Beruflich Dokumente
Kultur Dokumente
R E L AT I V I T Y , PA R T I C L E S , F I E L D S
Beachten Sie bitte, dass das Skript noch nicht gründlich Korrektur gelesen wurde und
es sicher noch Fehler gibt. Bitte schicken Sie sehr willkommene Vorschläge, Korrekturen,
Kommentare an:
andreas.weiler@tum.de
Thanks to Tobias Duswald, Nepomuk Ritz, Patrick Selle, Ennio Salvioni, Stefan Schuldt,
and Christian Schuster for helping to improve the script!
1 Foreword 9
1.1 Units and conventions 9
2 Introduction 11
2.1 Why quantum field theory? 11
2.1.1 Necessity of the field viewpoint: causality 12
2.1.2 General features of relativistic quantum field theories 13
2.2 String Theory 13
2.2.1 The discrete chain 13
2.2.2 The continuous string 14
2.2.3 Quantum string theory 16
2.2.4 Review of the simple harmonic oscillator 17
2.2.5 Fock space of the discrete chain 18
3 Relativity 23
3.1 Galileo and Newton 23
3.1.1 Maxwell vs. Newton 24
3.1.2 Derivation of the Lorentz transformations 24
3.2 Consequences of special relativity 25
3.2.1 Relativity of simultaneity 26
3.2.2 Time dilation 27
3.2.3 Lorentz contraction 27
3.2.4 Addition of velocities 28
3.3 Minkowski space 29
3.3.1 Fourvectors 30
3.3.2 Minkowski Metric 31
3.3.3 Discrete transformations 32
3.3.4 Motion of a free particle 32
4 andreas weiler, tum
5 Canonical Quantization 53
5.1 Free fields 53
5.1.1 The free scalar field 54
5.1.2 Infinities 55
5.1.3 Normal ordering 57
5.2 Particles 58
5.2.1 Multi-particle states and Bose statistics 59
relativity, particles, fields 5
9 Outlook 175
1
Foreword
and many more. I strongly suggest that you find a book (or books)
you like from the ones above and study it as a complement to these
lecture notes.
This course provides a hopefully gentle introduction into the
beautiful world of quantum field theory while it familiarizes you
with relativity. It is the first real theory course in that it is the
first course to tackle the theoretical framework that underlies all of
nature. Let’s go.
The whole point of units is that you can choose whatever units are
most convenient! In particle physics and cosmology, we use “natural
units”
~=c=1 (1.4)
and so
Ψa = Ψa (x, t) (2.1)
∆p ≥ ~/L
where in the last step we have used a general gaussian integral3 with 3
A very useful formula:
A = it/m 1 and B = i(x − x0 ) This amplitude is non-zero for all x Z −1
n
P
Aij xi xj +
n
P
Bi xi
2
n i,j=1 i=1
and t, which implies that a particle can propagate between any two d xe
where we have chosen the z-axis along the x − x0 direction. We now The main idea of stationary phase
want an approximate solution for this integral for x2 t2 (far out- methods relies on the cancellation
of sinusoids with rapidly varying
side the light-cone). We will use the stationary phase approximation. phase. If many sinusoids have the
The phase function same phase and they are added
together, they will add constructively.
p If, however, these same sinusoids
φ(p) ≈ px − t p2 + m2 (2.10) have phases which change rapidly
as the frequency changes, they will
add incoherently, varying between
constructive and destructive addition
at different times.
https://en.wikipedia.org/wiki/
Stationary_phase_approximation
relativity, particles, fields 13
• Spin-statistics
R=Na (2.20)
a → 0, N → ∞, R = N a = const. (2.21)
Z " 2 2 #
R
1 ∂q(x, t) ∂q(x, t)
L= dx ρ −σ (2.26)
2 0 ∂t ∂x
with ωn = c kn and
πn
kn = (2.33)
R
The normal-modes reproduce the discrete chain as long as n/N
!1.0
0
0.8
discrete chain
0.6
0.4
0.2
0.0
0 2 4 6 8
This is a first encounter with the important concept of renormalization. The cut-off frequency
ωC ∼ 1/a
is a theoretical necessity. The cut-off tells us until which scale, we can trust our continuum theory. A continuous
string at frequencies ω ωC is not a good description of the discrete chain: the minimal wavelength is set by the
atomic distance, below which there is nothing to oscillate. Without a cut-off the specific heat of the string would
diverge since at a temperature T each mode (equipartition theorem in statistical physics) contributes an amount
kT , where k is the Boltzmann constant.
Here, we cannot determine the value of the cut-off from the long-wavelength theory alone, because only the
combination
c ∼ a ωC
appears. We can absorb the cut-off in measurable parameters, as in Eq. (2.25). This is called renormalization. If
this can be done, a theory is called renormalizable. You will discuss this at length in the chapter of loop correc-
tions in quantum field theories.
If a theory is non-renormalizable, the behavior can be sensitive to short-distance physics, which in this case would
be the atomic motion. This behavior does appear random on macroscopic scales, as in the propagation of a crack
or the nucleation of a raindrop.
We will switch to periodic boundary conditions now, Think of the system as being on a
torus. The spacing of the frequencies
q̂j+N (t) = q̂j (t) (2.37) is doubled but each frequency ap-
pears two-fold. We are not interested
since they allow us to use just the exponentials in the behaviour at the endpoints
and we will be at large N .
N/2
1 X
q̂j (t) = √ Q̂n ei2πnj/N (2.38)
N n=−N/2
N/2
1 X
p̂j (t) = √ P̂n ei2πnj/N (2.39)
N n=−N/2
4κ n N/2 N/2
ωn2 = sin2 π (2.44) =
X
P̂n P̂−n =
X
P̂n P̂n†
m N
n=−N/2 n=−N/2
As you remember from your QM course, we can write the energy- N
X
eigenvalues of the harmonic oscillator with frequency ω as = P̂n† P̂n
n=1
1 where in the last step we have used
n = ~ω +n , n = 0, 1, . . . (2.45) that P̂n† and P̂n commute and that
2
the product P̂n† P̂n therefore is
symmetric n → −n.
where n is the occupation number. We can therefore write the energy- The orthogonality relation follows
eigenvalues of H as a set of occupation numbers from the geometric series.
N
N/2
X X
ei2π(n+m)j/N =
1 − ei2π(n+m)j
=0
1 1 − ei2π(n+m)j/N
Eα = ωn + αn (2.46) j=1
2
n=−N/2
because we have ei2π(n+m)j = 1
since n, m ∈ N and ei2π(n+m)j/N 6= 1
with αn = 0, 1, 2, . . .. for n + m 6= 0 . The case n + m = 0 is
trivial.
Ex: Show the result for the qj
2.2.4 Review of the simple harmonic oscillator
substitution!
Consider the quantum mechanical Hamiltonian
1 2 1 2 2
H= p + ω q (2.47)
2 2
with the canonical commutation relations
[q, p] = i (2.48)
[a, a† ] = 1 (2.53)
18 andreas weiler, tum
. . . , E − ω, E, E + ω, E + 2ω, . . . (2.61)
The energy is bounded from below and there must be a ground state
|0i which satisfies
a|0i = 0 (2.62)
1
H|0i = ω|0i (2.63)
2
With this definition we find that the ladder operators obey the
commutation relations.
h i r r
mωk i mωk0 i
ak , a†k0 = Qk + Pk , Q−k0 − P−k0
2 mωk 2 mωk0
√
m ωk ωk 0 i i 1
= [Qk , Q−k0 ] − [Qk , P−k0 ] + [Pk , Q−k0 ] + 2 [Pk , P−k0 ]
2 | {z } mωk0 mωk m ωk ωk0 | {z }
=0 =0
√ h i
m ωk ωk 0
i † i h † i†
= P 0 , Qk − P , Q 0
2 mωk0 k k k
| {z } mωk
=−iδkk0
√
m ωk ωk 0 δkk0 δkk0 mωk 2
= + = δkk0 = δkk0
2 mωk0 mωk 2 mωk
† †
where
h we
i have also used that P−k = Pk , Q−k = Qk and that
Pk†0 , Qk = −iδkk0 .
r r
mωk i mωk0 i
[ak , ak0 ] = Qk + Pk , Qk0 + Pk 0
2 mωk 2 mωk0
√
m ωk ωk0 i i 1
= [Qk , Qk0 ] + [Pk , Qk0 ] + [Qk , Pk0 ] − 2 [Pk , Pk0 ]
2 | {z } mωk mωk0 m ωk ωk0 | {z }
=0 =0
[ak , ak0 ] = 0.
N/2
X
1
H= ωk a†k ak + (2.69)
2
k=−N/2
20 andreas weiler, tum
The energy is bounded from below and there must be a ground state
|0i which satisfies
We started with a discrete chain for pedagogical reasons. Of course we do not believe that the fields observed
in Nature, such as the electron field or the photon field, are actually constructed of point masses tied together
with springs. A modern picture which was first used by Landau-Ginzburg and fully understood by S. Wein-
berg, is that we start with the desired symmetry, say Lorentz invariance or a conserved U (1) charge (more on
both later) decide on the fields we want by specifying how they transform under the symmetry (in this case we
decided on a scalar field φ(x, t) as the local displacement) and then write down the action involving no more than
two time derivatives – because we don’t know how to quantize actions with more than two time derivatives (we
usually end up with ghosts). See also the discussion in the statistical physics script (in particular the chapter:
Landau-Ginzburg Theorie)
3
Relativity
for fixed initial and final conditions, the second term is just an
irrelevant additional constant, like the last term. We can generalize
this to a many particle system with interactions
Z X 1 2 X
dxa
S = dt ma − V (xa − xb ) (3.12)
2 dt
a a6=b
The Michelson–Morley experiment was performed over the spring and summer of 1887 by Albert A. Michel-
son and Edward W. Morley at what is now Case Western Reserve University in Cleveland, Ohio, and published
in November of the same year. It compared the speed of light in perpendicular directions, in an attempt to detect
the relative motion of matter through the stationary luminiferous aether (‘’aether wind”). The result was negative,
in that the expected difference between the speed of light in the direction of movement through the presumed
aether, and the speed at right angles, was found not to exist; this result is generally considered to be the first
strong evidence against the then-prevalent aether theory, and initiated a line of research that eventually led to
special relativity, which rules out a stationary aether.
Michelson–Morley type experiments have been repeated many times with steadily increasing sensitivity. These
include experiments from 1902 to 1905, and a series of experiments in the 1920s. More recent optical resonator
experiments confirmed the absence of any aether wind at the 10−17 level. Together with the Ives–Stilwell and
Kennedy–Thorndike experiments, Michelson–Morley type experiments form one of the fundamental tests of special
relativity theory. https://en.wikipedia.org/wiki/Michelson-Morley_experiment
(ct)2 = x2 + y 2 + z 2 (3.14)
does not depend on the observer and is the same in the moving
inertial system
Let us assume that the origin is moving along the x-axis. We know Three useful hyperbolic function
then formulae
cosh2 η − sinh2 η = 1, (H1)
0 0 tanh η
y=y, z=z (3.16) sinh η = p , (H2)
1 − tanh2 η
1
The most general linear solution is, as can be verified by direct cosh η = p (H3)
1 − tanh2 η
substitution using (H1)
H2 and H3 can be easily shown by
! ! ! plugging in the definition of tanh η
0
x cosh η sinh η x and use of H1.
= (3.17)
ct sinh η cosh η ct0
Now divide :
x v
= tanh η = = β ⇒
ct c
v
c 1
sinhη = q , cosh η = q = γ,
v2 v2
1− c2 1− c2
where x = vt was used in the first step, (H2) and (H3) in the second,
which, when plugged back in (3.17), gives
x0 + vt0 t0 + v2 x0
x= q , t= q c , (3.20)
2 2
1 − vc2 1 − vc2
We see that the transformations for time and space are now symmet-
ric2 ! With 2
Very clear if you set c = 1.
Or,
1 x0 = γ(x − vt), (3.23)
γ=q (3.25)
v 2 vx
1−
c t0 = γ t − 2 . (3.24)
c
space-like
x
past
time-like
tA tB = 0
A B x’
t0 > 0
The stationary clock ticks faster! Why is the situation not symmet-
ric? An observer in B 0 will have to use two clocks to compare to the
for him moving clock of A.
We can now define the important quantity of proper time ∆τ :
it is the time measured by a clock which is at a fixed place. We can
observe time dilation in (almost) everyday experiments like the decay
of unstable particles, like the muons generated in the atmosphere by
cosmic radiation. See Ex1!
which means
r
0 v2
l =l 1− (3.38)
c2
time dilation
x’
t= t0 > t0
x’
Lorentz
contraction
x
0 0
l=l/ <l
dx dx0
Let the velocity seen in A be u = dt and in B 0 be u0 = dt0 . Then
dividing dx0 by dt0 . We obtain
3.3.1 Fourvectors
We define space-time points X using the components of a four-vector
0
x ct
x1 x
xµ ≡ 2 ≡ (3.48)
x y
x3 z
The components are defined relative to basis vectors eµ
3
X
X = xµ eµ = xµ eµ (3.49)
µ=0
x0µ = Λµ ν xν (3.50)
relativity, particles, fields 31
where we have moved one index to the right to make obvious, which
one is the first and which one is the second index of the matrix.
Since we have only changed the coordinates and the physical vectors
has to remain the same, we know
eν = Λµ ν e0µ (3.52)
with
1 0 0 0
0 −1 0 0
(ηµν ) = (3.56)
0 0 −1 0
0 0 0 −1
ds2 = ηµν dxµ dxν = ηµν dx0µ dx0ν = ηµν Λµ α Λν β dxα dxβ (3.57)
which we will use extensively later. This defines the set of all
Lorentz transformations.6 6
To get the representations of the
Lorentz group SO(1, 3).
32 andreas weiler, tum
Vµ W µ (3.59)
Parity and time reversal are special because they cannot be written
as a product of rotations and boosts7 . We can write them as 7
Because rotations and boosts
reduce to 14×4 for v → 0 or θ → 0.
1 0 0 0 −1 0 0 0
0 −1 0 0 0 1 0 0
P = , T = (3.62)
0 0 −1 0 0 0 1 0
0 0 0 −1 0 0 0 1
d2 X µ
=0 (3.64)
dτ 2
where X µ = X µ (τ ) denotes the location of a particle with τ being
the proper time. In another inertial frame X 0µ = Λµ ν X ν and so
d2 X 0µ 2 ν
µ d X
= Λ ν =0 (3.65)
dτ 2 dτ 2
and that dτ is the same for both observers. This motivates the
introduction of the four-velocity
dxµ
uµ ≡ (3.66)
dτ
relativity, particles, fields 33
p · q ≡ ηµν pµ q ν (3.69)
pν ≡ ηµν pµ (3.70)
Or in other words,
and therefore
p · q = pµ q µ = p0 q 0 − p · q (3.72)
I hope you are not confused by this trivial act of notational sloth
and the fact that we have ”two kinds of vectors”, contravariant
vectors pµ and covariant vectors pµ . There is nothing profound
going on here. Just a convenient notation.10 10
See however box below!
Indeed, we can also define an inverse for ηµν , (ηµν )−1 = η µν
As you can see it is numerically the same matrix11 and we will use 11
In general curved space-times the
the same symbol η with two covariant indices: ηµν . From this we can metric can be a function of xµ and
the inverse will in general not be the
see there is no such thing as δ µν or δµν , neither does ηνµ exist ! same matrix.
∂φ µ
δφ = δx (3.75)
∂xµ
34 andreas weiler, tum
and therefore
∂ ∂ ∂ ∂ ∂
∂µ = = , , , (3.76)
∂xµ ∂t ∂x ∂y ∂z
and
∂ ∂ ∂ ∂ ∂
∂µ = = ,− ,− ,− (3.77)
∂xµ ∂t ∂x ∂y ∂z
A useful mnemonic is to think of xµ carrying an upper index but
since in ∂x∂ µ it appears in the denominator, it should carry a lower
index.
We will also use
∂µ Aµ = ∂0 A0 + ∂j Aj (3.78)
and
∂2
∂µ ∂ µ = − ∇2 = (3.79)
∂t2
Finally, we will make use of the completely anti-symmetric tensor
µναβ , known as the Levi-Civita tensor. It is defined as
1,
if µναβ is an even permutation of (0, 1, 2, 3)
µναβ = −1, if µναβ is an odd permutation of (0, 1, 2, 3) (3.80)
0, if µναβ is not a permutation of (0, 1, 2, 3)
pµ ≡ muµ (3.84)
which we can write in an inertial system where the particle has the
three-velocity v(t)
! !
0
p c
pµ = = mγ(v) (3.85)
p v
relativity, particles, fields 35
p
n p
n
The dual space: The Minkowski metric defines a (not positive definite) norm for four vectors
Every vector space V has a dual space, often denoted V ∗ , which is simply the set of all linear maps from the
vector space into the real (or complex) numbers. We can find a basis eµ for the dual space with
p = mγ(v) v (3.86)
where a new term, the rest energy, appeared. We find for the
square of the momentum
p2 = pµ pµ = (p0 )2 − p2 = m2 c2 (3.88)
3.5.1 E = mc2
The energy and the momentum of a particle of mass m form a four-
vector
!
µ E
p = (3.89)
p
whose square is
p2 = E 2 − p2 = m2 (3.90)
E 2 = m2 + p2 (3.91)
µ+
⇡+
3.5.2 Application: weak decay of a pion
± ±
A charged π can decay to a muon µ and a muon anti-neutrino νµ .
The muon has about 3/4 the mass of a pion13 and the neutrino is
⌫¯µ
almost massless. We can calculate the energy and the velocity of the
decay products using the four-momentum conservation Figure 3.7: Feynman diagram of a π +
decay.
pµini = pµfinal
13
(3.92) mπ± = 139.57 MeV and mµ± =
105.658 MeV.
which is
! ! ! !
µ µ
Eini Eπ+ Eµ+ + Eν̄ Efinal
= = = (3.93)
pini pπ + pµ+ + pν pfinal
The almost mass-less neutrino has Eν = |pν | and because the initial
three-momentum was zero in the rest system of the pion, we get
m2π+ − m2µ+
|pν | = |pµ+ | = (3.95)
2mπ+
For the decay to an electron and an electron anti-neutrino, the
electron mass is negligible: we find the decay products back-to-back
and the rest-mass has been converted entirely into kinetic energy
with |pν̄ | = |pµ+ | = mπ+ /2.
4
Classical Field Theory
φa (xµ ) (4.1)
Remember that in classical mechanics, L depends on q and q̇. In Interactions with multiple derivatives,
field theory we analogously restrict to Lagrangians depending on e.g.
1
φ and φ̇ and not on φ̈. We could in principle write Lagrangian’s L= φ2 φ
M2
which depend on ∇φ, ∇2 φ, ∇3 φ and so on. But anticipating Lorentz-
may occur due to quantum effects in
invariance, we will only consider ∇φ and no higher derivatives. Also all but the simplest renormalizable
we will not include an explicit xµ dependence, all such dependence field theories. The are generic in all
effective field theories, as we may
comes only implicitly through φ and its derivatives. discuss at the end of the course, time
permitting.
In the following, we will use functional derivatives. A functional F [φ] for us is a map from a vector space of
δF [φ]
functions into the real numbers, like e.g. the action S[φ]. A functional derivative is denoted by δφ(x) and it is
defined as the linear term in an expansion
Z
δF [φ]
F [φ + η] = F [φ] + dx0 η(x0 ) + . . . (4.7)
δφ(x0 )
or
Z
δF F [φ + εη] − F [φ]
dx (x)η(x) = lim (4.8)
δφ ε→0 ε
d
= F [φ + η] , (4.9)
d =0
where η is an arbitrary function. In particular, for the functional F [φ] = φ(x), we find
δF [φ] δφ(x)
= = δ(x − x0 ) (4.10)
δφ δφ(x0 )
The quantity η is called the variation of φ. The differential (or variation or first variation) of the functional F [φ]
is
Z
δF
δη F (φ) = dx (x) η(x) (4.11)
δφ
This is similar in form to the total differential of a function. You should familiarize yourself with the mechanics of
functional derivatives on the next exercise sheet.
δS = 0 (4.12)
with
Z
∂L ∂L
δS = d4 x δφa + δ(∂µ φa ) (4.13)
∂φa ∂(∂µ φa )
Z
∂L ∂L ∂L
= d4 x − ∂µ δφa + ∂µ δφa
∂φa ∂(∂µ φa ) ∂(∂µ φa )
The last term is a total derivative and therefore its integral only
depends on the field values at the spatial and temporal boundaries.
It vanishes for any δφa that decays at spatial infinity and obeys Note, we will always make the
assumption that we can drop such
δφa (x, t1 ) = δφa (x, t2 ) = 0 (4.14) total derivatives from the Lagrangian.
It let’s us integrate by parts without
consequence. We will therefore
equate
∂µ ∂ µ φ + m2 φ = 0 (4.21)
4.1.6 Locality
All examples above are local Lagrangians. We have not written any
terms that look like
Z
L = d3 xd3 yφ(x)φ(y)
relativity, particles, fields 41
which would couple a field at x to a field at y. The closest we get is Again, effective field theories are
a coupling between φ(x) and φ(x + δx)through the gradient term special and can describe the correct
physics until some scale while being
(∇φ(x))2 . This property of locality is, as far as we know, a key non-local.
feature of all theories of Nature. Locality has a deep connection
to unitarity and one finds that non-local theories quickly violate
unitarity.
Z Z
−→ d4 y ∂µ0 φ(y)∂ 0µ φ(y) − m2 φ2 (y) = d4 y L(y) = S
∂
where we’ve denoted ∂µ0 = ∂y and used the fact that the Jacobian
is trivial
d4 x → det(Λ) d4 y = d4 y
with4 4
Ex: Show that
| det(Λ)| = 1
det(Λ) = 1
using Eq. (3.58).
for Lorentz transformations connected to the identity, which we
are using for now.
Aµ (x) → Λµ ν Aν (Λ−1 x)
42 andreas weiler, tum
∂µ j µ (x) = 0 (4.36)
or
∂L
∂µ Xa (x) − B µ =0 (4.42)
∂(∂µ φa )
This proves the theorem and we find Noether’s theorem was proved
for classical fields and one usually
extends it to quantum theory by
∂L replacing the fields by the corre-
∂µ j µ (x) = 0 with j µ (x) = Xa (x) − B µ (4.43) sponding quantized fields. This does
∂(∂µ φa ) however not always give a conserved
current in the quantum theory! A
non-zero divergence is called an
current anomaly. In the SM, e.g.
4.2.1 Conserved currents baryon and lepton number are global
symmetries which have a quantum
A conserved current implies a conserved charge Q with anomaly.
Z
Q ≡ d3 xj 0 (4.44)
we find5 5
∂V is the boundary area of V .
Z Z
dQV
=− d3 x ∇ · j = − dS · j (4.47)
dt V ∂V
with Stokes theorem. Any charge leaving V per time interval must
be equal to the flow of three-current out of the volume. This holds in
any local field theory.
L = ∂µ ψ ∗ ∂ µ ψ − V (ψ ∗ ψ) (4.48)
44 andreas weiler, tum
j µ = i(∂ µ ψ ∗ )ψ − iψ ∗ (∂ µ ψ) (4.52)
Once we quantize this field theory, we will see that it can be in-
terpreted as electric charge and implies a conservation of particle
number.
We see that all the terms are polynomials of the form φa φa = φ21 + φ22 ,
similar to r2 = x2 + y 2 which is invariant under rotations. We
reproduce the complex transformation with
! ! !
φ1 (x) cos α sin α φ1 (x)
→ (4.54)
φ2 (x) − sin α cos α φ2 (x)
or
! ! !
δφ1 (x) 0 α φ1 (x)
→ (4.55)
δφ2 (x) −α 0 φ2 (x)
det O = 1 (4.57)
relativity, particles, fields 45
j µ = iT ij (∂ µ φi )φj (4.61)
[T a , T b ] = iεabc T c (4.62)
∂L
(j µ )ν = ∂ν φa − δνµ L ≡ T µ ν (4.67)
∂(∂µ φa )
and
Z
Pi = d3 xT 0i (4.69)
j µ = i(∂ µ ψ ∗ )ψ − iψ ∗ (∂ µ ψ) (4.75)
We will discuss this issue in more detail once we tackle the issue of
gauge invariance.
where we have used Eq. (4.31). As you may remember from your
electro-dynamics lectures, we can freely choose a gauge for Aµ and
we will use Lorenz gauge ∂α Aα = 0 to obtain
Aν = jµ (4.80)
48 andreas weiler, tum
Ai (x) = 0 (4.83)
e
A0 (x) = δ 3 (x) (4.84)
The homogeneous solutions Aν = 0 , which are the electro-
magnetic waves, we will ignore here since they have nothing to
do with our source.
and so
Z
4 d4 k µ
δ (x) = (−k µ kµ ) eix kµ (4.91)
(2π)4
We will use this all the time: for you as a field theorist, means
”−k 2 ”.
relativity, particles, fields 49
1 1
L = − hh + λh3 + J · h (4.96)
2 3
where we have dropped the indices. This defines the non-linear
equations of motion. We solve perturbatively and we start with the
non-interacting theory with λ = 0
1
h0 = J (4.97)
We obtain the first oder correction h = h0 + h1 where h1 = O(λ)
x Π(x, y) = −δ 4 (x − y) (4.103)
where we have used integration by parts. We can solve the free field
with λ = 0 equation, by inserting y h0 (y) = J(y) to obtain
Z
h0 (x) = − d4 y Π(x, y)J(y) (4.108)
relativity, particles, fields 51
derive the most general λn expansion for a classical field with the
following classical Feynman rules:
2. Truncate with a source J(xi ) or branch into two lines and multi-
ply with a factor λ.
4. The final result for h(x) is given by all graphs up to some order
λn with the ends truncated by currents J(xi ) and the lines re-
placed by propagators Π(xi , xj ). All internal points are integrated
over.
As we will see, in quantum field theory the Feynman rules are almost
identical except for ~ → 0 (classical limit), lines can not close
in on themselves.17 In classical general relativity, these diagrams 17
No loops!
describe the post-Newtonian effects of the Sun on Mercury! The
lines represent gravitons and the sources are in this case the Sun.
The first diagram is the well-known Newtonian potential while the
√
second order represents the self-interaction λ ∼ GN which is a
prediction of general relativity.
The Hamilton density is obtained by a Legendre transformation The conjugate momentum π a (x)
should not be confused with the
H = π a (x)φ̇a (x) − L(x) (4.112) total momentum P i which is just
a number characterizing a field
configuration.
As in classical mechanics, we replace φ̇a (x) in favor of π a (x) every-
where. The Hamiltonian is simply
Z
H = d3 x H (4.113)
[qa , pb ] = i~ δba
[qa , qb ] = 0
[pa , q b ] = 0
We obtain for the Hamiltonian in terms of ap , a†q Ex: Assuming canonical com-
Z mutation relations for the fields,
1 show that the creation and
H= d3 x π(x)2 + (∇φ(x))2 + m2 φ(x)2
2 annihilation operators satisfy
Z r [ap , a†q ] = (2π)3 δ (3) (p − q).
1 3
3
d p d q 3
ωp ωq ix·p † −ix·p
ix·q † −ix·q
= d x 3 3
(−1) ap e − ap e aq e − aq e
2 (2π) (2π) 4
1
+p ip ap eix·p − ip a†p e−ix·p iq aq eix·q − iq a†q e−ix·q
4ωp ωq
1
+ m2 p ap eix·p + a†p e−ix·p aq eix·q + a†q e−ix·q
4ωp ωq
We will now collect the terms, integrate over d3 x and replace
e−ix·(p+p) → δ 3 (p − p) and e−ix·(p−p) → δ 3 (p + p) which allows us to
remove one d3 q integral.
Z
d3 p 1 h † †
i
H= ap a −p + a p a−p (−ωp2 + p2 + m2 )
(2π)3 4ωp
1
+ ap a†p + a†p ap (ωp2 + p2 + m2 )
4ωp
The first line vanishes because of ωp2 = p2 + m2 and we obtain
Z
1 d3 p
H= 3
ωp ap a†p + a†p ap (5.13)
2 (2π)
Z
d3 p † 1 3 (3)
= ω p a a
p p + (2π) δ (0) (5.14)
(2π)3 2
What is this!? We clearly have to work to interpret this properly.
There are two infinities: the delta-function has an infinite spike at 0
and the integral over ωp also diverges.4 4
The latter infinity is closely related
to the infinity of the specific heat
mentioned in the introduction.
5.1.2 Infinities
We will separate the infinities into two classes: infrared diver-
gencies which are caused by assuming infinitely large scales (or
arbitrarily low energies) and ultraviolet divergencies which come
from extrapolating to infinitely small scales (or arbitrarily high
energies).
We will better understand the problem by discussing the ground
state or the vacuum |0i. We define |0i as the state being annihi-
lated by all the ap
This is common for quantum field theories: infinities pop up left and
right. They tell us that we might be asking the wrong question or
are doing something wrong.
The infrared divergence or long-distance divergence can be taken
care of by putting the field in a box of length L or volume V = L3
and assuming periodic boundary conditions. We obtain
Z L Z L Z L
2 2 2
(2π)3 δ (3) (0) = lim dx dy dz eix·p =V (5.19)
L→∞ −L −L −L
2 2 2 p=0
and now
H 0 |0i = 0 (5.28)
If you are a bit taken aback by the lack of rigor implied by an ar-
bitrary redefinition of our Hamilton, note the following: there is
an inherent ambiguity of ordering transitioning from classical to
quantum. If we had defined the classical Hamiltonian as
0 1 1 ω
Hclassical = (ωq − ip)(ωq + ip) = p2 + q 2 = Hclassical
2 2 2
0
which is classically equivalent Hclassical = Hclassical . We would
naturally get after quantization10 10
Compare to the definition
r
ω i
0
HQM = ω a† a a=
2
q + √ p,
2ω
r
ω i
a† = q− √ p=
This method of exploiting the order ambiguity is called normal 2 2ω
ordering: we define it as the ordering in which all the creation .
The Cosmological Constant : As mentioned before, gravity can indeed measure energy differences. Gravity
sees everything! We would expect the sum of a all zero-point energies to contribute to the cosmological constant
Λ = E0 /V and appear on the right side of the Einstein-Hilbert equations
Cosmological observations tell us that ca. 70% of the energy density of the universe today is in a term that looks
like the cosmological constant with
Λ ∼ (10−3 eV)4
This is many orders of magnitude smaller than all the other scales in particle physics (the SM is valid to ∼ 200
GeV = 2 · 1011 eV). Why don’t zero-point energies contribute to Λ? What cancels them to such accuracy? We do
not know the answer. In fact, we don’t even know how to ask the right question.
5.2 Particles
What are the excitations of the field with respect to the vacuum |0i?
We know that
As in the case of the simple harmonic oscillator, we can construct Ex: Verify the relation
energy eigenstates by applying the creation operators a†p on the [H, a†p ] = ωp a†p
vacuum |0i. We define
1
a†p |0i = p |pi (5.33)
2Ep
p
The factor 2Ep constitutes a relativistic normalization – see
√
below.12 The energy of this state is 12
The 2 is for convenience only.
Z
d3 p p
: H : |qi = ωp a†p ap 2Eq a†q |0i
(2π)3
Z
p d3 p † † 3 (3)
= 2Eq ωp ap a q a p + (2π) δ (p − q) |0i
(2π)3
p
= ωq 2Eq a†q |0i
= ωq |qi
relativity, particles, fields 59
with
ωq2 = q 2 + m2 (5.34)
The first term vanishes because the term in the brackets is sym-
metric under p → −p which together with p makes this part of the
integrand anti-symmetric and therefore vanishes after integrating
over a symmetric interval. We find
Z
d3 p
:P : = p a† ap (5.35)
(2π)3 p
In summary we find
Z
d3 p µ †
: Pµ : = p ap ap (5.36)
(2π)3
J i |p = 0i = 0
has momentum
p1 + p2 + . . . + pn (5.39)
We interpret this as an n-particle state. Note that these are not localized
in space; a†pn creates a momentum
|p1 , . . . , pn i = N a†p1 a†p2 . . . a†pn |0i (5.40) eigenstate.
and we know that these particles are bosons. We can construct the
Fock space as the Hilbert space spanned by all possible combina-
tions of the a†pi
|0i, a†p1 |0i, a†p1 a†p2 |0i, a†p1 a†p2 a†p3 |0i, . . . (5.42)
Z
d3 p †
N= a ap (5.44)
(2π)3 p
which satisfies
[H, N ] = 0
Similarly
h0|φ(x)|0i = 0 (5.48)
We can obtain h0|φf (x)φf (x)|0i by observing that of the four terms
after expanding φ(x)2 only the one proportional to h0|ap a†q |0i ∼
δ 3 (p − q) will be non-zero. We have
Z Z 2
1 d3 p 3 (x − y)2
h0|φf (x)φf (x)|0i = d y exp − + ip(x − y)
8(aπ)6 2Ep a2
and realizing again that the only scale is |x − y|, we have to have Up to factors of order one, like π etc.
1
h0|φ(x)φ(y)|0i ∼
|x − y|2
62 andreas weiler, tum
Z [ψ] = n ψ ∼ (mass)n
S = d4 x ∂µ φ(x)∂ µ φ(x) + . . . which means that ψ has dimension
(mass)n ∼ (energy)n ∼ (length)−n .
[φ(x)] = 1
hp|pi = hp̃|p̃i
pµ pµ = m2 p20 = Ep2 = p2 + m2
which completes the proof. Now we know that the Lorentz-invariant Note, that
δ (3) -function for 3 vectors is: X δ(x − xi )
δ(g(x)) =
i
|g 0 (xi )|
2Ep δ (3) (p − q)
where the sum extends over all roots
of g(x) which are assumed to be
this holds because
simple. For example
Z
d3 p 1 h
2Ep δ (3) (p − q) = 1
i
δ x2 − α 2 =
δ (x + α)+δ (x − α) .
2Ep 2|α|
L = ∂µ ψ ∗ ∂ µ ψ − m2 ψ ∗ ψ
∂µ ∂ µ ψ + m2 ψ = 0
Z
d3 p 1
ψ(x) = 3
p bp eix·p + c†p e−ix·p (5.52)
(2π) 2Ep
Z
d3 p 1 † −ix·p
ψ † (x) = 3
p bp e + cp eix·p (5.53)
(2π) 2Ep
∂L
π= = ∂t ψ ∗
∂∂t ψ
and22 22
Recall, that we define π(x) in
analogy with the result for the
harmonic oscillator and that it
satisfies the canonical commutation
relations.
relativity, particles, fields 65
Z r
d3 p Ep
π(x) = 3
(+i) b†p e−ix·p − cp eix·p (5.54)
(2π) 2
Z r
3
d p Ep
π † (x) = (−i) bp eix·p − c†p e−ix·p (5.55)
(2π)3 2
j µ = i(∂ µ ψ ∗ )ψ − iψ ∗ (∂ µ ψ) (5.58)
Z
d3 p †
:Q: = c cp − b†p bp = Nc − Nb (5.60)
(2π)3 p
66 andreas weiler, tum
[H, Q] = 0
We can now discuss the time-evolution of the operator φ(x, t), e.g.24 24
We use
[A2 , B] = A[A, B] + [A, B]A = 2f A
∂t φ(y) = i[H(t), φ]
Z if the commutator is just a c-valued
i function
= d3 x π(x, t)2 + (∇φ(x, t))2 + m2 φ(x, t)2 , φ(y) [A, B] = f
2
Z e.g. f = δ(x − y).
= i d3 x π(x)(−i)δ (3) (x − y)
= π(y)
relativity, particles, fields 67
And
Z
i
∂t π(y) = i[H, π] = d3 x π(x)2 + (∇x φ(x))2 + m2 φ(x)2 , π(y)
2
Z
= i d3 x ∇x [φ(x), π(y)]∇x φ(x) + im2 φ(x)δ 3 (x − y)
Z
=− d x (∇x δ (x − y))∇x φ(x) − m2 φ(y)
3 3
∂µ ∂ µ φ + m2 φ = 0
We know that
Z
d3 p 1
φ(x) = 3
p ap e−ip·x + a†p eip·x (5.67)
(2π) 2Ep
68 andreas weiler, tum
where the exponent has the opposite sign, because we have used the
Lorentz-vectors
p · x = pµ xµ = Ep t − p · x
You can easily check that Eq. (5.67) indeed satisfies the Klein-
Gordon equation with Ep2 = p2 + m2 .
For completeness, we show the straight-forward result for the
complex scalar field of Eq. (5.52) in the Heisenberg picture
Z
d3 p 1
ψ(x) = 3
p bp e−ip·x + c†p eip·x (5.68)
(2π) 2Ep
Z
d3 p 1 † ip·x
ψ † (x) = 3
p bp e + cp e−ip·x (5.69)
(2π) 2Ep
We would like to study now: how does the commutator look like at
arbitrary space-time distances? This brings us back to the discussion
of causality in the introduction. Recall that we must require that
space-like separated events (x − y)2 < 0 do not overlap. Or in other
words, we should be able to find a basis in which matrix elements of
operators can be brought into diagonal form (for x and y space-like).
This is equivalent to requiring that space-like separated operators
commute This is the very important causality
condition.
d3 p 1 h −i(x−y)·p
Z i
∆(x − y) = 3
e − ei(x−y)·p
(2π) 2Ep
which is the desired result.
which is non-vanishing.
theories. Continuing with the causality problem, we could ask the d3 p 1 −ip·(x−y)
Z
D(x − y) = e
(2π)3 2Ep
question: what is the amplitude to create a particle at x and observe Z ∞
2π p2 eipr − e−ipr
it at y while leaving the vacuum |0i undisturbed? We calculate = dp p
(2π)3 0 2
2 p +m 2 ipr
Z ∞ ipr
−i pe
= dp p
2(2π)2 r −∞ p2 + m2
The integrand can be considered a
complex function with branch-cuts
on the imaginary axis (from the
√
. . .) starting at ±im. We evaluate
70 andreas weiler, tum
Z
d3 p d3 q 1
h0|φ(x)φ(y)|0i = p h0|ap a†q |0i e−ip·x+iq·y
(2π)3 (2π)3 4Ep Eq
Z
d3 p 1 −ip·(x−y)
= e
(2π)3 2Ep
≡ D(x − y) (5.74)
D(x − y) ∼ e−m|x−y|
cut
Iradial
Ibranch
Let us describe what this equation means: we know that for space-
like separation (x − y)2 < 0 of x and y, there is no Lorentz invariant
ordering of events: if a particle can travel from x −→ y, it can just +im
IR
as easily travel from y −→ x. In any measurement, the amplitudes
im
for the two processes cancel.
Let us consider this for a complex scalar-field, which has distinct
Figure 5.1: Since the integrand
particle and anti-particle excitations, see Eq. (5.52). When the has no poles or branch-cuts inside
complex scalar field ψ(x) is quantized, this contour, the integral has to
vanish. The integrand on the radial
• ψ(x) will create positive charged particles and destroy negatively boundary vanishes Iradial → 0. We
can therefore ’push the contour’ to
charged ones wrap around the branch cut. We
know IR + Ibranch−cut + Iradial = 0
• ψ † (x) will create negative charged particles and destroy positively and therefore IR = −Ibranch−cut
charged ones
In oder for these two processes to cancel, both particles must exist
and they must have the same mass.
relativity, particles, fields 71
φ(x)φ(y) x0 > y 0
T φ(x)φ(y) = (5.77)
φ(y)φ(x) y 0 > x0
The meaning of the ”contour” will become clear soon. This is the
first integral over 4-momenta. Until now, we have integrated only
over 3-momenta and kept the p0 on the mass shell with p0 = Ep .
Here, we have no such condition. The integral for ε = 0 is ill-defined,
because for each p the denominator has a pole at
p
p2 − m2 = (p0 )2 − p2 − m2 p0 = ±Ep = ± p2 + m2
+Ep
72 andreas weiler, tum
where in the step from the 2nd to the 3rd line, we have flipped the
integration variable p → −p which is valid since the rest of the
expression (including the boundaries) is symmetric in p. Also in this
case we reproduce the Feynman propagator. QED.
with ε > 0 and infinitesimal. This shifts the poles slightly off the real
axis, such that the integral along p0 is equivalent to the contour C.
This form of the propagator is called the iε-prescription.
relativity, particles, fields 73
+Ep i"
Note, that we did not need the contour in the derivation. For other
applications it is useful to pick alternative contours, which also result
in Green’s functions.
We can define a retarded Green’s function ∆R (x − y) with
D(x − y) − D(y − x) x0 > y 0
∆R (x − y) =
0 y 0 > x0
retarded advanced
Ep +Ep
Ep +Ep
φ + m2 φ = J(x)
∆F (x − y) = h0| T φ(x)φ(y)|0i
= θ(x0 − y 0 ) h0| φ(x)φ(y)|0i + θ(y 0 − x0 ) h0| φ(y)φ(x)|0i
and
We are still a long way from being able to do any real calculation,
since so far we have only talked about the free Klein-Gordon theory,
where the field equations are linear and there are no interactions.
On the other hand, the formalism here is extremely important since
the free theory forms the basis for doing perturbative calculations in
interacting theories.
without a source, this would be the solution for all time. We can con-
struct the solution with a source, by using the retarded Green’s
function
and may be dropped. Then the solution involves only the Fourier
transform of J(y)
Z
J(p) = d4 y eip·y J(y)
p2 = m2 (on-shell) (5.84)
Since all observables are built out of the fields, we have solved the
theory. We can now guess (or derive) the Hamiltonian after J(x) has
been switched on and off. The free Hamiltonian (with J(x) ≡ 0) is
Z
d3 p
H= Ep a†p ap
(2π)3
and
h0|H|0i = 0
If we act with J(y) and study the system in the far future (x0 →
+∞), we can get HJ looking at the above discussion of the retarded
Green’s function by just replacing
i
ap → ap + p J(p)
2Ep
i
a†p → a†p − p (J(p))∗
2Ep
to obtain
Z ! !
d3 p i i
HJ = Ep a†p −p (J(p))∗ ap + p J(p)
(2π)3 2Ep 2Ep
The energy of the system in the far future after the source has been
switched off, is therefore
Z
d3 p |J(p)|2
h0|HJ |0i = Ep (5.85)
(2π)3 2Ep
particles. This means that the expectation value of the total number
of particles created with momentum p is
|J(p)|2
dN (p) =
2Ep
We see from the condition Eq. (5.84), that only the Fourier modes
of J(x) which are in resonance (p2 = m2 ) with the on-shell Klein-
Gordon waves are effective at creating particles. This is just the
classical phenomenon of resonance occurring in the quantum field
theory setting.
|p| m
∂ ψ̃ 1 2
i =− ∇ ψ̃ (5.88)
∂t 2m
This looks like the Schrödinger equation, except it does not have
a probabilistic interpretation – it is just a free classical field, evolving
according to an equation of first order in the time derivatives.
In Sec. 4.1.4, we discussed a Lagrangian which was first order
in the time derivatives. Once again, we can derive Eq. (4.23) from
relativity, particles, fields 77
L = ψ ∗ (− − m2 )ψ
= |ψ̇|2 − |∇ψ|2 − m2 |ψ|2
= | − imψ̃ + ∂t ψ̃|2 − |∇ψ̃|2 − m2 |ψ̃|2
≈ im ψ̃ ∗ (∂t ψ̃) − ψ̃ (∂t ψ̃ ∗ ) − |∇ψ̃|2
∂L
π= = iψ ∗
∂(∂t ψ)
or
p2
H|pi = |pi
2m
The single particle states satisfy the non-relativistic dispersion
relation.
Comments
|xi = ψ † (x)|0i
so that
X |xi = x |xi
Let us now construct a state |ϕi which we would usually call the
Schrödinger wavefunction (in the position representation). We
obtain it by taking superpositions of one-particle states |xi
Z
|ϕi = d3 x ϕ(x) |xi
Let us check if it has the right properties. The position operator has
the right action, as we can see
Z
i
X |ϕi = d3 x xi ϕ(x) |xi (5.92)
Z
∂ϕ(x)
P i |ϕi = d3 x −i |xi (5.93)
∂xi
[X i , P j ] |ϕi = iδ ij |ϕi
and we find
∂ϕ 1 2
i = Hϕ = − ∇ ϕ (5.94)
∂t 2m
This is the same equation obeyed by the original (full quantum) field.
Only this time, it is really the Schrödinger equation with the usual
probabilistic interpretation for the wavefunction ϕ. In particular, the
total probability which is conserved in quantum mechanics arises
as the conserved charge of the Noether current:
Z
Q= d3 x |ϕ(x)|2
∆L = λ ψ ∗ (x)ψ ∗ (x)ψ(x)ψ(x)
which destroys two particles before creating two new ones. In the
following chapter we will explore interactions like there in detail for
relativistic theories.
6
Interacting quantum fields
These terms create and destroy particles and particle number will
in general not be conserved and we could calculate [H, N ] 6= 0
(even though the discrete parity would be conserved.)
relativity, particles, fields 83
This simple argument using dimensional analysis tells us that we can usually focus on the first few terms in the
interactions. It also tells us that it will be very difficult to figure out the high energy theory. These considerations
are a simplified version of what is called effective field theory and Wilson’s renormalization group. Due to the
unknown UV physics completing our QFT, we are generally required to write down all the allowed marginal and
relevant interactions to capture the IR behaviour of a theory with a scalar field.
This discussion is the modern viewpoint of what people previously discussed under the notion of ”renormalizable”
quantum field theories.
1 1
L = ∂µ ψ ∗ ∂ µ ψ − M 2 ψ ∗ ψ + ∂µ φ∂ µ φ − m2 φ2 − g ψ ∗ ψ φ
2 2
The coupling g has [g] = 1 and is a relevant coupling. For the per-
turbation to be small, we require g M, m. Particle number is
again not conserved anymore but since the Lagrangian is invariant
under a continuous phase symmetry (U (1))
ψ → eiα ψ
H = H0 + Hint (6.1)
We now define states and operators in the interaction picture as It is a hybrid: informally speaking it
is the Heisenberg picture for the free
H0 and the Schroedinger picture for
the Hint .
|ψ(t)iI = eiH0 t |ψ(t)iS (6.3)
iH0 t −iH0 t
OI (t) = e OS e (6.4)
d
i |ψ(t)iS = HS |ψ(t)iS
dt
relativity, particles, fields 85
d −iH0 t
i (e |ψiI ) = (H0 + Hint )S e−iH0 t |ψiI (6.6)
dt
d
i |ψiI = HI |ψiI (6.7)
dt
where we have first canceled H0 on both sides, multiplied the equa-
tion by eiH0 t and used Eq. (6.5).
When λ is small, this will give the most important part of the time-
dependence. Since H0 is diagonalized by the Fourier expansion, we
get
Z
d3 p 1 −ip·x
† ip·x
φI (x) = p ap e + ap e 0 (6.10)
(2π)3 2Ep x =t−t0
To verify, that this indeed the correct solution, take the derivative:
you find that each term gives the previous one ×(−i)HI (t). The For the first two terms:
initial condition is U (t0 , t0 ) = 1 is trivially satisfied. ∂
U (t, t0 )
We observe that that the factors of HI (ti ) stand in time order, ∂t
Z t
with the later operators to the left. We can now simplify using the = (−i) HI (t) + (−i)2 HI (t) dt2 HI (t2 ) + . . .
t0
time ordering T . Starting with HI2 Z t
= −iHI (t) 1 + (−i) dt1 HI (t1 ) + . . .
Z t Z t1 Z Z t
1 t t0
dt1 dt2 HI (t1 )HI (t2 ) = dt1 dt2 T {HI (t1 )HI (t2 )}
t0 t0 2 t0 t0
= −iHI (t)U (t, t0 )
The time-ordered double integral on the RHS just counts everything
twice, since in the t1 − t2 plane, the integrand T {HI (t1 )HI (t2 )} is
symmetric about the t1 = t2 line. Similarly, one can show
Z t Z tn−1 Z Z t
1 t
dt1 . . . dtn HI (t1 ) . . . HI (tn ) = dt1 . . . dtn T {HI (t1 ) . . . HI (tn )}
t0 t0 n! t0 t0
It’s not easy to draw this but you should be able to convince yourself
that this is in fact correct. We can now write the time evolution
operator in a very compact form:
Z t Z Z t
(−i)2 t
U (t, t0 ) = 1+(−i) dt1 HI (t1 ) + dt1 dt2 T {HI (t1 )HI (t2 )}
t0 2! t0 t0
Z Z t Z t
(−i)3 t
+ dt1 dt2 dt3 T {HI (t1 )HI (t2 )HI (t3 )} + . . .
3! t0 t0 t0
Z t
0 0
U (t, t0 ) = T exp −i dt HI (t ) (6.15)
t0
t
Z t Z t1
dt1 dt2
t0 t0
t0
t1
t0 t
where all the fields are in the interaction picture.7 This HI allows 7
Which means the field operators are
particles to morph into each other. To understand why that is, in the Heisenberg picture of the free
theory.
follow the time evolution of a state
c† b† a
cc† a ψ+φ→ψ
ψ̄ + ψ → φ → ψ̄ + ψ
[HI , N ] 6= 0
We will not just have two colliding protons but a very complicated
mess of protons, pions, photons, gluons, etc.
The result of the scattering process can have several outcomes.
Multiple initial particles could form a bound state.10 In this case, 10
e.g. p + p → 2 D, two protons
no matter how long we wait after the scattering, the final state will forming a deuterium nucleus.
1
t
T
with f (t) like in Fig. 6.2. The function f (t) = 0 for large |t| and
f (t) = 1 around 0 where the scattering happens. A long time
after the scattering has occurred, we turn the interaction off very
slowly (adiabatically) over a period ∆: we expect that the simple
states in the real theory slowly turn into eigenstates of H0 with unit
probability.
There will be a one-to-one correspondence between asymptotic
simple eigenstates of the full Hamiltonian H and the eigenstates of
the free H0 . 13 We should recover the full theory if 13
Again, we can’t consider bound
states, which are not eigenstates of
T → ∞, ∆ → ∞, ∆/T → 0 the free Hamiltonian.
Z
4
S = T exp −i d x HI (x) (6.20)
90 andreas weiler, tum
where q1,2 are the momenta of the decay products, a nucleon anti-
nucleon pair (ψ̄ψ). We can compute the amplitude for the decay
with Eq. (6.15) to leading order in g
Z ∞
hf |S|ii = hf |ii − ihf | dt HI (t)|ii + . . . (6.23)
Z −∞
= 0 − ighf | d4 x ψ † (x)ψ(x)φ(x) |ii (6.24)
We first expand φ ∼ a + a† . We will only need the a piece which The a† term turns |ii into a two
annihilates the initial meson state |ii into something ∼ |0i. We meson state with zero overlap with
the two nucleon state in |f i.
obtain,
Z ∞
hf |S|ii = ihf | dt HI (t) |ii + . . .
−∞
Z Z p
† d3 k 2Ep
4
= −ighf | d x ψ (x)ψ(x) 3
√ ak a†p e−ik·x |0i
(2π) 2Ek
Z
= −ighf | d4 x ψ † (x)ψ(x)e−ip·x |0i
and we have found our first quantum field theory amplitude! Notice
the δ-function which encapsulates the 4-vector conservation. In
particular, the decay will only take place if mφ ≥ 2Mψ .16 16
Ex: Show this!
We post-pone the steps to turn this into a something observable,
like a life-time of the meson. You can already see that it might in-
volve some additional insights since once we calculated probabilities,
we will get the square of a δ-function.
the vacuum
φ(x)φ(y) = h0|φ(x)φ(y)|0i
= h0|T {φ(x)φ(y)}|0i − h0| : φ(x)φ(y) : |0i
= h0|T {φ(x)φ(y)}|0i
= ∆F (x − y)
Z
† † d4 p i
ψ(x)ψ (y) = ψ (x)ψ(y) = e−ip·(x−y)
(2π)4 p2 − M 2 + iε
(6.28)
Note, that the only difference is the mass M of ψ. All other contrac-
tions vanish
+φ1 φ2 φ3 φ4 + φ1 φ2 φ3 φ4 + φ1 φ2 φ3 φ4 T {φ1 φ2 . . . φn }
−
= (φ+
1 + φ1 )(: φ2 . . . φn + contractions :)
+φ1 φ2 φ3 φ4 + φ1 φ2 φ3 φ4 + φ1 φ2 φ3 φ4 : The φ− †
1 ∼ a is already normal
ordered. To have the RHS as a
where a contraction across operators which are not adjacent, still normal ordered product, we need to
−
gives a factor of ∆F , e.g. commute φ+ 1 past the φk operators.
Each time φ+ 1 moves past one φ−
k , we
: φ1 φ2 φ3 φ4 : = ∆F (x1 − x3 ) : φ2 φ4 : pick up a factor
φ1 φk = ∆F (x1 − xk )
6.4.1 Example: nucleon scattering ψψ → ψψ Try it!
This proves Wick’s theorem by
We are scattering ψψ → ψψ and therefore have the in- and outgoing induction.
states
p p
|ii = |p1 p2 i = 2Ep1 2Ep2 b†p1 b†p2 |0i (6.29)
p p
|f i = |q1 q2 i = 2Eq1 2Eq2 b†q1 b†q2 |0i (6.30)
relativity, particles, fields 93
and we find
Z h
2 d4 k (2π)8
i(−ig) δ (4) (q1 − p1 − k)δ (4) (q2 − p2 + k)
(2π)4 k 2 − m2 + iε
i
+ δ (4) (q1 − p2 + k)δ (4) (q2 − p1 − k)
We finally obtain
1 1
i(−ig)2 + (2π)4 δ (4) (p1 + p2 − q1 − q2 )
(p1 − q1 )2 − m2 (p1 − q2 )2 − m2
22
We will get back to why we can
hf |S − 1|ii = i Af i (2π)4 δ (4) (pf − pi )
remove the δ-function later. We will
see it has to do with our assumption
where the factor i is by convention, since it reproduces the phase of free particles being in plane wave
states.
conventions for scattering in non-relativistic quantum mechanics.
P µ P µ
Further, pµf = µ
n pf,n and pi = n pi,n . We can ignore the iε,
because the denominator is never zero: in the center of mass frame,
we can write the momenta as23 23
Careful: p = |p| and not a 4-
vector!
p
p1 = ( p2 + M 2 , p, 0, 0) (6.33)
p
p2 = ( p2 + M 2 , −p, 0, 0) (6.34)
p
2 2
q1 = ( p + M , p cos θ, p sin θ, 0) (6.35)
p
q1 = ( p2 + M 2 , −p cos θ, −p sin θ, 0) (6.36)
and so
2 1 1
A=g 2 2
+ 2 (6.37)
2p (1 − cos θ) + m 2p (1 + cos θ) + m2
Note that the two terms are required because of Bose statistics. Scat-
tering into two identical particles at an angle θ is indistinguishable
from scattering at an angle θ − π, and so the probability must be
symmetrical under the interchange of the two processes. Since these
particles are bosons, the amplitude must also be symmetric.
This was a bit tedious. There is a much simpler way which we
ψ†
will discuss in a little bit using Feynman diagrams.
φ
6.5 Diagrammatic Perturbation theory: Feynman Diagrams
ψ † ψφ
Our final result for ψψ → ψψ scattering in Eq. (6.37) was remark- ψ
ably simple, even though the intermediate steps were a bit messy.
This motivates a diagrammatic short-hand called Feynman dia-
grams. These are pictures of the fields and contractions which we
Figure 6.3: Interaction vertex. Each
need to evaluate to give the matrix element. At the nth order in per- field in HI is represented by a line
turbation theory HI will act n times and so a Feynman diagram will emanating from the vertex. The
ψ † -line creates a nucleon or destroys
contain n interaction vertices. For our toy model the interaction an anti-nucleon, the ψ-line creates an
vertex looks like in Fig. 6.3. anti-nucleon or destroys a nucleon,
We need to distinguish ψ from ψ † and we will therefore draw an and the φ line creates or annihilates
a meson.
arrow on the line.
Next, contractions are the lines joining different vertices. Every
time there is a contraction, we join the lines of the contracted fields.
The example of Eq. (6.32) corresponds to the diagram Fig. 6.4.
φ1 φ2
Arrows will line up because ψ(x)ψ(y) = ψ † (x)ψ † (y) = 0 and
the arrows indicate the flow of the U (1) charge. Now, any fields
which are left uncontracted must either annihilate particles from the
incoming state or create particles from the outgoing state. If there q1
p1
are different ways of doing this, they correspond to indistinguishable
processes, so we must add the corresponding amplitudes: we write k
down a separate Feynman diagram for each distinct labeling of the p2 q2
external legs. For ψψ → ψψ scattering, this results in two Feynman
diagrams
We will revisit them below. p1 q2
k
6.5.1 Feynman rules p2 q1
1. For each particle in |ii and |f i, draw an external line. Assign a
directed momentum pi for each line, and add an arrow to denote Figure 6.5: Feynman diagrams
contributing at order g 2 to ψψ → ψψ
its charge. scattering, see Eq. (6.32). We show
the explicit momentum assignment
2. Join the external lines together with vertices as in Fig. 6.3. for the external and internal lines.
There are two distinct options.
3. At each vertex, write a factor of
X
(−ig)(2π)4 δ (4) ( ki )
i
P
with i ki is the sum of all momenta flowing into the vertex.
In fact, we can simplify things further and get rid of the trivial
delta-functions: we impose energy-momentum conservation on the
momenta flowing into each vertex and so we change
3.’ At each vertex, write a factor
(−ig)
i
k2 − m2 + iε
k
This is ok for graphs like the ones we have been considering.
However, there are also diagrams with closed loops where energy-
momentum conservation at the vertices does not fully fix all the
internal momenta! For example the matrix element obtained from p p+k p
Figure 6.6: Diagrammatic repre-
sentation of Eq. (6.38) where the
blue arrows show the momentum
assignments.
96 andreas weiler, tum
k is unconstrained . p+k
Figure 6.7: Diagrammatic repre-
sentation of Eq. (6.39) where the
blue arrows show the momentum
assignments.
relativity, particles, fields 97
(−ig)
2 i i
iA = (−ig) 2 2
+ (6.40)
(p1 − q1 ) − m (p1 − q2 )2 − m2
The interpretation of these diagrams is the following: the nucleons ψ
exchange a meson φ which has momentum k = p1 − q1 = p2 − q2 . This
meson does not satisfy the usual energy momentum relation, since as
we saw above k 2 = −2p2 (1 − cos θ) < 0 and therefore in particular
k 2 6= m2 ! The meson φ is called a virtual particle and which is
off-shell.25 Heuristically we can say, it cannot live long enough for 25
k2 = m2 would be on-shell.
its energy to be measured to great accuracy.
2 i i
iA = (−ig) + (6.41)
(p1 − q1 )2 − m2 (p1 + p2 )2 − m2
Note, since the diagrams are simply a shorthand for matrix elements
of operators in the Wick expansion, the orientation of the lines
inside the graphs have absolutely no significance. In the second
contribution, the virtual meson can go on-shell for certain values of
relativity, particles, fields 99
p2 k + q2 q2
i i
×
(k − q1 + p1 ) − M + iε (k + q2 ) − M 2 + iε
2 2 2
There are 3! contractions associated with this diagram, and the 1/3!
λ 3
HI = φ (x)
3!
The factor 1/3! is included since there are usually 3! different ways
of contracting the fields of a vertex with the fields of neighboring
vertices, and so the 1/3! gets canceled.
However, the 1/3! does not always get completely canceled be-
cause permutations of the lines might not give a new contraction.
We must therefore also include in the symmetry factor of a graph
the number of permutations of lines that leave the contraction
unchanged.
relativity, particles, fields 101
n!(η)n
S(G) =
r
S(G) = 1
2!(3!)2
S(G) = =1
(3 · 2)2 2
λ 4
• This is a 2nd order graph in 4! φ theory.
2!(4!)2
S(G) = =2
8·3·4·3·2
with r = 8 · 3 · 4 · 3 · 2 because we can connect the first external
point to 8 different lines emerging from the two vertices, after
102 andreas weiler, tum
that we can use the remaining 3 lines and the 4 · 3 lines from the
other vertex. The internal propagator is then fixed apart from a
two-fold ambiguity.
λ 4
• This is a 2nd order graph in 4! φ with two loops
2!(4!)2
S(G) = =6
8·4·3·2
with r = 8 · 4 · 3 · 3 because there are 8 · 4 ways to connect to the
external points and once we choose one of the internal lines we
have first 3 options, then 2 to connect it to the other vertex.
λ 3
• A 4th order 3! φ graph:
4!(3!)4
S(G) = =1
12 · 2 · 9 · 2 · 6 · 2 · 3 · 2
Most people never need to evaluate a diagram with a symmetry
factor greater than 2, so you should not worry too much about this
technicality.
6.7 Potentials
Let us look at the non-relativistic limit of our scattering amplitude
with the goal to treat it as a potential in non-relativistic quantum
mechanics.
( + m2 )φ(x) = δ 3 (x)
where we have skipped in the 2nd and 3rd line the steps which we
have derived carefully in Sec. 4.3.2. We close the contour in the
upper half plane p → +i∞ where we pick up the pole at p = +im,
which finally gives
1 −m r
φ(x) = e (6.46)
4πr
If you compare this to the ∼ 1/r of Coulomb’s law Eq. (4.95) you see
that the field now dies off exponentially at distances > 1/m, which is
the Compton wavelength of the meson.
Can we now understand the profile of the φ field as a force be-
tween ψ particles? Recall, that we treat the electro-static potential
sourced by a δ-function as the potential energy for another charged
(test) particle moving in this background.
Is there a classical limit of the scalar Yukawa theory with the ψ
particles as δ-function sources for φ, creating the Yukawa potential?
The answer is yes, at least in the limit M m, that is, when the
nucleons are much heavier than the mesons and act as quasi static
sources.
Recall the Born approximation 27 in non-relativistic quantum me- 27
See e.g. Cohen-Tannoudji, Quan-
chanics: at first order in perturbation theory, we find the amplitude tum Mechanics, Vol. II, Chapter
VIII, especially section B.4
to scatter an incoming state with momentum p of a potential U (r)
into an outgoing state with momentum q
Z
AQM (p → q) = hq|U (r)|pi = −i d3 r U (r)e−i(q−p)·r
i i
iA = −g 2 2 2
= g2 (6.47)
(p1 − q1 ) − m |p1 − q1 |2 + m2
where we have used the fact that in the center of mass frame the
energies of the incoming and scattered particles are the same, see
Eq. (6.33).
We need to account for the different normalizations of the rel-
ativistic and non-relativistic amplitudes and divide the result by
(2M )2 to account for the difference between relativistic and non-
relativistic normalization of states, see Sec. 5.4. We define the dimen-
sionless quantity
g
λ≡
2M
and obtain
Z
λ2
d3 r U (r)e−i(q1 −p1 )·r = − 2
|p1 − q1 | + m2
λ2 −mr
U (r) = − e
4πr
This is the Yukawa potential, where the minus sign tells us that it
is attractive. Yukawa made this potential28 the basis for his theory 28
To be precise: he used the fermion-
of the nuclear force and worked backwards from the range of the scalar version of this interaction
which however gives the same poten-
force (about 1 fm) to predict the mass (about 200 MeV) of the tial.
required meson, the pion.
We can directly from the scattering amplitudes for ψψ → ψψ, Ex: Check this!
ψ ψ̄ → ψ ψ̄, and ψ̄ ψ̄ → ψ̄ ψ̄ that the sign of the spin-0 generated
Yukawa term is always positive – it leads to a universally attractive
potential. Compare this to the Coulomb potential which is mediated
by the exchange of a spin-1 boson, which can be either attractive
and repulsive. Gravity, which is mediated by a spin-2 field is again
universally attractive.
Notice that quantum field theory has given us an entirely new
view of forces between particles. Rather than being a fundamental
concept, the force arises from the virtual exchange of other particles,
in this case the meson.
h i2
|hf |S − 1|ii|2 = |Af i |2 (2π)4 δ (4) (pF − pI ) (6.49)
Let us derive the familiar Fermi’s golden rule from Dyson’s formula
of Eq. (6.15). Consider two energy eigenstates |ni and |mi of H0 ,
with En 6= Em . We find in leading order in the perturbation Recall from Eq. (6.5), that the
interaction Hamiltonian in the
interaction picture is
HI = (Hint )I = eiH0 t (Hint ) e−iH0 t
relativity, particles, fields 105
Z t
hm|U (t)|ni = −ihm| dt0 HI (t0 )|ni
0
Z t
0
= −ihm|Hint |ni dt0 ei(Em −En )t
0
Z t
0
= −ihm|Hint |ni dt0 eiωt
0
eiωt − 1
= −hm|Hint |ni
ω
t2
with ω = Em − En . We obtain for the probability Pn→m (t) for the
sin2 (!t/2)
transition |mi → |ni after a time t, 4
!2
2 2 sin2 (ωt/2)
Pn→m (t) = |hm|U (t)|ni| = |hm|Hint |ni| 4
ω2
We show the function in brackets in Fig. 6.12. As you can see, after
a time t most of the transitions happen in a region between energy
eigenstates separated by !
⇡
2π t
∆E ≈
t sin2 (ωt/2)
h i
Figure 6.12: 4 ω2
vs ω.
As we take t → ∞, the function in brackets starts to approach a
δ-function. We get the normalization by
Z ∞
sin2 (ωt/2)
dω 4 = 2πt
−∞ ω2
and therefore
sin2 (ωt/2)
4 → 2πt δ(ω), for t → ∞
ω2
We consider now a transition to a cluster of states with density
ρ(E). In the limit t → ∞, we get
Z
sin2 (ωt/2)
Pn→m (t) = dEm ρ(Em )|hm|Hint |ni|2 4
ω2
→ 2πt |hm|Hint |ni|2 ρ(En )
|hf |S|ii|2
P =
hf |f ihi|ii
and
Y
hf |f i = 2Epl V
final states
1
P = |hf |S|ii|2 × (6.52)
hf |f ihi|ii
1 Y 1
= |Af i |2 (2π)4 δ (4) (pI − pF ) V T ×
2mV 2Epl V
final states
taking the T derivative (or just dividing by T ). One worry is still the hf |S|ii = −ig (2π)4 δ (4) (p − q1 − q2 )
sum over all final states. We proceed in two steps: first we put our = iAf i (2π)4 δ (4) (p − q1 − q2 )
theory in a box with side L assuming periodic boundary conditions and therefore
which allows us to count states. Then we integrate over all possible Af i = −g
momenta of all final states. in leading order.
relativity, particles, fields 107
where we have used Eq. (6.53) for the phase space of the final-states.
108 andreas weiler, tum
1 p1
dΠ2 = dΩ (6.56)
16π 2 ET
Here we have assumed that the particles A and B in the final state
are distinguishable and we treated the final states |A(p1 ), B(p2 )i and
|A(p2 ), B(p1 )i as distinct. Note, that if the final states are instead
identical, we would have double-counted by 2!. In general, we need
to multiply an n-body phase space of n identical particles in the final
states by a factor of
1
dΠn → dΠn
n!
iA = −ig
Figure 6.13: Pion decay into two
nucleons. This is kinematically
allowed in our toy model if m > 4M 2 .
Note, in nature the pion is the
lightest QCD resonance and would
not be able to decay to two nucleons.
relativity, particles, fields 109
N =F ·σ (6.58)
differential probability
dσ =
unit time · unit flux
|Af i |2 1
= · · dΠ
2E1 2E2 V F
We now need an expression for the unit flux F . Let us assume for
simplicity that we are in the center of mass frame of the collision.
We have so far considered a single particle per spatial volume V such
that the flux is given in terms of the 3-velocities as This is easy to see: consider a beam
of particles perpendicular to a plane
|v1 − v2 | of area A moving with v. If the
F = density of particles is n, then after a
V
time t, the total number of particles
We therefore get passing through the plane is
N = |v|A · t · n
1 1
dσ = · |Af i |2 · dΠ (6.59) The flux is therefore N/(At) = |v| · n.
4E1 E2 |v1 − v2 |
With our normalization, there is one
particle in the box with volume V , so
You can now use Eq. (6.59) and take your favorite scattering n = 1/V and the flux is
amplitude to compute the probability for particles to end up at
F = |v|/V
various angles.
If we have two colliding beams, the
probability of finding either particle
is 1/V but since the collision can
occur anywhere in the box, the total
flux is
which leads together with the two-body phase space of Eq. (6.56) to
1 1
dσ = · |Af i |2 · dΠ2
4E1 E2 |v1 − v2 |
1 E1 E2 1 pf
= |Af i |2 dΩ
4E1 E2 pi ET 16π 2 ET
1 1 pf |Af i |2
= dΩ
4pi ET 16π 2 ET
from which we obtain
dσ 1 1 pf
= |Af i |2 (6.60)
dΩ 64π 2 ET2 pi
H|Ωi = 0 (6.61)
hΩ|Ωi = 1
where we have used Eq. (6.61) in the first term. The second term
can also be simplified. In fact, it vanishes!
P R
The sum n is actually an integral d3 p . . . since the momentum
states form a continuum. We can now use the Riemann-Lebesgue Figure 6.14: The Riemann-Lebesgue
lemma. The function f (x) is multi-
lemma, which states that for any well-behaved function f (x), plied by a rapidly oscillating function.
Z β The product integrates to zero in the
limit of infinite oscillation frequency.
lim dx f (x) eiωx = 0
ω→∞ Proof:
α Z Z
f (x)e−iωx dx =
1 0 −iωx
We can alternatively derive this by shifting the integration slightly iω f (x)e dx
into the imaginary t → −∞(1 + iε), which exponentially suppresses all 1
Z
≤ |f 0 (x)| dx → 0 as ω → ±∞.
energy eigenstates except for the ground-state.29 The above result 29
See e.g. Peskin &|ω| Schroeder,
therefore gives, chapter 4.2
In particular, we see that the vacuum of the free theory and the
vacuum of the full theory are proportional to each other Since the result is true for any state
hΨ|, we have
|0i = hΩ|0i · |Ωi UH (t, −∞)|0i = |ΩihΩ|0i
†
|0i = UH (t, −∞)|ΩihΩ|0i
which justifies the somewhat hand-waving argument in Sec. 6.3.1 to
|0i = |ΩihΩ|0i
use free states, or creation operators acting on the free vacuum |0i,
in scattering processes.
If we insert the result of Eq. (6.65) in the RHS of Eq. (6.64), we find With
where we have used hΩ|Ωi = 1. This completes the proof of this very
important result which we can write in slightly more explicit form
4
R
h0|T φI (x1 ) . . . φI (xn ) e−i d x HI (x) |0i
hΩ|T φH (x1 ) . . . φH (xn )|Ωi = R
4
h0| e−i d x HI (x) |0i
(6.66)
2
2. Examine None⟨0|S|0⟩ todiagrams
of these in φ4totheory.
order λconnect Identifyand
external particles thewedifferent
there- diagrams with
forecontributions
call them vacuum 31
the different arising from We
bubbles. an will not show this
application of here
Wick’s , theorem.
31
See Confirm
however Sheet 7, Ex 2!
but the
2 pre-factors of the diagrams consisting of combinatoric and
that to order λ , the combinatoric factors work out so that the the vacuum to vacuum
symmetry factors conspire such that the whole series can be expo-
amplitude is given by the exponential of the sum of distinct vacuum bubble types,
nentiated
and we find that the amplitude for the vacuum of the free theory to
evolve into itself (the numerator of the RHS of Eq. (6.66))
3. Consider the Lagrangian for 3 scalar fields φi , i = 1, 2, 3, given by
3 h0|S|0i = exp(all distinct vacuum
3 bubbles) 3
! 1 1 ! λ ! 2 2
L= (∂µ φi )(∂ µ φi ) − m2 ( φ2i ) − ( φi ) (4)
i=1
2
The same combinatoric 2
simplification takes place for 8
i=1
the numerator
i=1
evaluating of Eq. (6.66) generic correlation functions. One can show See e.g. Schwartz 7.2.5 or Peskin 4.4.
Compare also to the exponentiation
Show that the Feynman propagator for the free
Xfield theory (i.e. λ = 0) isofof
thethe form
cluster expansion in statistical
h0|T φI (x1 ) . . . φI (xn ) S |0i = h0|S|0i · connected diagrams physics, see script 3.4.1 of SS17 of
the TH4A lecture by yours truly.
⟨0|T φi(x)φj (y)|0⟩ = δij DF (x − y) (5)
where again connected means no vacuum bubbles, or that all the
lines are joined to an external leg. We can now conclude that the
where DF (x − y) is the usual scalar propagator. Write down the Feynman rules of the
division by h0|S|0i has a very cute interpretation in terms of Feyn-
theory. Compute the amplitude for the scattering φi φj → φkFeynman
man diagrams. We only need to calculate the connected
φl to lowest order in λ.
graphs and we do not have to care for the vacuum bubbles. We
therefore obtain as promised Eq. (6.67)
X1
hΩ|T φH (x1 ) . . . φH (xn )|Ωi = connected Feynman graphs
g 3
Here we show another example for HI (x) = 3! φ (x) theory for a
two-point function
4
R
h0|T φI (x1 )φI (x2 ) e−i d x HI (x) |0i =
Z
1 1 1
D12 − g 2 3
D1x Dxx Dyy Dy2 + D12 Dxx Dxy Dyy + D12 Dxy + ... =
4 8 12
114 andreas weiler, tum
= +
[σi , σj ] = 2i εijk σk
are the exact same algebraic relations that the generator of infinites-
imal rotations satisfy, see Eq. (7.8) and the discussion below. We
anticipate the result and conclude that the changes induced by
rotations cancel in
B ·σψ
(1 ∂t − ∂i σi )ψ = 0 (7.4)
118 andreas weiler, tum
σ µ = (1, σ1 , σ2 , σ3 ) (7.5)
(σ µ ∂µ + m)ψ = 0 (7.7)
but this is not a Lorentz invariant equation. Please forget Eq. (7.7)
immediately (and remember only Eq. (7.6)).
In conclusion, we need to understand the intricate Lorentz prop-
erties of spin- 12 particles, discuss their representations. The Dirac
equation and its non-relativistic limit, the Schrödinger-Pauli equa-
tion, will follow.
gi × gj = gk
ηαβ = ηµν Λµ α Λν β
xµ → Λµ ν xν
and if you go back to Sec. 3.3.1, you’ll find examples of this embed-
ding as the set of 4 × 4 rotation and boost matrices. Our goal is now
to find all the representations.
Our goal is now to extract the properties of the group inde-
pendent of one particular representation. We will now focus on
continuously connected groups, which can be described by a set
of coordinates which are real numbers. The easiest approach is to A non-continuous group, is e.g. the
consider infinitesimal transformations: for any group G, group reflection group Z2 , which has the
elements 1, −1
elements g ∈ G can be written as
g = exp(iαig λi )
relativity, particles, fields 119
where the αig are just numbers and the λi are group generators. Gen-
erators λi form an algebra, because we can add and multiply them,
whereas group elements can only be multiplied. They can be ex-
tracted from the group elements using infinitesimal transformations
g = exp(iαig λi ) = 1 + iαig λi + . . .
or through
1 dg
λi =
i dαig αg =0
i
You can obtain J2 and J3 by straight-forward permutations. How Check that the commutator in
about the Lorentz-boosts? We recall Eq. (3.17) Eq. (7.8) is indeed satisfied!
! ! !
x cosh η sinh η x0
= (7.10)
t sinh η cosh η t0
x → x+ηt
t → t+ηx
and analogously
0 0 1 0 0 0 0 1
0 0 0 0 0 0 0 0
iK2 =
1
, iK3 =
0 0 0 0 0 0 0
0 0 0 0 1 0 0 0
What does that mean? This just reflects that the boost generators
Ki transform as a 3-vector under rotations, as expected since the
boost direction is a 3−vector characterized by the relative velocity v i .
We will now do one of the most important calculations in history. Under a rotation we would get for
Compute [K1 , K2 ]. What do we get? the boost group elements U =
exp(iη l Kl ) which are 4 × 4 matrices,
0 0 0 0 0 0 0 0 U → G−1 l
i U Gi = (1 − iJi )(1 + iη Kl )(1 − iJi ) + . . .
0 0 −1 0 0 0 0 0 = 1 + iη l Kl + η l [Kl , Ji ] + . . .
K1 K2 − K2 K1 =
0 0
−
0 0 0 −1 0 0
and so
0 0 0 0 0 0 0 0
Kl → i[Ji , Kl ]
0 0 0 0
0 0 −1 0
= = −i J3
0 1 0 0
0 0 0 0
7.1.1 Representations
We can now simply use our knowledge of SU (2) representations
to determine all representations of SO(1, 3). We know that the
representations are labeled by their spin
j = 0, 12 , 1, 32 , . . . (7.18)
ψm with m = −j, −j + 1, . . . , j − 1, j
1
Ji =
σi
2
1
iKi = σi
2
after solving the simple linear system. The equality here means
”represented by”.
We will similarly denote the ( 12 , 0) with the slightly exotic symbol We use dotted and undotted indices
to make sure we don’t accidentally
contract the α of ( 12 , 0) with an
ψR = χ̄α̇ , with α̇ = 1, 2 (7.20)
α̇ = 1, 2 of ( 12 , 0), which would not be
Lorentz invariant. This is called van
Do not be confused by the bar. For now it is just part of the typo- der Waerden notation. It will prove
graphic symbol. extremely useful in supersymmetry
and in superstring theory.
The two-component spinors ψα and χ̄α̇ are called Weyl spinors
and furnish perfectly good representations of the Lorentz group.
What about the Dirac spinor? If you have already encountered
it, you know that it has also 4 components. Why? The reason is
parity. We know that under parity, see XXXX, that
x → −x
t→t
and similarly for other 3-vectors like the 3-momentum or the boosts
p → −p
K → −K
Rotations are pseudo-vectors and they transform as Think of the angular momentum as
L = r × v or imagine a clock-wise
rotation in a mirrored coordinate
J →J
system.
( 12 , 0) ↔ (0, 12 )
( 21 , 0) ⊕ (0, 12 )
This reducible representation is thus given by stacking two Weyl QED and QCD are parity invari-
spinors to form a Dirac spinor ant, which means the equations of
motions and the Lagrangian are
! ! invariant under P transformations. If
ψL ψα had chosen to write the theory using
Ψ= = just the ( 12 , 0) spinor, we would have
ψR χ̄α̇
explicitly violated P. This is in fact,
what happens in the electro-weak
sector: SU (2)L gauge bosons only
couple to ( 12 , 0) spinors and not to
the (0, 21 ) representation.
relativity, particles, fields 123
Note, the very important minus sign, which distinguishes the two
irreducible representations entering a Dirac spinor.
Parity forced us to us a 4-spinor but we know that a electron has
only two physical degrees of freedom. We must project out two of
four the components described by Ψ(p). Let us go to the rest-frame
(pµ )r ≡ (m, 0)
P2 = P (γ 0 )2 = 1
ψL ↔ ψR
we cannot simply use the projection operator to set one of the fields
(e.g. χ̄α̇ ) to 0. Parity means that the two irreducibles should be on
the same footing. We choose the following representation of γ 0
!
0 0 12×2
γ =
12×2 0
P Ψ(pr ) = 0
ψL − ψR = 0 (7.21)
124 andreas weiler, tum
we define
γ µ pµ
≡ e−iη·K γ 0 eiη·K
m
to obtain the Dirac-equation
and with the unit vector η/|η| we get Ex: Show this, too! Use {σi , σj } =
σi σj + σj σi = 2δij
η
exp(η · σ) = cosh |η| + · σ sinh |η|
|η|
If we identify the relation between boost and the 3-momentum and
the relevant boost as Ex: Show this, too! You can imme-
diately convince yourself, that this is
η the correct result in 1 + 1 dimensions,
p=m sinh |η| (7.24)
|η| since
{γ µ , γ ν } = 2g µν 14×4 (7.25)
and we can write the Dirac matrices in the Weyl or chiral basis as
!
µ 0 σµ
γ = (7.27)
σ̄ µ 0
J ij = −i(xi ∇j − xj ∇i )
J µν = i(xµ ∂ ν − xν ∂ µ ) (7.28)
J i → J ij and K i → J 0i
126 andreas weiler, tum
ωµν = −ωνµ
and
i µν
Λ = e− 2 ωµν J
[J µν , J ρσ ] = i(η νρ J µσ − η µρ J νσ − η νσ J µρ + η µσ J νρ ). (7.29)
[S µν , S ρσ ] = i(η νρ S µσ − η µρ S νσ − η νσ S µρ + η µσ S νρ ). (7.35)
relativity, particles, fields 127
Ψ(x) → S[Λ]Ψ(Λ−1 x)
(γ 0 )† = γ 0 , (γ i )† = −γ i
We see that there is no way to pick the γ µ such that the exponent is
hermitian and the representation is therefore non-unitary.
Ψ(x) → S[Λ]Ψ(Λ−1 x)
Ψ† (x) → Ψ† (Λ−1 x)S[Λ]†
γ 0 γ µ γ 0 = (γ µ )†
Ψγ µ Ψ → Ψ S[Λ]−1 γ µ S[Λ] Ψ
Z
S= d4 x Ψ(x)(iγ µ ∂µ − m)Ψ(x) (7.42)
Aµ γ µ = A
/
(i∂/ − m)Ψ = 0
L = Ψ(iγ µ ∂µ − m)Ψ
† µ † µ † †
= iψL σ̄ ∂µ ψL + iψR σ ∂µ ψR − m(ψL ψR + ψR ψL ) (7.45)
iσ̄ µ ∂µ ψL = 0
iσ µ ∂µ ψR = 0
∂L
πΨ = = iΨ†
∂ Ψ̇
which is not proportional to the time-derivative of Ψ. Therefore, the
phase space of the spinor is parametrized by Ψ and Ψ† , while for a
scalar it is φ and π = φ̇. So the phase space is described by 8 real
dimensions and consequently the Dirac spinor has 4 real degrees of
freedom.
7.1.8 γ5
In our choice of basis for the γ µ matrices, the Lorentz-transformations
S[Λ] came out block-diagonal. What happens if we choose a different
representation of the Clifford algebra?
γ µ → U γ µ U −1 , and Ψ → UΨ
S[Λ] will now likely not be block-diagonal. Can we define the Weyl
spinors in an invariant way? We introduce a ”fifth” gamma-matrix8 8
You can convince yourself that we
can also write γ 5 as
γ 5 = iγ 0 γ 1 γ 2 γ 3 γ5 =
i
εµνσρ γ µ γ ν γ σ γ ρ
4!
which satisfies which in the Weyl basis looks like using the totally anti-symmetric four
index tensor with ε0123 = 1.
!
−1 0 γ 5 is called the ”fifth” gamma-
2 2
γ5 = (7.46) matrix, because with ΓA = γ A for
02 12 A = 0, 1, 2, 3 and Γ4 = iγ 5 it satisfies
the five-dimensional Clifford algebra
{ΓA , ΓB } = 2η AB
µ 5 5 2
{γ , γ } = 0, and (γ ) = +1
[Sµν , γ 5 ] = 0
1
PL/R = (1 ∓ γ 5 )
2
such that
PL2 = PL , PR2 = PR , PL PR = 04
with Eq. (7.46), we see that PL/R projects into the Weyl-basis
ψ L = PL Ψ
ψ R = PR Ψ
132 andreas weiler, tum
7.2.1 Parity
We already saw that under parity P the left-handed and right-
handed spinors are exchanged
P : ψL ↔ ψR
Using this, we can write the action of parity on the Dirac spinor
itself as
P : Ψ(x, t) → γ 0 Ψ(−x, t)
Notice that the Dirac Lagrangian conserves parity as you can easily
see from Eq. (7.45). Let us now discuss how our interaction terms
change under parity. The ”mass” bi-linear transforms as
Ψγ 5 Ψ, and Ψγ 5 γ µ Ψ
Ψ(x) Γ Ψ(x)
for various Γ matrices. The total number of bi-linears is 16, which What about Γ = γ [µ γ ν γ σ] or
is exactly what we could hope for from a 4 × 4 component object. Γ = γ [µ γ ν γ σ γ ρ] ? The brackets
indicate anti-symmetrization, since
We have now a set of terms to build covariant Lagrangians with. we do not need to consider the
If we add γ 5 to our terms then these terms will typically break symmetric part which is ∼ η µν (anti-
symmetrization also explains why we
consider at most 4-tensors). One can
show that γ [µ γ ν γ σ] = −iεµνσρ γρ γ 5
and γ [µ γ ν γ σ γ ρ] = −iεµνσρ γ 5 and
so they are linearly related to the
simpler bilinears with Γ = γ ρ γ 5 and
Γ = γ5.
relativity, particles, fields 133
1 scalar 1
γµ vector 4
σ µν = 2i [γ µ , γ ν ] tensor 6
γµγ5 pseudo-vector 4
γ5 pseudo-scalar 1
16
differently, is called a chiral theory. QCD and QED are vector-like, L ⊃ e†L heR
and the electro-weak sector including the Higgs are chiral theories.
Ψ = u(p)e−ip·x , with p2 = m2
(γ µ pµ − m)u(p) = 0
ξ†ξ = 1
√
and we added m for future convenience. The interpretation for the
ξ two spinor is given by the transformation under rotations, where
ξ transforms as an ordinary two-component spinor of the rotation
group, see Eq. (7.43). We have
ξ → exp( 2i θ · σ) ξ
then the particle has spin up (↑) along the z-direction. Notice again,
that after applying the Dirac equation we are free to choose only 2 of
the four components u(p).
Let us now apply a boost in the z-direction
! !
E m cosh η
=
p3 m sinh η
Now apply the same boost to u(p). With Eq. (7.44) we get Show this!
√ !
p·σξ
u(p) = √ (7.48)
p · σ̄ ξ
(p · σ)u2 = m u1 (7.49)
(p · σ̄)u1 = m u2 (7.50)
We can see that either of the two equations implies the other using
Let us try the ansatz u1 = p · σξ 0 for some ξ 0 then Eq. (7.50) immedi-
ately gives
u2 = mξ 0
Ψ = v(p) e+ip·x
√ !
p·σχ
v(p) = √ (7.51)
− p · σ̄ χ
7.3.2 Examples
We take again the positive frequency solution with mass m at rest
p=0
!
√ ξ
u(pr ) = m
ξ
This expression in fact also makes sense for a massless particle with
E = p3 . For a massless spinor we therefore obtain We picked the normalization such
that this limit would make sense.
0
√ 0
u(p) = 2E
1
0
7.3.3 Helicity
We finally explain the meaning of ”left-handed” for ψL and ”right-
handed” for ψR . The massless solutions are eigenstates of the helic-
ity operator
!
1 σi 0
h ≡ p̂ · S = p̂i (7.52)
2 0 σi
with
and similarly
v̄ r (p) · us (p) = 0
and the same holds for (p · σ̄)(p0 · σ̄). This means the two terms in
Eq. (7.54) cancel and we get
Note, the two spinors are not contracted, but placed back to back to
give a 4 × 4 matrix Similarly Think of |aihb| instead of hb|ai.
2
X
v s (p)v̄ s (p) = p
/−m
s=1
Now
2
X
ξ s ξ s† = 12
s=1
δΨ = εµ ∂µ Ψ
and use the standard formula Eq. (4.67) for the energy-momentum
tensor Recall
∂L
(j µ )ν = ∂ν φa − δνµ L ≡ T µ ν
T µν = iΨγ µ ∂ ν Ψ − L ∂(∂µ φa )
using the fact that L does not depend on ∂µ Ψ. Since the current
is only conserved on-shell, we can impose the equations of motions
directly on T µν and we won’t lose anything. Since the equations of
motions are first order, this affects the energy-momentum tensor. We
use (iγ µ ∂µ − m)Ψ(x) = 0 which means, we can set L = 0, to obtain
T µν = iΨγ µ ∂ ν Ψ
relativity, particles, fields 139
Ψ → eiα Ψ
jVµ = Ψγ µ Ψ (7.60)
L = Ψ(x)(iγ µ ∂µ − m)Ψ(x)
π = iΨ†
we obtain
and
n o
Ψα (x), Ψ†β (y) = δαβ δ (3) (x − y) (7.67)
r s†
bp , bq = (2π)3 δ rs δ (3) (p − q) (7.68)
r s†
cp , cq = (2π)3 δ rs δ (3) (p − q) (7.69)
for all creation operators. The Hamiltonian still has nice commuta-
tion relations with the anti-commuting b and c operators. You can
easily see with Eq. (7.65) that
[H, bs† s†
p ] = Ep bp
and exactly the same for b ↔ c. We can treat them therefore again
as ladder operators creating a tower of energy-eigenstates by acting
with bs† and cs† on the vacuum, just as in the bosonic case. The (not
relativistically normalized) one-particle states are
|p, ri = br†
p |0i
|p1 , r1 ; p1 , r1 i = 0
As for the scalar field (which had spin 0), we could act with the
angular momentum operator and confirm that a particle at rest
|p = 0, ri
and
Z
Q = d3 x Ψ† Ψ
φ(x)φ(y) = φ(y)φ(x)
and the order here doesn’t matter. For fermions this no longer
holds! If (x − y)2 < 0, Fermi fields anticommute. For (x − y)2 < 0 and
if we are in a frame where x0 > y 0
Ψ(x)Ψ(y) for x0 > y 0
T {Ψ(x)Ψ(y)} = (7.73)
−Ψ(y)Ψ(x) for y 0 > x0
in particular we have
Ψ = Ψ(+) + Ψ(−)
: Ψ1 Ψ2 Ψ3 Ψ4 : = − : Ψ1 Ψ3 Ψ2 Ψ4 : = − Ψ1 Ψ3 : Ψ2 Ψ4 :
∂Ψ
= i[H, Ψ]
∂t
which is the usual operator time evolution. As before, we can solve
this by expanding (replacing ix · p → −ix · p in the exponent)
2 Z
X d3 p
Ψ(x) = p bsp us (p)e−ip·x + cs† s
p v (p)e
ip·x
(7.74)
s=1
2Ep
2 Z
X
† d3 p
Ψ (x) = p bs† s†
p u (p)e
ip·x
+ csp v s† (p)e−ip·x (7.75)
s=1
2Ep
relativity, particles, fields 145
Inserting the expansion Eq. (7.74) we obtain We will in the following often drop
the indices, but you should remember
Z h
d3 pd3 q 1 −i(p·x−q·y)
that S(x − y) = {Ψ(x), Ψ(y)} is a
iSαβ (x − y) = p {bsp , br† s r
q } uα (p)ūβ (q)e 4 × 4 matrix.
(2π)6 4Ep Eq
i
+{cs† r s r
p , cq } vα (p)v̄β (q)e
+i(p·x−q·y)
in terms of the propagator for the real scalar field D(x − y). If we stay Recall, we can write
away from singularities, the propagator satisfies Z
d3 p 1 −ip·(x−y)
D(x − y) = e
(2π)3 2Ep
(i∂/x − m)S(x − y) = 0 with pµ = (Ep , p) or p2 = m2 .
and
Z
d3 p 1 +ip·(x−y)
h0|Ψβ (y)Ψα (x)|0i = / − m)αβ e
(p
(2π)3 2Ep
= (−i∂/x − m)D(y − x)
SF (x − y) = Ψ(x)Ψ(y)
146 andreas weiler, tum
or
h0|Ψ(x)Ψ(y)|0i x0 > y 0
SF (x − y) = h0|T Ψ(x)Ψ(y)|0i ≡
h0| − Ψ(y)Ψ(x)|0i y 0 > x0
= (i∂/x + m)φ(x)φ(y)
where the Feynman propagator for scalar fields is If you are worried about the time-
Z derivative of the θ function, you can
d4 p −ip·(x−y) i see that
φ(x)φ(y) = ∆F (x − y) = e
(2π)4 p2 − m2 + iε ∂t θ(x0 − y0 )D(x − y) = δ(x0 − y0 )D(x − y) = 0
The integral representation of the Feynman propagator for Dirac using the properties of the scalar
propagator.
spinors is therefore We can write Eq. (7.77) also as
d4 p e−ip·(x−y)
Z
Z SF (x − y) = i
d4 p −ip·(x−y) p
/+m (2π)4 p
/ − m + iε
SF (x − y) = i 4
e (7.77)
(2π) p − m2 + iε
2
which is a commonly used short
hand. We see that like for the scalar
field, the Feynman propagator is just
which satisfies the inverse of the Dirac equation
(with suitable ε pole structure).
(i∂/x − m)SF (x − y) = iδ (4) (x − y)
1. Propagators:
i
φ(x)φ(y) = =
p2 − µ2 + iε
i(p
/ + m)
Ψ(x)Ψ(y) = =
p2 − m2 + iε
2. Vertices:
−iλ =
p
= us (p)
p
= v̄ s (p)
p
= ūs (p)
p
= v s (p)
+ (7.82)
p p0 0
p p
"
i
iA = (−ig)2 ū(q 0 ) · u(q) ū(p0 ) · u(p)
(q − q 0 )2 − µ2
#
0 i 0
−ū(p ) · u(q) 0 ū(q ) · u(p) (7.83)
(p − q)2 − µ2
Where we have suppressed the spinor spin labels. The minus sign
between the diagrams is a reflection of Fermi-Dirac statistics.
+ p p0 (7.84)
p p0
p0 q0
p q
150 andreas weiler, tum
Z 0 0 0
4 d4 k k/ + m k/ + p
/ +m k/ + p
/ − /q + m k/ − p
/+m
iA = −(−iλ) Tr 2
(2π)4 k − m2 (k + p0 )2 − m2 (k + p0 − q0 )2 − m2 (k + p)2 − m2
where we have suppressed the +iε piece in the denominator. Note,
that the integral diverges logarithmically20 In the k → ∞ limit, the 20
Compare to the pure scalar case
integral behaves like which was convergent.
Z 4
d k
iA ∼ ∼ ln Λ
k4
You will have to come to the QFT course next semester to learn how
to make sense of results like these. There is an overall minus sign Like for the cosmological constant in
sitting in front the of the amplitude. Sec. ?? we will have to regularize and
renormalize the result in a systematic
In complicated diagrams, we can often find the minus signs, by way.
noting that the product (ΨΨ) or any other pair of fermions is a
commuting object. So,
λ2 e−µr
U (r) = −
4πr
If we repeat the calculation for ΨΨ → ΨΨ, two minus signs can-
cel and the Yukawa interaction once again leads to an attractive
potential.
8
Quantum Electro Dynamics
What are the Feynman rules? We can easily guess, even though the
theory behind them is quite involved.
1. Interaction vertex:
−ieγ µ =
2. Propagators:
−iηµν
Aµ (x)Aν (y) = =
q 2 + iε
p
= ∗µ (p)
Incoming
p
= µ (p)
Fµν = ∂µ Aν − ∂ν Aµ (8.3)
∂λ F µν + ∂µ F νλ + ∂ν F λν = 0
B =∇×A
∇·B =0
∂B
= −∇ × E
∂t
The equations of motions give the two remaining Maxwell equations
∇·E =0
∂E
=∇×B
∂t
Note, the interesting electro-magnetic
duality E → B and B → −E!
relativity, particles, fields 155
ψ → eiβ ψ
(ηµν (∂ λ ∂λ ) − ∂µ ∂ν )Aν = 0
156 andreas weiler, tum
Aµ , and Aµ + ∂µ λ
∂µ ∂ µ λ(x) = 0
∇·A=0 (8.7)
∇·E =0
∂µ ∂ ν A = 0
ξ(p) · p = 0
r (p) · p = 0
and
and
This is not correct! Why? It does not work with the gauge and
Gauss constraints. We need to have imposed on the operators
∇·A=∇·E =0
You can see this problem already in the classical Poisson bracket
structure.3 The correct Poisson bracket structure implies different 3
See e.g. P. Ramond - QFT, or
commutation relations in Coulomb gauge Dirac ”Lectures on Quantum me-
chanics”.
∂i ∂j
[Ai (x), E j (y)] = i δij − 2 δ (3) (x − y) (8.10)
∇
2 Z
X d3 p 1
A(x) = 3
p r (p) arp eip·x + arp † e−ip·x
r=1
(2π) 2|p|
2 Z p
X d3 p (−i) |p|
E(x) = 3
√ r (p) arp eip·x − arp † e−ip·x
r=1
(2π) 2
and
where you will have to use the outer product (or completeness
relation) for the polarization vectors
2
X pi pj
ir (p)jr (p) = δ ij −
i=1
p2
Where explicitly denote with tr, the transverse part of the photon.
Let us therefore explore the manifestly invariant Lorentz gauge.
∂ µ Aµ = 0
∂ µ ∂µ Aν = 0
1 1
L = − Fµν F µν − (∂µ Aµ )2
4 2
we obtain
∂µ F µν + ∂ ν (∂µ Aµ ) = ∂ µ ∂µ Aν = 0
non-vanishing
∂L
π0 = = −∂µ Aµ (8.11)
∂(∂t A0 )
∂L
πi = = ∂ i A0 − ∂ 0 Ai (8.12)
∂(∂t Ai )
We impose the equal-time commutation relations in Lorentz-gauge
and
Expanding again
3 Z
X d3 p 1
Aµ (x) = 3
p λµ (p) aλp eip·x + aλp † e−ip·x
(2π) 2|p|
λ=0
3 Z p
X d3 p +i |p| λ
πµ (x) = 3
√ µ (p) aλp eip·x − aλp † e−ip·x
(2π) 2
λ=0
Note that we here have (+i) in the momentum, rather than the
usual (−i). The reason is that the conjugate momentum is
π µ = −Ȧµ + . . .
which implies
0
(µ )λ (ν )λ ηλλ0 = ηµν
1 · p = 2 · p = 0 (8.14)
For later use, we note for the time-like and longitudinal polarizations
that
This means trouble. Let us see what it implies e.g. for one-particle
states. With the vacuum
aλp |0i = 0
∂L
π0 = = −∂µ Aµ
∂(∂t A0 )
∂µ Aµ |ψi = 0 (8.16)
on all the ”good” states. However this does not work either. It is
−
still too strong, since we can e.g. decompose Aµ = A+
µ + Aµ with
3 Z
X d3 p 1
A+
µ (x) = p λ (p) aλp e−ip·x
(2π)3 2|p| µ
λ=0
3 Z
X d3 p 1
A−
µ (x) = p λ (p) aλp † e+ip·x
(2π)3 2|p| µ
λ=0
A+
µ |0i = 0
∂ µ A−
µ |0i =
6 0
which shows that the vacuum as a physical state isn’t even one of
the good states.
where |ϕT i contains the transverse photons (created by a1p † and a2p † )
and |ϕL−0 i contains time-like and longitudinal photons (created by
a0p † and a3p † ).
relativity, particles, fields 163
What does this mean? Condition Eq. (8.18) requires all physical
states to contain pairs of timelike and longitudinal photons: for each
timelike photon with p, it must also contain a longitudinal photon
(n)
with the same p. We can write |ϕL−0 i as a sum over states |ϕL−0 i
containing n pairs of timelike and longitudinal photons
∞
X (n)
|ϕL−0 i = cn |ϕL−0 i (8.19)
n=0
(0)
where |ϕL−0 i = |0i is just the vacuum.
One can show that Eq. (8.18) indeed decouples the negative norm
states and the all the remaining timelike and longitudinal photons
have zero norm
(n) (n)
hϕL−0 |ϕL−0 i = δn,0 δm,0
Z " 3 #
d3 p X
i† i 0† 0
H= |p| ap ap − ap ap
(2π)3 i=1
∂µ F µν = Rν
∂ν ∂µ F µν = 0 = ∂ν Rν
∂µ j µ = 0
Ψ → eiα Ψ
j µ = Ψγ µ Ψ (8.21)
Ψ → e−ie λ(x) Ψ
which shows that the covariant derivative has the convenient prop-
erty that it only picks up a factor exp(−ieλ(x)), with the derivative
canceling the transformation of the gauge field. The whole La-
grangian Eq. (8.26) is then gauge invariant, since
Ψ → Ψe+ie λ(x)
/ invariant.
which leaves the Lagrangian and in particular iΨDΨ
Dµ = ∂µ + ieAµ
This tricks works for any theory. Replacing the derivatives by co-
variant derivatives in suitable representations works in almost any
theory. This procedure is called minimal coupling.
We find again that this is the number of particles minus the num-
ber of antiparticles. The value of the coupling constant in QED is
usually written in terms of the fine-structure constant There is a subtlety here, in that
these couplings are scale dependent
2
e 1
α= ≈ (8.28) α = α(µ)
4π~c 137
√ Once you include loop corrections,
which in natural units means for e = 4πα ≈ 0.3. the value of the fine-structure con-
stant grows logarithmically as
the energy scale is increased. The
8.7 Discrete symmetries of QED value here is associated with the
energy scale of the electron mass, so
In addition to the Lorentz transformations which are continuously α(µ = me ) ≈ 1/137, however e.g. at
the mass of the Z boson
connected to the 1, there are two other space-time operations which
can be symmetries of the Lagrangian: parity P and time-reversal α(µ = MZ ) ≈ 1/129
T . Additionally, there can be a particle-antiparticle symmetry called The scale dependence is encoded in
the renormalization group equation
charge conjugation C. We have already encountered parity as which you will discuss in the QFT
lecture
P : (t, x) → (t, −x)
∂e(µ) e(µ)3
≡ β(e) = + ...
which reverses handedness. Time reversal acts as ∂ ln µ 12π 2
where just show the leading order
T : (t, x) → (−t, x) dependence. The running of the
coupling is a loop effect as you see
both leave the Minkowski interval x2 = t2 − x2 invariant. Formally, from its e(µ)3 dependence.
we can say that the full Lorentz group breaks into four disconnected
subsets. We call the continuous Lorentztransformation the proper,
orthochronous Lorentz group L↑+
P
L↑+ L↑−
T T
P
L↓+ L↓−
relativity, particles, fields 167
8.7.1 Parity
We have already discussed parity extensively. Let us just add the
result for the gauge field and any four-vector
P : Ψ(x, t) → γ 0 Ψ(−x, t)
iγ µ (∂µ + ie Aµ )Ψ = 0
−iγ µ∗ (∂µ − ie Aµ )Ψ = 0
iγ µ (∂µ − ie Aµ )ΨC = 0
with
ΨC ≡ Cγ 0 Ψ∗
168 andreas weiler, tum
Cγ 0 γ µ∗ γ 0 C −1 = −γ µ
(γ µ )T = −C −1 γ µ C
C = −iγ 2 γ 0
ΨC = −iγ 2 Ψ∗ (8.30)
and vice versa. We can explicitly check that ΨC transforms as a This is due to the fact the represen-
spinor with the known Lorentz transformation, Eq. (7.36), Ψ → tations of SU (2) are pseudo-real.
i µν A representation and its complex
e− 4 ωµν σ Ψ we have for the complex conjugate conjugate are related to each other
µν ∗
by a simple transformation S, e.g.
i
Ψ∗ → e+ 4 ωµν (σ )
Ψ∗ for the complex doublet 2
2∗ = S −1 2S
hence
2 it is just
in this case with εij = iσij
2 + 4i ωµν (σ µν )∗ ∗
ΨC → −iγ e Ψ 2∗i = εij 2j
− 4i ωµν (σ µν )
=e ΨC Since for µ, ν 6= 2
∗
i
where we have used the fact, that [γ 2 , γ µ γ ν ] = 0 for µ, ν 6= 2 (σµν )∗ = [γµ , γν ]
2
and {γ 2 , γ 2 γ ν } = 0 for ν =
6 2. This shows that ΨC transforms i ∗ ∗
= − [γµ , γν ]
like a regular Dirac spinor and confirms our statement about the 2
i
representations of ψL and ψR being truly exchanging roles. = − [γµ , γν ]
2
= −σµν
since these γ matrices are real (only
γ 2 is purely imaginary). If one of the
indices µ = 2 or ν = 2, then e.g.
i
(σ2ν )∗ = − [(−)γ2 , (±)γν ]
2
relativity, particles, fields 169
/ = mΨC
i∂Ψ (8.31)
ΨT CΨ = Cαβ Ψα Ψβ
but since
!
−iσ2 0
C = −iγ2 γ0 =
0 iσ2
we have Cαβ = −Cαβ this seems to indicate that the term has
to vanish if the spinor fields Ψα commute. In your QFT lecture,
you will learn that spinors Ψ have to be treated as anticommuting
Grassmannian numbers,
Ψ1 Ψ2 = −Ψ2 Ψ1
- q
=
Z
-ie J
4
x e-iq·x
d−iq·x
= −ie d4 xe hΩ|T jµ(OJ T Jµ(x)(8.33)
(x)|Ωi JO), (10.5)
-
where jl' = Ψγisµ Ψ
= j'ljryl''lj;
where µ the electromagnetic
is the electro-magneticcurrent operator.The
current operator. But the vacuum
expectation value of jl' must vanish
µ by Lorentz invariance,
vacuum expection value of j has to vanish because of Lorentz- since otherwise it
would be a preferred
invariance: 4-vector.
otherwise j µ would be a preferred direction in Minkowski
Thespace.
photon one-point function also vanishes for a second reason: charge-
conjugation
Theinvariance.
photon one-point Recall that C
function hasistoa vanish
symmetry ofother
also for QED, rea-so C JO) = JO).
But jl'(x)
sons,changes
namely charge sign under charge
conjugation Since CCjl'(x)Ct
conjugation,
invariance. is symmetry = of-jl'(x), so its
vacuumQED,
expectation value must vanish:
we know that the vacuum |0i and the vacuum of the interact-
ing theory
C|Ωi = |Ωi
The same argument applies to any vacuum expectation value of an odd num-
are C invariant. A currents.
ber of electromagnetic diagram with n external photons
In particular, (and nothree-point
the photon external function,
fermions) is proportional to
Fig. 10.2d, vanishes. (This result is known as Furry's theorem.) It is not hard
to check explicitly that the hΩ|Tphoton
jµ1 (x1 ) . .one- and
. jµn (x three-point functions vanish in
n )|Ωi
the leading order of perturbation theory (see Problem 10.1).
TheIf remaining
n is odd thenamplitudes in Fig. 10.2 are all nonzero, so we must analyze
their structures in more detail. Consider, for example, the electron self-energy
hΩ|T j (x ) . . . j (x )|Ωi = hΩ|T CCj (x )CC . . . CCj (x )CC|Ωi
µ1 1 µn n µ1 1 µn n
This means that the a pure photon n-point function vanishes, for n
odd. This is known as Furry’s theorem.
1
with for concreteness think H = − 2m ∇2 − V (x). We now consider
the time reversal transformation
t → t0 = −t
Our goal is now to find ψ 0 (t0 ) which satisfies
∂
iψ(t0 ) = Hψ 0 (t0 ) (8.35)
∂t0
We write for the transformed field
ψ 0 (t0 ) = T ψ(t)
where T is an operator which we will determine in the following. Let Again, we can fix it up to an arbi-
us plug this into Eq. (8.35), we obtain trary phase.
∂
i T ψ(t) = HT ψ(t)
∂(−t)
∂
T −1 i T ψ(t) = T −1 HT ψ(t)
∂(−t)
Since H is time-independent by assumption, the time-inversion
operator has no effect and
T −1 H = H T −1 ,
which results in
∂
T −1 (−i)T ψ(t) = Hψ(t)
∂t
Comparing to Eq. (8.34), we are no forced to conclude
T −1 (−i)T = i (8.36)
We can define
T ≡UK
where K complex conjugates everything to the right. Applied to For real φ1 , φ2 we have
Eq. (8.36) we get with T −1 = KU −1 K(φ1 + iφ2 ) = φ1 − iφ2
Ψ0 (t0 ) = T Ψ(t)
KU −1 HU K = H
KU −1 γ 0 U K = γ 0
and
KU −1 (iγ 0 γ i )U K = iγ 0 γ i
U −1 γ 0 U = γ 0∗ , and U −1 γ i U = −γ i∗
and so
0 1
−1 0
Ψ(t, x) → γ 1 γ 3 Ψ(−t, x) = Ψ(−t, x)
0 1
−1 0
Thus T flips the spin of particles.
q q0
q
q0
+ (8.37)
p p0
p p0
relativity, particles, fields 173
"
(−iηµν )
2
iA = (−ig) ū(q 0 )γ µ u(q) ū(p0 )γ ν u(p)
(q − q 0 )2 − µ2
#
0 µ (−iηµν ) 0 ν
− ū(p )γ u(q) 0 ū(q )γ u(p) (8.38)
(p − q)2 − µ2
Compared to the scalar mediator we seem to find an overall minus
sign, but remember that the the expression is really positive for
µ, ν = 1, 2, 3.
p p0
174 andreas weiler, tum
which gives
" #
2 0 µ (−iηµν ) 0 ν
iA = −(−ig) v̄(q )γ v(q) ū(p )γ u(p) (8.39)
(q − q 0 )2
p
= −ie(p + q)µ , = +2ie2 ηµν
(8.40)
The momentum dependence comes from the ∂ µ term and the factor
of 2 is because of the identical particles Aµ Aµ appearing in the
Lagrangian Similar to the cancellation of the 1/4!
We can see that the difference in sign comes from the A0 piece in the φ4 vertex.
(p + p0 )µ (q + q 0 )ν (2m)2
= −iηµν (−ie)2 0
→ −i(−ie)2
(p − p) 2 −(p − p0 )2
If you found this interesting you should continue and attend the
QFT course in the next semester. You have yet to encounter many
of the truly awesome and deep insights that quantum field theory
contains. It is the language in which the laws of Nature are written.
As Sidney Coleman said: ”Not only God knows, I know, and by the
end of the semester, you will know.”