Lecture notes Particle Physics II

Quantum Chromo Dynamics

2. SU(2) and SU(3) Symmetry

Michiel Botje
Nikhef, Science Park, Amsterdam

November 12, 2013

 Symmetry in (particle) physics

• If the Lagrangian of the world would be fully known we could derive

the equations of motion from it, and the symmetries of nature and
the conservation laws would automatically follow.
• For instance the Maxwell Lagrangian yields, via the Maxwell equa-
tions, all the symmetries and conservation laws of electrodynamics.
• In subatomic physics the Lagrangians are not so obvious, and sym-
metry considerations provide essential clues to construct them.
• It can be shown that an invariance of the Lagrangian under a
symmetry operation leads to a conserved quantity (Noether’s the-
orem). Thus, if a symmetry is found, the hunt is open for the
related conservation law, and if a conservation law is found, the
hunt is open for the related symmetry. For instance we know that
electric charge is conserved in all reactions of elementary particles,
but what symmetry is responsible for this charge conservation?
(The answer will be given in the next lecture.)
• As will become clear later, it turns out that discrete symmetries
lead to multiplicative conserved quantum numbers (e.g. reflec-
tion symmetry ! parity conservation ! multiplication of parities)
while continuous symmetries lead to additive conserved quantum
numbers (e.g. rotation invariance ! angular momentum conser-
vation ! addition of angular momentum quantum numbers).
• We will now use some elementary non-relativistic quantum me-
chanics to establish the relation between symmetries and constants
of motion.

2–3
 When is an observable conserved?

• The expectation value of a quantum mechanical operator F is

hF i ⌘ h |F | i with Hermitian conjugate hF i⇤ ⌘ h |F †| i

• The expectation value of an observable is a real number so that

the operator of an observable should be Hermitian
F = F† if hF i is observable

• Because energy is an observable the Hamiltonian H is Hermitian.

We have for the Schrödinger equation and its Hermitian conjugate
@| i @h |
i = H| i and i = h |H † = h |H
@t @t
• This immediately leads to
@hF i
= i h |HF F H| i = 0 , HF FH = 0
@t

An observable constant of motion F is

Hermitian and commutes with the Hamiltonian

• When H is known, we can find observable constants of motion by

searching for Hermitian operators that commute with H.
• However, when H is not fully known, it is sufficient to establish
(or postulate) the invariance of H, or the Lagrangian, under a
symmetry operation, as we will now show.

2–4
 Symmetry operators

• A transformation operator U transforms one wave function

into another
| 0i = U | i
• Wave functions are always normalized so that we must have
h 0| 0i = h |U †U | i = 1

• It follows that the transformation operator must be unitary

U †U = U U † = I

• We call U a symmetry operator when | 0i obeys the same

Schrödinger equation as | i. Then, with U time independent,
@U | i @| i 1
i = HU | i ! i =U HU | i = H| i
@t @t Iwant

and thus
1
U HU = H or [H, U ] = 0

A symmetry operator U is unitary and

commutes with the Hamiltonian

• Thus U commutes with the Hamiltonian, as does a constant of

motion. However, we cannot identify U with an observable since
it is unitary, and not necessarily Hermitian.

2–5
 Discrete symmetries

• There is a class of unitary transformations with the property

U2 = I
Multiplying from the right with U † and using U U † = I we find
that U = U †: the operator is both unitary and Hermitian.
• Thus if U is a symmetry of (commutes with) the Hamiltonian we
can directly conclude that it is an observable constant of motion.
• Examples of this the are the charge conjugation operator C (ex-
change of particles and antiparticles) and the parity operator P
(reflection of the spatial coordinates).12
• Remark: C and P are not the only operators that are both unitary
and Hermitian. This is, for instance, also true for the Pauli spin
matrices, as is straight-forward to check.
✓ ◆ ✓ ◆ ✓ ◆
0 1 0 i 1 0
x = , y = , z =
1 0 i 0 0 1

• If | i is an eigenvector of both U1 and U2 then

U1,2| i = 1,2 | i and U1U2| i = U2U1| i = 1 2| i
The quantum numbers of a discrete symmetry are multiplicative.
• In these lectures we are not so much interested in discrete trans-
formations (like C, P , T ) but, instead, in continuous transfor-
mations. These transformations are unitary (by definition), but
not necessarily Hermitian. But the generator of a unitary con-
tinuous transformation is Hermitian, as we will see.
12
The time reversal operator T also has T 2 = I but it is antiunitary, and not unitary.

2–6
 Continuous transformations

• There is a large class of continuous transformations that

depend on one or more continuous parameters, say ↵
| 0i = U (↵)| i
An example is the transformation induced by a rotation over an
angle ↵ of the coordinate system (passive rotation), or of the wave
function (active rotation).
• Such transformations have the property that they can be written
as a succession of infinitesimal deviations from the identity
✓ ◆n
i↵
U (↵) = lim I + F = exp(i↵F )
n!1 n
The factor ‘i’ is a matter of definition but important (see below).
In the above, F is called the generator of U .13
• Now if U is unitary we have, to first order in ↵,
U †U = (I i↵F †) (I + i↵F ) = I + i↵(F F †) = I
so that F = F †. In other words,

The generator of a unitary operator is Hermitian

• Now we also understand the factor ‘i’ in the definition of a genera-

tor: without it the generator G ⌘ iF of a unitary operator would
not be Hermitian but anti-Hermitian:
G= G†
13 1 2
Exponentiation of an operator F should be interpreted as exp(i↵F ) = I + i↵F + 2! (i↵F ) + · · · But watch
out, the familiar relation eA eB = eA+B is only true when A and B commute.

2–7
 Generators as conserved observables

• We have seen that a symmetry operator U commutes with the

Hamiltonian so it remains to show that its generator will then also
commute with H. The proof is very simple:
• First, if U (↵) is a symmetry operator then the infinitesimal trans-
formation U (✏) will also be a symmetry operator. Expanding to
the first order in ✏ obtains
.
[H, U ] = [H, I + i✏F ] = [H, I] +i✏ [H, F ] = 0 ! [H, F ] = 0
| {z }
0

If U is a unitary operator that commutes with the

Hamiltonian then its generator F is a Hermitian
operator that also commutes with the Hamiltonian

• We now have the work plan to find the relation between a contin-
uous symmetry of H and the corresponding conserved observable:
1. Find the generator F of the symmetry transformation U .
2. The expectation value of F is a constant of motion
• Clearly a multiplication of continuous symmetry operators corre-
sponds to the addition of their generators in the exponent. The
conserved quantum numbers, which are related to F and not to U ,
are therefore additive.
• We will now proceed with the introduction of some concepts of
group theory which is the mathematical framework to system-
atically describe and classify symmetry operations.

2–8
 Exercise 2.1:

Show that (consult a quantum mechanics book if necessary)

(a) [0.5] Invariance for translations in space leads to the conservation

of momentum.
(b) [0.5] Invariance for translations in time leads to the conservation
of energy.
(c) [0.5] Rotational invariance leads to the conservation of angular
momentum.

2–9
 Group theory

• It is clear that a combination of two symmetry operations—that

each leaves the system unchanged—is again a symmetry operation.
And there is of course the trivial symmetry operation, namely, ‘do
nothing’. Furthermore, we can assume that each symmetry oper-
ation can be undone. We say, in fact, that symmetry operations
form a group.
• What is a group? It is a set of elements {gi},
– with a composition law gi · gj = gk
– that is associative (gi · gj ) · gk = gi · (gj · gk )
– with a unit element e such that e · gi = gi · e = gi
1
– and with an inverse gi such that gi · gi 1 = gi 1 · gi = e
• Examples:
The set {1, i, 1, i} under multiplication (discrete, 4 elements)
The set of integers under addition (discrete, infinite # elements)
Rotations in 3 dimensions (continuous, 3 parameters)
Lorentz transformations (continuous, 6 parameters: which ones?)
• A group is called Abelian when the group operation is commu-
tative gi · gj = gj · gi (e.g. 2-dim rotations). Non-commutative
groups are called non-Abelian (e.g. 3-dim rotations).
• A systematic study of symmetries is provided by a branch of math-
ematics called group theory. We will not present group theory
in these lectures, but only a few basic concepts.14

14
A nice summary of group theory can be found in A&H-II, Appendix M.

2–10
 Representation of a group

• In these lectures, we will be concerned with groups of matrices.

• It may be the case, of course, that the group ‘is’ a set of matrices.
For instance, the group SO(2) of orthogonal 2 ⇥ 2 matrices with
determinant 1, that describe 2-dimensional rotations.
• But a matrix representation may also come from mapping each
element gi of some group to an n ⇥ n matrix Mi (why must M be
square?), such that the multiplication structure is preserved
g1 · g2 = g3 ! M1 M2 = M3
This is called an n-dimensional representation of the group {g}.
Thus, SO(2) is defined by 2⇥2 matrices, (the fundamental rep-
resentation) but it has also representations in higher dimensions.
• Two groups with the same multiplication structure are said to be
isomorphic (⇠ =) if the elements map one-to-one. If the mapping
is not one-to-one, they are called homomorphic (⇠).
• Exercise 2.2: [0.5] Show that
⇢✓ ◆ ✓ ◆ ✓ ◆ ✓ ◆
1 0 0 1 1 0 0 1
{1, i, 1, i} ⇠
= , , ,
0 1 1 0 0 1 1 0
• From an n and an m-dimensional representation we can always
construct an (n + m)-dimensional representation through
" #
(n)
(n+m) Mi 0
Mi = (m) ⌘n m
0 Mi
but this does not classify as a new representation. The relevant
representations are the so-called irreducible ones which cannot
be decomposed in block diagonal form. It is a (non-trivial) task of
group theory to find all the irreducible representations of a group.
2–11
 Lie groups

• On page 2–10 we have encountered discrete groups (elements la-

belled by an index, or a set of indices) and continuous groups
where the elements are labelled by a set of continuous parameters
↵ = (↵1, ↵2, . . . , ↵m). Important groups of transformations U (↵)
are those which can be written as a succession of infinitesimal de-
viations from the identity transformation (see also page 2–7):
U (↵) = lim [ 1 + i(↵/n) · T ]n = exp(i↵ · T )
n!1

Such a group is called a Lie group,15 and the matrices T are

called the generators of the group.16 The number of generators
is equal to the number of parameters that label the group elements.
Example: Rotations are a Lie group but reflections are not since
these are not continuously connected to the identity.
• There is a theorem which states that the commutator of two gen-
erators is always a linear combination of the generators
[Ti, Tj ] = fijk Tk (summation over k implied)
These commutation relations are called the algebra, and the
(complex) numbers fijk are called the structure constants of
the group. It can be shown that these structure constants fully
characterise the multiplication structure of a Lie group.
• On page 2–7 we have shown that if U is unitary then Ti = Ti†. In
other words, the generators of a unitary operator are Hermitian.

15
The formal definition of a Lie group states first of all that the number of parameters is finite, and furthermore
that U (↵1 ) · U (↵2 ) = U (↵3 ), with ↵3 an analytic function of ↵1 and ↵2 .
16
Discrete groups also have generators: e.g. repeated rotation over 2⇡/n generates the cyclic group Zn .

2–12
 The 2-state nucleon system

• After the discovery of the neutron by Chadwick in 1932, the near

equality of its mass (939.5 MeV) to that of the proton (938.3 MeV)
suggested to Heisenberg that, as far as the strong interactions are
concerned, these are two nearly degenerate states of one particle:
the nucleon.
• This ‘isospin symmetry’ of the strong force is further supported
by, for instance, the observation of very similar energy levels in
mirror nuclei (the number of protons in one, is equal to number
of neutrons in the other, and vice versa, like in 137N and 136C).
• In addition, apart from the p-n doublet, there are other particles
that are nearly degenerate in mass, like the pion triplet (⇠140 MeV)
and the quadruplet of resonances (⇠1.23 GeV) ! Fig. This
looks like the doublet, triplet and quadruplet structure of spin- 12 ,
spin-1 and spin- 32 systems built from spin- 12 states, and is thus
strongly suggestive of hadronic substructure.
• We know today that hadrons are built up from quarks and we can
explain isospin symmetry from the fact that the strong interaction
is insensitive to the quark flavour. The mass di↵erences within the
nucleon, ⇡ and multiplets are, after electromagnetic correction,
believed to be due to the di↵erence in the u and d quark masses.
• The invariance for p to n transitions obeys the mathematics of
ordinary spin, hence the term ‘isospin’. The reason is that tran-
sitions in any 2-state quantum mechanical system are described
by the special unitary group SU(2), as will become clear next.

2–13
2–14
 Isospin symmetry

• We work in a 2-dim Hilbert space spanned by the basis vectors17

✓ ◆ ✓ ◆
1 0
|pi = and |ni =
0 1
The Hermitian conjugates are hp| = (1, 0) and hn| = (0, 1). An
arbitrary state is written as the linear combination
| i = ↵ |pi + |ni
Because |↵|2 is the probability to find the system in a |pi state and
| |2 the same for the |ni state we must have, for any state | i,
h | i = |↵|2 + | |2 = 1

• We have seen already that a transformation | 0i = U | i must

preserve the norm so that U must be unitary: U †U = 1.
• Taking determinants we find
det(U †U ) = det(U †) det(U ) = det(U )⇤ det(U ) = 1
Therefore det(U ) = ei with some arbitrary phase factor.
• So we may set U = ei V with det(V ) = 1. Invariance for phase
shifts is called a U(1) invariance and leads to charge conservation,
as we will see later. The charge conserved in the p-n case here is
not electrical charge, but baryon number
A = (Np Np̄) + (Nn Nn̄)

• Putting U(1) invariance aside, we have to deal with unitary 2 ⇥ 2

matrices V with unit determinant, that is, with the group SU(2).
17
When we talk about quarks we will use the notation |ui and |di instead.

2–15
 The group SU(2)

• The mathematics of SU(2) is well known from the treatment of

ordinary spin in quantum mechanics. A transformation can be
written as U = exp(i↵ · I) with the three generators I ⌘ ⌧ /2
given by the Pauli matrices
✓ ◆ ✓ ◆ ✓ ◆
0 1 0 i 1 0
⌧1 = , ⌧2 = , ⌧3 =
1 0 i 0 0 1
These generators are clearly Hermitian (⌧i† = ⌧i), as they should
be, since U is unitary. It can be shown (Exercise 2.3) that, quite
in general, det[exp(A)] = exp[Tr(A)] so that the traces of the ⌧i
vanish because the SU(2) transformations have unit determinant.

The generators of a unitary matrix group with unit

determinant are Hermitian and traceless

• By matrix multiplication you may check the commutation relations

[Ii, Ij ] = i ✏ijk Ik
with ✏ijk the antisymmetric tensor (+1 for cyclic permutations of
123 and 1 for cyclic permutations of 213, zero otherwise).
• SU(2) has one so-called Casimir operator that commutes with
all the generators, and is always some non-linear function of the
generators. For SU(2) this is the total isospin operator:
I 2 = I12 + I22 + I32
A state can then be a simultaneous eigenstate18 of I 2 with eigen-
value i(i + 1), i = 12 , 1, 32 , . . . and of I3 with eigenvalue m =
i, . . . , +i. The eigenvalues label the state, like | i = |i, mi.
18
A Hermitian matrix has the property that it can always be diagonalised by a unitary transformation.
Hermitian matrices can be simultaneously diagonalised by a single transformation if they commute.

2–16
 Exercise 2.3:

In this exercise we will review a few easy-to-prove properties of matrices

and of matrix transforms (also called similarity transforms)
defined by
A0 = SAS 1,
where S is a non-singular transformation matrix. Such transforms
can come in very handy in a calculation because they allow you to
transform matrices to convenient forms, such as a transformation to
diagonal form which is used for the proof in (e) below.

(a) [0.1] Show that Tr(AB) = Tr(BA).

(b) [0.2] Show that a matrix transform preserves the algebra of a Lie
group. Representations that are related by similarity transforma-
tions are therefore called equivalent.
(c) [0.2] Show that a matrix transform preserves the product, deter-
minant and trace, that is,
(AB)0 = A0B 0, det(A0) = det(A) and Tr(A0) = Tr(A).
?
What about Hermitian conjugation: (A0)† = (A†)0.
(d) [0.2] Show that a matrix transform preserves the terms in a power
series, that is,
(An)0 = (A0)n ! (exp A)0 = exp(A0).

(e) [0.3] Now show that

det[exp(A)] = exp[Tr(A)]
for all matrices A that can be brought into diagonal form.

2–17
 Exercise 2.4:

(a) [0.5] Show that ⌧i⌧j = ij + i"ijk ⌧k . Together with the fact that
the ⌧ are Hermitian, we thus have ⌧i† = ⌧i = ⌧i 1.
(b) [0.5] Now show that (a · ⌧ )(b · ⌧ ) = a · b + i⌧ · (a ⇥ b) and, from
this, that (✓ · ⌧ )2 = |✓|2.
(c) [0.5] Use the above, and the Taylor expansions of exp(), sin() and
cos(), to show that exp(i✓ · ⌧ ) = cos |✓| + i(✓ˆ · ⌧ ) sin |✓|. Here ✓ˆ
is the unit vector along ✓.
(d) [0.25] Instead of |pi and |ni we will write |ui and |di to reflect
isospin symmetry on the quark level. Verify that
I3 |ui = 12 |ui, I3 |di = 1
2 |di

and that the Casimir operator I 2 = I12 + I22 + I32 is a multiple of

the unit operator, with
I 2 |ui = 34 |ui, I 2 |di = 34 |di

(e) [0.25] Define the step operators I± = I1 ± i I2 and verify that

I+|ui = 0, I+|di = |ui, I |ui = |di, I |di = 0

We can now draw a, kind of trivial, weight diagram like

I+
d ! u
y y - I3
1
2 + 12
I

2–18
 Composite states

• The rules for addition of angular momenta from quantum mechan-

ics carry straight over to the addition of isospins. We will not derive
here the mathematics but will only indicate how it works.
• Addition of two states |i1, m1i and |i2, m2i, results in (2i1 + 1) ⇥
(2i2 + 1) di↵erent states which can be classified according to the
eigenvalue label i of the Casimir operator I 2 which ranges from
|i1 i2| to i1 +i2, and the eigenvalues m of the I3 operator that, for
each state i, range from i to +i. Here m = m1 + m2. Formally,
the combined state can be written as
X
|i, mi = hi1, i2, m1, m2|i, mi|i1, m1i|i2, m2i
The Clebsch-Gordan coefficients h·|·i can be found in the Particle
Data Book ! Fig. For a nucleon-nucleon system we get
p
|I, I3i = |0, 0i = (pn np)/ 2
= |1, 1i = pp p
= |1, 0i = (pn + np)/ 2
= |1, 1i = nn
• Exercise 2.5: [1.0] Use exchange symmetry arguments or the
(1) (2)
step operators I± ⌘ I± +I± to justify the decomposition above.19
Hint: See H&M Exercise 2.1.
• This splitting of the combination of two 2-component states into
a singlet and a triplet state is often written as 2 ⌦ 2 = 1 3.
The significance of such a decomposition is that under a SU(2)
transformation the substates of the 1 and 3 representation will
transform among themselves.
19
p
In full, the step operator is defined by I± |i, mi = i(i + 1) m(m ± 1) |i, m ± 1i.

2–19
 !

Clebsch-Gordan coefficients from the Particle Data Book. Given in the tables is the square of
the coefficients, so you should take the square root.

2–20
 SU(2)f for antiquarks

• If | i is a particle state then the complex conjugate is identified

with the corresponding antiparticle state:20 | ¯i ⌘ | i⇤. An anti-
quark state therefore transforms in the complex conjugate repre-
sentation of SU(2), denoted by 2⇤ or 2̄.
| ¯0i = U ⇤| ¯i = exp( i↵ · ⌧ ⇤/2) | ¯i ⌘ exp(i↵ · ⌧¯ /2) | ¯i
The two representations are thus related by ⌧¯ = ⌧ ⇤.
• To combine a quark with an antiquark we could calculate from
scratch the Clebsch-Gordan coefficients of 2 ⌦ 2̄ but we can save
us the e↵ort by using a trick that, by the way, only works for SU(2).
• Just replace ū by d̄ and d̄ by ū in | ¯i, that is, define
✓ ◆✓ ◆ ✓ ◆
0 1 ū d̄
| ˜i ⌘ C| ¯i = =
1 0 d̄ ū
It is now straight-forward to show (Exercise 2.7) that | ˜i trans-
forms as a quark state | ˜0i = U | ˜i so that we just can use the
Clebsch-Gordans of the 2 representation.21
• Exercise 2.6: [ ⇥ ] Take the |qqi states given on page 2–19 (sub-
stitute u for p and d for n), to arrive at |q q̄i meson states that
properly transform under SU(2):
p
! = |0, 0i = (uū + dd̄)/ 2
⇡ + = |1, 1i = ud̄ p
⇡ 0 = |1, 0i = (uū dd̄)/ 2
⇡ = |1, 1i = dū
20
We use here | ¯i to indicate an antiparticle; please do not confuse it with a conjugate Dirac spinor .
21
In fact, for SU(2) the generators ⌧¯i and ⌧i are related by the similarity transformation ⌧¯i = C 1 ⌧i C so that
they are equivalent, that is, they are not regarded as di↵erent representations, see also Exercise 2.3.

2–21
 Exercise 2.7:

(a) [1.0] Use isospin invariance to show that the ratio

(pp ! ⇡ +d)
=2
(pn ! ⇡ 0d)
Here the deuteron has isospin I = 0 and the pion isospin I = 1.
You may assume that the cross section is
2
X 2
X 2
0 0
⇠ |amplitude| = |hI , I3|A|I, I3i| = A 2
|hI 0, I30 |I, I3i| .
I I

Hint: See H&M Exercise 2.3.

(b) [0.2] Show that the generators ⌧¯ are a representation of SU(2).

1
(c) [ ⇥ ] Verify that I3(ū) = 2 and I3(d̄) = + 12 .
(d) [0.3] Show that
✓ ◆✓ ◆
0 1 ū
| ˜i = C| ¯i =
1 0 d̄
transforms as a particle state.

2–22
 The group SU(3)f I

• To accommodate strange quarks, our space has to be extended

0 1
✓ ◆ u
u @d A
from to
d
s

• Like in the (iso)spin case we can write a unitary transformation as

| 0i = U | i = exp(ia · /2) | i ⌘ exp(ia · T ) | i
but the generators are now Hermitian 3⇥3 matrices. A complex
3 ⇥ 3 matrix is characterised by 18 numbers but only 8 are inde-
pendent because the matrices are Hermitian, and traceless since
det U = 1. Thus there are 8 independent generators.
• The 8 Gell-Mann matrices (with Pauli matrices inside!) are
0 1 0 1 0 1 0 1
0 1 0 0 i 0 1 0 0 0 0 1
@1 0 0 A @ i 0 0 A @0 1 0 A @0 0 0A
0 0 0 0 0 0 0 0 0 1 0 0
| {z } | {z } | {z } | {z }
1 2 3 4
0 1 0 1 0 1 0 1
0 0 i 0 0 0 0 0 0 1 0 0
@0 0 0 A @0 0 1A @0 0 iA p1 @0 1 0A
3
i 0 0 0 1 0 0 i 0 0 0 2
| {z } | {z } | {z } | {z }
5 6 7 8

• The algebra of the SU(3) group is given by the commutation rela-

c
tion of the matrices Ta = a/2: [Ta, Tb] = ifab Tc

2–23
 The group SU(3)f II

c
• The structure constants fab are antisymmetric in the exchange of
two indices (see Exercise 2.8); the non-zero ones are
3
f12 =1
7 5 6 7 5 6 1
f14 = f16 = fp24 = f25 = f34 = f37 = 2
8
f45 8
= f67 = 12 3
• It is seen that 3 and 8 are simultaneously diagonal so that
we can label quark states by the simultaneous eigenvalues of the
isospin operator
p T3 p= 3/2 and the hypercharge operator
Y = 2T8/ 3 = 8/ 3. This gives rise to following weight di-
agram for the quark states (see Exercise 2.8 for antiquarks):

Y
6
1

y y
2( 1 ±i 2)

+ 13 d u
J ⌦
J ⌦
J ⌦
1
J ⌦ 1
2( 6 ±i 7) J ⌦ 2( 4 ±i 5)
J ⌦
2 y
JJ⌦⌦
3
s
- I3
1
2 0 + 12

• As mentioned on page 2–16 there is one Casimir operator for SU(2),

but there are two Casimirs for SU(3). By definition, these commute
P 2
with all the i. One of them is is the total ‘isospin’ operator i
while the other is a rather complicated trilinear function of the i
which can be found in A&H-II, Appendix M.5.

2–24
 Exercise 2.8:

(a) [0.5] The matrices are normalised such that Tr( a b) = 2 ab .

Check this for a few matrices a and b.
c
(b) [0.5] Use Tr(AB) = Tr(BA) to show that Tr( c[ a, b]) = 4ifab
and, by changing the order of the , that the structure constants
c
fab are antisymmetric in the exchange of two indices.
(c) [0.5] Plot the eigenvalues of the isospin and hypercharge operator
for the u, d and s quarks in an I3-Y diagram. Check the Gell-Mann
Nishijima formula Q = I3 + 21 Y and also that Y = S + B. Repeat
the exercise for antiquarks in the 3̄ representation.
(d) [0.5] Write down the matrices for the step operators 12 ( 1 ± i 2),
1 1
2 ( 4 ± i 5 ) and 2 ( 6 ± i 7 ) and justify their position in the weight
diagram on page 2–24.

2–25
 Exercise 2.9: The adjoint representation of SU(3)

• We have encountered the algebra of the groups SU(2) and SU(3)

in terms of the two-dimensional Pauli matrices and the three-
dimensional Gell-Mann matrices, respectively. These matrices are,
together with the 2- or 3-dim vectors on which they act, called the
fundamental representation of SU(2) or SU(3).
• However, the structure constants of a Lie group automatically gen-
erate a representation with a dimension that is equal to the number
of generators, e.g. 8⇥8 for SU(3). This is called the adjoint rep-
resentation. Below we let you find out how this works.

(a) [ ⇥ ] Verify the Jacobi identity for matrices A, B and C:

[[A, B], C] + [[B, C], A] + [[C, A], B] = 0

(b) [ ⇥ ] Now show that in terms of the SU(3) structure constants the
Jacobi identity reads
fijm fmk
n m n
+ fjk m n
fmi + fki fmj = 0

(c) [ ⇥ ] Verify that fijk = fjik

(d) [1.0] Define the 8 ⇥ 8 matrices Ci with elements
(Ci)kj = fijk
and show that the Ci obey the SU(3) algebra
[Ci, Cj ] = fijk Ck

In this way, we have constructed the adjoint representation of SU(3)

from its structure constants. We will see later that coloured quarks are
described by the fundamental representation of SU(3), of dimension 3,
and gluons by the adjoint representation, of dimension 8.
2–26
 The Eightfold Way

• Because our interest in SU(3) lies in the fact that it is an exact

(colour) symmetry of QCD, we will not present here how SU(3)f is
used to classify the hadrons (the Eightfold Way). This is treated
in great detail in H&M Chapter 2, and also in Griffiths Chapter 5.
• We just mention that the mesons |q q̄i can be grouped into octets
and singlets (3 ⌦ 3̄ = 8 1) and baryons |qqqi can be grouped
into decuplets, octets and singlets (3 ⌦ 3 ⌦ 3 = 10 8 8 1).
• Nevertheless, let us have a look at the spin 3/2 baryon decuplet,
because it provides us with an argument to introduce the colour
quantum number.

2–27
 The need for a colour quantum number

S
6
u u u u
0 + ++

0
ddd udd uud uuu

u u u
⌃⇤ ⌃⇤0 ⌃⇤+
1
dds uds uus

u u
⌅⇤ ⌅⇤0
2
dss uss

u
⌦
3
sss
- I3
3 1
2 2 + 12 + 32

• In this spin 3/2 baryon decuplet, the flavour wave functions at the
corners are obviously symmetric under the exchange of two quarks.
Although this is not apparent from the labels, all wave functions of
the decuplet are symmetric, as you will discover in Exercise 2.10.
• But now we have a problem: the total wave function
= space (L = 0) ⇥ spin (""") ⇥ flavour (q1 q2 q3 )

is symmetric under the exchange of two quarks, while it should be

anti-symmetric, since baryons are fermions (half-integer spin).
• The solution is to assign a ‘colour’ quantum number (r, g, b) to each
quark so that the quarks can be distinguished by their colour, pro-
vided, of course, that we do not allow two quarks in a baryon to
have the same colour. Thus the three colours are always present
and we say that baryons are ‘white’, or colour singlets (= invari-
ant under SU(3)c transformations). By anti-symmetrising the wave
function in colour space, over-all anti-symmetry is established.
2–28
 Exercise 2.10:

(a) [0.5] Use the step operators defined in the weight diagram on
page 2–24 (and also in Exercise 2.8d) to generate all quark states of
the baryon decuplet, starting from one of the corner states (ddd),
(uuu) or (sss). You will not obtain the correct normalisation in
this way, but that is not so important here (you can always nor-
malise the wave functions afterwards, if you wish). The point of
this exercise is to note that all wave functions that you obtain by
stepping through the diagram are symmetric in the exchange of
two quarks.
(b) [0.5] Construct a wave function colour(c1, c2, c3) that is fully anti-
symmetric in the exchange of two colours.

2–29
 Experimental evidence for colour I

e µ e q

e+ (a) µ+ e+ (b) q̄

• The cross section for the left diagram is given in PP-I section 8.3:

+ + 4⇡↵2
(e e ! µ µ ) =
3s
Here particle masses are neglected and if we do the same for the
right diagram, we obtain the cross section for q q̄ production simply
by putting the correct charge at the q q̄ vertex

+ 4⇡↵2e2i
(e e ! qi q̄i) =
3s
• Because quarks fragment with 100% probability into hadrons, we
can sum over all available quark species to get the observable
X 4⇡↵2e2
+ i
(e e ! hadrons) = Nc
i
3s

• Here the sum runs over all quark flavours that can be produced at a
p
given energy s, and Nc counts the number of coloured duplicates
of each quark. Thus Nc = 3 for the quark colours qr, qg and qb.

2–30
 Experimental evidence for colour II

p
• This plot shows, as a function of
s, measurements of the ratio
(e+e ! hadrons) X X
2 2
R= = N c e i = 3 e i
(e+e ! µ+µ ) i i

• The data are consistent with Nc = 3 and certainly exclude Nc = 1.

• Remark: There is quite some structure in this plot, in particular
around the thresholds of heavy quark production where q q̄ pairs
are produced with little relative momentum so that they can form
bound states, like the J/ family (cc̄) at about 3 GeV, and the ⌥
family (bb̄) at about 10 GeV.

2–31