PHY 422A : Mathematical Methods II

∗
Lecturer: Das
Last updated: 2020/02/17, 10:33:13

1 Group Theory Basics 1

2 Prelude to Lie groups 7

3 Bits of Representation theory 9

3.1 Some Great Theorems . . . . . . . . . . . . . . . . . . . . . . 11
3.1.1 Consequences of the orthogonality theorem . . . . . . . 12

4 Young tableaux for Sn 14

5 Group theory in Quantum Mechanics 15

6 Lie Groups 16
6.1 Geometry of SU (2) . . . . . . . . . . . . . . . . . . . . . . . . 18
6.2 Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.3 Certain aspects of SO(3) . . . . . . . . . . . . . . . . . . . . . 22
6.4 SU (2) and SO(3) . . . . . . . . . . . . . . . . . . . . . . . . . 26

7 Weights and Roots 28

7.1 SU(3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

8 Young Tableaus for SU (N ) 38

∗
Please send corrections to didas at iitk.ac.in

1
1 Group Theory Basics
Introduction
Symmetries play a very important role in dictating physics. They give strong
constraints on the observables of our theory (starting from the point of de-
ciding what are the good observables? For instance in a gauge theory mean-
ingful quantities are gauge invariant objects. Symmetries can be analyzed
nicely in the path integral formalism and give rise to Ward identities which
are constraints respected by the observables. However sometimes symmetries
go beyond the Lagrangian framework and can allow computation of observ-
ables. This is the case for instance in a conformal field theory.
Symmetries have a group structure. Once we understand symmetries via the
theory of groups, we can compute the physics unavailable to us via pertur-
bation. The group transformations can be represented as linear operations
in vector spaces, and this leads naturally to finding representations of the
group. Much of group theory, concerns classifying all possible representa-
tions..

Axioms
A group G, is a set with binary operation that assigns to every ordered pair,
(g1 , g2 ) of its elements, a third element g3 usually written as g3 = g1 g2 . This
binary operation of product follow the following rules,

1. Associativity : g1 (g2 g3 ) = (g1 g2 )g3 .

2. Existence of identity, e ∈ G s.t, eg = g ∀g ∈ G.

3. Existence of inverse, g −1 s.t., g −1 g = e.

There are many corollaries that follow from the above three axioms,

1. gg −1 = e, prove starting from g −1 g = e and use associativity.

2. ge = g, use eg = g, the above, and associativity.

3. e is unique.

4. Inverse is unique.

1
Two elements are said to commute if, g1 g2 = g2 g1 . If this is true ∀gi ∈ G
then the group is abelian, else it is non − abelian 1 . If G contains finite
number of elements, order of G, |G|, then it is called a f inite group.

Some examples
1. The integers Z under addition. – Infinite What is the identity?

2. The set of functions,

1 z−1
f1 (z) = z f2 (z) = f3 (z) = ,
1−z z
1 z
f4 (z) = f5 (z) = 1 − z f6 (z) = .
z z−1
The product rule is,
(fi , fj ) 7→ fi ◦ fj .

3. Integers modulo n under addition. – Finite.

4. Set of rotations in three dimensions, or set of 3-by-3 real matrices, O

with, OOT = I and det O = 1; this is SO(3). This is an example of a
Lie group.

5. Groups can be specified by list of generators and relations. Cyclic

group of order n, Cn is given by generator a and relation, an = e.

6. The group Zn is a finite group of order n. This is a cyclic group and

the elements can be represented as roots of unity. e.g., Z3 = {1, ω, ω 2 }.

7. Lorentz group generated by the transformation :

 0    0
ct cosh φ sinh φ ct
= .
x0 sinh φ cosh φ x0

Here φ is the boost angle. One can verify that the composition of two
transformations adds up the boost angles.

[End of Lecture 1]
1
There may be composition rules which are commutative but not associative. e.g., the
arithmetic mean (example suggested by Radhika Prasad)

2
Some basic properties
1. Subgroups : A subset of a group (lets call it H) that forms a group.
{e} and G itself are trivial subgroups, others are proper subgroups.
For a finite group, |G| is divisible by |H|. This is Lagrange0 s theorem.

To prove Langrange’s theorem, one starts by listing out A1 = H =

{h1 , . . . hm }. Then one picks g1 ∈ G and g1 ∈ / H and constructs the
set, A2 = {h1 g1 , h2 g1 , . . . , hm g1 }. If this exhausts G then one stops,
else one picks g2 ∈ G and g2 ∈ / H and g2 ∈ / A2 and constructs,
A3 = {h1 g2 , h2 g2 , . . . hm g2 }. One repeats this until all elements of G
are exhausted. Let the last list be, Ak = {h1 gk , h2 gk , . . . , hm gk }. Then
one can show that all such lists are distinct. This can be proven by
contradiction; let ha gj = hb gl , then gj = h−1 a hb gl . But this implies that
gj ∈ Al . However, this is not the case by construction. Thus all are
distinct. Now since k is an integer, we immediately have,

n = mk. (proved)

2. Direct Product : Given two groups F and G, we may arrive at a

group, F ⊗ G. The way to list elements of this is (f, g), where f ∈ F
and g ∈ G. The composition is via, (f, g) · (f 0 , g 0 ) = (f f 0 , gg 0 ). The
identity element is (ef , eg ).

An example is Z2 ⊗ Z2 = {(1, 1), (1, −1), (−1, 1), (−1, −1)}. Note that
this is different from the group, Z4 .

3. Multiplication Table : An useful way to list down the group properties

is by filling up the multiplication table :

e g1 ... gi ... gn
e e g1 ... gi ... gn
g1 g1 g12 ... g1 gi ... g1 gn
... ... ... ... ... ... ...
gj gj ... ... gj gi ... gj gn
... ... ... ... ... ... ...
gn gn ... ... gn gi ... gn2

3
 The multiplication table follows the once and only once theorem. Ev-
ery group element appears only once if we trace a particular row or a
coloumn. This can be proven once again by contradiction. If we as-
sume gj gk = gj gl , then this gives the contradiction gk = gl . This is a
simple yet a very useful theorem. This rule immediately tells us,
X X X
f (g) = f (gg 0 ) = f (g 0 g).
g g0 g0

This is known as the rearrangement lemma.

4. Homomorphism A map f : G → G0 which preserves the multiplicative

structure of the group. i.e., acting on the group elements, g, g 0 we have,
f (gg 0 ) = f (g)f (g 0 ).

A homomorphism becomes an isomorphism if the map is one-to-one

and onto. [End of Lecture 2]

5. N ormal/Invariant subgroup: H is a normal subgroup, if g −1 Hg = H.

There is scrambling going on.

6. Conjugacy : Two group elements g1 , g2 are said to be conjugate to

each other if, there is an element g ∈ G s.t., g2 = g −1 g1 g. We write
this as, g1 ∼ g2 . It is an equivalence relation.

Example : In the rotation group SO(3), the conjugacy classes are the
sets of rotations by same angle, but about different axis.

Permutation groups
The action of this group is to permute its elements. Since the once and only
once rule need to be satisfied, one can show easily that every finite group of
order n is some subgroup of the permutation group Sn .

4
The permutation group of n objets, Sn has order n!. If we order G =
{e, g1 , g2 , . . . } and then multiply from left by g ∈ G then the ordered list,
{g, gg1 , gg2 , . . . } is a permutation of G. Thus any group G is a subgroup of
the permutation group S|G| . This is Cayley 0 s theorem.

Permutation group is denoted using cycles. For example an element, π1

in S8 be denoted by,
 
1 2 3 4 5 6 7 8
π1 = ↓ ↓ ↓ ↓ ↓ ↓ ↓ ↓ = (123)(45)(67)(8).
2 3 1 5 4 7 6 8

The rhs defines the cycle notation. Any permutation with the pattern, (∗ ∗
∗)(∗∗)(∗∗)(∗) is in the same conjugacy class as π1 .

Problem
If ϕ : F → G is a group homomorphism, and if we define Ker(ϕ) as the set of
elements in F that map to eG , then show that Ker(ϕ) is a normal subgroup
of F .

Since it is a homomorphism, we also have, ϕ(eF ) = eG . Now let the Ker(ϕ) =

H = {eF , h1 , . . . , hn }. Consider, g · hi · g −1 , where hi ∈ H and g ∈ F , and
lets act on this by the mapping, ϕ. We have:

ϕ(ghi g −1 ) = ϕ(g)ϕ(hi )ϕ(g −1 ) = ϕ(g)eG ϕ(g −1 ) = ϕ(gg −1 ) = ϕ(eF ) = eG .

By definition,
eG = ϕ(hl ).
∴ hl = ghi g −1 (proved).
[End of Lecture 3]

5
 7. Cosets :Given, H ⊆ G with elements H = {h1 , h2 , . . . }, and g ∈ G,
the left coset is gH = {gh1 , gh2 , . . . }. If two cosets intersect, then they
coincide. A group can be written as,

G = g1 H + g2 H + · · ·

8. Quotient group : Given a normal subgroup H, take the set of cosets

G/H ≡ {g1 H, g2 H, . . . }. Define product (gi H)(gj H) by the coset where
this lies, i.e., identify : (gi H)(gj H) = gk H. As a result we now have a
group. This is called the quotient group. The identity element is given
by eG H. The inverse of gH is g −1 H. The order of the quotient group
is |G|/|H|.

Coxeter groups
This group is defined by the relation, g ∈ G, g 2 = 1 and also (gi gj )nij = 1
for nij ≥ 2. One can show that nij = nji . One starts with

gj2 = 1,
gj (gi gj )nij gj = 1,
gj (gi gj )(gi gj ) . . . (gi gj ) gj = 1,
| {z }
nij number of terms

(gj gi )(gj gi ) . . . (gj gi ) gj2 = 1,

| {z }
nij number of terms

(gj gi )nij = 1.

But by definition, (gj gi )nji = 1. Therefore one has, nij = nji .

9. Simple group : A group with no normal subgroups. Example : Cyclic

group, 16 families of Lie groups, 26 sporadic groups including the
M onster group.

6
2 Prelude to Lie groups
In contrast to finite groups, one can have infinite dimensional continuous
groups. Here we focus on the rotation group where the elements denote
rotations in space by angles. In 2 dimensions, the group is SO(2) and acts
via,  0   
x cos φ sin φ x
= . (2.1)
y0 − sin φ cos φ y
The matrix is denoted by R(φ). In addition to length of a vector, rotation also
leaves angles between two vectors invariant. Let us denote the vectors by, ~u, ~v .
Then under the rotation : ~u → ~u0 = R(φ)~u and similarly ~v → ~v 0 = R(φ)~v .
Since, |~u|, |~v |, as well as the angle between ~u and ~v stays invariant; this
implies: ~u · ~v = ~u0 · ~v 0 = (R(φ)~u)T (R(φ)~v ) = ~uT R(φ)T R(φ)~v . Thus the
condition on the transformation is,

R(φ)T R(φ) = 1. (2.2)

This condition is not only satisfied by rotations, but also for instance by re-
flections. Real matrices satisfying this relation are called orthogonal matrices.
We can check this explicitly for the matrices (2.1) by matrix multiplication :
    
cos φ − sin φ cos φ sin φ 1 0
= .
sin φ cos φ − sin φ cos φ 0 1

Next, using det AB = det A det B, and det A = det AT we can conclude that,

det RT R = (det R)2 = det I = 1.

Thus the determinant of the matrix is ±1. Orthogonal matrices also contain
reflections. For instance reflecting the x axis corresponds to the orthogonal
matrix :  
−1 0
0 1
which has determinant −1. We can select out rotations from the set of
orthogonal matrices by demanding determinant to be +1. Matrices with
positive unity determinant are called special.

7
Rotating infinitesimally
A key insight of Lie was that the aspects of rotation group can be captured
by focussing on infinitesimal rotations. Thus we write the small rotation
through the small angle, θ as an expansion:

R(θ) = I + A + O(θ2 ).

Here A is order θ. The equation (2.2) implies that,

A = −AT .

This means that A is antisymmetric. In two dimensions then the only pos-
sibility is,  
0 1
J=
−1 0
Thus near identity rotations look like:

R(θ) = I + θJ + O(θ2 )
 
1 θ
' . (2.3)
−θ 0

To get to the finite transformation, we may apply the above matrix operation
many many times. Let the small angle θ = Nφ where N → ∞. Therefore we
want to compute repeated action of R(θ) and see if after N → ∞ number of
times if we recover R(φ) from (2.1). We have :
 N
N φJ
lim (R(φ/N )) = lim 1 +
N →∞ N →∞ N
∞
X φn J n
= eφJ =
n=0
n!
∞
! ∞
!
X φ2m X φ2m+1
= (−1)m I+ (−1)m J
m=0
(2m)! m=0
(2m + 1)!
= cos φI + sin φJ
 
cos φ sin φ
= = R(φ). (2.4)
− sin φ cos φ

8
In the third equality above we used, J 2 = −1, therefore all even (= 2m)
powers are just (−1)m while odd (= 2m + 1) powers are (−1)m J.
[End of Lecture 4]

SO(3)
For SO(3), the rotations are in 3 dimensions. For 3 × 3, one may choose the
following antisymmetric matrices,
   
0 0 0 0 0 −1
J1 = −i 0 0 1 , J2 = −i 0 0
   0 (2.5)
0 −1 0 1 0 0
 
0 1 0
J3 = −i −1 0 1 .
 (2.6)
0 0 0

Thus analogous to R = eφJ for SO(2) in this case one can write a group
element as,
~ = ei 3i=1 θi Ji .
P
R(θ) (2.7)
Note that here we have chosen the generators to be Hermitian which is why
there is the factor of i in the exponent. The algebra of these generators is :

[Ja , Jb ] = −iabc Jc . (2.8)

3 Bits of Representation theory

The groups that we study here are conveniently represented using matrices.
For every group element g ∈ G, we associate a matrix D(g). In order to be
a valid representation, this mapping of group element into matrices must be
via a homomorphism, i.e.,

D(g1 )D(g2 ) = D(g1 g2 ). (3.9)

9
The identity element is the identity matrix. Inverse are simply matrix in-
verses. We already encountered some matrix representations with these prop-
erty in (2.4) and (2.7). For finite groups, the multiplication table provides
automatically a representation called the regular representation. In order to
write down the matrices from the multiplication table in this representation,
first order the group elements as gi for i = 1, 2, . . . , N (G).2 and represent ev-
ery gi as a column vector with all zeros except a 1 at the ith entry. Therefore
the matrix, D(gk ) in regular representation is just read from the character
table, as D(gk )ij = 1 if the gk -th row and gi -th column in the multiplication
table gives gj else it is zero. Thus the regular representation is a N (G) di-
mensional square matrix. The trivial representation is furnished by just the
number 1, i.e., D(g) = 1∀g ∈ G.
A given group may have many representations. However only some of them
are irreducible. An irreducible representation (written as irrep for short) is
a representation which does not have any invariant subspace.
A reducible representation may be obtained by stacking up irreps (Di (g)) in
block diagonal form,
 1 
D (g) . . . ...
D(g) =  . . . D2 (g) ... .
... ... D3 (g)

In this way, one can easily construct higher dimensional representations of a

group. The above example shows the following irrep decomposition :

D(g) = ⊕3i=1 Di (g). (3.10)

Character
Given a matrix representation, D(g), the trace is an important quantity. In
particular it depends on the representation used. The trace defines character
of the representation :
χ(g) = Tr D(g). (3.11)
A part of the utility of the character is inherited from the cyclicity property
of the trace. Suppose we transform our basis, then

D(g) → D0 (g) = SD(g)S −1 .

2
It is customary to choose g1 = e.

10
Under this similarity transformation, non-zero entries of the D(g) matrix
may become zero and vice-versa. These two representations should still be
equivalent to each other, since we just changed the basis. The trace is pre-
served under similarity transformations, since

Tr D0 (g) = Tr SD(g)S −1 = Tr S −1 SD(g) = Tr D(g) = χ(g).

Thus given two non-equivalent representations the traces will turn out to be
different. Another property of the character is that for the group elements
belonging to the same conjugacy class, it is the same. This is because, if g
and g 0 are in the same conjugacy class, then ∃g 000 , such that, g 0 = g 000 gg 000−1 .
Thus,

Tr D(g 0 ) = Tr D(g 000 gg 000−1 ) = Tr D(g 000 )D(g)D(g 000−1 ) (3.12)

000 000 −1 000 −1 000
= Tr D(g )D(g)D(g ) = Tr D(g ) D(g )D(g) = Tr D(g) = χ(g).

Thus if g, g 0 ∈ c where c is a conjugacy class then,

χ(c) = χ(g) = χ(g 0 ). (3.13)

Characters depend on representations and are functions of the conjugacy

class. Note:

χ(e) = dimension of the matrix representation.

[End of Lecture 5]

3.1 Some Great Theorems

Unitarity theorem : For finite groups we can always choose the matrix
representations to be unitary.
Schur’s Lemma : If D(g)A = AD(g), then A = λI.
Great Orthogonality Theorem:
X N (G) rs i k
D†r (g)ij Ds (g)kl = δ δl δj . (3.14)
g
dr

[End of Lecture 6]

11
3.1.1 Consequences of the orthogonality theorem
We start with equation :
X N (G) rs i k
D†r (g)ij Ds (g)kl = δ δl δj . (3.15)
g
dr

Setting j = i and k = l and summing, we obtain

X
χ∗r (g)χs (g) = N (G)δ rs . (3.16)
g

since character is a function just of the conjugacy class, the above can be
simplified further to,
X
n(c)χ∗r (c)χs (c) = N (G)δ rs . (3.17)
c

In the above the sum is over all conjugacy classes c, n(c) is the number of
elements in the conjugacy class c.
Viewing the characters as vectorial functions in the space of conjugacy classes,
we have an orthogonality of vectors indexed by the irrep indices r, s. This is
possible if the space of irreps form a subset of the space of classes.

N (R) ≤ N (C). (3.18)

One can also prove :

X N (G) cc0
χ∗r (c)χr (c0 ) = δ . (3.19)
r
n(c)

Following a similar logic and this time viewing the characters as vectorial
functions in the space of irreducible representations, we have an orthogonality
indexed by conjugacy classes. This implies,

N (R) ≥ N (C). (3.20)

The only way the two inequalities, (3.18), (3.20) can be satisfied is when

N (C) = N (R). (3.21)

12
Thus the number of conjugacy classes also give us the number of irreducible
representations.
Now suppose we are given the character of some matrix representation D
which in principle can be reducible. Let us assume that D contains nr number
of times the irreducible matrix representation Dr . In this case, we can write,
X
χ(c) = nr χr (c). (3.22)
r

Therefore plugging this into, the l.h.s of the equation below we get,
X X X X
n(c)χ∗r (c)χ(c) = n(c)χ∗r (c) ns χs (c) = N (G) ns δ rs ,
c c s c,s
= N (G)nr . (3.23)

Thus we see that the operator,

n(c) ∗r
χ (c).
N (G)

projects any reducible character into the irrep r subspace and gives the times
the irrep r is containedPupon summing over the class.
If we plug (3.22) into c n(c)χ(c)∗ χ(c) we get,
X X X
n(c)χ(c)∗ χ(c) = nr ns χ∗ r(c)χs (c) = N (G) δ rs nr ns
c c,r,s r,s
X
= N (G) n2r . (3.24)
r

For Regular representations the only non-zero character is for the identity
conjugacy class, and this just give the dimension of the representation, which
for this case is just the dimension of the group N (G)3 . Hence from (3.23) we
get,
χ∗r (e)N (G) = N (G)nr .
But χr (e) = dr , the dimension of irrep r. Therefore we obtain that in the
regular representation,
dr = nr . (3.25)
3
i.e, χregular (e) = N (G)

13
It is highly reducible, moreover the irrep r is contained dr number of times!
If we look at the regular representation for (3.24) we get:
X X
N (G)2 = N (G) n2r , or, n2r = N (G). (3.26)
r r

Plugging in (3.25) into (3.26) we obtain,

X
d2r = N (G). (3.27)
r

Character table : §II.3 of Group Theory in a Nutshell for Physicists,

A. Zee, Princeton University Press, 2016
[End of Lecture 7]

4 Young tableaux for Sn

The irreps of Sn are in one to one correspondence with Young tableaux. The
corresponding ones for the case of S5 are given by, , , , , ,
, . These also correspond to the cycle structure and thus the partition
of integers. For instance, we can write 5 = 5, 4 + 1, 3 + 2, 3 + 1 + 1, 2 + 2 +
1, 2 + 1 + 1 + 1, and, 1 + 1 + 1 + 1 + 1. The cycle structures respectively
are, (∗ ∗ ∗ ∗ ∗), (∗ ∗ ∗∗)(∗), (∗ ∗ ∗)(∗∗), (∗ ∗ ∗)(∗)(∗), (∗∗)(∗∗)(∗), (∗∗)(∗)(∗)(∗),
and, (∗)(∗)(∗)(∗)(∗). The first row of the Young tableaux represents the
largest cycle structure, second row the second largest and so on. The trivial
representation is the last one which corresponds to :

(1)(2)(3)(4)(5) =

The number 5 has been partitioned into integers in 7 different ways 4 , which
corresponds to Nc = 7 conjugacy classes, which is also the possible cycle
structures5 . Since we have, Nc = NR , these are also in correspondence with
the number of irreducible representations. The Young diagrams can be used
to calculate the dimensions of the irreducible representations using Hook’s
law.
4
In Problem Set 2 you have seen p(5) = 7.
5
The cycle structure is invariant under conjugation – see here

14
 • Starting from the first row, pass a line through the boxes, from the
right to the left.
• Next take a left-turn of 90o in order to exit the figure. This turn makes
a shape of hook.
• Count the number of boxes the hook passed through. Let the product
of all possible hook lengths be called H.
• The dimension of the corresponding irreducible representation of Sn is
given via
n!
dr = .
H
• Finally an example for the (∗ ∗ ∗)(∗∗) cycle structure of S5 :

The coloured lines in the above diagram show the non-trivial hooks,
we calculate
5!
d= = 5.
4×3×2
Calculating similarly we obtain: , d1 = 1, , d2 = 4, , d3 = 5,
, d4 = 6, , d5 = 5, , d6 = 4, , d7 = 1. We can immediately verify
(3.27), that,
X
n! = d2r = 5! = 120 = 12 + 42 + 52 + 62 + 52 + 42 + 12 .
r

For the use of Young tableaux for SU (N ) see §8. [End of Lecture 8]

5 Group theory in Quantum Mechanics

Follow the discussion in §III.1 of Group Theory in a Nutshell for Physicists,
A. Zee, Princeton University Press, 2016 [End of Lecture 9]

15
6 Lie Groups
Continuous groups have infinite group elements. However to parametrize
them one can often use finite number of real quantities. A convenient way to
describe the transformations enacted by the group is via matrix representa-
tions where the finite number of real quantities enter as variables. Here we
shall look at some important examples and count the number of the variables.
Since we are concerned with Lie groups, we will be looking close to the
identity, thus the matrix group is expanded around the identity matrix,

D(g) = 1 + X.

As we shall see in the next lecture, the matrices X are the generators, closing
under Lie algebra, and can be exponentially parametrized to generate finite
group transformations in the neighbourhood of identity.
The number of generators is same as the number of real variables describing
the group. Each variable can be thought of as the strength of the transfor-
mation away from identity, in the direction of the corresponding generator
element.
• the group GL(n, C) is the group of n × n complex matrices, which are
invertible. Since every entry of the matrix is an independent complex
number, one has a total of 2n2 number of variables.

• the group SL(n, C) is the above with the restriction that the determi-
nant is 1.

Note: S in front of a group acronym, stands usually for special. This

implies that the determinant is set to 1. This is a convenient way to normalize
the group volume as we shall see later. This constraint implies tracelessness
of the generators. This can be seen by using the expansion,

det(1 + A) = 1 + trA + O(2 ),

which can be derived, using

Y X
det(1 + A) = (1 + λi ) = 1 +  λi + O(2 ) = 1 + trA + O(2 ).
i i

16
where, λi ’s are the eigenvalues of the generator matrix X. Therefore, de-
manding det(1 + A) = 1, sets trX = 0.

• U(n) The Unitary group satisfies, U U † = 1. These are complex matri-

ces. For the generators the defining relationship implies,
(1 + X)((1 + X)† = 1, or, X = −X † .
Thus X is a n × n complex matrix whose entries are related as, Xij =
−Xji∗ . The diagonals can only be of the form, i × something real. Thus
there are n of them. The rest, come in pairs, that is, Xij determines,
Xji . Thus there are n(n − 1)/2 of these entries. Since these are com-
plex numbers the number of reals is, n(n − 1). Hence total number of
generators is = n + n(n − 1) = n2 . From the condition that
† ∗
Uij Ujk = 1, or, Uij Ukj = 1.
we can conclude that |Uij | ≤ 1. This boundedness of the norm of U
implies that U is a compact group.
• SU (n) In this case the determinant condition gives one constraint
among the entries. Thus the number of generators is n2 − 1.
• O(n) : The orthonormal matrix group is a real matrix of size n × n
satisfying, OOT = 1. This implies that the generators satisfy,
Xij = −Xji .
Thus the generators are real antisymmetric matrices, and hence there
are n(n−1)/2 of them. Since det AB = det A det B and det AT = det A,
we have, det OOT = (det O)2 = 1. Thus
det O = ±1.

• SO(n) : As the name suggests these special orthogonal groups are

matrix groups which are real and orthogonal and also have determi-
nant 1. However since the orthogonal group generators have all zeros
in the diagonal, the tracelessness condition is automatically satisfied.
Thus SO(n) also has, n(n − 1)/2 independent generators. This is the
symmetry group of rotations in the Euclidean space.

17
 • Sp(2n, R) : The real symplectic group of rank 2n is a 2n × 2n matrix
S which satisfies :
S T ωS = ω. (6.28)
where  
0n×n −In×n
ω= .
In×n 0n×n
Note that even for Sp(2n, C), when the entries are complex, the defining
relation is (6.28), with a transpose. Thus the generators satisfy :
X T ω = −ωX. (6.29)
Parametrizing the generator with the following n × n matrices :
 
an×n bn×n
X= .
cn×n dn×n
Now (6.29) gives the following constraints :
d = −aT b = bT , c = cT .
Thus we have one arbitrary real matrix, and two symmetric real matri-
ces. Therefore the total number of generators is, n2 + 2 × n(n + 1)/2 =
n(2n + 1).

6.1 Geometry of SU (2)

The matrix group A = SU (2) is a 2×2 complex matrix with unit determinant,
and with 22 − 1 = 3 independent real parameters. Additionally, there is the
defining property that A† = A−1 . A convenient way to parametrize A is by,
 0 
x + ix3 ix1 + x2
A= . (6.30)
ix1 − x2 x0 − ix3
The unit determinant condition is :
3
X
(xi )2 = 1. (6.31)
i=0

The inverse is,

 
−1 † x0 − ix3 −ix1 − x2
A =A = . (6.32)
−ix1 + x2 x0 + ix3

18
Equation (6.31) describes a three dimensional sphere, S 3 with unit radius.
The S 3 coordinates are x1 , x2 , x3 , and x0 is determined via (6.31) as,
v
u
X3
0
u
x = 1− t (xj )2 . (6.33)
j=1

Using the Pauli matrices6 ,

   
1 0 1 2 0 −i
σ = , σ = , (6.34)
1 0 i 0
 
1 0
σ3 = , (6.35)
0 −1

we can rewrite,
A = x0 I2×2 + ix1 σ 1 + ix2 σ 2 + ix3 σ 3 .
Away from identity, there are 3 independent directions along the 3 generators.
These near identity elements along the three directions are :

1 + iσ i , for i ∈ {1, 2, 3}.

Let us find how the 1 + iσ 3 acts on the group element A,

A → A · (1 + iσ 3 ) = (x0 − x3 ) + iσ 1 (x1 − x2 ) + iσ 2 (x2 + x1 ) + iσ 3 (x3 + x0 ).

Thus the change in the parameters xi (i ∈ {0, 1, 2, 3}), under the action of
(1 + iσ 3 ) can be written as,
 0  3
x −x
 x1  −x2 
δ x2  =   x1 
   (6.36)
x3 x0

Thus we can read-off the generator of this deformation in iσ 3 direction as,

∂ ∂ ∂
L3 = −x2 1
+ x1 2 + x0 3 (6.37)
∂x ∂x ∂x
6
Reminder : {σ i , σ j } = 2δ ij I2×2 and [σ i , σ j ] = 2iijk σ k

19
Similarly,
∂ ∂ ∂
L1 = x0 1
− x3 2 + x2 3
∂x ∂x ∂x
3 ∂ 0 ∂ 1 ∂
L2 = x +x −x (6.38)
∂x1 ∂x2 ∂x3
Now, using (6.33),
∂x0 xi
= − , for, i = 1, 2, 3.
∂xi x0
With the above, we can explicitly check that,

[Li , Lj ] = −2ijk Lk . (6.39)

This coincides with the albegra of iσ i ’s as it should, since both are represen-
tations of the generators of SU (2).
[End of Lecture 10]

6.2 Lie Algebra

Let the number of independent real quantities needed to describe the continu-
ous group G be N , and we call them as, αa with a = 1, 2, . . . , N . The group el-
ements are functions (here we consider smooth ones) of αa , g ∈ G, g = g(αa ).
We choose, αa=0 = e. Let the matrix representations of the group elements
be D(g(α)) = D(α). Thus with our choice about identity, we have,

D(0) = 1.

(note we interchangeably use 1 for the identity matrix I). Expanding in-
finitesimally around α = 0,

D(dα) = 1 + idαa Xa .

Here, dαa is a small variation of αa , and Xa ’s are the generators. We also

have the implicit summation over the a index. From the above equation it
is also clear that,
∂
Xa = −i D(α) .
∂αa αa =0

We shall mostly be focussed on groups which have unitary representations.

In that case, Xa are hermitian. Now following Lie, to generate a finite group

20
transformation, we can keep acting by the infinitesimal transformation many
number of times;
 k
α a Xa
D(α) = lim 1 + i = eiαa Xa .
k→∞ k
Note, that we have also used, dαa = αa /k. This is called the exponential
mapping, how from a generator, we can get to finite group transformations
(at least those connected to the identity element). This implies, that we can
focus our study to the generators, which is nice since the generators form a
vector space.

Now we want to see how the closure property of the group G, enforces
the Lie algebra of the generators X. Closure implies,
eiαa Xa eiβb Xb = eiδc Xx .
This is same as,
1
iδa Xa = log 1 + eiαa Xa eiβb Xb − 1 = log(1 + K) ' K − K 2 .


2
We have just added and subtracted 1 within the parentheses in the r.h.s.
Next we expand the exponentials in K = eiαa Xa eiβb Xb − 1, and keep till
quadratic order in αa , βa ’s. Finally we get,
[αa Xa , βb Xb ] = 2i(αc + βc − δc )Xc = iγc Xc .
Next with the choice, γc = αa βb fabc , we can rewrite the above as,
[Xa , Xb ] = ifabc Xc .
Clearly the l.h.s is antisymmetric in a, b, thus,
fabc = −fbac .
From the hermiticity of X (which is true given a unitary representation),
[Xa , Xb ]† = −ifabc
∗
Xc = [Xb , Xa ] = ifbac Xc = −ifabc Xc .
This shows that fabc are real quantities.
Also once can go on to show,
[Xa , [Xb , Xc ]] + [Xb , [Xc , Xa ]] + [Xc , [Xa , Xb ]] = 0.
This is called the Jacobi identity. [End of Lecture 11]

21
Adjoint Representation
It turns out that,
(Xa )bc = −ifabc , (6.40)
forms a representation of the generator. From the Jacobi identity one has,

fabd fdcg + fbcd fdag + fcad fdbg = 0.

Using the above, one can show that with the adjoint representation (6.40)
one has group closure property.

6.3 Certain aspects of SO(3)

Refer to §SO(3) : The representation (2.5) is the adjoint representation.
This is because one can check,

(Ja )bc = −iabc .

Now to construct other representations of SO(3) we focus on the algebra

generated by,
J± = Jx ± iJy , and Jz .
The relevant brackets give,

[J± , Jz ] = ∓J± and [J+ , J− ] = 2Jz (6.41)

Since the generators do not commute we cannot simultaneously diagonalize

all three of them. Let us choose to work with the Jz basis :

Jz |mi = m|mi. (6.42)

Using the algebra (6.41) we have :

Jz J+ |mi = (m + 1)J+ |mi.

Jz J− |mi = (m − 1)J− |mi. (6.43)

Thus we can rename:

J+ |mi = cm+1 |m + 1i,
and,
J− |mi = bm−1 |m − 1i.

22
Here, cm , bm are unfixed normalizations. The generators J± act as rais-
ing(lowering) operators. Since the generators are hermitian, we have (J± )† =
J∓ .
Since, hm + 1|J+ |mi = cm+1 , and hm|J− |m + 1i = bm ,
the hermiticity implies,
bm−1 = c∗m .
Thereby,
J− |mi = c∗m |m − 1i.
Thus we also have,
J+ J− |mi = |cm |2 |mi (6.44)
J− J+ |mi = |cm+1 |2 |mi.
Since we are after finite irreps of SO(3), we would need the states to terminate
somewhere, so that it cannot be raised further. This defines the highest
weight state. Let it have a Jz eigenvalue j. Then Jz |ji = j|ji. Now since
this is the highest weight,
hj|J− J+ |ji = 0.
However, using (6.41) this is also same as,
hj|J+ J− − 2Jz |ji = |cj |2 − 2j.
where we also use (6.44). Thus we immediately have,
|cj |2 = 2j. (6.45)
To figure out the rest of the normalizations, we use again the commutator,
[J+ , J− ] (6.41) along with (6.44),
hm|[J+ , J− ]|mi = hm|J+ J− − J− J+ |mi = |cm |2 − |cm+1 |2 = hm|2Jz |mi = 2m.
∴ |cn |2 = |cn+1 |2 + 2n. (6.46)
Now we can use the above recursion relation multiple times to determine,
|cj−1 |2 , |cj−2 |2 , . . . , |cj−s |2 using the highest weight normalization value (6.45).
|cj−1 |2 = |cj |2 + 2(j − 1) = 2(2j − 1)
|cj−2 |2 = |cj−1 |2 + 2(j − 2) = 2(3j − (1 + 2))
(6.47)

23
Identifying the pattern, in the general case we have,
s
X
2
|cj−s | = 2((s + 1)j − k) = (s + 1)(2j − s). (6.48)
k=1

This tells us now, that we hit zero after going down s = 2j steps, so, |c−j |2 =
0. Thus we have 2j + 1 states in total,

|ji, |j − 1i, . . . , |j − si, . . . , | − j + 1i, | − ji.

Thus this representation is 2j + 1 dimensional, i.e., the action of the SO(3)

group is described by a 2j +1×2j +1 matrix in this particular representation.
Since s = 2j needs to be an integer (number of times one lowers), =⇒ j ∈
1
2
Z. Thus j can be half-integers as well. Additionally we can also determine
the normalization constant |cm |2 from (6.48) by setting s = j − m. Since we
cannot determine the phase, we choose the cm ’s to be reals, then we have,
p
J+ |mi = (j + 1 + m)(j − m)|m + 1i
p
J− |mi = (j + 1 − m)(j + m)|m − 1i. (6.49)

A product representations of two such j and j 0 representations in general

will be composed of other irreps of SO(3). Let us see how this works out.
Given the product representation, j ⊗ j 0 , we can write a generic state in this
representation as,
|j, m; j 0 m0 i.
[End of Lecture 12]

The way a group element g ∈ G acts on the product states is the following:

D(r) (g) ⊗ D(s) (g) |j (r) i ⊗ |j (s) i = D(r) (g)|j (r) i ⊗ D(s) (g)|j (s) i .

(r)
In terms of generators, D(r) = 1+iθa Xa , thus plugging this in the generator
acts as,
Xa (|ji ⊗ |j 0 i) = Xa |ji ⊗ |j 0 i ⊕ |jiXa |j 0 i. (6.50)
Thus the Jz eigenvalue of the state |j, m; j 0 , m0 i is m + m0 which at the most
can be j + j 0 . Next, J− can act on this state to produce, |j, m − 1; j 0 , m0 i
and |j, m; j 0 , m0 − 1i, both of whose Jz eigenvalue is m + m0 − 1, which can

24
at the most be j + j 0 − 1. Thus this is part of the j + j 0 − 1 highest weight
representation. Inductively therefore,

j ⊗ j 0 = (j + j 0 ) ⊕ (j + j 0 − 1) ⊕ · · · ⊕ |j − j 0 |. (6.51)

One can check that all the states are accounted for. In both sides there are a
total of (2j + 1)(2j 0 + 1) states. In the following we look at an example with
j = j 0 = 21 . The highest weight state in the spectrum is, |1/2, 1/2; 1/2, 1/2i.
This has Jz value 1, and since this is highest √ weight, it is = |1, 1i. Acting by
Jz on this gives, using (6.49), J− |1, 1i = 2|1, 0i. However this should be
same as J− |1/2, 1/2; 1/2, 1/2i = |1/2, −1/2; 1/2, 1/2i + |1/2, 1/2; 1/2, −1/2i.
Thus we identify,
1
|1, 0i = √ (|1/2, −1/2; 1/2, 1/2i + |1/2, 1/2; 1/2, −1/2i) . (6.52)
2
Applying J− once again to both hand sides of the above equation gives :
1 1 1 1
|1, −1i = | , − ; , − i. (6.53)
2 2 2 2
This is the lowest weight state. The only other missing state in the decom-
position is the singlet |0i. One can easily convince that this is the state
:
1
|0i = √ (|1/2, −1/2; 1/2, 1/2i − |1/2, 1/2; 1/2, −1/2i) . (6.54)
2
Note, that the sign is undetermined. The coefficients relating one basis of
irreps to another are known as the Clebsch-Gordon coefficients,
j j 0
X X
|J, M i = |j, m; j 0 m0 ihj, m; j 0 , m0 |J, M i. (6.55)
m=−j m0 =−j 0

Here in the r.h.s one has inserted identity by using completeness and the
Clebsch-Gordon coefficients are the overlaps,

hj, m; j 0 , m0 |J, M i.

[End of Lecture 13]

25
6.4 SU (2) and SO(3)
Reminder: SO(3) generates rotations of 3 dimensional vectors (§SO(3)):

xi → x0i = Rij xj . (6.56)

Here the matrix R, is made up of real entries, with the property that RRT =
1, and det R = 1. Whereas, an element of SU (2) can be represented by the
2 × 2 complex matrix :  
a b
u= . (6.57)
−b∗ a∗
This has 4 real quantities (a, b are complex numbers). Also the property,
uu† = 1 needs to be satisfied. This holds provided,

|a|2 + |b|2 = 1. (6.58)

This is also the necessary condition to have, det u = 1. Now consider the
matrix :
h = σ i xi . (6.59)
Explicitly,  
x3 x1 − ix2
h= . (6.60)
x1 + ix2 −x3
And enact a SU (2) transformation on it,

h → h0 = uhu† (6.61)
   ∗ 
a b x3 x1 − ix2 a −b
=
−b∗ a∗ x1 + ix2 −x3 b∗ a
x03 x01 − ix02
 
= 0 0 (6.62)
x1 + ix2 −x03

From explicit matrix multiplication one can read off the transformed x0i ’s :
   
0 1 2 2 i ∗ 2 ∗ 2
x1 = (a − b ) + c.c. x1 + ((a ) + (b ) ) + c.c. x2 + (−ab + c.c.) x3 ,
2 2
     
0 i ∗ 2 ∗ 2 1 2 2 i ∗ ∗
x2 = ((b ) − (a ) ) + c.c. x1 + (a + b ) + c.c. x2 + a b + c.c. x3 ,
2 2 2
x03 = (a∗ b + c.c.) x1 + (ia∗ b + c.c.) x2 + |a|2 − |b|2 x3 .


(6.63)

26
One can check that this transformation has the properties of an SO(3) trans-
formation. Constructing the R matrix from above :
(a − b2 ) + c.c.
2i ((a∗ )2 + (b∗ )2 ) + c.c.
1 2



2
(−ab + c.c.)

R =  2i ((b∗ )2 − (a∗ )2 ) + c.c. 1 2 i ∗ ∗


2
(a + b2 ) + c.c. 2
a b + c.c.  .
∗ ∗
(a b + c.c.) (ia b + c.c.) (|a|2 − |b|2 )
(6.64)
For the above matrix all the entries are real. Furthermore,
 2 
|a| + |b|2 0 0
RRT =  0 |a|2 + |b|2 0 . (6.65)
2 2
0 0 |a| + |b|
But this is just the identity matrix, due to (6.58). Similarly,
det R = (|a|2 + |b|2 )3 = 1.
Thus R as defined in (6.64) represents SO(3). Hence from the representation
of SU (2) we can write down the R matrix of SO(3). Going back to (6.59)
and (6.61) and writing in terms of components (with summation implicit),
h0 = σj x0j = uσi xi u† . (6.66)
Now using the property of Pauli matrices :
Tr (σj σk ) = 2δjk

we can solve for x~0 as follows. First multiply b.h.s of (6.66) by σk and then
take trace:
Tr σk h0 = Tr (σk σj ) x0j = Tr σk uσi u† xi

or 2x0j δjk = Tr σk uσi u† xi

1
∴ x0j = Tr σk uσi u† xi .


(6.67)
2
Thus
1
Rji = Tr σj uσi u† .


2
Note from the above that both u and −u give the same matrix element Rij .
Thus SU (2) is not just SO(3), rather, it covers SO(3) twice. This choice of
sign is expressed in the following isomorphism relation:
SO(3) ' SU (2)/Z2 .

27
The upshot of this discussion is that the irreps of SO(3) are also irreps of
SU (2). The generators of SU (2) also satisfy the same Lie algebra as (2.8).
However we will follow standard SU (2) conventions and define the raising
and lowering operators are bit differently :
J1 ± J2
J± = √ . (6.68)
2
√
In comparison with (6.41) there is an extra prefactor of 1/ 2, due to which
(6.49) is also modified to :
p
(j + 1 + m)(j − m)
J+ |mi = √ |m + 1i
2
p
(j + 1 − m)(j + m)
J− |mi = √ |m − 1i. (6.69)
2
And, (6.41) is modified to,

[Jz , J± ] = ±J± and [J+ , J− ] = Jz (6.70)

7 Weights and Roots

The largest subset of commuting Hermitian generators are called the Cartan
subalgebra : Hi , where i = 1, 2, . . . , m (m is called the rank of the Lie
algebra). They thus satisfy,

Hi = Hi† , and , [Hi , Hj ] = 0.

Since these are simultaneously diagonalizable, we can choose a basis such

that,
Tr [Hi Hj ] = kD δij . (7.71)
Here D subscript denotes representation. In the case of SU (2) the Cartan
subalgebra, just consisted of J3 . (we could have also chosen any one of the
Ja ’s). Thus as in SU (2) we work with the states, that are eigenstates of Hi ,

Hi |µ, x, Di = µi |µ, x, Di. (7.72)

Here the eigenvalues, µi are called the weights, and in general the vector,
(µ1 , µ2 , . . . , µm ) is called the weight vector.

28
Adjoint Representations
Recall the definition of Adjoint representation from §AdjointRep . In this case
the states themselves can be labelled by the generators, Xa as, |Xa i. One
can also have a linear combination of generators as a state : |αXa + βXb i.
The inner product can be defined as,
1
Tr Xa† Xb .


hXa |Xb i =
λ
In the above, λ is just kD as in (7.71) for the adjoint representation. Let us
next see, how generators act on these states:

Xa |Xb i = |Xc ihXc |Xa |Xb i = |Xc i [Ta ]cb = ifabc |Xc i = |[Xa , Xb ]i. (7.73)

Thus the generators in the adjoint representation act as commutators.

[End of Lecture 14]
The weights in the adjoint representation are called roots. Let Eα be a
generator, then in adjoint |Eα i is a state with weight : Hi |Eα i = αi |Eα i.
Note that this implies :
[Hi , Eα ] = αi Eα . (7.74)
Next consider the state, |Eᆠi. Let us calculate its root. Conjugating (7.74),

[Eα† , Hi† ] = αi∗ Eα† ,

or, [Hi , Eα† ] = −αi Eα† . (7.75)

Thus from the above it is clear that we can rewrite,

Eα† = E−α . (7.76)

Though we do not prove it here, it can be shown that the label α is unique
for a generator, i.e., there cannot be a β 6= α such that, Eα and Eβ are the
same generator. This is called the uniqueness theorem for the roots. Since
the states, |Eα i are eigenstates of the Hermitian operators, Hi , we can and
will choose a basis such that these are normalized, i.e.,
1
hEβ |Eα i = Tr (E−β Eα ) = δαβ . (7.77)
λ
Note that we also have,
hHi |Hj i = δij .

29
Next we show that (7.74) implies that the generators E±α act as raising and
lowering operators (similar to J± for SU (2) ),

Hi E±α |µ, Di = E±α Hi |µ, Di ± αi E±α |µ, Di

= (µi ± αi ) E±α |µ, Di. (7.78)

In the above first equality we used the algebra, (7.74). This implies that the
state, Eα |E−α i has weight zero. Thus it must be linear combination of the
Cartan generators. Thus,

Eα |E−α i = βi |Hi i. (7.79)

But since |E−α i is in the adjoint representation, we also know :

Eα |E−α i = |[Eα , E−α ]i. (7.80)

Thus we conclude that,

|[Eα , E−α ]i = βi |Hi i. (7.81)

From the above using orthogonality relations we can write,

βi = hHi |Eα |E−α i

1
= Tr (Hi [Eα , E−α ])
λ
1
= Tr ([Hi , Eα ]E−α )
λ
αi 
†

= Tr Eα E−α
λ
αi 
†

= Tr E−α Eα
λ
= αi hEα |Eα i = αi . (7.82)

Thus we have shown that βi = αi . In the third and the fifth equalities we
used the cyclicity property of the trace. Next, we can define :
1 α·H
E± = E±α , E3 = . (7.83)
α α2
Then using (7.74), one can derive:

[E3 , E ± ] = ±E ± , and [E + , E − ] = E3 . (7.84)

30
The above is exactly the algebra of SU (2) (6.70), thus we see that given the
Cartan subalgebra and given a particular generator Eα we can construct an
SU (2) subalgebra for the group G. We shall call this algebra, SU (2)α .
[End of Lecture 15] Acting on a state in representation D,
α·µ
E3 |µ, x, Di = |µ, x, Di. (7.85)
α2
Since this is an SU (2) eigenvalue this must be either and integer or a half-
integer; therefore,
2α · µ
∈ Z. (7.86)
α2
Now let (E + )p |µ, x, Di be a highest weight state, which means that there is
some integer p for which,

(E + )p+1 |µ, x, Di = 0.

Now the weight of this highest weight state can be directly evaluated :
α · µ 
E3 (E + )p |µ, x, Di = + p |µ, x, Di. (7.87)
α2

[E3 , (E + )p ]: For the above computation we needed this commutator. The

strategy is to work step by step :

E3 (E + )p = E3 E + (E + )p−1 = E + E3 (E + )p−1 + (E + )p
= (E + )2 E3 (E + )p−2 + 2(E + )p = · · · = (E + )p E3 + p(E + )(7.88)
p
.

Thus,
[E3 , (E + )p ] = p(E + )p .

Thus we have :
α·µ
+ p = j. (7.89)
α2

31
Similarly we construct a lowest weight state with,
(E − )q |µ, x, Di,
whose E3 eigenvalue gives :
α·µ
− q = −j. (7.90)
α2
We can now add (7.89) and (7.90) and obtain:
α·µ 1
= − (p − q) . (7.91)
α2 2
Now, consider two generators, Eα and Eβ . Both has their own SU (2)s.
Consider first, SU (2)α . In this case (7.91) for the case when the state
|µ, x, Di = |Eβ i we have,
α·β 1
2
= − (p − q). (7.92)
α 2
Similarly for SU (2)β algebra with state, |Eα i we have,
β·α 1
2
= − (p0 − q 0 ). (7.93)
β 2
Next multiplying (7.92) and (7.93),
(β · α)2 1
2 2
= (p − q)(p0 − q 0 ) = cos2 θαβ . (7.94)
β α 4
In the above we see that the l.h.s is the angle between the root vectors θαβ
whose cosine squared appears. Thus the product, (p − q)(p0 − q 0 ) which
appears in the middle equality should be positive (since cosine squared is
positive) and also should be an integer (since each of p, q, p0 , q 0 are integers).
Thus there are only 4 cases :
(p − q)(p0 − q 0 ) θαβ
0 90o
1 60o or 120o
2 45o or 135o
3 30o or 150o
4 0o or 180o
In the above, the last case is again trivial. The zero degree case is ruled
out due to the uniqueness theorem of the roots, while the 180o case is the
generator E−α .

32
7.1 SU(3)
Let us look at the group SU (3) and how the above features fit in here. For
SU (3) there are 8 generators, which are traceless, hermitian matrices of rank
3. The standard basis for them are generalizations of Pauli matrices and are
called Gellmann matrices in the literature. This also furnishes us with the
fundamental representation of SU (3).
     
0 1 0 0 −i 0 1 0 0
λ1 = 1 0 0 λ2 =  i 0 0 λ3 = 0 −1 0
0 0 0 0 0 0 0 0 0
     
0 0 1 0 0 −i 0 0 0
λ4 = 0 0 0 λ5 = 0 0 0  λ6 = 0 0 1 (7.95)
1 0 0 i 0 0 0 1 0
   
0 0 0 1 0 0
1
λ7 = 0 0 −i λ8 = √ 0 1 0  .
0 i 0 3 0 0 −2

Conventionally, the generators of SU (3) are defined as,

1
Ta = λa . (7.96)
2
They satisfy,
1
Tr (Ta Tb ) = δab . (7.97)
2
One can check that the structure constants:
[Ta , Tb ] = ifabc Tc , (7.98)
take the explicit values :
f123 = 1,
1
f147 = −f156 = f246 = f257 = f345 = −f367 = (7.99)
√ 2
3
f458 = f678 = .
2
The Cartan subalgebra in this case is generated by T3 = H1 and T8 = H2 .
The eigenvalues are simply,
√ √
1 3 3
(h1 , h2 ) = (± , ), (0, − ). (7.100)
2 6 3
33
 H2

0.2

0.0 H1

-0.2

-0.4

-0.6

-0.4 -0.2 0.0 0.2 0.4

Figure 1: Weight diagram for SU (3) fundamental, also called 3 due to its dimensionality.

These form an inverted equilateral triangle in the H1 − H2 plane. This is

called the weight diagram.

The figure 1 is the weight diagram for SU (3) in the fundamental representa-
tion. The other 6 generators generate movements between the points in this
weight lattice. [End of Lecture 16]
One can thus determine the 6 roots :
1 1
E±1,0 = √ (T1 ± iT2 ) , E± 1 ,± √3 = √ (T4 ± iT5 ) , (7.101)
2 2 2 2
1
E∓ 1 ,± √3 = √ (T6 ± iT7 ) .
2 2 2
To find the combinations of the generators appearing above we can act by
the Cartans and fix the coefficients. For instance, to find E±1,0 , by definition
we need to have :

H1 |E±1,0 i = ±1|E±1,0 i, and, H2 |E±1,0 i = 0|E±1,0 i. (7.102)

This can be used to fix E±1,0 in terms of Ta upto a overall normalization.

The latter is fixed by demanding,

hE±1,0 |E±1,0 i = 1. (7.103)

34
 H2

1.0

0.5

0.0 H1

-0.5

-1.0

-1.0 -0.5 0.0 0.5 1.0

Figure 2: Weight diagram for SU (3) adjoint, also called 8.

Let us see how (7.101) satisfy the above requirements. We have already
identified, H1 = T3 and H2 = T8 . Thus,
1
H1 |E±1,0 i = |[H1 , E±1,0 ]i = |[T3 , √ (T1 ± iT2 )]i
2
1 i
= √ |[T3 , T1 ]i ± √ |[T3 , T2 ]i
2 2
1 i
= √ |if31j Tj i ± √ |if32j Tj i
2 2
1 i
= √ |if312 T2 i ± √ |if321 T1 i
2 2
i 1 1
= √ |T2 i ± √ |T1 i = ± √ (|T1 i ± i|T2 i)
2 2 2
= ±1|E±1,0 i. (7.104)

Clearly since the individual states, |Ta i are ortho-normalized the overall nor-
malization is √12 . The weight diagram for the adjoint representation (which
is 8 dimensional and also called 8) has the shape of a hexagon along with
two weights in the origin corresponding to the Cartans. See Fig.2.

Another important irrep of SU (3) is the anti-fundamental representation

35
 H2

0.6

0.4

0.2

0.0 H1

-0.2

-0.4 -0.2 0.0 0.2 0.4

Figure 3: Weight diagram for SU (3) anti-fundamental, also called 3̄ due to its dimen-
sionality and conjugation.

which also has dimensionality 3 and is obtained by conjugating the algebra.

√ √
1 3 3
(h1 , h2 ) = (∓ , − ), (0, ). (7.105)
2 6 3

Product representations
The weight diagrams can also be drawn for product representations. And
then one can use it to decompose it into the contained irreducible represen-
tations. To see how this is done, we first look at SU (2) where we know the
decomposition rule (6.51). Let us consider the example
1 1
⊗ .
2 2
In the case of SU (2) there is only one axis for the Cartan generator which
we choose to be J3 . The first step is to list all the possible weights contained
in the product representation. For the above example the weights are the
J3 eigenvalues of the four states | 12 , ± 12 ; 12 , ± 12 i. The J3 values simply add
up and gives us three possible values −1, 0, 1. However there are two zeros,
corresponding to addition of 12 and − 12 which happens twice. Thus we have

36
 0.10 0.10
0.05 0.05
-1.0 -0.5
-0.10
0.5 1.0
H
= -1.0 -0.5 0.5 1.0
H
⊕ -1.0 -0.5
-0.10
0.5 1.0
H

1 1
Figure 4: The weight diagram indicates 2 ⊗ 2 = 1 ⊕ 0.

the following weight diagram as given in l.h.s of Fig. 4. The next step is
to realize that in the weight diagrams of the irreducible representations of
SU (2) none of the weights are degenerate. For instance in the spin-j irrep of
SU (2), the possible J3 eigenvalues, −j, −j + 1, . . . , j − 1, j, occur only once.
Thus the decomposition of the diagram is into the two diagrams as shown in
the r.h.s of Fig. 4. These are nothing but the weight diagrams of j = 1 and
j − 0. Thus we have derived,
1 1
⊗ = 1 ⊕ 0.
2 2
For SU (3) it is a bit tricky since in the irreps of this group, weights can have
degenerate values. Clearly the adjoint weight diagram (Fig. 2) has double
degeneracy at the origin reflecting the rank 2 feature of SU (3). Let us look
at the following example for SU (3) for the case,
3 ⊗ 3̄.
In this case, following the first step let us list down all the possible weights in
this product representation. We simply need to do all possible additions of
the weights which are given in (7.100) to the weights in (7.105). We obtain
the set,
√ √ √ √
1 3 1 3 1 3 1 3
(0, 0), (−1, 0), (1, 0), (− , ), (− , − ), (− , − ), ( , − ).
2 2 2 2 2 2 2 2
In the above (0, 0) appears thrice, all others appear once. This gives the
weight diagram, l.h.s of Fig. 5 from which we can immediately see that it
is built out of Fig. 2 and the singlet, which is the trivial 1 dimensional
representation (just one weight in the origin). Thus we conclude:
3 ⊗ 3̄ = 8 ⊕ 1. (7.106)
[End of Lecture 17]

37
 H2 H2 H2

1.0 1.0 1.0

0.5 0.5 0.5

0.0 H1 = 0.0 H1 ⊕ 0.0 H1

-0.5 -0.5 -0.5

-1.0 -1.0 -1.0

-1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0 -1.0 -0.5 0.0 0.5 1.0

Figure 5: The weight diagram indicates 3 ⊗ 3̄ = 8 ⊕ 1.

8 Young Tableaus for SU (N )

A standard Young diagram has the following form:

...
...
.. .. ..
. . .

• Each tableaux corresponds to a irreducible representation.

• The trivial irrep (singlet) consists of N vertical boxes.

• There cannot be more boxes in a row below and on a column to the

right.

• The boxes can be associated with numbers that can be used to calculate
dimension of the irrep. The way numbers are assigned are as follows :

– The top left box is given a number N for SU (N ).

38
 – Numbers increase by one in a given row. The boxes in the top
row next to the left most will have numbers: N + 1, N + 2, . . . .
– Numbers decrease by one in a given column. In the left most
column, the boxes below the N box have numbers, N − 1, N −
2, . . . .
N N +1 N +2 ...
N −1 N N +1 ...
.. .. ..
. . .

• The conjugate of an irrep via Young diagram is found by i) first replac-

ing the j boxes in a column by N − j boxes, and ii) then flipping the
diagram w.r.t the left most vertical axis. For example in SU (4),
 ∗  flip
   
  =  =
   

• The dimension of a Young diagram for SU (N ) can be calculated using,

Factors
dr = .
Hooks
Here the numerator is the product of all the entries in the boxes of the
diagram, and the denominator is the product of all the hook numbers.
For example we find for the irrep given by, , For N = 3 i.e., SU (3)
we obtain dr = 3, this is the Young diagram,
∗
= = ,

which is the anti-fundamental representation 3̄ that as we know is three

dimensional.

39
 N 2 (N + 1)(N − 1)(N − 2)
: dr = .
4×3×2

Figure 6: Non-trivial hooks are drawn in the given irrep of SU (N ).

• The decomposition of product representations into irreps are obtained

by all legal joinings of the second irrep in the tensor product with the
first irrep. Ordering has to be maintained, for this see §12.2 of Lie
Algebras in Particle Physics by H. Georgi. For instance in the case of
3 ⊗ 3̄ we obtain :
⊗ = ⊕

Computing the dimensions show that this is,

3 ⊗ 3̄ = 8 ⊕ 1.

Another example can be for SU (3), 3 ⊗ 3,

⊗ = ⊕

Computing the dimensions show that this is,

3 ⊗ 3 = 6 ⊕ 3̄.

[End of Lecture 18]

40