above. which is how you made it this far without worrying about the distinction. x2 ) are defined by x1.6) so scalar products are always obtained by pairing upper with lower indices. it is simple to take the scalar product of two vectors. Note that for orthogonal axes in flat (Euclidean) space there is no distinction between covariant and contravariant coordinates.2 ≡ x · e1. in Minkowski space-time) the distinction is crucial.2 ˆ (I. The covariant coordinates (x1 . these distances are marked on the diagram. The relation between contravariant and covariant coordinates is straightforward to derive: xi = (x1 e1 + x2 e2 ) · ei ˆ ˆ ˆ = xj (ˆi · ej ) e ˆ ≡ gij xj where we have defined the metric tensor gij ≡ ei · ej . ˆ ˆ (I. Given the two sets of coordinates. which defines the contravariant coordinates x1 and x2 .8) (I.11 2 x 1 FIG.4 Non-orthogonal coordinates on the plane. I.7) .

and upper indices are always paired with lower indices (see Fig.13) . repeated indices are summed over.10) Minkowskian space is a simple situation in which we use non-orthogonal basis vectors. 2. (I. −r).5)). If in doubt. this notation was designed to make your life easier! Under a Lorentz transformation a four-vector transforms according to matrix multiplication: x µ (I. a µ b µ = aµ bµ .9) j j (note that gi = δi . It easily follows that this is Lorentz invariant. the Kronecker delta). The flat Minkowski space metric is  1 0 0 0     0 −1 0  µν gµν = g =    0 0 −1  0 0 0 −1   0   .12 Note that we are also using the Einstein summation convention: repeated indices (always paired upper and lower) are implicitly summed over. above. (I. 0    (I. The scalar product of two four-vectors is written as aµ bµ = aµ bµ = aµ gµν bν = a0 b0 − a · b. Note that as before. If you get an expression like aµ bµ (this isn’t a scalar because the upper and lower indices aren’t paired) or (worse) aµ bµ cµ dµ (which indices are paired with which?) you’ve probably made a mistake. Despite our location. (I. it’s sometimes helpful to include explicit summations until you get the hang of it. x. but since we will always be working in flat space in this course. z) where µ = 0. This ensures that the result of the contraction is a Lorentz scalar. The contravariant components of the four-vector xµ are (t. xi = g ij xj . Remember. because time and space look different. r) = (t. Note that some texts define gµν as minus this (the so-called “east-coast convention”). y. One can also define the metric tensor with raised and mixed indices via the relations k gij ≡ gik gj ≡ gik gjl g kl (I. we’ll adopt the “west-coast convention”. The metric tensor g ij raises indices in the natural way.12) = Λµ ν xν . we will use gµν and ηµν interchangeably. 1.11) Often the flat Minkowski space metric is denoted ηµν . 3. The metric tensor is used to raise and lower indices: xµ = gµν xν = (t.

Note that ∂ µ Aµ = ∂ 0 A0 + ∂ j Aj (I. .17) Thus. where the 4 × 4 matrix Λµ ν defines the Lorentz transformation. we note that the variation δφ = ∂φ µ δx ∂xµ (I. Special cases of Λµ ν include space rotations and “boosts”. .− . To see how derivatives transform under Lorentz transformations.5 Be careful with indices. Thus we define ∂µ ≡ and ∂µ ≡ ∂ = ∂xµ ∂ ∂ ∂ ∂ . I.15) is a scalar and we would therefore like to write it as δφ = ∂µ φδxµ .− .− . ∂/∂xµ transforms as a covariant (lower indices) four-vector.13 a b c FIG.18) ∂ = ∂xµ ∂ ∂ ∂ ∂ .16) (I.19) . The set of all Lorentz transformations may be defined as those transformations which leave gµν invariant: gµν = gαβ Λα µ Λβ ν . which look as follows:  1 0 0 0       0 cos θ − sin θ 0    µ Λ ν (rotation about z−axis) =      0 sin θ cos θ 0    0 0 0 1  γ −γv 0 0 γ 0 0     −γv  Λµ ν (boost in x direction) =    0  0   0 0   1 0   (I.14) 0 1 with γ = (1 − v 2 )−1/2 . ∂t ∂x ∂y ∂z (I. ∂t ∂x ∂y ∂z (I.

Fourier Transforms We will frequently need to go back and forth between the position (x) and momentum (or wavenumber) (p or k) space descriptions of a function. β) is an even permutation of (0. 3.1.1. ν. that is. β) is an odd permutation of (0.21) µναβ =  −1 if (µ. Also remember that in Minkowski space. You should is a relativistically invariant tensor. we will make use (particularly in the section of Dirac fields) of the completely antisymmetric tensor µναβ (often known as the Levi-Civita tensor).2. while dn x’s have no such factors.     0 if (µ.22) and (I. which is frequently a very useful thing to do. . α. that under a Lorentz transformation the properties (I.2.3). ν.23) We have introduced two conventions here which we shall stick to in the rest of the course: the sign of the exponentials (we could just as easily have reversed the signs of the exponentials in Eqs. As you should ˜ recall. via the Fourier transform. where E = k0 and t = x0 . so Fourier transforming a field will allow us to write it as a sum of modes with definite momentum.2. 0123 Note that you must be careful with raised or lowered indices. ν.21) still hold. (2π)n (I. (I. It is defined by       1 if (µ. k · x = Et − k · x. α.3).14 and ∂ µ ∂µ = ∂2 − ∂t2 2 = 2. α. Finally. the Fourier transform f (k) allows any function f (x) to be expanded on a continuous basis of plane waves. plane waves correspond to eigenstates of momentum.every time you see a dn k it comes with a factor of (2π)−n .20) The energy and momentum of a particle together form the components of its 4-momentum P µ = (E.3). (I.23)) and the placement of the factors of 2π. In n dimensions we therefore write f (x) = dn k ˜ f (k)eik·x . β) is not a permutation of (0. In quantum mechanics. (I.1. since verify that (like the metric tensor g µν ) µναβ =− 0123 = 1. p). The latter convention will prove to be convenient because it allows us to easily keep track of powers of 2π .22) ˜ It is simple to show that f (k) is therefore given by ˜ f (k) = dn xf (x)e−ik·x (I.

in this section we will illustrate with a simple example the somewhat abstract worries about causality we had in the previous section. A Na¨ Relativistic Theory ıve Having dispensed with the formalities.δ(xn ) which satisfies dn x δ (n) (x) = 1.15 4. x = 0. dx (I. however. and discover that the theory violates causality: a single free particle will have a nonzero amplitude to be found to have travelled faster than the speed of light. C. in n dimensions we may define the n dimensional delta function δ (n) (x) ≡ δ(x0 )δ(x1 ).25) We will also make use of the (one-dimensional) step function 1. (I.29) Note that the symbol x will sometimes denote an n-dimensional vector with components xµ .29) . The δ function can be written as the Fourier transform of a constant. . we will usually distinguish three-vectors (x) from four-vectors (x or xµ ). (I. as in Eq.26). We will construct a relativistic quantum theory as an obvious relativistic generalization of NRQM..28) (I.26) (I. Similarly. as in Eq. For clarity.30) x<0 x>0 (I. δ (n) (x) = 1 (2π)n dn p eip·x .. which satisfies ∞ dx δ(x) = 1 −∞ (I. θ(x) = 0.24) and δ(x) = 0. and sometimes a single coordinate.it should be clear from context. which satisfies dθ(x) = δ(x). The Dirac Delta “Function” We will frequently be making use in this course of the Dirac delta function δ(x).27) (I. (I.

it appears that we can make this theory relativistic simply by replacing the Hamiltonian in Eq.33) (I. (Note that in our notation.34) (I. while the components of k are just numbers.36) is | ψ(t ) = e−iH(t −t) | ψ(t) .35) (I.39) . spinless particle of mass µ. 2µ (I. P is an operator on the Hilbert space. (I.38) by the relativistic Hamiltonian Hrel = The basis states now satisfy Hrel | k = ωk | k (I.) These states are normalized k |k = δ (3) (k − k ) and satisfy the completeness relation d3 k | k k | = 1. the components of momentum form a complete set of commuting observables). ∂t (I.38) (I. In NRQM.40) |P |2 + µ2 .16 Consider a free.37) If we rashly neglect the warnings of the first section about the perils of single-particle relativistic theories. (I. An arbitrary state | ψ is a linear combination of momentum eigenstates |ψ = d3 k ψ(k) | k (I.32) ψ(k) ≡ k |ψ . The time evolution of the system is determined by the Schr¨dinger equation o i ∂ | ψ(t) = H| ψ(t) .36) where the operator H is the Hamiltonian of the system. The state of the particle is completely determined by its three-momentum k (that is. for a free particle of mass µ. The solution to Eq. (I.31) where P is the momentum operator. We may choose as a set of basis states the set of momentum eigenstates {| k }: P | k = k| k (I. H| k = |k|2 |k .

it is instructive to show this explicitly.43) Now let us imagine that at t = 0 we have localized a particle at the origin: | ψ(0) = | x = 0 . (2π)3/2 (I. (I. (I.44) and (I. X.46) Inserting the completeness relation Eq. (I. This theory looks innocuous enough.45) After a time t we can calculate the amplitude to find the particle at the position x. The angular integrals are straightforward. giving x |ψ(t) = − i (2π)2 r ∞ −∞ k dk eikr e−iωk t .42) |k|2 + µ2 . This is just x |ψ(t) = x |e−iHt | x = 0 .41) (remember.47) where we have defined k ≡ |k| and r ≡ |x|.33) and using Eqs. we are setting h = 1 in everything that follows).44) ∂ ψ(k) ∂ki (I.48) . (I. if we prepare a particle localized at one position. matrix elements ¯ of X are given by k |Xi | ψ = i and position eigenstates by k |x = 1 e−ik·x .40) we can express this as x |ψ(t) = = = 0 d3 k x |k k |e−iHt | x = 0 d3 k ∞ 1 ik·x −iωk t e e (2π)3 k 2 dk π dθ sin θ (2π)3 0 2π 0 dφ eikr cos θ e−iωk t (I.17 where ωk ≡ is the energy of the particle. In the {| k } basis. we introduce the position operator. To measure the position of a particle. (I. We will find that. Nevertheless. satisfying [Xi . (I. Pj ] = iδij (I. there is a non-zero probability of finding it outside of its forward light cone at some later time. We have already argued on general grounds that it cannot be consistent with causality.

49) means that for distances r 1/µ there is a negligible chance to find the particle outside the light-cone. (I. it is deformed to the dashed path (where the radius of the semicircle is infinite). Changing variables to z = −ik. The integral is along the real axis. For r > t. This is in accordance with our earlier arguments based on the uncertainty principle: multi-particle effects become important when you are working at distance scales of order the Compton wavelength of a particle. and the integrand is analytic everywhere in the plane except for branch cuts at k = ±iµ. so the theory is acausal. so the integral may be rewritten as an integral along the branch cut. The contour integral can be deformed as shown in Fig. for a point outside the particle’s forward light cone. (I.6 Contour integral for evaluating the integral in Eq.49) The integrand is positive definite. e−µr in Eq.48). i. so the integral is non-zero. Consider the integral Eq.18 k Im k+µ > 0 2 2 Im iµ k+µ < 0 2 2 .iµ FIG. Note the exponential envelope. We will see in a few lectures that one of the most striking predictions of QFT is the existence of antiparticles with the same mass as. the single-particle theory will not lead to measurable violations of causality. For r > t. the integrand vanishes exponentially on the circle at infinity in the upper half plane.6). The only contribution to the integral comes from integrating along the branch cut. we can prove using contour integration that this integral is non-zero. x |ψ(t) = − i (2π)2 r i −µr = e 2π 2 r ∞ µ ∞ µ √ √ 2 2 2 2 (iz)d(iz)e−zr e z −µ t − e− z −µ t dz ze−(z−µ)r sinh z 2 − µ2 t . (I.e. but opposite . arising from the square root in ωk .48) defined in the complex k plane. I. (I. How does the multi-particle element of quantum field theory save us from these difficulties? It turns out to do this in a quite miraculous way. so at distances much greater than the Compton wavelength of a particle. (I. The particle has a small but nonzero probability to be found outside of its forward light-cone. The original path of integration is along the real axis.

As it turns out. what appears to be a particle travelling from O1 to O2 in the frame on the left looks like an antiparticle travelling from O2 to O1 in the frame on the right.19 quantum numbers of.3). . In a Lorentz invariant theory. Therefore. the amplitudes exactly cancel. there is no Lorentz invariant distinction between emitting a particle at x and absorbing it at y. both processes must occur. we must add the amplitudes for these two processes. and they are indistinguishable. (I. Now. since the time ordering of two spacelikeseparated events at points x and y is frame-dependent. the corresponding particle. so causality is preserved. if we wish to determine whether or not a measurement at x can influence a measurement at y. and emitting an antiparticle at y and absorbing it at x: in Fig.

but instead is simply a parameter. k2 . the unphysical question “where is the particle at time t” is replaced by physical questions such as “what is the expectation value of the observable O (the electric field. Therefore.4) (II. CONSTRUCTING QUANTUM FIELD THEORY A. particles are defined analogously. k2 = δ (3) (k1 − k1 )δ (3) (k2 − k2 ) + δ (3) (k1 − k2 )δ (3) (k2 − k1 ) H| k1 . The first thing we need to do is define the states of the system. k2 |k1 . k2 = (k1 + k2 )| k1 . {| k }. .. The momentum operator is fine. but now this is only a piece of the Hilbert space.5) We will postpone the study of fermions until later on.20 II. these states are even under particle interchange1 | k1 . (II. momentum is a conserved quantity and can be measured in an arbitrarily small volume element. In other words.” Therefore. like the time t. position is no longer an observable. when we discuss spinor fields. Because the particles are bosons. causal quantum theory.4. 1 P|0 = 0 (II. we now proceed to set up the formalism for a consistent theory. k2 = (ωk1 + ωk2 )| k1 . k1 . etc. There is also a zero-particle state. The basis of two-particle states is {| k1 . we can’t use position eigenstates as our basis states. Multi-particle Basis States 1. k2 P | k1 . They also satisfy k1 . x)..) at the space-time point (t. In QFT. the energy density.3. Fock Space Having killed the idea of a single particle. relativistic. . k2 = | k2 .) However. k2 }.2) (II. The basis for our Hilbert space in relativistic quantum mechanics consists of any number of spinless mesons (the space is called “Fock Space”.3) (II. the vacuum | 0 : 0 |0 = 1 H| 0 = 0. we choose as our single particle basis states the same states as before. we saw in the last section that a consistent relativistic theory has no position operator.1) States with 2.

This is nice because the wavefunctions in the box are normalizable.. We need a better description. not any single k. k2 k1 . ny . As a pedagogical device. ky .6) (The factor of 1/2! is there to avoid double-counting the two-particle states. n(k ). An interaction term in the Hamiltonian which creates a particle will connect the single-particle wave-function to the two-particle wave-function.. An arbitrary state will have a wave function over the single-particle basis which is a function of 3 variables (kx .7) where the n(k)’s give the number of particles of each momentum in the state. | . (II. and the allowed values of k are discrete... .10) This is bears a striking resemblance to a system we have seen before. kz ). In terms of N (k) the Hamiltonian and momentum operator are H= k (II.21 and the completeness relation for the Hilbert space is 1 = |0 0| + d3 k| k k | + 1 2! d3 k1 d3 k2 | k1 . N (k)| n(·) = n(k)| n(·) . Since translation by L must leave the system unchanged.. . We can then write our states in the occupation number representation.8) 2πnx 2πny 2πnz . a wave function over the two-particle basis which is a function of 6 variables. L L L (II. 1 For a single oscillator. it will often be convenient in this course to consider systems confined to a periodic box of side L. HSHO = ω(N + 2 ).8) is written | n(·) where the (·) indicates that the state depends on the function n for all k’s. Fock space . where N is the excitation level of the oscillator. The number operator N (k) counts the occupation number for a given k.. nz integers. the two-particle to the three-particle.9) ωk N (k) P = k kN (k). (II. . (II. Sometimes the state (II. the simple harmonic oscillator. k2 | + .) This is starting to look unwieldy. .n(k). the allowed momenta must be of the form k= for nx . preferably one which has no explicit multi-particle wave-functions.. This will be a mess.. and so forth.

15) that Ha† | E = (E + ω)a† | E Ha| E = (E − ω)a† | E . q] = −i)..13) The raising and lowering operators a and a† are defined as q + ip q − ip a = √ . E − ω. If H| E = E| E .H. and up to an (irrelevant) overall constant..O. the Hamiltonians for the two theories look the same. N | n = n| n . it follows from (II.14) so there is a ladder of states with energies . Since ψ |a† a| ψ = |a| ψ |2 ≥ 0..15) (II. Review of the Simple Harmonic Oscillator The Hamiltonian for the one dimensional S. . it is easy to show that the constant of proportionality cn = 1/ n!. [H. | n = cn (a† )n | 0 ..12) (the transformation is canonical because it preserves the commutation relation [P. (II. . E + 2ω. 2µ 2 (II.. In terms of p and q the Hamiltonian (II. The higher states are made by repeated applications of a† . (II. We can make use of that correspondence to define a compact notation for our multiparticle theory..16) (II. E.17) √ Since n |aa† | n = n + 1. X → q = µωX µω (II. 2. there is a lowest weight state | 0 satisfying N | 0 = 0 and a| 0 = 0.11) is HSHO = ω 2 (p + q 2 ).11) We can write this in a simpler form by performing the canonical transformation P √ P → p = √ . is HSHO = P2 1 2 + ω µX 2 . a† ] = ωa† . a† = √ 2 2 and satisfy the commutation relations [a. [H. E + ω. a] = −ωa where H = ω(a† a + 1/2) ≡ ω(N + 1/2). 2 (II.22 is in a 1-1 correspondence with the space of an infinite system of independent harmonic oscillators. X] = [p. a† ] = 1.

| 0 . P | k = k| k . Define creation and annihilation operators ak and a† for each momentum k (remember. | k1 . . k (II.. any observable may be written in terms of creation and annihilation operators. define creation and annihilation operators in the continuum. In fact.22) At this point we can remove the box and.21) ωk a† ak . a† ] = 0.23 3. satisfies ak | 0 = 0 and the Hamiltonian is H= k (II. An Operator Formalism for Fock Space Now we can apply this formalism to Fock space.} form a perfectly good basis for Fock Space.19) (II. since the normalization and completeness relations clearly treat . The vacuum state. k = a† a† | 0 k k and so on. which is what makes them so useful.20) (II. a† ] = δkk . k2 . k k k The single particle states are | k = a† | 0 . with the obvious substitutions. This is not unexpected. but will sometimes be awkward in a relativistic theory because they don’t transform simply under Lorentz transformations. 4. we are still working in a box so the allowed momenta k are discrete). k the two-particle states are | k. [ak .18) (II.23) it is easy to check that we recover the normalization condition k | k = δ (3) (k − k ) and that H| k = ωk | k . We have seen explicitly that the energy and momentum operators may be written in terms of creation and annihilation operators.. ak ] = [a† . These obey the commutation relations [ak . Relativistically Normalized States The states {| 0 . a† ] = δ (3) (k − k ). [ak . a† ] = 0 k k k (II. ak ] = [a† . Taking [ak . | k .

29) (The factor of (2π)3/2 is there by convention .24). Eq.26) Since the completeness relation. k )| k (II. The components of the four-vector k µ = (ωk . (I. But this tells us nothing about the normalization of the transformed state. This is easy to see from the completeness relation. d3 k | k k | = we must have O(Λ)| k = ωk |k ωk (II.24 spatial components of k µ differently from the time component.33). As we will show in a moment. under the Lorentz transformation (II. for states which have a nice relativistic normalization. (II. our states don’t have a nice relativistic normalization. we can see how our basis states transform under Lorentz transformations by just looking at the single-particle states. Of course. holds for both primed and unprimed states. Unfortunately. Eq.) The states | k now transform simply under Lorentz transformations: O(Λ)| k = | k . it only tells us that O(Λ)| k = λ(k. Let O(Λ) be the operator acting on the Hilbert space which corresponds to the Lorentz transformation x µ = Λµ ν xν .33). k) transform according to k µ = Λµ ν k ν .30) . under a Lorentz transformation. λ would be one. (II.it will make factors of 2π come out right in the Feynman rules we derive later on.25) where k is given by Eq. and λ is a proportionality constant to be determined. (II. Since multi-particle states are just tensor products of single-particle states.24) Therefore. (I.28) d3 k | k k |=1 (II. a state with three momentum k is obviously transformed into one with three momentum k .24) the volume element d3 k transforms as d3 k → d3 k = ωk 3 d k. ωk (II.27) which is not a simple transformation law. because d3 k is not a Lorentz invariant measure. Therefore we will often make use of the relativistically normalized states |k ≡ √ (2π)3 2ωk | k (II.

26). Since a proper Lorentz transformation doesn’t change the direction of time. such as | k . whereas states with four-vectors. (II. but the four-volume element d4 k is.25 The convention I will attempt to adhere to from this point on is states with three-vectors. are relativistically normalized. such as | k . The easiest way to derive Eq. whereas the nonrelativistically normalized states obeyed the orthogonality condition k |k = δ (3) (k − k ) the relativistically normalized states obey k |k = (2π)3 2ωk δ (3) (k − k ). which suggests that the fundamental degrees of freedom in our theory should be .31) (Note that the θ function restricts us to positive energy states.33) (II.34) (II.32) The factor of ωk compensates for the fact that the δ function is not relativistically invariant. 2k (II. (II. As we argued in the last section.26) is simply to note that d3 k is not a Lorentz invariant measure. Since the free-particle states satisfy k 2 = µ2 . are non-relativistically normalized. B. (II.) Performing the k 0 integral with the δ function yields the measure d3 k .35) (II. we can restrict k µ to the hyperboloid k 2 = µ2 by multiplying the measure by a Lorentz invariant function: d4 k δ(k 2 − µ2 ) θ(k 0 ) = d4 k δ((k 0 )2 − |k|2 − µ2 ) θ(k 0 ) d4 k = 0 δ(k 0 − ωk ) θ(k 0 ). Finally. we expect that causality will require us to define observables at each point in space-time.T. Canonical Quantization Having now set up a slick operator formalism for a multiparticle theory based on the SHO. this term is also invariant under a proper L. we now have to construct a theory which determines the dynamics of observables. 2ωk Under a Lorentz boost our measure is now invariant: d3 k d3 k = ωk ωk which immediately gives Eq.

we get t2 δS = t1 dt a ∂L − pa δqa + pa δqa ˙ ∂qa t2 t1 . . The action. For the theory to be causal. t) = T − V . φ}). this gives t2 δS = t1 dt a ∂L ∂L δqa + δ qa .37) by parts. Define the Hamiltonian H(q1 .37) Define the canonical momentum conjugate to qa by pa ≡ ∂L .qn . (II. φ(y)] = 0 for (x − y)2 < 0 (that is. we must have [φ(x).40) An equivalent formalism is the Hamiltonian formulation of particle mechanics. Classical Particle Mechanics In CPM. qn . ˙ ∂qa (II.. qn . S. ∂ qa ˙ (II.. To see how to achieve this. .38) Integrating the second term in Eq.. is defined by t2 S≡ t1 L(t)dt.39) Since we are only considering variations which vanish at t1 and t2 . (II. In the quantum theory they will be operator valued functions of space-time. δS = 0. z} or {r. 1. a function of the qa ’s.36) Hamilton’s Principle then determines the equations of motion: under the variation qa (t) → qa (t) + δqa (t). Eq. φa (x). Since the δqa ’s are arbitrary. θ.37) gives the Euler-Lagrange equations ∂L = pa . p1 . y. let us recall how we got quantum mechanics from classical mechanics. ˙ ∂qa ∂ qa ˙ (II.. and the dynamics are determined by the Lagrangian. (II.26 fields. δqa (t1 ) = δqa (t2 ) = 0 the action is stationary. their time derivatives qa and the time t: L(q1 ... . where T is the kinetic energy and ˙ ˙ ˙ V the potential energy. for x and y spacelike separated). Explicitly.41) . (II.. ... ˙ (II. We will restrict ourselves to systems where L has no explicit dependence on t (we will not consider time-dependent external potentials).. q2 .. pn ) = a pa qa − L. the last term vanishes. q1 . the state of a system is defined by generalized coordinates qa (t) (for example {x.

42) gives Hamilton’s equations ∂H ∂H = qa . rather than the more familiar Schr¨dinger picture.42) = where we have used the Euler-Lagrange equations and the definition of the canonical momentum. This is because we are going to work in the Heisenberg picture2 .27 Note that H is a function of the p’s and q’s. H is the energy of the system (we shall show this later on when we discuss symmetries and conservation laws. we are considering operators with no explicit time dependence in their definition). Quantum Particle Mechanics Given a classical system with generalized coordinates qa and conjugate momenta pa .44) so H is conserved. in which o the states carry the time dependence. pa (t). Varying the p’s and q’s we find ˙ dH = a dpa qa + pa dqa − ˙ ˙ dpa qa − pa dqa ˙ ˙ a ∂L ∂L dqa − dqa ˙ ∂qa ∂ qa ˙ (II.in which states are time-independent and operators carry the time dependence.) 2. not the q’s.45) (recall we have set ¯ = 1). We will discuss this in a few lectures. Note that we have included explicit time dependence in the operators qa (t) and pa (t). In fact. (II. At this point let’s drop the ˆ’s on the operators . 2 Actually. qb (t)] = [ˆa (t). ˙ = −pa . Eq.it should be obvious h by context whether we are talking about quantum operators or classical coordinates and momenta. we will later be working in the “interaction picture”. ˙ ∂pa ∂qa (II. but for free fields this is equivalent to the Heisenberg picture. with the commutation relations ˆ ˆ [ˆa (t). we obtain the quantum theory by replacing the functions qa (t) and pa (t) by operator valued functions qa (t). Varying p and q separately.43) Note that when L does not explicitly depend on time (that is. . qb (t)] = −iδab p ˆ (II. pb (t)] = 0 q ˆ p ˆ [ˆa (t). (In both cases. its time dependence arises solely from its dependence on the qa (t)’s and qa (t)’s) we have ˙ dH = dt = a a ∂H ∂H pa + ˙ qa ˙ ∂pa ∂qa qa pa − pa qa = 0 ˙ ˙ ˙ ˙ (II.

47) Thus. In the o SP.48) Since physical matrix elements must be the same in the two pictures. Notice that Eq. The time dependence of the system is carried by the states through the Schr¨dinger equation o i d | ψ(t) dt S = H| ψ(t) S =⇒ | ψ(t) S = e−iH(t−t0 ) | ψ(t0 ) S . (II. OS = OH (0)).52) . Heisenberg states are related to the Schr¨dinger states via the unitary transformation o | ψ(t) H = eiH(t−t0 ) | ψ(t) S . not the states. qa (t)]. which carry the time dependence: OH (t) = eiHT OS e−iHt = eiHT OH (0)e−iHt (II. This is simply because we never measure states directly. (II. the HP will turn out to be much more convenient than the SP. (II.50) (since at t = 0 the two descriptions coincide.51) gives dqa (t) = i[H. there are many equivalent ways to define quantum mechanics which give the same physics. H].48) we see that in the HP it is the operators. dt (II. (II. (II.46) However. One such formalism is the Heisenberg picture (HP). any formalism which differs from the SP by a transformation on both the states and the operators which leaves matrix elements invariant will leave the physics unchanged. dt (II. operators with no explicit time dependence in their definition are time independent.51) Since we are setting up an operator formalism for our quantum theory (recall that we showed in the first section that it was much more convenient to talk about creation and annihilation operators rather than wave-functions in a multi-particle theory). (II. all we measure are the matrix elements of Hermitian operators between various states. Therefore.28 You are probably used to doing quantum mechanics in the “Schr¨dinger picture” (SP). In the HP states are time independent | ψ(t) H = | ψ(t0 ) H. This is the solution of the Heisenberg equation of motion i d OH (t) = [OH (t). S ψ(t) |OS | ψ(t) S = S ψ(0) |eiHt OS e−iHt | ψ(0) S = H ψ(t) |OH (t)| ψ(t) H.49) from Eq.

Thus. the most general Lagrangian we could write down for the fields could couple fields at different coordi- . F (q. p)] = i∂F/∂pa where F is a function of the p’s and q’s. H] = i∂H/∂pa and we recover the first of Hamilton’s equations. Therefore [qa . Of course. and this turns our equation among polynomials of quantum operators into an equation among classical variables. In general. but rather a label on the field. (II. fluctuations are small. The subscript a labels the field. pn q m = p n q m. this does not mean the quantum and classical mechanics are the same thing. q. But in a general quantum state. It is like t in particle mechanics.54) Since the Lagrangian for particle mechanics can couple coordinates with different labels a. or equivalently the vector and scalar potentials) are defined at each point in space-time. it is easy to show that pa = −∂H/∂qa . for fields which aren’t scalars under Lorentz transformations (such as the electromagnetic field) it will also denote the various Lorentz components of the field. In a classical field theory. the Heisenberg equations of motion for an arbitrary operator A relate one polynomial in p. We could label them just as before. Everything we said before about classical particle mechanics will go through just as before with the obvious replacements → a d3 x a δab → δab δ (3) (x − x ). Note that x is not a generalized coordinate. in the classical limit. and expectation values of products in classical-looking states can be replaced by products of expectation values. describing its position in spacetime. dqa ∂H . but instead we’ll call our generalized coordinates φa (x). qx. such as classical electrodynamics. and so the expectation values will NOT obey the same equations as the corresponding operators. p and ˙ q to another.53) Similarly. The generalized coordinates of the system are just the components of the field at each point x. = dt ∂pa (II. observables are constructed out of the q’s and p’s. the Heisenberg picture has the nice property ˙ that the equations of motion are the same in the quantum theory and the classical theory. Classical Field Theory In this quantum theory.a where the index x is continuous and a is discrete. However. 3.29 A useful property of commutators is that [qa . observables (in this case the electric and magnetic field. We can take the expectation value of this equation to obtain an equation relating ˙ the expectation values of observables. We will be rather cavalier about going to a continuous index from a discrete index on our observables.

x).59) The analogue of the conjugate momentum pa is the time component of Πµ . However. we will usually be sloppy and follow the rest of the world in calling it the Lagrangian. since we are trying to make a causal theory. Note that both L and S are Lorentz invariant.30 nates x.58) and the integral of the total derivative in Eq. ∂µ φa (x)) (II. (II.60) where H(x) is the Hamiltonian density. we don’t want to introduce action at a distance . since we are attempting to construct a Lorentz invariant theory and the Lagrangian only depends on first derivatives with respect to time.57) vanishes since the δφa ’s vanish on the boundaries of integration.56) The function L(t.55) where the action is given by t2 S= t1 dtL(t) = d4 xL(t. (II. we will only include terms with first derivatives with respect to spatial indices. . Π0 . x) is called the “Lagrange density”.57) where we have defined Πµ ≡ a ∂L ∂(∂µ φa ) (II. Thus we derive the equations of motion for a classical field. ∂L = ∂µ Πµ . while L is not. a ∂φa (II. Furthermore. We can write a Lagrangian of this form as L(t) = a d3 xL(φa (x).the dynamics of the field should be local in space (as well as time). Once again we can vary the fields φa → φa + δφa to obtain the Euler-Lagrange equations: 0 = δS = a d4 x = a = a ∂L ∂L δφa + δ∂µ φa ∂φa ∂(∂µ φa ) ∂L d4 x − ∂µ Πµ δφa + ∂µ [Πµ δφa ] a a ∂φa ∂L − ∂µ Πµ δφa d4 x a ∂φa (II. The Hamiltonian of the system is H= a d3 x Π0 ∂0 φa − L ≡ a d3 x H(x) (II. and we will often a a abbreviate it as Πa . however.

The Klein-Gordon equation was actually first written down by Schr¨dinger.62) For the theory to be physically sensible.64) (II. H must be bounded below.68) . at the same time he wrote down o i 1 ∂ ψ(x) = − ∂t 2µ 2 ψ(x). the conjugate momenta are Πµ = ±∂ µ φ so the Hamiltonian is H = ±1 2 d3 x Π2 + ( φ)2 − bφ2 . 2 (II. and we must have b < 0. (II.61) √ The parameter a is really irrelevant here. So let’s take instead L = ± 1 ∂µ φ∂ µ φ + bφ2 . we can easily get rid of it by rescaling our fields φ → φ/ a. 2 What does this describe? Well.31 Now let’s construct a simple Lorentz invariant Lagrangian with a single scalar field. (II. Defining b = −µ2 .67) This looks promising.66) may be made arbitrarily large.66) Each term in H is positive definite: the first corresponds to the energy required for the field to change in time. we have the Lagrangian (density) L= and corresponding Hamiltonian H= 1 2 1 2 ∂µ φ∂ µ φ − µ2 φ2 (II. (II. In fact. The simplest thing we can write down that is quadratic in φ and ∂µ φ is L = 1 a ∂µ φ∂ µ φ + bφ2 . and Eq. there must be a state of lowest energy. (II.65) d3 x Π2 + ( φ)2 + µ2 φ2 . this equation is called the Klein-Gordon equation.65) is the Klein-Gordon Lagrangian. (II.63) (II. The equation of motion for this theory is ∂µ ∂ µ + µ2 φ(x) = 0. (II. the overall sign of H must be +. the second to the energy corresponding to spatial variations. Since there are field configurations for which each of the terms in Eq. and the last to the energy required just to have the field around in the first place.

φ(x) is not a wavefunction. t)] = iδab δ (3) (x − y). The energy is unbounded below and the theory has no ground state. φa (x)].67). Then we’ll try and figure out what we’ve created. dt dt (II. This should not be such a surprise. t). t) and Πa (y. φb (y. Quantum Field Theory To quantize our classical field theory we do exactly what we did to quantize CPM. with little more than a change of notation. and we just showed that the Hamiltonian is positive definite. t)] = 0 [φa (x. t). p = k.71) . ∂µ ∂ µ + µ2 ψ(x) = 0. satisfying dΠa (x) dφa (x) = i[H. though.) The Hamiltonian will still be positive definite. and the negative energy solutions correspond to the annihilation of a particle of the same mass by the field operator. t)] = [Π0 (x. 4. it is a Hermitian operator. t).72) (II. Πa (x)].70) ∂2 + ∂t2 2 − µ2 ψ = 0 (II. since we already know that single particle relativistic quantum mechanics is inconsistent. E = ± p2 + µ2 . we know E = ω. so from E 2 = p2 + µ2 he also got o − or.32 In quantum mechanics for a wave ei(k·x−ωt) . So let’s quantize our classical field theory and construct the quantum field. so this equation is just E = p2 /2µ. φa (x. (II. Schr¨dinger knew about relativity. (It took eight years after the discovery of quantum mechanics before the negative energy solutions of the Klein-Gordon equation were correctly interpreted by Pauli and Weisskopf. In Eq. Replace φ(x) and Πµ (x) by operator-valued functions satisfying the commutation relations [φa (x. Of course. in our notation. Π0 (y. Π0 (y. t) are Heisenberg operators.67) correspond to the creation of a particle of mass µ by the field operator. (II.69) Unfortunately. It is a classical field. (II. this is a disaster if we want to interpret ψ(x) as a wavefunction as in the Schr¨dinger o Equation: this equation has both positive and negative energy solutions. Soon it will be a quantum field which is also not a wavefunction. It will turn out that the positive energy solutions to Eq. = i[H. b As before.

φ(y. (II. φ(x. 0) eik·x .66) that the operators satisfy ˙ ˙ φa (x) = Π(x). (II. (Since φ is a solution to the KG equation this is completely general.71). (2π)3 2ωk (II.78) † Using the equal time commutation relations Eq. we can calculate [αk . αk ]: † [αk .75) Recalling that the Fourier transform of e−ik·x is a delta function: d3 x −i(k−k )·x e = δ (3) (k − k ) (2π)3 we get d3 x † φ(x. 0)e−ik·x = αk + α−k (2π)3 d3 x ˙ † φ(x. 3 (2π) ωk (II. (II. 0)e−ik·x = (−iωk )(αk − α−k ) (2π)3 and so αk = † αk = 1 2 1 2 (II. 0)] e−ik·x+ik ·y 6 4 (2π) ωk ωk 1 d3 xd3 y i i [iδ (3) (x − y)] + [iδ (3) (x − y)] e−ik·x+ik ·y = − 6 4 (2π) ωk ωk 1 d3 x 1 1 = + e−i(k−k )·x 4 (2π)6 ωk ωk 1 = δ (3) (k − k ). We can therefore write φ(x) as φ(x) = † d3 k αk e−ik·x + αk eik·x (II. it must be Hermitian. 0) e−ik·x 3 (2π) ωk d3 x i φ(x.) The plane wave solutions to Eq.77) d3 x i φ(x.74) † where the αk ’s and αk ’s are operators. We can solve for αk and αk . Since φ(x) is going to be an observable. 0) − ∂0 φ(x. † † which is why we have to have the αk term. 0)] + [φ(y. Π(x) = 2 φ − µ2 φ (II. (II. 0). First of all. Let’s try and get some feeling for φ(x) by expanding it in a plane wave basis. αk ] = − 1 d3 xd3 y i i ˙ ˙ [φ(x. 0) = ∂0 φ(x. φ(x. 0).79) . 0) = † d3 k αk eik·x + αk e−ik·x † d3 k(−iωk ) αk eik·x − αk e−ik·x .33 For the Klein-Gordon field it is easy to show using the explicit form of the Hamiltonian Eq. 0) + ∂0 φ(x.67) are exponentials eik·x where k 2 = µ2 .76) (II.73) and so the quantum fields also obey the Klein-Gordon equation.

then [ak . k (II. we obtain H= d3 k 2 ak a−k e−2iωk t (−ωk + k 2 + µ2 ) 2ωk 2 + a† ak (ωk + k 2 + µ2 ) k 1 2 2 + ak a† (ωk + k 2 + µ2 ) k 2 + a† a† e2iωk t (−ωk + k 2 + µ2 ) . So the quantum field φ(x) is a sum over all momenta of creation and annihilation operators: φ(x) = d3 k √ 3/2 ak e−ik·x + a† eik·x .84) This is almost. k −k 2 Since ωk = k 2 + µ2 .85) Commuting the ak and a† in Eq.66)). ak ] = −ωk ak k k (II. we can substitute the expression for the fields in terms of a† and ak and the comk mutation relation Eq. they had k better have the right commutation relations with the Hamiltonian [H. the time-dependent terms drop out and we get (II. (II. If we define ak ≡ αk /(2π)3/2 2ωk .80) to obtain an expression for the Hamiltonian in terms of the a† ’s and k ak ’s.84) we get k H= 1 d3 k ωk a† ak + 2 δ (3) (0) . Then H= 1 2 k ωk ak a† + a† ak = k k k a† ak + k 1 2 (II.81) (2π) 2ωk Actually. a† ] = δ (3) (k − k ). [H. k (II. From the explicit form of the Hamiltonian (Eq.86) δ (3) (0)? That doesn’t look right. H= d3 k ωk a† ak .34 √ This is starting to look familiar.82) so that they really do create and annihilate mesons. After some algebra (do it!). (II. k (II. k k (II. if we are to interpret ak and a† as our old annihilation and creation operators.80) These are just the commutation relations for creation and annihilation operators. but not quite. a† ] = ωk a† . k (II. what we had before.87) . Let’s go back to our box normalization for a moment. (II.83) H= 1 2 d3 k ωk ak a† + a† ak .

k (II.35 so the δ (3) (0) is just the infinite sum of the zero point energies of all the modes. In general relativity the curvature couples to the absolute energy. φ2 (x2 ).φn (xn ): (II. which is that if you ask a silly question in quantum field theory. It’s just an overall energy shift. and these are finite. So instead of H. But there is a lesson to be learned here. In fact. φn (xn ). So by a judicious choice of ordering. Asking about absolute energies is a silly question3 .. not zero. and since there are an infinite number of modes we got an infinite 2 energy in the ground state. We can do this by noticing that the zero point energy of the SHO is really the result of an ordering ambiguity..the energy density is at least 56 orders of magnitude smaller than dimensional analysis would suggest). However.89) instead of the usual ω(a† a+1/2). . We won’t worry about gravity in this course. we should be able to eliminate the (unphysical) infinite zero-point energy.90) as the usual product. For a set of free fields φ1 (x1 ). define the normal-ordered product :φ1 (x1 ). this becomes HSHO = ω a† a (II. and so it is a physical quantity. we can use :H : and the infinite energy of the ground state goes away: :H:= d3 k ωk a† ak . Only energy differences have any physical meaning.. (II. this uniquely specifies the ordering. for reasons nobody understands. if you ask an unphysical question (and it may not 3 except if you want to worry about gravity.. In general in quantum field theory. you will get a silly answer.88) But when p and When p and q are numbers.91) That was easy. and it doesn’t matter where we define our zero of energy. The energy of each mode starts at 1 ωk . as do annihilation operators. but with all the creation operators on the left and all the annihilation operators on the right. This is no big deal. 2 ω 2 2 2 (p + q ). Since creation operators commute with one another. For example.. . when quantizing the simple harmonic oscillator we could have just as well written down the classical Hamiltonian HSHO = ω (q − ip)(q + ip). this is the same as the usual Hamiltonian q are operators. the observed absolute energy of the universe appears to be almost precisely zero (the famous cosmological constant problem . let’s use this opportunity to banish it forever. since the infinity gets in the way.

kaons. we find p |φ(x. The classical theory of a scalar field that we wrote down has nothing to do with particles.92) Thus. and from the energy-momentum relation we saw that the parameter µ in the Lagrangian corresponded to the mass of the particle. 0)| 0 = d3 k 1 −ik·x e |k . At this point it’s worth stepping back and thinking about what we have done. However. the quanta of the electromagnetic (vector) field are photons. p |x = e−ip·x (II. quantizing the classical field theory immediately forced upon us a particle interpretation of the field: these are generally referred to as the quanta of the field. however. For the scalar field.36 be at all obvious that it’s unphysical) you will get infinity for your answer. it pops out a linear combination of momentum eigenstates. while fermions like the electron are the quanta of the corresponding fermi field. (2π)3 2ωk (II. or the Higgs boson of the Standard Model). Hence. creates a particle . these commutation relations also ensured that the Hamiltonian had a discrete particle spectrum. acting on the vacuum. the field operator φ may still seem a bit abstract . when the field operator acts on the vacuum.81).93) = e−ip·x .) Taking the inner product of this state with a momentum eigenstate | p . these are spinless bosons (such as pions. The canonical commutation relations we imposed on the fields ensured that the Heisenberg equation of motion for the operators in the quantum theory reproduced the classical equations of motion. 0)| 0 . we have φ(x. At this stage. thus building the correspondence principle into the theory. From the field expansion Eq. there is not such a simple correspondence to a classical field: the Pauli exclusion principle means that you can’t make a coherent state of fermions. let us consider the interpretation of the state φ(x. (Think of the field operator as a hammer which hits the vacuum and shakes quanta out of it. To get a better feeling for it. so there is no classical equivalent of an electron field. (II.an operator-valued function of space-time from which observables are built. Taming these infinities is a major headache in QFT.94) we see that we can interpret φ(x. 0) as an operator which. it simply had as solutions to its equations of motion travelling waves satisfying the energy-momentum relation of a particle of mass µ. As we will see later on. In this latter case. Recalling the nonrelativistic relation between momentum and position eigenstates. 0)| 0 = d3 k 1 −ik·x e p |k (2π)3 2ωk (II.

but the convention was established by Heisenberg and Pauli. For convenience. Suppose we prepare a particle at some spacetime point y. What is the amplitude to find it at point x? From Eq.4 So let’s check this explicitly. C.97) (II. . t). because the equal time commutation relations [φ(x. t)] = iδ (3) (x − y) (II. thus. the amplitude to find it at x is given by the expectation value 0 |φ(x)φ(y)| 0 . let’s revisit the question we asked at the end of Section 1 about the amplitude for a particle to propagate outside its forward light cone.93).37 at position x.96) (II. it’s not obvious that the quantum theory is Lorentz invariant. Causality Since the Lagrangian for our theory is Lorentz invariant and all interactions are local. However.98) (the ± convention is opposite to what you might expect. we expect there should be no problems with causality in our theory. we create a particle at y by hitting it with φ(y). we first split the field into a creation and an annihilation piece: φ(x) = φ+ (x) + φ− (x) where φ+ (x) = φ− (x) = d3 k √ ak e−ik·x . so who are we to argue?). when it acts on an n particle state it has an amplitude to produce both an n + 1 and an n − 1 particle state. Π0 (y. Then we have 0 |φ(x)φ(y)| 0 = 0 |(φ+ (x) + φ− (x))(φ+ (y) + φ− (y))| 0 = 0 |φ+ (x)φ− (y)| 0 4 In the path integral formulation of quantum field theory. (2π)3/2 2ωk d3 k √ a† eik·x (2π)3/2 2ωk k (II. Since it contains both creation and annihilation operators. which you will study next semester.95) treat time and space on different footings. (II. Lorentz invariance of the quantum theory is manifest. First of all.

the function D(x − y) does not vanish for spacelike separated points. (II. .104) to boost x − y to equal times. φ− (y)] + [φ− (x). (II. From Eq. φ+ (y)] = D(x − y) − D(y − x).104) where Λ is any connected Lorentz transformation.5 5 Another way to see this is to note that a spacelike vector can always be turned into minus itself via a connected Lorentz transformation. t). we have D(x − y) ∼ e−µ|x−y| . O2 (x2 )] = 0 for (x1 − x2 )2 < 0.100) and hence. we can always use the property (II. spacelike separated observables must commute: [O1 (x1 ). Because d3 k/ωk is a Lorentz invariant measure. spacelike separated measurements can’t interfere and the theory preserves causality. Hence. (II. φ(y)] = 0 for all spacelike separated fields. How can we reconcile this with the result that space-like measurements commute? Recall from Eq. In fact. D(Λx) = D(x) (II. hence. this vanishes for spacelike separations. (I. t)] = 0 for any x and y.99).38 = 0 |[φ+ (x). a† ]| 0 e−ik·x+ik ·y = k 3 2√ω ω (2π) k k = d3 k e−ik·(x−y) (2π)3 2ωk ≡ D(x − y). by the equal time commutation relations. D(x − y) is manifestly Lorentz invariant . φ− (y)]| 0 d3 k d3 k 0 |[ak . D(x − y) = D(y − x) and the commutator of the fields vanishes. φ(y. This means that for spacelike separations.102) Since all observables are constructed out of fields. so we must have [φ(x).103) It is easy to see that.47) we studied in Section 1: d D(x − y) = dt d3 k −ik·(x−y) e . (I.99) Unfortunately. (2π)3 (II. (II. There is always a reference frame in which spacelike separated events occur at equal times. for two spacelike separated events at equal times. it is related to the integral in Eq. if they do.101) outside the lightcone.1) that in order for our theory to be causal. we have [φ(x). we know that [φ(x. unlike D(x − y). we just need to show that fields commute at spacetime separations. Now. (II. φ(y)] = [φ+ (x).

At higher order more complicated processes can occur. Writing φ(x) in terms of a† ’s and ak ’s. each normal mode evolves independently of the others. there will be a piece which looks like a† 1 a† 2 ak3 ak4 . A more general theory describing real particles must have additional terms in the Lagrangian which describe interactions. There is no scattering. which carries no charge and is its own antiparticle. we are going to derive some more exact results from field theory which will prove useful. because in Eq. and in fact. At second order in perturbation theory we can get 2 → 4 scattering. D. occurring with an amplitude proportional to λ2 . We will study charged fields shortly. containing two annihilation and two creation operators. But before we set up perturbation theory and scattering theory. and the amplitude for the scattering process will be proportional to λ. consider adding the following potential energy term to the Lagrangian: L = L0 − λφ(x)4 (II. The field now has self-interactions. This k k will contribute to 2 → 2 scattering when acting on an incoming 2 meson state. so the dynamics are nontrivial. no way to measure anything.65) is a free field theory: it describes particles which simply propagate with no interactions. which means that in the quantum theory particles don’t interact. The two amplitudes cancel for spacelike separations! Note that this is for the particular case of a real scalar field. This is where we are aiming. (II. and in that case the two amplitudes which cancel are the amplitude for the particle to travel from x to y and the amplitude for the antiparticle to travel from y to x. Classically. For example. .39 This puts into equations what we said at the end of Section 1: causality is preserved. For example. We will have much more to say about the function D(x) and the amplitude for particles to propagate in Section 4.103) the two terms represent the amplitude for a particle to propagate from x to y minus the amplitude of the particle to propagate from y to x. Interactions The Klein-Gordon Lagrangian Eq. or pair production. (II.105) where L0 is the free Klein-Gordon Lagrangian. To see how such a potential affects the dynamics of the field quanta. when we study interactions. we see that the interaction term k has pieces with n creation operators and 4 − n annihilation operators. consider the potential as a small perturbation (so that we can still expand the fields in terms of solutions to the free-field Hamiltonian).

” Given some general transfor- . As a simple example. A symmetry (L(qi + α. SYMMETRIES AND CONSERVATION LAWS The dynamics of interacting field theories. Since in quantum field theory we won’t be able to solve anything exactly. qi )) has resulted in a conservation law.2) (III. symmetry arguments will be extremely important. are extremely complex. there is a corresponding conserved quantity. It is useful because it allows you to make exact statements about the solutions of a theory without solving it explicitly. P ≡ p1 + p2 = − ˙ ˙ + . In this chapter we will look at this question in detail and develop some techniques which will allow us to extract dynamical information from the symmetries of a theory. q2 ). and from the Euler-Lagrange equations ˙ pi = − ˙ ∂V ∂V ∂V ˙ . L depends on t only through the coordinates qi and their derivatives). The resulting equations of motion are not analytically soluble. ˙ ˙ We also saw earlier that when ∂L/∂t = 0 (that is. H (the energy) is therefore a conserved quantity when the system is invariant under time translation. ˙2 1 ˙2 2 The momenta conjugate to the qi ’s are pi = mi qi . A. then dH/dt = 0. (??). we first need to define “symmetry. the particles aren’t attached to springs or anything else which defines a fixed reference frame) then the system is invariant under the shift qi → qi + α. in more complicated interacting theories it is often possible to discover many important features about the solution simply by examining the symmetries of the theory. such as φ4 theory in Eq.40 III. ˙ and ∂V /∂q1 = −∂V /∂q2 . To prove Noether’s theorem. This is a very general result which goes under the name of Noether’s theorem: for every symmetry. where the Lagrangian is L = T − V .1) If V depends only on q1 − q2 (that is. ∂qi ∂q1 ∂q2 (III. The total momentum of the system is conserved. so P = 0. Nevertheless. consider two particles in one dimension in a potential L = 1 m1 q1 + 2 m2 q2 − V (q1 . qi ) = L(qi . Classical Mechanics Let’s return to classical mechanics for a moment. free field theory (with the optional addition of a source term. as we will discuss) is the only field theory in four dimensions which has an analytic solution. In fact.

t). t2 ) − F (qa (t1 ). Actually. a transformation is a symmetry iff DL = dF/dt for some function F (qa . Why is this a good definition? Consider the variation of the action S: ˙ t2 t2 DS = t1 dtDL = t1 dt dF = F (qa (t2 ). Dr = e.4) so DL = dL/dt. DL = 0. pi = mi qi and Dqi = 1. 1. DL = a ∂L ∂L Dqa + D qa ˙ ∂qa ∂ qa ˙ pa Dqa + pa Dqa ˙ ˙ = a = d dt pa Dqa a (III. Space translation: qi → qi + α. even if the system looks nothing like particle mechanics. for example. qa (t) → qa (t + λ) = qa (t) + λdqa /dt + O(λ2 ). First of all. L(t. DL = dF/dt. ˙ ˙ dt (III. For time e ˆ ˆ translation..41 mation qa (t) → qa (t. ˙ But by the definition of a symmetry. Dqa = dqa /dt. (III. Therefore the additional term doesn’t contribute to δS and therefore doesn’t affect the equations of motion. for the transformation r → r + λˆ (translation in the e direction). You might imagine that a symmetry is defined to be a transformation which leaves the Lagrangian invariant. dt (III. Then DL = 0.. Let’s apply this to our two previous examples. doesn’t satisfy this requirement: if L has no explicit t dependence. λ).5) Recall that when we derived the equations of motion. define Dqa ≡ ∂qa ∂λ (III. qa (t2 ). δqa (t1 ) = δqa (t2 ) = 0. where qa (t. So more generally. qa (t1 ). Time translation. qa . so p1 + p2 = ˙ m1 q1 + m2 q2 is conserved.3) λ=0 For example. qa (t + λ)) = L(0) + λ ˙ dL + . So d dt So the quantity a pa Dqa pa Dqa − F a = 0. this is too restrictive. t1 ).6) where we have used the equations of motion and the equality of mixed partials (Dqa = d(Dqa )/dt).7) − F is conserved. We will call any conserved quantity associated with spatial ˙ ˙ translation invariance momentum. It is now easy to prove Noether’s theorem by calculating DL in two ways. λ) = L(qa (t + λ). . 0) = qa (t). we didn’t vary the qa ’s and qa ’s at the ˙ endpoints.

because not only are conserved quantities globally conserved. we find that the total charge Q is conserved. we have the stronger statement of current conservation. (III. in field theory conservation laws will be of the form ∂µ J µ = 0 for some four-current J µ. B.10) λ=0 ˙ A transformation is a symmetry iff DL = ∂µ F µ for some F µ (φa .9) where S is the surface of V . Eq. φa . (III. and defining QV = V d3 xρ(x).8). a transformation of this form doesn’t affect the equations . the same arguments must go through as well. Then Dqa = dqa /dt. λ). Integrating over some volume V . This works for classical particle mechanics. As before. I will leave it to you to show that. but not locally. However. we will call the conserved quantity associated with time translation invariance the energy of the system. In fact. they must be locally conserved as well. φa (x. just as in particle mechanics. This conserves charge globally. Again.8) This just expresses current conservation. stronger statements may be made in field theory. Recall from electromagnetism that the charge density satisfies ∂ρ + ∂t ·  = 0.42 2. Time translation: t → t + λ. we have dQV =− dt ·=− v S dS ·  (III. This is the Hamiltonian. justifying our previous assertion that the Hamiltonian is the energy of the system. Symmetries in Field Theory Since field theory is just the continuum limit of classical particle mechanics. Therefore. Since the canonical commutation relations are set up to reproduce the E-L equations of motion for the operators. 0) = φa (x). For example. Taking the surface to infinity. we consider the transformations φa (x) → φa (x. DL = dL/dt. in a theory which conserves electric charge we can’t have two separated opposite charges simultaneously wink out of existence. and define Dφa = ∂φa ∂λ . This means that the rate of change of charge inside some region is given by the flux through the surface. x). (III. it will work for quantum particle mechanics as well. F = L and so the conserved quantity is ˙ a (pa qa ) − L.

We now have DL = a ∂L Dφa + Πµ D(∂µ φa ) a ∂φa ∂µ Πµ Dφa + Πµ ∂µ Dφa a a = a = ∂µ a (Πµ Dφa ) = ∂µ F µ a (III. we have DL = ∂µ (eµ L).. The conserved current is therefore Jµ = a Πµ Dφ − F a Πµ eν ∂ ν φa − eµ L a a = = eν Πµ ∂ ν φa − g µν L a a ≡ eν T µν (III.43 of motion. Under a shift x → x + λe. so that no charge can flow out through the boundaries.12) satisfy ∂µ J µ = 0.13) 1. this gives the global conservation law dQ d ≡ dt dt d3 xJ 0 (x) = 0. we have φa (x) → φa (x + λe) = φa (x) + λeµ ∂ µ φa (x) + . so F = eµ L. where e is some fixed four-vector. so Dφa (x) = eµ ∂ µ φa (x). (III. (III.. Space-Time Translations and the Energy-Momentum Tensor We can use the techniques from the previous section to calculate the conserved current and charge in field theory corresponding to a space or time translation.11) so the four components of Jµ = a Πµ Dφa − F µ a (III.15) (III.16) . If we integrate over all space.14) Similarly. since L contains no explicit dependence on x but only depends on it through the fields φa .

a straightforward substitution of the expansion of the fields in terms of creation and annihilation operators into the expression for expression we obtained earlier for the momentum operator. it corresponds to the conserved quantity associated with time translation invariance. (III.18) where H is the Hamiltonian density we had before. So the Hamiltonian. (III. T µ0 is therefore the “energy current”. For the Klein-Gordon field. : P := d3 k k a† ak k (III. It is important not to confuse these two uses of the term “momentum.) Similarly. the conserved charge associated with space translation. it has the simple transformation law φ(x) → φ(Λ−1 x).44 where T µν = µ ν µν a Πa ∂ φa −g L is called the energy-momentum tensor. has nothing to do with the conjugate momentum Πa of the field φa .20) as discussed in the first section. eν = (1. really is the energy of the system (that is. x) then we will find the conserved charge to be the x-component ˆ of momentum. if we choose eµ = (0. we also have ∂µ T µν = 0.” 2. Lorentz Transformations Under a Lorentz transformation xµ → Λ µ ν xν a four-vector transforms as aµ → Λµ ν aν (III. Since ∂µ J µ = 0 = ∂µ T µν eν for arbitrary e.22) . as we had claimed.21) (III. Since a scalar field by definition does not transform under Lorentz transformations.19) d3 xT 01 gives the where again we have normal-ordered the expression to remove spurious infinities. 0). Note that the physical momentum P . and the corresponding conserved quantity is Q= d3 xJ 0 = d3 xT 00 = d3 x a Π0 ∂0 φa − L = a d3 xH (III.17) For time translation.

26) where in the third line we have relabelled the dummy indices. This will define a family of Lorentz transformations Λ(λ)µ ν .45 This simply states that the field itself does not transform at all. This could be rotations about a specified axis by an angle λ. From the fact that aµ bµ is Lorentz invariant. a vector field (spin 1) Aµ transforms as Aµ (x) → Λµ ν Aν (Λ−1 x).25) is antisymmetric. we will restrict ourselves to scalar fields at this stage in the course. As usual. To use the machinery of the previous section. This is good because there are six independent Lorentz transformations . Let us define µ ν (III. since the various components of the fields rotate into one another under Lorentz transformations. Since this holds for arbitrary four vectors aµ and bν . Fields with spin have more complicated transformation laws. (III. the value of the field at the coordinate x in the new frame is the same as the field at that same point in the old frame. which means there are 4(4 − 1)/2 = 6 independent components of .27) The indices µ and ν range from 0 to 3. let us consider a one parameter subgroup of Lorentz transformations parameterized by λ. . we must have µν =− νµ .24) Then under a Lorentz transformation aµ → Λµ ν aν . from which we wish to get Dφ = ∂φ/∂λ|λ=0 .23) ≡ DΛµ ν . we have Daµ = It is straightforward to show that we have 0 = D(aµ bµ ) = (Daµ )bµ + aµ (Dbµ ) = = = ( µ ν ν a bµ µν µ νa ν . (III. + aµ + µ ν bν ν µ µν a b µν νµ a ν µ b + νµ ) a ν µ b (III. For example. or boosts in some specified direction with γ = λ. (III.three rotations (one about each axis) and three boosts (one in each direction).

taking 01 → 10 cos λ − sin λ sin λ cos λ a1 a2 . it depends on x only through its dependence on the field and its derivatives.29) =− = +1 and all other components zero. (III. (III.30) =− 2 10 a Note that the signs are different because lowering a 0 index doesn’t bring in a factor of −1.32) ∂ φ(x). This is just the infinitesimal version of a0 a1 → cosh λ sinh λ sinh λ cosh λ a0 a1 . Take the other components zero. we find Dφ(x) = ∂ φ(Λ−1 (λ)µ ν xν ) λ=0 ∂λ ∂ = ∂α φ(x) (Λ−1 (λ)(x)α ∂λ = ∂α φ(x) D Λ−1 (λ)α β xβ = ∂α φ(x) − = − αβ x β α α β λ=0 xβ (III.46 Let’s take a moment and do a couple of examples to demystify this. we get 0 1 1a 1 1 0a Da0 = Da1 = = 1 01 a = +a1 = +a0 .33) Πµ αβ αβ x α β ∂ φ− αβ x α µβ g L (III.31) which corresponds to a boost along the x axis. (III.34) = Πµ xα ∂ β φ − xα g µβ L . (III. Using the chain rule. Then we have Da1 = Da2 = 1 2 2a 2 1 1a 12 =− 21 and all =− =− 2 12 a 2 21 a = −a2 = +a1 . . Since L is a scalar. Therefore we have DL = αβ x α β ∂ L β µα = ∂µ and so the conserved current J µ is Jµ = a αβ x g L (III. a1 a2 On the other hand. Now we’re set to construct the six conserved currents corresponding to the six different Lorentz transformations.28) This just corresponds to the a rotation about the z axis.

is J 12 = d3 x x1 T 02 − x2 T 01 . Particles with spin carry intrinsic angular momentum which is not included in this expression . the energy momentum tensor is T 0i (x. the part of the quantity in the parentheses that is antisymmetric in α and β must be conserved. The six conserved charges are given by the six independent components of J αβ = d3 x M 0αβ = d3 x xα T 0β − xβ T 0α . they make up the conserved quantities you learned about in . So for example J 12 . t) = pi δ (3) (x − r(t)) which gives J 12 = x1 p2 − x2 p1 = (r × p)3 (III. That takes care of three of the invariants corresponding to Lorentz transformations.38) This is the field theoretic analogue of angular momentum.35) where T µν is the energy-momentum tensor defined in Eq. Together with energy and linear momentum. the conserved quantity coming from invariance under rotations about the 3 axis.39) which is the familiar expression for the third component of the angular momentum. ∂µ M µαβ = 0 where M µαβ = Πµ xα ∂ β φ − xα g µβ L − α ↔ β = xα Πµ ∂ β φ − g µβ L − α ↔ β = xα T µβ − xβ T µα (III. Note that this is only for scalar particles. we will call the conserved quantity corresponding to rotations the angular momentum. In this case. (III.this is only the orbital angular momentum.47 Since the current must be conserved for all six antisymmetric matrices αβ .37) Just as we called the conserved quantity corresponding to space translation the momentum.40) (III.36) (III. Particles with spin are described by fields with tensorial character.16). (III. That is. which is reflected by additional terms in the J ij . (III. We can see that this definition matches our previous definition of angular momentum in the case of a point particle with position r(t).

since they require L to have the appropriate symmetry and so tend to greatly restrict the form of L. But there’s nothing in principle wrong with this. = t pi + pi − dt dt (III. there are a number of other quantities which are experimentally known to be conserved. baryon number and lepton number which are not automatically conserved in any field theory. The three conserved quantities partners of the angular momentum. (III. What about boosts? There must be three more conserved quantities. we see that in field theory the relation between the T 0i ’s and the first moment of T 00 is the result of Lorentz invariance. Experimental observation of these conservation laws in nature is crucial in helping us to figure out the Lagrangian of the real world. We could write down an expression for the energy-momentum tensor T µν without knowing the explicit form of L. Internal Symmetries Energy.” Although you are not used to seeing this presented as a separate conservation law from conservation of momentum.41) This has an explicit reference to x0 . the time. The x0 may be pulled out of the spatial integral. Therefore we get pi = d dt d3 x xi T 00 = constant. The centre of mass is replaced by the “centre of energy. is a constant. . d3 x xi T 00 (x) are the Lorentz C. and the second term. By Noether’s theorem. and the conservation law gives 0 = d 0i d t d3 x T 0i − d3 x xi T 00 J = dt dt d d d3 x T 0i + d3 x T 0i − d3 x xi T 00 = t dt dt d d d3 x xi T 00 . We will call these transformations which don’t correspond to space-time transformations internal symmetries. these must also be related to continuous symmetries. What are they? Consider J 0i = d3 x x0 T 0i − xi T 00 . which is something we haven’t seen before in a conservation law.48 first year physics. such as electric charge. However.43) This is just the field theoretic and relativistic generalization of the statement that the centre of mass moves with a constant velocity.42) The first term is zero by momentum conservation. momentum and angular momentum conservation are clearly properties of any Lorentz invariant field theory. pi . (III.

= This Lagrangian is invariant under the transformation φ1 → φ1 cos λ + φ2 sin λ φ2 → −φ1 sin λ + φ2 cos λ. (III.48) This isn’t very illuminating at this stage. and just as r2 = x2 + y 2 is invariant under real rotations. U (1) Invariance and Antiparticles Here is a theory with an internal symmetry: 2 2 L= 1 2 a=1 ∂µ φa ∂ φa − µ φa φa − g a µ 2 (φa ) 2 . (III. So the conserved current is J µ = Πµ Dφ1 + Πµ Dφ2 = (∂ µ φ1 )φ2 − (∂ µ φ2 )φ1 1 2 and the conserved charge is Q= d3 xJ 0 = ˙ ˙ d3 x(φ1 φ2 − φ2 φ1 ). We can write this in matrix form: φ1 φ2 cos λ = − sin λ sin λ cos λ φ1 φ2 . this symmetry isn’t terribly interesting. We say that L has an SO(2) symmetry. 1 2 these are invariant under the transformation (III. (III.46) In the language of group theory. with a common mass µ and a potential g g (φ1 )2 + (φ2 2 )2 .49) (III.47) Since F µ is a constant.44) 2 2 a (φa ) It is a theory of two scalar fields. Once again we can calculate the conserved charge: Dφ1 = φ2 Dφ2 = −φ1 DL = 0 → F µ = constant. (III. (III. this is known as an SO(2) transformation. But in the quantized theory it has a very nice interpretation in terms . we can just forget about it (if J µ is a conserved current.45). meaning that the transformation matrix has unit determinant. At the level of classical field theory. φ1 and φ2 . so is J µ plus any constant).45) This is just a rotation of φ1 into φ2 in field space. It leaves L invariant (try it) because L depends only on φ2 + φ2 and (∂µ φ1 )2 + (∂µ φ2 )2 .49 1. the O for “orthogonal” and the 2 because it’s a 2 × 2 matrix. The S stands for “special”.

it is easy to show that Q = i = d3 k (a† ak2 − a† ak1 ) k1 k2 d3 k (b† bk − c† ck ) k k (III. At this stage. c† | 0 = | k.50 of particles and antiparticles. 1 b† | 0 = √ (| k. They create linear combinations of states with type 1 and type 2 mesons. where i = 1. b† ≡ k1 √ k2 k 2 2 † a + ia† ak1 − iak2 √ ≡ . k 2 2 (III.50) into Eq. c . c† ≡ k1 √ k2 . (III. Q=i d3 k(a† ak2 − a† ak1 ). 1 .56) (III.53) It is easy to verify that the bk ’s and ck ’s also have the right commutation relations to be creation and annihilation operators. except for the fact that the terms are off-diagonal. let’s also forget about the potential term in Eq.49) gives.51) a† | 0 = | k.52) We are almost there. k2 (III. We will denote the corresponding creation and annihilation operators by a† and aki .44). k1 k2 (III. 1 − i| k. 2 ).55) = Nb − Nc . b . k 2 (III. Then we have a theory of two free fields and we can expand the fields in terms of creation and annihilation operators. which we denote ki by a† | 0 = | k. k1 Substituting the expansion φi = d3 k √ aki e−ik·x + a† eik·x ki 3/2 2ω (2π) k (III. 2. We can call them particles of type b and type c b† | 0 = | k. So let’s consider quantizing the theory by imposing the usual equal time commutation relations. They create and destroy two different types of meson. 2 . This looks like the expression for the number operator. k k In terms of our new operators. Let’s fix that by defining new creation and annihilation operators which are a linear combination of the old ones: bk ≡ ck a† − ia† ak1 + iak2 √ . after some algebra. (III. so let’s work with these as our basis states.54) Linear combinations of states are perfectly good states.

2 In terms of ψ and ψ † . ψ † ] = ψ † . [Q. We say that c and b are one another’s antiparticle: they are the same in all respects except that they carry the opposite conserved charge.58) in front). In terms of creation and annihilation operators. (III.51 where Ni = dk a† aki is the number operator for a field of type i. (III. so we clearly have b-type particles with charge +1 and c-type particles with charge −1. Now. Note that we couldn’t have a theory with b particles and not c particles: they both came out of the Lagrangian Eq. We can also see this from the commutator [Q.61) . L is L = ∂ µ ψ † ∂ µ ψ − µ2 ψ † ψ (note that there is no factor of ψ † have the expansions ψ= ψ† = d3 k √ 3/2 bk e−ik·x + c† eik·x k ck e−ik·x + b† eik·x k (III. (III.56) it is easy to show that [Q. ψ] = −ψ. ψ]: from the expression for the conserved charge Eq. whereas ψ † creates b-type particles and annihilates c’s.59) 1 2 (III. The total charge is therefore ki the number of b’s minus the number of c’s. then Q(ψ| q ) = [Q. | q is an eigenstate of the charge operator Q with eigenvalue q). ψ]| q + ψQ| q = (−1 + q)ψ| q (III. Thus ψ always changes the Q of a state by −1 (by creating a c or annihilating a b in the state) whereas ψ † acting on a state increases the charge by one. The existence of antiparticles for all particles carrying a conserved charge is a generic prediction of QFT. With the benefit of hindsight we can go back to our original Lagrangian and write it in terms of the complex fields 1 ψ ≡ √ (φ1 + iφ2 ) 2 1 ψ † ≡ √ (φ1 − iφ2 ). that was all a bit involved since we had to rotate bases in midstream to interpret the conserved charge.44).57) (III. ψ and (2π) 2ωk 2ωk (2π) d3 k √ 3/2 so ψ creates c-type particles and annihilates their antiparticle b.60) If we have a state | q with charge q (that is.

Π0 (y. not operators. The transformation Eq.66) (III. and is somewhat simpler to work with. we vary them independently and assign a conjugate momentum to each: Πµ = ψ Therefore we have Πµ = ∂ µ ψ ∗ .52 so ψ| q has charge q − 1.45) may be written as ψ = e−iλ ψ. t). we clearly recover the equations of motion for φ1 and φ2 .62) This is called a U (1) transformation or a phase transformation (the “U” stands for “unitary”. as we asserted. and two real degrees of freedom in ψ.. Still. That is.63) (these are classical fields. not ψ † . which may be independently . t)] = iδ (3) (x − y). In fact. Clearly ψ and ψ ∗ are not independent. this rule of thumb works because there are two real degrees of freedom in φ1 and φ2 . Adding and subtracting these equations.) We can quantize the theory correctly and obtain the equations of motion if we follow the same rules as before.64) Similarly.67) We will leave it as an exercise to show that this reproduces the correct commutation relations for the φ fields and their conjugate momenta. so the complex conjugate of ψ is ψ ∗ .. we find (2 + µ2 )ψ = 0. In terms of the classical fields.65) ∂L ∂L . but treat ψ and ψ ∗ as independent fields.. start with the Lagrangian L = ∂µ ψ ∗ ∂ µ ψ − µ2 ψ ∗ ψ (III. [ψ † (x. t)] = iδ (3) (x − y). t). we can now work from our ψ fields right from the start. ψ ψ (III. Πµ ∗ = . Πµ ∗ = ∂ µ ψ ψ ψ which leads to the Euler-Lagrange equations ∂µ Πµ = ψ ∂L → (2 + µ2 )ψ ∗ = 0. ψ ∂(∂µ ψ) ∂(∂µ ψ ∗ ) (III. ∂ψ (III.) Clearly a U (1) transformation on complex fields is equivalent to an SO(2) transformation on real fields. . (III. Π0 † (y. (III. We can similarly canonically quantize the theory by imposing the appropriate commutation relations [ψ(x.

. (III.. So we get the same equations of motion. gives A − A∗ = 0. since it only depends on the “length” of (φ1 . The correct way to obtain the equations of motion is to first perform a variation δψ which is purely real. Therefore the internal symmetry group is the group of rotations in n dimensions.68) where A is some function of the fields.68) get A = 0. (III. Then taking δψ = 0 we get A∗ = 0. Consider the EulerLagrange equations for a general theory of a complex field ψ. φa → b Rab φb (III. We first take δψ ∗ = 0 and from Eq. Non-Abelian Internal Symmetries A theory with a more complicated group of internal symmetries is n n 2 L= 1 2 a=1 ∂µ φa ∂ φa − µ φa φa − g a=1 µ 2 (φa ) 2 .53 varied.69) Then performing a variation δψ which is purely imaginary. Later on we will show that the only consistent way to couple a matter field to the electromagnetic field is for the interaction to couple a conserved U (1) charge to the photon field. φ2 . we get A = A∗ = 0. we find an expression for the variation in the action of the form δS = d4 x(Aδψ + A∗ δψ ∗ ) = 0 (III. Combining the two. We will refer to complex fields as “charged” fields from now on. δψ = δψ ∗ . We can see how this works to give us the correct equations of motion. Just as in the first example the Lagrangian was invariant under rotations mixing up φ1 and φ2 . A = A∗ = 0. so we can vary them independently. This gives the Euler-Lagrange equation A + A∗ = 0. 2.70) This is the same as the previous example except that we have n fields instead of just two. φn ).. at which point the U (1) charge will correspond to electric charge. A better analogue of the “charge” in this theory is baryon or lepton number..71) . If we instead apply our rule of thumb. “charged” only indicates that they carry a conserved U (1) quantum number. For a variation in the fields δψ and δψ ∗ . .. Note that since we haven’t yet introduced electromagnetism into the theory the fields aren’t charged in the usual electromagnetic sense. this Lagrangian is invariant under rotations mixing up φ1 . we imagine that ψ and ψ ∗ are unrelated.φn . (III. δψ = −δψ ∗ .

There are n(n − 1)/2 independent planes in n dimensions.3] = (∂ µ φ1 φ3 ) − (∂ µ φ3 φ1 ) µ J[2. or SU (n) × U (1). n dimensions).3] . and this theory has an SO(n) symmetry. for a theory with SO(3) invariance. For example.3] = (∂ µ φ2 φ3 ) − (∂ µ φ3 φ2 ) (III.74) the theory is invariant under the group of transformations ψa → b Uab ψb (III. For n complex fields with a common mass. Orthogonal. which is just multiplication of each of the fields by a common phase. neither do the currents or charges in the quantum theory.3] [Q[2.2] = (∂ µ φ1 φ2 ) − (∂ µ φ2 φ1 ) µ J[1.3] . We won’t be discussing non-Abelian symmetries much in the course.3] ] = iQ[1.73) This means that it not possible to simultaneously measure more than one of the SO(3) charges of a state: the charges are non-commuting observables. but we just note here that there are a number of non-Abelian symmetries of importance in particle physics. the currents are µ J[1.3] (III. We can write this as a product of a U (1) symmetry. Q[1.2] ] = iQ[2.2] ] = iQ[1.72) and in the quantum theory the (appropriately normalized) charges obey the commutation relations [Q[2. Q[1.3] . The symmetry group of the theory is the direct product of these transformations.54 where Rab is an n × n rotation matrix. and we can rotate in each of them. Q[1. and an n × n unitary matrix with unit determinant. The group of rotation matrices in n dimensions is called SO(n) (Special. and the charges of the strong .2] [Q[1. so there are n(n − 1)/2 conserved currents and associated charges. n n ∗ ∂µ ψa ∂ µ ψa a=1 2 L= −µ 2 ∗ ψa ψa −g a=1 |ψa | 2 (III. a so-called SU (n) matrix. This example is quite different from the first one because the various rotations don’t in general commute . The familiar isospin symmetry of the strong interactions is an SU (2) symmetry. A new feature of nonabelian symmetries is that.the group of rotations in n > 2 dimensions is nonabelian. just as the rotations don’t in general commute.75) where Uab is any unitary n×n matrix.

. Clearly. ... Uc |N (k1 ). 1. Charge Conjugation. C For convenience (and to be consistent with the notation we will introduce later).. let us refer to the b type particles created by the complex scalar field ψ † as “nucleons” and their c type antiparticles created by ψ as “antinucleons” (this is misleading notation. We could proceed much further here into group theory and representations. N (kn ) = |N (k1 ). χ = |N (k). since real nucleons are spin 1/2 instead of spin 0 particles. k k k Since this is true for arbitrary states | χ . but then we’d never get to calculate a cross section. χ . N (kn ) . “Grand Unified Theories” attempt to embed the observed strong. Consider some general state | χ and its charge conjugate | χ . N (k2 ).. χ = c† | χ = c† Uc | χ . D.78) . so Uc = Uc = Uc . (III.. Then b† | χ ≡ |N (k). c† → ck† = Uc c† Uc = b† . which are parameterized by some continuously varying parameter and can be made arbitrarily small. we must have Uc b† = c† Uc . electromagnetic and weak charges into a single symmetry group such as SU (5) or SO(10). k k k k k k (III. Discrete Symmetries: C. Three important discrete space-time symmetries are charge conjugation (C).. P and T In addition to the continuous symmetries we have discussed in this section. and denote the corresponding single-particle states by | N (k) and | N (k) . The charges of the electroweak theory correspond to those of an SU (2) × U (1) symmetry group. Given an arbitrary state |N (k1 ). .55 interactions correspond to an SU (3) symmetry of the quarks (as compared to the U (1) charge of electromagnetism). N (k2 ). parity (P ) and time reversal (T ). N (k2 ). and k Uc b† | χ = Uc |N (k). respectively. N (kn ) composed of “nucleons” and “antinucleons” we can define a unitary operator Uc which effects this discrete transformation. .76) 2 −1 † We also see that with this definition Uc = 1. theories may also have discrete symmetries which impose important constraints on their dynamics.77) (III. We can now see how the fields transform under C. hence the quotes). or k k † † b† → bk† = Uc b† Uc = c† . So we won’t delve deeper into non-Abelian symmetries at this stage. The discrete symmetry C consists of interchanging all particles with their antiparticles...

so Up | k = | − k (III. The amplitude for | ψ to evolve into | χ is therefore identical to the amplitude for | ψ to evolve into | χ : † † † χ(tf ) |Uc Uc eiH(tf −ti ) Uc Uc | ψ(ti ) = χ(tf ) |Uc eiH(tf −ti ) Uc | ψ(ti ) = χ(tf ) |eiH(tf −ti ) | ψ(ti ) . 2. which is easily seen by taking the complex conjugate of both equations. (III. since as long as ψ and ψ † always occur together in each term of the Hamiltonian it will be invariant. Similarly. Expanding the fields in terms of creation and annihilation operators. we immediately see that † † ψ → ψ = Uc ψUc = ψ † . In such a theory. however in theories with spin it gets more interesting). Denote the charge conjugates of these states by | ψ and | χ . t) → Up φ(x.79) † Why do we care? Consider a theory in which the Hamiltonian is invariant under C: Uc HUc = H (for scalars this is kind of trivial. As expected. for example. the amplitude for “nucleons” to scatter is exactly the same as the amplitude for “antinucleons” to scatter (we will see this explicitly when we start to calculate amplitudes in perturbation theory.80) So. momenta are reflected. C invariance immediately It is straightforward to show that any transition matrix elements are therefore unchanged by charge conjugation. x → −x. Parity. but C conjugation immediately tells us it must be true). (III. t)Up .56 A similar equation is true for annihilation operators. Clearly we will also have Up ak a† k † Up = a−k a† −k (III. P A parity transformation corresponds to a reflection of the axes through the origin. the transformation exchanges particle creation operators for anti-particle creation operators. Consider the amplitude for an initial state | ψ at time t = ti to evolve into a final state | χ at time t = tf . and vice-versa.81) where Up is the unitary operator effecting the parity transformation. ψ † → ψ † = Uc ψ † Uc = ψ.82) and so under a parity transformation an uncharged scalar field has the transformation † φ(x.

In this case.84) since that is also a symmetry of L. for example. Actually. t) → φ(−x.57 d3 k † ak eik·x−iωk t + a† e−ik·x+iωk t Up k 3/2 2ωk (2π) d3 k √ a eik·x−iωk t + a† e−ik·x+iωk t = −k 2ωk (2π)3/2 −k 3k d √ = ak e−ik·x−iωk t + a† eik·x+iωk t k 2ωk (2π)3/2 = φ(−x. Just as before. But this is just a question of terminology. (III..85) for some n × n matrix Rab . Eq. t) is not unique. we call φ a pseudoscalar. (??)) where we have added a nontrivial interaction term (of which we will have much more to say shortly). so the only sensible definition of parity is Eq. In other situations.83). for example L = L0 − gψ ∗ ψφ (which we shall discuss in much more detail in the next section). In fact. if you have a number of discrete symmetries of a theory there is always some ambiguity in how you define P (or C. t) → −φ(−x. Suppose we had a theory with an additional discrete symmetry φ → −φ. t) → ∂i φa (−x. t). for that matter). t) (III. (III. we call it a scalar. to be completely general. So long as this transformation is a symmetry of L it is a perfectly decent definition of parity. The point is. this transformation φ(x. t) → ±φa (−x. Under parity. but Eq. we could define a parity transformation to be of the form φa (x.φn . then ∂0 φa (x. or T . (III. t). In this case. The simplest example is 4 L= where µναβ 1 2 a=1 ∂ µ φa ∂µ φa − m2 φ2 − i a a µναβ ∂µ φ1 ∂µ φ2 ∂µ φ3 ∂µ φ4 0123 (III. (III.. theories with pseudoscalars look a little contrived. t) → φa (x. L = L0 − λφ4 (see Eq. t) (III. In some cases.84) is. if φa (x. φ → −φ is not a symmetry. t) = Rab φb (−x. we could equally well have defined the fields to transform under parity as φ(x. When there are only spin-0 particles in the theory.83) is not a symmetry of the theory. t) → ±∂0 φa (−x. When φ does not change sign under a parity transformation. The important thing is to recognize the symmetries of the theory. if we had a theory of n identical fields φ1 . and = 1. t) ∂i φa (x. any theory for which † Up LUp = L conserves parity.83) where we have changed variables k → −k in the integration. t) = Up √ (III.86) is a completely antisymmetric four-index tensor.87) .

However. it is somewhat awkward to express antilinear operators in this notation. is an operator which is anti-linear. p(t)]UT = UT iUT = i = −[q(−t). 3. Time Reversal. T . Therefore in order for parity to be a symmetry of this Lagrangian. The simplest anti-linear operator is just complex conjugation. three of the fields will be scalars. numbers are complex conjugated under an antilinear transformation.86) always contains three spatial derivatives and one time derivative because µναβ = 0 unless all four indices are different. linear transformation.58 where i = 1.89) and so † † UT [q(t).88) (III. We can see why this is the case by going back to particle mechanics and quantizing the Lagrangian 1 L = q2.92) . Then † UT q(t)UT = q(−t) dq(t) † † ˙ UT p(t)UT = UT U = −q(−t) = −p(−t) dt T (III. It doesn’t matter which. (III. Thus.91) That is. in which t → −t. a| ψ → Ω [a| ψ ] = a∗ | ψ (III. or else three must be pseudoscalars and one scalar. p(−t)] (III. T The last discrete symmetry we will look at is time reversal. Now. A more symmetric transformation is P T in which all four components of xµ flip sign: xµ → −xµ .90) and so we cannot consistently apply the canonical commutation relations for all time! Clearly UT can’t be a unitary operator. time reversal is a little more complicated than P and T because it cannot be represented by a unitary. ˙ 2 Suppose the unitary operator UT corresponds to T . Since Dirac notation is set up to deal with linear operators. Under an antilinear operator Ω. since parity reverses the sign of x but leaves t unchanged. an odd number of the fields φa (it doesn’t matter which ones) must also change sign under a parity transformation. 3. a| ψ → Ω [a| ψ ] = a∗ Ω| ψ . and one pseudoscalar. What we need. (III. 2. the interaction term in Eq. We need something else. in fact.

any of C.94) (III. relativistic field theory that the amplitude must be invariant under the combined action of CP T (this is called the CP T theorem). (III. First of all. However. it is a general property of any local. In a quantum field theory..95) (this is to be expected.97) . p(t)]Ω−1 = i∗ = −i = −[q(−t). −t). Hence this is exactly what is required for a P T transformation.90) and there is no contradiction: ΩT [aq(t)]Ω−1 = a∗ q(−t) T so Ωt [q(t).. t) → ΩP T φ(x. t)Ω† T P d3 k √ ak eik·x−iωk t + a† e−ik·x+iωk t Ω−1 = ΩP T PT k 2ωk (2π)3/2 d3 k √ = ak e−ik·x+iωk t + a† eik·x−iωk t k 2ωk (2π)3/2 = φ(−x. since time reversal flips the direction of all the momenta. ΩP T or on the states ΩP T |k1 .93) as required. ..kn = |k1 . It has no effect on the creation and annihilation operators. and a parity transformation flips them back).kn (III. (III.. so there is an extra minus sign in Eq.59 and in fact this is precisely what we need. The only thing it acts on is the i in the exponents occurring in the expansion of the fields φ(x. . P or T may be broken (we will see some examples of such theories later on). p(−t)] T (III.96) ak a† k Ω−1 = PT ak a† k (III. complex conjugation corresponds to the operator P T . it doesn’t screw up the commutation relations because ΩiΩ−1 = −i. In field theory.

Fix the sign of b by demanding the energy be bounded below. and write the energy. let’s pause and work through an example. ignoring the fact that ψ and ψ ∗ are complex conjugates. it is invariant under space-time translations and an internal U (1) symmetry transformation. (As explained in class.).) 2. and is a useful formalism for studying multi-particle quantum mechanics. but that’s all right: the action integral is real). . As the investigation proceeds. In the following years I gave it as a problem set. and find ω as a function of k. you just turn the crank. Although this theory is not Lorentz-invariant. I suggest you work through it before looking at the solution. Everything should turn out all right in the end: the equation of motion for ψ will be the complex conjugate of that for ψ ∗ . Find the Euler-Lagrange equations. The following problem was used as a midterm test the first time I taught this course (with rather bleak results . Find the plane-wave solutions. The Problem Consider a theory of a complex scalar field ψ L0 = iψ ∗ ∂0 ψ + b ψ ∗ · ψ. Don’t be.) (WARNING: Even though this is a non-relativistic problem.. Consider L0 as defining a classical field theory. Because the equations of motion are first-order in time. in dealing with complex fields. a conserved linear momentum and a conserved charge associated with the internal symmetry. you should recognize this theory as good old nonrelativistic quantum mechanics.. those for which ψ = ei(k·x−ωt) . where b is some real number (this Lagrange density is not real. It is only on these that you need to impose the canonical quantization conditions. Find these quantities as integrals of the fields and their derivatives. Treating the theory in this manner is called second quantization. don’t miss minus signs associated with raising and lowering spatial indices. Thus it possesses a conserved energy. our formalism is set up with relativistic conventions.) Identify appropriately normalized coefficients in the expansion of the fields in terms of plane wave solutions with annihilation and/or creation operators.60 IV. a complete and independent set of initial-value data consists of ψ and its conjugate momentum alone. (HINT: You may be bothered by the fact that the momentum conjugate to ψ ∗ vanishes. 1. and the conserved quantities will all be real. Canonically quantize the theory. EXAMPLE: NON-RELATIVISTIC QUANTUM MECHANICS (“SECOND QUANTIZATION”) To put some flesh on the formalism we have developed so far.

Treating ψ and ψ ∗ as independent fields. we find Π0 = ψ Π0 ∗ ψ ∂L = iψ ∗ . a (IV.7) .4) Recall that the conserved current is given in general by Jµ = a Πµ Dψa − F µ .5) (IV.) Find the equation of motion for the single particle state |k and the two particle state |k1 .2) Thus. The Euler Lagrange equations are ∂µ Πµ = a ∂L ∂φa (IV. (Normal-order freely.1) so we first need the momenta conjugate to the fields. as required. k2 in the Schr¨dinger Picture.3) (note that.61 linear momentum and internal-symmetry charge in terms of these operators.) This is a wave equations for ψ. the equations of motion gives the dispersion relation ωk = −b|k|2 . ∂(∂i ψ ∗ ) Πi = ψ Πi ∗ ψ ∂L = −b∂ i ψ ∗ ∂(∂i ψ) ∂L = = b∂ i ψ. the equations of motion for the two fields are i∂0 ψ = b i∂0 ψ ∗ = −b 2 ψ 2 ψ∗ (IV. expanding in normal modes ψ = ei(k·x−ωk t) . What physical quantities do b and the internal o symmetry charge correspond to? Solution 1. ∂(∂0 ψ) ∂L = = 0. The internal U (1) symmetry is (of course) ψ → e−iλ ψ. (IV. ψ ∗ → eiλ ψ ∗ . This is actually ensured by the fact that the action is real. the equations of motion are conjugates of each other.6) (IV. ∂(∂i ψ ∗ ) (IV.

)ψ − a0 b ψ ∗ · ψ.9) (IV. t). (IV. we need b < 0. Cancelling the i’s. and [ψ(x. t)] = [ψ ∗ (x. 2. we get [ψ(x. the conserved charge density is the time component of J µ J 0 = ψ∗ψ and the conserved charge Q is the integral of this quantity over all space. Hence.62 In our case. ψ ∗ (y. ψ F µ = aµ L. ψ(y. Q= d3 x J 0 = d3 x ψ ∗ ψ. ψ ∗ (x. Since the momentum conjugate to ψ ∗ vanishes. iψ ∗ . t)] = 0. where aµ is and arbitrary four vector (unit vector).8) For the invariance under space–time translations ψ(x) → ψ(x + λµ aµ ). (IV. the only surviving equal time commutation relation to impose is on ψ and its conjugate. ai = x. Pi = d3 x[iψ ∗ ∂ i ψ] It’s easy to see that both the energy and momentum are Hermitian. since DL = 0. We also have Dψ = −iψ. For a space translation: a0 = 0. t).10) For this energy to be bounded from below. H=E= d3 x[−b ψ ∗ · ψ] = −b d3 x | ψ|2 . Therefore J 0 = Π0 Dψ − a0 L = iψ ∗ aµ ∂µ ψ − a0 iψ ∗ ∂0 ψ − a0 b ψ ∗ · = −iψ ∗ (a · For a time translation: a0 = 1. ai = 0. . t)] = δ(x − y). we find Dψ = aµ ∂µ ψ. t). F µ = 0 (or equivalently a constant).

63 Now expand the fields in the plane wave solutions given in part (1) to get ψ(x. The field ψ therefore only annihilates a particle and ψ ∗ only creates particles. t) = ψ ∗ (y. k2 = −b |k1 |2 + |k2 |2 |k1 . Bk ] d3 k ei(k·x−ωt) e−i(k ·y−ωt) α2 δ(k − k ) d3 kAk ei(k·x−ωt) d3 kBk e−i(k·y−ωt) d3 keik·(x−y) = (2π)3 α2 δ(x − y) This means that α = 1 a† (2π)3/2 k 1 (2π)3/2 and Ak = 1 a (2π)3/2 k is an annihilation operator. t) = Assume that Ak = αak and Bk = αa† . ψ ∗ (y. Now we can go ahead and write the energy. whereas Bk = is a creation operator. The Hamiltonian acting on a one-particle state is therefore : H : |k = −b|k|2 |k and on the two-particle state is : H : |k1 . k2 . We find E = d3 x[−b ψ ∗ · d3 x ψ] d3 k d3 k a† e−i(kx−ωt) ak ei(k ·x−ω t) k · k k 1 (2π 3 ) 1 = −b (2π 3 ) = −b = −b Similarly we find d3 k d3 k a† ak e−iωt eiω t k · k (2π)3 δ(k − k ) k d3 k a† ak |k|2 k Q = Pi = d3 k a† ak k d3 k k i a† ak k This form for the momentum operator is to be expected. the momentum and the internal symmetry charge in terms of these creation and annihilation operators. t). then k [ψ(x. t)] = = = α2 d3 k d3 k d3 k ei(k·x−ωt) e−i(k ·y−ωt) [Ak . since a† ak is the usual number k operator.

The conserved charge Q= d3 k a† ak k is just the number operator. k2 = ∂t 2m ∂ |k|2 |k = |k ∂t 2m This is just the usual EOM for one. This clearly corresponds to the usual energy of one.64 (this is straightforward to show using the commutation relations of the creation and annihilation operators in the usual way). |k1 . The equations of motion for these states in the Schr¨dinger o picture are therefore i and i 1 ∂ |k1 |2 + |k2 |2 |k1 .and two-particle states if b = −1/2m.and two-particle states in NRQCD. since particle creation is a relativistic effect. k2 . This is a conserved quantity in a nonrelativistic theory. .

65 V. as we have seen. φ. This leads to the equation of motion ∂µ ∂ µ φ + µ2 φ = −ρ(x). the only equation of motion we have solved is the Klein-Gordon equation. we could expand the fields as sums of plane waves multiplied by creation and annihilation operators. Particle Creation by a Classical Source The simplest type of interaction we can introduce into the theory is to couple the φ field to a classical source: L = Lφ − ρ(x)φ(x) (V. (V. A. but it is incredibly dull because nothing happens.3) (2π) (2π) (2π) 2ωk 2ωk 2ωk We have expressions for the energy. We just have plane waves propagating.5) . but they never interact because the two Lagrangians are decoupled. (V. Although we have been talking about symmetries of general (possibly very complicated) Lagrangians. which is just a theory of free fields. known function of space and time which is only nonzero for a finite time interval. we had Lφ = 1 (∂µ φ∂ µ φ − µ2 φ2 ) 2 and for a complex field ψ we had Lψ = ∂µ ψ † ∂ µ ψ − m2 ψ † ψ. k (V. In the quantum theory.4) where ρ(x) is some fixed. φ(x) = ψ(x) = ψ † (x) = d3 k √ 3/2 d3 k √ 3/2 d3 k √ 3/2 ak e−ik·x + a† eik·x k bk e−ik·x + c† eik·x k ck e−ik·x + b† eik·x . this corresponds to a theory of noninteracting. We can make things more interesting by adding interaction terms to the Lagrangian. L = Lφ + Lψ is a theory of φ particles and ψ particles. INTERACTING FIELDS In this section we will put the formalism we have spend the past few lectures deriving to work. momentum and U (1) charge in our theory.2) (V. spinless bosons.1) Because we could solve the Klein-Gordon equation. For a real field.

A). just as a charge distribution is a source for electric field. satisfying (∂µ ∂ µ + µ2 )DR (x − y) = −iδ (4) (x − y) DR (x − y) = 0. so in four-vector notation we may write this as ∂µ ∂ µ Aν = 4πJ ν (V.7) where J µ = ( . Writing DR (x − y) = ˜ we find the algebraic equation for DR (k). is required so that the boundary condition φ(x) → φ0 (x) as x0 → −∞ is satisfied. The simplest way to find the Green function is to rewrite Eq.66 To realize why ρ(x) is a source term. as in Eq.11) d4 k −ik·(x−y) ˜ e DR (k) (2π)4 (V. (V. recall from classical electromagnetism that in the presence of a charge distribution (x.3).8) where DR (x − y) is the retarded Green function. and φ0 (x) may be expanded in terms of creation and annihilation operators. This theory is actually simple enough that we can solve it exactly. Before ρ(x) is turned on. we can construct the solution to the equation of motion as follows: φ(x) = φ0 (x) + i d4 y DR (x − y) ρ(y) (V. the theory is free.9) The second requirement. these two theories look quite similar. has no vector index and is a quantum field.10) . c ∂t c 2 ϕ− (V. If we start in the vacuum state.9) in momentum space. so we may interpret ρ(x) as a source for the φ field. Except for the fact that φ is massive. (V. t) the potentials obey the inhomogeneous wave equations 1 ∂2ϕ = −4π c2 ∂t2 1 ∂2A 4π 2 A − 2 2 = − . (V.6) ϕ and A form the components of the four-vector Aµ = (ϕ. that DR be the retarded Green function. what will we find at some time in the far future. t) and a current (x. After the source has turned on. after the source ρ(x) has been turned on and off again? We can answer this by solving the field equations directly. ˜ (−k 2 + µ2 )DR (k) = −i (V. x0 < y 0 . ) is the 4-current.

12) This doesn’t quite define DR : the k 0 integrand in Eq. we can close the contour in the bottom half plane. the vacuum expecation value of the commutator: DR (x − y) = θ(x0 − y0 )[φ(x). path of integration which passes above both poles. V. φ(y)]| 0 . (V. For x0 > y0 .12) has poles at k 0 = ±ωk .67 which immediately gives us DR (x − y) = i d4 k e−ik·(x−y) . or equivalently (since the commutator is a c-number. (V. the Green function vanishes for y0 > x0 . Thus. obtaining for the integral DR (x − y) x0 >y0 = d3 k (2π)3 1 −ik·(x−y) e 2ωk + k0 =ωk 1 e−ik·(x−y) −2ωk = = = d3 k (2π)3 k0 =−ωk 1 e−ik·(x−y) − eik·(x−y) 2ωk D(x − y) − D(y − x) [φ(x).14) .13) where the function D(x) was introduced in Section 2.1 The contour defining DR (x − y). φ(y)] (V. Then for y0 > x0 we can close the contour in the upper half plane. φ(y)] = θ(x0 − y0 ) 0 |[φ(x). not an operator). In order to define the integral. The retarded Green function DR (x − y) is therefore related to the commutator of two fields. giving zero for the integral since the path of integration doesn’t enclose any singularities. we must choose a path of integration around the poles. Let us choose a k0 −ωk ωk FIG. making this the appropriate contour for DR (x − y). 4 k 2 − µ2 (2π) (V.

The expectation value of the total number of particles ρ produced is dN = d3 k |˜(k)|2 . The Hamiltonian in the far future is now H= d3 k ωk a† − k i (2π)3/2 √ 2ωk ρ∗ (k) ˜ ak + i (2π)3/2 √ 2ωk ρ(k) ˜ (V. (V.15) where in the second line we have used the fact that if we wait until all of ρ(x) is in the past. ˜ (V. we have solved the theory. We have also defined the Fourier transform ρ(k) = ˜ d4 y eik·y ρ(y). we only need the second line in Eq.18) (this is obvious if you go back to the original derivation of H in terms of φ(x)) and so the expectation value of the energy of the system in the far future is 0 |H| 0 = d3 k 1 |˜(k)|2 . φ(x) = (2π) d3 k √ 3/2 2ωk ak + i (2π)3/2 √ 2ωk ρ(k) e−ik·x + h.8) gives φ(x) = x0 →∞ φ0 (x) + i φ0 (x) + i φ0 (x) + i d4 y d3 k θ(x0 − y0 ) e−ik·(x−y) − e−ik·(x−y) ρ(y) (2π)3 2ωk = = d3 k d4 y e−ik·(x−y) − eik·(x−y) ρ(y) (2π)3 2ωk d3 k e−ik·x ρ(k) − eik·x ρ(−k) ˜ ˜ (2π)3 2ωk (V. . (V. since in the far future we have free field theory again.19) Note that because we are in the Heisenberg representation. (V. after the source has been turned off.16) Thus we find. ρ (2π)3 2 (V. the spectrum of the Hamiltonian is just free particles. we are still in the ground state of the free theory – the state hasn’t evolved.17) Since all observables are built out of the fields. The time evolution of the system is all contained in the evolution of the fields. Now.68 For our present purposes. ρ (2π)3 2ωk (V.13). the theta function equals one over the whole domain of integration and may be dropped.c. Inserting this expression into Eq.21) . which means that the expectation value of the total number of particles created with momentum k is dN (k) = |˜(k)|2 ρ (2π)3 2ωk (V.20) and so each Fourier component of ρ produces particles with the corresponding four-momentum with a probability proportional to |˜(k)|2 .

which instructs us to place the operators that follow in order with the latest on the left. Choosing a path of integration which passes below both poles would give the advanced Green function. V. = θ(x0 − y 0 ) 0 |φ(x)φ(y)| 0 + θ(y0 − x0 ) 0 |φ(y)φ(x)| 0 ≡ 0 |T φ(x)φ(y)| 0 (V. This Green function. will be of central importance to scattering theory. Other paths of integration give Green functions which are useful for solving problems with different boundary conditions. x0 < y0 we close the contour above. since the poles are then at k 0 = ±(ωk − i ) and are displaced properly above and below . Another possibility is a path which goes below the pole at −ωk and above the pole at ωk . and we shall return to it shortly. This defines the Green function DF (x − y) = D(x − y).2 The contour defining DF (x − y). x0 < y0 . obeying GA (x − y) = 0 for x0 > y0 . It is convenient to write the Feynman propagator as DF (x − y) = where the limit d4 k i e−ik·(x−y) 4 k 2 − µ2 + i (2π) (V. When k0 −ωk ωk FIG. obtaining the same expression but with x and y interchanged.23) → 0+ is understood and the path of integration in the k 0 plane is now along the real axis.1).12) a bit more. called the Feynman propagator. The retarded Green function DR (x − y) was obtain by choosing the path of integration shown in Fig. D(y − x). when x0 > y0 we perform the k 0 integral by closing the contour below. let’s pause for a moment and study the expression (V. obtaining the result D(x − y) for the integral. x0 > y0 .69 B. More on Green Functions Since Green functions are of central importance to scattering theory.22) where the last line defines the time ordering symbol T. (V. This would be useful if we knew the value of the field in the far future and were interested in its value before the source was turned on. In this case.

25) The field equations are now coupled. Mesons Coupled to a Dynamical Source The Lagrangian Eq.4) is analogous to electromagnetism coupled to a current which is unaffected by the dynamics of the field. a source for the φ field. if g is small we can treat the interaction term as a small perturbation of free field theory. C. we will have to solve them perturbatively: that is. . Furthermore. and the resulting dynamics are quite complicated. We are therefore guaranteed that the interacting theory will also conserve charge.4). This model is much more complicated that the previous one.6 In fact. Note that the sign of the i term is crucial: if were negative. 6 Since g has dimensions of mass. but a full dynamical variable. just like ρ(x). so solving this theory is going to be much harder. In fact.24) Note that the potential only depends on ψ and ψ † in the combination ψ † ψ. we see that ψ † ψ is a current density. where M is some typical mass or energy in the problem. not their derivatives. We will be able to solve the equations of motion as a power series in g. comparing this with Eq. so the interaction term doesn’t break the U (1) symmetry. ∂µ ∂ µ ψ + m2 ψ = −gψφ. (V. and the time ordering would come out reversed. While this is in many cases a good approximation. most of the rest of this course will be concerned with applying perturbation theory to an assortment of different theories. (V. (V. the analogous situation is described by a potential which couples the two fields ψ and φ: L = Lφ + Lψ − gψ † ψφ. because there is a back-reaction: the current ψ † ψ in turn depends on the field φ. Instead. so the conjugate momenta are the same as they were in the free theory. Much of what is known about quantum field theory comes from perturbation theory. the contours would enclose the opposite poles. (V. The source is now not a prescribed function of space-time.70 the real axis. so the fields interact. the interaction depends only on the fields. as it was in the previous case. in the real world the current itself interacts with the electromagnetic field. the power series will actually be a series in g/M . however. In general we cannot solve this system of coupled nonlinear partial differential equations exactly. For scalar field theory. This equations of motion are ∂µ ∂ µ φ + µ2 φ = −gψ † ψ.

27) while in the Heisenberg picture the states are independent of time and the operators (and in particular. because in this form we know exactly how the fields act on the states of the theory. one of which carries a conserved charge. the solution to the Heisenberg equations of motion are no longer plane waves but instead something awful. In particular. (V. If the ψ fields were spin 1/2 fermions instead of spin 0 bosons we would have a theory of the strong interactions between nucleons. and the t depeno dence is carried entirely by the states OS (t) = OS (0) i d | ψ(t) dt S = H| ψ(t) S . We have already discussed the Schr¨dinger and Heisenberg pictures. (V. the operators don’t evolve with time. However. D. We’ll take advantage of this analogy and refer to the ψ particles as “nucleons” (in quotation marks) and the φ’s as mesons. the fields) carry the time dependence | ψ(t) H = | ψ(0) H . The interaction picture o combines elements of each. The Interaction Picture How do we set this problem up? First of all. This doesn’t look anything like the particles we see in the real world. Recall that in the Schr¨dinger picture.24) describes the interactions of two types of meson.71 The theory defined in Eq. but we will use it in this section as a toy model to illustrate our perturbative approach to scattering theory. All three pictures will coincide at t = 0: | ψ(0) S = | ψ(0) H = | ψ(0) I OS (0) = OH (0) = OI (0) (V. Unfortunately. We’ll call this our “nucleon”-meson theory. we have seen that the equations of motion look quite similar to the equations of motion of an electric field coupled to a current. We can fix this with a clever trick called the interaction picture. where the force is transmitted through the exchange of φ mesons. we would like to be able to write our fields in terms of creation and annihilation operators.26) where the subscript I refers to the interaction picture. we would like to make use of some of our previous results for free field theories. and O represents a generic operator with no explicit t dependence.

In our example. this would immediately give | ψ(t) | ψ(0) H = and the states would be independent of time.P. dt i (V. the Hamiltonian corresponding to the free-field Lagrangian). H]. just as in the Heisenberg picture. HI = 0. we find S ψ(t) |OS | ψ(t) S = I ψ(t) |OI (t)| ψ(t) I = S ψ(t) |e−iH0 t OI (t)eiH0 t | ψ(t) S (V. we see immediately that HI = −LI .31) ≡ eiH0 t | ψ(t) S .36) .72 d | OH (t) = [OH (t). From the equations of motion of the Schr¨dinger field.P. dt (V.32) = | ψ(t) H If we were dealing with a free field theory. This is the solution of the equation of motion i d OI (t) = [OI (t). a (V. (V.28) We showed earlier that matrix elements are the same in the two pictures.35) (V.26)). HI = −LI = gψ † ψφ. so we can continue to use all of our results for free fields. are defined by | ψ(t) I (V. H = H0 + HI (V.27). Demanding that matrix elements be identical in all three pictures. and HI contains the interaction term.P. In the interaction picture (IP) we split the Hamiltonian up into two pieces. H0 ]. All of the complications have been relegated to the equation of motion for the states. Eq. (V. States in the I.30) if LI contains no derivatives of the fields (so it doesn’t change the conjugate momenta from the free theory).34) This is useful because fields in an interacting theory in the I. Since H= d3 x a ˙ Π0 φ a − L = a d3 x a ˙ Π0 φa − L0 − LI .29) where H0 is the free Hamiltonian (that is. will evolve just like free fields in the Heisenberg picture. the operators evolve according to the free Hamiltonian: OI (t) = eiH0 t OI (0)e−iH0 t (where we have used Eq. we have o i ⇒ ⇒ d −iH0 t e | ψ(t) dt I I = HS e−iH0 t | ψ(t) + e−iH0 t i d | ψ(t) dt I I H0 e−iH0 t | ψ(t) i d | ψ(t) dt I = (H0 (0) + HI (0))e−iH0 t | ψ(t) I I = eiH0 t HI (0)e−iH0 t | ψ(t) = HI (t)| ψ(t) I (V.33) and so in the I. I (V.

and so on..) In particular.. with some particular momentum. Particles will be created and destroyed. (V. .34). annihilates a φ particle and creates a ψ particle and antiparticle: this corresponds to the decay process φ → ψψ. On the other hand. We say we are colliding two electrons. (V. photons. the Hamiltonian acts on the initial state. we start out with some initial state | i consisting of a number of isolated particles. we expect them to be eigenstates of particle number. or two protons. Scattering processes are particularly convenient to study because in many cases the initial and final states look like systems of noninteracting particles.37) These interactions don’t conserve particle number. The initial state looks simple. order by order in the interaction Hamiltonian. pions. just two colliding protons. the time dependence of the states is going to be taken into account perturbatively. In the first term. but a complicated mess of protons. for example. At this intermediate stage. and so forth. In this section we will set up a formalism to apply perturbation theory to scattering processes. and can contribute to a number of processes.36) in a complicated and non-linear way. at first order in perturbation theory the Hamiltonian can act once on the states. c† ca. .. and the states start to evolve according to Eq. even though N will not in general commute with the interaction Hamiltonian HI . Since the particles are widely separated. (V. E. The second corresponds to the absorption ψ + φ → ψ. eigenstates of the free Hamiltonian H0 . simply given by the free field equations. What do we mean by this? In a scattering process. bb† a† . We no longer have. bca† . producing more complicated processes like ψ + ψ → φ → ψ + ψ (ψ anti-ψ scattering through the creation of an intermediate φ). we don’t expect them to feel the effects of the potential in Eq. as expected from Eq. they begin to feel the potential. since HI in general doesn’t commute with N . and so they will look like free plane wave states (that is.24). As the particles approach one another. Since the Hamiltonian generates time evolution. The time dependence of the operators is trivial. Again we see explicitly that when HI = 0 the fields in the I. At second order in perturbation theory the Hamiltonian can acts twice on the state. or whatever. the system will look extremely complicated when expressed in terms of our basis of free particles. such as c† b† a. (V.P. are independent of time. The interaction term ψ † ψφ contains a collection of creation and annihilation operators.73 where HI (t) = eiH0 t HI (0)e−iH0 t . Dyson’s Formula We can already get an idea of how perturbation theory is going to work in the interaction picture.

Despite this. Similarly. because the interaction is responsible for the bound state. I should tell you that this is a bit of a fake. perhaps three protons. In fact. Several initial particles could collide and form a bound state. so our simple picture is not quite right. For processes where f(t) 1 t ∆ T ∆ FIG. (V. no matter how far you go into the past or future from a scattering process you never end up with a collection of free particles. Again it will look simple. If we turn off the interaction. our quick and dirty scattering theory will still work. the states will change. In the limit ∆ → ∞. When we quantize electromagnetism. You can see that this might be the case by imagining that instead of Eq. (V. In this case.3. This is the type of process we will be considering. V. we will see that this corresponds to a cloud of photons around the electron. such as p + p → 2 D (two protons fusing to form a deuterium nucleus). Before we go any further. as shown in Figure V. an electron still carries its electromagnetic field along with it. the bound state will fly apart. (V. We already know this from electromagnetism: long after the collision process.38) where f (t) = 0 for large |t| and f (t) = 1 for t near 0.3 The “turning on and off” function f (t) in Eq.24).74 We can imagine several results of the scattering process. we could have a process in which no bound states are formed. ∆/T → 0 we expect to recover the results of the original theory Eq. Instead. since when f (t) → 0 . The formalism we are going to develop for scattering theory will not be very useful in this situation. The scattering process occurs near t = 0.24). If we turn the interaction off. an antiproton and fourteen pions. bound states occur. The system will again look like a collection of noninteracting particles. f (t) clearly drastically changes the states in the far future. no matter how long we wait after the scattering process has occurred the final state will never look like an eigenstate of the free Hamiltonian. T → ∞. Then some long time after the interaction the system will consist of a bunch of widely separated particles.38). the “nucleons” in our toy model will always have a cloud of mesons around them. our theory is defined by the Lagrangian L = Lφ + Lψ − gf (t)ψ † ψφ (V.

We can solve for S iteratively: integrating both sides of Eq. If we define the scattering operator S | ψ(∞) = S| ψ(−∞) = S| i then the amplitude to find the system in some given state | f in the far future is f |S| i ≡ Sf i .75 the interaction turns off and the states will fly apart. Section 7. we find t | ψ(t) = | i + (−i) −∞ dt1 HI (t1 )| ψ(t1 ) .39) from t1 = −∞ to t. (V. In the limit T → ∞.2. but this would take us into technical details which we don’t have time for in this course. we expect that the simple states in the real theory will slowly turn into the eigenstates of the free Hamiltonian with unit probability.43) 7 See Peskin and Schroeder. ∆ → ∞ and ∆/T → 0 (the last requirement ensures that edge effects vanish) we should recover the full theory. (V.P. This description is really meant as a hand-waving way of justifying our approach in which the initial and final states are taken to be eigenstates of the free Hamiltonian. The hand-waving approach will have to suffice at this stage. for the proper treatment of this problem. which are not eigenstates of the free Hamiltonian.41) This is conventionally known as the “S-matrix element”. (V.39) 7 (we will drop the subscript I on the states.42) (V. we turn the interaction off very slowly (adiabatically) over a time period ∆. there must be a 1 − 1 correspondence between the asymptotic (simple) eigenstates of the full Hamiltonian and the eigenstates of the free Hamiltonian. In other words. if we imagine that a long time T /2 after the scattering process occurs. This means that we can’t consider bound states. But in cases where there are no bound states formed. (V. long after the collision has taken place. It is possible to justify this approach (more) rigorously. you might imagine that adding f (t) to the interaction won’t change the scattering amplitude at all. In particular. from now on) with the boundary condition | ψ(−∞) = | i . .40) We want to connect the simple description in the far past with the simple description in the far future. So we want to solve i d | ψ(t) = HI (t)| ψ(t) dt (V. since we will always be working in the I.

O2 (x2 )O1 (x1 ). (V.48) Notice that in the first term t2 > t1 . we obtain the following expansion for S: ∞ S= (−i)n n=0 ∞ −∞ t1 tn−1 dt1 −∞ dt2 . So the HI ’s are always ordered (a) (b) FIG.. (V. (V.. (V. while in the second t1 > t2 . As before.49) .. (V.46) and (b) Eq. −∞ dtn HI (t1 ). with the earlier one on the right.HI (tn ). Look at the n = 2 term. (V. we can write the term as ∞ −∞ ∞ ∞ dt2 t2 ∞ dt1 HI (t1 )HI (t2 ) dt2 HI (t2 )HI (t1 ).76 Iterating this gives t | ψ(t) = | i + (−i) + (−i)2 −∞ t dt1 HI (t1 )| i t1 −∞ dt1 −∞ dt2 HI (t1 )HI (t2 )| ψ(t2 ) . for example: ∞ −∞ t1 dt1 −∞ dt2 HI (t1 )HI (t2 ).46) This corresponds to integrating over the region −∞ < t2 < t1 < ∞ shown in part (a) of the figure. We can reverse the order of integration.45) There is a more symmetric way to write this.47). we define the time-ordered product T (O1 O2 ) of two operators O1 (x2 ) and O2 (x2 ) by T (O1 (x1 )O2 (x2 )) = O1 (x1 )O2 (x2 ).44) Repeating this procedure indefinitely and taking t → ∞.4 The shaded regions correspond to the region of integration in (a) Eq. (V. t1 = −∞ dt1 (V. V.. and noting that this is the same region of integration as in part (b) of the figure. t1 > t2 . t1 < t2 .47) so we can write the second term of the expansion as 1 2! ∞ −∞ ∞ ∞ t1 dt1 t1 dt2 HI (t2 )HI (t1 ) + −∞ dt1 −∞ dt2 HI (t1 )HI (t2 ) .

(V.. k2 (N ) . for n operators we define the time ordered product (or T -product) such that the operators are ordered chronologically. so there is no ambiguity in this definition.77 In terms of the time-ordered product. because then the only ordering which would contribute to this process would be ones with two “nucleon” annihilation operators on the right and two “nucleon” creation operators on the left. We can even be slick and write this series as a time-ordered exponential..53) where the time-ordering acts on each term in the series expansion. k4 (N ) . Since we know how the fields act on the states in the I.HI (tn )) (V. S = T e−i d4 xHI (x) ... (V. the earliest on the right and the latest on the left.P. we have | i = | k1 (N ). For example.. However. we can write the second term in the expansion of S as 1 2! ∞ −∞ ∞ dt1 −∞ dt2 T (HI (t1 )HI (t2 )). It would be much simpler if we could normal-order this expression. In fact. HI commutes with itself at equal times.51) dt1 ... Wick’s Theorem To evaluate the individual terms in Dyson’s formula we will have to calculate matrix elements of time ordered products of fields between the initial and final scattering states... The n’th term in the expansion of S may then be written as 1 n! and the expansion for S is then S = = (−i)n n! n=0 (−i)n n! n=0 ∞ ∞ ∞ −∞ ∞ ∞ −∞ ∞ dt1 . This is Dyson’s formula. −∞ dtn T (HI (t1 ). F. where k4 = k1 + k2 − k3 since our theory con- serves momentum.HI (xn ))..52) d4 x1 . in our meson-“nucleon” theory at second order in g we have to evaluate matrix elements of the form f |T (HI (x1 )HI (x2 ))| i = f |T (ψ † (x1 )ψ(x1 )φ(x1 )ψ † (x2 )ψ(x2 )φ(x2 ))| i . in this form it’s still rather messy.. (V. there is a relation between time-ordered and ..50) Similarly.HI (tn )) (V. −∞ dtn T (HI (t1 ).54) For the scattering process N + N → N + N (elastic scattering of two “nucleons”). d4 xn T (HI (x1 ). | f = | k3 (N ). because the T -product contains 16 arrangements of “nucleon” creation and annihilation operators.. this matrix element is straightforward to calculate.

55) between vacuum states to find that −− −− −− −− A(x)B(y) = 0 |A(x)B(y)| 0 = 0 |T (A(x)B(y))| 0 − 0 | : A(x)B(y) : | 0 = 0 |T (A(x)B(y))| 0 (V.φn : +... For the charged fields.. So we have found that the contraction of two fields is just the vacuum expectation value of the time ordered product of the fields.61) . (V.78 normal-ordered products.. −− −− A(x)B(y) ≡ T (A(x)B(y))− : A(x)B(y) : (V. We have already seen this object before .59) (V..it is the Feynman propagator for the field. Then T (A(x)B(y)) = (A(+) + A(−) )(B (+) + B (−) ) =: AB : +[A(+) . + φ1 φ2 .......56) −− −− so A(x)B(y) is a number (given by the canonical commutation relations).55) −− −− It is easy to see that A(x)B(y) is a number.φn ) = : φ1 . φ2 ≡ φa2 (x2 ). we can now state Wick’s theorem.. we define the contraction of two fields. For any collection of fields φ1 ≡ φa1 (x1 ). it is straightforward to show that the propagator is −−− −− −−− −− i d4 k ik·(x−y) ψ(x)ψ † (y) = ψ † (x)ψ(y) = e 4 2 − m2 + i (2π) k while other contractions vanish: −−− −−− −− −− ψ(x)ψ(y) = ψ † (x)ψ † (y) = 0. not an operator. so we can sandwich both sides of Eq. . 4 2 − µ2 + i (2π) k (V.57) since the vacuum expectation value of a normal ordered product of fields vanishes (the annihilation operators on the right annihilate the vacuum)... the T -product of the fields has the following expansion T (φ1 .. −− − φ(x)φ(y) = DF (x − y) = 0 |T (φ(x)φ(y))| 0 = where the lim →0+ d4 k ik·(x−y) i e . (V... : φn−3 φn−2 φn−1 φn : + .+ : φ1 φ2 . B (−) ] (V. therefore 0 |T (ψ(x)ψ(y))| 0 = 0.φn : +.. (V...φn : −− − −− −− + : φ1 φ2 φ3 .. Similarly. Consider first the case x0 > y 0 .) Having defined the propagator of a field.58) is implicit in this expression. which goes by the name of Wick’s theorem.60) (The last equation is true because ψ only creates c-type particles and annihilates b-type particles.. it is a number when x0 < y 0 .φn−1 φn : −− −− −− −− + : φ1 φ2 φ3 φ4 φ5 . To state Wick’s theorem.φn : + : φ1 φ2 φ3 .

Other combinations of annihilation and creation operators in this term can also contribute to N + N → N + N and N + N → N + N . whose matrix elements are easy to take without worrying about commutation relations.62) Wick’s theorem is true by definition for n = 2. (V. so let’s get a feeling for it by applying it to the expression for S at O(g 2 ) in our model: (−ig)2 2! d4 x1 d4 x2 T (ψ † (x1 )ψ(x1 )φ(x1 )ψ † (x2 )ψ(x2 )φ(x2 )).65) can contribute to elastic N N scattering. (V. You can also see that there is no combination of creation and annihilation operators that will contribute to N + N → N + N . Wick’s theorem has unravelled the messy combinatorics of the T -product. In its general form. The proof that this is true for all n is by induction.” So the operator −−−−−−−−−− −−−−−−−−−− −− −− :ψ † (x1 )ψ(x1 )φ(x1 )ψ † (x2 )ψ(x2 )φ(x2 ):≡:ψ † (x1 )ψ(x1 )ψ † (x2 )ψ(x2 ): φ(x1 )φ(x2 ) (V.61). . That is to say. leaving us with an expression in terms of propagators and normal-ordered products. N + N → N + N . and so not terribly illuminating. because the theory has a conserved U (1) charge which wouldn’t be conserved in this process. so we won’t repeat it here. the matrix element k3 (N ). to give a nonzero matrix element.66) is nonzero. The ψ field contains operators which annihilate a “nucleon” and create an “anti-nucleon. k4 (N ) | :ψ † (x1 )ψ(x1 )ψ † (x2 )ψ(x2 ): | k1 (N ).” The ψ † field contains operators which annihilate an “anti-nucleon” and create a “nucleon. and the ψ † fields can’t create anti-nucelons.79 On the right-hand side of the equation we have all possible terms with all possible contractions of two fields.64) This term can contribute to a variety of physical processes. It is reassuring to see that this actually works in practice. One of these terms is (−ig)2 2! d4 x1 −−−−−−−−−− −−−−−−−−−− d4 x2 : ψ † (x1 )ψ(x1 )φ(x1 )ψ † (x2 )ψ(x2 )φ(x2 ) : (V. We are also using the notation −−−− −−−− −− −− : A(x)B(y)C(z)D(w) : ≡ : A(x)C(z) : B(y)D(w) (V. The ψ fields would have to annihilate the nucleons. k2 (N ) (V. it looks rather daunting. Eq. We already knew this had to be the case.63) Wick’s theorem relates this to a number of normal-ordered products. because there are terms in the two ψ fields that can annihilate the two nucleons in the initial state and terms in the two ψ † fields that can create two nucleons.

69) Note that there are no arrows over the momenta in the states. A single term is able to contribute to a variety of processes like this because each field can either destroy or create particles. φ + φ → N + N .68) We really want S − 1. (V.70) . p2 (N ) |(S − 1)| p1 (N ). N + φ → N + φ. √ | k = (2π)3/2 2ωk | k . We can write these states as | k = a† (k)| 0 where the relativistically normalized creation operator a† (k) is defined as √ a† (k) ≡ (2π)3/2 2ωk a† k (V. For N N → N N scattering we want the matrix element p1 (N ). S matrix elements from Wick’s Theorem Having used Wick’s theorem to relate (unpleasant) T-products to products of normal ordered fields and contractions (which are easy to work with). p2 (N ) . N + N → φ + φ.67) This term can contribute to the following 2 → 2 scattering processes (you should verify this): N + φ → N + φ. We are now doing relativistic field theory in earnest and so we are going to use our relativistically normalized states from the first lecture. let’s now calculate the scattering amplitude for “nucleon”-“nucleon” scattering at first order in perturbation theory. G. but instead for the matrix element f |(S − 1)| i .71) (V.80 Another term in the expansion of the T -product is (−ig)2 2! d4 x1 −−−−−−− −−−−−− d4 x2 : ψ † (x1 )ψ(x1 )φ(x1 )ψ † (x2 )ψ(x2 )φ(x2 ) : (V. because we aren’t interested in processes in which no scattering at all occurs. which corresponds to the leading order term of the Wick expansion.72) (V. First note that for a given process we are interested not in having an expression for the operator S. (V. not S.

81 and the scalar field φ has the expansion φ(x) = d3 k a(k)e−ik·x + a† (k)eik·x .76) Now.78) From the explicit expansion of ψ in terms of b† (k) and c(k) and Eq.77) (since we only have nucleons in the initial and final states.75). (V.73) From Eqs.75) (V.74) Similar relations holds for the relativistically normalized “nucleon” and “anti-nucleon” creation and annihilation operators. (V. p2 | :ψ † (x1 )ψ(x1 )ψ † (x2 )ψ(x2 ): | p1 . So in equations. (2π)3 2ωk (V. to evaluate Eq. p1 . (V. (V. p2 (V. p2 . p2 = p1 . (V. (V. and using the nucleon creation terms in ψ † (x1 ) and ψ † (x2 ) to create the two nucleons in the final state. a† (k)]| 0 = (2π)3 2ωk δ (3) (k − k )| 0 and so d3 k a(k )| k = | 0 . p2 (N ) = b† (p1 )b† (p2 )| 0 .69) at second order in the Wick expansion we need to evaluate the matrix element in Eq.79) . so a relativistically normalized incoming two nucleon state is | p1 (N ). The only way to get a nonzero matrix element is by using the nucleon annihilation terms in ψ(x1 ) and ψ(x2 ) to annihilate the two incoming nucleons. I’m going to suppress the “N ” label on the states). p1 . p2 = e−ip1 ·x1 −ip2 ·x2 + e−ip1 ·x2 −ip2 ·x1 . Any other combination of creation and annihilation operators will give zero inner product. you can easily show that 0 |ψ(x1 )ψ(x2 )| p1 . p2 |ψ † (x1 )ψ † (x2 )| 0 0 |ψ(x1 )ψ(x2 )| p1 .71).70) and (V. (2π)3 2ωk (V.66). (V. we also find a(k )| k = a(k )a† (k)| 0 = [a(k ). p2 | :ψ † (x1 )ψ(x1 )ψ † (x2 )ψ(x2 ): | p1 .

This just enforces energy-momentum conservation for the scattering process. Since we are integrating over x1 and x2 symmetrically.80) Notice that the first two terms on the first line of the final answer differs by the interchange x1 ↔ x2 .84) . This factor of 2 cancels the 1/2! in Dyson’s formula. so this becomes (−ig)2 d4 k i 4 k 2 − µ2 + i (2π) (2π)4 δ (4) (p1 − p1 + k)(2π)4 δ (4) (p2 − p2 − k) (V.58). and pi is the sum of initial momenta. The x1 and x2 integrations are easy to do – they just give us δ functions. we can do the k integration using the δ functions. Using our expression for the φ propagator. Eq.81) × ei(p1 −p1 +k)·x1 +i(p2 −p2 −k)·x2 + ei(p2 −p1 +k)·x1 +i(p1 −p2 −k)·x2 . (V.82 Using this and its complex conjugate. it is traditional to define the invariant Feynman amplitude Af i by f |(S − 1)| i = iAf i (2π)4 δ (4) (pf − pi ). −− −− and since φ(x1 )φ(x2 ) is symmetric under x1 ↔ x2 . p2 = (eip1 ·x1 +ip2 ·x2 + eip1 ·x2 +ip2 ·x1 )(e−ip1 ·x1 −ip2 ·x2 + e−ip1 ·x2 −ip2 ·x1 ) = eip1 ·x1 +ip2 ·x2 −ip1 ·x1 −ip2 ·x2 + eip1 ·x2 +ip2 ·x1 −ip1 ·x2 −ip2 ·x1 +eip1 ·x2 +ip2 ·x1 −ip1 ·x1 −ip2 ·x2 + eip1 ·x1 +ip2 ·x2 −ip1 ·x2 −ip2 ·x1 (V. and we get (−ig)2 (2π)4 δ (4) (p1 + p2 − p1 − p2 ) i (p1 − p1 )2 − µ2 +i + i (p2 − p1 )2 − µ2 + i .82) +(2π)4 δ (4) (p2 − p1 + k)(2π)4 δ (4) (p1 − p2 − k) . Since energy and momentum are conserved in any theory with a time. p2 | :ψ † (x1 )ψ(x1 )ψ † (x2 )ψ(x2 ): | p1 . The same is true for the last two terms. we obtain the following expression for the second order contribution to N N scattering (−ig)2 −− −− d4 x1 d4 x2 φ(x1 )φ(x2 ) × eip1 ·x1 +ip2 ·x2 −ip1 ·x1 −ip2 ·x2 + eip1 ·x2 +ip2 ·x1 −ip1 ·x2 −ip2 ·x1 = (−ig)2 d4 x1 d4 x2 i d4 k 4 k 2 − µ2 + i (2π) (V. Finally. we find four terms contributing to the matrix element p1 .and space-translation invariant Lagrangian.83) Notice that performing the final integral over δ functions leaves us with a factor of (2π)4 δ (4) (pf − pi ) where pf is the sum of all final momenta. (V. these terms must give identical contributions to the matrix element. (V.

88) (p1 − p2 )2 = −2p2 (1 + cos θ) (V. we find the invariant Feynman amplitude for “nucleon”“nucleon” scattering to be A = −g 2 1 (p1 − p1 )2 − µ2 +i + 1 (p2 − p1 )2 − µ2 + i .87) (V. Feynman diagrams are essentially pictures of the scattering process . our final result for N N elastic scattering. pˆ ) e p2 = ( p2 + m2 . −pˆ ) e where e · e = cos θ. was remarkably simple. These are called Feynman Diagrams. First of all.83 The factor of i is there by convention. This immediately gives ˆ ˆ (p1 − p1 )2 = −2p2 (1 − cos θ). nobody ever bothers thinking about Dyson’s formula or Wick’s theorem when calculating scattering amplitudes because there is a very simple diagrammatic shorthand which has all of this formalism built into it. Scattering into two identical particles at an angle θ is indistinguishable from scattering at an angle π − θ. (V. the amplitude must also be symmetric. and so the probability must be symmetrical under the interchange of the two processes. Note that the two terms in Eq.or more precisely.85) In the centre of mass frame. it reproduces the phase conventions for scattering in nonrelativistic quantum mechanics. so a given Feynman diagram will contain n interaction vertices. Eq. (V. Since these particles are bosons.86) Here we’ve dropped the i because the denominator never vanishes. and so we get A = g2 2p2 (1 1 1 + 2 . pictures of the fields and contractions which must be evaluated to give the matrix element. H. at n’th order in perturbation theory the interaction Hamiltonian will act n times. 2 − cos θ) + µ 2p (1 + cos θ) + µ2 (V. pˆ) e e p2 = ( p2 + m2 .88) are required because of Bose statistics. Indeed.85). Diagrammatic Perturbation Theory While the intermediate steps were a bit messy. we can write the momenta as p1 = ( p2 + m2 . according to simple rules. These are diagrams representing the . Thus. They are very easy to construct. (V. and θ is the scattering angle. −pˆ) p1 = ( p2 + m2 .

5 Interaction vertex for the “nucleon”-meson theory.64).64) corresponds to the diagram in Fig. The arrows will always line up.7 Diagrammatic representation of the contraction in Eq. any fields which are left uncontracted must either annihilate particles from the incoming .6. To distinguish ψ’s from ψ † ’s.6 Diagrammatic representation of the contraction in Eq. (V. V. (V.5. V. contractions are represented by connecting the lines coming out of different vertices. join the lines of the contracted fields. So the term in Eq. (Since the arrows always line up. A single interaction vertex for our toy theory is shown in Fig. FIG. while the term in Eq. V. V. we have only drawn one arrow on the contracted nucleon lines).84 interaction Hamiltonian in which each field in a given term of HI is represented by a line emanating from the vertex.7. ψ(x)ψ(y) and ψ † (x)ψ † (y). (V. we can draw an arrow on the corresponding line. Now. are zero. FIG.67). (V.67) corresponds to the diagram in Fig. −−− −−− −− −− because the contractions for which they don’t. V. V. An unarrowed −− − line will never be connected to an arrowed line because ψ(x)φ(y) is clearly zero as well. Next. Any time there is a contraction. FIG.

For the first diagram for N N scattering.are also used in other books (as well as previous versions of these notes!). top to bottom. We could just as well say that the meson is emitted from the second nucleon and then absorbed by the first. If there are different ways of doing this. To distinguish it from a physical particle. Other conventions . V. you can say that a nucleon with momentum p1 comes in and interacts. Since the two incoming nuclei are identical. It therefore can’t exist as a real particle. scattering it into a nucleon with momentum p2 . Note that I am using the convention that Feynman diagrams are read from left to right . the direction of the arrow on the line indicates the direction of flow of the U (1) charge.) The second diagram must be there because of Bose statistics. it is referred to as a “virtual” meson. the graph has no time-ordering in it. and it is reabsorbed by a nucleon with momentum p2 . but the virtual meson doesn’t satisfy k 2 = µ2 . bottom to top . An incoming arrow in the initial state corresponds to a nucleon being annihilated. an outgoing arrow in the initial state corresponds to an anti-nucleon and an outgoing arrow in the final state corresponds to an outgoing nucleon. so we must add the corresponding amplitudes. These diagrams have a very simple physical interpretation. the meson must not live long enough for its energy to be measured to great enough accuracy to measure this discrepancy. Thus. shown in Fig. V. The arrows on the lines indicate the flow of conserved charge.8: FIG. In terms of the uncertainty principle. we write down a separate Feynman diagram for each distinct labeling of the external legs. the other arrows indicate the direction of momentum flow. For N N scattering.right to left.so the incoming states come in on the left and the outgoing states go out on the right. Similarly. scattering into a nucleon with momentum p1 and a meson with momentum k = p1 − p1 . they correspond to indistinguishable processes. this gives us the two Feynman diagrams.8 Feynman diagrams contributing to N N scattering at order g 2 . an incoming arrow in the final state corresponds to an anti-nucleon being created. but must be reabsorbed after a short time. (Note that although we are writing this as though there is a definite ordering to these events. Energy and momentum are conserved in this process.85 state or create particles from the outgoing state. it is in principle impossible to say which of the incident nuclei carries p1 and which . For nucleons.

Having written down our two diagrams and interpreted them. write down a factor of (−ig)(2π)4 δ (4) i ki where ki is the sum of all momenta flowing into (or out of. Then i k 2 − m2 + i i k 2 − µ2 + i . Note that there is no excuse for not getting the factors of (2π)4 right. it’s a bit simpler even than this: we can shortcut some of the trivial delta functions and integrations by simply imposing energy-momentum conservation on the momenta flowing into each vertex.86 carries p2 . We can incorporate these simplifications into our Feynman rules for iAf i . and every δ (4) function always comes with a factor of (2π)4 . and so the amplitude must sum over both of them. Note that Bose statistics are automatically built into our creation and annihilation operator formalism. as long as you’re consistent) the vertex. we can now evaluate their contributions to the scattering amplitude iAf i by the following rules (called the Feynman rules for the theory): (a) At each vertex. (b) For each internal line with momentum k flowing through it. Every factor of d4 k always comes along with a factor of (2π)−4 . After drawing all possible diagrams at each order. and D(k 2 ) = for a “nucleon”. write down a factor d4 k D(k 2 ) (2π)4 where D(k 2 ) is the propagator for the appropriate field: D(k 2 ) = for a meson. where pI and pF are the sums of the total initial and final momenta. if you like. (2π)4 δ(pF − pI ). respectively. Actually. That’s it. (c) Divide the final result by the overall energy-momentum conserving δ function. The processes occuring in the two graphs are indistinguishable. assign a momentum to each line (internal and external) and enforce energy-momentum conservation at each vertex.

V. (V. FIG. However. write down a factor of (−ig). there are also diagrams with closed loops for which energy-momentum conservation at the vertices is not sufficient to fix all the internal momenta. (V.9 corresponds to the matrix element obtained from the contraction −−− −−− −− −− p | : ψ † (x1 )ψ(x2 )ψ(x1 )ψ † (x2 )φ(x1 )φ(x2 ) : | p . Thus we add an additional Feynman rule for iA: . (b ) For each contracted line.90) corresponds to the two-loop graph in Fig. V. so we must integrate over both momenta.89) Enforcing energy-momentum conservation at each vertex is still not sufficient to fix the momentum FIG.87 (a ) At each vertex. write down a factor of the propagator for that field. This is fine for graphs like the ones we have been considering. and so we must keep the factor of d4 k (2π)4 Similarly.91) (V. the diagram in Fig. V.91). neither p nor k is constrained.10). k flowing through the loop. For example.89). the fully contracted term −−− −−− −−− −−− −− −− 0 | : ψ † (x1 )ψ(x2 )ψ(x1 )ψ † (x2 )φ(x1 )φ(x2 ) : | 0 (V. In this diagram.10 Feynman diagram corresponding to matrix element (V.9 Feynman diagram corresponding to matrix element (V.

(V.11 Feynman Diagrams contributing to N N → N N N (p1 ) + N (p2 ) → N (p1 ) + N (p2 ): There are two Feynman graphs contributing to this process.11). I.13). shown in Fig. The diagrams are the same in the two figures because they have the same arrangement of lines and vertices: in the first figure. we immediately read off iA = (−ig)2 i (p1 − p1 )2 − µ2 + i (p1 + p2 )2 − µ2 . with the two φ’s contracted. the vertices are N (p1 ) − N (p1 ) − φ and N (p2 ) − N (p2 ) − φ in both diagrams. and immediately read off the amplitude (V. Applying our Feynman rules to these diagrams. Since the diagrams are simply a shorthand for matrix elements of operators in the Wick expansion. V. which gives .88 (c) For each internal loop with momentum k unconstrained by energy-momentum conservation. Similarly. (V.92) It is important to be able to recognize which diagrams are and aren’t distinct.85).8.11 as in Fig. We could even be perverse and draw the second diagram as in Fig. (V. We could just as well have drawn the diagrams in Fig. V. N (p1 )+N (p2 ) → φ(p1 )φ(p2 ). The amplitude is given by the diagrams in Fig. or nucleon-nucleon annihilation into two mesons. (V.12). the second diagrams in both figures are identical. V. it is a simple matter to write down the two Feynman diagrams in Fig.(V. More Scattering Processes We can now write down some more Feynman diagrams which contribute to scattering at O(g 2 ): FIG. (2π)4 With these rules. the orientation of the lines inside the graphs have absolutely no significance.14). write down a factor of d4 k .

we could have drawn the first diagram as shown in Fig.94) Once again. clearly these are simply related to the analogous process with particles instead of antiparticles. 2 − m2 (p1 − p1 ) (p1 − p2 )2 − m2 (V. .89 FIG. which differ only by the exchange of the identical particles in the final state. Note that there are processes such as N N → N N and N φ → N φ which we didn’t write down.12 Alternate drawing of the Feynman diagrams in Fig. (V. + 2 − m2 (p1 − p2 ) (p1 + p2 )2 − m2 (V. or nucleon-meson scattering. From the two diagrams in Fig. iA = (−ig)2 i i + . (V.93) In this case we have virtual nucleons in the intermediate state. N (p1 ) + φ(p2 ) → N (p1 ) + φ(p2 ). (V. FIG.11). V. V. instead of virtual mesons. Indeed.15) we obtain iA = (−ig)2 i i . the fact that these amplitudes FIG. V.13 Alternate drawing of diagram the second diagram in the previous figure.14 Diagrams contributing to N N → φφ. Bose statistics are taken into account by the two diagrams.16) This completes the list of interesting scattering processes at O(g 2 ). Once again.

17). there are more complicated diagrams contributing to these scattering processes. .95) The reason for the factor of 4! in the definition of the coupling is made immediately clear by examining the perturbative expansion of the theory.17 Interaction vertex for φ4 interaction. k2 . This theory has a single interaction vertex. For example. V. At higher orders in perturbation theory. φφ → φφ scattering is the completely uncontracted term − λ k . 4! 1 2 (V. V. are identical to those with the corresponding antiparticles is a consequence of charge conjugation invariance (C). 4! (V. giving a total of 4! different combinations. any one of the remaining three can annihilate the second.15 Diagrams contributing to N φ → N φ.16 Alternate drawing of the first diagram in the previous figure.90 FIG. discussed in the previous chapter. FIG. In some cases there are additional combinatoric factors which must be incorporated into the Feynman rules. V. any one of the φ fields can annihilate the first meson. At O(λ) in perturbation theory.96) Now. k | : φ(x)φ(x)φ(x)φ(x) : |k1 . leaving either of the remaining fields to create either of the final mesons. (V. Consider the Lagrangian of a self-coupled scalar field L = L0 − λ 4 φ . shown in Fig. the factors of 4! cancel and we are left simply with −iλ. for N N → N N scattering in our meson-“nucleon” theory. For the Feynman rule for this vertex. the only term which contributes to FIG.

The first diagram arises from Wick FIG. however.99) The evaluation of integrals of this type is a delicate procedure.97) FIG. that for large k µ the integral behaves as d4 k k8 .91 at O(g 4 ) we have diagrams like the two shown in Fig. Note. (V. Because of the overall energy-momentum conserving δ function. it does not matter whether we label. φφ → φφ scattering. the bottom line by k − p2 or k + p1 − p1 − p2 . for example. and we won’t discuss it in this course. We can also see explicitly that energy-momentum conservation at the vertices leaves one unconstrained momentum k which must be integrated over. × 2 − m2 + i )((k − p )2 − m2 + i ) ((k + p1 − p1 ) 2 (V. (V.19 Diagram contributing to φφ → φφ scattering. V. contractions of the form −−− −−− −−− −−− −− −− −− −− : ψ † (x1 )ψ(x2 )ψ(x3 )ψ † (x4 )ψ(x1 )ψ † (x3 )ψ † (x2 )ψ(x4 )φ(x1 )φ(x2 )φ(x3 )φ(x4 ) : whereas the second diagram arises from Wick contractions of the form −−− −−− −−− −−− −− −− −− −− : ψ † (x1 )ψ(x2 )ψ(x4 )ψ † (x4 )ψ(x1 )ψ † (x3 )ψ † (x2 )ψ(x3 )φ(x1 )φ(x2 )φ(x3 )φ(x4 ) : At O(g 4 ) we also get a new process.19). (V. this last graph is iA = (−ig)4 d4 k i4 (2π)4 (k 2 − m2 + i )((k + p1 )2 − m2 + i ) 1 .18 Two representative graphs which contribute to N N → N N scattering at O(g 4 ).18). from the graph in Fig. V. momenta flowing through the internal lines in this figure have been explicitly shown.98) The (V. According to our Feynman rules.

(V.102) 8 See.100) In the centre of mass frame. for example. to account for the difference between the relativistic and nonrelativistic normalizations of the states. we also must divide the relativistic result by (2m)2 . First of all. Let’s look at the nonrelativistic limit of the “nucleon-nucleon” scattering amplitude and try to understand it in terms of NRQM. By a sufficiently shrewd redefinition of the parameters in the Lagrangian. (V.8): iA = −ig 2 ig 2 = (p1 − p1 )2 − µ2 |p1 − p1 |2 + µ2 (V. However. This was a serious problem in the early years of quantum field theory. both classicially and quantum mechanically. There is a well-defined procedure known as renormalization which accomplishes this feat. Potentials and Resonances People were scattering nucleons off nucleons long before quantum field theory was around. two-body scattering is simplified to the problem of scattering off a potential. all infinities in observable quantities may be eliminated. II. To compare the relativistic and nonrelativistic amplitudes. This is not generally the case: in many situations loop integrals diverge. J.92 and so is convergent. the amplitude for an incoming state with momentum k to scatter off a potential U (r) into an outgoing state with momentum k is proportional to the Fourier transform of the potential. giving infinite coefficients at each order in perturbation theory. Chapter VIII.100) with that from the first diagram in Fig. Diu and Lalo¨. Defining the dimensionless quantity λ = g/2m. Quantum Mechanics. 4 .101) where we have used the fact that in the centre of mass frame. especially section e B. recall8 the Born approximation from NRQM: at first order in perturbation theory. Vol. it turns out that these infinities are similar in spirit to the infinity we faced when we found a divergent vacuum energy. (V. Let us therefore compare the nonrelativistic potential scattering amplitude Eq. Cohen-Tannoudji. and at low energies they could describe scattering processes adequately with non-relativistic quantum mechanics. ANR (k → k ) = −i d3 r e−i(k −k)·r U (r). the energies of the initial and scattered “nucleons” are the same. this gives d3 rU (r)e−i(p1 −p1 )·r = − λ2 |p1 − p1 |2 + µ2 (V.

2. which is mediated by a spin-2 field.106) E1 = E2 = p2 + m2 (V.104) This is called the Yukawa potential. The potential falls off exponentially with a range of µ−1 . we have p1 = −p2 ≡ p. and so the amplitude is proportional to A∝ 4M 2 1 .particle creation and annihilation is a purely relativistic effect. You can see directly from the scattering amplitudes that the sign of the Yukawa term in the amplitude is the same in “nucleon”-“nucleon”. “antinucleon”-“antinucleon” and “nucleon”-“antinucleon” scattering. 4πr (V. Indeed. This is a generic feature of scalar boson exchange .93 Inverting the Fourier transform gives U (r) = −λ2 eiq·r d3 q (2π)3 |q|2 + µ2 2 ∞ q 2 eiqr − e−iqr λ dq 2 = − 2 4π 0 q + µ2 iqr iqr 2 1 ∞ qe λ dq 2 = − 2πr 2πi −∞ q + µ2 (V. Gravity. and can be either attractive or repulsive. The potential is attractive. In the centre of mass frame.103) and closing the contour of the integral in the upper half complex plane to pick up the residue of the single pole at q = +iµ gives U (r) = − λ2 −µr e . This is in contract to the electrostatic potential. (V. so this term is suppressed compared with the i potential scattering term.105) . Now let’s look at “nucleon”-“antinucleon” scattering. What about the second term? First. is again universally attractive. which arises from the exchange of a spin-1 boson. This shouldn’t be surprising . The first term in Eq. + 4p2 − µ2 + i (V.92) again corresponds to scattering in a Yukawa potential.it leads to a universally attractive potential. The first feature is generic for the exchange of any particle. this was how Yukawa predicted the existence of the pion. by working backwards from the observed range of the force (about 1 fm) to predict the mass (about 200 MeV). We note two features: 1. the Compton wavelength of the exchanged particle. we note that in the nonrelativistic limit (p1 + p2 )2 4m2 p2 .

We can therefore drop the +i . This is reflected as a resonance in the cross section. and so the corresponding cross section does not have a resonance. the decay is not kinematically allowed). and rendering the resulting probability finite at the peak. For example. Note the resonance in the cross sections corresponding to the intermediate Z 0 boson going on-shell. the denominator never vanishes. . The +i in the amplitude is therefore not relevant in this case. This is because. If µ < 2m. At this kinematic point.3). V. Both processes can proceed either through an intermediate photon or an intermediate Z 0 boson.94 There are two cases to consider. but there is a clear resonance at a centre of mass energy around 91 GeV. However.20 Cross sections for electron-positron scattering to various final states. and instead is a real propagating particle. At low energies the intermediate photon dominates and the cross-section is monotonically decreasing. (VII. the mass of the Z 0 boson.in fact it is only in diagram with closed loops that the +i is crucial. if µ > 2m. This is another general feature of resonances. The amplitude for e+ e− → γγ does not have an intermediate Z 0 boson. and the scattering amplitude is a monotonically decreasing function of p2 . but rather a peak of finite height. shifting the pole into the complex plane. corresponding to the intermediate meson going further offshell as p2 increases. for µ > 2m. the meson is unstable to decay into a “nucleon”-“antinucleon” pair via the diagram in Fig. the scattering amplitude has a pole. When treated correctly. corresponding to the intermediate meson going on mass-shell at k 2 = (p1 + p2 )2 = µ2 . Note that the Z 0 peak in the figure is not a pole. this instability adds a finite imaginary piece to the denominator of the scattering amplitude. 10 5 LEP Total Cross Section [pb] 10 4 e+e → hadrons CESR DORIS _ 10 3 PEP PETRA TRISTAN 10 2 e+e → 10 _ + _ _ e+e →γ γ 0 20 40 60 80 100 120 Center of Mass Energy [GeV] FIG. the intermediate meson is no longer virtual. the figure below shows the cross-section (roughly the amplitude squared) for e+ e− to annihilation to muon pairs as well as to quarks (the curve marked “hadrons”). Searching for resonances in cross sections is one way of discovering new particles at colliders. (For µ < 2m. either .

|x = d3 k −ik·x e |k . Since the second quantized formulation of NRQM is the same as we have been using for relativistic QM.1) As you are about to show. EXAMPLE (CONTINUED): PERTURBATION THEORY FOR NONRELATIVISTIC QUANTUM MECHANICS We can get some more practice in the techniques of the previous chapter by applying them to the second quantized form of NRQM. the energy of the two-particle state) is shifted by x3 . x2 = V (|x1 − x2 |) x3 . t)ψ(x. (This part wasn’t on the midterm. but did make it onto a problem set). k2 . hence. t)ψ ∗ (x. Since this is a non-relativistic theory. we can do perturbation theory using the same techniques. This potential is non-local in space. 1. coupled to ψ by a potential: L = L0 + iχ∗ ∂0 χ + b χ∗ · χ− d3 x V (|x − x |)χ∗ (x . Using Dyson’s formula. introduced at the end of Chapter 3.95 VI. as a function of V .and momentum-conserving δ functions in your amplitude). this is a repulsive potential which depends only on the distance between the particles. Show that this term does indeed correspond to a two-body potential V (|x − x |) by showing that in the presence of the interaction the two-particle matrix element of the Hamiltonian (i. The Problem Consider adding a second field χ to the theory. t)χ(x . (2π)3/2 2. .2) where ∆k is the change in momentum of one of the scattered particles. x4 |x1 . k4 |S − 1|k1 . x4 |HI |x1 . and so violates causality in a relativistic theory. t). find (to lowest order in V ) k3 . x2 (normal order freely). Show that in the centre of mass frame you recover the Born approximation of nonrelativistic quantum mechanics for scattering off a potential. the amplitude for two body scattering. you can use the usual position eigenstates from NRQM. iA ∝ d3 r V (|r|)e−i(∆k)·r (VI. the expectation value of the interaction Hamiltonian in the position state |χ(r1 )ψ(r2 ) is V (|r1 − r2 |). (You should also get energy. (VI. but since the theory is non-relativistic this doesn’t matter. for V positive.e.

r e−ika x e−ikb x eikc x eikd x 0|a† a a† b akc ak−d |0 k k 1 = d3 x d4 xV (|x − x |) d4 ka d4 kb d4 kc d4 kd (2π 6 e−ika x e−ikb x eikc x eikd x δ 4 (k1 − kc )δ 4 (k2 − kd )δ 4 (k3 − ka )δ(k4 − kb ) × 4 1 = d3 x d4 xV (|x − x |)e−ik1 x e−ik2 x eik3 x eik4 x × 4 (2π 6 = = d3 x d4 xV (|x − x |)e−ix (k1 −k3 ) e−ix (k2 −k4 ) × 4 1 (2π 5 d3 x d3 xV (|x − x |)eix (k1 −k3 ) eix(k2 −k4 ) δ(E1 + E2 − E3 − E4 ) × 4 The factor of four comes from the various ways I can get a non–zero matrix element. t)χ(x . if the particles were identical. consider the amplitude for the scattering of two distinguishable “nucleons” ψa and ψb with identical masses and couplings to the φ field. we don’t encounter any time ordered products. Returning now to relativistic quantum mechanics. 1 Now d3 xd3 x = 8 d3 rd3 s to get s=x+x ⇒x= r+s . Using Dyson’s formula to first order. In the non-relativistic limit. This would be 4!. In analogy with your answer to part (c). This is not clear from the question. we can consider this as the scattering of two “nuclei” due to an effective nucleon-nucleon potential induced via meson exchange. 2 x = s−r 2 0|S − 1|0 = i i 1 1 d3 rd3 sV (|r|)e 2 r(k2 +k3 −k4 )−k1 ) e 2 s(k1 +k2 −k3 )−k4 ) δ(E1 + E2 − E3 − E4 ) 52 (2π i 4 = d3 rV (|r|)e 2 r(k3 −k1 )×2 δ 4 (k1 + k2 − k3 − k4 ) (2π 2 4 = d3 rV (|r|)e−ir(∆k) δ 4 (k1 + k2 − k3 − k4 ) (2π 2 . t)ψ ∗ (x. Since the free Hamiltonian for the χ field is the same as for the ψ field. we can use the same expansion as before.96 3. t)ψ(x. t)|0 d3 x d4 xV (|x − x |) d4 ka d4 kb d4 kc d4 kd e−µr . so life is pretty easy 0|S − 1|0 = 0| = 1 (2π 6 d4 xd3 x V (|x − x |)χ∗ (x . Now do a change in variables r =x−x. show that the effective nucleon-nucleon potential has the Yukawa form V (r) ∼ g 2 Is the force attractive or repulsive? Solution 1.

one finds V (|r|) = −g 2 2π 3 = −g 2 2π 3 0 d3 q ∞ eiqr q 2 + µ2 π eiqrcosθ sinθdθq 2 dq 2 2 0 q +µ With π 0 eiqrcosθ sinθdθ = 1 eirp − e−irp irp We find V (|r|) = −g 2 2π 3 e−qr 1 ∞ qdq 2 ir −∞ q + µ2 e−µr = −g 2 4π 4 r The last step involved closing the contour above and picking up the pole at p = iµ. iA = ig 2 |q|2 1 (2π)4 δ(k1 + k2 − k3 − k4 ) + µ2 Comparing that with the result obtained in c). . The Feynman amplitude for this Feynman diagram is iA = (−ig)2 i (2π)4 δ(k1 + k2 − k3 − k4 ). q 2 − µ2 + iε q = ∆p In the nonrelativistic limit. Thus.97 Note that I didn’t have to make use of the centre of mass frame. therefore the change in the energy is neglible(q0 = 2m2 + |p1 |2 + |p3 |2 − 2 m2 + |p1 |2 m2 + |p3 |2 → 0). m2 2 p2 . 2. This is an attractive potential.

They are not normalizable because they are plane waves. existing at every point in space-time. in a box measuring L on each side with periodic boundary conditions. But it looks like the probability is going to be proportional to |Sf i |2 ∼ |δ (4) (pF − pI )|2 .1) instead of δ (3) (k − k ). As we discussed earlier in the course. No wonder we got divergent nonsense. we will get the transition probability/unit time. (VII.98 VII. Another approach. which is really what we are interested in.2) The integrals where nx . over momentum for the expansion of the fields therefore becomes a sum over discrete momenta. DECAY WIDTHS. Furthermore. This is clearly not what we wanted. f |(S − 1)| i = iA(2π)4 δ (4) (pF − pI ) but we have yet to make contact with anything measurable. . CROSS SECTIONS AND PHASE SPACE At this stage we are now able to calculate amplitudes for a variety of processes by evaluating Feynman diagrams. we must square the amplitudes and sum over all observed final states. the allowed values of momenta must be of the form kx = 2πny 2πnz 2πnx .1). our states are normalized to δ (3) functions. for all time. In order to calculate probabilities. which is simpler and will give the right answer. This will solve the normalization problem because plane wave states in the box are square-integrable. kz = L L L (VII. the states being normalized to k |k = δ k k (VII. V → ∞ and hope it makes sense (it will). Instead. |δ (4) (pF − pI )|2 ?? Squaring a delta function makes no sense. The proper way to solve this problem is to take our plane wave states and build up localized wave packets. Finally. ky = . we can take the limit T. and turn the interaction on for only a finite time T . ny and nz are integers. What happened? The problem is that we are not working with “square-integrable” states. if we divide our answer by T . is to return to our old crutch and put the system in a box of volume V . Thus the scattering process is in fact occurring at every point in space. which are normalizable and for which the scattering process really is restricted to some finite region of space-time. as shown in the kx − ky plane in Fig.

so squaring it is now sensible. f |(S − 1)| i VT = iAViT (2π)4 δV T (pF − pI ) × f f (4) √ 1 √ 2ωk V √ i 1 √ 2ωk V (VII. and the notation V T indicates finite volume and time. we note that no experimentalist can measure the cross section for the scattering process N (p1 ) + N (p2 ) → N (p1 ) + N (p2 ) for any particular values of the momenta since it is impossible to resolve a single state. T → ∞ limit. However. we see that each time a field creates or annihilates a state it will bring in an additional factor of e±ik·x √ √ 2ωk V (VII. (VII. Switching back to our non-relativistic normalization for our states. we have for finite V = L3 and T .1 Allowed values of kx and ky in a box of measuring L on each side. It is only possible to measure all states about . when we were working with relativistically normalized states in infinite volume). Each quantity in Eq. Thus.5) is finite.5) where the products are over final (f ) and initial (i) particles. since we want to make contact with the real world. and the scalar field φ has the expansion φ(x) = k a† eik·x ak e−ik·x √ +√k √ √ 2ωk V 2ωk V (VII.6) approaches a δ function in the V. VII.99 ky 2π/L } } 2π/L kx FIG.3) (you can check that this is the right expansion by seeing that the commutation relations for a† and k ak reproduce the correct canonical commutation relations for the fields). The function δV T (p) ≡ (4) 1 (2π)4 T /2 dt −T /2 V d3 xeip·x (VII.4) (in contrast to the factor of e±ik·x we had in the last set of lecture notes.

V → ∞: lim (2π)4 δV T (p) (4) 2 2 (4) 2 = 1 (2π)4 T /2 dx −T /2 V d3 x = VT .T →∞ = V T (2π)4 δ (4) (p). and we find d4 p δV T (p) So indeed δ (4) (p) T. (VII.13) 1 . so we might anticipate it will be proportional to a delta function. with a coefficient which diverges in the limit V. Squaring our expression for the amplitude. summing over all final states and dividing by the total time T .100 some small region ∆k in momentum space. From the figure. there are L L L V ∆kx ∆ky ∆kz = ∆kx ∆ky ∆kz 2π 2π 2π (2π)3 (VII. T → ∞. (VII.12) Substituting this into Eq.11) is proportional to a δ function. If there are N particles in the final state.7) states. The only tricky part of taking the limit V. (VII..9) Note that the factors of V cancel in the product over final particles.8) states. 2ωi V (VII.9) and taking the limit V.10) Performing the integral d4 p.. which must be summed over. T → ∞ is the |δV T (p)|2 function. the exponential factors just give us (2π)4 δ (4) (x − x ). in the infinitesimal region of size d3 p1 d3 p2 . we find w (4) = |Af i |2 V (2π)4 δV T (pF − pI ) × T final ≡ |Af i |2 V D initial particles particles f d3 pf (2π)3 2Ef initial particles 1 2Ei V i (VII. So let’s look at d4 p δV T (p) (4) 2 (4) = 1 (2π)8 d4 p T /2 T /2 dt −T /2 −T /2 dt V d3 x V d3 x eip·x−ip·x .d3 pN there will be V d3 pf (2π)3 f =1 N (VII. We will find that for decay rates and cross sections the V in the product over initial particles also cancel. This will approach a function which is infinitely peaked at the origin. it is clear that in a region of size ∆kx ∆ky ∆kz . so we can trivially do the integrals over t and x . 2Ei V i . (2π)4 (VII. we find the following expression for the differential transition probability per unit time wV T /T : 2 wV T 1 (4) = |AViT |2 (2π)8 δV T (pF − pI ) × f T T f d3 pf (2π)3 2ωp i 1 .

Γ = 1/τ . In the particle’s rest frame. so w 1 = |Af i |2 D. (VII. where τ is the particle’s lifetime (in natural units). corresponding to decays and 2 → N particle scattering.16) Then the total decay probability/unit time. So let’s examine each of these in turn. we will define the quantity dΓ as the differential decay probability/unit time: dΓ ≡ 1 |Af i |2 D. T 2E (VII.14) Note that D is manifestly Lorentz invariant. each δ (n) function comes with a factor of (2π)n . and we have defined the factor D by D ≡ (2π)4 δ (4) (pF − pI ) final particles f d3 pf . in fact we are only really interested in processes with one or two particles in the initial state (but still an arbitrary number of particles in the final state). Γ. Therefore. Thus. any measurement of its energy (or mass. is Γ= 1 2M |Af i |2 D. Γ is called the “decay width. so there is no excuse for getting the factors of 2π wrong. and each integration dn k comes with a factor of (2π)−n .17) all final states Since the probability of the particle decaying/unit time is Γ.” If we consider the uncertainty principle. since the measure d3 pf /(2π)3 2Ef is the invariant measure we derived earlier on. Decays For a decay process there is a single particle in the initial state. we see that it does in fact correspond to a width. Now. 2M (VII. after a time t the probability that the particle has not decayed is just e−Γt . Since the particle exists for a time τ . The relevant physical quantities we wish to calculate are lifetimes and cross sections. a series of measurements of the particle’s mass will have a characteristic spread of order Γ. (2π)3 2Ef (VII. Also note that just as in the case of our Feynman rules. as they must in order to have a sensible V → ∞ limit. (VII.15) Note that the factors of V have cancelled. in its rest frame) must be uncertain by ∼ 1/τ = Γ.101 where we are using E and ω interchangeably for the energies of the particles. as indicated in Fig. .2). A.

there is one particle in the box of volume V . in terms of which the flux is |v1 −v2 |/V .20) Therefore the flux is N/At = |v|d. The width of the distribution is proportional to Γ. From Eq. we have dσ = differential probability unit time × unit flux A2 i 1 f = D× 4E1 E2 V flux A2 i 1 f = D 4E1 E2 |v1 − v2 | (VII.19) where v1 and v2 are the 3-velocities of the colliding particles. where σ is the total cross section. (VII. Cross Sections In a physical scattering experiment. In the case of two beams colliding. VII.102 # of measurements M ~Γ FIG. so d = 1/V . With our normalization. If the density of particles is d. B. and a measurement is made of the number of particles incident on a detector. the total cross section is σ= 1 1 4E1 E2 |v1 − v2 | |Af i |2 D. (VII. and the flux is |v|/V . This is easy to see. an infinitesimal detector element will record some number dN scatterings/unit time dN = F dσ (VII. Consider first a beam of particles moving perpendicular to a plane of area A and moving with 3-velocity v. (VII. then after a time t. So for an incident flux F =# of particles/unit time/unit area.2 The result of a series of measurements of the mass of a particle with lifetime τ = 1/Γ.19). The total number of scatterings per unit time is then N = F σ. With this definition.18) where dσ is called the differential cross section. a beam of particles is collided with a target (or another beam of particles coming in the opposite direction). the probability of finding either particle in a unit volume is 1/V .21) all final states . the total number of particles passing through the plane is N = |v|Atd. but since the collision can occur anywhere in the box the total flux is |v1 − v2 |/V 2 × V = |v1 − v2 |/V .

1 2 1 ∂E1 p1 ∂E2 p1 = . C.22) In the centre of mass frame. D= d3 p2 d3 p1 (2π)4 δ (4) (p1 + p2 − pI ). ∂p1 E1 E2 E1 E2 (VII. D for Two Body Final States These formulas for the decay widths and the cross sections are true for arbitrary numbers of particles in the final states. but four of the variables are constrained by the energy-momentum conserving δ function δ (4) (p1 + p2 − pI ). leaving only two independent variables to integrate over. We have also written d3 p1 = p2 dp1 d cos θ1 dφ1 ≡ p2 dp1 dΩ1 . pI = 0. we can write D in a simpler form. the δ function of energy must be converted to a δ function of p1 .23) where we have performed the integral over p2 . and p2 = −p1 is now implicit. 2 2 Since E1 = p2 + m2 . we must include a factor of ∂(E1 + E2 ) ∂p1 −1 . where θ and φ are the polar angles of p1 . V → ∞.26) . 1 1 To eliminate the last dependent variable. For two particles. (2π)3 2E1 (2π)3 2E2 (VII. there are six integrals to do (d3 p1 d3 p2 ). Therefore D = d3 p1 d3 p2 (2π)3 δ (3) (p1 + p2 )(2π)δ(E1 + E2 − ET ) (2π)3 2E1 (2π)3 2E2 d3 p1 (2π)δ(E1 + E2 − ET ) ⇒ (2π)3 4E1 E2 1 = p2 dp dΩ1 (2π)δ(E1 + E2 − ET ) 3 4E E 1 1 (2π) 1 2 (VII. the total energy available in the process. and E2 = p2 + m2 = p2 + m2 (from p2 = −p1 ).25) (VII. = ∂p1 E1 ∂p2 E2 and so p1 (E1 + E2 ) p1 ET ∂(E1 + E2 ) = = . Using the general formula δ[f (x)] = x0 ∈zeroes of f 1 δ(x − x0 ) |f (x0 )| (VII. and Ei ≡ ET . the factors of V cancel and the result is well-behaved in the limit T.24) to change variables from E1 to p1 . For a two-body final state. Thus.103 Once again.

if the particles are identical then these states are in fact identical. Going back to our QMD theory. because we treated the final states |A(p1 ). E1 E2 E1 E2 (VII. p1 ).3 Leading contribution to µ → N N . they are valid in any theory. P2 = 1 ( p2 + m2 . so p1 = 1 µ2 − 4m2 /2. 16π 2 ET (VII. (VII. B(p1 ) as distinct. As a second example. we must multiply D by a factor of 1/n!. tial four-momentum is (µ. In general. we consider 2 → 2 particle scattering in the centre of mass frame.30) 1 1 + E1 E2 = p1 E2 + E1 p1 ET = .3). we have assumed that the particles A and B in the final state are distinguishable. and so we’ve double-counted by a factor of 2!.27) In this derivation. 0) and the final momenta of the nucleons are P1 = ( p2 + m2 .104 The desired result for a two body final state in the centre of mass frame is therefore D= 1 p1 dΩ1 . The 3-velocities are v1 = p1 /(γm1 ) = p1 /E1 . so iA = −ig (simple!) and the decay width of the φ is Γ = g 2 p1 2µ 16π 2 µ g 2 p1 = 8πµ2 dΩ1 (VII. In fact. for n identical particles in the final state. p1 is straightforward to compute from energy-momentum conservation. Now let’s apply this to a couple of examples. shown in Fig. Since the results which follow are just kinematics and don’t depend on the amplitude A. and v2 = p2 /E2 = −p1 /E2 . so that the decay φ → N N is kinematically allowed. −p1 ).28) since dΩ = 4π. so |v1 − v2 | = p1 This leads to dσ = 1 1 pf dΩ1 |Af i |2 4pi ET 16π 2 ET (VII. There is only one diagram contributing to this decay at leading order in perturbation theory. B(p2 ) and |A(p2 ). The ini- FIG. suppose µ2 > 4m2 .29) . VII.

E2 and E3 . The derivation is straightforward but more lengthy than for the 2 body final state. where φ12 is the angle between particles 1 and 2. φ1 and φ12 . D for 3 Body Final States For three body final states. 32π 3 (VII. the amplitude is independent of Ω1 and φ12 . respectively. so we will just quote the result here.33) . In terms of these variables. D. then we will choose the independent variables to be E1 . there are nine integrals to do and four constraints from the δ function. in this case.32) in the centre of mass frame. D= 1 dE1 dE2 dΩ1 dφ12 256π 5 (VII. E2 . If the outgoing particles have energies E1 . we can integrate over those three variables ( dΩ1 dφ12 = 8π 2 ) to obtain D= 1 dE1 dE2 . In some cases (such as the decay of a spinless meson).105 and so pf 1 dσ |Af i |2 = 2E2 p dΩ 64π T i (VII. leaving five independent variables. θ1 .31) where pi and pf are the magnitudes of the three-momenta of the incoming and outgoing particles.

k3 . k2 . we’ve had a rather straightforward way to interpret Feynman diagrams: with all the external lines corresponding to physical particles. The question we will answer in this section is the following: Can we assign any meaning to this blob if the momenta on the external lines are unrestricted. we now proceed to put scattering theory on a firmer foundation. we will give three affirmative answers to this question. VIII. In order to reformulate scattering theory. the external momenta do not obey p2 = m2 . the momenta flowing through the in which only one type of scalar meson appears on the external lines. Let us denote the sum of all Feynman diagrams with n external lines carrying momenta k1 . we will restrict ourselves to Feynman diagrams k1 k2 ~ G (4) (k1. . Feynman Diagrams with External Lines off the Mass Shell Up to now.1 The blob represents the sum of all Feynman diagrams. it is useful first to think a bit about Feynman diagrams in a somewhat more general way than we have thus far. . (For simplicity. .106 VIII. . A. MORE ON SCATTERING THEORY Having now gotten a feeling for calculating physical processes with Feynman diagrams. that is. it just clutters up the formulas with indices). each one of which will give a bit more insight into Feynman diagrams. and maybe not even satisfying k1 +k2 +k3 +k4 = 0? In fact. such quantities do not directly correspond to S matrix elements. to include Feynman diagrams where the external legs are not necessarily on the mass shell. Clearly. k4 ) k4 external lines is unrestricted. The extension to higher-spin fields is straightforward. kn directed inward by ˜ G(n) (k1 . . off the mass shell. To do that. they will turn out to be extremely useful objects. . we will have to generalize this notion somewhat. kn ) as denote in the figure for n = 4. . they correspond to S matrix elements. k3 FIG. Nevertheless. . .

So this gives us a sensible.2(a-d). Recalling our discussion of Feynman diagrams (a) (b) (e) (c) (d) FIG.3 Lowest order contributions to G(4) (k1 . k2 . k4 ): k1 ~ G (2) k4 k3 i (k1 . do the appropriate integrals. k3 .2(e). which all have the form shown in Fig. we could simply plug it into this graph. . Let’s say we were interested in calculating the graphs in Fig. here are a few contributions to G(4) (k1 . k4 ). Before we go any further. . kn ). k2 . One simple thing we can do with these blobs is to recover S-matrix elements. . We could also include or not include the overall energy-momentum conserving δ-function. where the blob represents the sum of all graphs (at least up to some order in perturbation theory). We’ll include them both. VIII. . VIII. VIII. k3 .107 1. Answer One: Part of a Larger Diagram The most straightforward answer to this question is that the blob could be an internal part of a more complicated graph. ˜ So. and have something which we do know how to interpret: an S-matrix element. VIII. k4 ) = = k2 ! k! k1 2 k4 k3 + ! k! k1 4 3 2 2 k2 k3 i 2 + ! k! k1 2 k3 k4 + O (g 2 ) = (2 ) 4 (4) (k +k ) k 1 4 2 1 2 + i (2 ) 4 (4) (k + k ) k 2 + i +(2 permutations) ˜ FIG. For example. and possibly even useful. with internal loops. interpretation of the blob. we should choose a couple of conventions.2 The graphs in (a-d) all have the form of (e). k2 . k3 . we would label all internal lines with arbitrary momenta and integrate over them. We cancel off the . So if we had a table of blobs. for example. we could include or not include ˜ the n propagators which hang off G(k1 .

. + ikn · xn )G(n) (k1 . . Thus. .108 external propagators and put the momenta back on their mass shells. k2 = 2 k r − µ2 ˜ G(−k3 .4) where ρ(x) is a specified c-number source.3) (hence the tilde over G defined in momentum space). which we can then give another meaning to. k4 |(S − 1)| k1 . its Fourier transform. L → L + ρ(x)ϕ(x) (VIII. k2 ). consider the vacuum-to-vacuum transition amplitude.1) Because of the four factors of zero out front when the momenta are on their mass shell. At n’th order in ρ(x). i r=1 4 (VIII. all the contributions to 0 |S| 0 come from diagrams of the form shown in the figure. we get k3 . Now. 2. .2) (again keeping with our convention that each dk comes with a factor of 1/(2π)). . We can use it to obtain another function. shown in Fig.. this adds a new vertex to the theory. not an operator. f (x) = ˜ f (k) = d4 k ˜ f (k)eik·x (2π)4 d4 xf (x)e−ik·x (VIII.4.. As you showed in a problem set back in the fall. VIII. 0 |S| 0 . −k4 . . . in this modified theory. Now. xn ) = d4 k1 . k i ~(k ) FIG. as expected (since they don’t contribute to scattering away from the forward direction). (2π)4 d4 kn ˜ exp(ik1 · x1 + . the graphs that we wrote out above do not contribute to S − 1. Using the convention for Fourier transforms.4 Feynman rule for a source term ρ(x). . . Answer Two: The Fourier Transform of a Green Function We have found one meaning for our blob. . kn ) (2π)4 (VIII. VIII. . k1 . consider adding a source to any given theory. we have G(n) (x1 .

. . . to all orders. . 0 |S| 0 ρ is a functional of ρ. ρ(−kn ) G(n) (k1 . .. d4 xn ρ(x1 ) . . Recall we already introduced the n = 2 Green function (in free field theory) in connection with the exact solution to free field theory with a source.. It comes up often enough that it gets a name. . . . .5) The reason for the factor of 1/n! arises because if I treat all sources as distinguishable. (2π)4 d4 kn ˜ ρ(−k1 ) . the n’th order (in ρ(x)) contribution to 0 |S| 0 to all orders in g is in n! d4 k1 . kn ) ˜ ˜ (2π)4 (VIII. . xn ) δρ(x1 ) . . ˜ ˜ (2π)4 (VIII. . . . Really.109 kn k4 k2 k5 k3 k1 FIG..8) . it is just a function of an infinite number of variables. Z[ρ] = 0 |S| 0 ρ . . . which is how mathematicians denote functions of functions. kn ). . VIII. in the infinite dimensional generalization of a Taylor series.6) d4 x1 . . .5 n’th order contribution to 0 |S| 0 in the presence of a source. δρ(xn ) (VIII. The Fourier transform of the sum of Feynman diagrams with n external lines off the mass shell is a Green function (that’s what the G stands for). (VIII. we have δ n Z[ρ] = in G(n) (x1 .. . the value of the source at each spacetime point. Let’s explicitly note that the vacuum-to-vacuum transition ampitude depends on ρ(x) by writing it as 0 |S| 0 ρ . ρ(xn ) G(n) (x1 . xn ). I overcount the number of diagrams by a factor of n!. ρ(−kn ) G(n) (k1 . Thus. .) Z[ρ] is called the generating functional for the Green function because. . (2π)4 d4 kn ˜ ρ(−k1 ) . This provides us with the second answer to our question. Thus. . . we have 0 |S| 0 = 1 + = 1+ in n! n=1 in n! n=1 ∞ ∞ d4 k1 .7) (The square bracket reminds you that this is a function of the function ρ(x). .

As discussed in Peskin & Schroeder. Thus. the coefficients of z n is the Legendre polynomial Pn (x): 1 f (x. and hence all S matrix elements (and so all physical information about the system) are encoded in the vacuum persistance amplitude in the presence of an external source ρ. the functional derivative obeys the basic axiom (in four dimensions) δ δ J(y) = δ (4) (x − y).10) The generating functional Z[J] will be particularly useful when we study the path integral formulation of QFT. 2 = P0 (x) + zP1 (x) + z 2 P2 (x) + . . ∂ ∂ xj = δij . . let us consider adding a source term to the theory. 298. the generating function for the Legendre polynomials is f (x. . of the rule for discrete vectors. the coefficients are a set of functions of the other. all Green functions. Once again. z) = 1 + xz + (3x2 − 1)z 2 + . 3.9) This is the natural generalization. (VIII. or ∂xi ∂xi xj kj = ki . The term “generating functional” arises in analogy with the functions of two variables. (VIII. to continuous functions. and HI contains the interactions. z) = √ 1 z 2 − 2xz + 1 (VIII. xn ). or δJ(x) δJ(x) d4 y J(y)ϕ(y) = ϕ(x). j (VIII.110 where the δ instead of δ once again reminds you that we are dealing with functionals here: you are taking a partial derivatives of Z with respect to ρ(x). you can break the Hamiltonian up into a “free” and interacting . holding a 4 dimensional continuum of other variable fixed. . there’s more. . .12) Similarly. Thus. when Z[ρ] in expanded in powers of ρ(x). the Hamiltonian may be written H0 + HI → H0 + HI − ρ(x)ϕ(x) (VIII.11) since when expanded in z. as far as Dyson’s formula is concerned. . which when you Taylor expand in one variable. Now. Answer Three: The VEV of a String of Heisenberg Fields But wait. .13) where H0 is the free-field piece of the Hamiltonian. . p. For example. This is called a functional derivative. the coefficient of ρn is proportional to the npoint Green function G(n) (x1 .

2. and 3.111 part in any way you please. . ϕH (xn )) | 0 G(n) (x1 . S-matrix elements (the physical observables we wish to measure). (VIII. You can’t define a contraction for these fields. One of the tasks of the next few sections is to see how these are related.8) or (VIII. Now. and so we will subscript them with an H. These fields aren’t free: they don’t obey the free field equations of motion. what is the relation between these objects in our new formulation of scattering theory? . the fields evolve according to ϕ(x. ϕH (xn )) | 0 . 0)e−iHt where H = (VIII.16) For n = 2. n-point Green functions. but it’s not quite the same thing. . . .16). Now we want to get rid of that crutch. . t) = eiHt ϕ(x. . By introducing a turning on and off function. the two-point Green function is defined in the interacting theory. defined via by Eq. B. . (VIII. This looks like our definition of the propagator. I put quotes around “free”. we have the two-point Green function 0 |T (ϕH (x1 )ϕH (x2 )) | 0 . the sum of Feynman diagrams (the things we know how to calculate). d4 xn ρ(x1 ) . Green Functions and Feynman Diagrams From the discussion in the previous section. Let’s take the “free” part to be H0 + HI and the interaction to be ρϕ.14) d3 xH0 + HI . and thus you can’t do Wick’s theorem. . . we find Z[ρ] = 0 |S| 0 ∞ ρ = 0 |T exp i d4 xρ(x)ϕH (x) | 0 (VIII. because in this new interaction picture. ρ(xn ) 0 |T (ϕH (x1 ) . They are what we would have called Heisenberg fields if there had been no source. The Feynman propagator was defined as a Green function for the free field theory. . we could show that these were all related.15) = 1+ and so we clearly have in n! n=1 d4 x1 . . xn ) = 0 |T (ϕH (x1 ) . . and define perturbation theory in a more sensible way. The question we will then have to address is. just from Dyson’s formula. we have three logically distinct objects: 1.

−∞)| Ω (VIII. Thus. which implicitly referred to S matrix elements taken between free vacuum states.112 We set up the problem as follows.e.21) . . −k2 . in δρ(x1 ) . k2 |S − 1| k1 . will be “yes”. and the evolution operator U (t1 . | Ω . k2 = a 2 ka − µ2 ˜ G(−k1 . . in the presence of a source ρ(x). fortunately. with the interactions defined with the turning on and off function. satisfies H| Ω = 0. . Now. different from our previous definitions of Z and G. is Z = ZF ? The answer. with a time independent Hamiltonian H (the turning on and off function is gone for good) whose spectrum is bounded below. xn ) = 1 δ n Z[ρ] . no massless particles yet). since we derived Feynman rules based on the action of interacting fields on the bare vacuum. In other words. . 2. t2 ) is the Schor¨dinger picture evolution operator for the Hamiltonian o We then define G(n) (x1 .20) d3 x (H − ρ(x)ϕ(x)). and the Hamiltonian has actually been adjusted so that this state.17) Ω |Ω = 1. . . Is G(n) defined this way the Fourier transform of the sum of all Feynman graphs? Let’s call the G(n) defined as the sum of all Feynman graphs GF (n) (n) and the Z which generates these ZF . Are S matrix elements obtained from Green functions in the same way as before? For example. the physical vacuum. δρ(xn ) (VIII. Imagine you have a well-defined theory. The question then is. k2 )? i (VIII. k1 . This is actually rather surprising. whose lowest lying state is not part of a continuum (i.19) where the ρ subscript again means. we can ask two questions: 1. (VIII. is k1 . we can compute Green functions just as we always did. at least in principle. The vacuum is translationally invariant and normalized to one P | Ω = 0. not the full vacuum.18) = Ω |U (∞. let H → H − ρ(x)ϕ(x) and define Z[ρ] ≡ Ω |S| Ω ρ ρ (VIII. as the sum of Feynman graphs. Note that this is. is G(n) = GF ? Or equivalently.

113 The answer will be. (VIII. since Eq.25) is manifestly symmetric under permutations of the xi ’s. which you should look at as well. . (VIII. ϕH (xn )) | Ω . we have to show that this is equal to Eq. .25) HI | 0 Now. Equivalently. First of all. satisfying H0 | 0 = 0. we can simply prove the equality for a particularly convenient time ordering. . . .22) ZF [ρ] = t± →±∞ lim 0 |T exp −i t− [HI − ρ(x)ϕI (x)] | 0 (VIII. xn ).23) where I remind you that | 0 refers to the bare vacuum. So let’s take t1 > t2 > . The object which had a graphical expansion in terms of Feynman diagrams was t+ (n) (VIII. . we know that we will have to adjust the constant part of HI with a vacuum energy counterterm to eliminate the vacuum bubble graphs when ρ = 0. xn ) = Ω |T (ϕH (x1 ) . Now let’s show that this is what we get by blindly summing Feynman diagrams. . “almost. This will take a bit of work. Now.22). . an easier way to get rid of them is simply to divide by 0 |S| 0 . . . The ˜ formula will hold. the bare field ϕ0 (x) no longer creates mesons with unit probability. ϕI (xn ) exp −i 0 |T exp −i t+ t− t+ t− [HI − ρϕI ] |0 . . . but only when the Green function G is defined using renormalized fields ϕ instead of bare fields. . (VIII. that in an interacting theory. . First of all. as we have already discussed. . . .26) Since the answer to this question doesn’t depend on using bare fields ϕ0 or renormalized fields ϕ.” The problem will be. since as argued in the previous chapter the sum of vacuum bubbles is universal. it is easy to show that G(n) (x1 . this gives 0 |T exp −i ZF [ρ] = (n) t± →±∞ lim t+ t− [HI − ρ(x)ϕI (x)] | 0 t+ t− . we do n functional derivatives with respect to ρ and then set ρ = 0: (n) GF (x1 . First we will answer the first question: Is G(n) = GF ?9 Our answer will be similar to the derivation of Wick’s theorem on pages 82-87 of Peskin & Schroeder. (VIII. we will neglect this distinction in the following discussion. xn ) = t± →±∞ lim 0 |T ϕI (x1 ) . > tn 9 (VIII. . using Dyson’s formula just as we did at the end of the last section. .24) 0 |T exp −i HI | 0 To get GF (x1 .

t− →∞ t− →∞ t− →∞ (VIII. ϕH (xn )UI (0. ti )ϕI (xi )UI (ti .31) Next. we can trivially convert the evolution operator to the Schr¨dinger picture. and get rid of those intermediate U ’s: GF (x1 . not a discrete sum. . We’re almost there. 0 |UI (t+ . tb ). since H0 | 0 = 0. The sum over eigenstates is actually a continuous integral. and refer to the mess to the left of it as some fixed state Ψ |. a lemma . First of all. Now. the sum (integral) on the right is zero. t2 )ϕI (x2 ) . (VIII. . . UI (0. and we have used the fact that H| Ω = 0 and H| n = En | n . . This next part is the important one. t− )| 0 (VIII. .the Riemann-Lebesgue lemma) which states that as long as Ψ| n n| 0 is a continuous function. t− )| 0 Now. ϕI (xn )UI (tn . H. the integrand oscillates more and more wildly. As t− → −∞.  t− →∞  lim Ψ |U (0. 0)ϕH (x1 )ϕH (x2 ) . (VIII. 0)† ϕI (xi )UI (ti . everywhere that UI (ta .33) . 0)UI (0. . . we can express the time ordering in GF as GF (x1 . t1 )ϕI (x1 )UI (t1 . t− ) | Ω Ω | + n=0 t− →−∞ | n n | | 0 (VIII. we can drop the T -ordering symbol from G(n) (x1 .114 In this case. rewrite it as UI (ta . ϕH (xi ) = UI (ti . 0) = UI (0. . . . since UI (tb . xn ). The Riemann-Lebesgue lemma may be stated as follows: for any “nice” function f (x). t− )| 0 . 0)UI (0.32) = Ψ| Ω Ω |0 + lim e n=0 iEn t− Ψ| n n| 0 where the sum is over all eigenstates of the full Hamiltonian except the vacuum. tb ) appears. . t− )| 0 . and in fact there is a theorem (or rather.29) t± →±∞ lim 0 |UI (t+ .27) is the usual time evolution operator. t− )| 0 = lim Ψ |UI (0. . t− )| 0 . . where the En ’s are the energies of the excited states.30) Let’s concentrate on the right hand end of the expression. t− )| 0 = t− →∞ lim Ψ |U (0. o lim Ψ |UI (0. xn ) = (n) (VIII.28) 0 |UI (t+ . t− )| 0 (in both the numerator and denominator). t− ) exp(iH0 t− )| 0 = lim Ψ |U (0. 0) to convert everything to Heisenberg fields. ta ) = T exp −i tb ta (n) d4 x HI (VIII. . . insert a complete set of eigenstates of the full Hamiltonian. and then use the relation between Heisenberg and Interaction fields. xn ) = (n) t± →±∞ lim 0 |UI (t+ . b µ→∞ a lim f (x) sin µx cos µx = 0. .

as shown in Fig. VIII. . . . .8 1 FIG. . the contributions from infinitesimally close states destructively interfere. xn ) = (n) 0| Ω Ω |ϕH (x1 ) . 0)| Ψ | 0 = 0| Ω Ω |Ψ (VIII. It is quite easy to see the graphically. . xn ). . what the lemma is telling you is that if you start out with any given state in some fixed region and wait long enough. So there is now no longer to distinguish between the sum of diagrams and the real Green functions. . . . Physically.2 0. . . the only trace of it that will remain is its (true) vacuum component. A similar argument shows that lim 0 |UI (t+ . . ls |S − 1| k1 . kr = . . The LSZ Reduction Formula We now turn to the second question: Are S matrix elements obtained from Green’s functions in the same way as before? By introducing a turning on and off function.34) t+ →∞ and applying this to the numerator and denominator of Eq. (VIII. VIII.6 The Riemann-Lebesgue lemma: f (x) multiplied by a rapidly oscillating function integrates to zero in the limit that the frequency of oscillation becomes infinite.6 0. ϕH (xn )| Ω Ω |0 0| Ω Ω |Ω Ω |0 (VIII. . So we’re essentially done. we were able to show that l1 . .4 0. . C.30) we find GF (x1 .6. All the other one and multiparticle components will have gone away: as can be seen from the figure. .115 10 5 f (x) 0 -5 f (x) sin x -10 0 0.35) = G(n) (x1 .

. (VIII. So FROM NOW ON all fields will be in the Heisenberg representation (no more interaction picture). in an interacting theory ϕ0 may also develop a vacuum expectation value.36) The real world does not have a turning on and off function. For free field theory. −ls . (VIII. we get from Feynman diagrams) which we will derive is called the LSZ reduction formula. relativistically normalized H| k = k 2 + µ 2 | k ≡ ωk | k .instead. the fields we have been discussion have actually been the bare fields ϕ0 which we discussed at the end of the last chapter. we no longer need to make any reference to free Hamiltonia. in terms of ϕ0 . ϕ(x). we have really been talking about bare Green functions. Is this formula correct? The answer is “almost. which we will now denote G0 (n): G0 (x1 . interaction picture fields. (VIII. bare vacua. (n) (VIII. etc. however. ϕ0 (xn )| Ω . xn ) = T Ω |ϕ0 (x1 ) . . |0 ≡ |Ω .38) We have actually been a bit cavalier with notation in this chapter. . . .39) The reason the answer to our question is “almost” is because the bare field ϕ0 which does not have quite the right properties to create and annihilate mesons. k |k = (2π)3 2ωk δ (3) (k − k ).40) .116 2 la − µ2 i a=1 s 2 kb − µ2 ˜ (r+s) G (−l1 . k1 . Thus. k |ϕ(x)| Ω = k |eiP ·x ϕ(0)e−iP ·x | Ω = eik·x k |ϕ(0)| Ω . In particular. and states will refer to eigenstates of the full Hamiltonian (although for the rest of this section we will continue to denote the vacuum by | Ω to avoid confusion) ϕ(x) ≡ ϕH (x).37) The physical one-meson states in the theory are now the complete one meson states. . We correct for these problems by defining a renormalized field. For interacting fields. i b=1 r (VIII. . it is normalized to obey the canonical commutation relations. as we just showed. Ω |ϕ(x)| Ω = 0. . kr ). where the amplitude to create a meson from the vacuum has higher order perturbative corrections. as we have discussed. . . Since its derivation doesn’t require resorting to perturbation theory. . these two properties were equivalent. these two properties are incompatible. . . By translational invariance. .” The correct relation between S matrix elements (what we want) and Green functions (what. it is not normalized to create a one particle state from the vacuum with a standard amplitude . Furthermore.

. for all different incoming multiparticle states created by ϕ(x).42) We can now state the LSZ (Lehmann-Symanzik-Zimmermann) reduction formula: Define the renormalized Green functions G(n) . which is normalized to have a standard amplitude to create one meson. the factors of G in ˜ Eq. Naı¨ely. ϕ(xn )) | Ω (VIII. S matrix elements are given by l1 . It is some number. In terms of renormalized Green functions. you can see that k |ϕ(0)| Ω is independent of k. . (VIII. and a vanishing VEV (vacuum expectation value) ϕ(x) ≡ Z 1/2 (ϕ0 (x) − Ω |ϕ0 (0)| Ω ) Ω |ϕ(0)| Ω = 0. . k1 .almost.1) should be G0 . . . ls |S − 1| k1 . . (VIII. these additional states can all be arranged to oscillate away via the Riemann-Lebesgue lemma. 1. However.43) ˜ and their Fourier transforms. . k |ϕ(x)| Ω = eik·x . and only in free field theory will it equal 1. this is perhaps not so surprising. and that these do not pollute the relation between Green functions and S-matrix elements. Proof of the LSZ Reduction Formula The proof can be broken up into three parts. as we shall show. I will show you how to construct localized wave packets. . Given that it is the ϕ which create normalized meson states from the vacuum. ϕ. you might v think that the Green function would be related to a sum of S-matrix elements. . .44) . The only difference is that the S matrix ˜ is related to Green functions of the renormalized fields . G(n) . . kr s 2 2 la − µ2 r kb − µ2 ˜ (r+s) = G (−l1 . (VIII. −ls . G(n) (x1 . .in our new notation. . . which for historical reasons is denoted Z 1/2 (and traditionally called the “wave function renormalization”). xn ) ≡ Ω |T (ϕ(x1 ) . what we had before. . What is more surprising is that even the renormalized field ϕ(x) creates a whole spectrum of multiparticle states from the vacuum as well. The wave packet will have multiparticle as well as single particle . much as in the last section. but not quite. . kr ) i i a=1 b=1 That’s it . Z 1/2 ≡ k |ϕ(0)| Ω .117 By Lorentz invariance. .41) We now can define a new field. . .(VIII. . In the first part. .

it trivially satisfies Ω |ϕf (t)| Ω = 0 and has the correct amplitude to produce a single particle state | f : k |ϕf (t)| Ω =i =i =i d3 x d3 x d3 k F (k ) k | ϕ(x)∂0 e−ik ·x − e−ik ·x ∂0 ϕ(x) | Ω (2π)2 2ωk d3 k F (k ) −iωk e−ik ·x − e−ik ·x ∂0 k |ϕ(x)| Ω (2π)2 2ωk d3 x ei(k −k)·x e−i(ωk −ωk )t (VIII. t)∂0 ϕ(x.47) This is precisely the operator which makes single particle wave packets. f (x) ≡ d3 k F (k)e−ik·x . t) − f (x.49) . 1. satisfying the Klein-Gordon equation with negative frequency.45) where F (k) = k |f is the momentum space wave function of | f .46) Note that as we approach plane wave states. the multiparticle components will be set up to oscillate away after a long time. First of all. | v → | k . to derive the LSZ formula. t)) . (2π)3 2ωk (VIII. however. Now. define the following odd-looking operator which is only a function of the time. k0 = ωk . so the operators carry the time dependence) ϕf (t) ≡ i d3 x (ϕ(x. In the third part of the proof.48) d3 k F (k )(−iωk − iωk ) (2π)2 2ωk = F (k) (VIII. Associate with each F a position space function. (2 + µ2 )f (x) = 0. f (x) → e−ik·x . t (recall again that we are working in the Heisenberg representation. In the second part of the proof. these will be called in and out states. t)∂0 f (x. How to make a wave packet Let us define a wave packet | f as follows: |f = d3 k F (k)| k (2π)3 2ωk (VIII. (VIII.118 components. we massage this expression and take the limit in which the wave packets are plane waves. I will wave my hands vigorously and discuss the creation of multiparticle states in which the particles are well separated in the far past or future. and we will find a simple expression for the S matrix in terms of the operators which create wave packets.

thus.52) Proceeding much as before. A similar derivation. (VIII. which is an eigenvalue of the momentum operator: P µ | n = pµ | n .119 where we have used d3 x ei(k −k)·x = (2π)δ (3) (k − k ) and we note that the phase factor e−i(ωk −ωk )t (VIII. Note that this result is independent of time. with one crucial minus sign difference (so that the factors of ωk and ωk cancel instead of adding). ϕf (t) behaves as a creation operator for wave packets. yields Ω |ϕf (t)k = 0. as far as the zero and single particle states are concerned.53) (VIII.54) Note that we haven’t had to use any information about n |ϕ(x)| Ω beyond that given by Lorentz invariance. n n |ϕ(0)| Ω (VIII. Now we will see that in the limit t → ±∞ all the other states created by ϕf (t) oscillate away.50) becomes one once the δ function constraint is imposed (this will change when we consider multiparticle states). Consider the multiparticle state | n . n (VIII. let us calculate the amplitude for ϕf (t) to make this state from the vacuum: n |ϕf (t)| Ω =i =i =i =i d3 x d3 x d3 x d3 k F (k ) n | ϕ(x)∂0 e−ik ·x − e−ik ·x ∂0 ϕ(x) | Ω (2π)2 2ωk d3 k F (k ) −iωk e−ik ·x − e−ik ·x ∂0 n |ϕ(x)| Ω (2π)2 2ωk d3 k F (k ) −iωk e−ik ·x − e−ik ·x ∂0 eipn ·x n |ϕ(0)| Ω (2π)2 2ωk −p0 )t n d3 k F (k )(−iωk − ip0 ) d3 x ei(k −pn )·x e−i(ωk n (2π)2 2ωk ωp + p0 0 n = n F (pn )e−i(ωpn −pn )t n |ϕ(0)| Ω 2ωpn where ωpn = p2 + µ2 .51) Thus. we haven’t had to know anything about the amplitude to create multiparticle .

How to make widely separated wave packets The results of the last section were rigorous (or at least.120 states from the vacuum. can have p = 0 if the particles are moving back-to-back. ϕf (t) acts on the true vacuum and creates states with one. as already shown. for example.53). By contrast. that is. Thus. we find that at t = 0 the only surviving components are the single-particle states. n (VIII. for the n single particle state the oscillating phase vanishes. which make up a localized wave packet. . we find  t→±∞ lim ψ |ϕf (t)| Ω = t→±∞ lim  ψ| Ω Ω |ϕf (t)| Ω d3 k ψ |k k |ϕf (t)| Ω + ψ |n n |ϕf (t)| Ω  (2π)3 2ωk n=0. So now we have the familiar Riemann-Lebesgue argument: take some fixed state | ψ . so that only these states survived after infinite time. states which in the far past or far future look like well separated wavepackets. We now wish to construct multiparticle states of interest to scattering problems. and consider ψ |ϕf (t)| Ω as t → ±∞. could be made so without a lot of work). two. In contrast. Similarly. The crucial point is the existence of the phase factor e−i(ωpn −pn )t in Eq. (VIII..57) t→±∞ for any fixed state | ψ . while the energy of the state p0 can vary between 2µ and n ∞. and the observation that for a multiparticle state with a finite mass. Thus. whereas it can never vanish for multiparticle states. in this section we will wave our hands violently and rely on physical arguments. n particles.55) 0 This should be easy to convince yourself of: a two particle state.. Taking the inner product of this state with any fixed state.56)  + = ψ| f + 0 where we have used the fact that the multiparticle sum vanishes by the Riemann-Lebesgue lemma.1 (VIII. Inserting a complete set of states. ωpn < p0 . The physical picture we . We have a similar interpretation as before: in the t → −∞ limit. one can show that lim Ω |ϕf (t)| ψ = 0 (VIII. our cunning choice of ϕf (t) was arranged so that the oscillating phases cancelled only for the single particle state. a single particle state with p = 0 can only have p0 = ωp = µ < p0 . 2.

f |f . d4 xn f4 (x4 )f3 (x3 )f2 (x2 )f1 (x1 ) (2r + µ2 ) Ω |T ϕ(x1 ) . the S matrix is just the inner product of a given “in” state with another given “out” state. f2 out (in) . k2 = d4 x1 . 3 4 1 2 3 4 1 2 Thus. we have shown that f3 . (VIII. f4 |S| f1 . . k4 |S − 1| k1 . . but we can do that with a bit of massaging. f . out f . in the distant past and future they correspond to widely separated wave packets. . we have k3 .59) t4 →∞ t3 →∞ t2 →−∞ t1 →−∞ Ω |ϕf4 † (t4 )ϕf3 † (t3 )ϕf2 (t2 )ϕf1 (t1 )| Ω . i . However.58) t→∞ (−∞) Now by definition.60): we have written an expression for S matrix elements in terms of a weighted integral over vacuum expectation values of Heisenberg fields. (VIII. If we take the limit in which the wave packets become plane wave states. . it doesn’t look much like the LSZ reduction formula yet. d4 xn eik3 ·x3 +ik4 ·x4 −ik2 ·x2 −ik1 ·x1 ×i4 r (2r + µ2 )G(4) (x1 . | fi → | ki . (VIII. ϕ(x4 )| Ω . since the first wavepacket is arbitrarily far away. . . −k3 . Massaging the resulting expression In principle. f2 = lim lim lim lim (VIII. but it’s not. f2 = × i4 r ∗ ∗ d4 x1 .60) 3.62) = r 2 kr − µ2 ˜ (4) G (k1 . we have achieved our goal in Eq.61) This looks messy and unfamiliar. . (VIII. Let us denote states created by the action of ϕf2 (t) on | f1 in the distant past as “in” states. fi (x) → eiki ·xi . f4 |S − 1| f1 . Then we have lim ψ |ϕf2 (t2 )| f1 = | f1 . f in = f . f |S| f .121 will rely upon is that if F1 (k) and F2 (k) do not have common support. and states created by the action of ϕf2 (t) on | f1 in the far future as “out” states. . . it shouldn’t matter if this state is the vacuum state or the state | f1 . . −k4 ). What we will show is the following f3 . k2 . Then when ϕf2 (t) acts on a state in the far past or future. x4 ) (VIII.

it is the latest. thus. Similarly. we obtain RHS = lim − lim t1 →−∞ t4 →+∞ t4 →−∞ t1 →+∞ t3 →+∞ lim − lim t3 →−∞ t2 →−∞ lim − lim t2 →+∞ × lim − lim Ω |T ϕf1 (x1 )ϕf2 (x2 )ϕf3 † (x3 )ϕf4 † (x4 )| Ω . we get zero. f4 |ϕf2 (t2 )ϕf1 (t1 )| Ω . we find RHS = lim − lim t2 →+∞ t1 →−∞ t1 →+∞ out f .122 which is precisely the LSZ formula.47). (VIII. Next.65) Now we can evaluate the limits one by one. Similarly.61).64) t→+∞ t→−∞ Note the difference in the signs of the limits. Thus. and acts on the vacuum state on the left.57). to give f4 |. Note that we have taken the plane wave limit after taking the limit in which the limit t → ±∞ required to define the in and out states. When t4 → +∞.66) − lim out f3 .63) where we have integrated once by parts. it is the earliest (in the order of limits which we have taken). (VIII. (VIII. we can show i d4 x f ∗ (x)(2 + µ2 )A(x) = lim − lim Af † (t). (VIII. (VIII. (VIII. f4 |. Doing this for each of the xi ’s. according to the complex conjugate of Eq. and so acts on the vacuum. only the t3 → ∞ limit contributes. when t4 → −∞. i d4 x f (x)(2 + µ2 )A(x) = i = i = = − = 2 d4 x f (x)∂0 A(x) + A(x)(− 2 + µ2 )f (x) 2 2 d4 x f (x)∂0 A(x) − A(x)∂0 f (x) dt ∂0 d3 x i (f (x)∂0 A(x) − A(x)∂0 f (x)) dt ∂0 Af (t) lim − lim t→+∞ t→−∞ Af (t) (VIII. First of all. let us prove a useful result: for an arbitrary interacting field A and function f (x) satisfying the Klein-Gordon equation and vanishing as |x| → ∞. f |ϕf1 (t )| f 3 4 1 2 . and Af (t) is defined as in Eq. Before showing this. creating the state out f3 . taking the two limits of t2 . in this order of limits even the plane wave in and out states are widely separated. We can now use these relations to convert factors of (2 + µ2 )ϕ(x) to ϕf (t) on the RHS of Eq.

f2 in − out f3 .69) (VIII.the simplest relation is Eq. The proof relied only on the properties Ω |ϕ(0)| Ω = 0. since some complicated nonrenormalizable Lagrangians (VIII.56)) that lim ψ |ϕf1 (t1 )| Ω = lim ψ |ϕf1 (t1 )| Ω = ψ |f1 . but there are a few important things to remember: 1.we’ve proved the reduction formula.123 Unfortunately. and so it’s not clear what the second term is.67) Then taking the t1 limits. Physically. but the Green functions of any field ϕ which satisfies the requirements (VIII. f4 |ϕf2 (t2 ).70) No other properties of ϕ were assumed.42). (VIII. f4 |f1 .71) .68) t1 →∞ t1 →−∞ So that’s it . ϕ was not assumed to have any particular relation to the bare field ϕ0 which appears in the Lagrangian . In particular. ϕ(0) and ϕ(0) only differ in their vacuum to multiparticle state ˜ matrix elements. (VIII. f4 |S − 1| f1 . k |ϕ(0)| Ω = 1.70) will give the correct S-matrix elements. f2 out − ψ |f1 + ψ |f1 = f3 . One appropriately renormalized. (VIII. we find RHS = out f3 . we don’t know how ϕf2 (t2 → ∞) acts on a multi-meson out state. Note that we have used the fact (from Eq. this has a very useful consequence: you can always make a nonlinear field redefinition for any field in a Lagrangian. For example. It was a bit involved. In some cases this is quite convenient. but that is irrelevant to the physics). ϕ(x) = ϕ(x) + 1 gϕ(x)2 ˜ 2 is a perfectly good field to use in the reduction formula. Practically. t2 →∞ (VIII. Let’s parameterize our ignorance and define ψ | ≡ lim out f3 . f2 as required. and it doesn’t change the value of S-matrix elements (although it will change the Green functions off shell. f4 |f1 .the multiparticle states created by the field are irrelevant. (VIII. this is again because of the Riemann-Lebesgue destructive interference. But this difference just oscillates away .

in QCD mesons have the quantum numbers of quarkantiquark pairs.70). 2. . So if q(x) is a quark field and q(x) an antiquark field.1) e G (k1 . One could use perturbation theory to calculate the scattering amplitudes of an e+ e− pair to the various states of positronium.74) where the renormalized product of fields satisfies meson |(qq)R (0)| Ω = 1. . q(x)q(x) should have a nonvanishing matrix elements to make a meson. the matrix element can be calculated in terms of Feynman diagrams: out k . which can then be renormalized to satisfy Eq. . For example. . A simple example of this is given on the first problem set. . This is a good result to remember. Most of these papers are idiotic . (VIII. A lot of papers have been written (even in the past few years) which claim that some particular field is the “correct” one to use in a given problem. kn .1) (x1 . . which had no way of dealing with bound states). . k |A(x)| Ω = 1 n 2 d4 k ik·x n kr − µ2 ˜ (2. nobody can calculate this T -product since perturbation theory fails for the strong interactions. x) is the Green function with n ϕ fields and one A field. . ϕ(xn )A(x)| Ω for any field A(x). but that’s not a problem of the formalism (as opposed to the turning on and off function. .the authors’ pet form of the Lagrangian has been obtained by a simple nonlinear field redefinition from the standard form. there is no problem in obtaining matrix elements for scattering bound states you just need a field with some overlap with the bound state. . We can also use the same methods as above to derive expressions for matrix elements of fields between in and out states (remembering that the S matrix is just the matrix element of the unit operator between in and out states). . 3. . . . k) (2π)4 i r=1 (VIII. .. . . In principle. . . out k .+ikn ·xn (VIII.124 may take particularly simple forms after an appropriate field redefinition. xn .73) where G(2.. . . k |A(x)| Ω = 1 n ×in r d4 x1 . Substituting the expression for the Fourier transformed Green function. if only to save a few trees. Of course. For example. and so is guaranteed to give the same physics. d4 xn eik1 ·x1 +. So “all” you need to calculate for meson-meson scattering from QCD is T Ω |(qq)R (x1 )(qq)R (x2 )(qq)R (x3 )(qq)R (x4 )| Ω (VIII.72) 2r + µ2 T Ω |ϕ(x1 ) .

The matrices D(Λ) form an n-dimensional representation of the Lorentz group: if Λ1 and Λ2 define two Lorentz transformations. Sy and Sz are given by 1 1 Sx = 2 σx .2) If φa has n components. φ could have a more complicated transformation law.7) . φ transforms according to φ(x) → φ (x) = φ(Λ−1 x).1) This simply states that the field itself does not transform at all. φa (x) → Dab (Λ)φb (Λ−1 x). we could have four fields φµ . σy and σz are the Pauli matrices 0 σx = 1 1 0 . Dab (Λ1 )Dbc (Λ2 ) = Dac (Λ1 Λ2 ). (IX. Recall that since φ is a scalar.. σy = 0 −i i 0 .4.5) (IX. (IX. Transformation Properties So far we have only looked at the theory of an interacting scalar field φ(x). we already know how such objects transform under rotations. the identity matrix. SPIN 1/2 FIELDS A. b (IX. D(Λ−1 ) = D(Λ)−1 .4) In addition. σz = 1 0 0 −1 .125 IX. a field will transform in some well-defined way under the Lorentz group. Sz = 1 σz 2 | ψ↑ | ψ↓ . a subgroup of the Lorentz group. (IX. which make up the components of a 4-vector. under a Lorentz transformation xµ → x µ = Λµ ν xν .6) where σx . Sy = 2 σy . In this case. µ = 1. In general. A spin 1/2 state | ψ has two components: |ψ = The spin operators Sx . the value of the field at the coordinate x in the new frame is the same as the field at that same point in the old frame. From our previous experience in quantum mechanics. (IX.3) (IX. Dab (Λ) is an n × n matrix. φµ will transform under a Lorentz transformation as φµ (x) → φ µ (x) = Λµ ν φν (Λ−1 x). For example. We are interested in describing particles of spin 1/2. and D(1) = I. In general.

. we will simply introduce a nice trick for obtaining the spinor representation from the vector representation. Diu and Lalo¨. We could find all possible representations of the group. But since we have only a small amount of time. vector (both of which we already understand) and spinor representations. but it will serve our purposes.. . Chapter VI (especially Come plement BVI ). θ) = e−iJ·ˆθ e 10 (IX. In general. Hence. θ)| ψ e where the rotation operator e UR (ˆ.11) where R is the rotation matrix for vectors.9) is unitary. Therefore. not just the rotation group. Cohen-Tannoudji. The most satisfying way to do this would be to pause for a moment from field theory and develop the theory of representations of the Lorentz group. This is only equal to the exponential of the entries in the matrix if M is diagonal. a state with total angular momentum 1/2 transforms under this rotation as11 e | ψ → e−iσ·ˆθ/2 | ψ (IX. and we are suppressing spinor indices.. and then finally restrict ourselves to the spinor representation. Quantum Mechanics. Recall that the exponential of a matrix M is defined by the power series eM = 1 + M + M 2 /2! + M 3 /3! + .126 For a particle with no orbital angular momentum. In a relativistic theory. and since we are really not interested in any representations of the Lorentz group beside the scalar. for example. Volume 1. This is not generalizable to other representations. corresponding to particles of spin 0. a quantum field u which creates and annihilates spin 1/2 particles will transform under rotations according to † e u (x) = UR u(x)UR = e−iσ·ˆθ/2 u(R−1 x) (IX.8) (IX.10) so the matrices e e−iσ·ˆθ/2 form the spinor representation of the rotation group. we must determine the transformation properties of spinors under the full Lorentz group. rotations are generated by the angular momentum operator J: a general state | ψ transforms under a rotation about the e axis by an angle θ as ˆ | ψ → UR (ˆ. the total angular momentum J is just given by the spin operators. 1 and 1/2 respectively. 10 11 See.

The rotations are the group of transformations (x. where U is a two by two unitary matrix with unit determinant. y. so Eq.16) (IX.18) . (IX. It is easy to show by direct matrix multiplication that e U = e−iσ·ˆθ/2 (IX. y .13) so this is another way of writing the transformation law of a vector under rotations.12) is the most general form for a two by two traceless Hermitian matrix.15) x − iy −z . (IX. and a spinor is defined to be a two-component column vector which transforms under rotations through multiplication by U : u = U u. Re b = Re c and Im b = −Im c. Then X is also a traceless. Hermiticity requires Re a = Re d.17) is the appropriate transformation matrix for a rotation about the e axis. Consider the transformation X = U XU † .12) has eight independent components (two for each complex c d entry). z) → (x . (IX.14) † =X (IX. Im a = Im d = 0. We can assemble the components of a 3-vector into a two by two traceless Hermitian matrix z X= x + iy a A two by two complex matrix b x − iy −z = xi σi . Hermitian matrix TrX = TrU XU † = TrU † U X = TrX X † = U XU † so in general we can write it as z X = x + iy Since detX = detU detXdetU † = detX we have x 2 + y 2 + z 2 = x2 + y 2 + z 2 (IX. (IX. ˆ The U ’s by themselves form a two-dimensional represenation of the rotation group. reducing the number of independent components to four.127 Let us return to the rotation subgroup of the Lorentz group. so we do not include reflections). and tracelessness reduces this to three. z ) which leave r2 = x2 + y 2 + z 2 invariant (and which retain the handedness of the coordinates.

We can extend this construction to the whole connected Lorentz group. detX = detX . Then the vector transforms as (1. (IX. (IX. (IX. where γ = √ sinh φ = − γ 2 − 1 (IX. defined by cosh φ = γ.21) and so the transformation corresponds to a (proper12 ) Lorentz transformation.22) v 2 − 1 parameterizes the boost. 0) under a boost in the z direction. so t 2 − x 2 − y 2 − z 2 = t 2 − x2 − y 2 − z 2 (IX. Then under the transformation Eq.19) where Q is no longer required to be unitary. − γ 2 − 1ˆ).128 This agrees with our previous assertion. . Consider the transformation of the 4-vector (1.10). We note that the matrix Q has six independent parameters. sinh φ).20) x − iy t−z .24) It is straightforward to verify that in our matrix representation. 0) → (γ. (IX. so Q† = Qz ): z X = eφσz /2 Xeφσz /2 eφ/2 0 = 0 e−φ/2 12 1 0 0 1 eφ/2 0 0 e−φ/2 The proper or connected Lorentz transformations do not include reflections or time reversal.23) (IX. z It is convenient to introduce the parameter φ. but still det Q = 1. ˆ (1. Removing the tracelessness condition on X increases the number of free parameters by one. 0) → (cosh φ. which is the same as a proper Lorentz transformation (three independent rotations and three independent boosts). so it now takes the general form t+z X= x + iy Now consider the transformation X = QXQ† (IX. Any proper Lorentz transformation may be written as a product of a rotation and a boost. Eq. this boost corresponds to the transformation matrix Qz = exp(σz φ/2) (note that there is no i in the exponential.20). not unitary. Qz is Hermitian.

129 eφ = = 0 0 0 cosh φ − sinh φ (IX.28) for all transformations Λ.27) where we have used the fact that the Q’s form a representation. so do the matrices Q∗ (Λ). those which transform according to Q .29) There is no physics in a change of basis. However there may or may not ˜ be any physical difference between the two representations. if no such matrix S exists. Under a boost. If we have a set of matrices Q(Λ) which form a representation of a group. This is easy to verify. Two representations Q and Q are said to be equivalent if there is some unitary matrix S such that ˜ Q(Λ) = S Q(Λ)S † (IX. In general. and there is a physical difference between fields transforming according the two representations. This construction is not unique. in fact. For the Lorentz group. the group of unitary two by two matrices with unit determinant (including the rotation matrices U ) form a representation of the connected Lorentz group. as do the matrices SQ∗ (Λ)S † for some unitary matrix S. the new representation preserves the group multiplication rule: SQ∗ (Λ1 )S † SQ∗ (Λ2 )S † = S [Q(Λ1 )Q(Λ2 )]∗ S † = SQ∗ (Λ1 Λ2 )S † (IX. Therefore there are two different types of spinor fields we can define. We will see a practical illustration of this shortly. so a set of fields transforming under Q are physically ˜ equivalent to a set of fields transforming under Q. (IX. then the two representations are inequivalent. On the other hand.26) The Q’s. (IX. the two representations Q and Q∗ are. as required. This is physically sensible because if two representations are equivalent. a spinor transforms as u → Qu. I can always transform an object u transforming under one representation to one transforming under the other representation by performing the change of basis u → Su. inequivalent. you can verify that a boost in the e ˆ direction is given by e Q = eσ·ˆφ/2 . for example.25) 0 e−φ cosh φ + sinh φ and so t = cosh φ and z = sinh φ.

130 and those which transform according to Q∗ . However, for the rotation subgroup, U and U ∗ can be shown to be equivalent representations, with S = iσ2 : U (R) = iσ2 U ∗ (R)(iσ2 )† = 0 −1 1 0 U ∗ (R) 0 −1 (IX.30) 1 0

for all rotation matrices U (R) = exp(−iσ · eθ/2). This is why we never encountered two different ˆ types of spinors when dealing only with rotations. However, for boosts, it can be shown that
e iσ2 e−σ·ˆφ/2 ∗ e e (iσ2 )† = eσ·ˆφ/2 = e−σ·ˆφ/2


and so the two representations Q and SQ∗ S † of the Lorentz group are not equivalent. Thus we can define two types of spinors which we shall denote by u+ and u− . They transform in the same way under rotations,
e u± → e−iσ·ˆθ/2 u±


but differently under boosts
e u± → e±σ·ˆφ/2 u± .


In group theory jargon, u+ is said to transform according to the D(0,1/2) representation of the Lorentz group and u− transforms according to the D(1/2,0) representation. In order to construct Lorentz invariant Lagrangians which are bilinear in the fields, we shall need to know how terms bilinear in the u’s transform. Not surprisingly, since in some sense the spinors were the “square roots” of the vectors, we can construct four-vectors from pairs of spinors. First consider the bilinear u† u+ . Under a Lorentz transformation, + u† u+ → u† Q† Qu+ . + + (IX.34)

If Q is purely a rotation, Q† = Q−1 (Q is unitary) so u† u+ → u† u+ . Therefore it is a scalar under + + rotations (but not under Lorentz boosts). The three components of u† σu+ ≡ (u† σx u+ , u† σy u+ , u† σz u+ ) + + + + form a three-vector under rotations (I have left this as an exercise for you to show). these together, you should also be able to show that the four components of V µ = (u† u+ , u† σu+ ) + + form a four-vector. A similar construction shows that the components of W µ = (u† u− , −u† σu− ) − − transform as a four vector under a proper Lorentz transformation. (IX.37) (IX.36) (IX.35) Putting

B. The Weyl Lagrangian

We will now promote our spinors to fields, that is spinor functions of space and time, transforming under Lorentz transformations according to u+ (x) → UΛ (Λ)† u+ (x)UΛ (Λ) = D(Λ)u+ (Λ−1 x) (IX.38)

and similarly for u− , where UΛ (Λ) is the unitary operator corresponding to Lorentz transformations, UΛ (Λ)| k = | Λk . (IX.39)

Note that the u’s are complex fields. We can now construct a Lagrangian for u+ , keeping in mind the following restrictions: 1. The action S should be real. This is because we want just as many field equations as there are fields. By breaking up any complex fields into their real and imaginary parts, we can always think of S as being a function only of a number of real fields, say N of them. If S were complex, with independent real and imaginary parts, then the real and imaginary parts of the resulting Euler-Lagrange equations would yield 2N field equations for N fields, too many to be satisfied except in special cases (see Weinberg, The Quantum Theory of Fields, Vol. I, pg. 300). 2. L should be bilinear in the fields, to produce a linear equation of motion. In the absence of interactions, we want u+ to be a free field with plane wave solutions, which requires linear equations of motion. 3. L should be invariant under the internal symmetry transformation u+ → e−iλ u+ , u† → + eiλ u† . This is because we want to have some contact with the real world, and all observed + fermions carry some conserved charge (like baryon number or lepton number). We’ve already seen that bilinears in u+ and u† form the components of a four-vector; to make + this a scalar we have to contract with another vector. The only other vector we have at our disposal is the derivative ∂µ . Hence, the simplest Lagrangian we can write down satisfying the above requirements is L = i u† ∂0 u+ + u† σ · + + u+ . (IX.40)

The i in front is required for the action to be real, which you can verify by integrating by parts. The sign of L is not fixed; we will take it at this point to be +. We will see later on that this

132 theory has problems with positivity of the energy no matter what sign we choose, so we will defer the discussion to a later section. The Lagrangian Eq. (IX.40) is called the Weyl Lagrangian. We can get the equation of motion by varying with respect to u† : + Πµ† =

∂L ∂(∂µ u† ) +



so the equation of motion is ∂L ∂u† + Multiplying this equation by ∂0 − σ · = 0 ⇒ (∂0 + σ · )u+ = 0. (IX.42)

and using the relation σi σj = i
k ijk σk

+ δij


gives us (σ ·

)2 =


and so
2 ∂0 − 2

u+ = 2u+ = 0.


Remember that u+ is a column vector, so both components of u+ obey the Klein-Gordon equation for a massless field. Defining the energy to be positive, k0 = there are two solutions for u+ (x): u+ (x) = u+ e−ik·x , u+ (x) = v+ eik·x (IX.46) |k|2 (IX.45)

where u+ and v+ are constant 2 component spinors. Based on our previous experience with complex fields, we expect that when we quantize the theory, u+ will multiply an annihilation operator for a particle and v+ will multiply a creation operator for an antiparticle. Substituting the positive energy solution into the Weyl equation gives (∂0 + σ · and so (k0 − σ · k)u+ = 0. (IX.48) )u+ (x) = (−ik0 + iσ · k)u+ (x) = 0 (IX.47)

0.56) 0 and so λ = 1/2. θ)| k = e−iλθ 0 |u+ (0)| k ∝ e−iλθ z z (IX.57) . 0 (IX.55) 1 (IX.52) It is straightforward to find λ. k = k0 z . Then we have (1 − σz )u+ = 0. k = (0. Consider a state | k moving in the positive z direction. (IX.51) 1 (IX. 0 (IX. † z z UR (ˆ.53) and therefore † 0 |UR u+ (0)UR | k = 0 |e−iσz θ/2 u+ (0)| k 1 ∝ e−iσz θ/2 = e−iθ/2 0 1 . in the quantum theory we expect that u+ will multiply an annihilation operator. or ˆ ˆ u+ ∝ 1 . θ)u+ (0)UR (ˆ. Then we expect that 0 |u+ (x)| k ∝ e−ik·x or 0 |u+ (0)| k ∝ 1 . k0 > 0. Since u is a spinor field. (IX. 0 (IX. kz ).49) What does this tell us about the states of the quantum theory? Well. θ)u+ (0)UR (ˆ. Jz : Jz | k = λ| k . we know how it transforms under rotations about the z axis by an angle θ. θ) = e−iσz θ/2 u+ (0) (IX.54) But since UR | k = e−iλθ | k UR | 0 = | 0 we also have † 0 |UR (ˆ.133 Consider k to be in the z direction.50) 0 It will turn out that this state is in an eigenstate of the z component of angular momentum.

neutrinos are massless fermions. the particle’s 3-momentum is in the opposite direction but its spin is unchanged. Similarly. and so neutrinos are described by the u− field. the Weyl Lagrangian violates charge conjugation invariance. Thus. Thus. while there are no corresponding states with particles (antiparticles) carrying spin antiparallel (parallel) to the direction of motion (since there is no corresponding solution to the equations of motion for the fields). spin along the direction of motion is only a good quantum number for massless particles. Therefore. we can show that v+ ∝ 1 .134 Therefore in the quantum theory the annihilation operator multiplying u+ will annihilate states with angular momentum 1/2 along the direction of motion.” while if the helicity is negative (antiparallel to the direction of motion) it is “left-handed. since charge conjugation will turn a left-handed neutrino . we would find that it annihilates left-handed particles and creates right-handed particles. Clearly the Weyl Lagrangian. In this frame. violates parity. In fact. Similarly. Since the field operator therefore changes the helicity of the state it acts on by −1/2. since for a massive particle it is always possible to boost to a frame going faster than the particle. it is not consistent for a massive particle to have only one spin state. For the u− (x) field. while the Weyl Lagrangian does not describe the familiar massive fermions like electrons or nucleons. A particle with positive helicity (along the direction of motion) is referred to as “right-handed. These states don’t look much like electrons. if the spin was parallel to the direction of motion in one frame. In fact. the quanta of this theory consist of particles carrying spin +1/2 in the direction of motion and antiparticles carrying spin −1/2 in the direction of motion. while its spin is unchanged. Thus parity interchanges left and right-handed particles. since electrons (or any other massive spin 1/2 particle) can have spin either parallel or antiparallel to the direction of motion. there is a particle in nature which exists only as left-handed particles and right-handed antiparticles: the neutrino. 0 and that v+ will multiply a creation operator that creates states with angular momentum -1/2 along the direction of motion.” Thus. in distinguishing right and left-handed particles. Right-handed neutrinos and left-handed antineutrinos have never been observed. As far as anyone has been able to measure. in the quantum theory we expect that u+ (x) will annihilate right-handed particles. it should also create left-handed antiparticles. and is usually called the helicity of the particle. Spin is usually reserved for massive particles to describe their angular momentum in the rest frame. it is antiparallel in the other. Under a parity transformation the three momentum of a particle flips sign.

Furthermore. the strong and electromagnetic interactions of electrons are observed to conserve parity (the weak interactions. it is easy to check explicitly that u† u− and u† u+ transform as scalars under Lorentz transformations. C. Thus we can define the action of parity on the u± fields to be P : u± (x. and as we shall see it describes massive spin 1/2 fields. so we have suggestibly called it + m. Therefore we can include + − the parity conserving term L = L0 − m u† u− + u† u+ + − = iu† (∂0 + σ · + )u+ + iu† (∂0 − σ · − )u− − m u† u− + u† u+ + − (IX. but it’s not what we set out to find. but conserves the product CP .60) The coupling multiplying the u† u− term has dimensions of mass. Therefore we would like to write down a free field theory of massive spin 1/2 particles which has a parity symmetry. We were really looking for a theory of electrons. (IX. (IX. t). We can again vary the fields and derive the equations of motion. we expect that the Weyl Lagrangian violates C and P separately. which does not exist. The Dirac Equation This is all very well.135 into a left-handed antineutrino.59) but this is nothing more than two decoupled massless spinors. both of which exist in the theory. although we haven’t quantized the theory to show this explicitly. This is the Dirac Lagrangian. In its current form it doesn’t look the way Dirac wrote it down. violate parity). The simplest Lagrangian is just L0 = iu† (∂0 + σ · + )u+ + iu† (∂0 − σ · − )u− (IX. which are certainly not massless. but we will be introducing some slick new notation shortly to put it in a more elegant form. However. However.58) A parity invariant theory must therefore have both types of spinors. which we shall study later. t) → u (−x. Thus. We find the coupled equations i(∂0 + σ · i(∂0 − σ · )u+ = mu− )u− = mu+ .61) . We have already argued that parity interchanges left and right-handed fields. the combined operation of CP will turn a left-handed neutrino into a right-handed antineutrino.

β= 1 0 0 1 (IX. If you prefer. t) → βψ(−x.66) ψ − mψ † βψ (IX.67) This is the Dirac equation. the Dirac Lagrangian is L = iψ † ∂0 ψ + iψ † α · where σ α= 0 0 −σ .65) u+ u− . Note that we can get the Dirac equation directly from Eq. t) and under a Lorentz boost e ψ → eα·ˆφ/2 ψ.136 Multiplying the first equation by ∂0 − σ · 2 (∂0 − 2 . a parity transformation is now P : ψ(x. (IX. We can group the two fields u+ and u− into a single four component “bispinor” field ψ: ψ≡ In terms of ψ. In terms of ψ.70) . (IX.69) (IX. you can leave the spinor indices explicit in this derivation. (IX.65) from the Euler-Lagrange equations for ψ: Πµ † = ψ ∂L =0 ∂(∂µ ψ † ) ψ − mβψ = 0.64) and each entry represents a two-by-two matrix. (IX.62) )u+ = −im(∂0 − σ · and so each of the components of u+ and u− obeys the massive Klein-Gordon equation (∂µ ∂ µ + m2 )u± (x) = 0.63) At this point we will introduce some notation to make life easier. we find )u− = −m2 u+ (IX. (IX. and so these are now matrix equations.68) ∂L = 0 ⇒ i∂0 ψ + α · ∂ψ † You just have to remember that ψ is now a four component column vector. (IX. The equation of motion is i(∂0 + α · )ψ = βmψ.

will be sufficient. Plane Wave Solutions to the Dirac Equation As in the Weyl equation. (IX. take the energy p0 to be positive. In this basis (the “Dirac” basis). β 2 = α1 = α2 = α3 = 1 (IX. This representation is not unique. We will need these solutions to canonically quantize the theory. ψ(x) = vp eip·x |p|2 + m2 . 1. Finally. which hold in any basis. B} is the anticommutator of A and B: {A. for example. we note that the components of (ψ † ψ. However. The Weyl basis turns out to be convenient for highly relativistic particles m E while the Dirac basis is convenient in the nonrelativistic limit m E. β= 0 1 1 0 . ψ † αψ) form a 4-vector. αi } = 0. Then we have (IX. they would still obey the anticommutations relations Eq.71) σ 0 where L = .137 Since u+ and u− transform the same way under rotations. However. (IX.76) . The anticommutation relations Eq. except the α’s and β would be different. {αi . B} = AB + BA.75) We could define other bases as well. as we shall see. in most situations we will never have to specify the basis.74) (IX. before proceeding to that let us finish our discussion of the plane wave solutions to the Dirac equation.72).72). 0 σ The α’s and β obey the relations 2 2 2 {β. However. (IX. have defined ψ by 1 ψ=√ 2 u+ + u− u+ − u− (IX.73) and all of these results would still hold. p0 = both positive and negative frequency solutions ψ(x) = up e−ip·x .72) where {A. We could. we find 0 α= σ 0 σ . Very shortly we will introduce some even more slick notation which will allow us to write all of our results in a Lorentz covariant form. ψ transforms under rotations as e R : ψ → eL·ˆθ ψ 1 2 (IX. αj } = 0 (i = j). since the plane wave solutions multiply the creation and annihilation operators in the quantum theory.

so E−m (r) α · e u0 ˆ 2m  (IX. cosh φ = γ = E/m and sinh φ = |p|/m. Instead of solving the Dirac equation for p = 0. Note that Mandl & Shaw do not include this factor in their definition of the plane wave states. The two solutions correspond to the two spin states.79) 0 (The factor of √ 0 2m in the normalization is a convention. u0 = 2m   . we get ˆ up = cosh Now.77) 0 −1 .82) (1 + cosh φ)/2 and sinh φ/2 =  up =  (r) E+m + 2m (IX.78) 0 and so two linearly independent solutions are   1   0     0 1 √ √     (1) (2) u0 = 2m   . Now we use our knowledge of the transformation properties of ψ to find the the plane wave solutions when p = 0. in both the Dirac and Weyl bases the spin operator is Sz = (1) (1) (2) (2) 1 2 σz 0 0 (IX.83) .     0 0     (IX.138 where up and vp are constant four component bispinors. 1 For definiteness.81) 2 where e = p/|p|. so we find   (IX. we find (p0 − α · p)up = βmup . both solutions are present for a massive field. cosh φ/2 = (r) φ φ (r) + α · e sinh ˆ u . Using αi = 1 and αi αj = −αj αi . we can just boost the coordinate system in the opposite direction e up = eα·ˆφ/2 u0 (r) (r) (IX.80) σz 1 so Sz up = + 2 up and Sz up = − 1 up . In the rest frame. As 2 we expected. p=0 and a   b   up = βup ⇒ up =     0   (IX. so β = 0 p0 = m.) Now. 2 2 0 (1 + sinh φ)/2. Substituting the first solution into the Dirac equation. we will work in the Dirac basis.

Time and space appear to be on a different footing. We find   0   0     0 0 √ √     (2) (1)   .84) 0 √ − E−m The case where p is not parallel to z is straightforward to compute from Eq. γ Matrices With all of these α’s and β’s.88) since the scalar is unaffected by Lorentz boosts.85) 0 √ E+m Notice that we have chosen our solutions to be orthonormal: u0 u0 = 2mδ rs . ˆ √   E+m  0     . v0 (r)† (s) (r)† (r)† (s) (r)† (s) v0 = 2mδ rs (IX. Therefore we can immediately write u0 βu0 = up βup = 2mδ rs v0 (r)† (r)† (s) (r)† (s) βv0 = vp (s) (r)† βvp = −2mδ rs (s) (IX. = √     p  E + m   0     (IX.     √      0 E+m    (1)  . although we know they’re not because L is a scalar. for p in the z direction. ˆ Similar arguments also allow us to find the solutions for the v’s. v = 2m   v0 = 2m     0 1 0     √ (1) vp 0 E−m   1 0    √    − E − m  0     (2) . u(2) =  up =  √    p   E − m 0    (IX. v =  .139 so in the Dirac basis we find. We can clean .86) βv0 = −2mδ rs (s) (IX.87) This second form is useful because we’ve already noted that ψ † βψ is a Lorentz scalar. the theory doesn’t look Lorentz covariant. D. v0 or u0 βu0 = 2mδ rs . (IX. which in the quantum theory we expect to multiply creation operators for antiparticles.83).

ψ1 αψ2 ) = (ψ 1 βψ2 .95) . It will also allow us to write down combinations of bispinors which transform in simple ways under Lorentz transformations. (IX. we know that the components of (ψ1 ψ2 .92) We can now use this technology to construct objects from the bispinors which transform in more complicated ways under the Lorentz group. ψ † = ψβ). by γ 0 ≡ β.93) (IX.90) (IX. The γ µ ’s are called the Dirac γ matrices. † We’ve already seen that for two bispinors ψ1 and ψ2 . For example. The index i on γ is a Lorentz index. It’s convenient to make use of this fact and define the Dirac Adjoint of a bispinor ψ ψ ≡ ψ † β.91) (Note that the label i on the α’s is not a Lorentz index and so there is no distinction between upper and lower indices on α. Since under a Lorentz transformation. Furthermore. where we have defined the Dirac adjoint of the operator D(Λ) D(Λ) = γ 0 D† (Λ)γ 0 .89) † † (note that since β 2 = 1. µ = 1. γ i ≡ βαi . Therefore ψ 1 ψ2 is a scalar: under a Lorentz transformation ψ 1 ψ2 → ψ 1 ψ2 (IX.) The components of the four vector are now simply written as ψ 1 γ µ ψ2 . (IX. ψ1 βψ2 is a Lorentz scalar. and so this equation defines γ µ with raised indices.94) (IX..140 things up a bit by introducing even more notation which makes everything manifestly Lorentz covariant. ψγ µ γ ν ψ transforms like a two index tensor: ψγ µ γ ν ψ → ψD(Λ)γ µ D(Λ)D(Λ)γ ν D(Λ)ψ = Λµ α Λν β ψγ α γ β ψ (IX. You will learn to know and love them. and so ψ = ψ † β → ψ † D† (Λ)β = ψ † ββD† (Λ)β ≡ ψD(Λ). Under a Lorentz transformation ψ → D(Λ)ψ. ψ 1 βαψ2 ) transform like the components of a four-vector. ψγ µ ψ → ψD(Λ)γ µ D(Λ)ψ = Λµ ν ψγ ν ψ we find that the γ matrices satisfy D(Λ)γ µ D(Λ) = Λµ ν γ ν . It’s convenient then to define the four matrices γ µ .4.

The Dirac Lagrangian may be written in a manifestly Lorentz invariant form // ψ(i∂ − m)ψ / and the Dirac equation is (i∂ − m)ψ = 0.100) (IX. the γ matrices all anticommute with one another.97) and aa = a2 .98) (IX. γ µ† = γµ = gµν γ ν = γ 0 γ µ γ 0 but they are self-Dirac adjoint (“self-bar”) γµ = γµ.141 (where we have used D(Λ)D(Λ) = 1. the Dirac equation / / / / / / implies that the plane wave solutions satisfy (p − m)up = 0 = (p + m)vp . (IX. we define a (“a-slash”) by / a = aµ γ µ . and (γ 0 )2 = −(γ 1 )2 = −(γ 2 )2 = −(γ 3 )2 = 1. / / (r) (r) (IX.99) Note that these are all four by four matrix equations. Another property of the γ matrices is that they are not all Hermitian. where we have suppressed matrix indices. / From the γ algebra it follows that a/ + /a = 2a · b /b b/ (IX. and the γ matrix algebra is simply a statement about matrix multiplication.101) (IX. up vp = 0 (r) (s) (r) (s) (r) (s) (IX.96) Thus.104) . which follows from the fact that ψψ is a scalar). not about quantum operators anticommuting. γ ν } = 2g µν . The commutation relations for the α’s and β may now be written in terms of the γ’s as {γ µ .102) (IX. Also. everything is still classical.103) and since i∂ up (x) = i(−ip)up (x) = pup (x) and i∂ vp (x) = i(ip)vp (x) = −pvp (x). The orthonormality conditions on the plane wave solutions are now up up = 2mδ rs = −v p vp . / (IX. For any four-vector aµ .

we next consider ψγ µ γ ν γ α γ β ψ.106) These will be very useful later on when we calculate cross sections. Bilinear Forms We have already seen that ψψ transforms like a scalar and that the components of ψγ µ ψ form a 4-vector.104) gives up (p − m) = 0 = v p (p + m). 1. the symmetric combination is simply 2g µν ψψ and so is not an independent bilinear form.142 Taking the Dirac adjoint of Eq. The antisymmetric combination is new. / (r) (r) (IX. But if any two indices are the same. We can choose the remaining eleven to transform simply under Lorentz transformations. but since any collection of γ matrices is simply a four by four matrix there are only be sixteen independent bilinears which can be constructed out of ψ and ψ. (IX. / / The plane wave bispinors also obey the completeness relations 2 r=1 (r) (r) (IX.105) / up up = p + m. γ ν ]ψ. Thus we define a new matrix γ5 : γ5 = iγ 0 γ 1 γ 2 γ 3 = i 4! µναβ γ µ ν α β γ γ γ ≡ γ5. γ ν }ψ and ψ[γ µ . Since {γ µ . However. We could go on indefinitely and construct n component tensors ψγ µ γ ν .γ α ψ. This bring the number of bilinears to eleven. We define i σ µν = [γ µ . r=1 (r) (r) 2 vp v p = p − m. this doesn’t give us anything new. Skipping to four component objects. we may split it up into symmetric and antisymmetric pieces: ψ{γ µ . This is a sixteen component object. so we need to find five more..the one component scalar and the four-vector. (IX..107) and then the six independent components of ψσ µν ψ transform like a two index antisymmetric tensor (note that some books define σ µν with an opposite sign to this). γ 0 γ 1 γ 0 γ 2 = −γ 0 γ 0 γ 1 γ 2 = −γ 1 γ 2 = iσ 12 . We already have five .109) . For example. (IX.108) So only the matrix γ 0 γ 1 γ 2 γ 3 and its various permutations are new. γ ν } = 2g µν . Consider first ψγ µ γ ν ψ. γ ν ] 2 (IX.

(IX. t) → −ψγ 0 γ5 ψ(−x. t). and so transforms like a pseudoscalar. (IX.116) The spatial components of ψγ µ ψ flip sign under a reflection whereas the time component is unchanged.3. t) → ψγ 0 γ µ γ 0 ψ(−x. (IX. t) → βψ(−x. ψγ5 ψ → ψγ5 ψ.. γ5 = γ5 = −γ 5 . t) = −ψγ5 ψ(−x. i = 1. γ µ } = 0.114) (IX.3 (IX.111) Since µναβ γ µγν γαγβ has no free indices. its transformation differs from that of ψψ when we consider parity transformations. ψγ5 ψ → ψγ 0 γ5 γ 0 ψ(−x.” It obeys † (γ5 )2 = 1.143 Here. t). it transforms like a scalar under boosts and rotations. t) ψγ i ψ(x. t). ψ(x.117) . and 0123 µναβ =1=− 1023 = 1032 = . t)γ 0 and so under a parity transformation ψψ(x. t) ψ(x. t) = −ψγ5 γ 0 γ 0 ψ(−x.. is a totally antisymmetric four index tensor. t) = γ 0 ψ(−x. t) → −ψγ i ψ(−x. Under parity. i = 1. t) → ψ(−x.. which is how a vector transforms under parity.. t)β = ψ(−x. {γ5 .113) (IX. On the other hand. t). However. we see that under a parity transformation ψγ µ ψ(x. the addition of the γ5 means that the axial vector transforms like ψγ 0 γ5 ψ(x. Again. and so ψγ 0 ψ(x. t) → ψγ i γ5 ψ(−x. However. t) ψγ i ψ(x.112) Thus ψγ5 ψ changes sign under a parity transformation. t) → ψγ 0 ψ(−x.110) γ5 is in many ways the “fifth γ matrix. t) → ψψ(−x.115) which make up an axial vector.. (IX. t) exactly as a scalar should transform. The final four independent bilinear forms are the components of ψγµ γ5 ψ (IX.

ψL and ψR are just the left and right-handed pieces of the Dirac bispinor in four component. / / 2 / / (IX. we have S = ψψ (scalar) (vector) (tensor) (pseudoscalar) (axial vector). PL = PL . PR PL = 0.121) . (IX. This interaction is parity violating because there is no way to define the transformation of W µ under parity such that this term is parity invariant. A scalar field φ (such as a meson) could couple like φψψ. 2 2 2 2 These satisfy the requirements for projections operators: PR = PR . we have chosen the sixteen bilinears which can be formed from a Dirac field and its adjoint to transform simply under Lorentz transformations. For example. Chirality and γ5 1 In the Weyl basis.144 which is the correct transformation law for an axial vector. (IX. PR +PL = 1.120) We also find that ψR ≡ ψP R = ψγ 0 PR γ 0 = ψPL and ψL = ψPR . and they project out the Weyl spinors u+ and u− from the Dirac bispinor: u+ 0 0 u− 1 = 2 (1 + γ5 )ψ = PR ψ ≡ ψR 1 = 2 (1 − γ5 )ψ = PL ψ ≡ ψL . Finally. whereas the coupling φψγ5 ψ conserves parity if φ transforms like a pseudoscalar.118) V µ = ψγ µ ψ T µν = ψσ µν ψ P = ψγ5 ψ Aµ = ψγ µ γ5 ψ Given these transformation laws it will be easy to construct Lorentz invariant interaction terms in the Lagrangian. An axial vector field B µ could couple in a parity conserving manner as B µ ψγµ γ5 ψ. We can define the projection operators (in any basis) (IX. in a parity violating theory (such as the weak interactions) a vector field W µ could couple to some linear combination of vector and axial vector currents: W µ ψγµ (a + bγ5 )ψ. notation. 2. The Weyl Lagrangian for right-handed particles may therefore be written L = ψi∂ PR ψ = ψi∂ PR ψ = ψPL i∂ PR ψ = ψR i∂ ψR . γ5 = 0 0 −1 . if we have a vector field Aµ (such as a photon). To summarize. Thus. PL = 1 (1 − γ5 ).119) PR = 1 (1 + γ5 ). a Lorentz invariant interaction is Aµ ψγµ ψ. rather than two component (u− and u+ ).

145 Similarly. which has a coupling of the form Z µ ψ L γµ ψL = Z µ ψ 1 γµ (1 − γ5 )ψ.R → e−iλ ψL.124) The independent symmetries are called chiral symmetries. where the term chiral denotes the fact that the symmetries has a “handedness”.123) When m = 0. (IX. the weak interactions involve the coupling of vector fields (the W ± and Z bosons) to only the left-handed components of spin 1/2 fields. it distinguishes left and right handed particles. ψL → e−iλ ψL . ψL → ψL . Chiral symmetries play an important role in the study of both the strong and weak interactions. the Dirac Lagrangian is invariant under the U (1) symmetry ψL. we see that without the mass term the Dirac Lagrangian just describes two independent helicity eigenstates. ψR → ψR and ψR → e−iλ ψR . for left-handed particles we have L = ψL i∂ ψL . However. We also note that we may write the parity violating Weyl Lagrangian describing left-handed neutrinos in the four-component form L = ψ L i∂ ψL . As we argued from physical grounds earlier. 2 Such a theory clearly violates parity.125) (IX.R . is the quantum of the Z µ vector field. the left and right handed fields must transform the same way under the internal symmetry. The Dirac Lagrangian is / / / L = ψL i∂ ψL + ψR i∂ ψR − m(ψL ψR + ψR ψL ).126) . (IX. For example. / (IX. The Z boson. so the helicity of the massive field is no longer a good quantum number. Because of the mass term. The mass term couples the right and left-handed fields.122) As we noticed before. since its helicity is no longer a Lorentz invariant quantity. when m = 0 this is no longer required. and the theory has two independent U (1) symmetries. this is exactly what must happen for a massive particle. (IX. for example. that is.

β= 1 0 (IX.129) − βm)ψ = 0. (IX. without proofs.127) where ψ is a set of four complex fields. Dirac Equation. β} = 0. You will find many of these results in Appendix A of Mandl & Shaw.146 E. {αi . Under a Lorentz transformation characterized by a 4 × 4 Lorentz matrix Λ. (IX. αj } = 2δij . Λ : ψ(x) → D(Λ)ψ(Λ−1 x). arranged in a column vector (a Dirac bispinor) and the α’s and β are a set of 4×4 Hermitian matrices (the Dirac Matrices).130) 0 −1 (where each component represents a 2 × 2 matrix). β 2 = 1. however. The corresponding equation of motion is (i∂0 + iα · The Dirac matrices obey the following algebra. (IX. Dirac Matrices The theory is defined by the Lagrange Density L = ψ † i∂0 + iα · − βm ψ.132) . {αi . 1.128) Two representations of the Dirac algebra that will be useful to us are the Weyl representation σ α= 0 and the standard (or Dirac) representation 0 α= σ 0 σ . β= 1 0 0 1 (IX.131) 0 −σ . they use a different normalization for the plane wave states. Dirac Lagrangian. Summary of Results for the Dirac Equation These pages summarize the results we have derived for the Dirac equation. (IX. 2. Space-Time Symmetries The Dirac equation is invariant under both Lorentz transformations and parity.

137) (IX. t).142) (IX. The γ matrices are not all Hermitian.136) 3. γ Matrices The Dirac adjoint of a Dirac bispinor is defined by ψ = ψ†β and the Dirac adjoint of a 4 × 4 matrix is A = βA† β.g. P : ψ(x. e.141) (IX. γ µ† = γµ = gµν γ ν = γ 0 γ µ γ 0 (IX. From these we can define the γ matrices with lowered indices by γµ ≡ gµν γ ν . (IX. ψAφ The γ matrices are defined by γ 0 = β. Under parity. Dirac Adjoint.147 For a boost characterized by rapidity φ in the e direction.139) .134) where L= 1 2 σ 0 0 (IX. (IX. γ i = βαi . These obey the usual rules for adjoints. ˆ e D (R(ˆθ)) = e−iL·ˆθ e (IX. t) → βψ(−x. ˆ e D (A(ˆφ)) = eα·ˆφ/2 e (IX.138) = φAψ.133) while for a rotation of angle θ about the e axis.140) ∗ (IX.135) σ in both the Weyl and standard representations.

We can choose these sixteen to form the components of objects that transform in simple ways under the Lorentz group and parity.147) (IX. we define a = aµ γ µ / and from the γ algebra it follows that a/ + /a = 2a · b. For any 4-vector a. These are S = ψψ (scalar) (vector) (tensor) (pseudoscalar) (axial vector) (IX.143) (IX. They obey the γ algebra {γ µ . Bilinear Forms (IX. γ ν } = 2g µν and also obey D(Λ)γ µ D(Λ) = Λµ ν γ ν .144) (IX.149) There are sixteen linearly independent bilinear forms we can make from a Dirac bispinor and its adjoint.148) (IX. /b b/ In this notation.146) (IX. the Dirac Lagrange density is ψ(i∂ − m)ψ / and the Dirac equation is (i∂ − m)ψ = 0.148 but they are self-Dirac adjoint (“self-bar”) γµ = γµ. / 4.150) V µ = ψγ µ ψ T µν = ψσ µν ψ P = ψγ5 ψ Aµ = ψγ µ γ5 ψ .145) (IX.

p = (m. we can choose the independent u’s and v’s in the standard representation to be √    (1) u0 =     2m 0 0 0   0   0   0 0 0 √ 2m     . (IX.152) is a totally antisymmetric four index tensor.154) 5.156) (IX. by applying a Lorentz boost.) We can construct the solutions for a moving particle.v = √  0  0   0      0   2m        (IX.158) 0 0 (Note that these are normalized differently than in Mandl & Shaw. the invariant phase space factor. . / / (IX.155) There are two positive-frequency and two negative-frequency solutions for each p. (IX.151) i 4! µναβ γ µ ν α β γ γ γ ≡ γ5.       √      2m   0   (2)   (1)   (2)  .153) γ5 is in many ways the “fifth γ matrix. The negative-frequency solutions are of the form ψ = veip·x .” It obeys † (γ5 )2 = 1. 0).157) For a particle at rest. and 0123 = 1. They omit the (r) (r) √ 2m from the normalization and instead include it in the definition of D. Plane Wave Solutions The positive-frequency solutions of the Dirac equation are of the form ψ = ue−ip·x where p2 = m2 and p0 = p2 + m2 .u =  . γ5 = γ5 = −γ 5 . {γ5 . (IX.149 where we have defined i σ µν = [γ µ . (IX. γ µ } = 0.v =  . The Dirac equation implies that (p − m)u = 0 = (p + m)v. γ ν ] 2 and γ5 = iγ 0 γ 1 γ 2 γ 3 = Here. µναβ (IX. up and vp .

161) . They obey the completeness relations 2 r=1 (r) (s) (r) (s) (r) (s) (IX. / (r) (r) (IX. up vp = 0. r=1 (r) (r) 2 vp v p = p − m.159) / up up = p + m.150 These solutions are normalized such that up up = 2mδ rs = −v p vp .160) Another way of expressing the normalization condition is up γ µ up = 2δ rs pµ = v p γ µ vp . This form has a smooth limit as m → 0. (r) (s) (r) (s) (IX.

ψ (X. We expect that the canonical commutation relations Eq. the Fourier components of the solution. that we need to impose canonical commutation relations. the b’s and c’s are numbers. and so we expect to be able to set up canonical commutation relations much in the same way as for the scalar field. any solution to the free field theory may be written as a sum of plane wave solutions 2 ψ(x. ψ † (y. c’s and their conjugates be creation and . t)] = iδab δ (3) (x − y). Since there are two spin states for the fields. The momentum conjugate to ψ is Π0 = ψ ∂L = iψ † ∂(∂0 ψ) (X. t)] = δ (3) (x − y). t). t) = r=1 2 d3 p (2π)3/2 2Ep d3 p (2π)3/2 2Ep bp up e−ip·x + cp vp eip·x bp up eip·x + cp vp (r)† (r)† (r) (r)† −ip·x (r) (r) (r)† (r) ψ † (x. (X. it is not a problem. (X. ψ † (y. QUANTIZING THE DIRAC LAGRANGIAN A. t)] = 0. t) = r=1 e .3) Just as in the case of the scalar field. Although this seems odd. (X. It is only on these fields. (Π0 )b (y. t). we would also need the time derivatives of the fields at the initial time). which completely define the state of the system.4) In the classical theory. and so the b’s and c’s carry a spin index. we have [ψ(x. we can find the state of the system at any following time (if the equations were second order in time.2) will require that the b’s. t)] = [ψ † (x. [ψ(x.2) Here we have explicitly displayed the spinor indices a and b. The equations of motion from the Dirac equation are first order in time. The u’s and v’s are the four component bispinors we found explicitly in the previous section. How Not to Quantize the Dirac Lagrangian We now wish to construct the quantum theory corresponding to the Dirac Lagrangian. t). the b’s and c’s are replaced by operators. Canonical Commutation Relations or.1) while the momentum conjugate to ψ † vanishes. ψ(y. Therefore we take [ψa (x. a general solution to the Dirac Equation has components with both spin states.151 X. and so ψ and ψ † form a complete set of initial value data. That is. In the quantum theory. if we know ψ and ψ † at some initial time. just as in the case of the scalar field. t). Suppressing the indices.

so to make things simpler let us make the ansatz [bp . Clearly if / B = −C. ψ † (y. ψ † (y.10) . i∂0 ψ = r (X. So choosing B = −C = 1.7) as required. we obtain [ψ(x. t).6) (r) (r) r vp v p up up = p+m and / (r) (r) = p−m. To see if this is a sensible quantum theory.8) (X. the pi γ i and m terms cancel. (X. Note. that the sign in the commutator for the c’s is opposite to what we might have expected. Here we have used the completeness relations r (r) (s) (X.152 annihilation operators. and the p0 = Ep in the numerator cancels the denominator.5) into the commutation relations gives [ψ(x. Substituting Eq. bp ] = Bδ rs δ (3) (p − p ) [cp . ψ Since ψ satisfies the Dirac equation. however. we should look at the Hamiltonian and see if the energy is bounded below: H = Π0 ∂0 ψ − L = iψγ 0 ∂0 ψ − iψγ µ ∂µ ψ + mψψ = −iψγ i ∂i ψ + mψψ.9) d3 p (2π)3/2 Ep (r) (r) −ip·x (r)† (r) bp up e − cp vp eip·x 2 (X.5) where B and C are constant which we shall solve for. and suggests that something may not be quite right here. cp ]vp v p γ 0 e−ip·x+ip ·y = 1 (2π)3 d3 p B(p + m)γ 0 e−ip·(x−y) / 2Ep −C(p − m)γ 0 eip·(x−y) / = 1 (2π)3 d3 p −ip·(x−y) e B(p0 γ 0 + pi γ i + m)γ 0 2Ep −C(p0 γ 0 − pi γ i − m)γ 0 . t)] = r. t). we can write this as H = iψγ 0 ∂0 ψ = iψ † ∂0 ψ. t)] = 1 (2π)3 d3 p e−ip·(x−y) = δ (3) (x − y) (X. In terms of the creation and annihilation operators. bp ]up up γ 0 eip·x−ip ·y (r) (s)† (r) (s) +[cp .s (2π)3 d3 pd3 p 2Ep 2Ep (r)† (s) [bp . cp ] = Cδ rs δ (3) (p − p ) (r) (s)† (r) (s)† (X.

with the Hamiltonian H = k d3 k ωk a† ak . There is indeed something seriously wrong with this theory .the Hamiltonian is unbounded from below! The c-type particles (antiparticles) carry negative energy. but the canonical commutation relations must be abandoned and replaced with something else.12) The δ (3) (0) will vanish when we normal order. so we just find H= r d3 pEp [Nb (p. the d3 x integral times the exponential becomes a delta function.s d3 x H d3 x ψ † i∂0 ψ d3 x d3 pd3 p (2π)3 Ep (r)† (r)† ip ·x (r) (r)† b up e + cp vp e−ip ·x × Ep p (r)† (r) bp up e−ip·x − cp vp eip ·x . (r)† (r) (X. (s) (s) (X. we can’t fix this problem simply by changing the sign of the Lagrangian. Canonical Anticommutation Relations We didn’t spend two lectures on spinors and γ matrices just to throw it all in at the first sign of trouble. This will simply force the b particles to carry negative energy. Unlike previous problems with positivity of the energy. C] + [A. r)] (X. (X. r) − Nc (p. Let k me remind you that this worked because of the following useful identity for commutators: [AB. and using up up = up γ 0 up = 2δ rs Ep = vp (r) (s) (r)† (s) vp . C] = A[B.11) (r)† (s) As usual.14) . Recall that for the scalar field theory we could interpret a† k and ak as creation and annihilation operators because of their commutation relations with the number operator N = d3 k a† ak (or equivalently. C]B. The theory therefore has no ground state. B.13) where Nb (p. since you can always lower the energy of a state by adding antiparticles.153 and so the Hamiltonian is H = = = r.). The theory can be rescued. There is therefore no way to obtain a sensible quantum theory from the Dirac Lagrangian using canonical commutation relations. r) and Nc (p. r) are the number operators for b and c type particles. we arrive at d3 pEp bp bp − cp cp r (r)† (r) (r)† (r) (r) (r)† H = = r d3 pEp bp bp − cp cp + δ (3) (0) .

154 This immediately gives [N, a† ] = k and also [N, ak ] = −ak . (X.16) d3 k [a† ak , a† ] = a† k k


Therefore a† acting on a state raises the eigenvalue of N by one and the energy by ωk , while k ak acting on the states lowers both eigenvalues, exactly as expected for creation and annihilation operators. However, there is another useful identity for commutators [AB, C] = A{B, C} − {A, C}B (X.17)

where {A, B} ≡ AB + BA is the anticommutator of A and B. This is extremely useful, because it means that if we were to impose anticommutation relations on the a† ’s and ak ’s, they could still k be interpreted as creation and annihilation operators. That is, let us impose the relations {ak , a† } = δ (3) (k − k ) k {ak , ak } = {a† , a† } = 0. k k We then find, using Eq. (X.17) [N, a† ] = k [N, ak ] = d3 k [a† ak , a† ] = k


d3 k a† {ak , a† } = a† k k k d3 k {a† , ak }ak = −ak k (X.19)

d3 k [a† ak , ak ] = −

exactly as required. This suggests we try the following anticommutation relations for the b’s and c’s: {bp , bp } = δ rs δ (3) (p − p ) {cp , cp } = δ rs δ (3) (p − p ) {bp , bp } = {bp , bp } = 0 {cp , cp } = {cp , cp } = {bp , cp } = ... = 0.
(r) (s) (r) (s) (r) (s)† (r) (s) (r)† (s)† (r) (s)† (r) (s)†


Not surprisingly, substituting these anticommutation relations into the field expansions we find that the equal-time commutation relations are replaced by now obey equal time anticommutation relations {ψ(x, t), ψ † (y, t)} = δ (3) (x − y) {ψ(x, t), ψ(y, t)} = {ψ † (x, t), ψ † (y, t)} = 0. (X.21)

155 Note that you have to be careful when dealing with anticommuting fields, since the order is always important. For example, ψ(x, t)ψ(y, t) = −ψ(y, t)ψ(x, t). The crucial step is now to see if this modification gives us an energy bounded from below. It is easy to see that it does, since the previous derivation of the Hamiltonian goes through completely unchanged up until the last line: H =

d3 pEp bp bp − cp cp
(r)† (r)

(r)† (r)

(r) (r)†


d3 pEp bp bp + cp cp + δ (3) (0) .

(r)† (r)


The anticommutation relations have given us a crucial sign change in the second term. Throwing away the δ (3) (0) as usual, we now have H=

d3 pEp [Nb (p, r) + Nc (p, r)]


which is bounded from below. Both b and c particles carry positive energy.

C. Fermi-Dirac Statistics

We have saved the theory, but at the price of imposing anticommutation relations on the creation and annihilation operators, and we must now examine the consequences of this. First consider the single particle states in the theory. We label these by the spin r (where r = 1 or 2 labels spin up and down, as we did when in the last chapter when writing down the explicit form of the plane wave solutions) as well as the momentum p. As usual, they are produced by the action of a creation operator on the vacuum (for definiteness, we consider particle states, not antiparticle states, although the arguments will clearly apply in both cases): | p, r = bp | 0 p , s |p, r = 0 |bp bp | 0 = 0 |{bp , bp }| 0 = δ rs δ (3) (p − p ) (X.24)
(s) (r)† (s) (r)† (r)†

and so the states have the correct normalization, just as they did in the scalar case. However, the multiparticle states are different from the spin 0 case. We find | p1 , r; p2 , s = bp1 bp2 | 0 = −bp2 bp1 | 0 = −| p2 , s; p1 , r
(r)† (s)† (s)† (r)†


156 and so the states of the theory change sign under the exchange of identical particles. Thus, the particle obey Fermi-Dirac statistics, instead of Bose-Einstein statistics. Consistency of the theory has demanded that we quantize the particles as fermions instead of bosons. In particular, the Pauli exclusion principle follows immediately from bp1
(r) 2

= − bp1

(r) 2



which means that there is no two-particle state made up of two identical particles in the same state | p1 , r; p1 , r = 0. Thus, it is impossible to put two identical fermions in the same state. It isn’t immediately obvious that a theory with Fermi-Dirac statistics will be causal. For bosons, we said that [φ(x), φ(y)] = 0 for (x−y)2 < 0 guaranteed that spacelike separated observables, which are constructed out of the fields, couldn’t interfere with one another: [O(x), O(y)] = 0, (x − y)2 < 0. (X.28) (X.27)

However, for fermi fields we now have the relation {ψ(x), ψ(y)} = 0 as well as {ψ(x), ψ † (y)} = 0 for (x − y)2 < 0 (this follows from the analogous calculation to that from which we derived ∆+ (x − y) = 0 for (x − y)2 < 0.) How do we see that this requirement guarantees causality in the quantum theory? The reason it does it that observables are always bilinear in the fields. For example, the energy, momentum and conserved charge are given by H = i Pi = −i Q = d3 x ψ † (x, t)∂0 ψ(x, t) d3 x ψ † (x, t)∂i ψ(x, t) (X.29)

d3 x ψ † (x, t)ψ(x, t).

This is not surprising. We know that spinors form a double-valued representation of the Lorentz group since they change sign under rotation by 2π. Observables, on the other hand, are unaffected by a rotation by 2π and so must be composed of an even number of spinor fields. Using the anticommutation relations (X.21), we can easily verify that observables bilinear in the fields commute for spacelike separation, as required. The fact that particles with integer spin must be quantized as bosons while particles with halfintegral spin must be quantized as fermions is a general result in field theory, and is known as

157 the spin-statistics theorem. We have, at least for spin 1/2 fields, demonstrated the second part of the theorem. The first part of the theorem, the fact that particles with integral spin must be quantized as bosons, follows from that observation that if we were to attempt to impose canonical anticommutation relations on the creation and annihilation operators for a scalar field we would find that the fields obeyed neither [φ(x), φ(y)](x−y)2 <0 = 0 nor {φ(x), φ(y)}(x−y)2 <0 = 0. The theory would therefore not be causal. This is the gist of the spin-statistics theorem: quantizing integral spin fields as fermions leads to an acausal theory, while quantizing half-integral spin fields as bosons leads to a theory with energy unbounded below (and so with no ground state).

D. Perturbation Theory for Spinors

Now that we understand free field theory for spin 1/2 fields, we can introduce interaction terms into the Lagrangian and build up the Feynman rules for perturbation theory. Let us consider a simple nucleon-meson theory (now that the nucleons are fermions, we no longer need to enclose the word in quotations) 1 µ2 / L = ψ(i∂ − m)ψ + (∂µ φ)2 − φ2 − gψΓψφ 2 2 (X.30)

where we either take Γ = 1, in which case φ is a scalar, or Γ = iγ5 , in which case φ is a pseudoscalar (we include the i so that the Lagrangian is Hermitian, L = L† .). The theory with Γ = 1 is known as Yukawa theory; it was originally invented by Yukawa to describe the interaction between real pions and nucleons. It turns out that Yukawa theory does not, in fact, provide the correct description of nucleon-meson interactions even at low energies (where the internal structure of the nucleons and pions is irrelevant, so they may be treated as fundamental particles). However, in modern particle theory the Standard Model contains Yukawa interaction terms coupling the scalar Higgs field to the quarks and leptons. Dyson’s formula and Wick’s theorem go through for fermi fields in almost the same way as for scalars. However, the anticommutation relations introduce a crucial difference. Recall that when (x − y)2 < 0, time ordering is not a Lorentz invariant concept. In one frame x0 > y0 while in another y0 > x0 . Nevertheless, the T -product of two scalar fields is Lorentz invariant because the fields commute when (x − y)2 < 0, so φ(x)φ(y) = φ(y)φ(x) and the order is unimportant. However, for fermions this no longer holds. If (x − y)2 < 0, fermi fields anticommute. So for spacelike separation, T (ψ(x)ψ(y)) = ψ(x)ψ(y) (X.31)

158 in the frame where x0 > y0 , but T (ψ(x)ψ(y)) = ψ(y)ψ(x) = −ψ(x)ψ(y) (X.32)

in the frame where y0 > x0 . Therefore we must modify our definition of the T-produce of fermi fields to make it Lorentz invariant. The solution is simple: just define the T-product to include a factor of (−1)N , where N is the number of exchanges of fermi fields required to time order the fields. Thus, for two fields ψ(x)ψ(y), T (ψ(x)ψ(y)) = −ψ(y)ψ(x), x0 > y0 ; y0 > x0 . (X.33)

since for y0 > x0 we must perform one exchange of fermi fields to time order them. When (x − y)2 < 0 the fields anticommute and the T -product is the same in any frame. Also, from Eq. (X.33) and the anticommutation relations, we have T (ψ1 (x1 )ψ2 (x2 )) = −T (ψ2 (x2 )ψ1 (x1 )). (X.34)

Therefore we treat fermi fields as anticommuting inside the time ordering symbol. (Note that in this discussion of T -products I am using ψ to represent any generic fermi field, including ψ † ). The normal-ordered product is defined as before. Writing ψ = ψ (+) +ψ (−) , where ψ (+) multiplies an annihilation operator and ψ (−) multiplies a creation operator, the normal-ordered product : ψ1 ψ2 : is : ψ1 ψ2 : = : ψ1 ψ2 = ψ1 ψ2
(+) (+) (+)

+ ψ1 ψ2



+ ψ1 ψ2



+ ψ1 ψ2



: (X.35)


− ψ2 ψ1


+ ψ1 ψ2


+ ψ1 ψ2


where the second term has picked up a factor of (−1) because of the interchange of two fermi fields. Just as for the T -product, fermi fields can be treated as anticommuting inside a normal ordered product, : ψ1 ψ2 := − : ψ2 ψ1 : . (X.36)

(Recall that bose fields commuted inside T -products and N -products; that is, their order was unimportant). With this modified definition of the time-ordered product, Dyson’s formula and Wick’s theorem go through as before. Note, however, that we must be careful with contractions in Wick’s theorem, for example for fermion fields A1 − A4 we have −−− −− − − : A1 A2 A3 A4 := − : A1 A3 A2 A4 := −A1 A3 : A2 A4 : (X.37)

159 and so pulling this particular contraction out of the normal-ordered product introduces a minus sign. In general, pulling a contraction out of a normal-ordered product introduces a factor of (−1)N , where N is the number of interchanges of fermi fields required.

1. The Fermion Propagator

−− −− The fermion propagator is obtained from the contraction ψ(x)ψ(y) (note that this is a four −−− −− by four matrix: Sab = ψ(x)a ψ(y)b . As with scalar fields, this is number (or rather a matrix of numbers) instead of an operator, so −− −− −− −− ψ(x)ψ(y) = 0 |ψ(x)ψ(y)| 0 = 0 |T (ψ(x)ψ(y))− : ψ(x)ψ(y) : | 0 = 0 |T (ψ(x)ψ(y))| 0 . (X.38)

First consider the case x0 > y0 . Then T (ψ(x)ψ(y)) = ψ(x)ψ(y). Putting in the explicit expressions for the fields and using the completeness relations for the plane wave states we find −− −− ψ(x)ψ(y) = d3 p e−ip·(x−y) (p + m) / (2π)3 2Ep = (i∂ x + m)i∆+ (x − y) (x0 > y0 ) /

up up = p + m /

(r) (r)


where we recall that i∆+ (x − y) = d3 p e−ip·(x−y) (2π)3 2Ep (X.40)

and ∆+ (x − y) = 0 when x0 = y0 . Performing a similar calculation for x0 < y0 , we find13 −− −− ψ(x)ψ(y) = θ(x0 − y0 )(i∂ x + m)i∆+ (x − y) + θ(y0 − x0 )(i∂ x + m)i∆+ (y − x) / / = (i∂ x + m) (θ(x0 − y0 )i∆+ (x − y) + θ(y0 − x0 )i∆+ (y − x)) / −− − = (i∂ x + m)φ(x)φ(y) / where −− − φ(x)φ(y) = d4 p −ip·(x−y) i e . 4 2 − m2 + i (2π) p (X.42)



Note that it is legitimate to pull the derivative outside of the θ function because the additional term which arises when a time derivative acts on the θ function vanishes: ∂θ(x0 − y0 )/∂x0 ∆+ (x − y) = δ(x0 − y0 )∆+ (x − y) = 0.

so the / / p i(-p+m) / p2-m2+iε FIG.160 We have now related the fermion propagator to the scalar propagator. Of course.2 The fermion propagator is odd in p. X. so it matters that p and the conserved charge (the p i(p+m) / p2-m2+iε FIG.1 The fermion propagator.44) (the i in the p + m term in the denominator does not affect the location of the pole. just as in the scalar theory. 4 p2 − m2 + i (2π) (X. X. propagator is often written as i(p + m) / i = (p + m − i )(p − m + i ) / / p−m+i / (X. (X. so in the / limit → 0 we may cancel this against the numerator).1). arrow on the propagator) are poiting in the same direction. contractions of fields which don’t create and then annihilate the same particle vanish.45) . Note that p2 − m2 + i = (p + m − i )(p − m + i ). We immediately see that this gives the Feynman rule for the fermion propagator shown in Fig. (X. When they point in opposite directions the sign of p is reversed (Fig. (X.43) (where we have explicitly included the matrix indices. and I is the four by four identity matrix).2)). −− − φ(x)ψ(y)a = 0 −−− −− ψ(x)a ψ(y)b = 0 −−− −− ψ(x)a ψ(y)b = 0. Note that the propagator is odd in p. Moving the derivative back inside the integral we have −−− −− ψ(x)a ψ(y)b = d4 p i(pab + mIab ) −ip·(x−y) / e .

For the relativistically normalized nucleon state | N (p. we get −−−−−−− −−−−−− : ψ a (x1 )Γab ψb (x1 )φ(x1 )ψ c (x2 )Γcd ψd (x2 )φ(x2 ) := −−− −−− ψb (x1 )ψ c (x2 ) : ψ a (x1 )Γab φ(x1 )Γcd ψd (x2 )φ(x2 ) : . Feynman Rules We can deduce the Feynman rules for this theory by explicitly calculating the amplitudes for several scattering processes. r) = e−ip·x2 × up and similarly N (p .161 2.47) Anticommuting the fields inside the normal-ordered product.50) (X.46) where we have included the spinor indices. r) (momentum p.48) since there are four permutations of fermion fields required (note that fermi fields commute with bose fields).51) . We can pull the propagator out of the first term. and since this involves an even number of exchanges of fermi fields (two). This only differs from the first time by the interchange of x1 and x2 . The O(g 2 ) term in Dyson’s formula is (−ig)2 2! d4 x1 d4 x2 T ψ a (x1 )Γab (x1 )ψb (x1 )φ(x1 )ψ c (x2 )Γcd ψd (x2 )φ(x2 ) (X. and we are using the convention that repeated spinor indices are summed over (so this is just matrix multiplication). and since we are symmetrically integrating over x1 and x2 . There are two contractions which contribute: −−−−−−− −−−−−− : ψ a (x1 )Γab ψb (x1 )φ(x1 )ψ c (x2 )Γcd ψd (x2 )φ(x2 ) : −−−−−−−−−−−−−−−−−−−− −−−−−−−−−−−−−−−−−−− : ψ a (x1 )Γab ψb (x1 )φ(x1 )ψ c (x2 )Γcd ψ d (x2 )φ(x2 ) : . r ) |ψ(x1 )| 0 = eip ·x1 × up This immediately gives us two additional Feynman rules: (r ) (r) (X. the two terms give identical contributions.49) The ψ field inside the normal product must now annihilate the nucleon. (X. First we consider nucleon-meson scattering. this cancels the 1/2! in Dyson’s formula. This term contributes to a variety of processes. we can rewrite the second term as −−−−−−− −−−−−− : ψ c (x2 )Γcd ψd (x2 )φ(x2 )ψ a (x1 )Γab ψb (x1 )φ(x1 ) : (−1)4 (X. N + φ → N + φ. (X. spin r) we have 0 |ψ(x2 )| N (p. Just as before.

but now the ψ field annihilates the incoming antinucleon and the ψ field . X. four component column vectors. as shown in Fig. (r) (X.3 Feynman rules for external fermion legs.) The vertices and fermion propagators are four by four matrices while the bispinors are -igΓ FIG. Diagrammatrically. X. (X.4). include a factor of up (see Fig.5). When calculating Feynman diagrams for spinors. This is almost identical to the previous process.52) We next consider antinucleon-meson scattering. each interaction vertex corresponds to a factor of −igΓ (see Fig. process to be iA = (−ig)2 up Γ (r ) (r ) (r) Applying the Feynman rules. and as before we get two Feynman diagrams corresponding to the two choices of which φ field creates or annihilates which meson. include a factor of up • For each outgoing fermion with momentum p and spin r. (X.r FIG.r u(r) p u(r) p _ p. p. we find the invariant amplitude for this i(p + / + m) / q i(p − / + m) / q + 2 − m2 + i (p + q) (p − q )2 − m2 + i Γup . (X. including each matrix as it is encountered along the fermion line. including each matrix as it is encountered along the fermion line. From Eq.3). (X. where Sab is the fermion propagator.) (r) (r) Finally.162 • For each incoming fermion with momentum p and spin r. so I’m going to say it again.4 Fermion-scalar interaction vertex. this just corresponds to starting at the head of the arrow and working back to the start. The φ fields act as they always did on the meson states. N + φ → N + φ. the order of matrix multiplication is given by starting at the head of an arrow and working back to the start.49) we see that the matrices are multiplied together in the order up ΓSΓup . That last point is important. The amplitude is given by multiplying all of these factors together.

p. the fermi minus signs will be significant in the next example.53) v(r) p v(r) p p. introducing a factor of −1 because of the interchange of the two fermion fields. (r ) (X.7). . The amplitude is iA = −(−ig)2 v p Γ (r) (r) (r) i(−p − / + m) / q i(−p + / + m) / q + 2 − m2 + i (p + q) (p − q )2 − m2 + i Γvp .163 q' a) p'.6 Feynman rules for external fermion legs.5 Feynman diagrams contributing to nucleon-meson scattering. • Include the appropriate minus signs from Fermi statistics.r p'.r' p. the fields now include factors of v and v: 0 |ψ(x1 )| N (p. r) = e−ip·x1 × v p N (p . instead of up ) and work along the line to the start of the arrow. The two diagrams contributing to the process are shown in Fig.r FIG. This time the matrices are multiplied by starting at the incoming antinucleon and working back to the outgoing nucleon.r b) q q' p. Diagrammatically. X. When acting on the external states. • For each outgoing antifermion with momentum p and spin r. X. it’s the same as before: start at the end of the arrow (with a factor of v p this time.54) The overall −1 is clearly irrelevant. because they will differ between the two diagrams. So in the normal-ordered product : ψ(x1 )Γφ(x1 )Γψ(x2 )φ(x2 ) : the ψ field has to be moved to the right of the ψ field in order to annihilate the incoming state.r _ (r) (r) (r) (r ) (X. creates the outgoing antinucleon.r' q FIG. include a factor of vp . r ) |ψ(x2 )| 0 = eip ·x2 × vp This leads to three more Feynman rules: • For each incoming antifermion with momentum p and spin r. However. include a factor of v p . (X. since it vanishes when the amplitude is squared.

(X.59) There are now two possibilities: either ψ(x1 ) or ψ(x2 ) can annihilate the nucleon with momentum q. X. N (q.7 Feynman diagrams contributing to antinucleon-meson scattering.56) |0 . N (q . This sets the convention. we consider nucleon-nucleon scattering.r b) q q' p. r). s ) | = (2π)3 (2ωq )1/2 (2ωp )1/2 0 |bp bq and so the matrix element in Eq. r ). we then obtain 0 |bp bq (r ) (s ) : ψΓψ(x1 )ψ(x2 )Γuq bp | 0 e−iq·x2 : ψ(x1 )Γψ(x2 )Γuq ψ(x1 )bp | 0 e−iq·x2 : ψ(x1 )Γup ψ(x2 )Γuq | 0 e−iq·x2 −ip·x1 (X. N (q . So for definiteness. First we choose the case where ψ(x2 ) annihilates the nucleon. N + N → N + N . r ). (X. and so we have now choice but to define the corresponding bra as N (p . (X. Using the field expansion for ψ. leaving us with the matrix element N (p .55) becomes 4(2π)6 (ωq ωp ωq ωp )1/2 0 |bp bq (r ) (s ) (r ) (s ) (X. N (q. s ) | : ψΓψ(x1 )ψΓψ(x2 ) : | N (p. The (relativistically normalized) state | N (p.60) (r) (s) (s) (r)† (s) (r)† = − 0 |bp bq (r ) (s ) (r ) (s ) = − 0 |bp bq .57) Because of Fermi statistics.r p'. r).58) : ψΓψ(x1 )ψΓψ(x2 ) : bq (s)† (r)† bp | 0 . In this case the φ fields are contracted.164 q' a) p'. Finally.r' q FIG. let us choose the first definition. s) . s) can be defined either as (2π)3 (2ωq )1/2 (2ωp )1/2 bq or (2π)3 (2ωq )1/2 (2ωp )1/2 bp bq (r)† (s)† (s)† (r)† bp | 0 (X. (X.55) Now we have to be careful.r' p. the two definitions differ by a relative minus sign.

corresponding to two independent matrix multiplications. Also note that in this case there are two sets of arrowed lines. q'. Again. then there is no relative minus sign. the rule is to follow each arrowed line from finish to start. X. and the two graphs have a relative minus sign. The crucial observation is that the two choices differ by a relative minus sign.s' q. As .62) We could now follow through the same line or reasoning. r q'. r FIG. the fields must be anticommuted. Since we are integrating over x1 and x2 symmetrically.63) Note that since the overall sign of the graphs is unimportant.8). It is only the relative minus sign which is significant. once again this cancels the 1/2! in Dyson’s formula. If ψ(x1 ) creates the nucleon with q . Therefore the amplitude for the process is  −ig 2  uq Γuq up Γup (s ) (s) (r ) (r) (q − q )2 − µ2 + i − uq Γup up Γuq (s ) (r) (r ) (s)   (q − p )2 − µ2 + i (X.r' q.61) (X. if ψ(x2 ) creates the nucleon.r' p. so it commutes with the fields) in the correct position as far as matrix multiplication goes. multiplying matrices as you come to them. except choosing ψ(x1 ) to annihilate the nucleon with momentum q. (X. Thus.8 Feynman diagrams contributing to nucleon-nucleon scattering. and there is an additional minus sign. and we would find the same result with the interchange x1 ↔ x2 . The two graphs differ only by the interchange of identical fermions in the final state. s p'. In these diagram you simply have to do this twice.s' p. The two terms clearly correspond to the diagrams in Fig.165 where in the last line we have put the factor of up (which is not an operator. However. s p'. Now there are two choices for which ψ field creates which nucleon. we find two terms −uq Γuq up Γup e−i((q−q )·x2 +(p−p )·x1 ) and uq Γup up Γuq e−i((q−p )·x2 +(p−q )·x1 ) . (s ) (r) (r ) (s) (s ) (s) (r ) (r) (r) (X.

To calculate the rate for a process with given initial and final spins. in many cases this is not necessary. In such situations we can use the completeness relations for the spinors to perform the averaging and summing without ever writing down the explicit form of the plane wave states. Γ = iγ5 . we could simply use the explicit forms of the u’s and v’s that we found earlier. From Eq. nucleon-meson scattering in the “pseudoscalar” theory. The two amplitudes interfere with a relative minus sign. (r) (X. It is straightforward to show. (X. However. where it hits the first γ5 and the two square to one. In a large number of experimental situations the spins of the initial particles are unknown. that two diagrams which differ by the exchange of a fermion in the final state and an antifermion in the initial state (or vice versa) also interfere with a relative minus sign. Also. using the techniques of this section. and so a given particle has a 50% chance to be in either spin state. This example also shows you several other tricks which are useful for evaluating amplitudes with Fermi fields. q 2 = q 2 = µ2 to write this as iA = ig 2 up /up q (r ) (r) 1 1 + 2p · q + µ2 + i 2p · q − µ2 + i . the final spins are often not measured. γµ } = 0.166 expected. by conservation of momentum. so we may rewrite the amplitude as iA = ig 2 up (r ) (−p − / + m) / q (−p + / + m) / q + 2 − m2 + i (p + q) (p − q)2 − m2 + i (r) (r) up . Spin Sums and Cross Sections Our Feynman rules allow us to calculate amplitudes in terms of plane-wave solutions u and v. E. We will demonstrate this in a worked example.65) Next we use the fact that the spinors obey (p − m)up = up (p − m) = 0 as well as the mass-shell / / conditions p2 = p 2 = m2 . (X.64) 2 Using γ5 = 1 and {γ5 .52) the invariant Feynman amplitude for nucleon-meson scattering is iA = ig 2 up γ5 (r ) (p + / + m) / q (p − / + m) / q + (p + q)2 − m2 + i (p − q )2 − m2 + i γ5 up . A similar situation arises in nucleon-antinucleon scattering. (r) (X. p − q = p − q. we can anticommute the second γ5 through the propagators. our theory automatically incorporates fermi statistics. so we are interested in cross sections or decay rates in which we sum over all possible spins of the final particles.66) . Similarly.

. if somewhat tedious. p . q)2 qµ qν Tr (p + m)γ µ (p + m)γ ν .167 Let us call the expression in the square bracket F (p. q)2 qµ qν Tr up up γ µ up up γ ν r.67) 2 where we have use the relations γ0 = 1 and γ0 γ µ γ0 = γ µ† . Some useful formulas are: Tr [γ µ γ ν ] = 4g µν Tr γ µ γ ν γ α γ β = 4(g µν g αβ + g µβ g να − g µα g νβ ) Tr [odd number of γ matrices] = 0 Tr [γ µ γ5 ] = Tr [γ µ γ ν γ5 ] = Tr [γ µ γ ν γ α γ5 ] = 0 Tr γ µ γ ν γ α γ β γ5 = 4i µναβ . q). (X. The collection of spinors and gamma matrices is simply a number (a one by one matrix) and so is equal to its trace.70) Applying these trace theorems to our expression gives 1 2 |A|2 = 2g 4 F (p.r 1 = g 4 F (p. p . (X. p . p .69) The traces of products of γ matrices have simple expressions. p . Therefore |A2 | = g 4 F (p.68) r |A|2 ) |A|2 ) we obtain (r ) (r ) (r) (r) 1 2 g 4 F (p. Now the completeness relations comes in. q)2 2(p · q)(p · q) − p · p µ2 + m2 µ2 . / / 2 (X. p . q)2 |up /up |2 q = g 4 F (p. to go to the centre of mass frame.r r 1 2 (X. p . q)2 qµ qν up γ µ up up γ ν up (r ) (r) (r) (r ) (r ) (r) (r)† (r ) (r ) (r) (X. substitute explicit expressions for the external momenta and perform the phase space integrals to obtain the total cross section for meson nucleon scattering. q)2 qµ qν Tr up γ µ up up γ ν up (r ) (r ) (r) (r) (r ) (r) (r) (r ) = g 4 F (p. q)2 qµ qν up γ µ up up γ 0 γ 0 γ ν† γ 0 up = g 4 F (p. p . q)2 qµ qν Tr up up γ µ up up γ ν . Squaring the amplitude we get |A|2 = g 4 F (p. which are straightforward to prove (you can find a discussion of traces in Appendix A of Mandl & Shaw). The reason for writing it in this way is that a trace of a product of matrices is invariant under cyclic permutations of the factors. q)2 qµ qν p µ pν + p ν pµ − g µν p · p + m2 g µν r. Averaging over initial spins (corresponding to and summing over final spins (corresponding to 1 2 |A|2 = r. p .71) At this point it is straightforward.r = 2g 4 F (p. p .

∂ µ Aν ∂ µ Aν . A. quantizing a massless vector field is a delicate procedure.168 XI. and so gives the same contribution to the action. In this section we will see that a classical free field theory of a massless vector field is simply Maxwell’s equations in free space. has no more than two derivatives (a simplifying assumption) and is Lorentz invariant. Aµ Aµ . Any other term may be written as a sum of these terms and a total derivative. The Classical Theory A vector field is a four component field whose components transform in the familiar way under Lorentz transformations. ∂ µ Aµ ∂ ν Aν . However. Quantum Electrodynamics. Quantizing the theory will give us the theory of the quantized electromagnetic field. (XI. up to total derivatives. The particles associated with the quantized vector field will be photons. . • 1 derivative: there are no possible Lorentz invariant terms in four dimensions. In this section we will finesse these problems by quantizing the theory of a massive vector field and then taking the massless limit and tackling any problems that arise at that stage. ∂µ Aν ∂ν Aµ ∼ Aν ∂µ ∂ν Aµ ∼ ∂ν Aν ∂µ Aµ .1) Since we already know how products of four-vectors transform. we want to construct the simplest L which is quadratic in the fields (so that the resulting equations of motion are linear). For example. VECTOR FIELDS AND QUANTUM ELECTRODYNAMICS Quantizing a scalar field theory led to a theory of mesons. we can go straight to writing down Lagrangians. due to complications arising from gauge invariance of the classical theory. This gives the following terms: • 0 derivatives: there is only one possibility. • 2 derivatives: there are two independent terms. while the quantized spinor field allowed us to describe the interactions of spin 1/2 fermions. A µ (x) = Λµ ν Aν (Λ−1 x). As before.

This type of solution looks exactly like a scalar field. This leads to the equations of motion −2Aν − a∂ν ∂µ Aµ + bAν = 0. ε ∝ k (4-D longitudinal) 2. In the rest frame of the field. ε · k = 0 (4-D transverse). 1.6) . T This solution describes a field of mass µ2 . It would be (XI. (4-D longitudinal) k 2 kν + ak 2 kν + bkν = 0 ⇒ (k 2 (1 + a) + b)kν = 0 −b ⇒ k2 = ≡ µ2 L 1+a This solution has the right dispersion relation for a particle of mass µ2 . since the ε’s clearly correspond to the three polarization state of a massive spin one particle. T The 4-D transverse solution appears to be what we are looking for.5) The solutions to Eq. ε). isn’t very interesting. however. The 4-D longitudinal solution.2) (XI. these two types of solution correspond to ε = (ε0 . L 2. (XI. The lead to the equations of motion 1.3) (XI.5) may be classified in a Lorentz invariant manner into two classes. respectively. (4-D transverse) k 2 εν + bεν = 0 ⇒ k 2 = −b ≡ µ2 .169 The most general Lagrangian satisfying these requirements is then 1 L = ± [∂µ Aν ∂ν Aν + a∂µ Aµ ∂ν Aν + bAµ Aµ ] 2 for some constants a and b.4) This leads to k 2 εν + akν k · ε + bεν = 0. (XI. this doesn’t lead to anything new. we look for plane wave solutions of the form Aν (x) = εν e−ik·x for some constant 4-vector ν. As before. Since we already know how to quantize scalar field theory.7) (XI. (XI. 0) and ε = (0.

Or. the Lagrangian is L=± and the equations of motion are ∂µ F µν + µ2 Aν = 0. Define the field strength tensor F µν ≡ ∂ µ Aν − ∂ ν Aµ . (XI.8) where µ2 ≡ µ2 . any k is a solution to Eq.14) Equations (XI. . (XI. At the level of these two equations the µ2 → 0 limit is smooth. the equation of motion Eq.170 nice to get rid of this solution altogether. Therefore the longitudinal solutions are absent from the Lagrangian L=± 1 (∂µ Aν )2 − (∂µ Aµ )2 − µ2 A2 2 (XI. if you prefer.14) are equivalent to the Proca equation. In terms of F µν . we derive the requirement that the field is transverse ∂µ ∂ν F µν = 0 ⇒ ∂µ Aµ = 0. It is this arbitrariness in the solution to the classical theory which makes the massless theory difficult to quantize. This leads to the equations of motion T 2Aν − ∂ν ∂µ Aµ + µ2 Aν = 0. Fµν = −Fνµ .11) (XI. when a = −1 and b = 0. (XI. Using the fact that Fµν is antisymmetric. (XI.10) Equation (XI. (2 + µ2 )Aν = 0. 2Aµ = 0. removing it from the spectrum.15) When a = −1 and b = 0. each component of Aµ is found to satisfy the massive Klein-Gordon equation. if the 4-D transverse field is massive).13) and (XI.9) This can be written in a more compact form by introducing some more notation.13) Substituting this condition into the Proca equation. however.6).6) has no solutions14 . (XI. (XI. setting a = −1 takes µL to ∞.12) 1 1 Fµν F µν − µ2 Aµ Aµ 4 2 (XI. This is simple enough to do: if b = 0 (that is.12) is known as the Proca Equation. although in this form it is not clear how to derive them from a Lagrangian. They look promising. 14 ∂µ Aµ = 0. (XI.

However. Therefore we will stick with finite µ2 for a while longer. there are three linearly independent polarization vectors εµ . Aµ = εµ e−ik·x .16). 0. 0. In any basis. −1 and 0 respectively. Returning to the plane wave solutions to the Proca equation. ε(3) = (0. things aren’t quite so simple. corresponds to the free-space Maxwell Equations ×B = while the remaining two equations.3. ∂t ·E =0 (XI. the basis states are chosen to obey orthonormality (r) εµ εµ(s)∗ = −δ rs (XI. while the components of the field strength tensor are the electric and magnetic fields E = − φ− B = By direct substitution.171 These are just Maxwell’s equations in free space. each component of Aµ satisfies the massless wave equation. we could choose the basis ε(1) = (0.20) (r) but in fact it is usually more convenient to choose the basis vectors to be eigenstates of Jz : 1 1 ε(1) = √ (0. 0). 0..17) Ez We may also verify directly that the massless Proca equation. 1. Recall that in classical electromagnetism the scalar and vector potentials φ and A make up the components of the four-vector Aµ = (φ. ∂t ·B =0 (XI. 0. 1.   E  x F µν =    Ey  By   −Bx   0 (XI. 0. i. The condition ∂µ Aµ = 0 could only be derived when µ2 = 0. 0 −Ex 0 Bz −By −Ey −Bz 0 Bx −Ez   . 1.18) immediately follow from the definitions Eq. ε(2) = √ (0. 1) 2 2 (XI. ε(2) = (0. 1) (XI. (XI.22) . we easily find  ∂A ∂t (XI. ×E =− ∂B .16) × A. r = 1. A). 0. 0). ε(3) = (0. ∂µ F µν = 0. 1. 0). 0). In the rest frame.19) ∂E . The vector field Aµ is thus the familiar vector potential of classical electrodynamics. −i.21) which have Jz = +1. In the gauge where ∂µ Aµ = 0.

172 and completeness
3 r=1 (r) εµ ε(r) = −gµν + ν ∗

kµ kν µ2


relations. The minus sign in Eq. (XI.22) arises because the polarization vectors are spacelike. The orthonormality and completeness relations are Lorentz covariant, so are true in any frame, not just the rest frame. The sign of the Lagrangian may be fixed by demanding that the energy be bounded below, as usual. Denoting spatial indices by Roman characters, the Lagrangian may be written as L=± 1 1 µ2 µ2 F0i F 0i + Fij F ij − Ai Ai − A0 A0 2 4 2 2 (XI.24)

and so the time components of the canonical momenta are ∂L = ±F 0i ∂(∂0 Ai ) ∂L = 0. ∂(∂0 A0 )


The fact that the momentum conjugate to A0 vanishes does not constitute a problem. Because ∂µ Aµ = 0, there are fewer degrees of freedom than one would na¨ ıvely expect, and the spatial Ai ’s and their canonical momenta are sufficient to define the state of the system. The Hamiltonian is H = ±F0i ∂ 0 Ai − L = ±F0i F 0i ± F0i ∂ i A0 − L = ±F0i F 0i ∂ i F0i A0 − L

= ±F0i F 0i µ2 A0 A0 − L 1 µ2 µ2 1 = ± F0i F 0i − Fij F ij + Ai Ai − A0 A0 2 4 2 2


where we have integrated by parts and used the equation of motion ∂µ F µ0 = ∂i F i0 = −µ2 A0 . The metric tensor obscures it, but each term in the square brackets is a sum of squares with a negative coefficient and so is negative (for example, −Ai Ai = Ai Ai > 0), and so the Lagrangian has an overall minus sign, 1 µ2 L = − Fµν F µν + Aµ Aµ . 4 2 (XI.27)

B. The Quantum Theory

Canonically quantizing the theory is a straightforward generalization of the scalar field theory case, so we will skip some of the steps. Since the spatial components Ai and their conjugate momenta form a complete set of initial conditions, it is only on these fields that we impose the canonical commutation relations
j [Ai (x, t), F j0 (y, t)] = iδi δ (3) (x − y)

[Ai (x, t), Aj (y, t)] = [F i0 (x, t), F j0 (y, t)] = 0.


Expanding the field in terms of plane wave solutions times operator-valued coefficients ak (r) and a† k


Aµ (x) =

(r) d3 k ∗ (r) √ ak (r) εµ (k)e−ik·x + a† ε(r) (k)eik·x µ k 3/2 2ω (2π) k


and substituting this into the canonical commutation relations gives, not unexpectedly, the commutation relations [ak (r) , a† k
(s) (s)

] = δ rs δ (3) (k − k )

[ak (r) , ak ] = [a† k The Hamiltonian also has the expected form : H :=

, a† k


] = 0.


d3 k ωk a† k


ak (r)


and so we can interpret a† k with polarization r.


and ak (r) as creation and annihilation operators for spin one particles

−−− −− The propagator Aµ (x)Aν (y) may be calculated in a similar manner as the spinor case. Proceeding as before, we write −−− −− Aµ (x)Aν (y) = 0 |T (Aµ (x)Aν (y))| 0 If x0 > y0 , −−− −− Aµ (x)Aν (y) = 0 |A(+) (x)A(−) (y)| 0 µ ν = 0 |[A(+) (x), A(−) (y)]| 0 µ ν = [A(+) (x), A(−) (y)] µ ν (XI.33) (XI.32)

174 where we have split Aµ into the piece containing the creation operator, Aµ

and a piece Aµ


containing the annihilation operator. From the expansion of Aµ (x), it is straightforward to show that [A(+) (x), A(−) (y)] = µ ν = = where i∆+ (x − y) = d3 k e−ik·(x−y) (2π)3 2ωk (XI.35) d3 k e−ik·(x−y) (2π)3 2ωk d3 k
(r) (r) εµ (k)εν (k) r ∗

kµ kν e−ik·(x−y) −gµν + 2 (2π)3 2ωk µ y ∂y ∂µ ν −gµν − 2 i∆+ (x − y) µ


y and ∂µ ≡ ∂/∂y µ . After including the y0 > x0 term we obtain y y −−− −− ∂µ ∂ν Aµ (x)Aν (y) = θ(x0 − y0 ) −gµν − 2 µ

i∆+ (x − y)
y y ∂µ ∂ν µ2

+θ(y0 − x0 ) −gµν − Now, the scalar propagator is

i∆+ (y − x).


−− − φ(x)φ(y) = θ(x0 − y0 )i∆+ (x − y) + θ(y0 − x0 )i∆+ (y − x) d4 k −ik·(x−y) i e = 4 2 − µ2 + i (2π) k and so we would like to commute the θ functions and derivatives in Eq. (XI.36) to obtain −−− −− Aµ (x)Aν (y) = = −gµν −
y y ∂µ ∂ν µ2


(θ(x0 − y0 )i∆+ (x − y) + θ(y0 − x0 )i∆+ (y − x)) e−ik·(x−y) i . k 2 − µ2 + i (XI.38)

kµ kν d4 k −gµν + 2 (2π)4 µ

This leads to the propagator for a massive vector field, which is represented by a wavy line: −i g µν −
kµ kν µ2

k 2 − µ2 + i



Note that the vector propagator carries Lorentz indices: one end of the line corresponds to a field created by Aµ while the other corresponds to the field created by Aν . While this is correct, the derivation was not quite right when µ = ν = 0. In this case, the time derivatives don’t commute with the θ functions and there is the additional term when one of the derivatives acts on the θ function, giving a factor of δ(x0 − y0 ), and the other acts on the ∆+

ν) = (0.40) is invariant.40) leads to the interaction vertex shown in Fig.2 Fermion-vector interaction vertex 15 This sort of difficulty arises in the canonical quantization procedure because it breaks manifest Lorentz invariance.40) where Γ has the general form Γ = a + bγ5 by Lorentz Invariance. The fact that this term does not contribute is not obvious in the canonical quantization procedure. If you like. In this case. puts this derivation on sounder footing. As we discussed before.15 Now consider adding a fermion such as an electron to the theory. XI. From our previous experience with interacting theories. µ We note at this stage that there is a simple -igγµΓ FIG.2).175 ν µ -i(gµυ-kµkν/µ2) k2-µ2+iε FIG.1 The propagator for a massive vector field. XI. This wasn’t a problem in the spinor case because there was only a single time derivative. (XI. 0) as well. (XI. when both a and b are nonzero this theory violates parity. the time derivative of ∆(x − y) does not vanish when x0 = y0 and so there is an additional term. the interaction term Eq. and then argue that by Lorentz invariance the result must have this form for (µ. respectively. ν) = (0. A parity conserving theory may have either Γ = 1 (vector coupling) or Γ = γ5 (axial vector coupling). . which we will not discuss in this course. and the term vanished because ∆(x − y) = 0 when x0 = y0 . The path integral formulation treats space and time in a symmetric fashion. you can use the derivation above for (µ. in which case the components of Aµ transform under parity as a vector or an axial vector. (XI. function. by treating temporal indices different from spatial indices. since there is no choice of transformation for Aµ under which the interaction term Eq. however. 0). The path integral formulation of quantum field theory. A simple interaction term between the fermi field ψ and Aµ is /Γψ LI = −gψγ µ ΓψAµ = −gψA (XI.

When all the fields in the interaction term are different. r) = r =1 3 (r) d3 k (r ) √ (r √ εµ ) (k )e−ik ·x 0 |ak 2ωk (2π)3/2 a† | 0 k 3/2 2ω (2π) k = r =1 3 d3 k d3 k r =1 ε(r) (k)e−ik·x µ ωk (r ) (r ) (r) εµ (k )e−ik ·x 0 |ak a† | 0 k ωk (r) ωk (r ) (r) εµ (k )e−ik ·x 0 |[ak . include a factor . In the quantum theory. εµ (k)  k µ µ k εµ(r)(k) εµ(r)*(k) FIG. From the field expansion. evaluating Dyson’s formula requires matrix elements of the Aµ field between incoming and outgoing vector meson states and the vacuum. The Massless Theory To obtain a quantum theory of electromagnetism.41) where | V (k.176 rule for writing down the Feynman rule associated with an interaction term in L. Finally.3 Feynman rules for external vector particles.41) and its complex conjugate lead to the Feynman rule incoming • For every of outgoing   (r) εµ (k)   (r) ∗ (r) vector meson with momentum k and polarization r. each incoming vector meson contributes a factor of εµ to the amplitude in addition to the usual exponential factor. corresponding to the n! different way of choosing which field corresponds to which line in the vertex. XI. 3 0 |Aµ (x)| V (k. or i times the interaction Lagrangian. Equation (XI. the resulting Feynman rule is just −i times the interaction Hamiltonian. the limit µ → 0 must be taken of the results in the previous section. A term with n identical fields has a combinatoric factor of n! in the Feynman rule. C. Therefore. This limit looks bad for several reasons. a† ]| 0 k ωk = = (XI. there is a . r) is a relativistically normalized single particle state containing a vector meson with momentum k and polarization r.

177 factor of k µ k ν /µ2 in the vector propagator. so is any multiple of j µ . Dψ = iψ.49) (XI. The equations of motion in this theory are ∂µ F µν + µ2 Aν = J ν which leads to ∂µ Aµ = 1 ∂µ J µ . Consider L coupled to a source J µ (x) L = L0 − Aµ J µ (x). Fortunately. This will turn out to be closely related to a problem which arises at the classical level. this looks bad in the limit µ → 0. so is (for example) φa → exp(−2iqa λ)φa . There is no physics in this ambiguity . that is. (XI. However. DL = 0 and the current jµ = a Πµ Dφa = −i a a Πµ qa φ a a (XI. Recall that Noether’s theorem ensures that any internal symmetry has an associated conserved current and charge. ψ → eiλ ψ.44) is well defined. not operators. The conjugate momenta are Πµ = iψγ µ . if φa → exp(−iqa λ)φa is a symmetry. In this case. Πµ = 0 ψ ψ (XI. the simplest example being a U (1) symmetry associated with the transformation φa (x) → e−iλqa φa (x) (XI. (XI. we’re old hands at finding conserved currents.45) is a symmetry. If Eq.45). (XI.48) (XI.46) is conserved.43) (XI. and the qa ’s are numbers (the charge of each field).if j µ is a conserved current. Note that the qa ’s are arbitrary up to a multiplicative constant. µ2 (XI. For example.47) . There is no implied sum over a in Eq.42) Again. it gives a clue to how to obtain a theory with a sensible µ → 0 limit: the limit exists only if Aµ couples to a conserved current.44) (XI. the Dirac Lagrangian is invariant under the transformation ψ → e−iλ ψ. Therefore the corresponding qa ’s are qψ = 1.45) for some set of fields (not necessarily scalar fields) {φa }. qψ = −1 and Dψ = −iψ. ∂µ J µ = 0 and the µ → 0 limit of Eq.

the axial vector current ψγ µ γ5 ψ isn’t associated with an internal symmetry and is not conserved. So we might try the following interaction terms: • Fermions: /ψ LI = −gψγ µ Aµ ψ = −gψA (XI. • Charged scalars: LI = −ig [(∂ µ ϕ∗ )ϕ − (∂ µ ϕ)ϕ∗ ] Aµ (XI. For massive fermions.R = ψ L. Since any linear combinations of 2 µ these currents are conserved.52) (XI.54) The situation here is not as nice as it was for fermions. (XI. For massless fermions.51). jL.178 and so the conserved current is j µ = ψγ µ ψ. only the vector current ψγ µ ψ is conserved. and therefore this theory is not expected to have a smooth µ → 0 limit.17 Therefore we expect that only the theory where the vector field couples to the vector current will have a smooth µ → 0 limit.50) Therefore. we switch our notation for charged scalars at this point from ψ to ϕ. 16 17 To avoid confusion with fermi fields ψ. the conserved current is no longer given by Eq. so it changes the canonical momenta of the theory.R = 1 ψγ µ (1 γ5 )ψ.R γ µ ψL. Πµ ∗ = ∂ µ ϕ ϕ ϕ and so the conserved current is j µ = −i(∂ µ ϕ∗ )ϕ + i(∂ µ ϕ)ϕ∗ . we might hope that if we couple a vector field Aµ only to these conserved currents we will obtain a theory with a well-defined µ → 0 limit. For a charged scalar field16 ϕ we have Πµ = ∂ µ ϕ∗ . it is possible to couple a massless vector field to the axial vector current in the special case of massless fermions.53) This is the interaction we had written down earlier. . thereby changing the expression for j µ . (XI. So although this theory still has a U (1) symmetry and a conserved current. The interaction term contains derivatives of the fields. but with Γ = 1.51) (XI. Thus. we saw in the chapter on the Dirac Lagrangian that the theory has two U (1) symmetries µ and therefore two conserved currents. The mass term breaks the axial U (1) symmetry associated with ψγ µ γ5 ψ but not the vector U (1). both ψγ ψ and ψγ µ γ5 ψ are conserved.

Dµ φa ). LM is still invariant under the U (1) transformation. Dµ φa ). ∂µ φa ) is invariant under the U (1) symmetry. This proves the first assertion. ∂µ φa ). For quantum electrodynamics. where Dµ φa ≡ ∂ µ φa + ieAµ qa φa (XI. if we choose the dimensionless coupling constant e to be the fundamental electric charge. Under a U (1) transformation. LM (φa . and 2.56) and so Dµ φa transforms in the same way as ∂µ φa . Dµ φa → Dµ e−iλqa φa = ∂µ e−iλqa φa + ieAµ qa e−iλqa φa = e−iλqa Dµ φa (XI. which is invariant under the U (1) transformation φa → e−iλqa φa . (Note that again there is an ambiguity in the qa ’s. .57) (XI. so is L(φa .) The resulting Lagrangian has the following two properties: 1. Therefore if L(φa . Fortunately. This is straightforward to show. because the coupling itself will in general change the expression for the current. this just corresponds to the freedom to choose the overall coupling constant for the interaction term. That is ∂LI = −ej µ ∂Aµ and ∂µ j µ = 0. then q will be the electric charge of the field measured in units of e. there is a magic prescription which guarantees that Aµ always couples to a conserved current. replace it by LM (φa . Dµ is called the gauge covariant derivative.179 We see from the scalar case that it’s not always so easy to ensure that Aµ always couples to a conserved current.55) (no sum on a). Given a Lagrangian as a function of the fields and their derivatives.58) (XI. It is called minimal coupling. Aµ is coupled to a conserved current. Minimal Coupling The minimal coupling prescription is very simple. 1.

Similarly. (XI. (XI.60) a ∂LM ∂(Dν φa ) ∂(Dν φa ) ∂Aµ ∂LM ieqa φa ∂(∂ν φa ) Πµ ieqa φa a = a = −ej µ as required. when acting on the piece of the field which creates an outgoing . the conserved current is jµ = a Πµ Dφa = a a Πµ (−iqa φa ). Na¨ ıve ıvely.61) Going back to our examples.63) linear in Aµ is slightly subtle. we notice that a derivative ∂µ acting on the piece of the field which annihilates an incoming state (and so has a factor of exp(−ip·x)) brings down a factor of −ipµ . with the Feynman rule in Fig. The Feynman rule for the term in Eq.63) with all the interaction terms given by minimal coupling has a well-defined limit as µ → 0. This will lead to a new kind of vertex.62) which is just what we had before. (XI. proving the second assertion. the minimally coupled Dirac Lagrangian for a fermion with charge q (in units of the elementary charge e) is L = ψ(iD − m)ψ = ψ(i∂ − eqA − m)ψ / / / (XI. we will denote charged scalars by dashed lines in this chapter). However. the minimally coupled scalar Lagrangian for a scalar with charge q is L = Dµ ϕ∗ Dµ ϕ − m2 Aµ Aµ = (∂µ − ieqAµ )ϕ∗ (∂ µ + ieqAµ )ϕ − m2 Aµ Aµ = ∂µ ϕ∗ ∂ µ ϕ − ieqAµ (ϕ∗ ∂ µ ϕ − ϕ∂ µ ϕ∗ ) + e2 q 2 Aµ Aµ ϕ∗ ϕ.59) From the definition of the gauge covariant derivative.63) The term linear in Aµ is what we had before. (XI. a (XI. we also have ∂(Dν φa ) µ = ieqa φa δν ∂Aµ and so we find ∂LI ∂LM = = ∂Aµ ∂Aµ = a (XI. (XI. but there is a new term quadratic in Aµ . the so-called “seagull graph” (to avoid confusion with fermion lines. but it turns out that the na¨ approach gives the correct answer.180 In terms of the canonical momenta.4). Only the theory defined by Lagrangian Eq.

64) Substituting the field expansion in terms of creation and annihilation operators. XI. XI. (XI. in Dyson’s formula the derivative cannot be pulled out of the time ordered product. However. µ p' iqe(p+p')µ p FIG. Therefore. it brings down a factor of ipµ . we expect the Feynman rule shown in Fig. it turns out (we won’t prove this here) that these two problems cancel one another. There are two problems with this derivation.181 ν µ 2ie2q2gµν FIG. ıve We now justify the assertion that the coupling constant e arising in the covariant derivative is just the elementary charge.5).4 The “seagull graph” for charged scalar-photon interactions. state. Second. (XI. We will show this for the Dirac field. The same is true for Dirac fields: the conserved charge is Q = = d3 x ψγ 0 ψ d3 x ψ † ψ. indicating that particle and antiparticle carried opposite charges. it is straightforward to show that this is Q = r d3 k b (r)† (r) b k k −c (r)† (r) c k k = Nb − Nc . Recall that in the scalar case the U (1) charge was just proportional to the number of particles minus the number of antiparticles. First of all.65) . (XI. and so changes the canonical commutation relations. and that the na¨ Feynman rule is actually correct.5 Feynman rule for the derivately coupled charged scalar. the derivative interaction changes the canonical momenta in the theory.

and the electromagnetic four-current µ is therefore je.m. = qeψγ µ ψ.66) µ je. and LM is said to have a U (1) gauge symmetry.70) for any space-time dependent function λ(x). where α= and so in natural units e= √ 4πα.m ν which is just Maxwell’s equations in the presence of an electromagnetic current je. = −e d3 x ψ † (x)ψ(x) (XI. = qeQ. . It is invariant under the strange-looking transformation λ(x) : φa (x) → e−iqa λ(x) φa (x) 1 Aµ (x) → Aµ (x) + ∂µ λ(x) e (XI. Gauge Transformations The minimally coupled Lagrangian LM (φa .m.035 (XI. Note that when λ(x) is a constant and not a function of space-time this is just the usual U (1) transformation on the φa ’s (note that the Aµ field is invariant if λ is constant).68) 2.m. The transformation Eq. The Euler-Lagrange equation for a massless vector field Aµ is therefore ∂µ F µν = −eψγ ν ψ ν = je.69) e2 1 = 4π¯ c h 137. ∂µ φa ).m. (XI. since λ is the same at all points.67) The elementary charge is often expressed in terms of the “fine-structure constant” α. The odd transformation law of the . This kind of symmetry is called a global U (1) symmetry. = −eψγ µ ψ. Dµ φa ) is invariant under a much larger group of symmetries than LM (φa . (XI. the electric charge and electromagnetic current are Qe.70) is called a local or gauge transformation. the theory is invariant under different U (1) transformations at each point in space-time. Thus. (XI. Since λ(x) is now a function of space-time. the total electric charge of the system is therefore Qe. for electrons (q = −1).182 If the particles annihilated by ψ have electric charge q in units of the elementary charge e.m.

e−iλ(x)qa φa (x) e (XI. since ψ∂ ψ picks up a term proportional to (∂µ λ) ψγ µ ψ. making it difficult to quantize the massless theory directly. − 1 Fµν F µν + 4 indices. Thus. the “matter” (fermions and scalars) Lagrangian. In quantum electrodynamics. the vector meson mass term 1 (∂µ ∂ν λ(x) − ∂ν ∂µ λ(x)) = Fµν . LM (φa . Therefore the problem of finding the time-evolution of the fields from some initial values is ill-defined.72) µ2 µ 2 Aµ A .74) for some arbitrary function λ(x). e is not gauge invariant: (XI. Since Fµν is antisymmetric in its µ2 µ 2 Aµ A 2 1 λ(x) : Aµ Aµ → Aµ Aµ + ∂µ λ(x)Aµ + ∂µ λ(x)∂ µ λ(x). The problem arises at the classical level: if {Aµ (x). it is also gauge invariant. I can never uniquely predict the field configuration at some later time.71) Therefore unlike the usual derivative ∂µ φa . if LM (φa . e e The complete Lagrangian is only gauge invariant when µ = 0. derivatives).. Therefore..183 Aµ fields is crucial here: the Dirac Lagrangian is not invariant under gauge transformations. Rather than being a help in solving the theory. their first. every time we use the minimal coupling prescription we end up with a theory in which LM is invariant under a gauge symmetry. λ(x) : Fµν → Fµν + However.73) (XI. the photon is massless and so the theory has exact gauge invariance. second. Dµ φa ) is invariant under a gauged U (1) transformation. third . the gauge covariant derivative Dµ φa transforms in the same way under a gauge transformation as it does under a global transformation. (XI. which is the vector theory we are really interested in. So far we have just looked at LM . φa (x)} is a set of fields which form a solution to the equations of motion then so is the set 1 Aµ (x) + ∂µ λ(x). ∂µ φa ) is invariant under a global U (1) transformation. No matter how much initial value data I have at t = 0 (the fields. and ignored the free part of the vector Lagrangian. since their exist an infinite number of gauge transformed solutions of the equations . this gauge invariance complicates things tremendously. The transformation property of Aµ is chosen / precisely to cancel this term: Dµ φa = (∂µ + ieAµ qa )φa → (∂µ + ieAµ qa + iqa ∂µ λ(x)) e−iqa λ(x) φa = e−iqa λ(x)) (∂µ − iqa ∂µ λ(x) + ieAµ qa + iqa ∂µ λ(x))φa = e−iqa λ(x) Dµ φa .

then another solution to the equations of motion is Aµ (x) = Aµ (x)+∂µ λ(x)/e. which satifies 1 ∂µ Aµ (x) = 2λ(x) = 0. any observable is gauge invariant. D. However. The Limit µ → 0 Since we are avoiding quantizing the gauge invariant massless theory directly. Two systems different by a gauge transformation contain identical physics. in the massless theory the condition ∂µ Aµ = 0 implied by the Proca equation is no longer implied by the equations of motion.75) The four dimensionally longitudinal mode which we had banished has come back to haunt us. With any luck the minimal coupling prescription 18 See Chapter 5 of Mandl & Shaw for a discussion of the Gupta-Bleurer method of canonically quantizing the massless theory. they just differ in the choice of description. Some popular gauge choices are · A = 0 (Coulomb gauge) ∂µ Aµ = 0 (Lorentz gauge) A0 = 0 (temporal gauge) A3 = 0 (axial gauge) The trick is then to canonically quantize the theory in the given gauge. ∂µ Aµ is no longer zero. Furthermore. as expected. that is. we will instead derive the Feynman rules for Quantum Electrodynamics by examining at the µ → 0 of the theory of a minimally coupled massive vector field. F µν is gauge invariant. subject to the corresponding constraint18 . and therefore so are the electric and magnetic fields E and B. as we will see in the next section. physical amplitudes are independent of gauge. In perturbation theory. Things are not actually so badly defined. In fact. e (XI. but arbitrary. So we can fix the description by fixing the gauge once and for all. These field configurations just differ by a gauge transformation which vanishes at t = 0. these terms in the propagator do not contribute in a minimally coupled theory and so. If Aµ (x) is a solution to the equations of motion satisfying ∂µ Aµ = 0. .184 of motion which also have the same initial value data. different choices of gauge result in extra terms in the photon propagator proportional to k µ k ν .

(XI. In the limit µ → 0 this is just the pair production process e+ e− → µ+ µ− in QED.6 Feynman diagram contributing to e+ e− → µ+ µ− . Therefore. The 1/µ2 term in the amplitude is p-'. Although we have just demonstrated it in one process. / / µ (k − µ2 + i ) p+ p− p− p+ i But by the Dirac equation. in this section we will see by direct calculation that the factors of 1/µ2 in the quantum theory do not contribute to amplitudes when the theory is minimally coupled. and that the massless theory does in fact have sensible Feynman rules. There is only one graph at O(g 2 ) which contributes to this process. shown in Fig.78) and so vk u = 0. / / / (r) (s) (r) (s) (r) (s) (XI. this is a very general / feature of minimal coupling. this is no accident (there are no accidents in field theory). XI. (XI. r' p+. It just follows from current conservation: ∂µ j µ = 0 ⇒ 0 |∂µ j µ | e+ e− = 0 |∂µ eγ µ e| e+ e− = ∂µ (v p+ γ µ up− e−i(p+ +p− )·x ) = −i(p+ + p− )v p+ γ µ up− e−i(p+ +p− )·x / = −iv p+ k up− e−i(p+ +p− )·x (r) (s) (r) (s) (r) (s) (XI. e2 (r) µ (s) kµ kν (r ) (s ) v γ up− 2 u γ ν vp 2 p+ + µ k − µ2 + i p − 2 e (r) (s) (r ) (s ) = i 2 2 v ku u kv . where e and µ are two different fermion fields (electrons and muons). with the propagator shown in Fig. First consider the process e+ e− → µ+ µ− .77) So this term vanishes before taking µ to zero. s p+'. v p+ k up− = v p+ (p− + p+ )up− = v p+ (me − me )up− = 0. s' p-. Of course. minimally coupled to a massive gauge boson.6). and it means that in such theories we can completely ignore the piece of the propagator proportional to k µ k ν . Indeed.76) (XI. . r FIG.185 will have solved the problems we previously noted in taking this limit.7). in the µ → 0 limit the vector boson is the photon of quantum electrodynamics.

) This corresponds to the fact that classical electromagnetic waves . Jz = ±1. µ ν Squaring and summing over final spins of the vector particles and averaging over initial spins will give a result proportional to Mµν Mαβ −gµα + kµ kα µ2 −gνβ + kν kβ µ2 (XI. you might worry about the factor of 1/µ2 in the polarization sum. kν Mµν = 0. Eq. Two diagrams contribute to this process. (s) (s ) (s ) (XI. For a massive particle you can always boost to its rest frame and perform a rotation to change a Jz = 1 state to a Jz = 0 state. this is only possible because the photon is massless. 1. the contributions from these terms vanish: kµ Mµν ∝ up = up = up = up (s ) k (p + k + m)γ ν / / / γ ν (p − k + m)k / / / + 2p · k −2p · k up (s) (s ) γ ν (mk − k p + 2p · k ) (s) / // (−p k + mk + 2p · k )γ ν // / − up 2p · k 2p · k (−mk + mk + 2p · k )γ ν / / γ ν (mk − k m + 2p · k ) (s) / / − up 2p · k 2p · k [γ ν − γ ν ] up = 0. Decoupling of the Helicity 0 Mode The result that kµ Mµν = 0 has another consequence in the µ → 0 limit. A massive spin 1 particle has three spin states. just as before. (XI.80) and so the terms proportional to kµ M µν and kν M µν look bad as µ → 0. and so the k µ k ν term doesn’t contribute to the polarization sum.7 The photon propagator. We can demonstrate this by looking at Compton scattering of a massive vector boson off an electron. 0.79) ≡ Mµν ε(r ) (k )ε(r) (k). whereas a massless gauge boson like the photon only has two helicity states.23). XI. However.81) Similarly. giving iA = −ie2 up (s ) ∗ γ µ (p + k + m)γ ν / / γ ν (p − k + m)γ µ (s) (r ) ∗ / / + up εµ (k )ε(r) (k) ν 2 − m2 (p + k ) (p − k)2 − m2 (XI. In a similar vein. but a similar thing happens here. ±1 (once again. V e− → V e− .186 ν µ -i gµυ k2+iε FIG.

) How is this apparently discontinuous behaviour possible if the µ → 0 limit of the theory is smooth? A massive vector particle travelling in the z direction has four-momentum k µ ( k 2 + µ2 . 0. satisfying ε(3) · ε(3) = −1. −i. The absence of a longitudinal mode corresponds to the absence of a (3dimensionally) longitudinal photon.85) k = − O µ µ → 0. µ ∗ (XI.82) ε(3) is the 3D longitudinal polarization state. ε(3) · k = 0. ∝ k. as a result of current conservation. (XI. We discovered that the massless limit is in general ill-defined. (Nothing special happens to the transverse modes in this limit). 0) 1 ε(3) = (k. 1. QED At this point it is worth summarizing our results. 0. We set out to find a quantum theory of a massless vector field. where ε(1) = (0. i. the amplitude to produce a 3D longitudinally polarized vector meson vanishes as µ → 0. . k) and three possible polarization states εµ .187 are always transverse. k Therefore. in the massless theory. and for µ = 0 is absent as a physical state. (We call mode satisfying ε ∝ k “3D longitudinal” to distinguish it from the 4D longitudinal mode discussed earlier. 0.83) The amplitude to produce the helicity 0 state is therefore proportional to ε(e) Mµ = µ ∗ 1 M3 −k 1 + O µ µ2 k2 → 0 as µ2 k2 +k 1+O µ2 k2 (XI. The amplitude for a longitudinal vector boson to be produced in any process (like the Compton scattering process from the previous section) is proportional to ε(3) Mµ µ where the tensor Mµ satisfies k µ Mµ = 0 ⇒ M0 = −M3 k = −M3 1 + O k 2 + µ2 µ2 k2 . 0. both of which are 3D transverse. The helicity zero mode smoothly decouples in this limit. the photon. The appropriate form of the polarization sum is 2 r=1 (r) ε(r) εν = −gµν . ε(3) ∝ k. 0) ε(2) = (0. there are only two physical polarization states of the vector boson. 1. Therefore. k 2 + µ2 ) µ (r) = (XI.84) ∗ (XI. ε ∝ k.86) E.

8 Feynman rules for QED .r p.r v(r) p v(r) p _ p Fermion Propagator p. The resulting theory is quantum electrodynamics. The QED Lagrangian for a theory with a single charged scalar ϕ and a single charged fermion ψ with charges qϕ and qψ respectively. the quantum theory of electromagnetism.r i(p+m) / p2-m2+iε Incoming/outgoing Fermions (particles) i p2-µ2+iε k µ -i gµυ p2+iε Incoming/outgoing Fermions (antiparticles) p Scalar Propagator ν p Photon Propagator µ µ k εµ(r)(k) εµ(r)*(k) Incoming/outgoing Photons FIG. This requirement gave us the interaction between the photon and charged scalars and Dirac fields (up to the overall coupling constant).188 unless the vector field couples to a conserved current. is 1 L = Dµ ϕ∗ Dµ ϕ − µ2 ϕ∗ ϕ + ψ (iD − m) ψ − F µν Fµν / 4 ∗ µ ∗ µ µ ∗ 2 = ∂µ ϕ ∂ ϕ − ieqϕ Aµ (ϕ ∂ ϕ − ϕ∂ ϕ ) + e2 qϕ Aµ Aµ ϕ∗ ϕ 1 +ψ (i∂ − m) ψ − eqψ ψA − F µν Fµν / /ψ 4 The Feynman rules for the Feynman amplitude iA in QED are illustrated in Fig.8).87) µ µ p' ν p 2ie2q2gµν µ -ieqγµ -ieq(p+p')µ Fermion-Photon Vertex Charged Scalar-Photon Vertices p.r u(r) p u(r) p _ p. XI. (XI. (XI.

or scalar-scalarphoton-photon) write down the appropriate factor. it was possible to take the µ → 0 limit. include the appropriate factor of the four-spinor or polarization vector.4 (lepton pair production). To see why this is so. they follow the fermion line from the end of an arrow to the start. the cancellation in Eq. Note than M&S use differently normalized fermion fields than we do. scalar-scalar-photon. and instead of being suppressed.6 (Compton scattering). reading from left to right. 3. 6. Renormalizability of Gauge Theories In this chapter we started with the theory of a massive vector boson and showed that. There are additional rules for diagrams with loops. 2. Multiply the expression by a phase factor δ which is equal to +1 (−1) if an even (odd) number of interchanges of neighbourhing fermion operators is required to write the fermion operators in the correct normal order. For each external fermion or photon. I will include a few worked examples for QED processes. In some future version of these notes.85) does not occur. that of gauge invariance.189 1. The four-momenta associated with the lines meeting at each vertex satisfy energy-momentum conservation. In the meantime. 4. (XI. If a vector boson is coupled to a non-conserved current. . 8. The spinor factors (γ matrices. which we have not considered because we are just working at tree-level in this theory. F. the final answers are independent of normalization. For each internal line. you should read Chapter 8 of Mandl & Shaw and work through Sections 8. in which case the theory had a larger symmetry. Now we will go one step further and assert that gaugeinvariance is required in order for a theory of vector bosons to make sense as a fundamental theory (I will explain what I mean by “fundamental” in a moment). the amplitude to produce a helicity 0 mode grows like k/µ. let’s go back to the discussion of the previous section. For each interaction vertex (fermion-fermion-photon.5 (Bhabha scattering) and 8. however. include a factor of the corresponding propagator. four-spinors) for each fermion line are ordered so that. 5. despite appearances.

What is fascinating is that the theory carries within itself the seeds of its own destruction. the vector boson could be revealed not to be fundamental. At some scale (typically set by the mass of the particle. and so its dynamics would change at the scale of order the size of the particle. Similarly. Without getting into the details of radiative corrections. since that’s the only dimensionful parameter in the theory). renormalizability is the extension of the above discussion to radiative corrections. is related to a property known as “renormalizability. this will affect loop graphs even for low-energy processes. You recall that internal loops in a Feynman diagram come with a factor of d4 k (2π)4 since the momentum running through the loop is unconstrained. As a result. If we are interested in fluid dynamics.for example. Many aspects of nuclear physics can be described by a field theory of nucleons and pions. In our case. what the theory is telling us is that a theory of massive vector mesons coupled to a nonconserved current is a perfectly fine theory at low energies. then. but as an effective field theory. but that it can’t be valid up to arbitrarily high energy scales (that is. despite the fact that we know that these particles aren’t “fundamental. It makes much more sense to consider the fluid as a continuous medium. because at some energy the probability will become greater than 1! (This is known as “unitarity violation”).not as a fundamental theory (valid up to arbitrary energy scales). There is nothing a priori wrong with this. for example. that it be valid up to arbitrarily high energies and so not predict its own demise. that if the theory breaks down at a certain scale. This property of a theory. we don’t have to consider the single atoms which make up the fluid. we just have to interpret our theory a bit differently . At this energy the theory has clearly stopped making sense (at least perturbatively). I will just assert that if one attempts to calculate loop graphs in a theory of a massive vector boson coupled to a non-conserved current one is again led to the conclusion that the theory cannot be . the probability of producing this mode grows with increasing energy without bound. if we are interested in the hydrogen atom we can treat the proton as a point object. down to arbitrarily short distances). this theory has to break down in some way .” Roughly speaking. This is in fact a Bad Thing. but to be a composite particle. This kind of thing is very familiar in physics.190 Thus. even if the process being considered is a low-energy process.” It all depends on the scale of physics we’re interested in. despite the fact that we know it is made up of quarks and gluons. It is perhaps not surprising. arbitrarily high momenta run through loop graphs.

You showed on an early problem set that in 4 dimensions. we can only do experiments at finite energies. a cross-section has units of area. there is no reason to only consider theories which are valid up to arbitrary energy scales. This is why we only considered very simple interaction terms . φ2 ψψ (dimension 6. Since the cross-section grows without bound. but interactions like ψψψψ. By dimensional analysis. Imagine a theory with a four-fermion interaction term LI = − g ψψψψ. respectively) are not.anything more complicated leads to a non-renormalizable theory. Therefore interaction terms like φ4 . Just by dimensional analysis. which I have made explicit by writing it as g/M 2 (g is dimensionless). Now. respectively) are allowed in a fundamental theory. φ5 . just by unitarity arguments. So what are the requirements for a theory to be renormalizable? It is easy to get an idea. we conclude that the dimensions of ψ are [mass]3/2 (so that mψψ has the right units). at high energies where we may ignore the fermion masses. once again the probability must grow without bound. Of course. Thus. From the Dirac Lagrangian. However. 5 and 5. This answers a question which may have been bothering you all along in this course: why do we always study theories with such simple interaction terms? Why can’t we have an interaction term like −gψψ cos ln (1 + ϕ/M )? The answer is that this is not a renormalizable interaction. you can see that this will happen in ANY theory with coupling constants which are inverse power of a mass. since higher-dimension operators come with coupling constants which are proportional to inverse powers of the scale at . and again the theory must break down at some energy scale set by M . for a theory to be renormalizable. ψψφ and φ3 (dimension 4. 4 and 3. so the coupling must have dimensions [mass]−2 . or [mass]−2 . ψψψψ has dimension [mass]6 . the Lagrangian density has dimensions of [mass]4 . the cross section for fermion-fermion scattering in this theory must be proportional to 1/M 4 . all terms in the Lagrangian must have mass dimension ≤ 4.191 fundamental. M2 Now let’s do some dimensional analysis.88) where s = (p1 + p2 )2 is the squared centre of mass energy of the collision. After all. Since the amplitude from the four-fermion interaction goes like 1/M 2 . we must therefore have σ∝ s M4 (XI.

there are four Higgs bosons. they couple to electrons and electron-neutrinos via the following interaction: + − LI = −g1 νγ µ (1 − γ5 )eWµ + eγ µ (1 − γ5 )νWµ −g2 Z µ (eγµ (gV − gA γ5 )e + νγµ (1 − γ5 )ν) (XI. gV and gA are coupling constants which are related to the electric charge and the ratio of the W ± and Z 0 boson masses.192 which the theory breaks down. So the W and Z clearly can’t be fundamental. and the fourth of which is just waiting to be .2 GeV and 91. Experimentally. Their effects are proportional to the ratio of the momentum of the process to the scale of new physics. The theory of a massive vector boson which we studied in this section. Furthermore. there certainly are massive vector bosons coupled to nonconserved currents in the world. Since the electron is massive. The gauge bosons associated with the weak interactions. Finally.” In the minimal theory. Furthermore. while the longitudinal components are made of a scalar particle.2 GeV. The question of what the W ’s and Z’s are made of is the foremost question in particle theory at the moment. or baryon number in GUTs) they can usually be safely ignored. g2 . as you may be aware. only theories with massless vector bosons are renormalizable. respectively. and so the corresponding quantum theory is nonrenormalizable. three of which are incorporated into the two W ’s and the Z. the current eγ µ γ5 ν is not conserved. The simplest possibility which leads to a renormalizable theory is that of the minimal Weinberg-Salam model. Now. while useful for obtaining the Feynman rules for QED. the W ± and Z 0 . since a vector meson mass term breaks the gauge symmetry.89) where g1 . is not a renormalizable theory. It is because of renormalizability that gauge symmetries are so crucial in field theory: the only way to couple a vector field to other fields in a renormalizable way is through a gauge covariant derivative. This is at the root of the difficulty of quantizing gravity: the graviton is a spin-2 field (corresponding to quanitizing the metric tensor gµν ). so we have a theory of gauge bosons coupled to nonconserved currents. known as the “Higgs Boson. the theory predicts that unitarity violation due to excessive production of longitudinal W ’s and Z’s will occur at a scale of about 3 TeV=3 × 103 GeV. it can also be shown that theories with fields of spin > 1 are also non-renormalizable. have masses of 80. in which the transverse components of the W and Z are fundamental (corresponding to massless vector bosons). the effects of these terms are negligible at low energies. Unless they break symmetries which are preserved by the renormalizable terms (such as parity in the case of the weak interactions.

.193 experimentally observed.there are many others. But this is only the simplest possibility .

You're Reading a Free Preview

/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->