Q. HO-KIM - Group Theory: A Problem Book

Group Theory
A Problem Book
Q. Ho-Kim
c Q. Ho-Kim (H Kim Quang). Group Theory: A Problem Book 2015-12.

Cover illustration from image by Jo Edkins, used with permission.
Group Theory: A Problem Book

This book contains suggested solutions to the end-of-chapter problems
in Group Theory: A Physicists Primer by the same author, to be
referred to as GTAPP, which is freely available at
http://www.scribd.com/doc/207786199/Group-Theory-A-Physicists-Primer.
Additional information on notations and definitions can be found in
the corresponding chapters of GTAPP.
Problems from Chapter 1
Subject Index
110
1127
2748
4870
7179
7986
87-99
100101
Chapter 1. Group Structure

1.1 Show that the set of all rotations Rz ( ) around the z axis by an angle
acting on the coordinates ( x, y ) for all values of forms a group.
SOLUTION 1.1 In an active transformation, r is rotated by an angle about
the coordinate origin to become r0 = ( x0 , y0 ) in the original frame ex , ey :
0
r 7 r = Rz ( ) r :
x0 = x cos y sin
y0 = x sin + y cos
Making use of the identities cos( a + b) = cos a cos b sin a sin b and sin ( a +
b ) = sin a cos b + cos a sin b, we obtain Rz ( ) Rz( ) = Rz ( + ) by ordinary
matrix multiplication. It follows that: (i) Rz ( + ) Rz( ) = Rz ( ) Rz( + );
(ii) Rz ( ) Rz(0) = Rz ( ), which means Rz (0) = 1 is the identity element;
(iii) Rz ( ) Rz( ) = Rz (0), which means Rz ( ) is the inverse of Rz ( ).
Since Rz ( )T Rz ( ) = Rz ( ) Rz( ) = 1, with T denoting transposition of
matrix, we have a continuous abelian group formed by a set of orthogonal
2 2 real matrices with matrix product as the group binary operation.
CHAPTER 1
1.2 Prove that a group is abelian if the order of any of its elements, except
the identity, is 2.
SOLUTION 1.2 Let a, b 6 = e be any two elements of a group G; then ab = c is
also in G. We want to prove that ab = ba. By assumption, a2 = b2 = c2 = e.
First, we have c b = ab b = ab2 = a c a = c cb = c2 b = b. And so
ba = ca a = c. Since ab = c, we have proved that ab = ba, for any a, b G.
Thus G is abelian.

1.3 Find the structure of the groups of order 4, 6, 8, and 10 with elements
all of order 2.
SOLUTION 1.3 We know from the preceding problem that any group with
all elements, other than the identity, of order 2 is an abelian group. So the
groups to be found must be abelian.
| G | = 4: G = { e, a, b, c }, with a2 = b2 = c2 = e. The product ab is
in G, but it cannot be e (because then a = b1 = b), nor can it be a or b
(because then b = e or a = e); so it must be c. We have then the four-group
V
= C2 C2.
| G | = 6: G = { e, a, . . . , c, d, . . .}. Consider the product of any two elements cd G. It cannot be e, c, or d because that would lead to contradictions (just as in the previous case), so it must be some distinct element, say
t. Then e, c, d, t form the group V. By Lagranges theorem (GTAPP 1.3), 4
must divide 6, but it does not. So an order-6 group with all order-2 elements
cannot exist.
| G | = 8: An argument similar to the case of order 4 leads to a group of
the form G = { e, a, b, ab, , a, b, ab} which is abelian and isomorphic to
the direct product V C2
= C2 C2 C2.
| G | = 10: An argument similar to the case of order 6 shows that this
group cannot exist, because 4 is not a divisor of 10.

1.4 Prove: the intersection of two subgroups of a group is also a subgroup.
SOLUTION 1.4 Let H1 and H2 be two subgroups of a group G, and H3 be the
set of common elements a, b, . . . (the intersection) of H1 and H2 . We want
to verify that H3 satisfies the group axioms. (i) Since H3 is a set in G, its
elements satisfy Gs multiplication law and associativity. (ii) The product
ab of two common elements of H1 and H2 belongs to both subgroups, and
therefore must belong to H3. (iii) The unique identity element of G is common to H1 and H2 , therefore is in H3 . (iv) The inverse a1 of any element
a H3 must be in both H1 and H2 , and therefore also in H3 . The intersection of any subgroups of a group contains at least one element in common,
the identity.

1.5 Use Cayleys theorem to find all distinct groups of order 6.
SOLUTION 1.5 It follows from Cayleys theorem that all groups of order n
are isomorphic to the regular-permutations subgroups of Sn . Since regular
PROBLEMS & SOLUTIONS
permutations, when resolved into independent cycles, must have all cycles
equal in length `, ` must be a divisor of the order n. In the present case,
n = 6 has divisors 6, 3, 2, 1.
First, suppose the group has a 6-cycle permutation p = (123456). Taking successive powers, we have p2 = (135)(246), p3 = (14)(25)(36), p4 =
(153)(264), p5 = (654321), and p6 = (1)(2)(3)(4)(5)(6) = e, which complete a period. So we have a structure isomorphic to C6 .
Next, suppose the group has no 6-cycles, then we need two 3-cycles or
three 2-cycles for each permutation. Let us start with pa = (123)(654),
which leads to p2a = (321)(456) and p3a = e. We need another generator,
which must have order 2. We take pb = (14)(25)(36), which has p2b =
e. By direct calculation, we prove that pb pa = (15)(26)(34), and pb pa =
1
p
a p b . Altogether, these relations suffice to define the group structure. It is
isomorphic to D3 .
Starting with other permutations leads to no new structures. So, up to
isomorphisms, there are just two groups of order 6, namely C6 and D3 .
1.6 Let G be a group of even order | G | that contains a cyclic subgroup H =
h a i of order | H |, such that | H | = | G | /2. Find the coset of H in G, and,
using Lagranges theorem (GTAPP 1.3), identify the structure of G for
| G | = 4, 6, 8.
SOLUTION 1.6 Since the subgroup H has order | H | = | G |/2, hence an index
k H = 2, it has only a single coset, which together with H itself gives back
G. As its left coset equals its right coset, the subgroup H is invariant in G.
We have the elements of H and its coset bH given by
H = { e = a p , a, . . . , a p1 };
def
p = | H |,
bH = { b, ba . . . , ba p1 } = { b0, b1 , . . . , b p1 }
Later we will use the notations bj+ p = bj , bj+m = bj am (with p = | H | and
j, m any integer). To solve this problem, we have to answer two questions:
What are the orders of bj for j = 0, . . . , p 1? What are the commutation
relations between the elements of G? We know, by Lagranges theorem
(GTAPP 1.3), that the order of an element is one of the divisors of | G |.
Whereas order 1 corresponds to the identity element, an order equal to | G |
is also ruled out in our case because of the presence of an element, a, of
order | G | /2. So we need to consider only divisors d such that 1 < d < | G |.
Order | G | = 4 and divisor 2. With p = 2, we simply have subgroup
H = { e = a2 , a }, coset bH = { b, ba}, and group G = { e = a2 , a, b, ba }.
Order of the elements: b2 must be one of the elements of G. It cannot
be b, because it would mean b = e. It cannot be ba, because it would mean
b = a. It cannot be a, because that would imply b4 = a2 = e, i.e. b would
have order 4. Finally, b2 = e leads to no contradictions, and this gives 2 as
CHAPTER 1
the order of b. By the same reasoning we also can show that ba also has
order 2. So we have a2 = b2 = ( ba )2 = e.
Commutation relation: ( ba )2 = e implies ba = a1 b1 = ab.
The resulting group G is isomorphic to V
= D2.
Order | G | = 6 and divisors 3, 2. Subgroup H = { e = a3, a, a2}, coset
bH = { b, ba, ba2}, and group G = { e, a, a2, b, ba, ba2}.
Order of the elements: For any j, we cannot have b2j = a or b2j = a2 ,
because in both cases, bj would have order 6. We cannot have b2j = bk (any
given j, some k), because it would mean bj would be one of the elements
e, a, a2. We are left with the only possibility, b2j = e for any j, which leads to
no contradictions.
Commutation relations: First, ( ba )2 = e means ba = a1 b1 = a2 b.
Second, ( ba2 )2 = e means ba2 = a2 b1 = ab.
G is isomorphic to D3 .
Order | G | = 8 and divisors 4, 2. Subgroup H = { e = a4, a, a2, a3 },
coset bH = { b, ba, ba2, ba3 } = { b0, b1 , b2 , b3 }, and group G = H + bH.
Order of the elements: For any j, we cannot have b2j = bk (any given j,
some k) because it would mean bj would be one of the elements e, a, a2, a3 . If
b2j = a, then b8j = a4 = e and bj has order 8. If b2j = a3, then b8j = a12 = e and
bj has order 8. All these possibilities are ruled out. Of the two remaining,
the first b2j = a2 implies b4j = e, i.e. 4th order elements; while the second,
b2j = e, or second-order elements.
If b2j = a2 for all j, we can derive from the general relation bj+m = bj am
several equivalent results:
bj bj+m = am+2 , then bj a2 = am+2 bj+m , a2 bj+m = bj am+2 . (Note: These
relations should be used only when all bj have order 4.)
If b2j = e for all j, we have bj bj+m = am , then bj am = am bj . (Note:
These relations should be used only when all bj have order 2.)
Therefore, there are three cases to be considered: All elements bj of
order 4; some bj of order 4, and some of order 2; and lastly, all bj of order 2.
. Case 1: b2j = a2, j = 0, 1, 2, 3. The canonical commutation relation is
bj a2 = a2 bj . This group is isomorphic to the quaternion group Q8 by the
correspondence a 7 k and b 7 j where j and k (together with i) are units
of the quaternion algebra (see GTAPP 1.1 and 1.4).
. Case 2: b02 = b22 = e, b12 = b32 = a2. We can verify that bj a = abj , from
which bj am = am bj for j = 0, 1, 2, 3. Hence { e = a4 , a, a2, a3, b, ba, ba2, ba3 } is
isomorphic to the direct product C4 C2 .
. Case 3: b2j = e, for j = 0, 1, 2, 3. The canonical commutation relation is
bj a = a3 bj . The group G is isomorphic to D4 .
1.7 Show that there are five distinct groups of order 8, namely, the cyclic
group C8, the group of symmetries of the square D4 , the quaternion group
Q8 , and two product groups, C4 C2 and D2 C2.
SOLUTION 1.7 A group G of order 8 contains elements of order 1, 2, 4, 8.
The only element of order 1 is the identity, now identify elements of order
8, 4, 2.
If there is at least one element of order-8, we have the cyclic group C8 .
Now assume there is no order-8 elements, but there is at least one element a of order 4. Then G contains the subgroup H = { a, a2, a3, a4 = e },
which is invariant in G with index k H = 2. The problem reduces to determining its coset bH = { b, ba, ba2, ba3}, already considered in Problem 1.6.
We have learned then that G must be Q8 , or D4 , or C4 C2 .
Lastly, if there are no elements of order 8 or 4, then all elements, except
the identity, must have order 2. From the answer to Problem 1.3, we have
learned that G is an abelian group isomorphic to D2 C2 .

1.8 Enumerate the conjugacy classes, subgroups and normal subgroups, as
well as the factor groups, of the symmetric groups S2, S3 , and S4.
SOLUTION 1.8
(a) Group S2 has two elements, e = (1)(2) and (12), forming two classes,
[ e ] and [(12)]. It has no proper subgroups, and is isomorphic to C2: S2
= C2.
(b) Groups S3 has six elements divided into 3 conjugate classes, [ e ], [(12),
(23), (31)], and [(123), (321)]. It contains four subgroups, three of which
have order 2, and one of order 3, which are listed below together with their
respective left cosets:
H21 = { e, (12)};
(23) H21 = (321) H21 = {(23), (321)}
H22 = { e, (23)};
(12) H22 = (123) H22 = {(12), (123)}
H23 = { e, (31)};
(12) H23 = (321) H23 = {(12), (321)}
(31) H21 = (123) H21 = {(31), (123)}
(31) H22 = (321) H22 = {(31), (321)}
(23) H23 = (123) H23 = {(23), (123)}

H3 = { e, (123), (321)};
(12) H3 = (23) H3 = (31) H3 = {(12), (23), (31)} .
Of these subgroups of S3 , there is just one invariant, H3

= A3 (alternating
group). The corresponding factor group S3 /H3 = S3 /A3 is isomorphic to
C2 , or to any of the order-2 subgroups H2i .
(c) Group S4 , the symmetric group of degree 4, is of order 4! = 24.
Its 24 elements are divided into 5 classes, each with a characteristic cycle
structure:
CHAPTER 1
[ e ];
[(12), (13), (14), (23), (24), (34)];
[(12)(34), (13)(24), (14)(23)];
[(123), (132), (124), (142), (134), (143), (234), (243)];
[(1234), (1243), (1324), (1342), (1423), (1432) ].
We can check the number of classes obtained by counting the ways in
which the degree n = 4 is partitioned into sums of integers:
4 = 1 + 1 + 1 + 1 = 2 + 1 + 1 = 2 + 2 = 3 + 1 = 4 + 0.
The order of a proper subgroup of S4 must be one of the divisors of 24,
that is 12, 8, 6, 4, 3, 2. We give below a partial list of the S4 subgroups:
Order 12: A4, isomorphic to the tetrahedral group T and clearly invariant in S4 since it contains elements in complete classes:
T
= { e, (12), (13), (14), (23), (24), (34),
(123), (132), (124), (142), (134), (143), (234), (243)}.
Order 8: A subgroup isomorphic to D4
D4
= { e, (13), (24), (13)(24), (12)(34), (14)(23), (1234), (1432)}.
Order 6: Four subgroups, all isomorphic to S3 and composed of permutations of any three of the four objects, leaving the fourth unchanged.
Order 4: Three subgroups isomorphic to C4, of the form { e, p, p2, p3 },
where p is (1234), or (1243), or (1324) with e = p4 ; plus one invariant
subgroup isomorphic to Kleins four-group:
V
= { e, (12)(34), (13)(24), (14)(23)}.
Order 3: Four of the form { e = p3, p, p2}, with p any 3-cycle.
Order 2: Three subgroups with two 2-cycles of the form { e, (ij)(kl )},
plus six with one 2-cycle of the form { e, (ij)}.
From the two invariant subgroups, T
= A4 and V
= D2, we construct
their cosets and the factor groups of S4 , which leads to
S4/V = { V, (12)V, (23)V, (13)V, (123)V, (321)V }, isomorphic to S3.
S4/T = { T, ( ij) T} with ( ij) any 2-cycle. S4 /T is isomorphic to C2.
1.9 The dihedral group D4 is the group of the symmetry transformations
of a square, generated by a rotation b through an angle about a diagonal, and a counterclockwise rotation a through an angle /2 about the
axis perpendicular to the square plane and passing through its center. The
generators satisfy the relations a4 = e, b2 = e, ba = a1 b.
(a) Give the multiplication table.
(b) Identify the subgroup of S8 isomorphic to D4 .
(c) Enumerate all the conjugacy classes.
(d) Enumerate all the subgroups, identifying the normal subgroups.
(e) Give the cosets of the subgroups and the factor groups associated
with the normal subgroups.
SOLUTION 1.9 (a) D4 has 8 elements, e, b, ba, ba2, ba3, a, a2, a3 , of respective
orders 1, 2, 2, 2, 2, 4, 2, 4. Their pairwise products can be derived from the
basic relations between the generators of the group: a4 = e, b2 = e, ba =
a1 b, from which we also have ab = ba1. The group table is as follows:
D4
e
b
ba
ba2
ba3
a
a2
a3
e
e
b
ba
ba2
ba3
a
a2
a3
b
b
e
a3
a2
a
ba3
ba2
ba
ba
ba
a
e
a3
a2
b
ba3
ba2
ba2
ba2
a2
a
e
a3
ba
b
ba3
ba3
ba3
a3
a2
a
e
ba2
ba
b
a
a
ba
ba2
ba3
b
a2
a3
e
a2
a2
ba2
ba3
b
ba
a3
e
a
a3
a3
ba3
b
ba
ba2
e
a
a2
(b) By Cayleys theorem, the order-8 group D4 must be isomorphic to

a subgroup of S8 , formed by the regular permutations of 8 symbols. Each
element g of D4 is mapped to such a permutation p g , as shown below (see
GTAPP 1.6 for additional details):
initially
e 7 pe
b 7 pb
ba 7 pba
ba2 7 pba2
ba3 7 pba3
a 7 pa
2
a 7 p a2
a 3 7 p a3
1
1
2
3
4
5
6
7
8
2
2
1
8
7
6
5
4
3
3
3
6
1
8
7
2
5
4
4
4
7
6
1
8
3
2
5
5
5
8
7
6
1
4
3
2
6
6
3
4
5
2
7
8
1
The eight permutations are

pe = (1)(2) . . . (7)(8) ,
pb = (12)(36)(47)(58) ,
pba = (13)(28)(46)(57) ,
pba2 = (14)(27)(38)(56) ,
pba3 = (15)(26)(37)(48) ,
pa = (1678)(5432) ,
pa2 = (17)(24)(35)(68) ,
pa3 = (2345)(8761) .
7
7
4
5
2
3
8
1
6
8
8
5
2
3
4
1
6
7
CHAPTER 1
By the rules of permutation multiplication, we obtain the following relations:

1
p4a = p2b = pe ; and pb pa = p
a pb ;
2
3
p a = p a2 , p a = p a3 ;
pb pa = pba , pb pa2 = pba2 , pb pa3 = pba3 .
They show that the eight permutations pi form a group isomorphic to D4 .
(c) One quick way of identifying conjugate elements is to place the grow and g-column, one above the other. The boxes that have the same entries in the row and the corresponding column indicate that the column
label and the row label are possible conjugates. Take for example the two
pairs of rows and columns corresponding to elements b and ba, we see that
e and a2 are self-conjugate, and a, a3; b, ba2; ba, ba3 are pairs of conjugates.
labels
row b
column b
row ba
column ba
e
b
b
ba
ba
b
e
e
a3
a
ba
a
a3
e
e
ba2
a2
a2
a
a3
ba3
a3
a
a2
a2
a
ba
ba3
ba2
b
a2
ba2
ba2
ba3
ba3
a3
ba3
ba
b
ba2
Proceeding in this way for all elements g will produce the conjugate classes,
Order-1 element : [ e ] ;
Order-2 elements: [ a2], [ b, ba2], [ ba, ba3] ;
Order-4 elements: [ a, a3] .
(d) Examining the multiplication table, we can identify the following
subgroups of D4 :
Order-2 subgroups: H21 = { e, a2}, { e, b}, { e, ba}, { e, ba2}, { e, ba3} ;
Order-4: H41 = { e, a, a2, a3}, H42 = { e, b, ba2, a2}, H43 = { e, ba, ba3, a2} .
The subgroups H21 , H4i (i = 1, 2, 3), which contain complete classes, are
the invariant subgroups in D4 .
(e) Cosets of invariant subgroups and factor groups:
Cosets of H21 = { e, a2}: This subgroup has an index equal to 4, and so
has 3 cosets: bH21 = { b, ba2}, aH21 = { a, a3}, baH21 = { ba, ba3}. The square
of each coset is the subgroup itself, which means we can map H21 7 e0 ,
aH21 7 a0 , bH21 7 b0 , baH21 7 c0 , with a02 = b02 = c02 = e0 . The factor
group D4 /H21 = { e0, a0 , b0 , c0 } is isomorphic to V.
H41 , H42 , H43 have an index equal to 2, and so have one coset each:
Coset of H41: bH41 = { b, ba, ba2, ba3}, with ( bH41)2 = H41.
Coset of H42: aH42 = { a, a3, ba, ba3}, with ( aH42)2 = H42.
Coset of H43: aH43 = { a, b, ba2, a3 }, with ( aH43)2 = H43.
In these cases the factor groups D4 /H4i are isomorphic to C2 .

1.10 Let Dn be the dihedral group generated by a of order n and b of order
2. Show that the subgroup h a i is normal in Dn .
SOLUTION 1.10 Let A = h a i = { ak; k = 0, . . . , n 1}. Each element of Dn

is of the form g = ai b j , with i = 0, . . . , n 1 and j = 0, 1, defined so that
( ab )2 = e. So we want to prove that gak g1 A. The case of j = 0 is trivial,
so we just consider j = 1.
From ( ab )2 = e, we have bab1 = bab = a1 A. We also have ak =
nk
a , where k = 0, . . . , n 1. So that:
(i) bak b1 = ( bab)k A.
(ii) ( ab ) ak( ab )1 = a ( bakb ) a1 A.
(iii) ( ai b ) ak ( ai b )1 = ai ( bakb ) ai A.
This proves that gak g1 A for every g Dn , and so A = h a i is a normal
subgroup in Dn .

1.11 The quaternion group Q8 is generated by two order-4 elements a and
b, such that ( bam )2 = a2 , with m = 0, 1, 2, 3.
(a) Give the multiplication table.
(b) The name of the group arises from the fact that its elements can be
represented by 4 objects, e.g. 1, x, y, z, and its negatives; or four independent 2 2 matrices, such as the identity I matrix and the Pauli matrices:
1 =

0 1
1 0
2 =

0 i
i 0
3 =

1 0
0 1
Making use of the relations i j = I + ieijk k , establish the correspondence

between the three representations.
(c) Identify the conjugacy classes and subgroups of Q8 . For each normal
subgroup, give the factor group of Q8 .
SOLUTION 1.11
(a) Let us define bj = ba j, where j = 0, 1, 2, 3 and bj + 4 = bj . By
assumption b2j = a2 for all j. From the general relation bj+m = bj am, we can
derive several equivalent relations:
bj bj+m = am+2 , then bj a2 = am+2 bj+m , a2 bj+m = bj am+2 .
With these relations, the group table can be constructed:
Q8
e
b0
b1
b2
b3
a
a2
a3
e
e
b0
b1
b2
b3
a
a2
a3
b0
b0
a2
a
e
a3
b3
b2
b1
b1
b1
a3
a2
a
e
b0
b3
b2
b2
b2
e
a3
a2
a
b1
b0
b3
b3
b3
a
e
a3
a2
b2
b1
b0
a
a
b1
b2
b3
b0
a2
a3
e
a2
a2
b2
b3
b0
b1
a3
e
a
a3
a3
b3
b0
b1
b2
e
a
a2
10
CHAPTER 1
(b) We note in particular the relations:

b02 = b32 = a2
b3 b0 = a2 a, b0 a = a2 b3 , ab3 = a2 b0 .
Given the properties of the Pauli matrices, the results suggest the mapping:
e 7 I, b0 7 i2 , b3 7 i1 , a 7 i3 .
Alternatively, we can also map
e 7 1, a2 7 1;
b0 7 y, b3 7 x, a 7 z;
where the new symbols satisfy the relations
x2 = y2 = z2 = 1, xy = z, yz = x, zx = y.
With the order of the elements slightly changed to make the symmetries
hidden in the table more apparent, the multiplication table of Q8 with the
new symbols becomes
Q8
1
x
y
z
1
x
y
z
1
1
x
y
z
1
x
y
z
x
x
1
z
y
x
1
z
y
y
y
z
1
x
y
z
1
x
z
z
y
x
1
z
y
x
1
1
1
x
y
z
1
x
y
z
x
x
1
z
y
x
1
z
y
y
y
z
1
x
y
z
1
x
z
z
y
x
1
z
y
x
1
(c) The group Q8 has the following characteristics:

. Order of elements: The orders of 1, 1, x, y, z are respectively 1,
2, 4, 4, 4.
. Five conjugate classes: [1], [1], [ x, x], [ y, y], [ z, z ].
. Four invariant subgroups: H2 = {1, 1}, H41 = {1, 1, x, x }, H42 =
{1, 1, y, y}, H43 = {1, 1, z, z }.
. Cosets and factor groups: The invariant group H2 is isomorphic to
the second-order inversion group V2 . Its cosets are { x }, { y }, { z }, the
square of each of which is H2 , so the factor group Q8 /H2, consisting of H2
and its three cosets, is isomorphic to Kleins group V. Each of the order-4
invariant subgroups H4i has an index equal to 2, and a single coset, whose
square is the invariant subgroup itself. Its factor group, formed by H4i and
its coset, is isomorphic to the inversion group V2 C2.
11
Chapter 2. Group Representations

2.1 (Representations in function space) In a two-dimensional Euclidean
space, define a transformation R on the Cartesian coordinates ( x, y ) by
0

x
x
= D( R )
.
0
y
y
The set of all homogeneous scalar functions of degree 2 in x, y invariant
under R form a linear vector spaceV 3 of dimension 3 spanned by the basis
vectors 1 ( x, y ) = x2, 2 ( x, y ) = 2xy and 3 ( x, y ) = y2. Given D( R ) as
one of the following matrices:

1 0
cos sin
1/2 3/2
( a)
, (b)
, (c)
,
0 1
sin
cos
3/2 1/2
find the corresponding matrix representation of R in V 3 = { 1, 2 , 3 }.
SOLUTION 2.1 First recall that a representation matrix M in representation
space V n is defined by the action of an operator ( g) on a basis { i } of V n
such that
( R)j =
i Mij ( R) .
i=1
In addition, if is a scalar function of a variable x, then it also satisfies the

relation ( R ) ( D ( R) x) = ( x ) or, equivalently, ( R ) ( x) = ( D ( R1 ) x ).
Applying it to the definition of M, we have for ( R ) j ( x ) = j0 ( x ) the
relation
j0 ( x ) = j ( x00 ) =
i ( x ) Mij ( R)
where x00 = D ( R1 ) x .
i=1
Keep in mind two points: (i) x00 = D ( R1 ) x, and not x0 , is present; (ii) each
calculation of j gives the entries of a column (not a row).
(a) In this case we have x00 = x, y00 = y, so that
00
1 ( x, y ),
10 ( x, y ) =
x 2 = x2 =
00
00
20 ( x, y ) = 2x y = 2xy = 2( x, y ), and
00
30 ( x, y ) = y 2 = y2 = 3 ( x, y ).
It follows that the representation matrix is
1 0 0
Ma ( R ) = 0 1 0 .
0 0 1
12
CHAPTER 2
(b) In this case, D( R1 ( )) = D( R ( )) and so we have x00 = cx + sy,

y00 = sx + cy, where c = cos , s = sin . To obtain the representation
matrix M, we calculate
00
( R ) 1( x, y ) = x 2 = x2 c2 + 2csxy + y2 s2 = 1 c2 + 2cs 2 + 3 s2 ,
( R ) 2( x, y ) = 2x00 y00 = 1 2cs + 2 ( c2 s2 ) + 3 2cs ,
00
( R ) 3( x, y ) = y 2 = s2 x2 2scxy + c2 y2 = 1 s2 2sc 2 + 3 c2 .
The representation matrix in V 3 is therefore
2cs
s2
c
Mb ( R ) = 2cs c2 s2 2cs .
s2
2cs
c2
(c) x00 and y00 are given by

x00
x
1/2
3/2 x
1
=
D
(
R
)
=
.
y00
y
3/2 1/2 y
The transformed basis functions are then calculated to be
1
= ( 1 6 2 + 3 3 )
4
i
1 h 1
( R ) 2( x, y ) = 2x00 y00 =
6 2 2 6 3
4
i
00 2
1h 1 2
3
( R ) ( x, y ) = y =
3 + 6 + 3 .
4
( R ) 1( x, y ) = x
00 2
From this follows the representation matrix
1
6 3
1
Mc ( R ) =
6 2
6 .
4
3
6 1
It can be checked that in every case the matrix Mi is orthogonal, MTi Mi = 1.

The reason comes from our choice
of the basis functions, and in particular
of the numerical factor in 2 = 2xy. It makes the transformed vectors in
1
2
3
the given basis
Forexample, in case (c) the three
{ , , } orthogonal.
vectors (1, 6, 3), ( 6, 2, 6) and (3, 6, 1) are mutually orthogonal.
2.2 (Block diagonalization) Find the basis that reduces the representation
matrix M( R ) of Problem 2.1 Case (c) to a block-diagonal form.
SOLUTION 2.2 The
2.1 was formed by the functions
basis used3 in Problem
1
2
2
2
= x , = 2xy and = y . We could try a basis in which one
13
vector is quadratic and completely symmetric in x,

y, that is x2 + y2 . So we
1
1
3
2
2
3
take as a new
basis the functions = ( + ) / i2, = , and =
1
3
( ) / 2. From the transformation rules for under M( R ) already
derived (Case (c) of Problem 2.1), we can check that 1 transforms into itself
and so is the basis of a one-dimensional subspace invariant under M( R ),
whereas 2 and 3 transform among themselves, and therefore constitute
the basis of a two-dimensional subspace invariant under M( R ). So M( R ) is
reduced to a block diagonal form by the similarity transformation S defined
by the new basis:
1 0
1
1
S = S1 = 0
2 0 .
2 1 0 1
By matrix multiplication, we can check that
1
0
0
1
6 3
1
1/2 3/2 .
SM( R )S1 = S 6 2
6 S1 = 0
4
0
3/2 1/2
3
6 1
2.3 (Regular representations) The regular representation of a group G is

defined by the left action of its group elements (called g, h, q, . . .):

1 : qg = h
qg = hDhg ( q ) , Dhg ( q ) =
0 : qg 6 = h
h
(where we have used group elements as indices to label matrix elements).
But it can also be defined by the right action

1 : gq = h
gq = D gh ( q ) h , D gh ( q ) =
0 : gq 6 = h
h
(a) How are D and D related when G is abelian?
(b) Find their relationship when G is a finite group.
(c) Find D, D, and X for the groups C3 and D3 .
SOLUTION 2.3 (a) If G is abelian, then qg = gq and D gh ( q ) = Dhg ( q ) =
DTgh ( q ), that is D is the transpose of D.
(b) In general, if we can find a matrix X independent of group elements
such that D = XDX1 , then we will have shown that D and D are equivalent.
Therefore, if
Dhp (q ) X pg = Xhp Dpg (q ) ,
p
then X must have elements such that Xhq,g = Xh,qg . This equation says that
two matrix entries of X are equal if the products of their row index and their
14
CHAPTER 2
column index (as group elements) are equal: ( hq ) g = h ( qg) = g0 , where g0

is the resulting group element. This suggests that we may build X by fixing
g0 (e.g. g0 = e) in the following way: Xpq = 1 if pq = e, and Xpq = 0 if
pq 6 = e. By letting g0 take all allowed values, we will have | G | possible X.
(c) To find D, D, and X, it is simplest to start from the group table and
use its labels for all the matrices in an identical manner.
Group C3 . This abelian group has three elements: e, a, b = a2 , corresponding to indices 1, 2, 3. As usual, the rows and columns of the multiplication table are both labeled in that order:
1 2 3
e 1 2 3
a 2 3 1
b 3 1 2
From this table, we obtain the (left-action) representation matrices Dij ( g)
1 0 0
D( e ) = 0 1 0 ,
0 0 1
0 0 1
D( a ) = 1 0 0 ,
0 1 0
0 1 0
D ( b ) = 0 0 1 .
1 0 0
The (right-action) representation matrices Dij ( g) are simply the transposes

of Dij ( g). A similarity transformation is given by the matrix
1 0 0
X = 0 0 1 .
0 1 0
Group D3 . The multiplication table of D3 is shown below
e
b
ba
ba2
a
a2
1
1
2
3
4
5
6
2
2
1
6
5
4
3
3
3
5
1
6
2
4
4
4
6
5
1
3
2
5
5
3
4
2
6
1
6
6
4
2
3
1
5
The nonzero matrix elements of Dhg ( q ) are found by using the group
table: Look at row q of the table, which has entries h corresponding to the
column labels g (being given in the first row), then Dhg ( q ) = 1. Similarly,
for the right action regular representation Dhg ( q ): At column q, the entries
are h all the way down the column, which correspond to the row labels
15
g, then Dhg ( q ) = 1. In this way, we obtain the following nonzero matrix

elements (all = 1), which we give in the order they appear in the multiplication table:
Dii ( e ) :
Dij ( b ) :
i = 1, . . . , 6.
( ij) = (21), (12), (53), (64), (35), (46).
Dij ( ba ) :
( ij) = (31), (62), (13), (54), (45), (26).
Dij ( ba ) :
( ij) = (41), (52), (63), (14), (25), (36).
Dij ( a ) :
( ij) = (51), (42), (23), (34), (65), (16).
Dij ( a ) :
( ij) = (61), (32), (43), (24), (15), (56).
Dii ( e ) :
i = 1, . . . , 6.
Dij ( b ) :
( ij) = (12), (21), (36), (45), (54), (63).
Dij ( ba ) :
( ij) = (13), (25), (31), (46), (52), (64).
Dij ( ba ) :
( ij) = (14), (26), (35), (41), (53), (62).
Dij ( a ) :
( ij) = (15), (23), (34), (42), (56), (61).
Dij ( a ) :
( ij) = (16), (24), (32), (43), (51), (61).
A matrix X that makes the transformation D = XDX1 is given by
1
0
0
X=
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
1
0
0
0
0
0
0
0
1
0
0
0
.
0
1
0
As X is independent of g, it defines the equivalence of the representations

D( g) and D( g) for every g G.
2.4 (Inner products) In GTAPP . 2.4 we have defined an inner product {, }

by the relation
def
{ x, y} = | G |1
h D ( h ) x | D (h ) y i,
hG
x, y V ,
where D ( G ) is a representation of a group G of order | G | in a linear space

V , and h|i is a complex scalar product. Show that {, } satisfies the scalarproduct axioms, i.e., linearity, complex conjugation, and positivity of norm.
SOLUTION 2.4 To simplify notations, we drop the overall factor | G |1.
16
CHAPTER 2
(a) Complex conjugate:
{ x, y} =
=
h D ( h ) x | D ( h )y i
hG
h D(h )y|D(h) xi = { y, x } .
hG
(b) Linearity:
{ x, ay + bz } = hG h D ( h ) x| D ( h)[ ay + bz ]i
= hG [ a h D ( h) x | D (h) yi + b h D ( h ) x| D ( h)z i]
=a
h D ( h ) x | D (h ) y i + b h D ( h ) x | D ( h )z i
hG
= a { x, y} + b { x, z} .
hG
However, note that just as for the product h x | yi, the product { x, y} is antilinear in x: { ax + bz, y } = a { x, y } + b { z, y}.
(c) Positivity of the norm:
{ x, x } =
h D ( h ) x | D (h ) x i 0 ,
hG
since each term satisfies h D ( h ) x| D ( h) xi 0. Also { x, x } = 0 if and only if

all h D ( h ) x| D (h) x i = 0, which in turn is satisfied iff x = 0.
2.5 (D3 two-dimensional representation) Group D3 can be realized in a

two-dimensional Euclidean space by the symmetry transformations that
leave an equilateral triangle invariant. Let us label the vertices of such a
triangle by 1, 2, 3 counterclockwise.
(a) To begin, choose a Cartesian basis e1 , e2 in which the origin coincides
with the triangle center, the axis y = e2 passes through a vertex , and the
axis x = e1, perpendicular to y in the conventional sense. Find the matrix
representation of the six elements of D3 . Is it unitary, is it reducible?
(b) Now take a new basis defined by a reference frame with axes u1
and u2 each passing through a vertex. Find the similarity transformation
that takes one basis to the other, Sim = h ei | um i. Find the equivalent group
representation in this new basis. Is it unitary, is it reducible?
SOLUTION 2.5 (a) In the Cartesian basis { e1, e2 }, the matrix representation is defined by ( g) e j = ei Dij ( g). A clockwise rotation of 2/3 of the
reference frame produces x 0 = ( a )e1 = 21 e1 +
23 e1
3 2
2 e ,
y 0 = ( a )e2 =
21 e2. This gives Dij ( a ). On the other hand a reflection about the y
We identify this transformation with
axis produces x 0 = x and y 0 = + y.
group element ba, and so we get Dij ( ba ). Then we have three matrices
17

D( e ) =
1 0
0 1

1/2 3/2
D( a ) =
3/2 1/2
D( ba) =

1 0
.
0 1
The other three representation matrices are obtained from matrix multiplication and group properties: D( a2 ) = D( a )D( a ), D( b ) = D( ba )D( a2),
and D( ba2) = D( ba )D( a), yielding the result

1
1 3
1
3
3
D( a2) = 12
D( b ) = 12
D( ba2) = 12
.
3 1
3 1
3 1
All six matrices are unitary, as seen by matrix multiplication: D D =
1. We have a unitary representation because the representation basis is
orthonormal. To check whether the representation is irreducible or not, we
use the relation
c |G | ,
where the equality sign is satisfied if and only if the representation is irreducible. We calculate from the above matrices
[1] = [ e ]: c1 = 1, 1 = 2;
[2] = [ b ]: c2 = 3, 3 = 0;
[3] = [ a ]: c3 = 2, 2 = 1;
so that c = 1.22 + 3.02 + 2(1)2 = 6 = | G |, which implies irreducibility. (If y = e2 passes through vertex 1 and the vertices are labeled
123 counterclockwise, we can identify D3 elements with the permutations
of the labels: a = (123), a2 = a1 = (321), b = (12), ba = (23), ba2 = (13)).
(b) Let u2 coincide with e2 , passing through vertex
1, and u1 a unit vector
through vertex 3. The bases are related by u1 = 12 ( 3e1 e2 ) and u2 = e1 .

So that we have the similarity transformation matrix S and its inverse

2/ 3 0
3/2 0
1
S=
S =
1/2 1
1/ 3 1
= S1 DS. Explicitly,
The representation in the ui basis is given by D

1 0
0 1
1 1
2
D( e ) =
D( a) =
D( a ) =
0 1
1 1
1 0

1 1
1 0
0 1
2
D(b) =
D( ba ) =
D ( ba ) =
.
0 1
1 1
1 0
This representation is not unitary because the basis being used is not orthogonal. However, it is still irreducible because the characters remain unchanged in similarity transformations. If we want to check it explicitly, we
must use the sum c 0 , where [ 0 ] designates the class of the inverse
elements of [ ]. In the present case [ 0 ] = [ ] for all , and so 0 = .
2.6 (Representation of sum of class elements) Let [ ] be a nontrivial conjugacy class of a finite group G. Prove that an irreducible representation of
18
CHAPTER 2
the sum of the group elements in [ ] is a multiple of the identity, I, and

determine the constant factor .
SOLUTION 2.6 Define q = h[] h. In any irreducible representation of
G, q is represented by D ( q ) = h[] D ( h ). Taking its trace, we get
Tr D ( q ) =
Tr D ( h ) = c ,
h[]
where c is the number of elements in [ ]. For any g G and any h [ ],

we have ghg1 [ ]. So the operator q commutes with every g G, gq =
qg. It follows
D ( g ) D ( q ) = D ( q ) D ( g )
for all g G .
By Schurs Lemma D ( q ) must be a multiple of the identity: D ( q ) = I.

Taking the trace of both sides leads to c = d , where d = TrI is the
dimension of the representation. This gives us the constant = c /d
and the representation D ( q ) = ( c /d ) I.
2.7 (Group algebra & completeness relation) (a) Consider a finite group G
and define the set C ( G ) = { gG a g g} with complex numbers a g . Show
that C ( G ) is closed under addition and multiplication of any two members.
A linear vector space that is closed under some multiplication law (of two
vectors) is called an algebra.
c
(b) Let K = i gi be the sum of all elements of the conjugacy class [ ]
of size c 1, where = 1, . . . , nc . Show that K K = c K , where c
are positive integers. In the following = 1 indicates the identity.
(c) Consider any irreducible representation D ( G ), and define, for any
conjugacy class [ ], the sum of matrices D = g[] D ( g). Show that D
commutes with D ( h ) for every h G, and therefore (why?) D = I.
Calculate .
(d) Show that = N 1 , where 1 is the character of the
identity e, and N is a number that depends only on the group, not its
representations. Give an interpretation of this result.
(e) Show that the inverses of the elements of any class [ ] form another
complete class [ 0 ], and prove that the coefficients defined in (b) are such
that c1 = c if [ 0 ]; otherwise, it is zero. From this and (b), show that
c = | G | 0 ,
where is the number of distinct simple representations. If D is unitary,

D ( g1 ) = ( D ( g)) , which implies 0 = . What does this result mean
19
about ; what does it say about the values of (the number of representations) and nc (the number of classes)?
SOLUTION 2.7 (a) Let A and B be any two members of C ( G ). Then
A + B = ( a g + bg ) g =
g
cg g = C ,
g
AB = a g bh gh
gh
= ( a g0 h1 bh ) g0 = D .
g0
We have clearly, by definition, C and D in C ( G ).

(b) K is by definition invariant under h G, that is hK h1 = K . So is
the product K K for any classes , , that is K K = h K K h1 . But in this
equation, h permutes the elements in the classes, and so K K must contain
complete classes, which means K K = c K , where c are positive

integers, characteristic of the group structure.
(c) For the same reason that we have hK h1 = K for any h G, we
also have, for any irreducible representation,
D ( h )D D ( h1 ) =
g[]
g0 []
D ( h ) D ( g ) D ( h 1 )
D ( g0 ) = D .
So D commutes with every D ( h ). Schurs Lemma tells us that D = I.

To find the constant take the trace on both sides: the left side gives c ,
whereas the right side gives d = 1 , resulting in = c /1
(d) Corresponding to K K = c K we have the matrix representa

tion D D = c D . From the result obtained in (c) we have =
c . Substituting the values for , we have
c c = 1 c c .
Thus, the result is of the form = N 1 . The product of two

primitive characters of different classes in the same representation is given
by a linear combination of the characters of all classes of that representation, the coefficients being independent of the representation, apart from
an overall factor, d = 1 .
(e) Let g, g0 be conjugates under G, i.e. hgh1 = g0 for all h G. If we
take the inverse of both sides of this equation, hg1 h1 = g01 , we see that
g1 and g01 are also conjugates of each other. If [ ] is a conjugacy class,
20
CHAPTER 2
the inverses of its elements also form a complete class, called [ 0 ], with the
same size, c0 = c . The product [ ][0] contains the identity c times, but
the product [ ][ ] of any two classes such that 6 = 0 has no element e. We
encapsulate this result in the equation c1 = c 0 . Now take the equation
given in (d) and sum both sides over = 1, 2, . . . , d :
c c = c c 1 = | G | c1c1 ,
where the second = is based on a result obtained in GTAPP 2.7. Now, we

have c1 = 1 and c1 = c 0 , and hence the relation
c = | G | 0 .
(Recall that 0 on the right-hand side indicates the class of the inverses of
the elements of the class present on the left-hand side.) This equation
holds for both unitary and non-unitary representations. If D is unitary,
we have D ( g1 ) = ( D ( g)) , which implies 0 = , and the above
equation becomes
c = | G | (for unitary representations).
This equation can be interpreted as a statement of the orthogonality of the

nc vectors , labeled by classes, in a -dimensional space. Since there can
be no more linear independent vectors than the dimension, we have nc .
Taking this result with the inequality nc derived in GTAPP 2.7, we
conclude that = nc . This proof of completeness differs from that given in
the Chapter in that it relies only on the characters; it does not depend on
the completeness of the representation matrices.

2.8 (Sums of characters) Show that the sum of the simple characters of the
elements of a finite group vanishes in any irreducible representation except
the trivial one-dimensional representation. Show also that the sum over the
simple representations, d with 6 = [ e ] and d being the representation dimensions, also vanishes.
SOLUTION 2.8 The stated results follow from the orthogonality between the
first row in the character table and every other row, and the orthogonality
between the first column and every other column. Orthogonality between
representation = 1 (1 = 1) and representation 6 = 1 gives: c =

gG ( g) = 0, 6 = 1. Orthogonality between class = 1 (1 = d ) and
any class 6 = 1 gives d = 0, ( 6 = 1). These two linear (rather
21
than quadratic) relations are useful in finding a character if all the others
are known.
2.9 (D3 representation in V 4) Let V 4 be the four-dimensional vector space
of the functions of two real variables f ( x, y) = c1 x3 + c2 x2 y + c3 xy2 + c4 y3 ,
with a basis consisting of 1 = x2, 2 = x2 y, 3 = xy2, and 4 = y3 .
It is assumed that the variables x, y are Cartesian coordinates that linearly
transform according to x0i = Dij ( g) x j (x1 = x, x2 = y, similarly for x01, x02)

0
x
x
= D ( g)
,
y0
y
where D ( g), with g = e, b, ba, ba2, a, a2, are the two-dimensional representation matrices of the group G = D3 (given in the SOLUTION 2.5).
(a) Find the representation matrices M( G ) in the basis { i } of V 4.
(b) Calculate the characters in this representation.
(c) Find the transformation that reduces M( G ) to direct sums of the irreducible representations of D3 .
(d) Find the bases of the fully reduced representation in terms of { i }.
SOLUTION 2.9 (a) Recall that representation matrix M( g) is defined by
( g) j ( x ) = j ( x00 , y00 ) =
i ( x ) Mij ( g) ,
x00 = D ( g1 ) x .
We need consider just the group generators, since the other group elements
can be obtained by combinations. We take a and ba as the group generators,
and so we need the matrices

1 1
3
1 0
1
1
2
D( a ) = D( a ) =
, D ((ba ) ) = D ( ba) =
.
0 1
2 3 1
First consider ( a ), where x00 = (1/2)( x + 3y ), y00 = (1/2)( 3x + y ).

( a ) 1 = ( x00 )3
1
= ( 1 + 3 3 2 9 3 + 3 3 4 ) ,
8
2
( a ) = ( x00 )2 y00
1
= ( 3 1 + 5 2 3 3 3 4) ,
8
3
( a ) = x00 ( y00 )2
1
= (3 1 + 3 2 + 5 3 + 3 4) ,
8
( a ) 4 = ( x00 )3
1
= (3 3 1 9 2 3 3 3 4 ) .
8
22
CHAPTER 2
Next, for ( ba ), we have x00 = x and y00 = y, and
( a ) 1 = ( x00 )3 = 1
( a ) 2 = ( x00 )2 y00 = 2
( a ) 3 = x00 ( y00 )2 = 3
( a ) 4 = ( x00 )3 = 4 .
The representation matrices can be read off from these results:
3 3 3
1 0 0
1 3
0 1 0
1
3
5
3
9
3

M( a ) =
M( ba ) =
0 0 1
8
9
3
5
3
3
0 0 0
3 3 3
3
1
0
0
0
1
Of course M( e ) = diag [1, 1, 1, 1] is the usual diagonal matrix.

(b) As we know, D3 has three conjugacy classes, namely, [1] = { e }, [2] =
{ b, ba, ba2}, and [3] = { a, a2}, with respective element numbers c = 1, 3, 2.
Note that every class coincides with the class of the inverses of its elements,
i.e. [ 0 ] = [ ]. Their characters in this representation are given by the
traces of the matrices M( g): = (4, 0, 1) on the conjugacy classes. To check
whether this representation is irreducible or not we calculate the sum
c 0 = 1 42 + 3 0 + 2 12 = 18 .
This sum being greater than | G | = 6, the representation is reducible.

We further know (GTAPP 2.3) that D3 has three irreducible representations of dimensions (1,1,2) and characters 1 = (1, 1, 1), 2 = (1, 1, 1),
and 3 = (2, 0, 1). We can then apply the formula
a = | G |1 c
to obtain the multiplicities for the representation M( G ), which are a = 1

for every . Therefore, M can be reduced to the direct sum D ( G ).
(c) The question is to find a matrix S independent of g such that
1 0 0 0
( ba ) = SD( ba )S1 = D ( ba ) = 0 1 0 0
D
0 0 1 0
0 0 0 1
and
( a ) = SD( a )S1
D
0
= D ( a ) =
0
0
0
0
0
1
0
0 .
0
1/2 3/2
0
3/2 1/2
23
( g)S, the equaWith these transformation equations written as SD( g) = D

tion for g = ba requires S to be of the form
a 1 0 c1 0
0 b1 0 d1
S=
a 2 0 c2 0 .
0 b2 0 d2
The equation for g = a imposes further restrictions on Sij , leaving three elements undetermined, which can be arbitrarily chosen, leading to the matrix
S that we give below together with its inverse:
1 0 1 0
1 0 1 0
1 0 1 0 1
S1 = 1 0 3 0 1 .
S=
0 1 0
0 1 0
2 3
2 3
0
1 0 3
0
1 0 1
( g) = SD( g)S1 are expressed

(d) The reduced representation matrices D
j
i
in the transformed basis functions = ( S1 )ij , which are

1 = 1 + 3 3 = x ( x2 3y2 ) ,
2 = 3 2 + 4 = y (3x2 y2 ) ,
3 = 1 + 3 = x ( x2 + y2 ) ,
4 = 2 + 4 = y ( x2 + y2 ) .
Using the transformation rules derived in (a), one can prove that 1, 2,
and { 3, 4 } span invariant subspaces, two of dimension 1 and one of dimension 2, which are the spaces of the irreducible representations of D3 .
Remarks: It is useful to note the method applied in (c) to find a similarity transformation matrix: One should start with an element g whose
( g) is diagonal, then proceed to the next simplest matrix
representation D
0
( g ). Proceeding in this order simplifies your work. In addition, if S1
D
is eventually needed, but not S itself, it would be wiser to determine the
saving
inverse (rather than S) directly from the calculation DS1 = S1 D,
you from extra work.
2.10 (Character table for D4 ) Construct the character table for D4 (the group
of symmetries for the square) studied in Problem 1.10.
SOLUTION 2.10 The dihedral group D4 has 8 elements divided into 5 conjugacy classes: [1]1 = { e }, [2]1 = { a2}, [3]2 = { a, a3}, [4]2 = { b, ba2}, and
[5]2 = { ba, ba3} (with sizes indicated as subscripts). As nc = 5, there are
5 distinct irreducible representations of dimensions d that satisfy the condition d2 = 8, which means d = (1, 1, 1, 1, 2) on the simple representations. Besides an invariant subgroup of order 2, which does not concern
24
CHAPTER 2
us, D4 has three invariant subgroups of order 4. The factor groups derived
from them are:
D4 /H 1 = { H 1, bH 1 } = {[1], [2], [3]; [4], [5]} ,
D4 /H 2 = { H 2, aH 2 } = {[1], [2], [4]; [3], [5]} ,
D4 /H 3 = { H 3, aH 3 } = {[1], [2], [5]; [3], [4]} ,
all of which are isomorphic to C2. Knowing the characters of the cyclic
group, we can now determine the characters of the one-dimensional representations of D4 . For example, the mapping H 1 7 e C2 and bH 1 7 c
C2 shows that 21 = 22 = 23 = 1 and 24 = 25 = 1. Finally, the characters
of the two-dimensional representation, calculated with the orthogonality
and completeness relations, allow us to complete the character table.
D4 [1]1 [2]1 [3]2 [4]2 [5]2
1
2
3
4
5
1
1
1
1
2
1
1
1
1
2
1
1
1
1
0
1
1
1
1
0
1
1
1
1
0
2.11 (Character table of T

= A4) T is the group of rotations that leave a
regular tetrahedron invariant (this four-sided pyramid has 4 three-fold axes
and 3 two-fold axes, found in C H4 for example). Using the group structure
and the orthogonality and completeness relations, construct the character
table for T. Note that T is a subgroup of S4, isomorphic to the alternating
group A4 (see Problem 1.8).
SOLUTION 2.11 T is a subgroup of S4 with 12 elements, divided into 4
classes:
[1] = { e } (c1 = 1),
[2] = {(12)(34), (13)(24), (14)(23)} (c2 = 3),
[3] = {(123), (142), (134), (243)} (c3 = 4)
[4] = {(132), (124), (143), (234)} (c4 = 4)
It has an invariant subgroup: N = { e, (12)(34), 13)(24), (14)(23)} consisting of all elements of classes [1] and [2]. The corresponding factor group
of T is T/N = { N, (123) N, (321) N }, isomorphic to C3 .
Since there are 4 classes, there are also 4 inequivalent simple representations, and the equation d2 = 12 admits the unique solution d =
(1, 1, 1, 3): there are 3 representations of one-dimension, and 1 rep of three
dimensions. To obtain the characters for the mentioned one-dimensional
hom
iso
irreps ( = 1, 2, 3), we may use T 7 T/N 7 C3 , that is, e, g [2] 7 N;
g [3] 7 (123) N; g [4] 7 (321) N; and known results for C3 .
25

The partially filled table is ( = exp (i2/3))
[1]1 [2]3 [3]4 [4]4
1
42 ?
43 ?
44 ?
The characters for = 4 are then determined by d = 0 with =

2, 3, 4. This completes the table. Alternatively, the permutational representation P on 4 letters for T has the character P = (4, 0, 1, 1); since P 1
has | |2 = 1, it is simple, and so must be
= 4 .
e
1
2
3
4
T
:1
:2
:3
:4
[1]1
1
1
1
3
(12)(34) (123) (132)

[2]3
[3]4
[4]4
1
1
1
1
2
1
2
1
0
0
2.12 (Character table for S4) Find the characters for the group S4 .
SOLUTION 2.12 The 24 elements of S4 are divided into 5 classes:
[1] = { e } (c1 = 1),
[2] = {(12), (13), (14), (23), (24), (34)} (c2 = 6),
[3] = {(12)(34), 13)(24), (14)(23)} (c3 = 3),
[4] = {(123), (132), (124), (142), (134), (143), (234), (243)} (c4 = 8),
[5] = {(1234), (1243), (1324), (1342), (1423), (1432)} (c5 = 6).
Since there are 5 classes, there are also 5 irreducible representations,
whose dimensions must satisfy d2 = 24. This equation admits the
unique solution d = (1, 1, 2, 3, 3).
S4 has two invariant subgroups, T (order 12) and V (order 4).
T contains the complete classes [1], [3], [4], and its coset (12) T contains
all the elements of [2], [5]. The factor group S4 /T = { T, (12) T } is isomorphic to C2 . This allows us to determine the characters of the 1-dimensional
representations
1 = (1, 1, 1, 1, 1)
2 = (1, 1, 1, 1, 1) .
The second invariant subgroup of S4 is V {[1], [3]}. The associated
factor group of S4 is S4 /V = { V, (12) V, (23)V, (13)V, (123)V, (321)V }, iso-
26
CHAPTER 2
morphic to D3 . Recalling that D3 has 3 classes, which we call [ e ] D, [ b ] D and

[ a ] D, we have the isomorphic mapping S4/V 7 D3 :
[ 1 ] [ 3 ] 7 [ e ] D, [ 2 ] [ 5 ] 7 [ b ] D, [ 4 ] 7 [ a ] D .
Since the 2-dimensional representation of D3 has the character (2, 0, 1)
on the conjugacy classes [ e ] D, [ b ] D and [ a ] D, we have for symmetric group
S4 the character 3 = (2, 0, 2, 1, 0).
The remaining characters, in representations = 4, 5, are then determined from d = 0 ( 6 = 0) and | |2 = 24/c. The results are
tabulated below, with a representative of each class given on the top line.
Table 2.1: Character Table of S4
1
2
3
4
5
S4
:1
:2
:3
:4
:5
e
[1]1
1
1
2
3
3
(12) (12)(34) (123) (1234)

[2]6
[3]3
[4]8
[5]6
1
1
1
1
1
1
1
1
0
2
1
0
1
1
0
1
1
1
0
1
2.13 Find the permutational representation on 4 letters for S4 ; find its decomposition. What is the product 2 4 (where 2 is a nontrivial onedimensional representation and 4 a three-dimensional simple representation of S4 )? On the basis of these results, show a way to obtain the characters for S4 .
SOLUTION 2.13 The character of P is P = (4, 2, 0, 1, 0) on the conjugacy
classes. As P = 1 + 4 (from the character table for S4 ), we have P
=
1 + 4 . On the other hand we have 2 4 = (3, 1, 1, 0, 1), of norm 1, and
so 2 4 is simple and has dimension 3. It is equivalent to 5 .
These results suggest another way of obtaining the character table for
S4. Once the two one-dimensional representations 1 and 2 are known,
we obtain 4 = P 1 , then 5 = 2 4 . The remaining 3 is determined
by orthogonality.
2.14 (Matrix representation of product group) Suppose that a finite group
G is the product group of its two subgroups, G = H K. Using the matrices and not the characters, show that the direct product D ( H ) D ( K ) of
two simple representations of H and K is a simple representation of G.
SOLUTION 2.14 By definition, hk = kh for any h H any k K. Let L1
be the representation space of H in which the irreducible representation
27
D ( H ) is defined, and L2 the representation space of K for the irreducible

representation D ( K ). We assume that L1 and L2 do not have non-trivial
invariant subspaces for H and K. Let us proceed in three steps.
(i) First we want to prove that D ( H ) D ( K ) is a representation of G.
Let h, h0 H and k, k 0 K. If we have the correspondence from hk, h0k 0 G
to reps D ( hk ) = D ( h ) D ( k ) and D ( h0 k 0 ) = D ( h0 ) D ( k 0 ); then we
have the correspondence from hkh0 k 0 = hh0 kk 0 G to reps
D ( hk ) D (h0k 0 ) = D ( h ) D ( k ) D ( h0 ) D ( k 0 )
= D ( h ) D ( h 0 ) D ( k ) D ( k 0 )
= D ( hh0 ) D ( kk 0)
= D ( hh0 kk 0).
So the set D ( hk ) forms a representation of G, to be denoted by D ( G ) =

D ( H ) D ( K ), in the representation space L = L1 L2 . If x L1 and
y L2 , then x y is in L.
(ii) We next want to show that D ( G ) is irreducible, which is equivalent
to show that the only possible invariant subspace in L is a trivial subspace.
If there is a nonzero invariant subspace L0 in L with respect to D ( G ),
then L0 = { x y } must contain all vectors of L1 y because D ( H )
D ( e ) D ( G ). Furthermore L0 must also contain all vectors in L1 L2
because D ( e ) D ( K ) D ( G ). So L0 and L1 L2 coincide, which implies
L0 = L.
(iii) If either D or D is replaced by a different inequivalent irreducible
representation, then the irreducible representation of the direct product is
also changed to another distinct irreducible representation. The number
of classes of the direct product G is given by the product of the numbers
of classes of the subgroups H and K, and is equal to the product of the
numbers of distinct irreducible representations of the subgroups. So each
irreducible representation of the direct product H K is a direct product
of irreducible representations of the two subgroups, D ( H ) D ( K ).
Chapter 3. Lie Groups & Lie Algebras

3.1 (Rotations, O(2), SO(2)) A rotation R ( ) in the plane R2 about a reference origin through an angle , is defined by a real matrix Rij , such that

cos sin
i
j
R( ) e = e R ji ( ) ,
R( ) =
sin cos
where e1 , e2 are the reference orthonormal unit vectors in R2 .
(a) Show that the set hR( )i form a Lie group: specify its product law,
identity element, and inverse element. What is the range of values of ?
28
CHAPTER 3
(b) The group O(2) is defined as the set of 2 2 real nonsingular matrices A obeying the condition AT A = I. What physical situations do the
two possibilities, det A = 1, correspond to? Define SO (2) = {A O (2) :
det A = 1}. Show that there is a one-to-one correspondence between rotations in a plane and SO(2) matrices.
SOLUTION 3.1 (a) With usual matrix multiplication, we verify R ( 0 ) R ( ) =
R ( 0 + ), R1 ( ) = R ( ) and R (0) = I (the 2 2 identity matrix), so that
the continuous set { R ( )} satisfies the group axioms for any real value of .
In addition, although and + 2n with any integer n, correspond to different group elements, they represent the same transformation; therefore
we require on geometrical and physical grounds that R ( + 2n ) = R ( ).
Then, in this parameterization, the underlying one-dimensional manifold
is defined by the real variable in the interval 0 < 2. The composition function ( 0, ) = 0 + is analytic in both arguments. So we have a
one-dimensional abelian Lie group. The range of values of is determined
by the product rule: if the range is smaller, there would be rotations whose
product is outside the range; if it is greater, there would be rotations given
more than once. The group parameter space is the entire unit circle.
(b) From the condition AT A = I that the 2 2 real matrices of O(2)
must satisfy, we have (det A )2 = 1, which implies det A = 1 or det A =
1. When A acts on a two-dimensional coordinate space, it represents all
transformations that conserve the scalar product of any two real vectors:
a pure rotation if det A = 1; and a rotation times a mirror reflection if
det A = 1. Group O(2) is a mixed (continuous and discrete) group.
An arbitrary matrix A of O(2) has the form:

a b
A=
a, b, c, d R.
c d
The condition AT A = I yields: (i) a2 + c2 = 1, (ii) b2 + d2 = 1, (iii) ab + cd =
0. From (i) we have a = cos and c = sin ; from (ii), d = cos and
b = sin ; and finally from (iii), sin ( + ) = 0 or = n , for any
integer n = 0, 1, . . .. So that, in general, A is one of the two forms

cos sin
cos sin
A=
,
A=
.
sin cos
sin cos
So O(2) is specified by a continuous parameter, 0 < 2, and a discrete
number, taking two values 1. It represents rotations in the plane adjoined
to mirror reflections. The special orthogonal group SO(2) contains only
matrices A with det A = 1, parameterized by , and represents proper
rotations in the plane. SO(2) is connected to the identity, A 7 1 as 7 0,
and is infinitely connected since there are an infinite number of operators
29
obeying the identity R (2n ) = exp(i2n ) = 1 for all integer nthere are
closed paths on the unit circle which wind around it n times, for any integer
n, and which cannot be continuously deformed into each other.

3.2 (O(2), SO(2), and U(1)) (a) Find the conjugacy classes of O(2). (b) Prove
that the transformations that leave the magnitude of complex numbers invariant form a one-parameter Lie group (called U(1)) isomorphic to SO(2).
SOLUTION 3.2 (a) O(2) contains a Lie subgroup, namely SO(2). In addition,
I = diag[1, 1], I1 = diag[1, 1], I2 = diag[1, 1], Ir = diag[1, 1] are all
elements of O(2), but I1 and I2 are not in SO(2). If R is in SO(2), then, for
any g O (2), so is gRg1 because det gRg1 = det R = 1, so SO(2) is an
invariant subgroup. Take any A O (2) with det A = 1 (and so clearly
not in SO(2)), then det( I2 A ) = 1 which means I2 A = R for some R SO (2).
It follows A = I2 R. In conclusion any A O(2) is either R SO (2) or I2 R,
and O(2) is a disjoint union of { R : R SO (2)} and { I2 R : R SO (2)}.
These are the cosets making up the quotient group O(2)/SO(2) isomorphic
to C2
= h I, I2i.
(b) Consider the transformation z 7 z0 = z, where z, z0 are complex
variables and a complex-valued parameter. The invariance | z0 | = | z |
is satisfied if and only if | | = 1. This condition is equivalent to =
exp(i ), for real . These transformations form an abelian group because we
have the multiplication rule U ( 1 )U ( 2 ) = U ( 1 + 2 ), the identity element
U (0) = 1, and the inverse element U ( )1 = U ( ). To each value of the
parameter, there must correspond a single group element, which implies
exp[i( + 2n )] = exp (i ) for any integer n. It follows that 0 < 2. If
the range were smaller, there would be group transformations whose product is outside the range; if it were greater, there would be transformations
given by more than one value within the range.
Since this group, called U(1), has the same number of parameters defined on the same range of values as SO(2) it must be isomorphic to it.
3.3 (U(2) and SU(2)) (a) Find the general parameterization of the U(2) and
SU(2) matrices. (b) In a complex vector space C2 , the inner product is defined by h x, y i = 2i=1 xi yi . Show that the group of linear transformations
in C2 that preserve the inner product is an U(2) group.
SOLUTION 3.3 (a) An arbitrary complex matrix A of U(2),

a b
A=
a, b, c, d C,
c d
satisfies the unitarity condition A A = I, or explicitly : (i) | a |2 + | c |2 = 1,
(ii) | b |2 + | d |2 = 1, (iii) a b + c d = 0. Eq. (i) has the solution
a = cos ei ,
c = sin ei
30
CHAPTER 3
where 0 /2 and 0 , < 2. From (ii) we have

b = sin ei,
d = cos ei .
With these expressions, (iii) becomes

cos sin ei() = sin cos ei() .
Equating the magnitudes of the two sides, we obtain sin ( ) = 0, which
has the only solution = in the allowed ranges of values of , . Equating
the phases of the two sides, we have = or + = + =
2 (modulo 2 ). This leads to the restrictions on the parameters:
= + ,
= + ,
= ,
= ,
(all modulo 2), where , , are arbitrary real phases. Therefore, the general unitary matrix A U (2) has the form

i
sin ei
i cos e
A=e
.
sin ei cos ei
The four essential parameters have the ranges of values: 0 /2,
0 < , and 0 , < 2.
SU(2) is the subgroup of U(2), with det A = 1. This condition is satisfied
if and only if ei = 1, or = 0. A general matrix U SU (2) is specified by
three parameters, as in the expression

cos ei sin ei
U=
.
sin ei cos ei
(b) Let x, y be vectors in a two-dimensional complex vector space C2 . A
linear operator T in C2 acts on x, y to give x0 = Tx, y0 = Ty. Then the
condition that the inner product be preserved under T means h x0 | y0 i =
h x | T T| y i = h x | yi. It is satisfied if and only if T T = I. That is, the
group of linear transformations that preserve the inner products in C2 is
a U(2) group. The special unitary group SU(2) arises when we require
that the volume element in C2 be also preserved: dx01 dx02 = dx1 dx2; as
dx01 dx02 = | ( x01 x02 ) / ( x1, x2 )|dx1 dx2, this implies det T = 1 (where xi are
the coordinates of x, and Tij = x0i /x j).

3.4 (O(1;1) and SO(1;1)) Write down the general matrices in O (1; 1) and
SO (1; 1).
SOLUTION 3.4 A matrix A in O (1; 1) on a space with metric g =
a diag
[1, 1]
T
b
is real, and satisfies the relation A gA = g. Writing A = c d , we get
31
a2 c2 = 1, d2 b2 = 1, and ab cd = 0. The solution to each of the

first two equations is a hyperbola in the plane ( a, c ) and ( d, b ) respectively:
a = cosh t, c = sinh t, d = cosh s, b = sinh s, where , , , are the
signs. The third equation ab cd = 0 requires s = t, and determines
the possible signs. It turns out that A may be in one of the four forms:

ch t sh t
ch t sh t
ch t sh t
ch t sh t
,
,
,
,
sh t ch t
sh t ch t
sh t ch t
sh t
ch t
where ch = cosh, sh = sinh, and t is a real number in ( , ). Matrices of the first two forms have det A = +1; those of the last two forms
have det A = 1; they compose the four disjoint components of the group
O (1; 1). Matrices A with determinant equal to one form the group SO (1; 1)
which consists of two disjoint components: one with c = cosh t is connected
to the identity (this identity component is associated with proper Lorentz
transformations rotations plus boosts), and the other with c = cosh t (it
is associated with time-reversal transformations).
3.5 (Dimension of orthogonal groups) Find the dimension of the orthogonal Lie groups O(n, R) and SO(n, R).
SOLUTION 3.5 O ( n ) = O ( n, R ) is the group of all real n n matrices A
satisfying the orthogonality condition AT A = I, which can be written in
terms of the matrix elements as k aki akj = ij , where i, j = 1, 2, . . . , n.
Taking i = j these relations give n distinct conditions; and with i 6 = j
they give ( n2 n ) /2 additional conditions. So we have n + ( n2 n ) /2 =
n ( n + 1) /2 distinct conditions on the n2 matrix elements. There remain
n2 n ( n + 1) /2 = n ( n 1) /2 independent matrix elements. So the dimension of the Lie group O ( n ) is n ( n 1) /2.
AT A = I implies det A = 1. To select either possibility does not
impose a new condition. So SO ( n ) = SO ( n, R ), with det A = 1, has the
same dimension as O ( n ), which is n ( n 1) /2.

3.6 (Dimension of unitary groups) Find the dimension of the unitary Lie
groups U(n, C) and SU(n, C).
SOLUTION 3.6 U ( n ) = U ( n, C ) is the group of all n n complex matrices A

satisfying the unitarity condition A A = I, which can be written in terms
of the matrix elements as k aki akj = ij , where i, j = 1, 2, . . . , n. Setting
i = j, we have n real relations; requiring i 6 = j we have ( n2 n ) /2 complex
relations or equivalently 2( n2 n ) /2 real conditions. So we have a total of
n2 real functional relations among the 2n2 matrix elements. The number of
independent parameters is 2n2 n2 = n2 : this gives the dimension of the
group U ( n ). Note that for any A U ( n ), we have det A = ei , where is a
real phase.
32
CHAPTER 3
SU ( n ) is a subgroup of U ( n ) restricted by det A = 1. This additional

constraint eliminates one parameter from U(n) by fixing the phase parameter, e.g. = 0. So SU(n) is of dimension ( n2 1).
3.7 (Dimension of symplectic groups) (a) Show that Sp ( n, F ) contains also

the inverse and the transpose of every one of its elements. Find the dimension of Sp(n, R). (b) Same questions for the compact symplectic group
Sp ( n ). Assume F = R or C.
SOLUTION 3.7 (a) The metric J for symplectic groups is such that J 2 = I2n
and J T J = I2n , where I` = diag[1, 1, . . . , 1] with ` entries.
If S Sp ( n, F ), then ST JS = J, which may be rewritten as JS1 = ST J,
or still ( ST )1 JS1 = J ( S1 )T JS1 = J (where ( AT )1 = ( A1 )T for
any invertible matrix). This shows if S Sp ( n, F ), then S1 Sp ( n, F ).
Noting S1 = JST J (using J 2 = I), we rewrite ( S1 )T JS1 = J as
J = ( JST J )T J ( JST J ) = J T S JST J
= S JST = J
= ( ST )T JST = J .
So if S is in Sp ( n, F ), then so is ST .
Contrary to the convention used in GTAPP, we take the m = 2n indices
2, 2,
. . . , n, n (a may be either or with
in the following order: a = 1, 1,
1 n) and define J as follows:
Jab
1 :
=
1 :
0 :
0
1
J=
0
0
0
a = , b =
b=
a = ,
otherwise
1 0
0 0
0 0
0 1
0 0
0
0
1
0
0
0
0
0
0
..
0
0
0
0
..
..
..
..
..
..
(J has the same properties as before: J 2 = I and J T J = I.) Then writing

out ST JS = J in terms of the components S = [ sab ], we get
ST JS = J
[sa sb sa sb ] = Jab .
Now assume F = R. When a = b, Jaa = 0, we have an identity leading

to no conditions. But when a 6 = b, there are ( m2 m ) /2 real relational
functions among the m2 real matrix elements. Therefore, the number of
independent parameters are m2 ( m2 m ) /2 = m ( m + 1) /2, where m =
2n. So the dimension of Sp(n, R) is n (2n + 1).
(b) We can prove as in (a) that (for U = U 1 ) U 1 and U T satisfy
T
U JU = J if U does. Note also: U = U 1 = JU T J, from which U =
JU J.
33

From U = JU J, we have the following relations:
u = u ,
u = u
.
(# )
Each gives 2n2 real constraints; so together they give 4n2 real constraints.
Next, unitarity, c uca ucb = ab , imposes the following:
u = 1. Whether a = or a = ,
we have
(i) a = b: ua ua + ua
a
the same set of n equations (to see this, we use Eqs. (#) So this case yields n
independent real constraints.
(ii) a = , b = 6 = : u u + u
= 0. These are n ( n 1 ) /2
u
complex constraints. They are the same as the relations obtained with a =
and b = 6 = as well as the relations obtained by exchanging a and b
(again use Eqs. (#)).
u u + u
(iii) a = , b = 6 = :
u = 0. Again, n ( n 1 ) /2
complex constraints. The indices a = , b = 6 = yield no new conditions.
Recalling that one complex condition is equivalent to two real conditions, we have the total number of real constraints given by
2 2n2 + n + 2 2 12 n ( n 1) = 6n2 n .
As the number of real parameters is 2(2n )2 = 8n2 , the number of independent real parameters in Sp(n) is 8n2 (6n2 n ) = n (2n + 1). So the
dimension of the Lie group Sp ( n ) is n (2n + 1).
3.8 (Product law) A Lie group is defined in terms of the complete set of its
generators X , obeying the brackets [ X , X ] = c X . Find the product
law for its elements exp( X ) and exp ( X ) up to the total third order of
, .
SOLUTION 3.8 The product of two elements defined by exponential maps
must be an exponential map given by exp( X ) exp ( X ) = exp( X ).
Using the CH formula given GTAPP 3.5 and the bracket relation between
the generators, we obtain for the product element exp ( X ):
1
1
= + + c + c c ( ) + . . . .
2
12
3.9 (Identities with matrix exponentials) (a) Prove eX eY = eX+Y for arbitrary n n commuting matrices. (b) Prove exp ( SXS1 ) = S exp( X) S1 for
arbitrary X M( n, F ) and some invertible matrix S M( n, F ).
SOLUTION 3.9 (a) Use the power series for eX and eY , regroup terms of
equal total powers in X, Y, and use the binomial formula (you may assume
34
CHAPTER 3
XY YX):
X2
Y2
+ XY +
+...
2!
2!
X m k Y k
( X + Y )m
=
= eX+Y .
( m k ) !k! m=0
m!
eX eY = I + ( X + Y ) +
m =0 k =0
(b) We have S (1 + X + 21 X2 + . . .) S1 = 1 + SXS1 + 21 SX2 S1 + . . .,

and this equals exp ( SXS1 ) because SXm S1 = ( SXS1 )m .

3.10 (Polar decomposition of a matrix) (a) Prove that for any A SL ( n, R )
there is a unique decomposition A = RS, where R SO ( n ), and S a symmetric positive definite matrix of determinant one. (b) Given A SL (2, R ),
determine R and S.
SOLUTION 3.10 (a) If A = RS with R SO ( n ), then AT A = ST RT RS = ST S
since RT R = I. Requiring ST = S, we have AT A = S2 . If S is positive
definite (h x, Sx i > 0 for x 6 = 0), then S may be regarded as the length
of A. The condition det S = 1 follows automatically from det A = 1 and
det R = 1.

(b) Let A = where det A = = 1, and R = sc cs where
2
2
det
1. Then define S = R1 A = RT A, or explicitly S =
R = c +s =
c + s c + s
. So that we get det S = ( c2 + s2 )( ) = 1, while
c s c s
ST = S produces c ( ) + s ( + ) = 0, which yields s/c = ( ) / ( +
). If we let c = cos and s = sin , then = tan1 ( ) / ( + ). Thus,
R and S are completely determined in terms of A.

3.11 (Nilpotent matrices) Find the lowest power for which the following
matrices A, B, C nullify; and compute e A , e B and eC .

0 a
A=
,
0 0
SOLUTION 3.11
0
C2 =
0
0
0
0
0
0
0
0 a b
0
B = 0 0 c , C =
0
0 0 0
0
a
0
0
0
b c
d e
.
0 f
0 0
0 0 ac
(a) A2 = 0. B2 = 0 0 0 , and B3 = 0.
0 0 0
ad ae + b f
0 0 0 ad f
0 0 0 0
0
df
3
4
,
C
=
0 0 0 0 , and C = 0
0
0
0
0
0 0 0 0
35

1 a
.
0 1
1 a b + 12 ac
.
eB = I + B + 12 B2 = 0 1 c
0 0 1
1 a b + 12 ad c + 12 ( ae + b f ) + 61 ad f
0 1 d
e + 12 d f
eC = I + C + 12 C 2 + 16 C 3 =
0 0 1
f
0 0 0
1
(b) e A = I + A =

3.12 (Computing eX ) Compute eX for the following real matrices X:

0 a
0 a
a b
a b
a 0
1 1
,
,
,
,
,
,
a 0
a 0
0 a
b a
b a
0 0
SOLUTION 3.12 We shall use matrices obeying J 2 = I, K 2 = I, N 2 = 0:

1 0
0 1
0 1
0 1
I=
,J=
,K=
,N=
.
0 1
1 0
1 0
0 0

0 a
(a) X =
= a J. Summing separately even and odd powers of X,
a 0
and using J 2 = I, we get:
eX = I (1 + 12 a2 + ) + J ( a +
1 3
a + . . .) = I cosh a + J sinh a.
3!

0 a
= aK. Summing separately even and odd powers of
(b) X =
a 0
X, and using K 2 = I, we get:
eX = I (1 12 a2 + . . .) + K ( a

1 3
a + ) = I cos a + K sin a.
3!

a b
(c) X =
= aI + bN. Remarking that N I = I N and N 2 = 0, we
0 a
have
a

a

e
0 1 b
e bea
X
aI+bN
aI bN
= ea ( I + bN ).
e =e
=e e =
=
0 ea 0 1
0 ea

a b
(d) X =
= aI bK. Using IK = K I and results from (b):
b a

cos b sin b
X
aIbK
a
e =e
=e
= ea ( I cos b K sin b )
sin b cos b
36
CHAPTER 3

a 0
= ( aI + bN )T . Using results from (c), we get
b a
a

e
0
X
T
T
e = exp( aI + bN) = (exp( aI + bN )) =
= ea ( I + bN T ).
bea ea

1 1
(f) X =
. As X is idempotent (X2 = X), we have
0 0

1
1
e e1
X
e = I + X (1 + + + ) = I + X (e 1 ) =
.
0
1
2 3!
(e) X =

3.13 (Is given Y equal e with real X?) Are the following (real) matrices of
the form eX ? If so, compute X.

1 0
0 1
a 0
0 1
1 1
1 1
,
,
,
,
,
,
0 0
1 0
0 b
1 0
0 1
1 1
X
SOLUTION 3.13 Recalling that det exp( X) = exp(TrX), we know we must

have det eX> 0 for
real X.
1 0
(a) Y =
. Since det Y = 0, Y cannot be of the form eX .
0

0
(b) Y =
1

a
(c) Y =
0
0

1
. We have det Y = 1, and so Y 6 = eX .
0

0
. If a, b are positive real then = log a, = log b, and
b
Y = e where
= diag[ , ].
0 1
(d) Y =
. We have seen in Problem 12 (b) that, with X = aK,
1
eX
= I cos a + K sin a. In order to have Y = K, we fix a = /2, which gives

us X = ( /2
) K.
1 1
= I + N . But I + N = eN , since N 2 = 0; so Y = eN .
0 1

1 1
(f) Y =
= I K . Referring to Problem 12 (d), we know that if
1 1
(e) Y =
b = /4, then

ea
ea
1 1
a
/4
X
= Y.
X=
= e =
/4
a
2 1 1
2
1

log 2 /4
So we have Y = 2eXa = eZ , where Z = 2
.

/4 21 log 2
3.14 (Dimension-three algebras) Find the generators of rotation acting on
(a) three-dimensional Euclidean space of metric g = diag[1, 1, 1]; and (b)
three-dimensional space of metric g = diag[1, 1, 1].

37
SOLUTION 3.14 To determine the generators of a Lie group, we need to consider its properties near the identity I, so that we write A I + M for any
group element A, and define the generators X by M = + X , where
are the independent essential parameters in the group space. (NB: This
definition of X differs in sign, for all , from the definition in GTAPP Chapter 3; the algebras for the two options are isomorphically equivalent under
X 7 Xa .) Under the action of A, the Cartesian coordinates transform
i . This is equivalent
as x0i = Aij x j = xi + dxi, where dxi = Mij x j = Mx
= i dxi ( /xi ) (where M
is now an operator on the coordinates).
to M
Writing in terms of X , we have
X = dxi xi
= X = +
(dxi)
.
xi
(Here again, there is a sign difference with GTAPP, as explained above.)

(a) Let A SO (3), so that AT A = I and det A = 1. Writing A I + M,
we have MT = M (antisymmetric matrix). Following are the general
expression for M and the variations dxi:
0
c b
dx = cy bz
M = c 0
a
dy = cx + az
b a 0
dz = bx ay
Using the formula for X given above, we get

Xa = z
y ,
y
z
Xb = x
z ,
z
x
Xc = y
x ,
x
y
which bracket according to [ Xa , Xb ] = Xc , [ Xb , Xc ] = Xa , [ Xc , Xa ] = Xb .

(b) Any B SO (2; 1) obey BT gB = g, where g = diag[1, 1, 1]. With
B I + L, the matrix L satisfies LT g + gL = 0. So gL is antisymmetric, and
we can take L = gM, with M SO (3). Therefore,
0 c b
dx = cy bz
L = c 0 a
dy = cx + az
b a 0
dz = bx + ay
So that the generators of SO (2; 1) are:

Xa = z
+y ,
y
z
Xb = x
z ,
z
x
Xc = y
x ,
x
y
which obey [ Xa , Xb ] = Xc , [ Xb , Xc ] = Xa , [ Xc , Xa ] = Xb .

3.15 (Exponential of rotation) (a) Referring to Problem 3.1, write the rotation matrix for a very small angle as R( ) = I + R: sum the expansion series that defines exp( R ). (b) When the rotation is applied on the
38
CHAPTER 3
Cartesian coordinates x1, x2 of a two-dimensional Euclidean space, we can

define the corresponding Lie group of rotations, generated by a differential operator X. Find X and calculate X ( ) = exp ( X ) (the minus sign
in exp ( X ) arises because the operator is applied on the coordinates, not
the basis vectors)
SOLUTION 3.15 (a) For a very small angle , we have from Problem 3.1

1
0 1
R( )
= I + R = R =
.
1
1 0
Now the power series that defines exp ( R ) for arbitrary real is
exp( R ) =
n=0
Noting that
R2
1
1
( R )2n +
( R )2n+1
(2n ) !
(
2n
+
1
)
!
n=0
= I, the series can be summed to give
exp ( R ) =
n=0
2n
2n+1
(1)n + R
(1)n
(2n ) !
(
2n
+
1
)
!
n=0
= cos + R sin ,
which reproduces the original rotation matrix R( ).
(b) The transformation functions are given by
f 1 ( ; x1, x2 ) = x1 cos x2 sin ,
f 2 ( ; x1, x2 ) = x1 sin + x2 cos .
From the general formula, the generator is defined as

f 2
f 1
= x2
x1
,
X=

=0 x1
=0 x2
x1
x2
so that Xx1 = x2 , Xx2 = x1. The exponential map X ( ) = eX , defined

by the usual series of powers of X, is applied on x1, x2:
1 2 2
1
X x 1 3 X3 x 1
2!
3!
= x1 cos x2 sin = f 1 ( ; x1, x2 ) ;
1
1
X ( ) x2 = x2 Xx2 + 2 X2 x2 3 X3 x2
2!
3!
= x1 sin + x2 cos = f 2 ( ; x1, x2 ) .
X ( ) x1 = x1 Xx1 +
In the proof, we have used the following identities

X2n+1 x1 = (1)n x2,
X2n+2 x1 = (1)n+1 x1,
X2n+1 x2 = (1)n+1 x1,
X2n+2 x2 = (1)n+1 x2.
39
So the exponential map X ( ) reproduces ( f 1, f 2 ) as it should.

3.16 (Lorentz boost) (a) Referring to the relevant discussion in the Chapter,
show that for any one-dimensional Lie group, with the composition function ( , ), one can find a new parameter 0 such that the composition
function is additive, i.e. ( 0, 0 ) = 0 + 0 (one-parameter Lie groups are
abelian).
(b) In a two-dimensional space with metric g = diag[1, 1], define the
Lorentz boost L ( v ) along the x axis by the matrix

p
v
L(v ) =
where 0 v < 1, = 1/ 1 v2.
v
(i) Show that L ( v2 ) L ( v1) = L ( ( v2, v1 )), with ( v2, v1 ) to be found. (ii)
Find the parameter ( v ), such that L ( 2 ) L ( 1 ) = L ( 2 + 1 ). (iii) Find
the group generator X such that the exponential map exp ( X ) reproduces
L ( ).
SOLUTION 3.16 (a) Assuming one dimensional group space, we use the following formula found in GTAPP 3.3
dx ( )
= ( ) Xx( ),
d
where
() =
"
# 1
( , )
.
=0
If ( ) is a representation of acting on x so that x ( ) = ( ) x(0), we have

d
d
= ( ) X() , (0) = 1,
= 0 (0) = X.
d
d 0
This differential equation can be solved to give

( ) = exp[ ( ) X ] ,
() =
Z
0
( s)ds .
Since for = + , we have the representation ( ) = ( ) + ( ),

the corresponding exponential map exp [ ( ) X ] implies the composition
rule ( ) = ( ) + ( ).
(b) By matrix multiplication we can check that

L ( v 2 ) L ( v 1) =
= L ( v2, v1 ) ,

p
where = 1/ 1 2 , which implies ( v2, v1 ) = v1 v2 / (1 + v2 v1 ), and so

( v2, v1 )
1
= 1 v21 = ( v) =
.

v2
1 v2
v 2 =0
40
CHAPTER 3
Hence the new parameter is (with v < 1)

(v ) =
Z v
0
1
ds =
1 s2
1
2
log
1+v
= tanh1 ( v ),
1v
or equivalently v = tanh , = cosh , and v = sinh . With this new

variable, we have

cosh sinh
L(v ) L( ) =
,
sinh cosh
and check directly that this special Lorentz group is abelian: L ( 2 ) L ( 1 ) =
dL
L ( 2 + 1 ) = L ( 1 ) L ( 2 ). Expanding to first order L ( ) = 1 + d
, we
define the generator X (again, with a + sign) by

d L
0 1
X=
=
1 0
d 0
Now, we sum the power series of the exponential, noting X2 = I:
3 3
2 2 4 4
X +
X + + X +
X +
exp( X) = I +
2!
4!
3!

2
3
2
= X 1+
+ +X +
+
2!
3!
= I cosh + X sinh ,
and recover the matrix L ( ) obtained before. Note that, as the L ( ) matrix
elements have no upper limits, the Lorentz group is non compact.

3.17 (Heisenberg group) Define the sets of matrices:
1 a b
H = 0 1 c ; a, b, c R , h = 0 0 ; , , R .
0 0 1
0 0 0
(a) Show that H is a Lie group, and find its center (invariant abelian
subgroup). (b) Show that for any X h, the exponential eX H, and
conversely if X is any matrix such that etX H, then X = detX /dt |t=0 is in
h. Define a basis for h, together with its algebra. (c) Find the center C (h) of
h. Using the CH formula, calculate exp X for an arbitrary X h.
SOLUTION 3.17 (a) It is evident that I = diag[1, 1, 1] H. If A, A0 H,
matrix arithmetic shows that AA0 and the inverse of A are in H:
1 a + a0 b + b0 + ac0
1 a ac b
AA0 = 0
1
c + c0 , A1 = 0 1
c .
0
0
1
0 0
1
41
The limit of matrices of the form in H is also of that form, so H is a closed

subgroup of GL (3, R ) and a (connected, simply connected) matrix Lie group.
From the product law, we see that matrices of the form
1 0 d
0 1 0
0 0 1
commute with each other and with every A H. So they form the center
C ( H ) of H, defined as C ( H ) = { A H; AB = BA for all B H }.
0 0
(b) Every X h is nilpotent: X2 = 0 0 0 , X3 = 0. So we have:
0 0 0
etX
0 t t + 12 t2
.
= I + tX + 12 t2 X2 = 0 1
t
0 0
1
For every X h, eX is in H. On the other hand, if A ( t ) H such that

a ( t ) = ta0 (0) + , b ( t ) = tb0 (0) + , and c ( t ) = tc0 (0) + , then
limt0 [( A ( t ) I ) /t] is strictly upper triangular and so belongs to h. Thus,
the Lie algebra of H (a space of upper triangular matrices) is h (a space of
strictly upper triangular matrices); conversely the Lie algebra h is exponentially mapped in a unique way to the Lie group H.
(c) The center
C (h) = { X h; [ X, Y ] = 0, Y h} consists of matrices
0 0
0 0
of the form 0 0 0 with R. If X, Y h, then [ X, Y ] C (h), and so
[[ X, Y ], T ] = 0 for all T h. In particular, [[ X, Y ], X ] = 0 and [ X, Y ], Y ] = 0.

Choose for basis in h
0 1 0
0 0 0
0 0 1
P = 0 0 0 , Q = 0 0 1 , E = 0 0 0 ,
0 0 0
0 0 0
0 0 0
with the commutation relations [ P, Q ] = E, [ E, P ] = 0, [ E, Q ] = 0, and

the center C (h) = h Ei. The only non-zero bilinear product is PQ = E.
Any X h can be written as X = P + E + Q, with real , , . As we
have seen eX can be calculated using the fact that X is nilpotent. Now we
recalculate it making use of the CH formula. First note that as E C (h),
exp X = exp E exp ( P + Q ). Further, as [ P, Q ] C (h) the CH formula
simplifies so that exp P exp Q = exp ( P + Q + 12 [ P, Q]).
42
CHAPTER 3
We have now
1
eX = eP+E+Q = e( 2 )E eP eQ
= [ I + ( 12 ) E][ I + P ][ I + Q ]
= I + ( 21 ) E + P + Q + PQ
= I + ( + 21 ) E + P + Q.
This agrees with result obtained in (b).
3.18 (Euclidean group E(2)) In classical physics, space is assumed to be homogeneous and isotropic, so that physical phenomena should not depend
on the specific location or orientation of the physical system. In mathematical terms, this means a physical Euclidean space, with its symmetries
described by the Euclidean group, which consists of uniform translations
and uniform rotations. Take, as an example, the Euclidean group E(2) defined by the following transformations of the Cartesian coordinates x1, x2
in R2 :
x01 = x1 cos x2 sin + 1
x02 = x1 sin + x2 cos + 2 ,

in terms of a displacement distance ( < i < ; i = 1, 2) and a rotation
angle (0 2). We call A ( , ) the operator effecting this transformation: A ( , ) xi = x0i = Rij ( ) x j + i .
(a) Verify that A ( , ) form a Lie group, and that A ( , 0) and A (0, )
are its two subgroups. Calculate the transformation functions as well as
the composition functions of E(2).
(b) Calculate the generators (X0 , X1 , X2 ) of E(2) and their brackets.
SOLUTION 3.18 We shall call the rotation angle 0 , and use the notation
( 0 , 1, 2 ) ( 0, ). Similarly for or .
(a) Apply A ( ) on x0 = A ( ) x, we have
x00i = Rij ( 0 ) x0j + i
= Rij ( 0)[ R jk ( 0) xk + j ] + i
= Rik ( 0 + 0 ) xk + Rij ( 0) j + i ,

which implies the product rule: A ( ) A( ) = A 0 + 0, R ( 0 ) + . With
the identity A (0, 0) = e, the inverse follows: A ( )1 = A ( 0, R (0) ).
43

Let us also check associativity:

A ( ) ( A ( ) A ()) = A ( ) A 0 + 0 , R ( 0) +

= A 0 + 0 + 0 , R ( 0 + 0 ) + R ( 0 ) + ;

( A ( ) A( )) A ( ) = A 0 + 0, R ( 0) + A ( )

= A 0 + 0 + 0 , R ( 0 + 0 ) + R ( 0 ) + .

Hence A ( ) A ( ) A( ) = A ( ) A( ) A ( ).
The transformation functions are given by
f i ( ; x) = Rij ( 0 ) x j + i ;
i = 1, 2 ,
and the composition functions

0 ( , ) = 0 + 0 ;
i ( , ) = Rij ( 0 ) j + i ,
i = 1, 2 .
Closure and associativity can be readily verified for these functions.

Since the composition functions are analytic in its arguments, they define a
Lie group. We remark that
A ( , 0) A ( , 0) = A ( + , 0) ,
A (0, ) A (0, ) = A (0, + ) ,
which imply closure for the sets { A ( , 0)} and { A (0, )}. Since the general group properties are also satisfied, they form subgroups of E(2): the
subgroup of rotation R = { A ( , 0)} and the subgroup of translation T =
{ A (0, )}.
(b) The generators are obtained from the transformation functions by
the formula (with a conventional sign):

f j ( ; x)

X ( x ) =
= 0, 1, 2 .
=0 x j
The nonzero derivatives are f 1 /|0 = x2, f 1 /1 = 1, f 2 /|0 = x1,
f 2 /2 = 1. So we have the generators in geometric space
+ x2
,
x2
x1
X1 ( x ) =
,
X2 ( x ) =
,
x1
x2
X0 ( x ) = x1
which bracket according to [ X0 , X1 ] = X2 , [ X0 , X2 ] = X1 , [ X1 , X2 ] =

0. These operators are anti-hermitian in an inner-product complex vector
44
CHAPTER 3
space. In physics, it is customary to define hermitian operators instead:

def
J = iX0 ( x ) = i x1
x2
,
x2
x1
def
Pi = iXi ( x ) = i
, ( i = 1, 2) ,
xi
together with the brackets [ J, P1] = iP2 , [ J, P2] = iP1, [ P1, P2 ] = 0

3.19 (Linearization
of E(2)) Let L be the vector space spanned by vectors

x1
x
v = x2 , where 1 is a vector in two-dimensional Euclidean space. The
x2
1
object of this problem is to show that E(2) is isomorphic to a matrix Lie

group, by showing that E(2) group and algebra elements (Problem 3.18)
are expressible as 3 3 matrices in space L, denoted M(3, R ).
(a) Find the generators of E(2) and their brackets as elements of M(3, R ).
(b) Prove in M(3, R ) that A ( , ) = A (0, ) A ( , 0).
(c) Show that the exponential series of exp ( X0 ) and exp( 1 X1 + 2 X2 )
sum to A ( , 0) and A (0, ) respectively.
(d) What is A ( , 0) A(0, )? Compare it to A (0, ) A ( , 0).
(e) In question (b), we have proved that exp ( 1 X1 + 2 X2 ) exp( X0 ) is
identical to the group element A ( , ). But the theory tells us that, in general, it is the map exp( X0 + 1 X1 + 2 X2 ) that should reproduce the group
elements. How do you explain the difference?
SOLUTION 3.19 We will use the same symbols as in Problem 3.18 although
now they refer to their matrix forms. Thus, the general transformation of
E(2) acting on R3 vectors is given by the 3 3 real matrix

cos sin 1
R2 ( )
A ( , 1, 2 ) = sin cos 2
,
0
1
0
0
1
and the elements of the (abelian) rotation group R = { A ( , 0)} and abelian
translation group T = { A (0, )} are
cos sin 0
1 0 1
A ( , 0) = sin cos 0 , A (0, ) = 0 1 2 .
0
0
1
0 0 1
(a) The group generators are identified with the first-order terms in the
expansion series of the group element, A ( , 1, 2 ) = I + X0 + 1 X1 +
2 X2 + ,
"
#
"
#
"
#
0 1 0
0 0 1
0 0 0
X0 = 1 0 0 , X1 = 0 0 0 , X2 = 0 0 1 .
0
0 0
0 0 0
45
With matrix arithmetic, we find Xi X0 = 0, X0 X1 = X2 , and X0 X2 = X1 .

Hence the brackets: [ X0 , X1 ] = X2 , [ X0 , X2 ] = X1 , and[ X1, X2 ] = 0.
(b) Rewrite R( ) = A ( , 0) and T( ) = A (0, ) in the form

R2 ( ) 0
I
R( ) =
T( ) = 2
,
0
1
0 1
where R2 ( ) and I2 are 2 2 matrices. Now take the product

I2 R2 ( ) 0
R2 ( )
T( )R( ) =
=
= A ( , ),
0 1
0
1
0
1
Another way to obtain the same result is to apply the group product rule:
A ( , ) A ( , 0) = A ( , ) T( )
A ( , ) = T( ) A ( , 0) = T( )R( ).
(c) To sum the series for exp ( X0 ), note first the identities (simple to
verify): X02n = ()n+1 X02 and X02n+1 = ()n X0 , and also
"
#
"
#
0 0 0
1 0 0
I = E X02 , E = 0 0 0 , X02 = 0 1 0 .
0
0 1
Then, the power series in question is
exp( X0 ) =
n=0
1
1
( X0 )2n +
( X0 )2n+1
(2n ) !
(
2n
+
1
)
!
n=0
= E X02
n=0
(1)n 2n
(1)n
( ) + X0
( )2n+1
(2n ) !
(
2n
+
1
)
!
n=0
X02
= E
cos + X0 sin
= A ( , 0) = R( ).
As for exp( 1 X1 + 2 X2 ), it suffices to notice that X12 = X22 = 0 and
X1 X2 = X2 X1 = 0, so that
exp ( 1 X1 + 2 X2 ) = (1 + 1 X1 )(1 + 2 X2 )
= I + 1 X1 + 2 X2
= A (0, ) = T( ).
So exp ( X0 ) generates the subgroup of rotation R, and exp( 1 X1 +
2 X2 ) the subgroup of translation T .
(d) Applying group multiplication, we get A ( , 0) A (0, ) = A [ , R2( ) ].
So we have here R( )T( ) = A [ , R2( ) ], whereas in (b) we obtained
46
CHAPTER 3
T( )R( ) = A ( , ). The matrices R( ) and T( ) are both elements of E(2),

and so their products must also be elements of the group. As they do not
commute, their products depend on the order of the factors.
(e) In Question (b), we have proved that exp ( 1 X1 + 2 X2 ) exp( X0 ) is
identical to the group element A ( , ). But the theory tells us that, in general, it is exp( X0 + 1 X1 + 2 X2 ) that should reproduce the group elements. The first few powers of the power series of this exponential are:
M = X0 + 1 X1 + 2 X2 ,
M2 = 2 X02 2 X1 + 1 X2 ,
M3 = 3 X0 2 1 X1 2 2 X2 ,
M4 = 4 X02 + 3 2 X1 3 1 X2 ,
M5 = 5 X0 + 4 1 X1 + 4 2 X2 .
It follows the exponential series

1 2
M +
2!

3
4
= ( E X02 ) + X0 + X02 2 + +
3!
4!

1
3
2 2
4
+ X1
+
+
3!
2!
4!

3
2
2
1
4
+ X2
+ +
+
3!
2!
4!
exp M = I + M +
= E + X0 sin X02 cos + X1 1 + X2 2.
where

sin
1 cos
1 = 1
2
S1j ( ) j,

1 cos
sin
2 = 1
+ 2
S2j ( ) j,
Thus we have shown that the exponential map of the algebra h X0 , X1 , X2 i

gives
exp( X0 + 1 X1 + 2 X2 ) = E + X0 sin X02 cos + X1 1 + X2 2
= A ( , 1, 2 ) = A [ , S( ) ].
That is, exp M generates the same Lie group, E(2), but the elements are
parameterized differently from the original definition, A ( , ). The two
parameterizations are equivalent, being related by a nonsingular similarity
47
transformation. (Let s = sin ( /2) /( /2) and = S/s, then T = I, and

so 1 = T .)

3.20 (Invariant subgroup and factor group of E(2)) With the notations and
results in Problem 3.19, answer the following questions:
(a) Prove that eX0 T( ) eX0 = T[ R ( )]. From this result, discuss the
meaning of the commutation relations between X0 and X1 , X2 .
(b) Prove that T = { A (0, )} is the invariant subgroup of E(2), and that
the factor group E2 / T is isomorphic to SO(2).
(c) Is E(2) simple, semi-simple, compact?
SOLUTION 3.20
(a) We use associativity of group product to calculate
eX0 T( ) eX0 = A ( , 0) A (0, ) A (, 0)
= A ( , 0) A ( , )
= A [0, R2( ) ] T[ R2( ) ].
From this result and T( ) = I + 1 X1 + 2 X2 , we have
2
eX0 ( j Xj ) eX0 =
j=1
Xk Rkj ( ) j.
j,k=1
Since j are arbitrary, it follows eX0 Xj eX0 = 2k=1 Xk Rkj ( ) which is equivalent to the bracket relations
[ X0, Xj ] = e jk Xk ,
( e12 = e21 = 1).
These commutation relations between X0 and X1 , X2 say that ( X1 , X2 ) transforms as a vector under rotation.
(b) First, use the group product rule
A ( , )T( ) A ( , ) = A (0, ) A ( , 0) T( ) A ( , 0)1 A (0, )1
= A (0, )T[ R2( ) ] A (0, )1
= T( )T[ R2( ) ]T( )1 = T[ R2( ) ].

The last line arises from the commutativity of translations. Hence the abelian
subgroup T = { A (0, )} is the center of E(2).
(b) As T is an invariant subgroup, we have E(2) = T + T A ( , 0) (with
6 = 0 in the second term). Letting range over all allowed values, we
have a continuous family of cosets in one-to-one correspondence with the
elements of { A ( , 0)}
= SO(2). So the factor group E(2) / T is isomorphic
with SO(2).
48
CHAPTER 4
(c) As E(2) has a normal subgroup, T , it is not simple; and since T is

abelian, E(2) is not semi-simple. As the range of values of the translation
parameters 1 , 2 is unbounded, the Lie group is not compact.

3.21 (The adjoint representation) Consider a Lie algebra g of dimension .
When it has a p-dimensional subalgebra, say h, we denote the first p
basis vectors by X , and the last p vectors by Xi , , which span h. Give
the structure of the matrices R( Xi ) and R( X ) in the adjoint representation
in the following situations: (a) g is abelian. (b) g has a subalgebra h. (c) g
has an invariant subalgebra h. (d) g is semisimple. Note: This problem is not
in GTAPP.
SOLUTION 3.21 In general the matrices are of the form
k

R
(
X
)
R
(
X
)
k
11
12
R( X ) = c c =
R21 ( X ) R22 ( X )
j cj ckj
R( Xi ) = ci
j cij
k

R
(
X
)
R
(
X
)
k
11
12
i
i
ci =
R21 ( Xi ) R22 ( Xi )
ckij
(a) g is abelian: R( X ) = 0, R( Xi ) = 0.
(b) h is a subalgebra: [h, h] h, and so R21 ( Xi ) = 0, and

R11 ( X ) R12 ( X )
R( X ) =
,
R21 ( X ) R22 ( X )
R11 ( Xi ) R12 ( Xi )
R( Xi ) =
0
R22 ( Xi )
(c) h is an invariant subalgebra: [g, h] h, and so R21 ( X ) = 0 and

R11 ( Xi ) = R21 ( Xi ) = 0.
(d) g is semisimple:

R11 ( X ) R12 ( X )
R( X ) =
,
R22 ( X )

R12( Xi )
R( Xi ) =
,
R22( Xi )
where indicates that not all the elements in this block can be zero for any
choice of the basis.
Chapter 4. sl(2, C )
4.1 (Casimir invariant) Let h, e, f be the canonical basis of sl(2, C ). Find
the quadratic Casimir invariant in any representation, and its value in an
irreducible representation. Hint: Use the angular momentum of su (2)C .
49
SOLUTION 4.1 In any representation of sl(2, C ), we have z 7 Z, with

Z related to the angular momentum operators Ji of su(2)C by J3 = 1/2H,
J1 = 1/2( E + F ), J2 = i/2( E F ). Now, Ji satisfies [ Ji, Jj ] = ieijk Jk , where
eijk is the usual antisymmetric unit tensor. Define J2 = Ji Ji , then for any
j = 1, 2, 3, we have [ J2 , Jj ] = Ji [ Ji, Jj ] + [ Ji, Jj ] Ji = ieijk ( Ji Jk + Jk Ji ) = 0. So
for any X = C J1 + C J2 + C J3 , we also have [ J2 , X] = 0. Conclusion: J2 is a
quadratic Casimir operator in su(2)C . Going to sl(2, C ), it has the form J2 =
1/4 H 2 + 1/2( EF + FE), which satisfies [ J 2 , Z ] = 0 for any Z (sl(2, C )).
In the simple representation of highest weight n (wrt H), sl(2, C ) acts on the
highest weight vector v according to Hv = nv, Ev = 0, and EFv = [ E, F ] v =
nv, and so J2 v = ( 1/4n2 + 1/2n ) v, which gives the invariant value of J2 in
the simple representation of highest weight n as 1/4n ( n + 2) = j( j + 1),
with n = 2j.

4.2 (Brackets) Let h, e, f be the canonical basis of sl(2, C ). Calculate the
brackets [ h, em], [ h, f m], [ f , em], and [ e, f m].
SOLUTION 4.2
(a) [ h, em] = 2mem

(b) [ h, f m] = 2m f m
(c) [ f , em] = mem1 h m ( m 1) em1
(d) [ e, f m] = m f m1 h m ( m 1) f m1
4.3 (Spin-one representation) Let , C2 transform into 0 = a + c,

0 = b + d underSU (2) (so that d = a , c = b, and aa + bb = 1).
Let + = 2, 0 = 2 , and = 2 . Show that the s transform as a
spin-one representation; find the representation matrix D1 in terms of the
Euler angles.
SOLUTION 4.3 With the s and the transformation of , as given, we
have
0
+
= 02 = a2 + 2ab 0 + b2
00 = 2 0 0 = 2ab + + ( aa bb ) 0 2a b
= 02 = b2 + + 2a b 0 + a2
The first line gives:
D1,1 = a2 = ei c2 ei ;
D0,1 = 2ab = 2cs ei;
D1,1 = b2 = e+i s2ei ;
50
CHAPTER 4
where we have used a = ei(+)/2 c, b = ei()/2 s, c = cos( /2), and

s = sin ( /2). The last two lines give:
D1,0 = 2ab = 2cs ei ; D0,0 = aa bb = c2 s2
D1,0 = 2a b = 2csei
D1,1 = b2 = ei s2 ei ; D0,1 = 2a b = 2cs ei

D1,1 = a2 = ei c2 ei .
1
0
1
With Dmm
0 ( , , ) = exp i( m + m ) d mm0 ( ), we have
2cs
s2
2 s 2 2cs .
d1mm0 ( ) = 2cs c
s2
2cs
c2
4.4 (Quaternion groups) A quaternion q is a number that can be written

as q = 3=0 x q where x are real numbers, and q = { q0, qi } (with =
0, . . . , 3 and i = 1, 2, 3) are defined such that
q 0 q 0 = q 0,
q0 qi = qi q0 = qi ,
qi q j = ij q0 + eijk qk ;
and q0 = q0 , qi = qi under complex conjugation.

(a) Let Q be a one-dimensional quaternion (q)-valued vector space, with
typical vector . The set of all non-zero mappings 0 = Q form a Lie
group, called GL(1,Q). Find its generators and its Lie algebra.
(b) Show that { Q GL (1, Q )|Q Q = 1} is a Lie group. Find its associated Lie algebra.
(c) Evaluate the exponential mapping of the above Lie algebras.
SOLUTION 4.4 (a) Let Q be the mapping Q 7 Q: 7 Q, where is a q
number (one-dimensional vector) in Q, and Q = 3=0 q a q-valued operator, depending on the real parameters . Clearly QQ0 = Q00 (closure),
and I = 1 q0 is the identity operator. So the set of all non-zero Q forms a
four-dimensional non compact Lie group, called GL(1,Q). Near the identity, Q I + Q 1 q0 1/2 q , so that by definition X = 1/2q , with
= 0, . . . , 3, are the generators of GL(1,Q). From the defining properties of
q , the X s obey the bracket rules
[ X0, Xi ] = 0,
[ Xi , Xj ] = eijk Xk ,
and so define a Lie algebra gl(1, Q ) which is manifestly a direct sum of two
commuting subalgebras: gl(1, Q ) = h X0 i h X1 , X2 , X3 i.
(b) { Q GL (1, Q )|Q Q = 1} is a closed and bounded subgroup of
GL (1, Q ), and so is a Lie group (called SL (1, Q )). If Q is required to be
unimodular, Q Q = 1, we have near the identity
Q Q (1 + Q ) (1 + Q ) = (1 0 ) q0 + O( 2).
51
It follows that 0 = 0, i.e. 0 is fixed, leaving 1, 2 , 3 as the remaining

free parameters. The Lie algebra (gl(1, Q )) spanned by X1 , X2 , X3 has the
product rules [ Xi , Xj ] = eijk Xk .
(c) First, define 2 = i i2 and
= /, then we have
i Xi
i
2
1
= i2 Xi2 + 1/2 i j ( Xi Xj + Xj Xi ) = 2 q0,
4
i
ij
which means that i i Xi is cyclic, so that the mapping exp sl(1, Q ), which
gives SL (1, Q ), has a closed form:
exp
i Xi
i
=
n
1
n!
i Xi
i
n
= q0 cos
q sin .
2
2
On the other hand, when mapping gl(1, Q ) onto GL (1, Q ), we may use
the identity exp ( X0 + Xi ) = exp X0 exp Xi since [ X0 , Xi ] = 0, and so we
have

q sin
exp X = exp( 0X0 ) exp( i Xi ) = e0 /2 q0 cos
.
GL (1, Q ) is non compact because 0 has unlimited range, but in contrast SL (1, Q ) is compact, because exp( i Xi ) involves periodic functions
of , and its elements are such that ( ,
) = ( + 4,
) = ( + 2,
).
We may choose the parameter space to be all points such that 0
2, so that the group elements are in one-to-one correspondence with elements of the algebra within a radius of 2 of the origin. Since further
(2,
) = (0,
) = q0, all points on the surface of this sphere are identified with a single group operation, q0. Note the isomorphism of GL (1, Q )
and U(2) on the one hand, and SL (1, Q ) and SU(2) on the other hand.
4.5 (det R = 1) Show that the invariance of the tensor eijk under rotations
is equivalent to det R = 1 for any orthogonal real 3 3 matrix R.
SOLUTION 4.5 Invariance of eijk under rotation means that for any rotation
R, we have
Ri` R jm Rkn e`mn = eijk .
First, let ( ijk ) = (123), then R1` R2m R3n e`mn = 1. The LHS is just det R, and
so the equation is: det R = 1. Secondly, if ( ijk ) = (11k ) for any k, then the
RHS= 0, and the LHS is
Rk1 ( R12 R13 R13 R12 ) + Rk2 ( R11 R13 R13 R11 ) + Rk3 ( R11 R12 R12 R11 )
which is identically equal to zero, no new information. And so the invariance relation for eijk is just equivalent to the condition det R = 1.
52
CHAPTER 4
4.6 (Invariant axis) Show that a proper rotation in an odd-dimensional space

possesses an invariant axis.
SOLUTION 4.6 Let R be an n n real orthogonal matrix with det R = 1.
Its eigenvalue problem is Ru = u, yielding the characteristic equation
det( R I ) = 0. It is a polynomial equation in of order n, with realvalued coefficients (given by the real values entries of R); if is a root, so
is . The roots must be either real or come in pairs of complex conjugates.
Since R is unitary (real orthogonal),
h u, u i = h Ru, Rui = | |2h u, u i,

so that | | = 1 is a condition that all the roots must satisfy. It follows that
the root = 1 can occur only when n is odd. Then the corresponding
eigenvector u(1) satisfies the equation Ru(1) = u(1) and the eigenvector u(1)
defines the direction of the axis of rotation, which is of course the invariant
direction.

4.7
Consider
the group of pure rotation in R3 :
(Similarity3 transformation)

T
R M(3, R )| R R = I; det R = 1 .
(a) Find the direction of the axis and the angle of rotation in terms of
the matrix elements Rij .
(b) Find the transformation U that diagonalizes R, so that U 1 RU =
is a diagonal matrix.
(c) Find the transformation that gives in a real orthogonal basis.
SOLUTION 4.7 From the preceding problem, we know that R must have
the eigenvalues 1 = ei , 2 = ei (0 ), and 3 = 1 (so that the
corresponding eigenvector u(3) is the direction of the axis of rotation).
(a) Consider Ru(3) = u(3) . Since RT R = I, we also have RT u(3) = u(3) . It
follows that ( R RT ) u(3) = 0. Writing this out in components, we have
(3)
( R21
( R31
(3)
R12 ) u1
(3)
R13 ) u1
( R12 R21 ) u2
(3)
+( R32 R23 ) u2
(3)
+( R13 R31 ) u3
(3)
+( R23 R32 ) u2
=0
=0
= 0.
These equations give the components of u(3) :

(3)
(3)
(3)
u1 : u2 : u3 = ( R23 R32 ) : ( R31 R13 ) : ( R12 R21 ) ,

which specify the direction of the axis of rotation. As for the rotation angle
, we take the trace of R, which is also equal to its diagonalized form:
Tr R = R11 + R22 + R33 = ei + ei + 1 = 1 + 2 cos .
This gives in terms of matrix elements.
53
(b) To find the two other eigenvectors, first note that u(1) = u(2) from
the fact that R = R and 1 = 2 , and so we only need to calculate the
vector u(1) ; it can be found from the condition that it is orthogonal to u(3) .
The evaluation of the eigenvectors is completed by normalizing them to 1.
We define a matrix U such that the entries of its columns are given by
(k)
the components of the eigenvectors: Ujk = u j . Its unitarity follows from
the orthonormality of the eigenvectors. The eigenequations then appear as
(k)
Rij u j
(k)
= k u i
(j)
= ui jk k ,
or in matrix form, RU = U, where = diag [ 1 , 2 , 3 ] expressed in the

complex basis { u(i) }.
3) , we now define a real basis
(c) Noting that u(1)= u(2) and u(3) = u(
with e1 = ( u1 + u2 ) / 2, e2 = i( u1 u2 ) / 2, and e3 = u3 . The matrix,
V = [ Vma ], defined by ea = m um Vma , connecting the 2 bases is given by
1/ 2 i/ 2 0
V = 1/ 2 i/ 2 0 .
0
0
1
V transforms the diagonal matrix in the
diagonal matrix R in the real basis { e }:
cos sin
V1 V = sin cos
0
0
complex basis { u } into a non

0
0 R ( , e3) .
1
Putting this together with the result in (b), we have
V1 U 1 R ( , u3)UV = R ( , e3)
Thus, the original R parameterized in angle and axis u3 is first diagonalized by U in the { u } basis, then transformed by V into a matrix in a real
basis. Put it another way, S = UV converts a rotation R ( , u3) about the
axis u3 into a rotation about the axis e3 through the same angle .

4.8 (Angle-axis rotation matrix) Write out the 3 3 matrix of rotation R ( , n)
in terms of the rotation angle and the spherical coordinates , of the axis
n; or, alternatively in terms of and the Cartesian components n1 , n2 , n3 of
n in an orthonormal basis { ei }, with n2 = 1.
SOLUTION 4.8 We use the formula R ( , n) = S ( , ) R( , e3 ) S1 ( , ),
where
n = s c e1 + s s e2 + c e3 = ei ni ,
S ( , ) = R ( , e3 ) R ( , e2 )
c = cos ;
s = sin ;
etc.
54
CHAPTER 4
The matrix S and its inverse (in the Cartesian basis { ei }) are:
c c s s c
c c c s s
S = c s c s s ,
S 1 = s c
0 .
s
0
c
s c s s c
In addition, we know the matrix for R ( , e3 ) from GTAPP 4.4. So we can

evaluate, by matrix multiplication, the elements Rij ( , n) Tij :
T11 = s2 c2 (1 c ) + c
T12 = s2 c s (1 c ) c s
T22 = c + s2 s2 (1 c )
T23 = s c s (1 c ) s c s
T13 = s c c (1 c ) + s s s
T31 = s c c (1 c ) s s s
T33 = 1
s2 (1 c ).
T21 = s2 s c (1 c ) + c s
T32 = s c s (1 c ) + s c s
In terms of the Cartesian components ni , expressed in spherical coordinates, n1 = s c , n2 = s s , and n3 = c , we have in the Cartesian basis
n21 (1 c ) + c
n 1 n 2 (1 c ) n 3 s n 1 n 3 (1 c ) + n 2 s
T = n 2 n 1 (1 c ) + n 3 s
n22 (1 c ) + c
n 2 n 3 (1 c ) n 1 s
n 3 n 1 (1 c ) n 2 s n 3 n 2 (1 c ) + n 1 s
n23 (1 c ) + c
We can do some checking here. First, calculate the trace of T, which is

T11 + T22 + T33 = 1 + 2c = 1 + 2 cos , to be compared with
Tr R ( , n) = Tr ( SR ( , e3 ) S1 ) = Tr R ( , e3 ) = 1 + 2 cos .
Secondly, we can use the invariance property R ( , n) n = n to verify that
Tij n j coincides with ni as it should.

4.9 (Euler angles rotation matrix) Write out the rotation matrix R ( , , )
in a Cartesian basis in terms of the Euler angles.
SOLUTION 4.9 In terms of the Euler angles , , , the rotation matrix can
be written as
R ( , , ) = R ( , e3 ) R ( , e2 ) R ( , e3 ).
Making use of the expressions for R ( , ei ) given in 4.4, we can calculate
the rotation matrix by simple matrix multiplications, leading to the result:
c c c s s c c s s c c s
R ( , , ) = s c c + c s s c s + c c s s ,
s c
s s
c
where c = cos , s = sin , etc..
55
4.10 (Relations between parameters) Find the relations between the angleaxis parameters and the Euler angles for the same rotation.
SOLUTION 4.10 This proof relies on the results given in the solutions 4.8
and 4.9. The two matrices T R ( , n) and R R ( , , ) are equal, component by component. We need three independent relations between the
two sets of parameters. One such relates the traces of T and R: We already
have Tr T = 1 + 2c . In addition,

+
2
Tr R ( , , ) = R11 + R22 + R33 = 1 + 4 cos
cos2 .
2
2
This determines via c = cos :
2
c = 1 + 2 cos
+
2
cos2
.
2
(4.1)
Next we have (by trigonometry)

R11 = c c c s s = c+ cos2 ( /2) c sin2 ( /2),
and with (4.1)s help, we get
T11 = 1 + 2c2 sin2 ( /2) + 2 cos2 ( /2) cos2 (( + ) /2).
Then 2c2 = 1 + cos(2 ) and T11 = R11 together yield
cos(2 ) + cos( ) = 0
+
.
2
(4.2)
Finally, consider T33 = R33 , which gives 1 s2 (1 c ) = c , leading to

tan2 = s2 / (1 s2 ) = (1 c ) / [(1 c ) (1 c )]. With the expression
(4.1) for c , we get
tan( /2)
tan =
.
(4.3)
sin (( + ) /2)
The equations (4.1)(4.3) relate the two sets of rotation parameters.
Another tactic is to first derive (4.1) as above, then use the invariance of
n, that is R ( , , )n = n, with the components expressed in , variables,
to obtain (4.2) and (4.3). This approach requires knowledge of the functions
Rij ( , , ) only, and not of Tij ( , , ). Conversely, one can evaluate the
Euler angles from the angle and axis parameters:
= 2
cos = 1 (1 cos ) sin2

1 cos 2 cos
cos( + ) =
.
1 + cos
56
CHAPTER 4
4.11 (R ( , , ) at = 0, ) Show that R ( , 0, ) is function of the sum +

only, and R ( , , ) of the difference only.
SOLUTION 4.11
(i) R ( , 0, ) = R ( , e3 ) R (0, e2 ) R ( , e3 ) = R ( , e3 ) R ( , e3 )
= R ( + , e3 ).
(ii) We evaluate by matrix multiplication
R ( , , ) = R ( , e3 ) R ( , e2 ) R ( , e3 )
c s 0
1 0 0
c s 0
= s c 0 0 1 0 s c 0
0
0
1
0 0 1
0
0
1
cos( ) sin( ) 0
= sin ( ) cos( ) 0 .
0
0
1
4.12 (To rotate a vector) Show that the rotation R ( , n) of a position vector
x in R3 produces the following vector (with n, n2 = 1):
x0 = x +
sin
1 cos
( x) +
[ ( x)].
(4.4)
SOLUTION 4.12 Let x = x/ | x|, and define an orthonormal basis formed

by the unit vectors n x / sin , u n ( n x ) / sin , and n, where is
the angle between n and x . In this basis, x = sin u + cos n, and the
rotation in question has the expression
cos sin 0
R ( , n) = sin cos 0 .
0
0
1
The rotated vector x0 is given by

x0 = | x| R ( , n) x
= (sin ) n x cos sin | x| u + cos | x| n

= cos x + sin ( n x) + (1 cos )( n x) n
or equivalently, with ( x) = ( x) + 2 x,
x0 = x +
1 cos
sin
( x) +
[ ( x)].
Therefore, to the lowest order of the rotation angle, the rotation by the directed angle applied on x produces the vector x0 = x + ( x) + O ( 2).
57
4.13 (Commutation of angular momentum) (a) Show that the sequence of

infinitesimal rotations by the directed angles , , , and on the position vector x is equivalent to the rotation by the directed angle on x.
(b) In Hilbert space, a rotation by is represented by the unitary operator
U ( ) = e A, where A = X with Xi denoting the generators of rotation.
Following the argument used in (a), derive the commutation relations for
Xi (see also 3.3 GTAPP).
SOLUTION 4.13 (a) The four successive infinitesimal rotations in question

applied on x give
A : x 7 x0 = x + x
B : x0 7 x00 = x0 + x0 = x + x + ( x) + x
A : x00 7 x000 = x00 x00 = x + x + ( x) ( x)

B : x000 7 x0000 = x000 x000 = x + ( ) x .
So the sequence A, B, A, B is effectively equivalent to a rotation by the
directed angle , transforming x into x + ( ) x.
(b) A rotation R ( ) is represented by the unitary operator U ( ) = e A
in Hilbert space, where A = X with Xi denoting the (anti-Hermitian)
generators of rotation, related to the Hermitian operators by Xi = iJi,
with i = 1, 2, 3.
The sequence A, B, A, B considered in (a) is therefore represented
by the operator U ( ) U ( ) U ( ) U ( ) in Hilbert space. To the first few
terms, U ( ) = e A 1 + A, and this product of exponentials is given by
U ( ) U ( ) U ( ) U ( ) I + [ X, X ] .
On the other hand, as we have seen in (a), this sequence of rotations taken
together is equivalent to a rotation by the directed angle , represented
by the operator U ( ) I + ( ) X. Equating the two results, we
have
[ X, X ] = ( ) X + i ( , ) I ,
where the extra additive constant term, ( , ) I = (, ) I, is necessary
for generality. In Cartesian components it follows
[ Xi , Xj ] = eijk ( Xk + i k I )
(summing over repeated index).
Now we can redefine Xk + i k I Xk for k = 1, 2, 3, then we obtain the

commutation relations for the rotation generators
[ Xi , Xj ] = eijk Xk .
58
CHAPTER 4
If we now introduce the Hermitian operators Ji via Xi = iJi, we obtain
[ Ji, Jj ] = ieijk Jk .
These are the bracket relations for the angular momentum operators.

4.14 (Exponential mapping) Re-express R ( , n) = exp(ink Lk ) in a nontranscendental form by summing the power series of the exponential. Lk
are 3 3 matrices of the angular momentum operators in a Cartesian basis.
SOLUTION 4.14 The 3 3 matrices Lk , with k = 1, 2, 3, that define so(3) are

given by
0 0 0
0 0 i
0 i 0
L1 = 0 0 i L2 = 0 0 0 L3 = i 0 0 .
0 i 0
i 0 0
0 0 0
With ( n1 , n2 , n3 ) denoting the Cartesian components
dinary matrix operations produce
0
in3 in2
1 n21
2
n L = in3
0
in1
( n L ) = n1 n2
in2 in1
0
n1 n3
of a unit vector n, or
n1 n2 n1 n3
1 n22 n2 n3 ,
n2n3 1 n23
and ( n L)3 = ( n L). So n L is a cyclic matrix, and the power series for
exp(in L ) divides into even and odd powers of n L:
exp(in L ) = 1 +
m =0
(in L)2m+1
(in L)2m+2
+
(2n + 1) !
(2n + 2) !
m =0
= 1 in L sin ( n L)2 (1 cos ),

which is the result we want. An easier way is to use R ( , n) = SR ( , e3) S1 ,
where
R ( , e3) = exp (iL3) = 1 iL3 sin L23 (1 cos ),
S = R ( , e3) R ( , e2) = SL3 S1 = n L,
(the vector n is in direction ( , )). Then, we get

R ( , n) = SR ( , e3) S1
= 1 i SL3 S1 sin SL23 S1 (1 cos ),
= 1 i n L sin ( n L)2 (1 cos ).
This result is in agreement with the summation of the power series obtained above, and with with the solution to Problem 4.8.
59
4.15 (From Cartesian to spherical basis) Show that the matrices Jk D1 ( Jk )

in the spin j representation (in which D1 ( J3 ) is diagonal) are equivalent by
a similarity transformation to the defining matrices Lk of the algebra so (3)
given in the preceding problem.
SOLUTION 4.15 The matrices Jk D1 ( Jk ) in the j = 1 angular-momentum
representation are given by
"
#
"
#
"
#
1 0 0
1 0 1 0
i 0 1 0
J1 = 1 0 1
J= 1 0 1
J3 = 0 0 0 .
2 0 1 0
2 0 1
0
0 0 1
We want to calculate the matrix U that transforms L3 into the diagonal
matrix J3. This is equivalent to solving the eignevalue equation L3 u = u .
There are three roots = +1, 0, 1 with corresponding normalized vectors

1
0
1
1
1
0
i
u = 0
u = i .
u =
2 0
2 0
1
()
The entries in each column of the transformation matrix {Ui } { ui } are

given by the components of each eigenvector (+0 for both rows, columns):
1 0
1
1 i 0
1
i
i 0 i
U=
U 1 = 0 0
2 .
2 0 2 0
2 1 i 0
Then U 1 Lk U = Jk for k = 1, 2, 3, as we can check. U converts a Cartesian
basis into the angular-momentum canonical basis in three dimensions. Another way of finding U, given Li and Ji , is to use the equations
L3 U = U J3
and
L1 U = U J1 ,
to determine all the elements of U up to an overall constant. Requiring that

U be unitary, we recover the matrix obtained by the first method.
As a corollary, we can relate the defining matrix R ( , , ) of SO(3) (cf.
Problem 9) to the irreducible representation D1 [ R ( , , )] as follows:
U 1 R ( , , )U = D1 ( , , ) .
Of course, this is true for any parameterization; e.g. in ( , n), knowing
R ( , n) = exp(ink Lk ) from Problem 14, we can get exp (ink Jk ):
D1 [ R ( , n)] = U 1 R ( , n)U
= 1 i n U 1 LU sin ( n U 1 LU )2 (1 cos )
= 1 i n J sin ( n J )2 (1 cos )
60
CHAPTER 4
gives the j = 1 representation of rotation R ( , n) in the angle-axis parameterization.

4.16 (Non standard parameterization) In GTAPP we emphasize the standard parameterization ( 1, 2, 3 ) of SU(2). But there are times when it is
more advantageous to use a non-standard parameterization (e.g. to make
the exponential mapping simpler). In SU(2) the use of a parameterization associated with the operators J = J1 iJ2 has a further advantage
when one has to study transitions between states belonging to the same irreducible representation. In this problem, we examine just such mappings
involving the elements J , J3 .
(a) Evaluate Uo ( +, , 3 ) = exp[i( + J+ + J + 3 J3 )] and relate the parameters +, , 3 to the standard parameters 1, 2, 3 .
(b) Evaluate Ua ( + , , 3 ) = exp [i( + J+ + J )] exp[i3 J3 ], and
discuss the characteristics of the parameters.
(c) Repeat with Ub ( +, , 3 ) = exp [i + J+ ] exp[i 3 J3 ] exp[i J ].
SOLUTION 4.16 We can rewrite an element of su(2)C in the form
1 J1 + 2 J2 + 3 J3 = + J+ + J + 3 J3
1
J = J1 iJ2 ,
= ( 1 i2 )
2
1 , 2, 3 real
+ =
q
q
= 12 + 22 + 32 = 4+ + 32 .
In the standard parameterization and labeling, each element i Ji is
mapped onto
U ( , n) = exp[i( 1 J1 + 2 J2 + 3 J3 )]
sin(/2)
sin /2
cos 2 i3
i( 1 i2)
=
.
/2)
i( 1 + i2) sin/2 cos 2 + i3 sin(
(a) The exponential mapping of + J+ + J + 3 J3 leads to
Uo ( +, , 3 ) = exp[i( + J+ + J + 3 J3 )]
sin(/2)
sin /2
cos 2 i3
2i+
.
=
sin(/2)
sin /2
2i
cos 2 + i3
61

(b) We have in this case
Ua ( + , , 3 ) = exp [i( + J+ + J )] exp [i3 J3 ]
sin /2
i3 /2
2i
cos
e
0
+
2
=
sin /2
+
i3 /2
0
e
2i
cos 2
i3 /2
sin /2 +i3 /2
cos 2 e
2i+ e
.
=
sin /2 i3 /2
+i3 /2
2i e
cos 2 e
Here = 2 + , and the parameters 3 = 3 and + = are related to

and 3 (and hence to the Cartesian parameters i ) for the same transformation (group element) through the Cayley-Klein parameters.
(c) In this case, we have
Ub ( +, , 3 ) = exp (i + J+ ) exp (i 3 J3 ) exp (i J )

1 i + ei3 /2
0
1
0
=
0
1
0
e+i3 /2 i 1
i /2

e 3 + e+i3 /2 i + e+i3 /2
=
.
i e+i3 /2
e+i3 /2
Here all three variables are complex, constrained by the conditions
d = a :
c = b :
e+i3 /2 = ei3 /2 + ei3 /2 ,
i e+i3 /2 = +i+ ei3 /2 .
4.17 (Matrices in non standard parameters) Find

U found in Problem 4.16.
j
Dm0 m [U ]

for the different
j
SOLUTION 4.17 The most direct way is to apply the formula Dm,m ( a, b )
given in GTAPP 4.5, with the appropriate a, b found in Problem 4.16. In
the cases where U is expressed in terms of the standard parameters i , or
in terms of and 3, the coefficients a and b consist each of two terms,
which make their binomial expansions rather messy. Thats why it is useful
to consider alternatives. For Ua and Ub , the elements a (or a ) and b consist
each of a single term, and the general formula applies without difficulty.
4.18 (j = 1/2 rotation matrices) (a) Write down the matrices for the same
general rotation in the Euler angles, R ( , , ), and the angle-axis parameters, R ( , n), both in the 2-dimensional j = 1/2 angular-momentum basis.
(b) By comparing the expressions for D1/2 [ R ( , , )] and D1/2 [ R ( , n)],
derive the relationship between the two sets of parameters.
62
CHAPTER 4
SOLUTION 4.18 (a) For the Euler angles, we will use the matrix given in
GTAPP 4.5 (both rows and columns are labeled in the order (+, )):
0
1/2
im 1/2
Dm
dm0 m ( ) eim
0 m [ R ( , , )] = e
0
= eim h 12 m0 | ei2/2 | 21 m i eim

i(+)/2

e
cos( /2) ei()/2 sin ( /2)
=
.
e+i() /2 sin ( /2)
ei(+)/2 cos( /2)
For the , n parameters, we will use the matrix found in Solution 4.11,
in which 3 is diagonal:

cos ( /2) i sin ( /2) cos
i sin( /2) sin ei
1/2
Dm0 m [ R ( , n)] =
.
i sin( /2) sin ei
cos( /2) + i sin ( /2) cos
(b) These two matrices are directly comparable. If they represent the
same rotation, they are identically equal, and we can then derive a relationship between their respective parameters. We shall call R D1/2 [ R ( , , )]
and T D1/2 [ R ( , n)].
(i) TrR = TrT yields
cos
= cos
cos .
2
2
2
From this equation we obtain a result to be used in the next step:

sin2 ( /2) sin2 ( /2) = sin2
cos2 .
2
2
(ii) R+ R+ = T+ T+ yields
sin2 =
sin2 ( /2)
sin2 ( /2)
from which we have

tan2 =
sin2
sin2
=
1 sin2
sin2 ( /2) sin2 ( /2)
tan2 ( /2)
.
sin2 (( + ) /2)
(iii) R+ + R+ = T+ + T+ yields
2i sin
+
sin cos = 2i sin cos
,
2
2
2
which implies = + /2. The results found here agree with those
obtained in Problem 4.10.
63
4.19 (Schwingers model) J. Schwinger gave a model of the algebra of angular momentum based on the algebra of two independent harmonic oscillators represented by ( a+, a+ ) and ( a , a ), which satisfy the commutation
relations typical of uncoupled harmonic oscillators:

a+ , a+ = 1,
a , a = 1,
(4.5)

a+ , a = 0,
a , a+ = 0.
(4.6)
(a) Define the number operators N+ = a+ a+ and N = a a . Show

that they commute with one another, and thus have common eigenkets
| n+, n i of respective eigenvalues n+ , n where the vacuum ket |0, 0i |0i
is defined by a+ |0i = 0, and a |0i = 0. Construct the normalized ket
| n+, n i from the normalized vacuum state |0, 0i.
(b) Define the operators J+ = a+ a , J = a a+, J3 = 21 ( a+ a+ a a ),
2
J = 12 ( J+ J + J J+ ) + J32. Derive their bracket rules.
(c) Evaluate J 2 | n+ , n i, J3 | n+ , n i, J+ | n+, n i, and J | n+ , n i. Interpret | n+ , n i as a vector | jm i describing a system of spin j in state m.
(d) From the rotation property of the angular momentum kets | jm i, obj
tain the explicit formula for the rotation functions dm0 m ( ).
SOLUTION 4.19 (a) N+ and N are clearly Hermitian, and further commute
by (4.6). They have common eigenkets, and their eigenvalues are real:
N+ | n+, n i = n+ | n+, n i,
N | n+ , n i = n | n+, n i .
These kets are orthogonal to one another and assumed normalized to 1. In

addition, since the norm of a | n i is positive and n = h n | N | n i, the
numbers n must be both
non negative: n 0. From (4.5), N also satisfy

[ N , a ] = a, and N, a = a . Consider first the (+) operators:
h
i
N+ a+ | n+ i = ( N+, a+ + a+ N+ )| n+ i = ( n+ + 1) a+ | n+ i.
And similarly, N+ a+ | n+ i = ( n+ 1) a+ | n+ i. Therefore, a+ | n+ i is an eigenket of N+ of eigenvalue n+ 1, and so must be proportional to | n+ 1i,

up to a normalization factor. Similarly a+ | n+ i is proportional to | n+ + 1i.
To find the normalization factor, first notice that h n+ | a+ a+ | n+ i = n+ . So
that we have a+ | n+ i = n+ | n+ 1i If we keep on applying a+ on the ket

2
3
| n+ i, we will get ap
+ | n+ i, a+ | n+ i,pa+ | n+ i, . . . , with successively smaller
coefficients n+ ,
n+ ( n+ 1), n+ ( n+ 1)( n+ 2), . . ., until the sen
quence terminates
with
a++ | n+ i = n+ ! |0i. If the process is reversed with
a+ | n+ i =
n+ + 1| n+ + 1i, then starting from the vacuum, we will get
( a+ )n+ |0i = n+ ! | n+ i. And so, assuming |0i is normalized, we have the
eigenkets of N+, all normalized to one,
| n+ i =
1
( a+ )n+ |0i .
n+ !
64
CHAPTER 4
Obviously the same reasoning applies to the () type oscillators as well,

and we have the simultaneous normalized eigenkets for N+ and N given
by
1
| n+ , n i =
( a+ )n+ ( a )n |0i.
n+ !n !
(b) Given (4.5)(4.6), we can prove [ J3, J ] = J and [ J+, J ] = 2J3,

which are the usual angular-momentum commutation relations. In addition, defining N = N+ + N , we have the quadratic Casimir operator:

1
N N
2
2
J = ( J+ J + J J+ ) + J3 =
+1 .
2
2
2
Let n = n+ + n , we can easily obtain

n n
J 2 | n+ , n i =
+ 1 | n+ , n i,
2 2
1
1
J3 | n+, n i = ( N+ N )| n+, n i = ( n+ n )| n+, n i,
2
2
q
J+ | n+, n i = a+ a | n+, n i =
J | n+, n i = a a+ | n+, n i =
n ( n+ + 1)| n+ + 1, n 1i,
n+ ( n + 1)| n+ 1, n + 1i.
Notice that these operations leave the sum n = n+ + n unchanged.

We can interpret the quantum oscillators as particles of spin 1/2: states
of spin up (+ 1/2) are associated with type (+) operators, and states of spin
down ( 1/2) with type () operators; thus | n+, n i represents a state with
n+ spin up particles and n spin down particles. This is expressed by the
fact that the eigenvalue of J3 in this state is 1/2( n+ n ) m and the
eigenvalue of J 2 is n/2( n/2 + 1) j( j + 1). The operator J+ destroys one
unit of spin down and creates one unit of spin up, thus increasing the z
component of angular momentum by one unit. The operator J destroys
one particle of spin up and creates one particle of spin down, which has
the effect of reducing the z component of the angular momentum by one
unit.
Now making the correspondence n/2 7 j, ( n+ n ) /2 7 m, we write
| jm i = Njm ( a+ ) j+m ( a ) jm |0i,
Njm = p

1
( j + m)! ( j m)!
(4.7)
as the eigenket of J 2 and J3, with eigenvalues j( j + 1) and m.

From the theory of angular momentum we know that when we add,
for example, the spins of two spin 1/2 particles, we obtain states with spin
j = 1 and 0; if we have three such particles, we have states with spin j = 3/2
65
and 1/2. Generally, adding the spins of 2j spin 1/2 particles, we obtain states
with angular momentum j, j 1, j 2, . . . , 0 (or 1/2). In Schwingers model
of angular momentum, we have a construct of 2j spin 1/2 particles, and
obtain only states of angular momentum j. The reason is that the operators a , a that create or destroy quanta of spin 1/2 are bosonic operators,
satisfying commutation (not anticommutation) relations, and the n particle
quantum states (4.7) are totally symmetric under permutation, with total
angular momentum j = n 1/2 = n/2.
j
(d) As we are concerned only with the rotation function dm0 m ( ), we
need to consider rotations through the angle about the y-axis:
U [ R (0, , 0)] = exp(iJ2 ).
We want to examine how (4.7) transforms under U [ R ]:
U [ R ]| jmi = Njm
U [ R ] a+U 1 [ R ]
j+m
U [ R ] aU 1 [ R ]
jm
U [ R ]|0i .
(4.8)
(Notice that on the LHS, U [ R ] should be regarded as a rank 2j tensor operator applied on a direct-product vector.) Examine now each of the factors on the RHS. First, U [ R ]|0i = |0i, because the vacuum is invariant (i.e.
Ji |0i = 0). Next evaluate various commutators of a+ with J2 = ( a+ a
a a+ ) / (2i):
i
i
1
1 h
a a+ , a+ = a ,
J2, a+ =
2i
2i
h
i

i
1 h
1
J2, J2, a+ =
J2, a = a+ .
2i
4
h
Keeping on in this way bracketing more factors of J2 , we obtain either a+ or

a . Summing terms in the exponential series of U [ R ] a+U 1 [ R ], we obtain
(and in a similar way for a ):
U [ R ] a+U 1 [ R ] = cos( /2) a+ + sin ( /2) a ,
U [ R ] aU 1 [ R ] = cos( /2) a sin ( /2) a+ .
These results are equivalent to giving the rules for rotating spin-up and
spin-down vectors ( a |0i ) through the angle about the y axis. Substituting them into (4.8) and using the binomial expansion twice, we get
p
( j + m)! ( j m)!
U [ R ]| jm i =
[ a+ cos ( /2)]j+m
( j + m ) !!( j m ) !!
[ a sin ( /2)][ a+ sin ( /2)]jm [ a cos( /2)] |0i .
66
CHAPTER 4
Let us replace by m0 , defined as m0 = j , then the above becomes

U [ R ]| jm i =
( a+ ) j+m ( a ) jm
p
0 ( j + m0 )! ( j m0 )!
m
p
( j + m0 ) ! ( j m0 ) ! ( j + m ) ! ( j m ) !
m 0 m +
()
! ( j m0 ) !( j + m ) ! (m0 m + ) !
(cos( /2))2j+mm 2 (sin( /2))2+m m |0i .
By definition, the matrix representation of U [ R ] in the angular momentum

basis is given by the usual equation:
U [ R ]| jm i =
0 | jm0 i Dm0m [ R] .
m
In our case, R = R (0, , 0), and so Dm0 m [ R (0, , 0)] = dm0 m ( ), and hence
the above result should be compared with
U [ R (0, , 0)]| jmi =
0 | jm0 i dm0m ( ) ,
m
which yields Wigners formula for the rotation functions:

j
d m0 m ( )
= ()
m 0 m +
( j + m0 ) ! ( j m0 ) ! ( j + m ) ! ( j m ) !
! ( j m0 ) !( j + m ) ! (m0 m + ) !
0
[cos( /2)]2j+mm 2 [sin( /2)]2+m m .

Its transpose is given by
j
dmm0 ( ) = ()mm dm0 m ( )

p
( j + m0 ) ! ( j m0 ) ! ( j + m ) ! ( j m ) !
= ()
! ( j m0 ) ! ( j + m ) !( m0 m + ) !
[cos( /2)]2j+mm 2 [sin( /2)]2+m m .
This agrees with the formula obtained by other means in the Chapter.

4.20 (Weight diagrams) We have studied in this Chapter how to decompose tensor products of the kind Syma V (1) Symb V (1) , where V (1)
= C2
is the standard (two-dimensional) representation of sl(2, C ). In this problem we want to apply the same approach to find the decomposition of
symmetric powers of V (2) , i.e. Syma V (2) . Here, V (n) = Symn V (1) is the
(n + 1)-dimensional irreducible representation of sl(2, C ). In order to do
67
this, first find the decomposition of Sym2 V (2) and Sym3 V (2) ; then generalize to Syma V (2) .
SOLUTION 4.20 (a) Although Sym2 V (1) is irreducible, Sym2 V (2) is not, as
we will see now. Take as a basis for V (2) the three objects x, z, y, then
Sym2 V (2) consists of the six symmetric objects xx, xz, xy, zz, zy, yy (symmetric means xy = yx, xz = zx, etc.). So we have in all six eigenvalues
for J3, namely, 2, 1, 0 (twice), 1, 2. The presence of a multiplicity of two
tells us that this representation is not irreducible. The quintet 2, 1, 0, 1, 2
composes V (4) , while the remaining 0 is V (0) . In terms of spherical harmon2
ics, the first set corresponds to Dm
and the second to D00 . The decomposition is Sym2 V (2) = V (4) + V (0) :
s
2
s
1
se
0
s
2
se
1
se
0
s
1
s
2
(b) As for Sym3 V (2) , it consists of the symmetric triple products of x, z, y,

namely, x3 , x2 z, x2 y, xz2, xzy, xy2, z3 , z2 y, zy2 and y3 , corresponding to the
eigenvalues 3, 2, 1 (twice), 0 (twice), 1 (twice), 2, 3. It is most likely
(dimensional considerations) that they belong to the irreducible representations V (6) and V (2) . So we have the relation Sym3 V (2) = V (6) + V (2) :
s
3
se
1
s
2
s
3
(c) Finally, Symn V (2) , for some positive integer n, consists of all the homogeneous n order polynomials in x, y, z. The number of symmetric monomials of the form xa yb zc with a + b + c = n is d(n) = 21 ( n + 1)( n + 2), for
any integer n. In other words, there are d(n) eigenspaces m , with m =
n, n + 1, . . . , n 1, n, with various multiplicities.
The maximally symmetric space contained in Symn V (2) is V (2n) , having
the eigenvalues n, n + 1, . . . , n 1, n. Its complement must have the
dimension
1
d(n) (2n + 1) = ( n 1) n = d(n2) ,
2
corresponding to the number of the remaining eigenvalues. The maximally
symmetric subspace of V (2n) is therefore V (2(n2)) ; it has dimension 2n 3.
The number of remaining eigenvalues is given by
d(n2) (2n 3) =
1
( n 3)( n 2) = d(n4) .
2
The representation of dimension 2( n 4) + 1 is V (2(n4)) . The sequence

terminates either with V (2) or V (0) . So the symmetric n-powers of V (2) is
fully reducible, with the decomposition into a direct product given by
Symn V (2) = V (2n) V (2n4) V (2) ( or V (0) ) .
68
CHAPTER 4
We can check the result by calculating the dimensions. As we have seen,

Symn V (2) has dimension d(n) = 21 ( n + 1)( n + 2). It must match the sum of
the dimension on the RHS. When n is even, that sum is
n/2
n/2
n/2
1
(
2n
4k
+
1
)
=
(
2n
+
1
)
1
k = 2 (n + 1)(n + 2) .
k =0
k =0
k =1
When n is odd, the sum has the upper limit k = ( n 1) /2, and we still
obtain
(n1)/2
k =0
(2n 4k + 1) = (2n + 1)(
(n1)/2
k =0
1) 4(
(n1)/2
k =1
The two sides agree in dimensions.
k) =
1
( n + 1)( n + 2) .
2
4.21 (Reduction of direct product) Calculate the ClebschGordan (CG) coefficients for the direct product of irreducible representations of SO(3): (a)
D1/2 D1 , and (b) D1 D1 . A good way is to make use of the properties
of states of highest weight and the raising/lowering operators.
SOLUTION 4.21 Consider the addition rule J = j I + I j0 of angular mo0
mentums, corresponding to the reduction of the product D j D j = D J
(J being summed over | j j0 |, | j j0 | + 1, . . . , j + j0 ). In the representation
0
space of D j D j , the basis state is | jm i| j0 m0 i | jm, j0 m0 i, while in the coupled representation D J , the basis state |( jj0 ) J M i; both are eigenstates of J3 ,
respectively, of eigenvalue m + m0 and M. The state of the highest weight in
0
D j D j is | jj, j0 j0 i; it is also the state of highest weight in the representation
0
space of D j+j , so that
|( jj0 ) j + j0 , j + j0 i = | jj, j0 j0 i .
Applying powers of J on this state, we can obtain all vectors | j + j0 , M i. We
proceed similarly in the subspace orthogonal to the representation space
0
of D j+j , seeking the highest weight state belonging to a different D J and
using J to generate |( jj0 ) J M i for M = J, J 1, . . . , J. As j, j0 are fixed,
we can use the simplified notations: | jm, j0 m0 i | m : m0 i and |( jj0 ) J M i
| J, M i, suppressing explicit reference to j, j0 . (Note that the two bases are
differentiated by the distinctive signs (: and ,).) The following formula will
69

be useful:
J | J, M i = [( J + M )( J M + 1)]1/2 | J, M 1i
= ( j I + I
=
mm
0
j
)
j0
m= j m = j0
| m : m0 ih jm, j0 m0 | J M i
[( j + m )( j m + 1)]1/2 | m 1 : m0 i
o
+ [( j0 + m0 )( j0 m0 + 1)]1/2 | m : m0 1i h jm, j0 m0 | J M i
(a) Product D1/2 D1 = D3/2 D1/2 .

. Representation J = 3/2:
| 3/2, 3/2i = | 1/2 : 1i,

| 3/2, 1/2i = (1/ 3) J | 3/2, 3/2i = (1/ 3)[ 2| 1/2 : 0i + | 1/2 : 1i]
| 3/2, 1/2i = (1/2) J| 3/2, 1/2i
= (1/ 12)[2| 1/2 : 1i + 2| 1/2 : 0i + 2| 1/2 : 0i]
= (1/ 3)[| 1/2 : 1i + 2| 1/2 : 0i],

| 3/2, 3/2i = | 1/2 : 1i .
. Representation J = 1/2: The highest weight state is | 1/2, 1/2i, which

must be orthogonal to | 3/2, 1/2i. Its component | 1/2 : 0i has a positive coefficient, consistent with the phase convention adopted for the CG coefficients.
| 1/2, 1/2i = (1/ 3)[| 1/2 : 0i 2| 1/2 : 1i

| 1/2, 1/2i = J | 1/2, 1/2i

= (1/ 3)[ 2| 1/2 : 1i + | 1/2 : 0i 2| 1/2 : 0i]

= (1/ 3)[ 2| 1/2 : 1i | 1/2 : 0i] .
(b) Product D1 D1 = D2 D1 D0 .
. Representation J = 2:
|2, 2i = |1 : 1i,
|2, 1i = (1/2) J|2, 2i = (1/ 2)[|1 : 0i + |0 : 1i]
|2, 0i = (1/ 6)|2, 1i
= (1/ 6)|1 : 1i + |0 : 0i + |0 : 0i + | 1 : 1i]
= (1/ 6)|1 : 1i + 2|0 : 0i + | 1 : 1i]
|2, 1i = (1/ 2)[|0 : 1i + | 1 : 0i]

|2, 2i = | 1 : 1i .
70
CHAPTER 4
. Representation J = 1: The state of highest weight |1, 1i can be found

from its orthogonality to |2, 1i found above, and the phase convention of
the C-G coefficients. Then, apply J on |1, 1i.
|1, 1i = (1/ 2)[|1 : 0i |0 : 1i]
|1, 0i = (1/ 2) J |1, 1i
= (1/ 2)[|1 : 1i + |0 : 0i |0 : 0i | 1 : 1i]
= (1/ 2)[|1 : 1i | 1 : 1i]
|1, 1i = (1/ 2) J |1, 0i = (1/ 2)[|0 : 1i | 1 : 0i]
. Representation J = 0: This representation has just one basis state

|0, 0i; it is a linear combination of three | m : m0 i vectors:
|0, 0i = a |1 : 1i + b |0 : 0i + c | 1 : 1i.
J+ annihilates this state:
(1/ 2) J+ |0, 0i = a |1 : 0i + b |1 : 0i + b |0 : 1i + c |0 : 1i = 0,
which yields a = b = c. The normalized state is
|0, 0i = (1/ 3)|1 : 1i |0 : 0i + | 1 : 1i.

It is orthogonal to both |2, 0i and |1, 0i as it should.
j
Tm
4.22 (Rotation of tensor product) Show that

= h` , s| jm i X` Ys transforms as a rank j tensor if X` and Ys transform as tensors of rank ` and s.
SOLUTION 4.22 X` and Ys transform as tensors of rank ` and s means
U [ R ] Ys U [ R1 ] =
U [ R ] X` U [ R1 ] = 0 X` 0 D` 0 [ R ],
0 Ys0 Ds 0 [ R].
We have to show that Tj transforms like a tensor of rank j. Making use of

the CG series for D j and the orthogonality of the CG, we obtain under an
arbitrary rotation R:
j
UTm U 1 = h` , s| jmiUX` UU 1 Ys U 1
0 0 h` , s| jmi D` 0 Ds 0 X` 0 Ys0
0 0 h` , s| jmi 0 h` 0, s0| J M0ih` , s| J Mi DM0M X` 0 Ys0
JMM
h` 0, s0| J M0i DM0 m X` 0 Ys0 = 0 TM0 DM0 m .

0 0 0
This completes our proof.
71
Chapter 5. sl(3, C )
5.1 (Matrices Eij ) The matrices Eij have a single non-zero entry, equal to 1,
at row i, column j, i.e. ( Eij )ab = ia jb . Rewrite the basic elements of sl(3, C )
in terms of 3 3 matrices Eij , and check their commutation relations.
SOLUTION 5.1 The basic eight elements of sl(3, C ), in terms of the matrices
Eij , are: T1 = E11 E22 , T2 = E22 E33 , E1 = E12 , E2 = E23 , E3 = E13 ,
F1 = E21 , F2 = E32 , and F3 = E31 .
From their definition, Eij satisfy Eij Eab = Eib ja . In particular, with no
summation over repeated indices, Eaa Eab = Eab . These relations can be
used to check the commutation relations among the Ti , Ei , and Fi .
5.2 (Spin and hypercharge) In physics, the CSA of su (3)C is often defined
in a different basis: tz = 1/2t1 and y = 1/3t1 + 2/3t2 . The other elements of
the algebra remain the same as we have in GTAPP, with new notations:
t+ = e1 , t = f 1, u+ = e2, u = f 2 , v+ = e3 , and v = f 3 . In the
standard representation, where z 7 Z, they are related to the 3 3 Gell
Mann matrices i by Tz = 1/23 and Y = 1/ 38 , T = 1/2( 1 i2 ),
U = 1/2( 6 i7 ), and V = 1/2( 4 i5 ).
(a) Calculate the quadratic traces Tr( hi h j ) in the two bases of the CSA.
(b) Calculate the roots in the physical basis h = { tz, y }.
(c) Calculate the Killing forms ( y : y ), ( tz : tz ), and ( y : tz ).
(d) Calculate hi from its definition, and then the inner products h i , j i,
with i, j = 1, 2, 3.
SOLUTION 5.2
(a) In our basis Tr T1 2 = Tr T2 2 = 2, Tr( T1 T2 ) = 1. In the physical
basis: Tr Tz2 = 1/2, Tr Y 2 = 2/3, Tr (YTz ) = 0.
(b) We display the commutation relations of tz and of y with various
elements of the algebra in the following table
tz
y
tz
0
0
y
0
0
t+
t+
0
t
u+
t 1/2u+
0
u+
(00) (00) (10) (01)
1/2
1
1/2u
u
1/2
( 1 )
v+
1/2v +
v+
1/2
1
v
1/2v
v
1
(1/2 )
t , u , v are simultaneous eigenvectors of ad tz and ad y with nonzero

eigenvalues, and so are the root elements of the algebra. From the first two
rows, the roots are obtained and given in the last row. The positive roots,
relative to tz , y, are 1 = (1, 0), 2 = ( 1/2, 1), and 3 = ( 1/2, 1) = 1 + 2 ;
the other three are their negatives.
(Note: H. Georgi Ch. VII uses H1 = tz and H2 = 3y/2 for the

CSA. So
i
1
2
1
3
our roots are written in his basis as 7 (1, 0), 7 ( /2, /2), and
72
CHAPTER 5
3 7 ( 1/2, 3/2). Further, he chooses 3 and 2 for the simple

roots, so that 1 = + . His H1 axis is along 1 and H2 perpendicular to
it, and so is perpendicular to the line 2 3.)
(c) We calculate the Killing form directly from its definition, ( x : y ) =
Tr (adx ady ), with adx ady = [ x, [ y, ]]. For ( tz : tz ), we have
[ tz , [ tz, tz ]] = 0, [ tz, [ tz, y ]] = 0, [ tz, [ tz, t ]] = t , [ tz, [ tz, u ]] = 1/4u ,
[ tz, [ tz, v ]] = 1/4v . So that
( tz : tz ) = 0 + 0 + 1 + 1 + 1/4 + 1/4 + 1/4 + 1/4 = 3. Similarly from the
commutation relations, and in the same order, we have
( y : y ) = 0 + 0 + 0 + 0 + 1 + 1 + 1 + 1 = 4,
( tz : y ) = 0 + 0 + 0 + 0 12 12 + 12 + 12 = 0. If the bilinear forms among
the t1 and t2 are assumed known, then the above results are arrived at more
quickly. Note that ( tz : y ) = 0 tells us that tz y.
(d) Let hi = ci tz + di y with i = 1, 2, 3. We know, on one hand, ( hi :
k ) = i ( k ), and, on the other hand, ( ci tz + di y : tz ) = 3ci and ( ci tz + di y :
y ) = 4di. With the roots known from (b), we obtain ci and di , leading to
h1 = 1/3tz , h2 = 1/6tz + 1/4y, h3 = 1/6tz + 1/4y. The inner products
on h0 are obtained via the Killing forms, h i , j i = ( hi : h j ), leading to
h i , i i = 1/3 (with i = 1, 2, 3), h 1, 3 i = h 2, 3 i = h 1, 2i = 1/6. Clearly
these properties of the roots are independent of the CSA basis used to do
the calculations.

5.3 (Casimir operator) Find the quadratic Casimir operator in sl(3, C ), and
calculate its invariant value in the irreducible representation of highest weight
( n, m ). Hint: Use the properties of the Gell-Mann matrices i .
SOLUTION 5.3 We have sl(3, C ) = h h1 , h2, e1 , f 1, e2, f 2 , e3, f 3 i in GTAPPs notations. In the standard representation z 7 Z, they are related to the
Gell-Mann matrices in su(3)C by 3 = H1, 8 = 1/ 3( H1 + 2H2 ), 1 =

E1 + F1 , 2 = i( E1 F1 ), 4 = E3 + F3 , 5 = i( E3 F3 ), 6 = E2 + F2 ,
7 = i( E2 F2 ). Define Ji = 1/2i , with i = 1, . . . , 8, then they have the
properties: (i) [ Ji, Jj ] = i f ijk Jk, and (ii) Tr ( Ji Jj ) = N ij . Remark: These two
relations are true in any representation, only the value of N varies with the representation . For = (1, 0), with the Gell-Mann matrices, N1,0 = 1/2. For
= (1, 1), with ( Ji ) jk = i f ijk , one has N1,1 = 3. It follows that f ijk is antisymmetric wrt all three indices: f ijk = f jik because of (i); and f ijk = f ikj
because i/2 f ijk = Tr ( Ji Jj Jk Jj Ji Jk ) = Tr ( Jk Ji Jj Ji Jk Jj ) by the trace cyclicity.
Define J2 = Ji Ji . Then [ J2 , Jj ] = Ji [ Ji , Jj ] + [ Ji , Jj ] Ji = i f ijk ( Ji Jk + Jk Ji ) =
0 for all j = 1, . . . , 8. So J2 is a quadratic Casimir operator of su (3)C .
To rewrite it in terms of the canonical basis in sl(3, C ), we use the correspondence written above, and obtain J2 = 1/3( H12 + H22 ) + 1/6( H1 H2 +
H2 H1 ) + 1/2( E1 F1 + F1 E1 + E2 F2 + F2 E2 + E3 F3 + F3 E3 ). The Casimir invariant for the irreducible representation of highest weight ( n, m ) (relative
73
to H1, H2 ) can be calculate by applying J2 on the highest weight vector

v, making use of ( H1 , H2 ) v = ( n, m ) v, H3 v = ( n + m ) v, and Ei v = 0,
Ei Fi v = [ Ei , Fi ] v = Hi v (i = 1, 2, 3). This gives us J2 v = C v, where
Cn,m = 1/3[( n + m )( n + m + 3) nm ].

5.4 The structure constants N of a simple Lie algebra are defined in the
equation [ e, e ] = N e+ if + is a non-zero root. Assuming this is the
case and going into a finite representation, prove the symmetry relations:
N = N = N, = N, = N, .
SOLUTION 5.4 Let E = ( e ) be defined in a finite representation cho = N . The relation N
sen such that E = E and N
= N follows
trivially from the defining equation. Applying Hermitian conjugation on
E
both sides of the same equation leads to [ E , E ] = N
, which
yields: N = N, . The invariance of the Killing form expressed in

the identity [ a, b ] : c ) = ( a : [ b, c ] , or in our case, [ E , E ] : E = E :

[ E , E ] gives us N ++,0 = N ++,0 , or simply N = N, ,
and interchanging the indices, it is equivalent to N = N, . This completes the proof.

5.5 (Weight vectors) Let s = h h, e, f i be an sl(2, C ) subalgebra of sl(3, C ),
and let ( h ) = H, ( e ) = E, and ( f ) = F. Further, let [ H, E] = ( h ) E,
[ H, F ] = ( h) E, and [ E, F ] = H; and finally Hv = ( h )v. (Hint: Review
GTAPP 5.2 Weights and Weight Vectors) Prove for any integer k 0:
(a) [ H, Ek ] = k ( h) Ek and [ H, Fk ] = k( h) Fk;
(b) HEk v = [ ( h) + k (h )] Ekv and HFk v = [ ( h) k (h )] Fk v;
(c) [ E, Fk ] = kFk1 H xk ( h ) Fk1 ; find xk .
S
5.5 (a) By definition, we have [ H, E] = ( h ) E. We assert that
OLUTION

H, Ek = k ( h) Ek, which can be proved by induction:

H, Ek+1 = E H, Ek + H, E Ek = E( k ( h) Ek) + ( h ) EEk
= ( k + 1 ) ( h ) E k +1 .
Similarly, starting from [H, F ] =

(h) F, we have [H, Fk ] = k (h) Fk.

(b) HEk v = Ek Hv + H, Ek v = ( h ) + k ( h) Ek v . This equation tells
us that Ek v is a weight vector with weight ( h) + k ( h).
(c) With the given result, we compute [ E, Fk+1 ] = ( k + 1) Fk H ( xk +
k ) Fk. For the given relation to hold we must have xk+1 = xk + k. Because
[ E, F ] = H, we have x1 = 0, and so xk = k ( k 1) /2.

5.6 (2,0 is group invariant) Consider the irreducible representation of highest weight (2, 0) of sl(3, C ), to be called 2,0 . Enumerate its weights and
weight vectors, and prove that the rep space is group invariant.
SOLUTION 5.6 Let b = { H1, H2 , E1 , E2 , E3 , F1 , F2 , F3 } be the basis of sl(3, C ),
satisfying the standard bracket relations. We call s1 = { H1, E1 , F1 } and
74
CHAPTER 5
s2 = { H2, E2 , F2 }. From GTAPP 5.2, the representation ((2, 0), V ) has

weights (2, 0), (0, 1), (1, 1), (2, 2), (1, 0), and (0, 2), relative to H1
and H2 . They correspond, respectively, to the weight vectors v, F1 v, F2 F1 v,
F12 v, F2 F12 v, and F22 F12 v. We call this set W , and V = 0 W . By definition
Ei v = 0 for i = 1, 2, 3.
To prove that V is group invariant, it suffices to prove that it is invariant
under each of the basis elements b. Each of the vectors in W is a joint
eigenvectors of H1 and H2 , and so W is invariant under these two elements.
(i) F2 and Ei nullify v, and we have F1 v W and F3 v = [ F2 , F1 ] v =
F2 F1 v W . So the action of b leaves v in V .
(ii) Consider now w2 = F1 v. Clearly, Fi w2 W for i = 1, 2. Now,
F3 w2 can be calculated in two ways. First, F3 F1 v = F1 F2 F1 v making use of
[ F3, F1 ] = 0. Second, F3 F1 v = ( F2 F12 F1 F2 F1 ) v making use of F3 = [ F2, F1 ].
Equating the two results gives F2 F12 v = 2F1 F2 F1 v. The end result is F3 w2
W . As for the action of Ei , we can use the bracket relations [ E1, F1 ] = H1,
[ E2, F1 ] = 0, and [ E3, F1 ] = E2 to show that E1 w2 = ? v, whereas E2 and
E3 nullify w2. This takes care of F1 v of W .
(iii) As for the remaining vectors wi of W , one argues by induction.
Assuming the assertion is true for wi , then using the bracket relations, one
proves that the action of b on wi+1 either gives 0 or a vector in W .

5.7 (Representation of highest weight (2, 0)) Construct the irreducible representation of highest weight (2, 0) of sl(3, C ) (also called a2 ). Give the matrix elements of the lowering operators F1 and F2 in a normalized basis of
the representation.
SOLUTION 5.7 The construction is based on the notion of string of weights,

Si ( ) = { + pi , . . . , , . . . , qi }, where i is a simple root, and p, q
are non-negative integers related by q p = ( hi ). The algebra a2 has two
simple roots, 1 = (2, 1) and 2 = (1, 2); and every weight is of the
form = ( 1, 2 ) where i = ( hi ). As = (2, 0) is the highest weight
(corresponding to the weight vector v) of the representation , all other
weights of must be of the form i j .
(i) The strings Si ( ) must have q = 0 since + i is not a weight for
both i = 1, 2. The formula p = q + i produces p = 0 + 2 = 2 for 1 and
p = 0 + 0 for 2 , and we obtain two weights 1 = 1 = (0, 1) (weight
vector F1 v) and 2 = 21 = (2, 2) (weight vector F12 v). On the other
hand, 2 and 31 are not weights of .
(ii) Consider S2 ( 1 ): here p = 0 because 1 + 2 = 1 + 2 is not a
weight, hence q = p + 12 = 1. So 3 = 1 2 = 1 2 = (1, 1) is
a weight (for the vector F2 F1 v).
(iii) Consider S2 ( 2 ): we have here q = p + 22 = 0 + 2 = 2, and so two
new weights, 4 = 2 2 = (1, 0) (vector F2 F12 v) and 5 = 2 22 =
(0, 2) (vector F22 F12 v).
75
(iv) From 3 = (1, 1), we can build two strings. For S1 ( 3), we have
q = p + 31 = 0 + 1 = 1 (p = 0 because 2 is not a weight), but 3 1 =
(1, 0) = 4 is not new. For S2 ( 3) we have q = p + 32 = 1 1 = 0, and
so 3 2 = 1 22 is not a weight.
(v) From 4 = (1, 0), we can check that the string S1 ( 4 ) has q =
p + 41 = 1 1 = 0, and S2 4 ) has q = p + 42 = 1 + 0 = 1, but 4 2 =
(0, 2) is not new.
(vi) And finally, from 5 = (0, 2), we can check that the string S1 ( 5 )
has q = p + 51 = 0, and S2 ( 5 ) has q = p + 52 = 2 2 = 0.
This completes the description of 2,0 . The results are summarized in
the following table ( = (2, 0) is the highest weight defining the irrep 2,0 ):
(2, 0)
(0, 1)
(2, 2), (1, 1)
(1, 0)
(0, 2)
v0
F1 v0
F12 v0 , F2 F1 v0
F2 F12 v0
F22 F12 v0
| , (2, 0)i
| , (0, 1)i
| , (2, 2)i, | , (1, 1)i
| , (1, 0)i
| , (0, 2)i
The last column gives the normalized kets. (2, 2), (0, 1), (2, 0) form a s1
triplet, whereas (0, 1), (1, 1) a doublet and (2, 2), (1, 0), (0, 2) a triplet
of s2 . To find the matrix elements of Fi we generalize a formula in Problem 1. For a string of weights , i , . . . , qi giving a complete multiplet of si , we have
q
Fi | si i = | ( s + 1) i i ( q s )( s + 1).
This formula is valid for one-dimensional weight spaces. In our case, it
gives
F1 | , (2, 0)i = | , (0, 1)i 2, F1 | , (0, 1)i = | , (2, 2)i 2,

F2 | , (0, 1)i = | , (1, 1)i,
F2 | , (2, 2)i = | , (1, 0)i 2, F2 | , (1, 0)i = | , (0, 2)i 2.

5.8 (The adjoint representation (1, 1)) Construct the irrep of highest weight
= (1, 1) of sl(3, C ). Find the matrix elements of F1 and F2 .
SOLUTION 5.8 Proceeding as in the preceding problem, we find
(1, 1)
(1, 2), (2, 1)
(0, 0)
(1, 2), (2, 1)
(1, 1)
v0
F1 v0 , F2 v0
F1 F2 v0 , F2 F1 v0
F22 F1 v0 , F12 F2 v0
F1 F22 F1 v0
| , (1, 1)i
| , (1, 2)i, | , (2, 1)i
| , (0, 0)1i, | , (0, 0)0i
| , (1, 2)i, | , (2, 1)i
| , (0, 2)i .
Here (1, 1) is one-dimensional because F1 F22 F1 = F2 F12 F2 , and so the

matrix elements of F1 and F2 can be found from formula given in the preceding problem. The difficulty lies in the (0, 0) states. As F1 F2 6 = F2 F1 ,
76
CHAPTER 5
(0, 0) is a two-dimensional weight space spanned by independent combinations of F1 F2 v0 and F2 F1 v0. Choose the combinations such that one, called
(0, 0)0, is a singlet of s1 , and the other, called (0, 0)1, is part of the triplet
(2, 1), (0, 0)1, (2, 1) of s1 . Then the (0, 0) state that is part of the triplet
(1, 2), (0, 0), (1, 2) of s2 must be a linear combination of the eigen states
of s1 (0, 0)0 and (0, 0)1:
F2 | , (1, 2)i = a | , (0, 0)1i + b | , (0, 0)0i.
Using the results found in Problem 1, we have a2 + b2 = 2. As (0, 0)0
and (0, 0)1 belong to the singlet and triplet of s1 , we have
E1 F2 | , (1, 2)i = aE1 | , (0, 0)1i + bE1 | , (0, 0)0i = a 2| , (2, 1i.
On the other hand, since E1 F2 = F2 E1 , we also have
E1 F2 | , (1, 2)i =
F2 E1 | , (1,2)i = F2 |(1, 1)i = |(2, 1)i. Hence a = 1/ 2, and, picking a
phase, b = 3/2.
State (0, 0) can also be reached via E2 | , (1, 2)i = c | , (00)1i + d|, (00)0i,
2
2
where
c + d = 2. Proceeding in the same way, we obtain c = 1/ 2 and
d = 3/2. Now, as F2 E2 = E2 F2 H2 , we have the equation
F2 | , (0, 0)1i + 3F2 | , (0, 0)0i = 2 2| , (1, 2)i ,
leading to 2F2 | , (0, 0)1i = | , (1, 2)i, 2F2 | , (0, 0)0i = 3| , (1, 2)i.

5.9 (Dual representation) Let G be a Lie group, and g its Lie algebra.
(a) Consider the dual space V defined wrt some bilinear form h u, v i for
all v V and u V. Let be a representation, : G GL ( V ), and the
: G GL ( V
). Show that ( g) = ( g1 )T for all
dual representation
T
g G, where indicates transposition.
(b) Let be the representation of g on V (i.e. dual to ). What is ( X)
for X g? Check that preserves the Lie algebra products.
(c) Now take the case of sl(3, C ). How are the weights of and related? If ( n, m ) is the highest weight of the irreducible representation ,
what is the highest weight of its dual ?

SOLUTION 5.9 (a) The bilinear form h u, v i must be invariant for any v V,
T
u V and g G: h u, v i = h ( g) u, ( g) vi = h u, ( g) ( g) vi. It follows
T
that ( g) ( g) = I, or ( g) = T ( g1 ) for any g G.
(b) Let g = ex , with x g; then ( g) = e (x) . Near the identity, the
relation h u, v i = h ( g) u, ( g) vi becomes h ( x ) u, vi + h u, ( x) vi. and
so ( x ) = ( x)T . Let g be defined by the products [ xi, x j ] = cijk xk , or
[ ( xi), ( x j)] = cijk ( xk) in any . Making the transposition the product
77
law becomes [ ( xi)T , ( x j )T ] = cijk ( xk)T . But this is [ ( xi ), ( x j )] =

cijk ( xk ), and so the product law is preserved (meaning that is indeed a
representation).
(c) Let be an arbitrary representation. Then in the weight vector basis,
H is diagonal, that is, H = = diag [ 1, 2 , . . .] for H = ( h ) with h h.
Since H T = T = , H = ( h ) = H T has the same set of eigenvalues,
or has the same set of weights as , only with a sign change for all the
weights. On the other hand E = ET = F and F = FT = E, the
and vice
highest weight vector in becomes the lowest weight vector in ;
each invariant
versa. Since to every vector in corresponds a vector in ,
subspace in V corresponds to an invariant subspace in V . Therefore if is
irreducible, so is .
As in GTAPP, call the fundamental representations V (1,0) = h 1, 2 , 3 i
and V (0,1) = h 1, 2 , 3 i, where i have the weights 1 = (1, 0), 2 = (1, 1),
and 3 = (0, 1), and i have the weights i for i = 1, 2, 3 (all with respect to the canonical basis h h1, h2 i for the CSA). Think of the simple representation V (n,m) as being built up from the tensor product Symn V (1,0)
Symm V (0,1) , then the highest weight vector in V (n,m) is 1n 3 m having the
weight n 1 m 3 = ( n, m ). The vector 3n 1 m is the lowest weight vector
having the weight n 3 m 1 = ( m, n ). On the other hand, in V (m,n) ,
the vector 1m 3 n has the highest weight given by ( m, n ). But V (m,n) is the
dual to V ( n, m ), i.e. V
(n,m )
(n,m )
= V (m,n) , and so we see that the highest weight
of V
is the negative of the lowest weight of V (n,m) .
5.10 (Mesons and baryons) Let i correspond to the quarks u, d, s, and i to
d, s. As in Problem 2, the isospin is Tz = 1/23 , and the
the anti-quarks u,
hypercharge Y = 1/ 38 ; in addition, the (electric) charge is Q = Tz + 12 Y.

Given that mesons are pairs of quark-antiquark, and baryons are made up
of three quarks, find the quark content of the meson and of the baryon
octets, and give the values of Tz , Y, and Q for their members.
SOLUTION 5.10 (a) Any tensor product i j in 3 3 may be split into two
parts of definite symmetry of the form: i j = Si j + Mi j , where S = 1/3 i i
and Mi j = i j 1/3i j i i . In terms of the quark flavors, ( 1, 2 , 3 ) =
d, s), the singlet S 1 and the octet Mi j 8
( u, d, s) and ( 1, 2, 3 ) = ( u,
take the form
S = 1/3( u u + d d + s s)
1/3(2u u d d s s)
u d
u s
1/3( u u + 2d d s s)
.
M=
d u
d s
1
s u
sd
/3( u u d d + 2s s)
The quarks u, d, s have, respectively, isospin Tz = ( 1/2, 1/2, 0), hypercharge Y = ( 1/3, 1/3, 2/3), and charge Q = ( 2/3, 1/3, 1/3), whereas the
78
CHAPTER 5
d, s just have the same values with opposite signs in all cases.
antiquarks u,
The singlet S has zero quantum numbers; the different entries of M have
the values for Tz , Y, Q as shown:
1/2, 1, 1
0, 0, 0
1, 0, 1
Tz , Y, Q = 1, 0, 1
0, 0, 0
1/2, 1, 0 .
1
1
/2, 1, 1 /2, 1, 0
0, 0, 0
The values of Tz indicate that there are 4 isospin subrepresentations: T = 0,

T = 1/2 (2 times), and T = 1. The two isospin doublets can be assigned
unambiguously to two pairs of mesons (provided their spin and parity are
also known), and similarly for the two charged isospin vectors (1, 0, 1 and
1, 0, 1). As for its neutral component, it is admixed with the isospin
singlet in the three diagonal entries (there are only two independent fields
because Tr M = 0). For example, if M is to describe the pseudo-scalar
mesons, then we will have the following assignment for the entries of the
matrix:
0
1/ 2 + 1/6
+
K+
M=
K0 .
1/ 2 0 + 1/ 6
K
K 0
2/ 6
(b) 3 3 3 can be reduced in two steps. In the first step the tensor
product i j in 3 3 is written as i j = 12 eijk k + 12 Sij , which combines a
3 tensor, k = ekmn m n , with a 6 tensor, Sij = i j + j i . In the second
step, the product 3 3 and 6 3 are decomposed into 1 8 and 8 10
respectively:
1 c
1
k ( m m ) + [ k c c k ( m m )]
3
3
1
1 ack bm
bck am n
ab c
S = ( e S + e S ) emnk + ( Sab c + Sbc a + Sca b ).
3
3
Putting altogether, we have the product
k c =
1
1
1
i j k = eijk S + eijm Bk m + ( eikm N j m + e jkm N i m ) + Dijk .
6
2
6
The four terms correspond to the representations
1, 8, 8, 10, respectively.
a
a
In particular, the B b octet is given by: 2B b = b a 31 a b ( m m ), or
explicitly in terms of u, d, s:
1
1
2B 1 = uds + usd + dus sud
2B 2 = ( su us ) u
1
2
2B 3 = ( ud du ) u
2B 1 = ( ds sd ) d
2
2
2B 2 = uds dsu + dus + sdu
2B 3 = ( ud du ) d
3
3
2B 3 = usd dsu sud + sdu
2B 2 = ( su us ) s
79
The table of the values of Tz , Y, Q for M is valid for B as well, and so is our
remark about the mixing of the neutral fields. The following table shows
explicitly the different subrepresentations in the octet of the spin 21 baryons:
Y T Tz 1 1/2 0
0 1
0
1 1/2
n
1
1 /2
0 0
0
1/2
1
+
p
0
Chapter 6. Simple Lie Algebras: Structure

6.1 (Find for b2 ) Find all the roots of Lie algebra b2 .
SOLUTION 6.1 The associated Dynkin diagram B2 indicates there are two
simple roots, 1 (short) and 2 (long), producing the Cartan integers a12 =
a21 = 1. The simple roots 1 and 2 are the only roots at level one. From
these, we construct two chains: (i) 1 + p2, with p = a21 = 1, which tells
us that 1 + 2 is a root, but not 1 + 22 ; (ii) 2 + p1, with p = a12 = 2,
which says that 21 + 2 is a root, but not 31 + 2 . So 1 + 2 is the only
(positive) root at level 2, and 21 + 2 is the only (positive) root at level 3.
At the next level, (21 + 2 ) + 1 = 31 + 2 is not a root (as we already
know); and (21 + 2 ) + 2 = 2( 1 + 2 ) is not a root. So there are no roots
at level 4 and higher. So the root system of so (5, C )
= sp(2, C ) is:
1
2
1
2
1
2
= { , }, = { , , ( + ), (21 + 2 )}.
6.2 (Find + for d4 ) We have seen in GTAPP 6.5 that the set of roots of a
simple Lie algebra g can be found from its given fundamental system and
Cartan matrix A. Use this recursive method to find all positive roots of the
d4 Lie algebra.
SOLUTION 6.2 We recall that any positive root can be written = i i k i

where the k i are non-negative integers; the sum i k i = n is called the level
of . The method is based on the fact that if all the roots up to level n are
known, then the roots at higher levels can be found by examining strings of
roots of the type S j ( ) = { + k j | q k p }, where p = q i a ji k i ,
with known q and k i . There are new (positive) roots if and only if p > 0;
they are + j , . . . , + p j .
The Lie algebra d4 has four simple roots, i ; i = 1, 2, 3, 4, such that the
80
Cartan matrix is
CHAPTER 6
2 1
0
0
1
2 1 1
.
A=
0 1
2
0
0 1
0
2
. Level 1: 1, 2, 3, 4.
. Level 2: All roots are of the form i + p j . Since q = 0 and k i = 0, 1,
we will have the possibilities:
1 + p2: p = a21 = 1 1 1 + 2 ,
1 + p3: p = a31 = 0 ,
1 + p4: p = a41 = 0,
2 + p3: p = a32 = 1 2 2 + 3 ,
2 + p4: p = a42 = 1 3 2 + 4 ,
3 + p4: p = a43 = 0 .
. Level 3: Roots at this level are of the form i + p j . Note that 1 + 22
is not a root since we already know that p = 1 for the string 1 + p2; and
similarly 2 + 23 and 2 + 24 are not roots. Hence we need to consider
only the following cases:
1 + p1: p = 1 a11 a12 = 0 ,
1 + p3: p = a31 a32 = 1 1 + 3 = 1 + 2 + 3 ,
1 + p4: p = a41 a42 = 1 1 + 4 = 1 + 2 + 4 ,
2 + p1: p = a12 a13 = 1 2 + 1 = 2 + 3 + 1 ,
2 + p2: p = 1 a22 a23 = 0 ,
2 + p4: p = a42 a43 = 1 2 + 4 = 2 + 3 + 4 ,
3 + p1: p = a12 a14 = 1 3 + 1 = 2 + 4 + 1 ,
3 + p2: p = 1 a22 a24 = 0 ,
3 + p3: p = a32 a34 = 1 3 + 3 = 2 + 4 + 3 .
There are three roots at level 3: 1 = 1 + 2 + 3 , 2 = 1 + 2 + 4 ,
and 3 = 2 + 3 + 4 .
. Level 4: Roots at this level are of the form i + p j .
1 + p2: p = a21 a22 a23 = 0 ,
1 + p4: p = a41 a42 a43 = 1 1 + 4 = 1 + 2 + 3 + 4,
2 + p2: p = a21 a22 a24 = 0 ,
2 + p3: p = a31 a32 a34 = 1 2 + 3 = 1 + 2 + 4 + 3,
3 + p1: p = a12 a13 a14 = 1 3 + 1 = 2 + 3 + 4 + 1,
3 + p2: p = a22 a23 a24 = 0 .
There is a single root at level 4: = 1 + 2 + 3 + 4,
. Level 5: As 21 + 2 + 3 + 4, 1 + 2 + 23 + 4 and 1 + 2 + 3 +
24 are not roots, we need to consider only 1 + 2 + 3 + 4 + p2. Here
the integer p = a21 a22 a23 a24 = 1, so that the only root at level 5 is
= 1 + 22 + 23 + 4 , which also implies that 1 + 32 + 23 + 4 is not
a root. So that at the next level, we need to consider only 3 cases.
81
. Level 6:
+ p1: p = a11 2a12 a13 a14 = 0 ,
+ p3: p = a31 2a32 a33 a34 = 0 ,
+ p4: p = a41 2a42 a43 a44 = 0 .
There are no roots at level 6, or higher.
In conclusion, d4 has 12 positive roots: 1 , 2 , 3 , 4; 1 + 2 , 2 + 3 ,
2 + 4 ; 1 + 2 + 3 , 1 + 2 + 4, 2 + 3 + 4 ; 1 + 2 + 3 + 4 ; and
1 + 22 + 23 + 4 . Thus, d4 has 24 roots, is of rank 4, and order 28. This is
the root system of so(8, C ).

6.3 (Rules for Dynkin diagrams) Identification of the allowed Dynkin diagrams (those that are consistent with the definition of a fundamental system of roots (FSR)) is based on a set of general rules derived from the definition of FSR of semisimple Lie algebras. Prove:
(a) Rule 1: A Dynkin diagram contains more vertices than joined pairs.
It cannot have closed polygons. (A closed polygon, or loop, is a set of points
sequentially connected and the last point is also connected to the first.)
(b) Rule 2: The maximum number of lines issuing from a vertex is 3.
(c) Rule 3: If an allowed Dynkin diagram contains an A2 simple chain
as a sub diagram, then the diagram obtained from by replacing the simple
chain with a vertex point is also an allowed Dynkin diagram.
SOLUTION 6.3 Let = { 1, 2 , . . . , ` } be an FSR, and the corresponding
Dynkin diagram.
Further let ui = i / | i|, so that h ui , ui i = 1 and 2h ui , u j i =
0, 1, 2, or 3 for i 6 = j. Note that 4h ui , u j i2 is the number of lines

joining i and j; if it is equal to 0 the two vertices i, j are not connected.
(a) Let u = `i=1 ui . Since h u, u i > 0, we have ` > 2 i<j h ui , u j i n,
where n is the number of joined vertices. Hence n < ` or Rule 1.
If a diagram contains a subdiagram (1, 2, . . . , k ) that is a loop, e.g. ui is
linked with ui+1 , and uk linked with u1 , then in this subdiagram the number
of vertices is equal to the number of joined pairs, contradicting Rule 1.
(b) Given Rule 1 assume you have a diagram in which normalized vertex u is linked with normalized vertices v1, . . . , vk . No two vi are connected
(otherwise there would be loops); this means h vi , v j i = 0 with i 6 = j. Since
u, v1, . . . , vk are linear independent, they span a space in which there is a
vector v0 normalized to 1 and orthogonal to all other vi . So { v0, . . . , vk }
forms an orthonormal basis, and u may be written as u = ki=0 vi h vi, u i.
Further h u, v0i 6 = 0, otherwise one could write u in terms of the vi , contrary
to assumption. We have then
1 = h u, u i = h u, v0i2 + h u, v1i2 + . . . + h u, vk i2 .
So h u, v1 i2 + . . . + h u, vk i2 < 1, or ki=1 4h u, vi i2 < 4, which tells us that the
number of lines joining vertex u to all vertices vi must be less than 4.
82
CHAPTER 6
(c) Let diagram contain a simple chain, consisting of (normalized)

vertices v1, . . . , vk linked pairwise by basic A2 links, such that 2h vi, vi+1 i =
1, with i = 1, . . . , k 1. Let v = ki=1 vi ; then h v, v i = k + 2 i<j h vi , v j i =
k ( k 1) = 1. Let u be a vertex of , but not in the chain, connected
to any one of the vi , say v j (but to no other, by Rule 1). Then we have
h u, v i = h u, i vi i = h u, v j i, so that 4h u, v i2 = 4h u, v ji2 = 0, 1, 2, or 3. These
two results, h v, v i = 1 and 4h u, v i2 3, mean that v behaves like any other
vertex. The simple chain v1 , . . . , vk is shrunk to vertex v, and a vertex u of
connected to a v j by n lines is now connected to v by n lines.

6.4 (Non allowed diagrams) Show that configurations Fig.6.7(c.1), Fig.6.8
(d.1)(d.2) in GTAPP are not allowed by showing that the bilinear form h, i
defined in the respective spaces h0 is not positive definite.
SOLUTION 6.4 By definition an FSR is a basis of an inner-product Euclidean space V = h0 , and so any non-zero vector v V must be such
that h v, vi > 0.
Configuration Fig.6.7(c.1): Let u1 , . . . , u5 the normalized vectors (roots)
3 4
corresponding to the vertices of the diagram
such that h u1 , u2 i = h u
,u i =
4
5
2
3
1
h u, u i = /2, and h u , u i = 1/ 2. Consider the vector v = 2u1 +
2 2u2 + 3u3 + 2u4 + u5 , which is nonzero. But h v, v i = 0, contrary to the
definition of an FSR.
Configuration Fig.6.8(d.1): We label the vertices such that h ui , ui+1 i =
1/2 for i = 1, 2, 3, 4, 6, and h u3, u6 i = 1/2. Now, take v = 3u3 + 2( u2 +
u4 + u6 ) + u1 + u5 + u7 . But h v, v i = 0, in contradiction.
Configuration Fig.6.8(d.2): We label the vertices such that h ui , ui+1 i =
1/2 for i = 1, 2, . . . , 7 and h u6, u9 i = 1/2, and consider the nonzero vector
v = u1 + 2u2 + 3u3 + 4u4 + 5u5 + 6u6 + 4u7 + 2u8 + 3u9 . It turns out that
h v, v i = 0, in contradiction.
6.5 (Generators of g2 ) Given that the Lie algebra g2 is defined by a 2 2

Cartan matrix, with elements a11 = a22 = 2, a12 = 3, and a21 = 1,
identify its generators.
SOLUTION 6.5 We know that from the Cartan matrix of g2 we can determine its simple (positive) roots 1 and 2, as well as its other four positive roots: 3 = 1 + 2 , 4 = 1 + 3 = 1 + 1 + 2 , 5 = 1 + 4 =
1 + 1 + 1 + 2, and 6 = 2 + 5 = 2 + 1 + 1 + 1 + 2. Note that each
is of the form i1 + . . . + ik such that the partial sums i1 + i2 + . . . + ir
are also roots for each r k.
Applying the structure theory, we know that associated with the simple
roots, 1 and 2 , we may define the subalgebras si = h hi, ei , f i i
= sl(2, C )
with i = 1, 2, such that [ ei, f j ] = ij hi , [ hi, e j ] = aij e j, and [ hi, f j ] = aij e j , so
that the simple roots relative to hi are j ( hi ) = aij .
We need 8 more generators associated with the remaining roots in .
83
We can define them as e3 = [ e1, e2 ], e4 = [ e1, e3 ], e5 = [ e1, e4 ], e6 = [ e2, e5 ],

and similarly f 3, f 4, f 5 , f 6. According to the general recipe, h1 , h2, e1 , . . . , e6,
f 1, . . . , f 6 form a basis for the 14-dimensional algebra g2 , with h1 and h2 a
basis for h, ei a generator of the root space gi , and f i a generator of the root
space gi , for i = 1, 2, . . . , 6.

6.6 (Multiplication table for g2 ) With the same definitions as in Problem 5,
establish the multiplication table for the generators of the Lie algebra g2 .
SOLUTION 6.6 Using a simplified notation: [ ab ] [ a, b ], wefirst
recall
the
Jacobi identity, written in equivalent forms: x [ ab ] = [ xa] b + a [ xb] and

[ ab ] x = a [ bx] b [ ax ] = ada adb x adb ada x.
We begin by calculating the products of h1 , h2 with the other generators,
with the result that ei , f i are eigen vectors with eigenvalues i in the root
system .
For i, j = 1 or 2, we know that [ hi, e j ] = aij e j , and [ hi, f j ] = aij e j, with
the simple roots relative to hi given by j ( hi ) = aij , where aii = 2, a12 = 3,
a21 = 1. For the remaining root vectors, we have for any h = ah1 + bh2:

[ he3] = [ he1] e2 + e1 [ he2] = ( 1( h ) + 2 ( h ))[e1e2] = 3( h ) e3

[ he4] = [ he1] e3 + e1 [ he3] = ( 1( h ) + 3 ( h ))[e1e3] = 4( h ) e4

[ he5] = [ he1] e4 + e1 [ he4] = ( 1( h ) + 4 ( h ))[e1e4] = 5( h ) e5

[ he6] = [ he2] e5 + e2 [ he5] = ( 2( h ) + 5 ( h ))[e2e5] = 6( h ) e6
And similarly for [ h f j] = j ( h ) f j. Since all roots j are known in terms of
j ( hi ) = aij , with i, j = 1 or 2, we obtain the products listed in the following
table:
h1
h2
e1
e2
e3 e4
e5 e6
f1
f2
f3
f4
f5
f6
2e1 3e2 e3 e4 3e5 0 2 f 1
3 f2
f 3 f 4 3 f 5
0
e1
2e2
e3 0 e5 e6
f 1 2 f 2 f 3
0
f5 f6
Since each ei is in root space gi , its adjoint action carries e j to 0 or gi + j ,

such that [ ei e j ] = 0 if i + j 6 . Similarly, we must have [ f i f j ] = 0 if
( i + j ) 6 and [ ei f j ] = 0 if i j 6 . The following tables record
these results, with * denoting the products still to be determined.
84
CHAPTER 6
e1
e2
e3
e4
e5
e6
e1 e2 e3 e4 e5 e6
0 e3 e4 e5 0 0
0 0 0 e6 0
0 0 0
0 0 0
0 0
0
f1 f2
e1 h1 0
e2 0 h2
e3
e4 0
e5 0
e6 0
f1
f2
f3
f4
f5
f6
f3
f4
f1
0
f2
f3
0
f5
0
0
f6
0
f3
f4
0
0
f4
f5
0
f5
0
f6
0
0
0
f6
0
0
0
0
0
0
g2 decomposes under s1 and s2 , and so the adjoint action of e1, f 1 and

e2, f 2 gives direct information on the elements that span the subrepresentations. Therefore, we start the calculation with ade2 and ad f 2 acting on the
other basis vectors of s2 subrepresentations:
[e2 f 3] = [[ e2 f 1 ] f 2] + [ f 1[ e2 f 2]] = 0 + [ f 1h2 ] = f 1

[e2 f 6] = [[ e2 f 2 ] f 5] + [ f 2[ e2 f 5]] = [ h2 f 5 ] + 0 = f 5
[ f 2 e3] = [[ f 2e1] e2] + [ e1[ f 2e2]] = 0 [ e1h2] = e1
[ f 2 e6] = [[ f 2e2] e5] + [ e2[ f 2e5]] = [ h2e5] + 0 = e5
In a similar way, we consider ade1 and ad f 1 acting on the other basis
vectors of s1 subrepresentations:
[e1 f 3 ] = [[ e1 f 1 ] f 2] + [ f 1[ e1 f 2]] = [ h1 f 2 ] = 3 f 2
[e1 f 4 ] = [[ e1 f 1 ] f 3] + [ f 1[ e1 f 3]] = [ h1 f 3 ] + 3[ f 1 f 2] = 4 f 3
[e1 f 5 ] = [[ e1 f 1 ] f 4] + [ f 1[ e1 f 4]] = [ h1 f 4 ] + 4[ f 1 f 3] = 3 f 4
[ f 1 e3] = 3e2
[ f 1 e4] = 4e3
[ f 1 e5] = 3e4
As for the remaining entries, we calculate in the same way, proceeding from the simple to the more complicated cases. For example, for the
products involving e3, f 3 , we have
[ e3 f 3] = ade1 ade2 f 3 ade2 ade1 f 3 = h1 3h2

[ e3 f 4] = ade1 ade2 f 4 ade2 ade1 f 4 = 0 4ade2 f 3 = 4 f 1
85
[ e3 f 6] = ade1 ade2 f 6 ade2 ade1 f 6 = ade1 f 5 0 = 3 f 4

[e3e4] = ade1 ade2 e4 ade2 ade1 e4 = 0 4ade2 e5 = e6
[ f 3 f 4] = ad f 1 ad f 2 f 4 ad f 2 ad f 1 f 4 = 0 ad f 2 f 5 = f 6
[ f 3e4] = ad f 1 ad f 2 e4 ad f 2 ad f 1 e4 = 0 ad f 2 e3 = 4e1
[ f 3e6] = ad f 1 ad f 2 e6 ad f 2 ad f 1 e6 = ad f 1 e5 0 = 3e4
We calculate another one:
[ e4 f 4] = ade1 ade3 f 4 ade3 ade1 f 4 = 4h1 + 4( h1 + 3h2) = 8h1 + 12h2.
The completed results are given in the following table.
e1 e2 e3
e4 e5 e6
f1
f2
f3
f4
f5
f6
e1 0 e3 e4
e5 0 0
h1
0
3 f2
4 f3
3 f4
0
e2
0 0
0 e6 0
0
h2 f 1
0
0
f5
e3
0 e6 0 0 3e2
e1 h3
4 f1
0 3 f4
e4
0 0 0 4e3
0 4e1
h4 12 f 1 12 f 3
e5
0 0 3e4
0
0
12e1
h5 36 f 2
e6
0
0 e5 3e4 12e3 36e2
h6
f1
0
f3
f4
f5
0
0
f2
0
0
0
f6
0
f3
0
f6
0
0
f4
0
0
0
f5
0
0
f6
0
where h3 = h1 + 3h2 , h4 = 8h1 + 12h2, h5 = 36( h1 + h2 ), h6 = 36( h1 + 2h2 ).
6.7 (Positivity of norms in classical series) Check that vectors in the root
spaces associated with Dynkin diagrams A` , B` , C` , and D ` have positive
definite norm squares, as expected of Euclidean spaces.
SOLUTION 6.7 Let i , with i = 1, . . . , ` be the FSR corresponding to a given
allowed diagram, ui = i / | i| be the normalized simple roots, and any
vector in h0 , written as = i xi i =
i | ui , where xi are arbitrary real
i x i |
i
j
numbers. We have 2h u , u i = 1, 2, 3 , with i 6 = j, respectively for
the A2, B2 and G2 basic links. We will show that in each case h , i may be
written as a sum of squares of reals.
Define ak = 1/2[( x1 x2 )2 + ( x2 x3 )2 + + ( xk1 xk )2 ].
. A` : h , i = i x2i ( x1 x2 + + x`1 x` )
= 1/2x21 + a` + 1/2x2` .
. B` : All vertices have weights 2, except ` of weight 1.
1
2
h , i = 2( x21 + + x2`1 ) + x2` `
i=1 x i x i+1 = x 1 + 2a` .
. C` : All vertices have weights 1, except ` of weight 2.
2
h , i = x21 + + x2`1 + 2x2` `
i=1 x i x i+1 2x `1 x `
= 1/2x21 + 2a`1 + 1/2( x`1 2x` )2.
86
CHAPTER 6
. D ` : The branch point `2 is linked with the endpoints `1 and ` .

2
h , i = i x2i `
i=1 x i x i+1 x `2 x `
= a`2 + 1/2x21 + ( 1/2x`2 x`1 )2 + ( 1/2x`2 x` )2 .

6.8 (Positivity of norms in the exceptional algebras) Same question as in the
preceding problem for the five exceptional algebras.
SOLUTION 6.8
. G2: h , i = x21 + 3x22 3x1 x2 = ( x1 3/2x2)2 + 3/4x22.
. F4: h , i = x21 + x22 + 2x23 + 2x24 x1 x2 2x2 x3 2x3 x4 = 1/4[ x21 +
( x1 2x4)2 + ( x1 2x3 + 2x4)2 + ( x1 2x2 + 2x3)2 ].
. E6: h , i = 6i=1 x2i x1 x3 x2 x4 x3 x4 x4 x5 x5 x6 = 1/2[( 1/2x1 +
x2 x3)2 + ( 1/2x1 + x2 + x3 x4 )2 + ( 1/2x1 + x4 x5 )2 + ( 1/2x1 + x5
x6 )2 + ( 1/2x1 + x6 )2 + 3/4x21 ].
. E7: h , i = 7i=1 x2i x1 x3 x2 x4 x3 x4 x4 x5 x5 x6 x6 x7 =
1/2[( 1/2 x 1 + x 2 x 3 )2 + ( 1/2 x 1 + x 2 + x 3 x 4 )2 + ( 1/2x 1 + x 4 x 5 )2 +
( 1/2x1 + x5 x6 )2 + ( 1/2x1 + x6 x7)2 + ( 1/2x1 + x7)2 + 1/2x21 ].
. E8: h , i = 8i=1 x2i x1 x3 x2 x4 x3 x4 x4 x5 x5 x6 x6 x7
x7 x8 = 1/2[( 1/2x1 + x2 x3 )2 + ( 1/2x1 + x2 + x3 x4 )2 + ( 1/2x1 + x4
x5 )2 + ( 1/2x1 + x5 x6 )2 + ( 1/2x1 + x6 x7 )2 + ( 1/2x1 + x7 x8 )2 +
( 1/2x1 + x8 )2 + 1/4x21 ].

6.9 (Maximal roots) Find the maximal roots of e6 , e7 , and e8 .
SOLUTION 6.9
e6 : 1 + 22 + 23 + 34 + 25 + 6 = 1/2( 1 +
2 + 3 + 4 + 5 + 36 ).
e7 : 21 + 22 + 33 + 44 + 35 + 26 + 7 = 27 .
e8 : 21 + 32 + 43 + 64 + 55 + 46 + 37 + 28 = 7 + 8 .

1
6.10 (Data on g2 ) Rank-two
Lie1 algebra g2 is defined by the simple roots
2
2
and related by | | = 3| | and separated by an angle of 150 . Introduce the basis ei (i = 1, 2, 3) for h0 and the dual basis i (i = 1, 2, 3) for h0
obeying i ( e j ) = ij and 1 + 2 + 3 = 0. Express the fundamental co-roots
hi and the fundamental roots and weights i and i , where i = 1, 2 and
i ( h j ) = ij , in the appropriate basis.
SOLUTION 6.10 Choose 1 = 2, then 2 = 1 2 . As h1 and h2 must be

of the form ae1 + be2 + ce3, we can determine the coefficients a, b, c from
j ( hi ) = aij (the Cartan integers) and ( 1 + 2 + 3 )( h ) = 0, leading to h1 =
e1 + 2e2 e3 and h2 = e1 e2. Since 1 and 2 in h0 may be written
m1 + n2 + q3 = ( m q ) 1 + ( n q )2, we determine the coefficients from
i ( h j ) = ij , leading to 1 = 1 + 2 and 2 = 21 + 2 = 1 3 .
87
Chapter 7. Simple Lie Algebras: Representations

7.1 Weyl group for a2 . Consider the Weyl group for a2 (see GTAPP 7.1 Example 1). (a) Give the transformations of the positive roots and an arbitrary
weight under the Weyl group. (b) Give the orbits of the dominant weights
(0, 0), (1,0), (0,1), (1,1), (2,0), (0,2), (3,0) and (0,3).
SOLUTION 7.1 (a) Let ( h ) = ( m1 , m2 ) where mi = ( h i ) be an arbitrary
weight on h0 . Roots weights of the adjoint representation can be denoted
in the same way. From the general definition w = ( h ) , it follows
for the simple roots 1 = (2, 1) and 2 = (1, 2) the basic relations
w1 ( m1 , m2 ) = ( m1 , m2 ) m1 (2, 1) = ( m1, m1 + m2 )
w2 ( m1 , m2 ) = ( m1 , m2 ) m2 (1, 2) = ( m1 + m2 , m2 ).
Hence the results for (a):
wi
w0
w1
w2
w3
w4
w5
w i ( m 1, m 2 )
( m 1, m 2 )
( m1, m1 + m2 )
( m1 + m2 , m2 )
( m2, m1 )
( m1 m2, m1 )
( m 2, m 1 m 2 )
wi 1
1
1
3
2
2
3
wi 2
2
3
2
1
3
1
wi 3
3
2
1
3
1
2
(b) The Weyl orbit of a weight consists of the distinct equivalent weights
of the given weight. From Weyl reflections w ( m1 , m2 ) given above we
obtain the following results (s indicates the orbital size):
(0, 0) (0,0) [s = 1];
(1, 0) (1,0) (-1,1) (0,-1) [s = 3]; (0, 1) (0,1) (1,-1) (-1,0) [s = 3];
(1, 1) (1,1) (-1,2) (2,-1) (-2,1) (1,-2) (-1,-1) [s = 6];
(2, 0) (2,0) (-2,2) (0,-2) [s = 3]; (0, 2) (0,2) (2,-2) (-2,0) [s = 3];
(3, 0) (3,0) (-3,3) (0,-3) [s = 3]; (0, 3) (0,3) (3,-3) (-3,0) [s = 3].

7.2 Weyl group for b2 . (a) Find the Weyl group for Lie algebra b2 . (b) Give
the orbits of the dominant weights (0,0), (1,0), (0,1), (1,1), (2,0), (0,2).
SOLUTION 7.2 The Lie algebra b2 is defined by the Cartan matrix A (or its
inverse A1 )

2 1
1 1/2
1
A=
,
A =
.
2 2
1 1
Its two simple roots are 1 = (2, 2) and 2 = (1, 2) (or equivalently 1 =
2 1 2 2 = 1 2 and 2 = 1 + 2 2 = 2 ). The system of positive
roots + consists of 1, 2, 3 = 1 + 2 = (1, 0), and 4 = 1 + 22 = (0, 2),
the latter being the maximal root.
88
CHAPTER 7
(a) Let = ( m1 , m2 ) be an arbitrary weight, w1 = w1 , and w2 = w2

the Weyl reflections with respect to the simple roots, defined, as usual, by
wi ( m1, m2 ) = ( m1 , m2 ) mi i (no summation intended). Explicitly
w1 ( m1, m2 ) = ( m1, 2m1 + m2 )
w 2 ( m 1, m 2 ) = ( m 1 + m 2 , m 2 )
w1 w2 ( m1, m2 ) = ( m1 m2 , 2m1 + m2 )
w2 w1 ( m1, m2 ) = ( m1 + m2 , 2m1 m2 )
w21 ( m1, m2 ) = ( m1 , m2 ) = w22 ( m1 , m2 ).
Thus we have in general w21 = w22 = 1, w1 w2 = w2w1. We can pick w1

and w2 to generate the elements of the Weyl group W for b2 : 1, w1, w2,
and w1 w2. The element 1 has order 1; whereas 1, w1, w2 have order
2; and w1w2 have order 4. Element w1 can be regarded as reflection in
the horizontal of the square (root diagram), and w1 w2 four-fold rotation
(through an angle of /2) about the (vertical) z-axis. W is of order 8, and is
isomorphic to the symmetry group D4 of the square.
(b) Weyl orbits and their sizes s:
(0, 0) (0,0) [s = 1];
(1, 0) (1,0) (-1,2) (1,-2) (-1,0) [s = 4]
(0, 1) (0,1) (1,-1) (-1,1) (0,-1) [s = 4]
(1, 1) (1,1) (-1,3) (2,-1) (-2,3) (2,-3) (1,-3) (-2,1) (-1,-1) [s = 8]
(2, 0) (2,0) (-2,4) (2,-4) (-2,0) [s = 4]
(0, 2) (0,2) (2,-2) (-2,2) (0,-2) [s = 4].
The orbits of the dominant weights (1,0), (0,1), (2,0) and (0,2) are square,
while the orbit of (1,1) is octagonal.

7.3 Properties of Weyl reflections. Given a Lie algebra g, prove the following
properties of its Weyl reflections:
(a) The Weyl reflections are isometries on h , i.e transformations that
preserve the scalar product.
(b) The reflection w on the hyperplane perpendicular to a simple root
permutes all other positive roots.
(c) Let = 1/2 + , i.e. half the sum of the positive roots, and w 6 = 1
be an arbitrary Weyl reflection. w is a sum of distinct positive roots.
(d) defined in (c) is the sum of the fundamental weights, and so is a
weight.
SOLUTION 7.3 (a) For a Weyl reflection w in any root , direct application
of the definition w = 2 h , i/ h , i produces h w , w i = h , i
for arbitrary , h .
89
(b) Recall first that any root has a unique integral linear expansion in
terms of the simple roots, = i k i i with all k i 0 (if is positive) or all
k i 0 (if is negative). Assume now that 0 (if 0 consider instead
), and let i be a given simple root.
If = i , then wi = . So assume positive 6 = i from now on,
then
wi = ( k i ( h i )) i + k j j .
j6=i
Since 6 = i , there must be some k j > 0, with j 6 = i; since wi is a root, all

k j 0, and so wi is a positive root.
(c) Consider first a reflection wi in a simple root, then from the definition of it follows wi = i, or wi = i . More generally, for any
w W different from the identity element, $ = w obeys w$ = $
and is given by
$ = 1/2
( w ) = 1/2 (1 1) ,
0
0
and so w is the sum of the positive roots that obey w = .

(d) Applying the isometry property in h on h , ii with the reflection
wi for any simple root i , we have h wi , wi i i = h , ii, or 2h , ii =
h i , i i. But this implies ( h i ) = 1 for any and therefore every i . This
means = 1/2 + = i i is half the sum of positive roots or, equivalently, the sum of the fundamental weights. It follows that is a (dominant)
weight, written in the basis { i } as (1, 1, . . . , 1, 1).

7.4 Dual representations. We know that the necessary and sufficient condition for an irrep with highest weight to be self-dual is that its lowest
weight is . Determine the self-duality of the irreps of a2 , b2 , and g2 .
SOLUTION 7.4 a2 : Irrep 10 has highest weight 1 = (1, 0) and lowest

weight 2 = (0, 1), whereas 01 has highest weight 2 = (0, 1) and
lowest weight 1 = (1, 0). Irrep nm has highest weight n 1 + m 2,
that is n times highest weight of 10 plus m times highest weight of 01.
Similarly its lowest weight is equal to n times lowest weight of 10 plus m
times lowest weight of 01, that is n ( 2) + m ( 1) = m 1 n 2, or
( m, n ). For self-duality, ( n, m ) = ( m n ), which is satisfied iff n =
m. If w3 is Weyl reflection wrt positive root 3 , then w3 ( n, m ) = ( m, n ).
So w3 = 1 when restricted to weight subspace ( n, n ).
b2 : Irrep 10 has lowest weight (1, 0), and so is self-dual (or, equivalently, self-conjugate). Irrep 01 has lowest weight (0, 1), and so is selfdual. The lowest weight of irrep nm is n (1, 0) + m (0, 1) = ( n, m ),
and so nm is self dual for every dominant weight. If we look at the Weyl
group W of b2 , we see that there is an element w W such that w2 = 1.
90
CHAPTER 7
Calling w1 and w2 the Weyl reflections wrt the fundamental roots, then
w = w1 w2 . Since w2 ( n, m ) = ( n, m ) for every dominant weight ( n, m ),
every irrep of b2 is self-dual.
g2 : The same reasoning applies. Irrep nm has lowest weight ( n, m )
and so is self-dual. Alternatively there is an element y W of g2 equal to
1, so that y ( n, m) = ( n, m ) for any non negative integers n, m, and so
every irrep of g2 is self-dual. That element y is ( w1 w2 )3 , where w1 = w1
and w2 = w2 are the Weyl reflections wrt the simple roots 1 and 2 .

7.5 Inner products of weights. Inner products of weights can be calculated in
the same way as for roots, since the two entities live on the same space, h0 .
We take sl(3, C ) to illustrate, and set = ( m1, m2 ) for any weight, where
mi = ( hi) and hi are the fundamental coroots. (a) Calculate h defined by
( h : h ) = ( h ), where h is any vector of h0 . (b) Calculate h , 0 i for any
weights , 0 h0 .
SOLUTION 7.5 Let + = { 1, 2, 3 } be the set of positive roots of sl(3, C ),
in which 3 = 1 + 2 , the sum of the two simple roots. Let hi be the
fundamental coroots so that i ( hi ) = 2 with i = 1, 2. Let h = ah1 + bh2
be any vector on h0 . We know that 1 ( h ) = 2a b and 2 ( h ) a + 2b. In
addition from ( h : h0 ) = 2 + ( h ) ( h0 ), we deduce ( h1 : h1 ) = ( h2 :
h2 ) = 12 and ( h1 : h2 ) = 6, or more generally ( h : h0 ) = 12( aa0 + bb0 )
6( ab0 + ba0 ).
(a) From this expression for ( h : h0 ) and the defining relation ( h : h ) =
( h ), we obtain a system of two linear equations which yield the desired
solution: h = ch1 + dh2 = 1/18(2m1 + m2 ) h1 + 1/18( m1 + 2m2 ) h2.
(b) The inner product h , 0i for any weights , 0 h0 can be calculated
from h , 0i = 0 ( h ) = c 0 ( h1 ) + d 0 ( h2 ) = cm01 + dm02, with c, d found in
(a). Hence h , 0i = 1/18(2m1 m01 + 2m2 m02 + m1 m02 + m2 m01 ). Note that the
weight ( m1 , m2 ) has norm square h , i = 1/9( m21 + m22 + m1 m2 ), which is
maximal for a highest weight (with both m1 , m2 0) or a lowest weight
(with both m1 , m2 0).

i
j
7.6 Fundamental weights. Find the expansion coefficients of in { } for
the algebras b3 , d4 , g2 , f4 , e6 , e7 , and e8 .
SOLUTION 7.6 For each algebra, let us define the ` ` matrices A = { aij },
B = A1 , C = { cij }, and D = CB = { dij }. Given that i = j j a ji and
i = i j c ji (where aij and cij are known), we must have i = j j d ji ,
where D is to be found. In the following we list for each algebra C, B and
D = CB.
. b3 :
1
0 0
1 1 1/2
1 1 1/2
C = 1 1 0
B = 1 2 1
D = 0 1 1/2 .
0 1 1
1 2 3/2
0 0 1/2
91
The entries of each column of C give the expansion coefficients in 1 =

1 2 , 2 = 2 3 , 3 = 3 . The entries of each column of B give the
expansion coefficients in 1 = 1 + 2 + 3 , 2 = 1 + 2( 2 + 3 ), 3 =
1/21 + 2 + 3/23 . And finally, the entries of each column of D give the
expansion coefficients in 1 = 1 , 2 = 1 + 2 , 3 = 1/2( 1 + 2 + 3 ).
. d4 :
1
0
0 0
1 1 1/2 1/2
1 1
1 2 1
0 0
1
C=
B=
0 1 1 1
1/2 1 1 1/2
1/2 1 1/2
0
0 1 1
1
1 1 1/2 1/2
0 1 1/2 1/2
D=
0 0 1/2 1/2 .
0 0 1/2 1/2
. g2 :

3/2
1
C=
,
3/2
0

1
/2
0
D=
.
3/2
3

2 3
B=
,
1 2
. f4 :
1/2
1/2
. e6 :
1
/2
1/2
1/2
C=
1/2
1/2
3/2
0 0
0
0 0
1
C=
1/2 0 1 1 ,
1/2 1 1 0
1 3/2
0 1/2
D=
0 1/2
0 1/2
1
1
0
0
0
0
1 0
0
1 1 0
0
1 1
0
0
1
0
0
0
0
0
0
0
0
1
0
D=
2
0
0
4/3
2
1
1
0
2
2
B=
2
1
1
1
.
0
0
0
0
0
,
1
1
0
1
1
1
1
1
3
1
0
1
0
1
2
1
2
1
2
5/3 2 3
3
6
4
2
4
8
6
3
2
4
3
2
4
3
1
5
B=
3
6
4
2
3
6
6
9
6
3
5 6 4 2
6 9 6 3
10 12 8 4
12 18 12 6
8 12 10 5
4 6 5 4
0
0
0
2
2
4/3
0
0
0
.
0
2
2/3
92
CHAPTER 7
. e7 :
C=
1/2
1/2
1/2
1/2
1/2
1/2
1/2
1
1
0
0
0
0
0
1
0
0
0
0
1 1
0
0
0
0
1 1
0
0
0
0
1 1
0
,
0
0
0
1 1
0
0
0
0
1
0
0
0
0
0
1
D=
2
. e8 :
C=
1/2
1/2
1/2
1/2
1/2
1/2
1/2
1/2
1
1
0
0
0
0
0
0
B=
2
2
2
1
2
7/2
4
6
9/2
3
3/2
0
1
1
0
0
1
1
0
0
1
1
2
0
1
1
2
0
1
1
2
0
1
1
2
2 2 2 2 3 2 4 2
0
0
0
2
2
2
3 2
B=
1
0
0
0
0
0
1 1
0
0
0
0
0
1 1
0
0
0
0
0
1 1
0
0
,
0
0
0
1 1
0
0
0
0
0
1 1
0
0
0
0
0
1
0
0
0
0
0
0
0 1 1 0 0
0 1
1 0 0
0 1
1 2 0
1
0
1
1 2 2
D=
0
1
1 2 2
2
0 1
1 2 2
0 1
1 2 2
4 5
7 10 8
0
0
0
0
2
2
2
6
3
4
6
8
6
4
2
4
6
8
12
9
6
3
3
9/2
6
9
15/2
5
5/2
0
0
0
0
0
2
2
0
0
0
0
2
2
2 2
2
3
4
6
5
4
2
0
0
0
0
0
0
2
2
4 5 7 10 8
5 8 10 15 12
7 10 14 20 16
10 15 20 30 24
8 12 16 24 20
6 9 12 18 15
4 6 8 12 10
2 3 4 6 5
0
0
0
0
0
2
2
4
1
3/2
2
3
5/2
2
3/2
6
4 2
9
6 3
12 8 4
18 12 6
15 10 5
12 8 4
8
6 3
4
3 2
7.7 Representations of b2 . Find the irreducible representations of b2 with

highest weights (0,1), (1,0),(0,2) and (2,0). Discuss the representations of
the tensor products 0,1 0,1 , 1,0 1,0, and 0,1 1,0.
SOLUTION 7.7 As in similar problems for a2 (Chapter 4), a solution may

make use of the notion of string of weights spaced by simple weights and
the formula q = p + ( hi). As usual, call si the subalgebra { Hi, Ei , Fi }, with
i = 1, 2, and the simple roots 1 = (2, 2) and 2 = (1, 2).
93
(a) Representation of highest weight = (0, 1). We have successively the

strings of weights:
S2 (0, 1) : (0, 1)(1, 1) ( ( h2 ) = 1: s2 doublet);
S1 (1, 1) : (1, 1)(1, 1) ( ( h1) = 1: s1 doublet);
S2 (1, 1) : (1, 1)(0, 1) (s2 doublet).
The process ends here because the last weight, (0,-1) has no positive
components. So 0,1 has dimension 4, a fact confirmed by the multiplicity
and the W-orbital size of the only dominant weight (0,1). Note that 0,1 is
self-conjugate because its lowest weight is the negative of its own highest
weight; (0, 1) = (0, 1).
The matrix elements of Fi in normalized
p kets can be calculated from the
formula Fi | ki i = | ( k + 1) i i ( q k )( k + 1), with 0 k q,
obtained in GTAPP Chapter 7 Sec. 7.2. In our case, we have
F2 | , (0, 1)i = | , (1, 1)i, F1 | , (1, 1)i = | , (1, 1)i ,
F2 | , (1, 1)i = | , (0, 1)i.
(b) Representation of highest weight = (1, 0).
Strings of weights:
S1 ( 1, 0) : (1, 0)(1, 2) (s1 doublet);
S2 (1, 2) : (1, 2)(0, 0)(1, 2) ( ( h2)=2: s2 triplet);
S1 (1, 2) : (1, 2)(1, 0) (s1 doublet);
The process ends with the weight (1, 0), which has no positive component. The weight (0,0) can be reached from (1,0) only through one path,
and so must have multiplicity 1. There are two dominant weights, (1,0) and
(0,0), each of multiplicity one, and W-orbits of size 4 for one and 1 for the
other. Hence the dimension of 1,0 is 5.
The matrix elements of Fi are given by
F1 | , (1, 0)i = | ,(1, 2)i,
F2 | , (1, 2)i = 2| , (0, 0)i, F2 | , (0, 0)i = 2| , (1, 2)i,

F1 | , (1, 2)i = | , (1, 0)i.
(c) Representation of highest weight = (0, 2).
S2 ( 0, 2) : (0, 2)(1, 0)(2, 2) (s2 triplet);
S1 ( 1, 0) : (1, 0)(1, 2) (s1 doublet);
S2 (1, 2) : (1, 2)(0, 0)(1, 2) (s2 triplet);
S1 (2, 2) : (2, 2)(0, 0)(2, 2) (s1 triplet);
S1 (1, 2) : (1, 2)(1, 0) (s1 doublet);
S2 (2, 2) : (2, 2)(1, 0)(0, 2) (s2 triplet).
We have reached the weights (1, 0) and (0, 2), which signal the end
of the process. There are three dominant weights (0,2), (1,0) and (0,0). The
weight (0,2) being the highest must be single, and so are the equivalents on
its W-orbit. (1,0) can be reached from (0,2) by one path only, and so must
also be single, as are its Weyl equivalents. The weight (0,0) has multiplicity
94
CHAPTER 7
2, because it can be reached via two independent paths. Hence, we have

dim 0,2 = 4 1 + 4 1 + 1 2 = 10. It contains, apart from (0,0), the
weights (2, 2) = 1 , (1, 2) = 2 , (1, 0) = 1 + 2, (0, 2) = 1 + 22 and
their negatives, i.e. all of , and so it is the adjoint representation of b2 . As
(0, 2) = (0, 2), it is self-conjugate.
The calculation of the matrix elements of Fi has only one difficulty in
the double multiplicity of (0,0) as a weight of , but it is a problem we
have met in GTAPP Chapter 7 Sec. 7.2 with the adjoint representation of a2 .
We just follow the same approach. The results (for = (0, 2)) are:
F2 | , (0, 2)i = 2| , (1, 0)i, F2 | , (1, 0)i = 2| , (2, 2)i,

F1 | , (1, 0)i = | ,(1, 2)i,
F2 | , (1, 2)i = 2| , (00)1i,
F1 | , (2, 2)i = | , (00)1i + | , (00)0i,
F1 | , (00)1i =
| , (2, 2)i,
F1 | , (00)0i = | , (2, 2)i,
F2 | , (00)1i =
2| , (1, 2)i,
F2 | , (2, 2)i = 2| , (1, 0)i,
F2 | , (1, 2)i = | , (1, 0)i, F2 | , (1, 0)i = | , (0, 2)i.
It is useful here to have a graphical aid: One may keep track of the
progress of the calculation with, for example, a diagram on the weight lattice, as shown below for 0,2 .
0,2[b2 ]
0, 2
c
@
@
1, 2
@ c1, 0
s
@
@
@
@
@ 2, 2
2, 2
0, 0
@
@s
s
cf
@
@
@
@
@ 1, 2
1,
0
@s
@s
@
@
@ 0, 2
@s
(d) Representation of highest weight = (2, 0). From explicit construction, we see that 20 has four dominant weights: (2,0), (0,2) and (1,0) with
multiplicities equal to one and W-orbital sizes equal to four; and (0,0) with
multiplicity two and W-orbital size one. So it is a 14-dimensional representation, as shown in the accompanying figure.
95
s
s
2,0[b2 ]
s
dq
s
s
s
s
(e) Tensor products 0,1 0,1 , 1,0 1,0 , and 0,1 1,0 . Weights in the
product representations are obtained as pairwise sums of weights of the
product factors. The resulting weights are partitioned into the weights of
irreducible representations, and in this way a likely decomposition is surmised. Confirmation is made by the usual CartanWeyl construction with
the ladder operators Ei , Fi applied on the maximal dominant weight. From
the weights contained in the products, we infer the following decompositions:
0,1 0,1 = 2 0,1 Sym2 0,1,
2 0,1
= C 0,1,
Sym2 0,1
= 0,2;
1,0 1,0 = 2 1,0 Sym2 1,0,

2 1,0
= 0,2
= Sym2 0,1,
Sym2 1,0
= C 2,0.
The Sym2 1,0 weight diagram is presented in the figure (see next page).
As for 0,1 1,0, it contains (1, 1) as the maximal dominant weight, hence
the tensor product must be reducible with the 16-dimensional 1,1 in its
decomposition, the remaining weights belong to 0,1, according to
0,1 1,0
= 0,1 1,1 .
7.8 Representations of g2 . (a) Find the Weyl group for the Lie algebra g2 .
(b) Give the orbits of the dominant weights (1,0), (0,1), (1,1), (2,0), (0,2).
96
CHAPTER 7
s
s
Sym2 1,0 [b2 ] s
s
ce
q
s
s
s
s
(c) Give the weights of the irreducible representation having (0,1) as the
highest weight.
SOLUTION 7.8 The Lie algebra g2 has the following characteristics:
: 1 = (2,
1), 2 = (
3,22) in the basis of the fundamental weights

1
1
2
2
1
3
= /2 + /2 , = 3 .
+ : 1, 2 , 1 + 2 , 21 + 2 , 31 + 2, 31 + 22 .
(a) Call w1 = w1 , w2 = w2 :
w1 ( m1 , m2 ) = ( m1, m1 + m2 ),
w2 ( m1 , m2 ) = ( m1 + 3m2 , m2 ).
Both w1 and w2 are of order 2, whereas w1 w2 is of order 6. Now let
a = w1, b = w2 and c = w1 w2. Then the Weyl group for g2 has the elements:
1, a, c, c2 = ba, c3 = 1, c4 = c, c5 = ba, ac = b, ac2 = bc, ac3 = a,
ac4 = b, and ac5 = bc. The order of the elements are as follows: order
1:1; order 2: 1, a, b, bc; order 6: c, ba. The multiplication table,
with columns and rows labeled in order 1, a, b, c, d = ba, f = bc, 1, a,
b, c, d, f , is divided into quarters (++, +, +, ), each obtained
from any other by changes of signs. The ++ quarter is given by:
1
a
b
c
d
f
1
a b
c
d
f
1
a b
c
d
f
a
1 c
b f d
b
d 1
f
a
c
c f a d
1
b
d
b f
1 c a
f c d a
b
1
The action of the Weyl group on the weights ( m1, m2 ) of g2 is shown in

the accompanying table (e.g. w1 ( m1, m2 ) = ( m1, m1 + m2 ); w1 ( m, 0) =
( m, m ), etc.).
97
1
m1 , m2
m, 0
0, m
m, m
a = w1
m 1, m 1 + m 2
m, m
0, m
m, 2m
b = w2
m1 + 3m2 , m2
m, 0
3m, m 4m, m
c = w1 w2 m1 3m2 , m1 + 2m2 m, m 3m, 2m 4m, 3m
d = w2 w1 2m1 + 3m2 , m1 m2 2m, m 3m, m 5m, 2m
f = w2 w1 w2 2m1 + 3m2 , m1 2m2 2m, m 3m, 2m 5m, 3m
The table must be completed with the results of the action of the six negative elements 1, w1, w2, w1w2 , w2 w1, and w2 w1 w2.
(b) From the results of (a), the W-orbits of the given dominant weights
a
are as follows ( 0 means a 0 ):
a
(1, 0) (1, 1) (2, 1) (2, 1) (1, 1) (1, 0)

b
a
b
a
b
(0, 1) (3, 1) (3, 2) (3, 2) (3, 1) (0, 1)
(2, 0) (2, 2) (4, 2) (4, 2) (2, 2) (2, 0)

b
a
b
a
b
(0, 2) (6, 2) (6, 4) (6, 4) (6, 2) (0, 2)
(1, 1) (1, 2) (5, 2) (5, 3) (4, 3) (4, 1) (1, 1)
(1, 1) (4, 1) (4, 3) (5, 3) (5, 2) (1, 2) (1, 1).

The W-orbital size is 12 for (1, 1), and 6 for the others.
(c) Representation 0,1 : Starting from weight (0,1), we have the strings:
S2 : (0, 1), (3, 1); s2 doublet;
S1 : (3, 1)), (1, 0), (1, 1), (3, 2); s1 quadruplet;
S2 : (1, 1), (2, 1); s2 doublet;
S1 : (2, 1), (0, 0), (2, 1); s1 triplet;
S2 : (3, 2), (0, 0), (3, 2); s2 triplet;
S1 : (3, 2), (1, 1), (1, 0), (3, 1); s1 quadruplet;
S2 : (3, 1), (0, 1); s2 doublet.
The dominant weights (0, 1) and (1, 0) have multiplicity 1 and W-orbits
of size 6; whereas (0, 0) as a weight of 0,1 has multipliciy two and orbital
size equal to one. Hence the dimension of 0,1 is 14. All non-zero weights
are also the roots of the algebra, and so 0,1 is the adjoint representation of
g2 , illustrated in the accompanying diagram (see next page).
7.9 Find the integral expansion i k i i (where i are simple roots) for the
positive roots of the classical algebras a` , b` , c` , and d` . (An error in GTAPP
Ver. 2014-12 has been corrected: positive weights positive roots.)
SOLUTION 7.9
. a` = sl(` + 1, C )
: i i i+1 (i = 1, . . . , `)
+ (`(` + 1) /2):
98
CHAPTER 7
s
s
0,1[g2 ]
s
s
s
sf
s
s
ij i j (1 i < j ` + 1).
= ij = i + i+1 + + j1 (1 i < j ` + 1).
. b` = so(2` + 1, C )
: i = i i+1 , (i = 1, . . . , ` 1), ` = ` .
+ (`2 ):
ij
ij
i j ; + i + j (1 i < j `); i0 i (1 i `).
ij
= = i + i+1 + + j1 (1 i < j `).
ij
+ = i + + j1 + 2 j + + 2` (1 i < j `).
i0 = i + i+1 + + ` (i = 1, . . . , ` 1), `0 = ` .
. c` = sp(`, C )
: i i i+1 , (i = 1, . . . , ` 1), ` 2` .
+ (`2 ):
ij
ij
i j ; + i + j (1 i < j `); i0 2i (1 i `).
ij
= = i + i+1 + + j1 (1 i < j `).
ij
+ = i + + j1 + 2 j + + 2`1 + ` (1 i < j ` 1).
i+` = i + + `1 + ` (1 i ` 1).
i0 = 2i + + 2`1 + ` (i = 1, . . . , ` 1), `0 = ` .
. d` = so (2`, C )
: i i i+1 , (i = 1, . . . , ` 1), ` `1 + ` .
+ (`(` 1)):
ij
ij
i j ; + i + j (1 i < j `);
ij
= = i + i+1 + + j1 (1 i < j `).
ij
+ = i + + j1 + 2 j + + 2`2 + `1 + ` (1 i < j ` 2);
i,`1
+
= i + + `1 + ` (i = 1, . . . , ` 2);
i,+` = i + + `2 + ` (i = 1, . . . , ` 2);
1,`
`
= ` .

+
99
7.10 Dimension of the irreps of a` . Find the dimension of the irreps of a` for
arbitrary `.
SOLUTION 7.10 All positive roots of a` are of the form ij = i j , with
1 i < j ` + 1. As the simple roots are defined to be i = i,i+1 , with
i = 1, . . . , `, all the positive roots may be written
i,j = i + i+1 + + j1 ,
1 i < j ` + 1.
The coefficients k s in the expansion = i,j = s k s s (in terms of simple

roots s ) are given by k s = 1 if i s j 1, and k s = 0 otherwise. So that
the dimension formula for an irrep of a` takes the form
`
dim [a` ] =
`+1
ij
s k s ( m s + 1 )
ij
s k s
i=1 j=i+ 1
`+1
i=1 j=i+ 1
j1
s=i ( m s + 1 )
.
ji
It is the product of factors listed below, with the corresponding ( ij) for each
factor shown in the first column:
(12)(23) . . . (`, ` + 1)
(13)(24) . . . (` 1, ` + 1)
(1, `)(2, ` + 1)
(1, ` + 1)
:
:
:
:
:
:
m 1 +1 m 2 +1
m ` +1
1
1 1
m
+
m
m 1 +m 2 +2 m 2 +m 3 +2
`1 2 ` +2
2
2
m1 +m2 ++m`1 +`1 m2 +m3 ++m` +`1

`1
`1
m1 +m2 ++m` +`
`
We can check that this general formula reproduces the values for ` =
1, 2 given in GTAPP Sec. 7.3. Let us take two further examples:
In sl(4, C ), irrep m1 ,m2 ,m3 has the dimension given by
1
12 ( m1 + 1 )( m2 + 1 )( m3 + 1 )( m1 + m2 + 2 )( m2 + m3 + 2 )( m1 + m2 + m3 + 3 ),
whereas in sl(5, C ), irrep m1 ,m2 ,m3 ,m4 has the dimension

1
288 ( m1 + 1 )( m2 + 1 )( m3 + 1 )( m4 + 1 )( m1 + m2 + 2 )( m2 + m3 + 2 )( m3 + m4 + 2 )
( m1 + m2 + m3 + 3)( m2 + m3 + m4 + 3)( m1 + m2 + m3 + m4 + 4).

In particular, dim111 = 26 and dim1111 = 210 , of the general form 2 p
where p = `(` + 1) / is the number of positive roots of the algebra a` .
Index
a1, sl (2, C ), 4870
a2, sl (3, C ), 7179
representation, 7379, 87
a` , sl (` + 1, C ), 97, 99
A4, 6
character table, 24
abelian group, 2
angular momentum, 57
b2 , so(5, C ), 79, 87
representation, 9296
b` , so (2` + 1, C ), 98
baryon, 77
block diagonalization, 12
c` , sp(`, C ), 98
C 2 C 2, 2
C2 C2 C2 , 2
C3, 14
C 4 C 2, 5
C 6, 3
C 8, 5
Cartan matrix, 80
Casimir operator, 49, 72
Cayley theorem, 2, 7
CH, CampbellHausdorff, 33, 40, 42
character of representation, 17
ClebschGordan coefficient, 68
d4, so (8, C ), 79, 81
d` , so(2`, C ), 98
D2, 4
D2 C2, 5
D3, 3, 4
multiplication table, 14
representation, 16, 21
D4, 5, 6
character table, 24
conjugacy class, 6
normal subgroup, 7
Dn , 8
dimension of
orthogonal group, 31
irreducible representation, 99
symplectic group, 32
unitary group, 31
dual (of representation), 76, 89
Dynkin diagram, 81
e6e8 , 86
Euclidean group, 4248
exponential map, 3336, 38, 44, 58
f4, 90
g2 , 8286, 88, 96
Gell-Mann matrices, 71
group algebra, 18
Heisenberg group, 40
hypercharge, 71
inner product, 15, 90
Lagrange theorem, 2, 3
level (root), 79
Lie bracket, 33, 49
Lorentz group, 39
matrix representation, 27
maximal root, 86
meson, 77
nilpotent matrix, 34
O(1;1), SO(1;1), 31
O(2), SO(2), 27, 29
O(3), SO(3), 59
100
101
INDEX
Pauli matrices, 9
permutation, 3
positivity of norm, 15
in classical algebras, 85
in exceptional algebras, 86
product of representations, 27
Q 8, 4, 5, 9
normal subgroup, 9
quaternion (Lie) group, 50
representation, 11
adjoint, 48
dual, 76, 77, 89
fundamental, 77
irreducible, 17, 73
regular, 13
root, 71, 74, 79
rotation, 1, 27, 38, 5162
S 2 , S 3, S 4 , 5
character table, 26
permutational representation, 26
Schur Lemma, 18, 19
Schwinger model, 63
self duality, 89
SL(1, Q ), 51
SL( n, R ), 34
SO(2;1), 37
structure constant, 73
SU(2), 29, 51
subgroup, 2
normal, 7, 9
symmetric power of representation, 67
T, see A4
tensor product of representation, 67, 77
U(1), 29
U(2), 29, 51
V, 2, 4
weight, 49, 73
diagram, 67
fundamental, 90
Weyl group for
a2, 87
b2 , 87
g2 , 88, 96
102
INDEX
c Q. Ho-Kim (H Kim Quang). Group Theory: A Problem Book Ver. 2015-12.

Cover illustration from image by Jo Edkins, used with permission.
This book contains suggested solutions to the end-of-chapter problems in
Group Theory: A Physicists Primer by the same author, available at
http://www.scribd.com/doc/207786199/Group-Theory-A-Physicist-s-Primer.

Q. HO-KIM - Group Theory: A Problem Book

Uploaded by

Copyright:

Available Formats

Q. HO-KIM - Group Theory: A Problem Book

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Q. HO-KIM - Group Theory: A Problem Book

Uploaded by

Copyright:

Available Formats

What is the purpose of this book?

What is the purpose of this book?

What chapters are covered in the book?

What chapters are covered in the book?

Group Theory

c Q. Ho-Kim (H Kim Quang). Group Theory: A Problem Book 2015-12.

Group Theory: A Problem Book

Chapter 1. Group Structure

PROBLEMS & SOLUTIONS

PROBLEMS & SOLUTIONS

(23) H21 = (321) H21 = {(23), (321)}

(12) H22 = (123) H22 = {(12), (123)}

(12) H23 = (321) H23 = {(12), (321)}

(31) H21 = (123) H21 = {(31), (123)}

(31) H22 = (321) H22 = {(31), (321)}

(23) H23 = (123) H23 = {(23), (123)}

Of these subgroups of S3 , there is just one invariant, H3

PROBLEMS & SOLUTIONS

(b) By Cayleys theorem, the order-8 group D4 must be isomorphic to

The eight permutations are

By the rules of permutation multiplication, we obtain the following relations:

PROBLEMS & SOLUTIONS

SOLUTION 1.10 Let A = h a i = { ak; k = 0, . . . , n 1}. Each element of Dn

Making use of the relations i j = I + ieijk k , establish the correspondence

(b) We note in particular the relations:

(c) The group Q8 has the following characteristics:

PROBLEMS & SOLUTIONS

Chapter 2. Group Representations

In addition, if is a scalar function of a variable x, then it also satisfies the

(b) In this case, D( R1 ( )) = D( R ( )) and so we have x00 = cx + sy,

( R ) 2( x, y ) = 2x00 y00 = 1 2cs + 2 ( c2 s2 ) + 3 2cs ,

The transformed basis functions are then calculated to be

From this follows the representation matrix

It can be checked that in every case the matrix Mi is orthogonal, MTi Mi = 1.

vectors (1, 6, 3), ( 6, 2, 6) and (3, 6, 1) are mutually orthogonal.

PROBLEMS & SOLUTIONS

vector is quadratic and completely symmetric in x,

2.3 (Regular representations) The regular representation of a group G is

column index (as group elements) are equal: ( hq ) g = h ( qg) = g0 , where g0

The (right-action) representation matrices Dij ( g) are simply the transposes

Group D3 . The multiplication table of D3 is shown below

PROBLEMS & SOLUTIONS

g, then Dhg ( q ) = 1. In this way, we obtain the following nonzero matrix

( ij) = (31), (62), (13), (54), (45), (26).

( ij) = (41), (52), (63), (14), (25), (36).

( ij) = (51), (42), (23), (34), (65), (16).

( ij) = (61), (32), (43), (24), (15), (56).

( ij) = (12), (21), (36), (45), (54), (63).

( ij) = (13), (25), (31), (46), (52), (64).

( ij) = (14), (26), (35), (41), (53), (62).

( ij) = (15), (23), (34), (42), (56), (61).

( ij) = (16), (24), (32), (43), (51), (61).

A matrix X that makes the transformation D = XDX1 is given by

As X is independent of g, it defines the equivalence of the representations

2.4 (Inner products) In GTAPP . 2.4 we have defined an inner product {, }

where D ( G ) is a representation of a group G of order | G | in a linear space

since each term satisfies h D ( h ) x| D ( h) xi 0. Also { x, x } = 0 if and only if

2.5 (D3 two-dimensional representation) Group D3 can be realized in a

PROBLEMS & SOLUTIONS