Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
students
Arjeh Cohen
Rosane Ushirobira
Jan Draisma
October 18, 2007
2
Contents
1 Introduction 5
1.1 Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Basic notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2 Permutation Groups 17
2.1 Cosets and Lagrange . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2 Quotient groups and the homomorphism theorem . . . . . . . . . 18
2.3 Loyd’s puzzle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Action of groups on sets . . . . . . . . . . . . . . . . . . . . . . . 28
3 Symmetry Groups in Euclidean Space 33
3.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2 Isometries of nspace . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3 The Finite Subgroups of O(2, R) . . . . . . . . . . . . . . . . . . 40
3.4 The Finite Subgroups of O(3, R) . . . . . . . . . . . . . . . . . . 42
4 Representation Theory 51
4.1 Linear representations of groups . . . . . . . . . . . . . . . . . . 51
4.2 Decomposing Displacements . . . . . . . . . . . . . . . . . . . . . 59
5 Character Tables 63
5.1 Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2 Orthogonality of irreducible characters . . . . . . . . . . . . . . . 64
5.3 Character tables . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.4 Application: symmetry of vibrations . . . . . . . . . . . . . . . . 77
5.5 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
6 Some compact Lie groups and their representations 81
6.1 Some examples of Lie groups . . . . . . . . . . . . . . . . . . . . 81
6.2 Representation theory of compact Lie groups . . . . . . . . . . . 84
6.3 The 2dimensional unitary group . . . . . . . . . . . . . . . . . . 86
6.4 The 3dimensional orthogonal group . . . . . . . . . . . . . . . . 88
3
4 CONTENTS
Chapter 1
Introduction
1.1 Symmetry
Group theory is an abstraction of symmetry
Symmetry is the notion that an object of study may look the same from diﬀerent
points of view. For instance, the chair in Figure 1.1 looks the same as its
reﬂection in a mirror that would be placed in front of it, and our view on the
wheel depicted next to the chair doesn’t change if we rotate our point of view
over π/6 around the shaft.
But rather than changing viewpoint ourselves, we think of an object’s sym
metry as transformations of space that map the object ‘into itself’.
What do we mean when we say that an object is symmetric? To answer this
question, consider once more the chair in Figure 1.1. In this picture we see a
plane V cutting the chair into two parts. Consider the transformation r of three
dimensional space that maps each point p to the point p
constructed as follows:
Figure 1.1: Bilateral and rotational symmetry. From: [8].
5
6 CHAPTER 1. INTRODUCTION
let l be the line through p and perpendicular to V . Denoting the distance of a
point x to V by d(x, V ), p
is the unique point on l with d(p
, V ) = d(p, V ) and
p = p
. The map r is called the reﬂection in V . Now apply r to all points in the
space occupied by the chair. The result is a new set of points, but because of
the choice of V it equals the old space occupied by the chair. For this reason, we
call the chair invariant under r, or say that r is a symmetry (transformation)
of the chair.
Transforming any point in space to its reﬂection in the mirror W and next
rotating it over an angle of π around the axis V ∩W, gives the same result as the
reﬂection r in V . This is an illustration of what we observed about symmetry:
the change of view on the chair from the original chair to its mirror image, was
‘neutralized’ to the original look, but then on a transformed chair (transformed
by means of r). Symmetry then is the phenomenon that the image chair is
indistinguishable from the original chair.
Next, consider the wheel depicted in Figure 1.1. It also has a symmetry,
but of a diﬀerent type: the rotation s around the wheel’s axis m over 2π/6
moves all points occupied by the wheel to other points occupied by it. For this
reason we call s a symmetry of the wheel. This type of symmetry is called
rotational symmetry. Note that the rotations around m over the angles 2kπ/6
are symmetries of the wheel as well, for k = 1, 2, 3, 4, 5. The wheel also has
reﬂection symmetry: the reﬂection t in the plane W perpendicular to m and
cutting the wheel into two thinner wheels of the same size also maps the wheel
to itself.
Symmetry registers regularity, and thus records beauty. But it does more
than that, observing symmetry is useful. Symmetry considerations lead to eﬃ
ciency in the study of symmetric objects. For instance,
• when designing a chair, the requirement that the object be symmetric may
reduce drawings to half the chair;
• when studying functions on the plane that respect rotational symmetry
around a ‘centre’ p in the plane, we can reduce the number of arguments
from the (usual) two for an arbitrary function on the plane to one, the
distance to p.
Symmetry conditions
A few more observations can be made regarding the previous cases.
1. A symmetry of an object or ﬁgure in space is a transformation of space
that maps the object to itself. We only considered the space occupied by
the chair and the wheel, respectively, and not the materials involved. For
example, if the left legs of the chair are made of wood and the right ones
of iron, one could argue that the chair is not symmetric under reﬂection
in V . Usually, we shall speak about symmetries of sets of points, but this
subtlety should be remembered.
1.1. SYMMETRY 7
O
Figure 1.2: The Hexagon
2. Symmetry transformations are bijective, i.e., they have an inverse trans
formation. In the case of r, it is r itself. In the case of s, it is rotation
about m over −2π/6, or, which amounts to the same, 10π/6.
3. Two transformations can be composed to obtain another transformation;
if both translations are symmetries of some object, then so is their com
position. In the case of s, we see that s◦s (s applied twice) is a symmetry
of the wheel as well; it is rotation over the angle 2 (2π/6). Also, t ◦ s
(ﬁrst apply s, then t) is a symmetry of the wheel. Usually, we will leave
out the composition sign ◦ and write ts instead of t ◦ s, and s
2
for ss.
4. There is one trivial transformation which is a symmetry of any object,
namely the identity, sending each point to itself. It is usually denoted by
e (but also by 0 or 1, see below).
Groups
The transformations under which a given object is invariant, form a group.
Group theory was inspired by these types of group. However, as we shall see,
‘group’ is a more general concept. To get a feeling for groups, let us consider
some more examples.
Planar groups
The hexagon, as depicted in Figure 1.2, is a twodimensional object, lying in
the plane. There are lots of transformations of the plane leaving it invariant.
For example, rotation r around 0 over 2π/6 is one of them, as is reﬂection s in
the vertical line l through the barycentre. Therefore, the hexagon has at least
8 CHAPTER 1. INTRODUCTION
Figure 1.3: A frieze pattern. From: [11].
the following symmetries:
e, r, r
2
, r
3
, r
4
, r
5
, s, sr, sr
2
, sr
3
, sr
4
, sr
5
.
In fact, these are all symmetries of the hexagon. Thus, each is a (multiple)
product of r and s. These two elements are said to generate the group of
symmetries.
Exercise 1.1.1. Verify that the elements listed are indeed distinct and that
r
6
= s
2
= e, and srs = r
5
.
Exercise 1.1.2. Show that sr is a reﬂection. In which line?
Figure 1.3 shows the palmette motif, which is very frequently used as a dec
oration on the upper border of wallpaper. Imagine this pattern to be repeated
inﬁnitely to the left and right. Then the translation t over the vector pointing
from the top of one palmette to the top of the second palmette to its right is a
symmetry of the pattern, and so is the reﬂection r in the vertical line l through
the heart of a given palmette. With these two, all symmetries of the frieze
pattern can be listed. They are:
e, t, t
2
, t
3
, . . . and r, rt, rt
2
, rt
3
, . . . .
We ﬁnd that the symmetry group of the pattern is inﬁnite. However, when
ignoring the translational part of the symmetries, we are left with essentially
only two symmetries: e and r.
Exercise 1.1.3. Find frieze patterns that have essentially diﬀerent symmetry
groups from the one of Figure 1.3.
A pattern with a more complicated symmetry group is the wallpaper pattern
of Figure 1.4, which Escher designed for ‘Ontmoeting’ (Dutch for ‘Encounter’).
It should be thought of as inﬁnitely stretched out vertically, as well as hori
zontally. The translations s and t over the vectors a (vertical from one nose
1.1. SYMMETRY 9
Figure 1.4: Wallpaper pattern for Ontmoeting. From: M.C.Escher, 1944.
to the next one straight above it) and b (horizontal, from one nose to the next
one to its right) leave the pattern invariant, but this is not what makes the
picture so special. It has another symmetry, namely a glidereﬂection g, which
is described as: ﬁrst apply translation over
1
2
a, and then reﬂect in the vertical
line l equidistant to a left oriented and a neighbouring right oriented nose. Note
that g
2
= s.
Exercise 1.1.4. List all symmetries of the pattern in Figure 1.4.
Space groups
Having seen symmetries of some 2dimensional ﬁgures, we go over to 3dimen
sional ones.
The cube in Figure 1.5 has lots of symmetries. Its symmetry group is gen
erated by the rotations r
3
, r
4
, r
4
whose axes are the coordinate axes, and the
reﬂection t through the horizontal plane dividing the cube in two equal parts.
Exercise 1.1.5. What is the size of the symmetry group of the cube?
Closely related is the methane molecule, depicted in Figure 1.6. Not all of
the symmetries of the cube are symmetries of the molecule; in fact, only those
that map the hydrogen atoms to other such atoms are; these atoms form a
regular tetrahedron.
Exercise 1.1.6. Which of the symmetries of the cube are symmetries of the
methane molecule?
Finally consider the cubic grid in Figure 1.7. Considering the black balls to
be natrium atoms, and the white ones to be chlorine atoms, it can be seen as the
structure of a saltcrystal. It has translational symmetry in three perpendicular
10 CHAPTER 1. INTRODUCTION
Figure 1.5: The cube.
Figure 1.6: The methane molecule. From: [7].
1.1. SYMMETRY 11
Figure 1.7: Structure of a crystal
directions along vectors of the same size. To understand its further symmetry,
consider all those symmetries of the crystal that leave a given natrium atom
ﬁxed. Such a symmetry must permute the 6 chlorine atoms closest to the
given natrium atom. These atoms form a regular octahedron, as depicted in
Figure 1.8, and therefore we may say that the symmetry group of salt crystals
is generated by that of the regular octahedron and three translations.
Exercise 1.1.7. This exercise concerns symmetries of onedimensional ﬁgures.
Considere therefore the real line R with the Euclidean metric, in which the
distance between x, y ∈ R is [x − y[. A map g from R to itself is called an
isometry if [g(x) −g(y)[ = [x −y[ for all x, y ∈ R.
1. Show that any isometry of R is either of the form s
a
: x → 2a−x (reﬂection
in a) or of the form t
a
: x → x +a (translation over a).
2. Prove that ¦e¦ and ¦e, s
0
¦, where e denotes the identity map on R, are
the only ﬁnite groups of isometries of R.
To describe symmetries of discrete subsets of R, we restrict our attention to
groups G whose translational part ¦a ∈ R [ t
a
∈ G¦ equals Z.
3. Show that if such a group G contains s
0
, then
G = ¦t
a
[ a ∈ Z¦ ∪ ¦t
a
2
[ a ∈ Z¦.
4. Conclude that there are only two ‘essentially diﬀerent’ groups describing
the symmetries of discrete subsets of R having translational symmetry.
12 CHAPTER 1. INTRODUCTION
Figure 1.8: The octahedron.
Exercise 1.1.8. Analyse the symmetry groups of discrete subsets of the strip
R [−1, 1] in a manner similar to that of Exercise 1.1.7. Compare the result
with your answer to Exercise 1.1.3.
1.2 Basic notions
Formal deﬁnition of a group
Having seen some examples of groups, albeit from the narrow point of view of
symmetry groups of ﬁgures in the plane or in threedimensional space, we are
ready for some abstract deﬁnitions that should reﬂect the experiences we had
in those examples.
Deﬁnition 1.2.1. A group is a set G, together with an operation , which maps
an ordered pair (g, h) of elements of G to another element g h of G, satisfying
the following axioms.
1. The operation is associative, i.e., for all g, h, k ∈ G we have g (h k) =
(g h) k.
2. G contains an identity element, i.e., an element e that satisﬁes e g =
g e = g for all g ∈ G.
3. Each element of G has an inverse, i.e., for each g ∈ G there is an h ∈ G
such that g h = h g = e. This element is denoted by g
−1
.
The cardinality of G is called the order of the group, and often denoted by
[G[; it may be inﬁnite. If the operation is not only associative but commutative
as well (meaning g h = h g for all g, h ∈ G), then G is called an Abelian group.
1.2. BASIC NOTIONS 13
Exercise 1.2.2. Prove that a group G cannot have more than one identity.
Also, the notation g
−1
for the inverse of g seems to indicate the uniqueness of
that inverse; prove this.
Let us give some examples of groups.
Example 1.2.3.
1. All sets of transformations found in '1.1 form groups; in each case, the
composition ◦ serves as operation . Some of them are Abelian; like the
the chair, and some aren’t, like those of the cube.
2. The real numbers R form a group with respect to addition +, the unit
element being 0. They do not form a group with respect to ordinary
multiplication, as 0 does not have an inverse. Leaving out 0 we do obtain
an Abelian group (R ` ¦0¦, ), the unit element being 1.
3. Also Q and Z form additive groups. Again Q ` ¦0¦ is a multiplicative
group. However, Z cannot be turned into a multiplicative group, the only
invertible elements being ¦±1¦.
4. Let X be a set. The bijections of X form a group. Composition is again the
group operation. The group inverse coincides with the ‘functional’ inverse.
The group is often denoted by Sym(X). If X is ﬁnite, say [X[ = n, then
[Sym(X)[ = n!. If two sets X and Y have the same cardinality, then the
groups Sym(X) and Sym(Y ) are essentially the same. For ﬁnite sets of
cardinality n, the representative Sym(¦1, ..., n¦) of this class of groups is
denoted by Sym(n) or S
n
.
Its element are also called permutations. For X = ¦1, 2, 3¦, we have
Sym(X) = ¦e, (1, 2), (2, 3), (1, 3), (1, 2, 3), (3, 2, 1)¦.
This notation is explained in Section 2.3.
5. Let V be a vector space. The set of all linear transformations of V , often
denoted by GL(V ), together with composition of maps, is a group. By
ﬁxing a basis of V , its element can be written as invertible matrices.
Let us make notation easier.
Deﬁnition 1.2.4. Usually, we leave out the operation if it does not cause
confusion. However, if the operation is written as +, it is never left out. For g
in a group G we write g
n
for the product
g g . . . g
. .. .
n
.
Our ﬁrst lemma concerns the inverse of a product.
Lemma 1.2.5. For two elements g, h in a group G we have (gh)
−1
= h
−1
g
−1
,
and (g
n
)
−1
= (g
−1
)
n
. For the latter we also write g
−n
.
14 CHAPTER 1. INTRODUCTION
This is very natural, as Weyl points out in [11]:
‘When you dress, it is not immaterial in which order you perform
the operations; and when in dressing you start with the shirt and
end up with the coat, then in undressing you observe the opposite
order; ﬁrst take oﬀ the coat and the shirt comes last.’
Exercise 1.2.6. Show that the notation for fractions which is usual for integers,
does not work here:
h
g
could stand for g
−1
h as well as for hg
−1
, and the latter
two may diﬀer.
With the methane molecule in Figure 1.6, some of the symmetries of the
cube form a smaller group, namely the symmetry group of the tetrahedron. We
found an example of a subgroup.
Deﬁnition 1.2.7. A subgroup of a group G is a subset H of G satisfying the
following conditions.
1. H is nonempty.
2. H is closed under the operation, i.e., for all g, h ∈ H we have gh ∈ H.
3. H is closed under taking inverses. This means that, for all h ∈ H, we have
h
−1
∈ H.
Exercise 1.2.8. Prove that a subgroup of a group contains the group’s identity.
Prove that for a subset H of G to be a subgroup of G it is necessary and suﬃcient
that H be nonempty and that gh
−1
∈ H for all g, h ∈ H.
Exercise 1.2.9. Prove that for a subset H of a ﬁnite group G to be a sub
group, it is necessary and suﬃcient that H be nonempty and closed under the
operation of G.
Generation
The following lemma is needed to formalize the concept of a subgroup generated
by some elements.
Lemma 1.2.10. The intersection of a family of subgroups of a group G is a
subgroup of G.
Deﬁnition 1.2.11. As a consequence of Lemma 1.2.10, for a subset S of G
the intersection of all groups containing S is a subgroup. It is denoted by 'S`
and called the subgroup generated by S. Instead of '¦g
1
, . . . , g
n
¦` we also write
'g
1
, . . . , g
n
`. The elements g
1
, . . . , g
n
are called generators of this group.
Some groups are generated by a single element. For example, Z is gener
ated by 1, and the symmetry group of the triquetrum depicted in Figure 1.9 is
generated by a rotation around its center over 2π/3.
1.2. BASIC NOTIONS 15
Figure 1.9: The triquetrum.
(0,1) (1,1)
(0,0) (1,0)
Figure 1.10: The ﬂat torus.
Deﬁnition 1.2.12. If there exists an element g ∈ G for which G = 'g`, then G
is called a cyclic group.
Consider the integers modulo n, where n is a positive integer. In this set,
which is denoted by Z/nZ for reasons that will become clear later, two integers
are identiﬁed whenever their diﬀerence is a multiple of n. This set forms a
group, addition modulo n being the operation. The order of Z/nZ is n, and it
is a cyclic group. In fact, one could say that it is the cyclic group of this order,
as all such groups are the same in a sense to be made precise later. Thus the
symmetry group of the triquetrum (Figure 1.9) is ‘the same’ as Z/3Z.
Deﬁnition 1.2.13. Let g be an element of a group G. The order of 'g` is called
the order of g. It may be inﬁnite.
Exercise 1.2.14. Prove that a cyclic group is Abelian.
Exercise 1.2.15. Consider the ﬂat torus T. It can be seen as the square
S = [0, 1] [0, 1] with opposite edges glued together, such that the arrows in
Figure 1.10 match. (Thus, the set is really in bijection with [0, 1) [0, 1).)
The torus has the following group structure: given two points p = (p
1
, p
2
) and
q = (q
1
, q
2
) on the torus, their sum corresponds to the point r on S with the
property that r − (p
1
+ q
1
, p
2
+ q
2
), the latter term of which may not be in S,
has integral entries.
Determine which elements of T have ﬁnite order, and which elements have
inﬁnite order.
16 CHAPTER 1. INTRODUCTION
Chapter 2
Permutation Groups
The realization of a group by means of symmetries (more generally, bijections
of a set) is closely related to the structure of its subgroups. Let us ﬁrst focus
on the latter aspect.
2.1 Cosets and Lagrange
Recall the notion of subgroup from Deﬁnition 1.2.7.
Deﬁnition 2.1.1. For subsets S and T of a group G, we write ST for the set
¦st [ s ∈ S, t ∈ T¦.
For s ∈ G and T ⊂ G we also write sT for ¦s¦T and Ts for T¦s¦.
Deﬁnition 2.1.2. For a subgroup H of G the sets of the form gH with g ∈ G
are called left cosets of H in G, and those of the form Hg are called right cosets
of H in G. The set of all left cosets of H in G is denoted by G/H.
It would be logical to denote the set of right cosets of H in G by H ` G.
Indeed, this is sometimes done, but this meaning of ` should not be confused
with the usual settheoretic one. We shall only use G/H in these notes.
Example 2.1.3.
1. G = Z, the additive group, and H = 'm` = mZ, the subgroup generated
by m, for some m ∈ Z. The set of left cosets is ¦H, 1 +H, 2 +H, . . . , m−
1 +H¦. Since x +H = H +x, left cosets coincide with right cosets.
2. G = S
3
= ¦(1), (12), (13), (23), (123), (132)¦, the symmetry group of an
equilateral triangle (check). Take the subgroup H = '(123)`. Then the
set of left cosets is ¦H, (12)H¦, where (12)H = ¦(12), (13), (23)¦. Observe
that (12)H = H(12) = G` H, so left and right cosets coincide.
17
18 CHAPTER 2. PERMUTATION GROUPS
3. G = R
n
, the (additive) translation group of a real vector space and H is a
linear subspace of G. Then the left cosets are the aﬃne subspaces parallel
to H.
4. A left coset need not be a right coset. Take for instance G = S
3
and
H the cyclic subgroup '(12)`. Then (13)H = ¦(13), (123)¦ and H(13) =
¦(13), (132)¦.
It is easy to verify that left cosets are equivalence classes of the equivalence
relation
g ∼ h ⇔ g
−1
h ∈ H,
and similarly for right cosets. So we have the following lemma.
Lemma 2.1.4. The left cosets of a subgroup H in the group G form a partition
of the group G. For g ∈ G, the map h → gh is a bijection between H and the
left coset gH.
A similar statement holds for right cosets.
Exercise 2.1.5. Show that (13)'(12)` is not a right coset of '(12)` in S
3
, and
that (12)'(123)` is a right coset of '(123)` in S
3
.
Deﬁnition 2.1.6. If the number of left cosets of a subgroup H in a group G is
ﬁnite, this number is called the index of H in G and denoted by [G : H].
Example 2.1.7. In the previous examples, we have [Z : 'm`] = m and
[S
3
: '(123)`] = 2 .
Theorem 2.1.8 (Lagrange’s Theorem). If G is a ﬁnite group and H a subgroup,
then [H[ divides [G[.
Corollary 2.1.9. If G is a ﬁnite group and g ∈ G, then the order of g divides
that of G.
Exercise 2.1.10. If H and K are subgroups of a ﬁnite group G, then H ∩ K
is also a subgroup of G. Prove that [HK[ =
HK
H∩K
.
2.2 Quotient groups and the homomorphism the
orem
If two groups G and H and some of their elements are in the picture, we shall
always make it clear to which of the two groups each element belongs. This
enables us to leave out the operation symbol throughout, as the product of two
elements of H is automatically understood to denote the product in H and not
in G. Also, we shall use e for the unit element of both groups.
Next, we will see the functions between two groups that preserve the struc
ture of groups.
2.2. QUOTIENT GROUPS AND THE HOMOMORPHISM THEOREM 19
Deﬁnition 2.2.1. Let G and H be groups. A homomorphism from G to H is
a map φ : G → H satisfying
φ(g
1
g
2
) = φ(g
1
)φ(g
2
) for all g
1
, g
2
∈ G,
and φ(e
G
) = e
H
. If, in addition, φ is a bijection, then φ is called an isomorphism.
An automorphism of G is an isomorphism from G to itself. By Aut(G) we denote
the set of all automorphisms of G.
If there exists an isomorphism from a group G to a group H, then G is
called isomorphic to H. this fact is denoted by G
∼
= H, and it is clear that
∼
=
is an equivalence relation. We claimed earlier that the symmetry group of the
triquetrum (Figure 1.9) is ‘the same as’ Z/3Z. The precise formulation is, that
both groups are isomorphic.
Deﬁnition 2.2.2. For a group homomorphism φ : G → H the subset
¦g ∈ G [ φ(g) = e¦
of G is called the kernel of φ, and denoted by ker(φ). The set
φ(G) = ¦φ(g) [ g ∈ G¦ ⊆ H
is called the image of φ, and denoted by im(φ).
Exercise 2.2.3. Let φ : G → H be a homomorphism. Then φ is injective if
and only if ker φ = ¦e¦.
Example 2.2.4. For g ∈ G, we deﬁne φ
g
: G → G by φ
g
(h) = ghg
−1
. Then
φ
g
is an element of Aut(G), called conjugation with g. An automorphism of G
is called an inner automorphism if it is of the form φ
g
form some g; it is called
outer otherwise.
Exercise 2.2.5. Let φ be a group homomorphism from G to H. Prove that
ker(φ) and im(φ) are subgroups of G and H, respectively. Show that for every
g ∈ G we have g ker(φ)g
−1
⊆ ker(φ).
Exercise 2.2.6. Prove:
1. Aut(G) is a group with respect to composition as maps;
2. the set Inn(G) = ¦φ
g
[ g ∈ G¦ is a subgroup of Aut(G). Here φ
g
denotes
conjugation with g.
Example 2.2.7. For G = GL(V ), the group of invertible linear transformations
of a complex vector space V , the map g → det(g) is a homomorphism G →C
∗
.
Its kernel is the group SL(V ).
Example 2.2.8. The map Z → Z/mZ, x → x + mZ, is a homomorphism of
additive groups. Its kernel is mZ.
20 CHAPTER 2. PERMUTATION GROUPS
Example 2.2.9. There is an interesting homomorphism between SL(2, C) and
the Lorentz group L, which is the group of all linear transformations of the
vector space R
4
that preserve the Lorentz metric
[x[ := x
2
0
−x
2
1
−x
2
2
−x
2
3
.
To x ∈ R
4
we associate a 2 2matrix ψ(x) as follows:
ψ(x) =
x
0
+x
3
x
1
−ix
2
x
1
+ix
2
x
0
−x
3
,
so that [x[ = det(ψ(x)). Now, the map ϕ : SL(2, C) → L given by ϕ(A)(x) =
ψ
−1
(Aψ(x)A
∗
) is a group homomorphism. Here A
∗
= (¯ a
ji
)
ij
if A = (a
ij
)
ij
.
Indeed, ψ is a linear isomorphism from R
4
onto
ψ(R
4
) =
x y
z u
x, y, z, u ∈ C and ¯ x = x, ¯ y = z, ¯ u = u
,
and one checks that this space is invariant under M → AMA
∗
for any A ∈
GL(2, C). Also, for A ∈ SL(2, C), ϕ(A) preserves the metric due to the multi
plicative properties of the determinant:
[ϕ(A)(x)[ = det(Aψ(x)A
∗
) = det(A) det(ψ(x)) det(A
∗
) = det(ψ(x)) = [x[.
Exercise 2.2.10. Show that ϕ in Example 2.2.9 satisﬁes ϕ(AB) = ϕ(A)ϕ(B),
and thus ﬁnish the proof that ϕ is a homomorphism SL(2, C) → L.
Exercise 2.2.11. Determine the subgroups of (Z/2Z) ⊕ (Z/2Z). Is (Z/2Z) ⊕
(Z/2Z) isomorphic to Z/4Z?
Exercise 2.2.12. Take the group Z/4Z, the subgroup '(1234)` of S
4
and the
multiplicative subgroup generated by i in C. Verify that they are all isomorphic.
As we have seen before, it is not always the case that left cosets and right
cosets of a subgroup H are the same. When that happens, we say that H is
normal.
Deﬁnition 2.2.13. A subgroup H of G is called normal if every left coset of
H in G is a right coset of H in G, or, equivalently, if gHg
−1
⊂ H for all g ∈ G.
We denote it by H G.
If H is normal, then we can deﬁne an operation on the set G/H of all left
cosets of H in H, as follows:
(gH) (kH) = (gk)H.
This operation turns G/H into a group, called the quotient group of G by H.
The map G → G/H, g → gH is a homomorphism G → G/H.
2.2. QUOTIENT GROUPS AND THE HOMOMORPHISM THEOREM 21
Exercise 2.2.14. Let H be a normal subgroup of the group G (cf. Exercise
2.2.13) Verify that the operation on G/H is welldeﬁned, i.e., does not depend
on the choice of representatives g and k. Verify that G/H does indeed become
a group with this operation. Note that one really needs normality of H.
Exercise 2.2.15. Prove that Inn(G) (cf. Example 2.2.4) is a normal subgroup
of Aut(G) (cf. Deﬁnition 2.2.1).
Example 2.2.16.
1. The subgroup '(12)(34), (13)(24)` is normal in S
4
.
2. Let G be the group of Euclidean motions in R
3
(see Chapter 3). Write an
element of G as
a v
0 1
with a ∈ SO(3) and v ∈ R
3
. This presentation is convenient as the oper
ation in G is now given by the multiplication of 4 4 matrices. Consider
the subgroup N of all translations; thus, N consists of all matrices
I v
0 1
.
Then N is a normal subgroup of G.
Lemma 2.2.17. The kernel of a group homomorphism φ : G → H is a normal
subgroup of G.
See Exercise 2.2.5 for the proof.
Exercise 2.2.18. Prove that any subgroup of index 2 is normal.
Theorem 2.2.19. Let φ: G → H be a group homomorphism. Then G/ ker φ
∼
=
imφ.
Proof. Let N := ker φ. We claim that π : G/N → imφ, gN → φ(g) is an
isomorphism. Indeed, it is welldeﬁned: if g
N = gN, then g
−1
g
∈ N so that
φ(g
−1
g
) = e, hence φ(g
) = φ(g). Also, π is a homomorphism:
π((gN)(kN)) = π((gk)N) = φ(gk) = φ(g)φ(k) = π(gN)π(kN).
Finally, if π(gN) = e, then φ(g) = e, hence gN = eN; it follows that ker(π) =
¦eN¦, and Exercise 2.2.3 shows that π is injective. It is also surjective onto its
image, hence it is an isomorphism.
Theorem 2.2.20 (First Isomorphism Theorem). Let G be a group, and H, K
normal subgroups of G such that H ⊇ K. Then K is a normal sugroup of H,
H/K a normal subgroup of G/K, and G/H
∼
= (G/K)/(H/K).
22 CHAPTER 2. PERMUTATION GROUPS
A B C D
E F G H
I J K L
M N O
Figure 2.1: Initial state of Sam Loyd’s 15puzzle.
Proof. For any g ∈ G, we have gKg
−1
⊆ K, so a fortiori this holds for
g ∈ H. Hence, K is normal in H. For any g ∈ G and h ∈ H, we have
(gK)(hK)(gK)
−1
= (ghg
−1
)K ∈ H/K as H is normal in G. Hence, H/K is
normal in G/K.
Consider the map π : G → (G/K)/(H/K), g → (gK)(H/K). It is clearly a
surjective homomorphism, and its kernel equals ¦g ∈ G [ gK ∈ H/K¦ = H. By
Theorem 2.2.19, G/H
∼
= (G/K)/(H/K) as claimed.
Theorem 2.2.21 (Second Isomorphism Theorem). Let G be a group, H a
subgroup of G, and K a normal subgroup of G. Then HK is a subgroup of G,
K a normal subgroup of HK, H∩K is a normal subgroup of H, and HK/K
∼
=
H/(H ∩ K).
Exercise 2.2.22. Prove the second Isomorphism Theorem.
2.3 Loyd’s puzzle
A MessedUp Puzzle
Suppose that a friend borrows your 15puzzle just after you solved it, so that it
is in the initial state depicted in Figure 2.1. You leave the room to make some
coﬀee and when you come back your friend returns the puzzle to you, sighing
that it is too diﬃcult for him, and asking you to show him how to solve it. At
this time, the puzzle is in the state of Figure 2.2. Conﬁdently you start moving
around the plastic squares; for readers unfamiliar with the puzzle a possible
sequence of moves is shown in Figure 2.3. As time passes, you start getting
more and more nervous, for your friend is watching you and you don’t seem to
be able to transform the puzzle back into its initial state. What is wrong?
To answer this question, we shall look at some basic properties of permuta
tions.
2.3. LOYD’S PUZZLE 23
A B C D
E F G H
I J K L
M O N
Figure 2.2: Messed up state of the 15puzzle.
A B C D
E F G H
I J K L
M O N
A B C D
E F G H
I J K
M O N L
A B C D
E F G H
I J
M O N L
K
Figure 2.3: Possible moves of the 15puzzle.
24 CHAPTER 2. PERMUTATION GROUPS
Permutations and their Signs
Recall the deﬁnition of a permutation and the symmetric groups S
n
. Permuta
tions can be represented in several ways, three of which we shall discuss here.
Tworow notation: for a permutation π : ¦1, . . . , n¦ → ¦1, . . . , n¦ one writes
a 2 n matrix, in which both rows contain the numbers 1, . . . , n. The
image of i is the number that appears below i in the matrix. An example
is
π =
1 2 3 4 5
5 4 1 2 3
∈ S
5
.
Disjoint cycle notation: To explain this notation, consider the permutation
π from the previous example. We have π(1) = 5, π(5) = 3 and π(3) equals
1 again. The permutation that satisﬁes these equations and that ﬁxes the
other elements is denoted by
(1, 5, 3) = (5, 3, 1) = (3, 1, 5).
A permutation of this form is called a 3cycle. But π does not ﬁx 2 or 4;
indeed, π(2) = 4 and π(4) = 2. We ﬁnd that
π = (1, 5, 3)(2, 4) = (2, 4)(1, 5, 3),
i.e., π is the product of the 2cycle (2, 6) and the 3cycle (3, 5, 1). These
are called disjoint, because each ﬁxes the elements occurring in the other,
and this fact makes them commute. Any permutation can be written as a
product of disjoint cycles, and this factorization is unique up to changes
in the order. In this notation, it is usual not to write down the 1cycles.
Leaving these out, it is not always clear to which S
n
the permutation
belongs, but this abuse of notation turns out to be convenient rather than
confusing.
Permutation matrix notation: consider once again the permutation π ∈ S
5
.
With it we associate the 5 5matrix A
π
= (a
ij
) with
a
ij
=
1 if π(j) = i,
0, otherwise.
It can be checked that A
πσ
= A
π
A
σ
, where we take the normal matrix
product in the righthand side. Hence S
n
can be seen as a matrix group,
multiplication being the ordinary matrix multiplication.
Diagram notation: in this notation, a permutation is represented by a dia
gram as follows: draw two rows of dots labeled 1, . . . , n, one under the
other. Then draw a line from vertex i in the upper row to vertex π(i) in
the lower. In our example, we get the picture in Figure 2.4.
2.3. LOYD’S PUZZLE 25
1 2 3 4 5
1 2 3 4 5
Figure 2.4: The diagram notation.
Exercise 2.3.1. Show that the 2cycles in S
n
generate that whole group. Can
you give a smaller set of 2cycles that generates the whole group? How many
do you need?
Exercise 2.3.2. Determine the order of a kcycle. Determine the order of π =
(2, 6)(1, 5, 3). Give and prove a general formula for the order of a permutation,
given in disjoint cycle notation, in terms of the lengths of the cycles.
Related to permutation matrices we have the following important deﬁnition.
Deﬁnition 2.3.3. For π ∈ S
n
we deﬁne the sign of π by
sgn π := det A
π
.
The sign of a permutation π is either 1 or −1. In the former case, π is called an
even permutation, in the latter an odd permutation.
Exercise 2.3.4. Show that sgn : S
n
→ ¦±1¦ is a group homomorphism.
The sign of a permutation can be interpreted in many ways. For example,
in the diagram notation, the number of intersections of two lines is even if and
only if the permutation is even.
Exercise 2.3.5. Prove that the sign of a kcycle equals (−1)
k+1
.
Exercise 2.3.6. Prove that the sign of a permutation equals (−1)
N
where N
is the number of crossings of the lines drawn in the diagram notation of the
permutation.
The even permutations in S
n
form a subgroup, called the alternating group
and denoted by A
n
. It is the kernel of the homomorphism sgn : S
n
→ ¦±1¦ of
Exercise 2.3.4.
26 CHAPTER 2. PERMUTATION GROUPS
1
5
12 10 9
13 14 15 16
11
2
6
3
7
4
8
Figure 2.5: Numbering of the positions.
No Return Possible
Let us return to Sam Loyd’s 15puzzle. It can be described in terms of states
and moves. The Figures 2.1 and 2.2 depict states of the puzzle. Formally, a
state can be described as a function
s : ¦1, . . . , 16¦ → ¦A, B, . . . , O, ∗¦.
Here, s(p) is the square occupying position p in the state s and ∗ stands for the
empty square. The positions are numbered as in Figure 2.5.
For p ∈ ¦1, . . . , 16¦, we denote by M(p) ⊆ S
16
the set of possible moves in a
state s with s(p) = ∗. A move m ∈ M(p) should be interpreted as moving the
square from position j to position m(j), for j = 1, . . . , 16. For example
M(1) = ¦(1, 2), (1, 5)¦,
M(2) = ¦(1, 2), (2, 3), (2, 6)¦, and
M(6) = ¦(2, 6), (5, 6), (6, 7), (6, 10)¦.
Moves can be combined to form more general transformations. The set of all
possible transformations starting in a state with the empty square in position
p is denoted by T(p), and it is deﬁned as the set of all products (in S
16
) of the
form m
k
m
k−1
. . . m
1
, where
m
j+1
∈ M(m
j
◦ m
j−1
◦ . . . ◦ m
1
(p))
for all j ≥ 0. This reﬂects the condition that m
j+1
should be a possible move,
after applying the ﬁrst j moves to s. In particular, in taking k = 0, we see
that the identity is a possible transformation in any state. Also, if t ∈ T(p) and
u ∈ T(t(p)), then u ◦ t ∈ T(p). Finally, we have t
−1
∈ T(t(p)).
Moves and, more generally, transformations change the state of the puzzle.
Let s ∈ S with s(p) = ∗ and let t ∈ T(p). Then t changes the state s to the
2.3. LOYD’S PUZZLE 27
Figure 2.6: Ordering of the nonempty squares in each state.
state t s, which is deﬁned by
(t s)(q) = s(t
−1
(q)).
In words, the square at position q after applying t is the same as the square at
position t
−1
(q) before that transformation.
Now that we captured Sam Loyd’s 15puzzle in a comfortable notation, we
can formalize our question as follows: is there a t
0
∈ T(16) that transforms the
messed up state s
1
of Figure 2.2 into the initial state s
0
of Figure 2.1, i.e., that
satisﬁes
t
0
s
1
= s
0
?
The answer is no, and we shall presently make clear why. In each state, order
the nonempty squares according to the path in Figure 2.6. That is, the jth
nonempty square one encounters when following this path in a particular state
s is called the jth square in s, for j = 1, . . . , 15. Now consider any position
p, any t ∈ T(p) and some s ∈ S with s(p) = ∗. Then we deﬁne a permutation
π
t
∈ S
15
by
π
t
(i) = j ⇔ the ith square in state s is the jth square in state t s.
Note that this is independent of the choice of s. We have that
π
u◦t
= π
u
◦ π
t
(2.1)
whenever the lefthand side makes sense, i.e., whenever u ∈ T(t(p)).
We see that the t
0
we would like to ﬁnd equals (14, 15), and then π
t0
also
equals (14, 15), considered as a permutation in S
15
. This is an odd permutation.
Now consider the move m depicted in Figure 2.7. It interchanges the squares in
position 6 and 10 in Figure 2.5, so the move equals (6, 10). The corresponding
permutation π
m
is the cycle (10, 9, 8, 7, 6). A moment’s thought leads to the
28 CHAPTER 2. PERMUTATION GROUPS
6 7 8
9 10
Figure 2.7: Moves correspond to odd cycles.
conclusion that π
m
is an odd cycle for any move m, hence an even permutation.
As a consequence of this and equation (2.3), a sequence of such moves will also
be mapped to an even permutation by π. Therefore, such a transformation
cannot possibly be (14, 15).
You conclude that your friend must have cheated while you were away mak
ing coﬀee, by lifting some of the squares out of the game and putting them back
in a diﬀerent way.
Exercise 2.3.7. Show that the set ¦π
t
[ t ∈ T(p) for some p¦ forms a subgroup
of S
15
.
Exercise 2.3.8. Prove that the set
¦t ∈ T(16) [ t(16) = 16¦
is a subgroup of A
16
. Use this to give another argument why your friend must
have cheated.
2.4 Action of groups on sets
Although the transformations in the 15puzzle do not quite form a group, their
‘acting’ on the puzzle states reﬂects an important phenomenon in group theory,
namely that of an action.
Deﬁnition 2.4.1. Let G be a group and M a set. An action of G on M is a
map α : GM → M satisfying
1. α(e, m) = m for all m ∈ M, and
2. α(g, α(h, m)) = α(gh, m) for all m ∈ M and g, h ∈ G.
2.4. ACTION OF GROUPS ON SETS 29
If the action α is obvious from the context, we leave α out and write gm for
α(g, m). Also, for subsets S ⊂ G and T ⊂ M we write ST for ¦α(g, m) [ g ∈
S, m ∈ T¦. For ¦g¦T we write gT and for S¦m¦ we write Sm.
Remark 2.4.2. Given an action α, we have a homomorphism G → Sym(M)
given by g → (m → α(g, m)). Conversely, given a homomorphism φ : G →
Sym(M), we have an action α : GM → M given by
α(g, m) = φ(g)m.
In other words, an action on M is nothing but a homomorphism G → Sym(M).
Let us give some examples of a group acting on a set.
Example 2.4.3.
1. The group S
n
acts on ¦1, . . . , n¦ by π i = π(i), and on the set of all
twoelement subsets of ¦1, . . . , n¦ by π(¦i, j¦) = ¦π(i), π(j)¦.
2. The group GL(3, R) acts on R
3
by matrixvector multiplication.
3. Consider the group of all motions of the plane, generated by all trans
lations (x, y) → (x + α, y) for α ∈ R in xdirection and all inﬂations
(x, y) → (x, βy) for β ∈ R
∗
. This acts on the solutions of the diﬀerential
equation
(
d
dx
)
2
(y) = −y.
Example 2.4.4. Consider G = S
4
and put a = ¦12, 34¦, b = ¦13, 24¦, c =
¦14, 23¦. Here 12 stands for ¦1, 2¦, etc. There is a natural action (as above) on
the set of pairs, and similarly on the set ¦a, b, c¦ of all partitions of ¦1, 2, 3, 4¦
into two parts of size two. Thus, we ﬁnd a homomorphism S
4
→ Sym(¦a, b, c¦).
It is surjective, and has kernel a group of order 4. (Write down its nontrivial
elements!)
Example 2.4.5. The group SL(2, C) acts on the complex projective line P
1
(C)
as follows. Denote by π : C
2
` ¦0¦ →P
1
(C) the map sending a pair of homoge
neous coordinates to the corresponding point. The linear action of SL(2, C) on
C
2
by left multiplication permutes the ﬁbers of π, hence induces an action on
P
1
(C). On the aﬃne part of P
1
(C) where the second homogeneous coordinate
is nonzero, we may normalize it to 1, and the action on the ﬁrst coordinate x
is given by
a b
c d
x =
ax +b
cx +d
.
Note, however, that this aﬃne part is not invariant under the group.
Lemma 2.4.6. Let G be a group. Then G acts on itself by conjugation, that
is, the map α : GG → G deﬁned by α(h, g) = hgh
−1
is an action.
30 CHAPTER 2. PERMUTATION GROUPS
Proof. We check the two conditions of Deﬁnition 2.4.1. First,
α(e, g) = ege
−1
= g
for all g ∈ G, and second,
α(h, α(k, g)) = h(kgk
−1
)h
−1
= (hk)g(hk)
−1
= α(hk, g).
The following lemma allows us to partition a set M on which a group G acts
into orbits.
Lemma 2.4.7. Let G be a group acting on a set M. Then the sets Gm = ¦gm [
g ∈ G¦ for m ∈ M partition the set M. The set Gm is called the orbit of m
under G.
Deﬁnition 2.4.8. An action of a group G on a set M is said to be transitive if
it has only one orbit, or, stated diﬀerently, Gm = M for any m ∈ M.
To deﬁne symmetry in a general context, we look at sets that are mapped
to themselves by a given group.
Deﬁnition 2.4.9. Let G be a group acting on M. A subset T ⊆ M is called
invariant under S ⊆ G if ST ⊆ T. In this case, the elements of S are called
symmetries of T.
Deﬁnition 2.4.10. Let G be a group acting on M and let m ∈ M. Then the
set
G
m
:= ¦g ∈ G [ gm = m¦
is called the stabilizer of m.
Lagrange’s theorem has an important consequence for problems, in which
one has to count the number of symmetries of a given set. It can be formulated
as follows.
Theorem 2.4.11. Let G be a group acting on a set M and let m ∈ M. Then
G
m
is a subgroup of G, and the map
f : Gm → G/G
m
, g m → gG
m
is well deﬁned and bijective. As a consequence, if G is ﬁnite, then
[G[ = [Gm[[G
m
[.
Remark 2.4.12. One must distinguish the elements of G from their actions on
M. More precisely, the map sending g ∈ G to the map α(g, ) : M → M is a
homomorphism, but need not be an isomorphism.
We now want to make clear how the action of groups can help in classiﬁcation
and counting problems, and we shall do so by means of the following problem.
2.4. ACTION OF GROUPS ON SETS 31
1 2
4 3
1 2
3
4
1
2
3
4 5
Figure 2.8: Three graphs, two of which are isomorphic.
Given is a ﬁnite set X with [X[ = v. What is the number of non
isomorphic simple graphs with vertex set X?
This natural, but seemingly merely theoretical, question has strong analogue
in counting all isomers with a given chemical formula, say C
4
H
10
(see [6]).
Of course, the question is formulated rather vaguely, and the ﬁrst step in
solving it is to formalize the meanings of the words graph and nonisomorphic.
A simple graph is a function
f :
X
2
→ ¦0, 1¦.
Here
X
2
stands for the set of all unordered pairs from X. If the function f
takes the value 1 on such a pair ¦x, y¦, this means that the pair is an edge in
the graph. Often, graphs are associated with a picture like the ones in Figure
2.8. The set of all graphs on X is denoted by Graphs(X).
The ﬁrst two graphs in Figure 2.8 are essentially the same; to make them
exactly the same one has to relabel the vertices. Such a relabeling is in fact a
permutation of X. Formally, this leads to an action of Sym(X) on Graphs(X)
in the following way. First of all Sym(X) acts on
X
2
by
π ¦x, y¦ = ¦π(x), π(y)¦.
Furthermore, we deﬁne π f by the commutativity of the following diagram.
X
2
f
−−−−→ ¦0, 1¦
π·
Id
X
2
−−−−→
π·f
¦0, 1¦
This diagram tells us that e ∈
X
2
is an edge in the graph f if and only if π e
is an edge in the graph π f. In formula,
π f(x) = f(π
−1
x).
32 CHAPTER 2. PERMUTATION GROUPS
Graphs that can be obtained from each other by the action of some π ∈
Sym(X) are called isomorphic. Now our original question can be rephrased
very elegantly as follows.
How many orbits does the action of Sym(X) have on Graphs(X)?
The following lemma answers this question in general.
Lemma 2.4.13 (CauchyFrobenius). Let G be a ﬁnite group acting on a ﬁnite
set M. Then the number of orbits of G on M equals
1
[G[
¸
g∈G
ﬁx(g).
Here ﬁx(g) denotes the number of m ∈ M with g m = m.
Exercise 2.4.14. Prove Lemma 2.4.13.
To apply this lemma to counting graphs, we need to know for some π ∈
Sym(X), the number of f ∈ Graphs(X) that are ﬁxed by π. The function f is
ﬁxed by π if and only if f is constant on the orbits of 'π` in
X
2
. This reﬂects
the condition that π should map edges to edges and nonedges to nonedges.
Denoting the number of orbits of 'π` on
X
2
by o(π), we ﬁnd that the number
of nonisomorphic graphs on X equals
1
v!
¸
π∈Sym(X)
2
o(π)
,
where v = [X[.
Exercise 2.4.15. Consider the permutation π = (1, 2)(3, 4, 5) on ¦1, 2, 3, 4, 5¦.
Calculate the number o(π). Try to ﬁnd a general formula in terms of the cycle
lengths of π and the number v. Observe that o(π) only depends on the cycle
type.
Example 2.4.16. For v = 4, we have o(e) = 6, o((12)) = 4, o((123)) = 2,
o((1, 2, 3, 4)) = 2 and o((12)(34)) = 4; the conjugacy clases of each of these
permutations have sizes 1, 6, 8, 6 and 3, respectively. Hence, the number of
nonisomorphic graphs on 4 vertices equals
(1 2
6
+ 6 2
4
+ 8 2
2
+ 6 2
2
+ 3 2
4
)/24 = 11.
Can you ﬁnd all?
Chapter 3
Symmetry Groups in
Euclidean Space
This chapter deals with groups of isometries of ndimensional Euclidean space.
We set up a framework of results that are valid for general n, and then we
restrict our attention to n = 2 or 3. In these two cases, we classify the ﬁnite
groups of isometries, and point out which of these are crystallographic point
groups, i.e., preserve a 2 or 3dimensional lattice, respectively. First, however,
we consider an example that motivates the development of this theory.
3.1 Motivation
In low dimensions, there are few ﬁnite groups of isometries and this enables us,
when given a two or threedimensional object, to determine its symmetry group
simply by enumerating all axes of rotation and investigating if the object admits
improper isometries such as inversion and reﬂections. As an example, suppose
that we want to describe the symmetry group G of the CH
4
molecule depicted
in Figure 3.1. The hydrogen atoms are arranged in a regular tetrahedron with
the Catom in the center.
As a start, let us compute [G[. Consider a ﬁxed Hatom a. Its orbit Ga
consists of all four Hatoms, as can be seen using the rotations of order 3 around
axes through the Catom and an Hatom. Applying Lagrange’s theorem we ﬁnd
[G[ = 4[G
a
[. The stabilizer of a consists of symmetries of the remaining 4 atoms.
Consider one of the other H atoms, say b. Its orbit under the stabilizer G
a
has
cardinality 3, and the stabilizer (G
a
)
b
consists of the identity and the reﬂection
in the plane through a, b and the Catom, so that we ﬁnd
[G[ = [Ga[[G
a
[ = [Ga[[G
a
b[[(G
a
)
b
[ = 4 3 2 = 24.
Only half of these isometries are orientationpreserving.
To gain more information on the group G, we try to ﬁnd all rotation axes:
33
34 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
H
C
H
H
H
Figure 3.1: The CH
4
molecule.
Figure 3.2: A regular tetrahedron inside a cube.
1. There are 4 threefold axes, namely those through an Hatom and the
center of the opposite side of the tetrahedron.
2. There are three twofold axes, namely those through the centers of two
perpendicular sides.
Later on, we shall see that this information determines the orientation pre
serving part of G uniquely; it is denoted by T. From the fact that G does
not contain the inversion, but does contain other improper isometries, we may
conclude that G is T ∪ i(W ` T), a group that can be described by Figure 3.2.
Here we see a regular tetrahedron inside a cube. Half of the proper isometries
of the cube are isometries of the tetrahedron, and the improper isometries of
the tetrahedron may be obtained as follows: take any proper isometry of the
cube which is not a isometry of the tetrahedron, and compose it with the in
version i. Note that T ∪ i(W`) and W are isomorphic as abstract groups, but
their realizations in terms of isometries are diﬀerent in the sense that there is
no isometry mapping one to the other. This point will also be clariﬁed later.
3.2. ISOMETRIES OF NSPACE 35
Exercise 3.1.1. Consider the reﬂection in a plane through two Hatoms and
the Catom. How can it be written as the composition of the inversion and a
proper isometry of the cube?
The ﬁnite symmetry group of the tetrahedron and the translations along
each of three mutually orthogonal edges of the cube in Figure 3.2 generate the
symmetry group of a crystallike structure. For this reason, the group G is
called a crystallographic point group.
3.2 Isometries of nspace
Consider the ndimensional Euclidean space R
n
with the distance function
d(x, y) = x −y = (x −y, x −y)
1/2
= (
¸
i
(x
i
−y
i
)
2
)
1/2
.
Deﬁnition 3.2.1. An isometry is a map g : R
n
→R
n
such that
d(g(x), g(y)) = d(x, y) for all x, y ∈ R
n
.
The set of all isometries is a subgroup of Sym(R
n
), which we denote by AO(n, R).
The following deﬁnition and lemmas will explain the name of this group.
Deﬁnition 3.2.2. For a ﬁxed a ∈ R
n
, the isometry t
a
: x → x +a is called the
translation over a.
Lemma 3.2.3. For any g ∈ AO(n, R), there exists a unique pair (a, r) where
a ∈ R
n
and r ∈ AO(n, R) ﬁxes 0, such that g = t
a
r.
Proof. Set a := g(0). Then t
−1
a
g(0) = 0, so t
−1
a
g is an isometry ﬁxing 0, and it
is clear that a = g(0) is the only value for which this is the case.
Lemma 3.2.4. Let r ∈ AO(n, R) be an isometry ﬁxing 0. Then r is an orthog
onal Rlinear map.
Proof. Let x, y ∈ R
n
, and compute
(rx, ry) = (−d(rx, ry)
2
+d(rx, 0)
2
+d(ry, 0)
2
)/2
= (−d(x, y)
2
+d(x, 0)
2
+d(y, 0)
2
)/2 = (x, y),
where we use that r is an isometry ﬁxing 0. Hence, r leaves the inner product
invariant. This proves the second statement, if we take the ﬁrst statement for
granted. For any z ∈ R
n
, we have
(z, rx +ry) = (z, rx) + (z, ry) = (r
−1
z, x +y) = (z, r(x +y)),
so that rx + ry − r(x + y) is perpendicular to all of R
n
, hence zero. Similarly,
one ﬁnds r(αx) = αrx for α ∈ R. This proves the ﬁrst statement.
36 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
Exercise 3.2.5. Prove that an isometry is bijective. (The assertion is used
in the second part of Deﬁnition 3.2.1.) Hint: If r is an isometry, then t
−r0
r
is an isometry ﬁxing 0, so it suﬃces to prove the assertion for r with r0 = 0.
Now the proof of Lemma 3.2.4 uses r
−1
, so we need another proof. Set v =
r(λx + µy) − λr(x) − µr(y) and consider (v, v). Using bilinearity of the inner
product and orthogonality of r, it can be shown that (v, v) = 0.
The group of all orthogonal linear maps R
n
→ R
n
is denoted by O(n, R).
An isometry is thus an orthogonal linear map, followed by a translation; hence
the name aﬃne orthogonal transformation.
For a group G, we write
T(G) := ¦a ∈ R
n
[ t
a
∈ G¦;
this is an additive subgroup of R
n
, which we shall identify with the group of all
translations in G.
Using the linearity of isometries ﬁxing 0, one can prove the following.
Proposition 3.2.6. For a subgroup G ⊆ AO(n, R), the map R : G → O(n, R)
deﬁned by
g = t
a
R(g) for some a ∈ R
n
is welldeﬁned, and a group homomorphism. Its kernel is T(G), and this set is
invariant under R(G).
Proof. Lemma 3.2.3 and 3.2.4 show that R is welldeﬁned. Let r
1
, r
2
∈ O(n, R)
and a
1
, a
2
∈ R
n
, and check that
t
a1
r
1
t
a2
r
2
= t
a1+r1a2
r
1
r
2
.
This proves that R is a homomorphism. Its kernel is obviously T(G). Let
r ∈ R(G) and a
1
∈ R
n
such that t
a1
r
1
∈ G, and let a
2
∈ T(G). Then, for
x ∈ R
n
,
t
a1
r
1
t
a2
r
−1
1
t
−a1
x = t
a1
r
1
t
a2
(r
−1
1
x −r
−1
1
a
1
)
= t
a1
(x −a
1
+r
1
a
2
)
= t
r1a2
x.
This shows shows that T(G) is invariant under R(G).
We are interested in certain groups that act discretely.
Deﬁnition 3.2.7. A subset V ⊆ R
n
is called discrete if for every point p ∈ V
there is an open neighbourhood U ⊆ R
n
such that
U ∩ V = ¦p¦.
Deﬁnition 3.2.8. A discrete subgroup of R
n
is called a lattice, and the dimen
sion of its Rlinear span is called its rank.
3.2. ISOMETRIES OF NSPACE 37
Lemma 3.2.9. Let L ⊆ R
n
be a lattice. Then L is a closed set.
Proof. As 0 is an isolated point, there is an > 0 such that v ≥ for all
v ∈ L`¦0¦. But then v −w ≥ for all distinct v, w ∈ L. Hence, any sequence
(v
n
)
n
in L with v
n
−v
n+1
converging to 0, is eventually constant.
Proposition 3.2.10. Let a
1
, . . . , a
k
∈ R
n
be linearly independent over R. Then
the set
L =
¸
i
Za
i
is a lattice. Conversely, every lattice is of this form.
Proof. To prove that an additive subgroup of R
n
is a lattice, it suﬃces to prove
that 0 is an isolated point. For L as in the lemma, let m be the minimum of
the continuous function
x → 
¸
i
x
i
a
i

on the unit sphere ¦x ∈ R
k
[ x = 1¦. As the a
i
are linearly independent,
m > 0. For any nonzero x ∈ Z
k
, we ﬁnd

¸
i
x
i
a
i
 = x
¸
i
x
i
x
a
i
 ≥ xm ≥ m,
proving the ﬁrst statement.
For the converse, we proceed by induction on the rank of the lattice. Clearly,
a lattice of rank 0 is of the desired form. Now suppose that every lattice of rank
k − 1 is of that form, and let L ⊆ R
n
be a lattice of rank k. First let us show
that L is a closed subset of R
n
.
Being discrete and closed, L has only ﬁnitely many points in any compact
set. In particular, one can choose a v
1
∈ L of minimal nonzero norm. Let
π : R
n
→ v
⊥
1
denote the orthogonal projection along v
1
. Then π(L) is an
additive subgroup of v
⊥
1
. Let v ∈ L be such that π(v) = 0. After subtracting an
integer multiple of v
1
from v, we may assume that v −π(v) ≤ v
1
/2. Using
v ≥ v
1
 and Pythagoras, we ﬁnd π(v) ≥
√
3/2v
1
. This proves that π(L)
is discrete, hence a lattice. It has rank k −1, so there exist v
2
, . . . , v
k
∈ L whose
images under π are linearly independent, and such that
π(L) =
k
¸
i=2
Zπ(v
i
).
Now v
1
, . . . , v
k
are linearly independent. For v ∈ L we can write
π(v) =
k
¸
i=2
c
i
π(v
i
)
38 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
with c
i
∈ Z. Then v−
¸
k
i=2
c
i
v
k
∈ L is a scalar multiple of v
1
, and by minimality
of v the scalar is an integer. Hence,
L =
k
¸
i=1
Zv
i
.
Deﬁnition 3.2.11. A subgroup G ⊆ AO(n, R) is said to act discretely if all its
orbits in R
n
are discrete.
If G ⊆ AO(n, R) is a group that acts discretely, then T(G) does so, too. In
general, this does not hold for R(G), as the following example shows.
Example 3.2.12. Let α ∈ R be an irrational multiple of π, and deﬁne r ∈
O(3, R) by
r =
¸
cos α sin α 0
−sin α cos α 0
0 0 1
¸
.
Consider the group G generated by g = t
e3
r. From g
n
= t
ne3
r
n
it is clear that
G acts discretely, and that R(G) = ¦r
n
[ n ∈ Z¦. As r has inﬁnite order, we
have T(G) = ¦0¦, and R(G) does not act discretely: the R(G)orbit of (1, 0, 0)
T
is an inﬁnite set in the circle ¦x ∈ R
3
[ x
3
= 0, x = 1¦.
This phenomenon does not occur if we require T(G) to have the maximal
possible rank.
Theorem 3.2.13. Let G be a subgroup of O(n, R) leaving invariant a lattice L
of rank n. Then G is ﬁnite.
Proof. Let a
1
, . . . , a
n
∈ L be linearly independent. As L is discrete, it contains
only ﬁnitely many vectors of norm a
i
, for each i. Hence, the Gorbits Ga
i
are ﬁnite. As the a
i
span R
n
, the permutation representation of G on the ﬁnite
set X :=
¸
n
i=1
Ga
i
is faithful. Hence, G is embedded into the ﬁnite group
Sym(X).
Deﬁnition 3.2.14. A subgroup G ⊆ AO(n, R) for which T(G) is a lattice of
rank n, is called a crystallographic group in n dimensions. A group G ⊆ O(n, R)
leaving invariant a lattice of rank n, is called a crystallographic point group in
n dimensions.
Corollary 3.2.15. Let G ⊆ AO(n, R) be a crystallographic group. Then R(G)
is a crystallographic point group. Any crystallographic point group is ﬁnite.
Proof. By Proposition 3.2.6, T(G) is invariant under R(G). Now apply Theorem
3.2.13.
3.2. ISOMETRIES OF NSPACE 39
This corollary allows for the following approach to the classiﬁcation of crys
tallographic groups in O(n, R): ﬁrst classify all ﬁnite subgroups of O(n, R), then
investigate for each of them whether it leaves invariant a lattice of rank n. Fi
nally, for each pair (R, L) of a crystallographic point group R and a lattice L of
rank n left invariant by R, there may be several nonequivalent groups G with
T(G) = L and R(G) = R. Among them is always the semidirect product LR:
the subgroup of AO(n, R) generated by L and R.
In this setting, ‘classiﬁcation’ means ‘up to equivalence under AO(n, R)’, i.e.,
the groups G and hGh
−1
for some h ∈ AO(n, R) are considered the same. This
is a ﬁner equivalence relation than isomorphism, as is shown in the following
example.
Example 3.2.16. The groups C
2
and D
1
(see Section 3.3) are isomorphic as
abstract groups, but not equivalent under AO(2, R). The same holds for the
groups W and T ∪ i(W ` T) (see Section 3.4).
Exercise 3.2.17. Show that the symmetry group G of Figure 1.4 is not the
semidirect product of R(G) and T(G).
In the next sections, we shall carry out this classiﬁcation in the cases n = 2
and 3, up to the point of determining the crystallographic point groups. The
determinant of an orthogonal map turns out to be a useful tool there.
Deﬁnition 3.2.18. An isometry r ∈ O(n, R) is called proper or orientation
preserving if det r = 1. The set of all proper isometries in O(n, R) is denoted
by SO(n, R). An isometry g ∈ AO(n, R) is called proper if R(g) is proper.
Otherwise, it is called improper.
Exercise 3.2.19. Prove that, in any subgroup G ⊆ AO(n, R) containing im
proper isometries, the proper ones form a subgroup of index 2.
Exercise 3.2.20. Consider the following map:
r
a
: v → v −2(v, a)a,
where a ∈ R
n
has norm 1 (that is, [[a[[ = 1). Prove that
1. it belongs to O(n, R);
2. ﬁxes each vector in the hyperplane a
⊥
;
3. r
a
(a) = −a;
4. det(r
a
) = −1;
5. r
a
has order 2;
Such an element is called a reﬂection.
Exercise 3.2.21. Show that any ﬁnite subgroup G ⊆ AO(n, R) ﬁxes a point
in R
n
. Hint: consider the ‘average’ of an arbitrary orbit.
40 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
Exercise 3.2.22. Let G be a ﬁnite subgroup of the group GL(n, R) of all invert
ible n nmatrices. Show that G leaves invariant a positive deﬁnite symmetric
bilinear form, and that it is hence conjugate to a subgroup of O(n, R). Hint: let
(., .) denote the standard inner product, and deﬁne
'x, y` :=
1
[G[
¸
g∈G
(gx, gy)
for x, y ∈ R
n
. Show that '., .` is symmetric and positive deﬁnite. Let A be the
matrix with entries a
ij
= 'e
i
, e
j
`, and B ∈ GL(n, R) such that B
T
B = A. Show
that BGB
−1
⊆ O(n, R).
3.3 The Finite Subgroups of O(2, R)
In nature, as well as in art and many other areas, many plane ﬁgures with
certain symmetries exist. One may think of ﬂowers (seen from above), frieze
patterns on picture frames, Escher’s plane tilings and decorative wall papers.
In order to describe these symmetries, we investigate the ﬁnite subgroups of
O(2, R).
In two dimensions, we have two diﬀerent types of isometries.
Lemma 3.3.1. The linear map r
φ
: R
2
→R
2
with matrix
cos φ −sin φ
sin φ cos φ
with respect to the standard basis is an element of SO(2, R), called rotation
around 0 over φ. Conversely, any element of this group equals some r
φ
. Any
improper isometry in O(2, R) is a reﬂection in a line, see Exercise 3.2.20.
The proof of this lemma is straightforward. Let us describe two classes of
ﬁnite subgroups of O(2, R).
1. The cyclic group of order n, denoted by C
n
, and generated by r
2π/n
.
2. The dihedral group of order 2n, denoted by D
n
, and generated by r
2π/n
and the reﬂection in a ﬁxed line.
In fact, these are the only two classes, as we shall now prove. Let G be a
ﬁnite subgroup of O(2, R). There are two possibilities:
1. G ⊆ SO(2, R). By Lemma 3.3.1, it consists of rotations r
φ
. Let φ > 0
be minimal such that r
φ
∈ G. As G is ﬁnite, φ = 2π/n for some integer
n > 0. If θ is any other angle such that r
θ
∈ G, then θ = kφ +σ for some
integer k and angle σ ≥ 0 and < φ. By minimality of φ, σ = 0. Hence G
is generated by r
2π/n
, and equal to C
n
.
3.3. THE FINITE SUBGROUPS OF O(2, R) 41
2. G contains a reﬂection g in a line l. Then G ∩ SO(2, R) is a subgroup of
index 2 in G, and by the ﬁrst case generated by r
2π/n
for some integer
n > 0. Now G is generated by the g and r
2π/n
, and conjugate to D
n
.
We have thus proved the following theorem.
Theorem 3.3.2. Let G be a ﬁnite subgroup of O(2, R). Then G is conjugate
to one of the groups C
n
or D
n
for some integer n ≥ 1.
Exercise 3.3.3. Show that O(n, R) is generated by reﬂections. Hint: show ﬁrst
that it is generated by elements that ﬁx an (n −2)dimensional subspace of R
n
pointwise, and that O(2, R) is generated by reﬂections.
Using this classiﬁcation, we can ﬁnd which of the ﬁnite groups are crystal
lographic point groups.
Theorem 3.3.4. The following groups are crystallographic point groups:
C
1
, C
2
, C
3
, C
4
, C
6
,
D
1
, D
2
, D
3
, D
4
, D
6
.
Moreover, any crystallographic point group G in two dimensions is conjugate to
one of the above.
Proof. For the ﬁrst part, one only needs to construct invariant lattices for D
4
and D
6
(see 3.3.5), as the other groups are contained in one of these see.
For the second statement, suppose that G is a crystallographic point group.
According to Theorem 3.2.13, G must be ﬁnite. The lattice L is invariant under
the subgroup K of O(2, R) generated by G∪¦i¦, where i denotes the inversion.
As the inversion commutes with all elements of G, K is ﬁnite (either equal to
G or twice as large). Therefore, according to Theorem 3.3.2, K is conjugate to
C
n
or D
n
for some n; as K contains the inversion, n must be even. So either
n ≤ 6, in which case G is of one of the types mentioned in the theorem, or
n ≥ 8. Suppose that the latter were the case and choose a vector c in the
lattice of shortest possible length. Now r
2π/n
c ∈ L and c − r
2π/n
c ∈ L. A
little calculus shows that the latter vector is shorter than c because the rotation
angle too small (for n = 8, this can be seen from Figure 3.3), and we arrive at
a contradiction.
Exercise 3.3.5. Show that D
4
and D
6
are crystallographic point groups.
Note that we have only classiﬁed all crystallographic point groups, and not
the crystallographic groups. There are 17 of the latter, and only 10 of the
former; see [2].
42 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
c
hc
chc
Figure 3.3: Why n cannot not be 8.
3.4 The Finite Subgroups of O(3, R)
In this section, we shall repeat the classiﬁcation of '3.3, but now for three
dimensions.
Lemma 3.4.1. Any element g ∈ SO(3, R) ﬁxes a line, i.e., with respect to some
orthonormal basis g has the matrix
1 0
0 M
,
where M is the matrix of an element of SO(2, R).
Proof. Let λ
1
, λ
2
, λ
3
be the eigenvalues of g. As g leaves the norm invariant,
all have norm 1. As g is proper, we have λ
1
λ
2
λ
3
= 1. Now either all are real,
in which case at least one of them is 1, or we may assume that λ
1
is real and
λ
3
= λ
2
. But then λ
2
λ
3
= [λ
2
[
2
= 1, so λ
1
= 1.
An element of SO(3, R) is called a rotation, the line ﬁxed by it its axis and
the angle corresponding to M its rotation angle. We present some classes of
ﬁnite subgroups of SO(3, R).
1. Fix an axis l and consider the proper rotations around l over the angles
2kπ/n for k = 0, . . . , n−1. They form a group, denoted by C
n
, due to its
analogy to the class denoted by C
n
in the twodimensional setting.
2. Start with the group C
n
of rotations around l and choose a second axis m
perpendicular to l. Consider the group generated by C
n
and the rotation
around m over π. It is twice as large as C
n
, and denoted by D
n
. Note that
it contains only proper isometries. The proper symmetries of an nprism
form a group of type D
n
. Note that D
1
is conjugate to C
2
.
3.4. THE FINITE SUBGROUPS OF O(3, R) 43
3. The ﬁve platonic solids (tetrahedron, cube, octahedron, dodecahedron
and icosahedron) with vertices on the unit sphere yield examples of ﬁnite
subgroups of SO(3, R). However, only three distinct classes arise:
• T, the group of all proper rotations leaving a tetrahedron invariant.
• W, the group of all proper rotations that are symmetries of a cube.
• P, the group of all proper rotations leaving a dodecahedron invariant.
The reason is that the octahedron and the cube are polar ﬁgures: it is
easily seen that the proper symmetries of a cube are exactly those of the
octahedron which has its vertices in the middles of the faces of the cube.
A similar statement holds for the dodecahedron and the icosahedron.
Theorem 3.4.2. Any ﬁnite subgroup of SO(3, R) is conjugate to one of the
following.
C
n
for n ≥ 1,
D
n
for n ≥ 2,
T, W, P.
The following proof is due to Euler.
Proof. Let G be a ﬁnite subgroup of SO(3, R). Consider the set S deﬁned by
S := ¦p ∈ R
3
[ [[p[[ = 1, there exists g ∈ G such that g = e and g(p) = p¦.
Count the number N of pairs (g, p), with e = g ∈ G and p on the unit sphere,
that satisfy g(p) = p. On one hand, each g = e ﬁxes exactly two antipodal
points, hence this number is
N = 2([G[ −1).
On the other hand, for each p ∈ S, the number of g = e ﬁxing p equals [G
p
[ −1.
Hence
N =
¸
p∈S
([G
p
[ −1).
If gp = p and g = e, and h ∈ G, then (hgh
−1
)hp = hp and hgh
−1
= e, so S is
invariant under G. Partition S into orbits of G. The value of [G
p
[ is constant
on such an orbit o, namely [G[/[o[, due to Lagrange’s theorem. Hence we ﬁnd
2([G[ −1) =
¸
o
[o[([G[/[o[ −1),
where the sum is taken over all orbits of G in S. Hence, dividing both sides by
[G[, we get
2 −
2
[G[
=
¸
o
(1 −
[o[
[G[
).
44 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
Group type Order Number of orbits on S Their sizes rotation axes
C
n
n 2 1, 1 1 n
D
n
2n 3 2, n, n 1 n, n 2
T 12 3 4, 4, 6 4 3, 3 2
W 24 3 8, 6, 12 4 3, 3 4, 6 2
P 60 3 12, 20, 30 6 5, 10 3, 15 2
Table 3.1: The ﬁnite groups of proper rotations in space.
Hence, if we set a
o
:= [G[/[o[ for each orbit o on S, then we have
¸
o
(1 −1/a
o
) < 2.
As each s ∈ S has nontrivial stabilizer, each orbit in S has size at most [G[/2;
hence each a
o
is at least 2. Now Exercise 3.4.3 leads to Table 3.1. That is, it
provides the numerical data for it; it remains to check that this data determines
the group up to conjugation.
Exercise 3.4.3. Consider the following function in GAP.
extend:=function(a,s)
#Precondition: a is a nonempty nondecreasing list of integers >=2,
#and s equals (sum i: (11/a[i])). This procedure ‘extends’ a in all
#possible ways, under the condition that this sum remains smaller than
#2.
local b,g,o;
if s >= 2 then return false;
else g:=2/(2s); o:=List(a,x>g/x);
Print("Orbit sizes:",o," G:",g,"\n");
b:=a[Length(a)];
while extend(Concatenation(a,[b]),s+(11/b)) do
b:=b+1;
od;
return true;
fi;
end;
Explain why the function call extend([2,2],1) results in an endless loop. On
the other hand, when called as extend([2,3],1/2+2/3), the function does halt,
and prints
Orbit sizes:[ 6/5, 4/5 ] G:12/5
Orbit sizes:[ 6, 4, 4 ] G:12
Orbit sizes:[ 12, 8, 6 ] G:24
Orbit sizes:[ 30, 20, 12 ] G:60
3.4. THE FINITE SUBGROUPS OF O(3, R) 45
Figure 3.4: The orbits of W on the points on rotation axes.
Compare this with Table 3.1.
Exercise 3.4.4. In Figure 3.4 we see a cube with vertices on the unit sphere.
Its group of proper symmetries contains nontrivial rotation axes through the
points drawn in that picture; they form the set S in the proof of Theorem 3.4.2.
The points in the diﬀerent orbits are given diﬀerent shades of grey. Make similar
pictures for the group types C
n
, D
n
and T. Describe the symmetry axes for the
icosahedron.
Now let us try to include improper rotations. Let G be a ﬁnite subgroup of
O(3, R) containing improper rotations. Then H := G∩ SO(3, R) has index 2 in
G. If G contains the inversion i, then we have a disjoint union
G = H ∪ iH.
Note that this is isomorphic to H C
2
as an abstract group. Now assume that
G does not contain i. Then the set
G
:= H ∪ i(G` H)
is a subgroup of SO(3, R) in which H has index 2, and one can recover G by
G = H ∪ i(G
` H).
Conversely, if G
, H ⊆ SO(3, R) are ﬁnite subgroups, and H has index 2 in G
,
then this equation deﬁnes a ﬁnite subgroup of O(3, R), which is isomorphic to
G
as an abstract group, but not equivalent to it.
This enables us to classify all ﬁnite subgroups of O(3, R).
Theorem 3.4.5. Any ﬁnite subgroup of O(3, R) is conjugate to one of the
following:
C
n
, C
n
∪ iC
n
, C
n
∪ i(C
2n
` C
n
), for n ≥ 1,
D
n
, D
n
∪ iD
n
, D
n
∪ i(D
2n
` D
n
), C
n
∪ i(D
n
` C
n
), for n ≥ 2,
T, W, P, T ∪ iT, W ∪ iW, P ∪ iP, T ∪ i(W ` T).
46 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
We do not give all details of the proof, but restrict ourselves to the following
remarks.
1. Note that D
1
and C
2
describe the same class; for this reason the second
list starts with n = 2.
2. The last group in the list is made possible by the fact that a group of
type W has a subgroup of type T of index 2. This can be seen as follows:
consider the group of proper symmetries of a cube with center 0, and
look at the lines between antipodal vertices of the cube; they are axes of
rotations of order 3 leaving the cube invariant; these proper rotations form
the symmetry group of a regular tetrahedron with center 0 and points on
those axes.
In applications, one wants to determine the (ﬁnite) symmetry group of a
given object. The following procedure, based on Table 3.1 and the list in The
orem 3.4.5, can then be helpful.
1. List all rotation axes and their orders. They determine the subgroup H
of G of all proper rotations.
2. Check whether G contains any improper rotations. If not, then G = H.
3. Suppose that G contains improper rotations. Check whether i ∈ G. If so,
then G = H ∪ iH.
4. Suppose that G contains improper rotations, but not the inversion. Then,
according to the list of Theorem 3.4.5, there is only one possibility for G,
unless H is of type C
n
.
5. Suppose that H = C
n
and that G contains improper rotations, but not
the inversion. Then if G is cyclic, it is of type C
n
∪i(C
2n
`C
n
). Otherwise,
it is of type C
n
∪ i(D
n
` C
n
).
Exercise 3.4.6. Consider the molecule depicted in Figure 3.5. Identify the
class of its symmetry group in the list of Theorem 3.4.5.
Exercise 3.4.7. Consider the full symmetry group of the tetrahedron, including
its improper rotations. Which class in the list of Theorem 3.4.5 does it belong
to?
Exercise 3.4.8. Determine the symmetry group of the CH
4
molecule depicted
in Figure 3.1.
Exercise 3.4.9. Consider the ethane molecule in staggered conﬁguration as
depicted in Figure 3.6. What is the type of its symmetry group?
Exercise 3.4.10. Determine the symmetry group of the N
4
P
4
molecule, de
picted in Figure 3.7.
3.4. THE FINITE SUBGROUPS OF O(3, R) 47
H
H H
N
Figure 3.5: The NH
3
 molecule.
H H
C
H
C
H
H
H
Figure 3.6: The ethane molecule in staggered conﬁguration.
48 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
N
N
N
N
P
P
P
P
Figure 3.7: The ethane molecule in staggered conﬁguration.
Most of the above examples are taken from [2].
Finally, the question arises which of the ﬁnite subgroups of O(3, R) are crys
tallographic point groups.
Theorem 3.4.11. Let G be a crystallographic point group in 3 dimensions.
Then G is conjugate to one of the following groups:
C
m
, C
m
∪ iC
m
m = 1, 2, 3, 4, 6
D
m
, D
m
∪ iD
m
m = 2, 3, 4, 6
C
m
∪ i(C
2m
` C
m
) m = 1, 2, 3
D
m
∪ i(D
2m
` D
m
) m = 2, 3
C
m
∪ i(D
m
` C
m
) m = 2, 3, 4, 6
T, W, T ∪ iT, W ∪ iW, T ∪ i(W ` T)
Like in the twodimensional case, a crystallographic point group may leave
several lattices invariant. Indeed, there are 230 mutually nonequivalent crys
tallographic groups in 3 dimensions, while there are only 32 crystallographic
point groups, see [2].
The proof runs like that of Theorem 3.3.4. For the sake of compatibility
with other texts on crystallographic groups, especially texts on the applications
in physics and chemistry, we include table 3.4. It is organized as follows: in
the ﬁrst column, one ﬁnds lattices classiﬁed up to their symmetry groups. The
second column contains the groups leaving that lattice invariant (but no lattice
with a smaller symmetry group), in our notation. The last row for each lattice
is the full symmetry group of that lattice. There is only one exception to this
rule: a trigonal lattice is really hexagonal. The third and fourth column contain
two commonly used sets of symbols for them.
3.4. THE FINITE SUBGROUPS OF O(3, R) 49
Type of Lattice Our notation Schoenﬂies International
Triclinic C
1
C
1
1
C
1
∪ iC
1
C
i
= S
2
1
Monoclinic C
2
C
2
2
C
1
∪ i(C
2
` C
1
) C
s
= C
1h
m
C
2
∪ iC
2
C
2h
2/m
Orthorhombic D
2
D
2
= V 2 2 2
C
2
∪ i(D
2
` C
2
) C
2v
m m 2
D
2
∪ iD
2
D
2h
= V
h
m m m
Trigonal C
3
C
3
3
D
3
D
3
3 2
C
3
∪ iC
3
C
3i
= S
6
3
C
3
∪ i(D
3
` C
3
) C
3v
3 m
D
3
∪ iD
3
D
3d
3 m
Tetragonal C
4
C
4
4
D
4
D
4
4 2 2
C
4
∪ iC
4
C
4h
4/m
C
2
∪ i(C
4
` C
2
) S
4
4
C
4
∪ i(D
4
` C
4
) C
4v
4 m m
D
4
∪ i(D
4
` D
2
) D
2d
= V
d
4 2 m
D
4
∪ iD
4
D
4h
4/m m m
Hexagonal C
6
C
6
6
D
6
D
6
6 2 2
C
6
∪ iC
6
C
6h
6/m
C
3
∪ i(C
6
` C
3
) C
3h
6
C
6
∪ i(D
6
` C
6
) C
6v
6 m m
D
3
∪ i(D
6
` D
3
) D
3h
6 m 2
D
6
∪ iD
6
D
6h
6/m m m
Isometric T T 2 3
W O 4 3 2
T ∪ iT T
h
m 3
T ∪ i(W ` T) T
d
4 3 m
W ∪ iW W
h
m 3 m
Table 3.2: A Dictionary of Crystallographic Group Names.
50 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
Chapter 4
Representation Theory
4.1 Linear representations of groups
An abstract group can be represented in various ways. In Chapter 2, we en
countered permutation representations; in the present chapter we will start the
theory of linear representations. In this chapter, G denotes a group and K a
ﬁeld of characteristic zero. All vector spaces will be over K.
The notion
We introduce the concept of a linear representation of G.
Deﬁnition 4.1.1. A (linear) representation of a group G on a vector space V
is a homomomorphism
ρ : G → GL(V ).
If ρ is injective, it is called faithful.
After a choice of basis of V , we can view a representation as a homomorphism
G → GL(n, K),
where GL(n, K) is the group of invertible matrices with entries in K; such a map
is called matrix representation of G.
If ρ : G → GL(V ) is a representation of G on V , then we often write gv
instead of ρ(g)v, if no confusion is possible. Also, G is said to act linearly on V ,
and V is called a Gmodule. The dimension of the representation is by deﬁnition
the dimension of V . Note that (g, v) → ρ(g)v is indeed an action of G on V in
the sense of Deﬁnition 2.4.1.
Exercise 4.1.2. To experience the use of matrix representations of groups, try
to guess which orthogonal linear map is the product ρ
x
ρ
y
ρ
z
. Here ρ
x
stands
for rotation over π about the xaxis, and similarly for y and z. To check your
guess, represent the rotations by matrices and compute the product.
51
52 CHAPTER 4. REPRESENTATION THEORY
Example 4.1.3. Let G be the cyclic group C
n
of order n with generator c. It
has onedimensional complex representations r
l
(l ∈ Z) deﬁned by
r
l
: C
n
→ GL(1, C), c
k
→ (e
kl2πi/n
).
Note that r
l
is faithful if and only if n and l are coprime.
Example 4.1.4. Recall (2.3) the permutation matrix presentation of the sym
metric group S
5
given in Chapter 2: to each permutation π ∈ S
5
, we associate
the 5 5 matrix M
π
with M
π
e
i
= e
π(i)
. This is a matrix representation of S
5
.
The same thing works for S
3
. The permutation (1, 2) can be represented by
the matrix
M
(1,2)
=
¸
0 1 0
1 0 0
0 0 1
¸
,
and (1, 2, 3) can be represented by
M
(1,2,3)
=
¸
0 0 1
1 0 0
0 1 0
¸
.
We know that (1, 2)(1, 2, 3) = (2, 3) and this can be seen from the matrices as
well:
M
(1,2)
M
(1,2,3)
=
¸
1 0 0
0 0 1
0 1 0
¸
.
Exercise 4.1.5. Prove that the correspondence given above between S
3
and
GL(n, K) is a group homomorphism, that is, a matrix representation of S
3
.
Example 4.1.6. Recall from 3.3 the dihedral group D
n
of order 2n. It contains
the cyclic group C
n
of order n with generator c as a subgroup of index 2 and has
a reﬂection a (of order 2) with aca = c
−1
. The group D
n
has a twodimensional
real representation r : D
n
→ GL(2, R) determined by
c
k
→
cos(2kπ/n) −sin(2kπ/n)
sin(2kπ/n) cos(2kπ/n)
, a →
0 1
1 0
.
Example 4.1.7. An important example of a linear representation of a ﬁnite
group G is the regular representation. Let G = ¦g
1
, . . . , g
n
¦ and take V to be
the vector space of dimension n and basis ¦e
g1
, . . . , e
gn
¦. Deﬁne ρ: G → GL(V )
by ρ(g
i
)(e
gj
) = e
gigj
, extended linearly to all of V .
Example 4.1.8. The above examples can be summarized in a more general
construction. Let π : G → S
n
be a permutation representation. Then π gives
rise to a linear representation M
π
: G → GL(V ), where V is a complex vector
space with (formal) basis e
i
(i = 1, . . . , n), in the following way:
M
π
(g)e
j
= e
π(g)j
(g ∈ G; j ∈ ¦1, . . . , n¦).
4.1. LINEAR REPRESENTATIONS OF GROUPS 53
The regular representation of Example 4.1.7 is M
π
with π : G → Sym(G)
the left regular permutation representation of G.
The matrix representation of Example 4.1.4 is M
π
, where π is the natural
permutation representation of S
n
(actually, for n = 5, 3, respectively).
Example 4.1.9. An important class of constructions of new representations
from known ones arises from function space constructions. Let r : G → GL(V )
be a linear representation of G on V . Consider the vector space F(V ), of all
maps V → K with pointwise addition (f + g)x = f(x) + g(x) and pointwise
scalar multiplication (λf)x = λf(x) (where λ ∈ K, x ∈ V , f, f
: V →K). Then
r induces the following representation r
∗
: G → GL(F(V )):
r
∗
(g)(f)x = f(r(g)
−1
x) (x ∈ V, f ∈ F(V ), g ∈ G).
As done before, we usually unwind the heavy notation by dropping r wherever
possible, and write (gf)(x) = f(g
−1
x) instead.
Note that F(V ) has the dual space V
∗
of Klinear functions as an invariant
subspace.
Exercise 4.1.10. Let R[x]
≤3
be the linear space of all polynomials in x of degree
≤ 3. For a ∈ R we consider the Rlinear representation ρ
a
: R → GL(R[x]
≤3
)
deﬁned as ρ
a
(r)p = p(x −ar) (r ∈ R, p ∈ R[x]
≤3
). Find a corresponding matrix
representation.
Equivalence and Subrepresentations
Just as with groups, we are mainly interested in representations up to isomor
phism.
Deﬁnition 4.1.11. Let r and s be representations of a group G on vector
spaces V and W, respectively. A linear map T : V → W is said to intertwine r
and s if s(g)T = Tr(g) for all g ∈ G. In this case, T is called a homomorphism
of Gmodules. The space of all homomorphisms of Gmodules is denoted by
Hom
G
(V, W).
The representations r and s are called equivalent if there exists a linear
isomorphism T : V → W intertwining them. In this case, T is an isomorphism
of Gmodules.
It is straightforward to prove that this really deﬁnes an equivalence relation
∼ on representations.
Exercise 4.1.12. Consider the linear representations r
l
of C
n
deﬁned in Exam
ple 4.1.3. For which pairs (l, m) are the representations r
l
and r
m
equivalent?
Exercise 4.1.13. Consider the linear representations ρ
a
of Exercise 4.1.10.
Show that ρ
a
∼ ρ
b
if and only if ab = 0.
In Example 4.1.9, we saw that V
∗
is a invariant under the action of G on
F(V ). This gives rise to the following deﬁnition.
54 CHAPTER 4. REPRESENTATION THEORY
Deﬁnition 4.1.14. Let r be a representation of a group G on V . A subspace
W is invariant if r(g)W ⊂ W for all g ∈ G. In this case, r[
W
: G → GL(W)
given by r[
W
(g) = r(g)[
W
, is called a subrepresentation of r, and W is called a
subGmodule of V .
If G is clear from the context, we leave it out and call W a submodule of V .
Example 4.1.15. Suppose that T : V → W is a Gmodule homomorphism
of G on V and W. Then ker T is a Ginvariant subspace of V and imT an
invariant subspace of W.
Example 4.1.16. Consider the twodimensional matrix representation ρ: R →
GL(R
2
) deﬁned by
M
t
=
1 t
0 1
.
Then it is not hard to see that the 1 dimensional subspace spanned by (1, 0) is
an invariant subspace.
Example 4.1.17. Let V be a ﬁnitedimensional vector space over K, and let
ρ : G → GL(V ) be a representation of a group g. Let P(V ) be the set of all poly
nomial functions on V , i.e., the set of all elements of F(V ) (see Example 4.1.9)
that can be written as a Klinear combination of products of elements of V
∗
.
The subspace P(V ) of F(V ) is invariant under G, and so are its homogeneous
components P(V )
d
, deﬁned by
P(V )
d
= ¦f ∈ P(V ) [ f(λx) = λ
d
f(x) for all x ∈ V and λ ∈ K¦.
We ﬁnd that P(V )
0
is the onedimensional space of constant functions, and
P(V )
1
= V
∗
has dimension n := dimV . Choosing a basis x
1
, . . . , x
n
of V
∗
, we
ﬁnd that P(V )
d
is precisely the Klinear span of monomials x
a1
1
x
an
n
where
(a
1
, . . . , a
n
) ∈ N
n
satisﬁes a
1
+. . . +a
n
= d. This implies dimP(V )
d
=
n+d−1
d
;
prove this!
To make all this more explicit, suppose that n = 2, and consider an element
A ∈ GL(V ) whose matrix is
a b
c d
with respect to the basis e
1
, e
2
dual to x
1
, x
2
.
We compute the matrix of A on P(V )
1
with respect to x
1
, x
2
. We have
A
−1
= (ad −bc)
−1
d −b
−c a
,
so
Ax
1
(e
1
) = x
1
(A
−1
e
1
) = (ad −bc)
−1
x
1
(de
1
−ce
2
) = (ad −bc)
−1
d
Ax
1
(e
2
) = x
1
(A
−1
e
2
) = (ad −bc)
−1
x
1
(−be
1
+ae
2
) = −(ad −bc)
−1
b
Ax
2
(e
1
) = x
2
(A
−1
e
1
) = (ad −bc)
−1
x
2
(de
1
−ce
2
) = −(ad −bc)
−1
c
Ax
2
(e
2
) = x
2
(A
−1
e
2
) = (ad −bc)
−1
x
2
(−be
1
+ae
2
) = (ad −bc)
−1
a
4.1. LINEAR REPRESENTATIONS OF GROUPS 55
from which we conclude Ax
1
= (ad − bc)
−1
(dx
1
− bx
2
) and Ax
2
= (ad −
bc)
−1
(−cx
1
+ax
2
). Hence A has matrix
(ad −bc)
−1
d −c
−b a
on the basis x
1
, x
2
of P(V )
1
. Note that this is the tranposed inverse of A.
Finally, let us compute the matrix of A on P(V )
2
with respect to the basis
x
2
1
, x
1
x
2
, x
2
2
. Now
Ax
2
1
= (Ax
1
)
2
= (ad −bc)
−2
(dx
1
−bx
2
)
2
= (ad −bc)
−2
(d
2
x
2
1
−2bdx
1
x
2
+b
2
x
2
2
)
A(x
1
x
2
) = (Ax
1
)(Ax
2
) = (ad −bc)
−2
(dx
1
−bx
2
)(−cx
1
+ax
2
)
= (ad −bc)
−2
(−cdx
2
1
+ (ad +bc)x
1
x
2
−abx
2
2
)
Ax
2
2
= (Ax
2
)
2
= (ad −bc)
−2
(−cx
1
+ax
2
)
2
= (ad −bc)
−2
(c
2
x
2
1
−2acx
1
x
2
+a
2
x
2
2
),
so the matrix sought for is
(ad −bc)
−2
¸
d
2
−cd c
2
−2bd ad +bc −2ac
b
2
−ab a
2
¸
.
In conclusion, we have found a matrix representation of GL(2, K) on K
3
sending A to the 3 3 matrix just computed. Can you determine its kernel?
Exercise 4.1.18. Find the invariant subspaces of the matrix representation
ρ: R → GL(R
2
) deﬁned by
M
t
=
e
t
e
t
−1
0 1
.
Lemma 4.1.19. Suppose that G is ﬁnite, and that ρ is a representation of
G on a vector space V . Let U be a subGmodule of V . Then there exists a
subGmodule W of V such that V = U ⊕W.
Proof. Let W
be any vector space complement of U in V , and let π
be the
projection of V onto U with kernel W
. Deﬁne a new map
π :=
1
[G[
¸
g∈G
ρ(g)π
ρ(g
−1
)
on V . We claim that π is a projection commuting with all ρ(h), h ∈ G. To see
that it does indeed commute with ρ(h), consider
πρ(h) =
1
[G[
¸
g∈G
ρ(g)π
ρ(g
−1
h);
56 CHAPTER 4. REPRESENTATION THEORY
by the dummy transformation g → hg this equals
1
[G[
¸
g∈G
ρ(hg)π
ρ((hg)
−1
h)
=
1
[G[
¸
g∈G
ρ(h)ρ(g)π
ρ(g
−1
h
−1
h)
= ρ(h)
1
[G[
¸
g∈G
ρ(g)π
ρ(g
−1
)
= ρ(h)π,
as claimed. To verify that π is a projection, we ﬁrst note that ππ
= π
; indeed,
both are zero on W
and the identity on U. Compute
π
2
= π
1
[G[
¸
g∈G
ρ(g)π
ρ(g
−1
)
=
1
[G[
¸
g∈G
ρ(g)ππ
ρ(g
−1
)
=
1
[G[
¸
g∈G
ρ(g)π
ρ(g
−1
)
= π,
where the second equality follows from the fact that π commutes with each
ρ(g). This concludes the proof that π is a projection. This implies that V =
im(π)⊕ker(π), where both im(π) and ker(π) are subGmodules of V by Exercise
4.1.15. Now note that im(π) = im(π
) = U, so that we may take W = ker(π)
to conclude the proof of the lemma.
Remark 4.1.20. Let ρ be a linear representation of G on a ﬁnitedimensional
complex vector space V . Suppose that V is endowed with a Hermitian inner
product (., .) satisfying
(ρ(g)(v), ρ(g)(w)) = (v, w) ∀v, w ∈ V, g ∈ G.
Now if U is a Ginvariant subspace of V , then the orthogonal complement U
⊥
of U in V is an invariant subspace of V complementary to U.
It is important note that, if G is ﬁnite, one can always build an invariant
Hermitian inner product (.[.) from an arbitrary one (., .) as was done in Exercise
3.2.22 for positive deﬁnite inner products:
(v[w) =
¸
g∈G
(ρ(g)(v), ρ(g)(w)).
Thus, the above argument gives an alternative proof of the lemma in the case
where V is a ﬁnitedimensional complex vector space.
4.1. LINEAR REPRESENTATIONS OF GROUPS 57
Exercise 4.1.21. Show that in Exercises 4.1.10 and 4.1.18, there are invariant
subspaces that have no invariant complements.
Deﬁnition 4.1.22. A Gmodule V is called reducible if it has subGmodules
other than 0 and V ; it is called irreducible otherwise. If V is the direct sum of
irreducible subGmodules, then V is called completely reducible.
This terminology is also used for the corresponding representations. By def
inition, each irreducible representation is completely reducible. Exercise 4.1.18
gives an example of a reducible completely reducible representation. In Exer
cise 4.1.27 we shall see that not every reducible representation is completely
reducible.
Exercise 4.1.23. Consider the cyclic group C
n
of order n with generator c. It
has a twodimensional real representation deﬁned by
r : C
n
→ GL(2, R), c
k
→
cos(2kπ/n) −sin(2kπ/n)
sin(2kπ/n) cos(2kπ/n)
.
Show that it is irreducible for n > 2. If we replace R by C, we obtain a complex
representation. Show that this representation is completely reducible.
Exercise 4.1.24. Show that any irreducible representation of an Abelian group
G over an algebraically closed ﬁeld is onedimensional.
For ﬁnite groups, irreducible representations are the building blocks of all
representations.
Theorem 4.1.25. For a ﬁnite group G, every ﬁnitedimensional Gmodule is
completely reducible.
Proof. Proceed by induction. Suppose that the stament holds for all representa
tions of dimension smaller than n, and let V be an ndimensional representation.
If V is irreducible, then we are done. Otherwise there exists a subGmodule U
with 0 U V . By Lemma 4.1.19, U has a invariant complementary subG
module W. As both dimU and dimW are smaller than n, we may apply the
induction hypothesis, and ﬁnd that U is the direct sum of irreducible subG
modules U
1
, . . . , U
k
of U and W is the direct sum of irreducible subGmodules
W
1
, . . . , W
l
. But then
V = U
1
⊕. . . ⊕U
k
⊕W
1
⊕. . . ⊕W
l
.
Remark 4.1.26. The representation of R in Example 4.1.16 is reducible, but
it is not the direct sum of irreducibles. Exercise 4.1.21 also gives examples of
this phenomenon. This shows that we need at least some condition on G (like
ﬁniteness) for Theorem 4.1.25 to hold.
58 CHAPTER 4. REPRESENTATION THEORY
Exercise 4.1.27. Let p be a prime, and consider the cyclic group C
p
with
generator c. The map C
p
→ GL
2
(Z/pZ) deﬁned by
c
k
→
1 k
0 1
is a matrix representation of C
p
over Z/pZ. The onedimensional subspace
spanned by (1, 0)
T
is invariant, but it has no invariant complement.
In general, representation theory of a group G over a ﬁeld of characteristic
p tends to be much harder if p divides [G[ than otherwise.
An important result about irreducible representations known as Schur’s
Lemma.
Lemma 4.1.28 (Schur’s Lemma). Suppose that K is algebraically closed, and
let V be an irreducible ﬁnitedimensional Gmodule over K. Then any homo
morphism of Gmodules V → V is a scalar.
In other words, a matrix commuting with all matrices of an irreducible ma
trix representation is a multiple of the identity matrix.
Proof. Let T : V → V be a homomorphism of Gmodules, and let λ ∈ K be
an eigenvalue of T. Then T − λI is also a homomorphism of Gmodules. By
Example 4.1.15, its kernel is a subGmodule of V , and it is nonzero as λ is an
eigenvalue of T. As V is irreducible, ker(T −λI) equals V .
Exercise 4.1.29. Compute the set of all 2 2 matrices commuting with C
n
in one of the 2dimensional complex representations given in Example 4.1.6.
Conclude that the representation is reducible.
Exercise 4.1.30. Prove that the following converse of Schur’s lemma holds:
Let r : G → GL(V ) be a completely reducible representation of a group G. For
r to be irreducible, it suﬃces that any linear map A : V → V that commutes
with all r(g), g ∈ G, is a scalar.
Exercise 4.1.31. Suppose V is a vector space of dimension n with a subspace
W of dimension k where 1 < k < n, and let G be the group of elements in
GL(V ) that leave W invariant. Then the embedding r : G → GL(V ) is clearly
a faithful reducible representation of G on V .
1. Establish that r is not completely reducible.
2. Show that the only Gmodule homomorphisms A : V → V are scalar
multiplications.
3. Conclude that the converse of Schur’s lemma does not necessarily hold if
G is not completely reducible, cf. Exercise 4.1.30.
4.2. DECOMPOSING DISPLACEMENTS 59
3
2 1
Figure 4.1: A system with 3 point masses and springs.
3
2 1
Figure 4.2: A displacement.
4.2 Decomposing Displacements
Consider the system of Figure 4.1. It depicts three point masses (labelled with
P = ¦1, 2, 3¦) at positions (−
1
2
, −
1
2
√
3), (−
1
2
,
1
2
√
3), (0, 1) in twodimensional
space E = R
2
, and these masses are connected with springs of equal lengths.
The system can be thought of as vibrating (in two dimensions) about its equi
librium state, in which the point masses form an equilateral triangle.
A displacement of the system is a function f : P → R
2
which attaches to
each point mass a velocity vector. Displacements can be depicted as in Figure
4.2. The set of all displacements is an Rlinear space, which we denote by Γ.
Its dimension is clearly 3 2 = 6, i.e., there are 6 degrees of freedom.
We would like to decompose Γ into subspaces in a manner that is consis
tent with the symmetry group D
3
that acts on the system. To be precise, we
60 CHAPTER 4. REPRESENTATION THEORY
3
2 1
1 2
3
(1,2,3)
Figure 4.3: The action of (1, 2, 3) on a displacement.
have a permutation representation π : D
3
→ Sym(P). The map is in fact an
isomorphism, and we shall identify D
3
with Sym(P) = S
3
. For example, the
permutation (1, 2) corresponds to reﬂection in the line bisecting the edge 12 and
meeting point 3.
The group D
3
also acts linearly on E. The corresponding linear representa
tion D
3
→ GL(E) is given by
(1, 2) →
−1 0
0 1
and (1, 2, 3) →
1
2
−1
√
3
−
√
3 −1
.
The space Γ becomes a D
3
module by setting
(gf)(i) = gf(g
−1
i), i ∈ P, f ∈ Γ, g ∈ G.
Our goal can now be formalized as follows: try to write Γ as a direct sum of
irreducible submodules. Theorem 4.1.25 tells us that this is possible.
Here, we shall try to solve this problem ‘by hand’, without using too much
representation theory. In this way, the reader will appreciate the elegant char
acter arguments which are to be given in Example 5.2.23.
As a ﬁrst step, we can write each displacement in a unique way as a purely
‘radial’ displacement and a purely ‘tangental’ displacement (see Figure 4.4), and
the spaces Γ
rad
and Γ
tan
of radial and tangental displacements, respectively, are
threedimensional submodules of Γ.
The submodule Γ
rad
contains a onedimensional submodule Γ
1
rad
, consisting
of displacements with the ‘same’ velocity vector in all three directions, such as
the one depicted in Figure 4.5. To be precise: this space consists of all radial
displacements f for which (1, i)(f(1)) = f(i).
4.2. DECOMPOSING DISPLACEMENTS 61
3
2 1
3
2 1
Figure 4.4: A radial (left) and a tangental (right) displacement.
3
2 1
Figure 4.5: A special radial displacement.
62 CHAPTER 4. REPRESENTATION THEORY
3
2 1
3
2 1
Figure 4.6: Two more special radial displacements.
The space Γ
1
rad
must have an invariant complement in Γ
rad
, and looking at
the pictures we ﬁnd that the space Γ
2
rad
of radial displacements spanned by the
two in Figure 4.6 is also invariant. Note that this space consists of all radial
displacements for which (1, 3)f(1) + (1, 2)f(1) +f(1) = 0.
One may wonder if the twodimensional module Γ
2
rad
can be decomposed
further, but this is not the case, as we shall see in Example 5.2.23. The rep
resentation Γ
tan
can be decomposed in a similar fashion. However, instead of
going into details rightaway, we will ﬁrst treat more representation theory, and
come back to this application in Example 5.2.23.
Notes
This section is based on a similar (but more diﬃcult) example in [10]. If your
interest is aroused by this example, you are encouraged to read the pages in
that book concerned with the displacemements of a tetrahedral molecule.
Chapter 5
Character Tables
In this chapter, G denotes a ﬁnite group and K an algebraically closed ﬁeld of
characteristic 0, e.g. C.
5.1 Characters
An extremely useful notion regarding matrices is the trace. The trace of a matrix
A = (a
ij
)
n×n
is tr(A) = a
11
+ +a
nn
.
Exercise 5.1.1. Prove that for two matrices A and B, one has tr(AB) =
tr(BA).
Exercise 5.1.2. Show that in the case of a permutation representation π: G →
S
n
, the number tr(M
π
(g)) (see Example 4.1.8) is equal to the number of ﬁxed
points of the permutation π(g).
Deﬁnition 5.1.3. Let r be a representation of a group G on a ﬁnitedimensional
vector space V . The character of r is the function χ
r
deﬁned on G by
χ
r
(g) = tr(r(g)) (g ∈ G).
If r is irreducible, then the character χ
r
is also called irreducible.
Thus, the character of r on g is the sum of the eigenvalues of r(g) counted
with multiplicities.
Remark 5.1.4. If G has ﬁnite order n, and ρ : G → GL(V ) is a ﬁnite
dimensional representation, then ρ(g)
n
= I for all g ∈ G. Hence, over the
algebraic closure of K, each ρ(g) is diagonalizable, and all its eigenvalues are
nth roots of unity. If K = C, then all eigenvalues have absolute value 1.
We next give some properties of complex characters.
Lemma 5.1.5. For any complex representation r : G → GL(V ) of a ﬁnite
group, we have:
63
64 CHAPTER 5. CHARACTER TABLES
1. χ
r
(e) = tr(I) = dimV ;
2. χ
r
(g
−1
) = χ
r
(g);
3. χ
r
(ghg
−1
) = χ
r
(h) for all g, h ∈ G.
Proof. The ﬁrst part is easy. As for the second part, we have seen that all the
eigenvalues λ
1
, . . . , λ
n
of r(g) have absolute value 1. Therefore:
χ
r
(g
−1
) = tr(r(g
−1
)) = tr(r(g)
−1
) =
¸
λ
−1
i
=
¸
λ
i
= tr(r(g)) = χ
r
(g).
The last part follows directly from Exercise 5.1.1.
The last part states that characters are constant on conjugacy classes of G.
Example 5.1.6. Take G = GL(2, K). The trace of an arbitrary element
A =
a b
c d
is a+d. This is the character of the identity, which in this case is a representation
(often called the natural representation).
But, as we have seen in Example 4.1.17, there are also linear actions on
P(V )
1
and on P(V )
2
. We read oﬀ the corresponding characters as traces of the
matrices found in 4.1.17. They are (a +d)/(ad −bc) on P(V )
1
, and (a
2
+ad +
bc +d
2
)/(ad −bc)
2
on P(V )
2
.
Lemma 5.1.7. Let V
1
, . . . , V
s
be representations of G with characters respec
tively, χ
1
, . . . , χ
s
. Then the character of the representation V
1
⊕ ⊕V
s
is the
sum of characters χ
1
+ +χ
s
.
5.2 Orthogonality of irreducible characters
The following theorem is very important in the representation theory of ﬁnite
groups.
Theorem 5.2.1. Let G be a ﬁnite group. The irreducible complex characters of
G form an orthonormal basis for the space of class functions on G, with respect
to the Hermitian inner product ( [ ).
We need a deﬁnition to make clear what this theorem states.
Deﬁnition 5.2.2. A class function on a group G is a function f : G →C that is
constant on conjugacy classes, or, equivalently, that is ﬁxed by the linear action
of G on C
G
deﬁned by
(g f)(h) := f(ghg
−1
).
The space of all class functions is denoted by H. The Hermitian inner product
is given by
(χ [ ψ) =
1
[G[
¸
g∈G
χ(g)ψ(g)
for χ, ψ ∈ H.
5.2. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 65
We have noticed earlier that characters are class functions. We divide the
proof of Theorem 5.2.1 into two parts: ﬁrst, we will show the orthonormality
of the irreducible characters, and then their being a full basis of the space of
class functions. For our ﬁrst purpose, we introduce the tensor product of two
representations.
Deﬁnition 5.2.3. Let V and W be vector spaces over K. Then V ⊗W, called
the tensor product of V and W, is the quotient of the free vector space over K
with basis ¦v ⊗ w [ v ∈ V, w ∈ W¦ by its subspace spanned by all elements of
the following forms (for v, v
1
, v
2
∈ V , w, w
1
, w
2
∈ W and λ ∈ K):
(v
1
+v
2
) ⊗w −v
1
⊗w −v
2
⊗w,
v ⊗(w
1
+w
2
) −v ⊗w
1
−v ⊗w
2
,
(λv) ⊗w −λ(v ⊗w), and
v ⊗(λw) −λ(v ⊗w).
Exercise 5.2.4. Show that if v
1
, . . . , v
m
and w
1
, . . . , w
n
are bases of V and W,
respectively, then
¦v
i
⊗w
j
[ i = 1, . . . , m, j = 1, . . . , n¦
is a basis of V ⊗W.
It is not hard to see that if U, V, and W are vector spaces over K, then
the spaces (U ⊗ V ) ⊗ W and U ⊗ (V ⊗ W) are canonically isomorphic. We
will identify them, leave out parentheses, and write T
d
(V ) for the dfold tensor
product of V with itself. The direct sum of all T
d
(V ) for d ∈ N (including 0,
for which T
d
(V ) is isomorphic to K) forms an associative algebra with respect
to the operator ⊗; this algebra is called the tensor algebra on V and denoted
by T(V ).
Deﬁnition 5.2.5. Let V be a vector space over Kand let d be a natural number.
Then the quotient of T
d
(V ) by the subspace spanned by the set
¦v
1
⊗v
2
⊗. . . ⊗v
d
−v
π(1)
⊗v
π(2)
⊗. . . ⊗v
π(d)
[ π ∈ S
d
¦
is denoted by S
d
(V ), and called the dth symmetric power of V .
Just like T(V ), the direct sum of all S
d
(V ) for d ∈ N forms an associative
algebra, called the symmetric algebra on V and denoted by S(V ). This algebra
is commutative, and to stress this fact we usually leave out the operator ⊗ from
expressions in S(V ).
Exercise 5.2.6. Show that if v
1
, . . . , v
n
is a basis of V , then the set
¦v
m1
1
v
mn
n
[ m
1
, . . . , m
n
∈ N, m
1
+. . . +m
n
= d¦
is a basis for S
d
(V ). Prove that S
d
(V
∗
) is canonically isomorphic to P(V )
d
(see
Example 4.1.17).
66 CHAPTER 5. CHARACTER TABLES
Deﬁnition 5.2.7. Let r : G → GL(V ) and s : G → GL(W) be linear represen
tations of G. Then we deﬁne a linear representation r ⊗s : G → GL(V ⊗W) of
G by
(r ⊗s)(g)(v ⊗w) := (r(g)v) ⊗(s(g)w),
extended linearly to all of V ⊗W. This representation is denoted by r ⊗s and
called the tensor product of r and s.
Lemma 5.2.8. Let G, r, V, s, and W be as in Deﬁnition 5.2.7. Then the char
acter of r ⊗s satisﬁes χ
r⊗s
= χ
r
χ
s
.
Proof. Fix bases v
1
, . . . , v
m
and w
1
, . . . , w
n
of V and W, respectively. Consider,
for a ﬁxed g ∈ G, the matrix A = (a
ij,kl
) of (r ⊗s)(g) with respect to the basis
v
i
⊗w
j
of V ⊗W; We have
(r ⊗s)(g)(v
k
⊗w
l
) =
¸
i,j
a
ij,kl
v
i
⊗w
j
. (5.1)
On the other hand, by the deﬁnition of the tensor representation, the lefthand
side equals
(r(g)v
k
) ⊗(s(g)w
l
) = (
¸
i
r
ik
v
i
) ⊗(
¸
j
s
jl
w
j
) =
¸
i,j
r
ik
s
jl
v
i
⊗w
j
. (5.2)
Comparing the righthand sides of equations (5.1) and (5.2) we obtain
a
ij,kl
= r
ik
s
jl
,
and therefore
χ
s⊗r
(g) =
¸
i,j
a
ij,ij
=
¸
i,j
r
ii
s
jj
= χ
r
(g)χ
s
(g),
concluding the proof.
Exercise 5.2.9. Let V be a Gmodule, and consider the linear isomorphism
σ : V ⊗V → V ⊗V determined by σ(x ⊗y) = y ⊗x for x, y ∈ V . Show that σ
has precisely two eigenspaces, with eigenvalues 1 and −1, and that both spaces
are Ginvariant. Here G acts on V ⊗ V via the tensor product representation
of the linear representation of G on V (given by the Gmodule structure of V )
with itself.
Another important ingredient in our proof of Theorem 5.2.1 is the socalled
dual representation.
Deﬁnition 5.2.10. Let r : G → GL(V ) be a representation of the group G.
Then we write V
∗
for the dual of V (in the notation of Example 4.1.17, this is
P(V )
1
), and r
∗
for the corresponding representation, given by
(r
∗
(g)f)(v) := f(r(g)
−1
v).
5.2. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 67
Exercise 5.2.11. Show that χ
r
∗
= χ
r
.
One ﬁnal lemma before proceeding with the proof of Theorem 5.2.1.
Lemma 5.2.12. Let r, V, s, W, G be as in Deﬁnition 5.2.7. Deﬁne a repre
sentation t of G in Hom(V, W), the space of all Clinear maps from V to W
by
t(g)H := s(g)Hr(g)
−1
, H ∈ Hom(V, W).
This representation is equivalent to s ⊗r
∗
.
Proof. Deﬁne T : W ⊗V
∗
→ Hom(V, W) by
T(w ⊗f)(v) = f(v) w, v ∈ V, f ∈ V
∗
, w ∈ W,
and extend it linearly. It is a matter of elementary linear algebra to see that T
is a linear isomorphism. We claim that it intertwines the representations t and
s ⊗r
∗
. The best way to see this is to do the computation yourself, so we leave
it as an exercise for the reader.
Exercise 5.2.13. Prove that the map T deﬁned in the proof of Lemma 5.2.12
is really an intertwining map.
If we choose a basis (v
j
)
j
of V with dual basis (v
j
)
j
of V
∗
, and a basis (w
i
)
i
of W, then the map T sends w
i
⊗v
j
to the linear map V → W having the matrix
E
ij
(the matrix with 1 on position (i, j), and zeroes elsewhere) with respect to
the bases (v
j
)
j
and (w
i
)
i
.
Having completed all preparations, of which the tensor product is a very
useful tool in its own right, we proceed with the ﬁrst part of the proof of this
section’s main theorem.
Proof of Theorem 5.2.1, ﬁrst part. Let r, V, s, W, G be as in Deﬁnition 5.2.7 and
t as in Lemma 5.2.12, and moreover, assume that both representations r and
s are irreducible. We wish to compute the inner product (χ
s
[ χ
r
). It can be
written in a particularly nice form, using tensor products.
Deﬁne the linear map P : Hom(V, W) → Hom(V, W) by
P =
1
[G[
¸
g∈G
t(g).
The trace of P equals
1
[G[
¸
g∈G
χ
t
(g) =
1
[G[
¸
g∈G
χ
s
(g)χ
r
∗
(g)
=
1
[G[
¸
g∈G
χ
s
(g)χ
r
(g))
= (χ
s
[ χ
r
).
68 CHAPTER 5. CHARACTER TABLES
Here we used Lemmas 5.2.12, 5.2.8 and Exercise 5.2.11.
We claim that P is a projection from Hom(V, W) onto the space Hom
G
(V, W).
To show that it is indeed a projection, let H ∈ Hom(V, W), and compute
P
2
H =
1
[G[
¸
g
1
[G[
¸
h
s(g)s(h)Hr(h
−1
)r(g
−1
)
=
1
[G[
¸
g
1
[G[
¸
h
s(gh)Hr((gh)
−1
)
=
1
[G[
¸
g
1
[G[
¸
h
s(h)Hr(h
−1
)
=
1
[G[
¸
g
PH
= PH,
where in third equality the dummy h is replaced by g
−1
h. This proves that P
is a projection. To see that im(P) is contained in Hom
G
(V, W), let g
0
∈ G,
H ∈ Hom(V, W), and v ∈ V , and compute
(PH)(g
0
v) =
1
[G[
¸
g
s(g)Hr(g
−1
g
0
)v
=
1
[G[
¸
g
s(g
0
g)Hr(g
−1
g
−1
0
g
0
)v
= s(g
0
)
1
[G[
¸
g
s(g)Hr(g
−1
)v
= g
0
(PHv),
where in the second equality, the dummy g is replaced by g
0
g. Finally, if H ∈
Hom
G
(V, W), then
PH =
1
[G[
¸
g
s(g)Hr(g
−1
)
=
1
[G[
¸
g
s(g)s(g
−1
)H
= H,
so that im(P) = Hom
G
(V, W). This proves the claim. We now distinguish two
cases.
First, suppose that r and s are not equivalent. Then, by Schur’s lemma, the
space Hom
G
(V, W) is zerodimensional, so that P and, a fortiori, its trace are
0. In this case (χ
s
[ χ
r
) = 0.
Alternatively, suppose that r and s are equivalent. As we only wish to
compute the inner product of their characters, we may assume that they are
5.2. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 69
equal: W = V and s = r. In this case, the image of P is the linear span of the
identity map I on V . Taking any basis of Hom(V, V ) that starts with I, we see
that the matrix of P with respect to that basis has only nonzero elements on
the ﬁrst row. The only contribution to the trace of P is therefore the element in
the upper left corner. Since PI = I, that element is 1. Hence (χ
r
[ χ
r
) = 1.
This proves that the characters are an orthonormal system in the space of
class functions on G.
Corollary 5.2.14. The number of nonisomorphic irreducible representations
of G is ﬁnite.
Proof. The dimension of H is the number of conjugacy classes of G. The irre
ducible characters form an orthonormal, whence independent, set in this ﬁnite
dimensional space by the above.
Lemma 5.2.15. Let r be a complex representation of G on a space V , and
write
V = V
1
⊕ ⊕V
t
,
where the V
i
are irreducible submodules. Let s be an irreducible representation of
G on a space W. Then the number of i for which V
i
is isomorphic to W is equal
to (χ
r
[χ
s
). In particular, this number does not depend on the decomposition in
irreducibles.
Proof. If χ
i
is the character of V
i
then the character χ
r
of V can be written as
χ
r
= χ
1
+ +χ
t
. Hence
(χ
r
[χ
s
) = (χ
1
[χ
s
) + + (χ
t
[χ
s
),
and by the above (χ
i
[χ
s
) is one if V
i
∼
= W, and zero otherwise.
Corollary 5.2.16. Two representations of a ﬁnite group are equivalent if and
only if they have the same character.
Recall that the regular representation ρ: G → GL(V ) of the group G is
deﬁned by ρ(g)(e
h
) = e
gh
, with V the vector space of dimension [G[ and basis
¦e
g
[ g ∈ G¦.
Proposition 5.2.17. The character of the regular representation ρ of G satis
ﬁes ρ(g) = δ
g,e
[G[ (g ∈ G).
Proof. Since eg = g, the transformation ρ(e) ﬁxes each element g of G. There
fore, tr(ρ(e)) = n = [G[. If g = e, then e
h
= e
gh
for each h ∈ G. This implies
that tr(ρ(g)) = 0.
Lemma 5.2.18. Every irreducible representation W of G is a subrepresentation
of the regular representation with multiplicity dimW.
70 CHAPTER 5. CHARACTER TABLES
Proof. From Proposition 5.2.17 we have
(χ
ρ
[ χ
s
) =
1
[G[
¸
h∈G
χ
ρ
(h)χ
s
(h) =
1
[G[
[G[χ
s
(e) = dim(W),
and the result follows from Lemma 5.2.15.
Corollary 5.2.19. Let r
1
, r
2
, . . . , r
s
be the irreducible representations of G.
Then
[G[ =
¸
n
2
i
where n
i
is the dimension of the representation r
i
.
Proof. We saw in the lemma that the regular representation ρ can be written
as a sum of all irreducible representations r
i
of G. Moreover, each of these
representations appears n
i
times in the decomposition. So
ρ =
n1 times
. .. .
r
1
⊕. . . ⊕r
1
⊕. . . ⊕
ns times
. .. .
r
s
⊕ ⊕r
s
.
Therefore [G[ = χ
ρ
(e) =
¸
s
i=1
n
i
χ
ri
(e) =
¸
s
i=1
n
2
i
.
Suppose that we manage to construct some nonisomorphic irreducible rep
resentations of G. Then Corollary 5.2.19 can be used to decide whether we have
already ﬁnd all.
In order to ﬁnish the proof of Theorem 5.2.1, we still need to show that the
irreducible characters span H.
Lemma 5.2.20. Suppose that r : G → GL(V ) is an irreducible ndimensional
complex representation of G, and let f ∈ H. Then the linear map
H :=
1
[G[
¸
g∈G
f(g)r(g)
equals
1
n
(f [ χ
r
)I.
Proof. First of all, note that H veriﬁes r(g)H = Hr(g) for all g ∈ G. Hence, as
r is irreducible, H must be multiplication by a scalar λI. Its trace equals nλ.
On the other hand, it equals
1
[G[
¸
g∈G
f(g)χ
r
(g) = (f [ χ
r
).
Consequently,
λ = (f [ χ
r
)/n.
5.2. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 71
Proof of Theorem 5.2.1, second part. In order to prove that the irreducible char
acters form a basis of H, it suﬃces to prove that the orthoplement of their linear
span in H is zero. Hence, let f ∈ H be orthogonal to all irreducible characters.
According to Lemma 5.2.20, the linear map
H =
1
[G[
¸
g∈G
f(g)r(g)
is zero for all irreducible representations r. But, writing an arbitrary represen
tation as a direct sum of irreducible ones, we see that this linear map is zero for
any representation r. Let us apply this knowledge to the regular representation.
Compute
He
h
=
1
[G[
¸
g∈G
f(g)r(g)e
h
=
1
[G[
¸
g∈G
f(g)e
gh
.
Thus, for H to be zero, f must be identically zero. This concludes the proof.
Theorem 5.2.21. The number of irreducible characters equals the number of
conjugacy classes of G.
Proof. The dimension of H equals the number of conjugacy classes of G, and
the irreducible characters form a basis of this linear space.
Half of the following proposition is Exercise 4.1.24, and can be done without
character theory.
Proposition 5.2.22. The group G is Abelian if and only if all of its irreducible
representations are one dimensional.
Proof. Denote by s the number of irreducible characters of G and denote their
dimensions, as before by n
i
.
If G is Abelian, all conjugacy classes consist of a single element of G. Then
by the theorem above, there are [G[ irreducible representations of G. Hence by
Corollary 5.2.19, all n
i
’s are equal to 1. This settles one implication.
As for the converse, suppose that n
i
= 1 for all i. Then [G[ = s and so there
are [G[ distinct conjugacy classes of G. Therefore, G is Abelian.
Example 5.2.23. Recall from Section 4.2 the D
3
module Γ of displacements.
Let us employ character theory to ﬁnd the decomposition of Γ into irreducible
modules. First of all, note that D
3
has three conjugacy classes, with represen
tatives (1), (12) and (123). Hence D
3
has three distinct irreducible characters
by Theorem 5.2.21. Let V = R
3
, and let s : D
3
→ GL(V ) be the linear rep
resentation corresponding to the permutation representation of D
3
on ¦1, 2, 3¦
(see Example 4.1.8). Then
χ
s
((1)) = 3, χ
s
((1, 2)) = 1, and χ
s
((1, 2, 3)) = 0.
72 CHAPTER 5. CHARACTER TABLES
We have (χ
s
[χ
s
) = (9 + 3 ∗ 1)/6 = 2, so that V is reducible. Indeed, the line
spanned by (1, 1, 1) ∈ V is invariant. Let r
1
be the representation of D
3
on this
line, and let r
2
be the representation of D
3
on an invariant complement. Then
χ
r1
((1)) = 1, χ
r1
((1, 2)) = 1, and χ
r1
((1, 2, 3)) = 1
and
χ
r2
((1)) = 2, χ
r2
((1, 2)) = 0, and χ
r2
((1, 2, 3)) = −1.
Both χ
r1
and χ
r2
have squared norm 1, so they are irreducible. The represen
tation of D
3
on E is equivalent to r
2
, as can be read of from the traces of the
matrices on page 4.2.
An element of Γ can be extended in a unique way to a linear map V → E, so
that we may identify Γ with the space Hom
K
(V, E) of Klinear maps from V to
E. By Lemma 5.2.12, Hom
K
(V, E) is isomorphic to V
∗
⊗E; let r : D
3
→ V
∗
⊗E
be the corresponding representation. Lemma 5.2.8 shows that χ
r
= χ
s
χ
r2
.
Hence,
χ
r
((1)) = 6, χ
r
((1, 2)) = 0, and χ
r
((1, 2, 3)) = 0.
First, we can decompose χ
r
= (χ
r1
+ χ
r2
)χ
r2
= χ
r1
χ
r2
+ (χ
r2
)
2
= χ
r2
+
(χ
r2
)
2
. The character (χ
r2
)
2
has squared norm 3, so it is reducible. We have
(χ
r1
[(χ
r2
)
2
) = 1 and also (χ
r2
[(χ
r2
)
2
) = 1, so that (χ
r2
)
2
−χ
r1
−χ
r2
is the char
acter of a third irreducible representation, say r
3
, of dimension 1. Its character
equals
χ
r3
((1)) = 1, χ
r3
((12)) = −1, and χ
(r3)
(123) = 1,
that is, χ
r3
= sgn . By Theorem 5.2.21, the characters r
1
, sgn , r
2
exhaust all
irreducible characters of D
3
. We conclude that the character χ
r
of D
3
on Γ
decomposes into χ
r1
+ 2χ
r2
+χ
r3
.
5.3 Character tables
Let χ
1
, χ
2
, . . . , χ
s
be the characters of the distinct irreducible representations of
G. Let C
1
, C
2
, . . . , C
s
be the distinct conjugacy classes of G, with representatives
g
1
, . . . , g
s
, respectively.
Construct a table with rows labelled by χ
1
, χ
2
, . . . , χ
s
and columns labelled
by g
1
, g
2
, . . . , g
s
; under each g
i
we write the size of C
i
. The (i, j)entry of the
table is χ
i
(g
j
). It is common to take C
1
= ¦e¦ = ¦g
1
¦, so that the ﬁrst column
contains the dimensions of the respective characters, and also to take χ
1
to be
the trivial representation, so that the ﬁrst row contains only ones.
This table is called the character table of G.
Example 5.3.1. Let us check the character table of S
3
. First notice that there
are three conjugacy classes: C
1
= e, C
2
= ¦(12), (13), (23)¦ consisting of all
transpositions and C
3
= ¦(123), (132)¦ consisting of all 3cycles. By Theorem
5.2.21, we know that there are three irreducible representations.
By convention, χ
1
is the trivial character. For the second one we take the
sign representation. Recall this is the one that assigns to each cycle its sign
5.3. CHARACTER TABLES 73
(2cycles are odd, 3cycles are even). Since it is a 1dimensional representation,
it is equal to its character.
Finally for the third representation, we proceed as follows. The permutation
representation π of S
3
on C
3
has an invariant subspace U spanned by (1, 1, 1).
Its orthogonal complement is also an invariant subspace, say V ; it consists of
all (z
1
, z
2
, z
3
) ∈ C
3
such that z
1
+ z
2
+ z
3
= 0. The latter representation has
dimension 2 and is irreducible; it is called the standard representation of S
3
.
To ﬁnd its character, we observe that χ
V
= χ
π
−χ
U
. But U is just the trivial
representation and χ
π
is equal to 3 on C
1
, 1 on C
2
and 0 on C
3
(why? see
Exercise 5.1.2)
We have found the complete character table.
S
3
(1) (12) (123)
1 3 2
trivial U 1 1 1
sign 1 −1 1
standard V 2 0 −1
Proposition 5.3.2. The standard representation of S
n
is irreducible for all
n > 1.
Proof. Let χ be the character of the ndimensional linear representation corre
sponding to the permutation representation of S
n
on ¦1, . . . , n¦ (see Example
4.1.8), and let χ
1
be the trivial character of S
n
. We wish to show that χ−χ
1
is
irreducible; to this eﬀect, it suﬃces to prove that (χ[χ) = 2. Denoting by ﬁx(σ)
the set of ﬁxed points of σ ∈ S
n
on ¦1, . . . , n¦, we have χ(σ) = [ ﬁx(σ)[. Hence,
we ﬁnd
(χ[χ) =
1
n!
¸
σ∈Sn
[ ﬁx(σ)[
2
.
To evaluate the righthand side of this equation we count the number of per
mutations having precisely k ﬁxed points (0 ≤ k ≤ n). This number is clearly
equal to
n
k
F(n −k),
where F(m) denotes the number of ﬁxedpointfree permutations in S
m
. A
standard inclusionexclusion argument shows that
F(m) =
m
¸
i=0
(−1)
i
m!
i!
.
74 CHAPTER 5. CHARACTER TABLES
Combining the above considerations we ﬁnd
(χ[χ) =
1
n!
n
¸
k=0
k
2
n
k
n−k
¸
i=0
(−1)
i
(n −k)!
i!
=
n
¸
k=0
k
2
1
k!
n−k
¸
i=0
(−1)
i
i!
=: M
n
.
It is readily seen that M
2
= 2. To prove that M
n
= 2 for all n ≥ 2, we note
that for n ≥ 3 we have:
M
n
−M
n−1
=
n
¸
i=0
(n −i)
2
(−1)
i
1
(n −i)!i!
,
and it is not hard to prove that the righthand side is zero for n ≥ 3.
Example 5.3.3. Consider the cyclic group C
n
with generator c, and let ζ ∈ C
be a primitive nth root of unity. Deﬁne, for j = 0, . . . , n − 1, the map χ
j
:
C
n
→ C by χ
j
(c
k
) := ζ
jk
. Then the χ
j
are characters of C
n
, and irreducible
as they are onedimensional. By Theorem 5.2.21 there are no other characters,
so that the character table of C
n
is as follows.
C
n
e c c
2
. . . c
n−1
1 1 1 . . . 1
χ
1
1 1 1 . . . 1
χ
2
1 ζ ζ
2
. . . ζ
n−1
χ
3
1 ζ
2
ζ
4
. . . ζ
2(n−1)
. . .
. . .
. . .
χ
n
1 ζ
n−1
ζ
2(n−1)
. . . ζ
(n−1)
2
Example 5.3.4. To compute the character table of S
4
, we ﬁrst count the
number of conjugacy classes. They are the identity, the 2cycles, the 3cycles,
the 4cycles and the product of two distinct 2cycles; ﬁve in total. Hence, by
Theorem 5.2.21, S
4
has ﬁve distinct irreducible characters. It has three of them
in common with S
3
(or, indeed, with any symmetric group; see Proposition
5.3.2), namely the trivial character, the sign and the standard character. They
are listed in the following partial character table.
S
4
(1) (12) (123) (1234) (12)(34)
1 6 8 6 3
χ
1
(trivial) 1 1 1 1 1
χ
2
(sign) 1 −1 1 −1 1
χ
4
(standard) 3 1 0 −1 −1
5.3. CHARACTER TABLES 75
By Lemma 5.2.19, the squares of the dimensions of the irreducible characters
sum up to [S
4
[ = 24, and we only have 1 +1 +9 = 11 so far. We conclude that
the two remaining irreducible characters have dimensions 2 and 3. The latter is
the product χ
5
:= χ
2
χ
4
, which is irreducible as it has norm 1. The remaining 2
dimensional character χ
3
can be found using the orthogonality relations among
characters. We then ﬁnd the following character table of S
4
.
S
4
(1) (12) (123) (1234) (12)(34)
1 6 8 6 3
χ
1
1 1 1 1 1
χ
2
1 −1 1 −1 1
χ
3
2 0 −1 0 2
χ
4
3 1 0 −1 −1
χ
5
= χ
2
χ
4
3 −1 0 1 −1
Exercise 5.3.5. Compute the number of elements in each conjugacy class of
S
4
.
Exercise 5.3.6. Consider the permutation representation of S
4
on partitions
of ¦1, 2, 3, 4¦ into two sets of size 2; that is, the action of S
4
on the set
¦¦A, B¦ [ [A[ = [B[ = 2, A∪ B = ¦1, 2, 3, 4¦, A∩ B = ∅¦.
Show that the character of the corresponding linear representation is χ
1
+ χ
3
;
this explains the character χ
3
.
A more or less explicit description of the irreducible representations of S
n
is
known for general n. It involves the beautiful combinatorics of Young tableaux,
a good textbook on which is Fulton’s book [5].
Exercise 5.3.7. Construct the character table of D
4
.
Exercise 5.3.8. Compute the character table of the Klein group (Z/2Z) ⊕
(Z/2Z).
Example 5.3.9. Consider the subgroup T of the orthogonal group SO
3
(R)
consisting of the proper rotations of the tetrahedron. We wish to compute its
character table. The conjugacy classes are those of (1), (234), (243) and (13)(24)
(make a picture that corresponds to this numbering), and their cardinalities are
1, 4, 4 and 3, respectively. Thus, there must be 4 irreducible characters, of
representations with dimensions m
1
≤ m
2
≤ m
3
≤ m
4
. The sum of their
squares should equal [T[ = 12. Their turns out to be only one solution, namely
m
1
= m
2
= m
3
= 1 and m
4
= 3. Hence, apart from the trivial character, there
should be two more characters of degree 1. Let r be one of them. As r is nothing
but a homomorphism from T to C
∗
, the complex number r((234)) should have
multiplicative order 3; hence it is either ω = e
2πi/3
or ω
2
. The image of (243) is
ﬁxed in both cases, and so is the image of (13)(24), as χ
r
should be orthogonal
to the trivial representation. This explains the ﬁrst three lines in Table 5.1. The
last line can be obtained from the orthonormality conditions.
76 CHAPTER 5. CHARACTER TABLES
T (1) (234) (243) (13)(24)
1 4 4 3
χ
1
1 1 1 1
χ
2
1 ω ω
2
1
χ
3
1 ω
2
ω 1
χ
4
3 0 0 −1
Table 5.1: The character table of the tetrahedral group
Exercise 5.3.10. Find a representation corresponding to the last line of Table
5.1.
Exercise 5.3.11. Let G be a group and let H be a normal subgroup of G.
Let r be a representation of G with the property that H ⊆ ker r. Show that r
induces a representation of G/H. Conversely, any representation of G/H can
be lifted to one on G. Apply this principle to the group T, which has the Klein
4group as a normal subgroup.
Exercise 5.3.12. Prove the column orthogonality of the character table: for
g, h ∈ G, and χ
1
, . . . , χ
s
a complete set of irreducible characters of G,
s
¸
i=1
χ
i
(g)χ
i
(h) = [C
G
(h)[δ
g,h
.
This formula is also useful for the construction of character tables.
Exercise 5.3.13. Let χ be the character of a complex linear representation of
G on V . Show that the character of G on the symmetric square S
2
(V ) of V
equals
g →
χ(g)
2
+χ(g
2
)
2
.
What is the character of G on S
3
(V )? Use these results to ﬁnd new characters
of A
5
from the irreducible 4dimensional character occurring in the standard
permutation representation of degree 5.
Exercise 5.3.14. Consider the alternating group A
5
on 5 letters.
1. Write down representatives for each of its conjugacy classes.
2. Determine the sizes of the corresponding conjugacy classes.
3. Show that A
5
has only one irreducible character of degree 1. (Hint: if
there were one more, then there would have to be a normal subgroup N
such that A
5
/N is abelian.)
5.4. APPLICATION: SYMMETRY OF VIBRATIONS 77
4. The rotation group of the icosahedron is A
5
. This gives an irreducible
representation ρ of A
5
of degree 3. Write down its character.
5. Verify that the composition of ρ with conjugation by (1, 2) gives a non
equivalent representation and determine its character.
6. Apply the symmetric square formula of Exercise 5.3.13 to ρ to ﬁnd a
6dimensional character. Show that this character splits into the trivial
character and an irreducible 5dimensional character.
7. Conclude from the orthogonality of the character table that there is one
more irreducible character. Compute this character, and compare it with
the character of the restriction of the standard representation of S
5
to A
5
.
Exercise 5.3.15. Up to isomorphism, there are precisely two groups of order 8.
They have identical character tables. Exhibit these groups and their character
tables.
5.4 Application: symmetry of vibrations
There exist many spectroscopic techniques in chemistry that can be used to mea
sure symmetry properties of molecules of a given substance. In this section, we
describe two of them: infrared (IR) spectroscopy and Raman spectroscopy. In
the former method, the substance is exposed to radiation of various frequencies,
and the intensity of the light leaving the substance is measured. If the radiation
excites the molecules from the ground state into vibration, then it is absorbed,
giving rise to a (downward) peak in the intensity (plotted as a function of the
frequency). As we have seen in Section 4.2, there are several essentially diﬀerent
vibrations, corresponding to irreducible submodules of the representation of the
symmetry group of the molecule on the space of all displacements. As with Ra
man spectroscopy—another technique whose physical details we do not discuss
here—only actual vibrations may contribute a peak to the intensity spectrum,
i.e., the (inﬁnitesimal) isometries of threedimensional space do not contribute.
However, not even all irreducible modules in the module of vibrations can be
observed by IR spectroscopy; due to technical properties of this technique, only
those irreducibles occurring also in the natural 3dimensional representation do
actually contribute a peak in the spectrum. This distinguishes IR from Ra
man spectroscopy, which allows to observe all irreducibles that occur in the
symmetric square of the aforementioned natural representation.
Hence, to predict the number of peaks using either of the above methods,
for a given molecule with n molecules, we proceed as follows.
1. Determine the symmetry group G of the molecule at hand, and compute
or look up its character table.
2. Determine the character χ
dis
of G on the 3ndimensional module of dis
placements of the molecule, as well as its character χ
iso
on the 6dimensional
78 CHAPTER 5. CHARACTER TABLES
Cl Cl
Cl
Cl
P
Cl
Figure 5.1: The PCl
5
molecule
submodule of (inﬁnitesimal) isometries of R
3
. Then
χ
vib
:= χ
dis
−χ
iso
is the character of G on vibrations of the molecule.
3. Compute the character χ
1
of G on R
3
, and the character χ
2
of G on the
symmetric square of R
3
.
4. Then the number of peaks in IR spectroscopy is the number of irreducible
characters in χ
vib
(counted with multiplicity) that also occur in χ
1
, and
and the number of peaks in Raman spectroscopy is the number of irre
ducible characters in χ
vib
(counted with multiplicity) that also occur in
χ
2
.
Example 5.4.1. We want to predict the number of peaks in IR and Raman
spectroscopy when applied to the molecule PCl
5
of Figure 5.1. In these chemical
applications, it is common to use the Schoenﬂies notation (see Table 3.2), and
we will adhere to this tradition in the present example.
1. The symmetry group of this molecule is D
3h
, whose character table is as
follows (E is the identity, C
i
stands for a rotation of order i, σ
h
and σ
v
are reﬂections in a horizontal and vertical plane, respectively, and S
3
is
the composition of a C
3
and a σ
h
).
5.4. APPLICATION: SYMMETRY OF VIBRATIONS 79
D
3h
E C
3
3C
2
σ
h
S
3
σ
v
1 2 3 1 2 3
A
1
1 1 1 1 1 1
A
2
1 1 1 1 1 1
E
2 1 0 2 1 0
A
1
1 1 1 1 1 1
A
2
1 1 1 1 1 1
E
2 1 0 2 1 0
2. Let χ
1
denote the character of D
3h
in its natural representation on R
3
,
and let χ
3
denote the character of D
3h
corresponding to its permutation
representation on the 6 molecules of PCl
5
(see Example 4.1.8). Their
values are easily calculated to be as follows.
D
3h
E C
3
3C
2
σ
h
S
3
σ
v
χ
1
3 0 1 1 2 1
χ
3
6 3 2 4 1 4
As in Example 5.2.23, the character of the module of displacements equals
the dual of χ
3
times χ
1
. Having only real values, χ
3
is selfdual, so that
we ﬁnd the following values for χ
dis
.
D
3h
E C
3
3C
2
σ
h
S
3
σ
v
χ
dis
18 0 2 4 2 4
This character can be decomposed into irreducible ones as follows:
χ
dis
= 2A
1
+A
2
+ 4E
+ 3A
2
+ 2E
The character χ
iso
is the sum of the characters of D
3h
on translations and
on rotations. The character of D
3h
on translations equals χ
3
= E
+ A
2
,
and the character χ
4
on rotations turns out to be as follows.
D
3h
E C
3
3C
2
σ
h
S
3
σ
v
χ
4
3 0 1 1 2 1
We ﬁnd that χ
4
= A
2
+E
. We conclude that
χ
vib
= 2A
1
+ 3E
+ 2A
2
+E
.
3. We have already computed χ
1
. Using Exercise 5.3.13 we ﬁnd the following
values for its symmetric square χ
2
.
D
3h
E C
3
3C
2
σ
h
S
3
σ
v
χ
2
6 0 2 2 2 2
We ﬁnd that χ
2
= 2A
1
+E
+E
, while χ
1
= E
+A
2
.
4. From the above, we conclude that only vibrations with characters E
or
A
2
are observable with IR spectroscopy; the number of such irreducible
characters in χ
vib
is 5, so that is the predicted number of peaks when
using this technique. Using Raman spectroscopy, only vibrations with
characters A
1
, E
and E
can be observed; this leads to 6 peaks.
80 CHAPTER 5. CHARACTER TABLES
PdCl (NH )
2 3 2
PdCl (NH )
2 3 2
N N
Cl Cl
N
Cl N
Cl
cis− trans−
Pd Pd
Figure 5.2: Two conﬁgurations of PdCl
2
(NH
3
)
2
The following exercises shows how spectroscopic techniques may or may not
be used to recognize molecules.
Exercise 5.4.2. The square planar coordination compound PdCl
2
(NH
3
)
2
can
be prepared with two diﬀerent conﬁgurations; see Figure 5.2. Is it possible
to discern the cis and the transconﬁgurations using information concerning
Pd −Cl stretch vibrations in IR spectroscopy?
Exercise 5.4.3. Is it possible to deduce the symmetry of Ni(CO)
4
(tetrahedron
or square plane) from the IR and/or Raman spectra of the COstretches?
Exercise 5.4.4. Let T be the octahedron whose 6 vertices are joined by strings.
Determine the decomposition of the 18dimensional space of vibrations of irre
ducibles of the group of rotational symmetries of T (of order 24). Interpret
the result. Which spaces fuse into irreducibles for the group of all (proper or
improper) symmetries of T (of order 48)?
5.5 Notes
Most of the content of this chapter can be found in Serre’s book [9]. The last
section is entirely based on Theo Beelen’s lectures in the course for which these
notes were prepared.
Chapter 6
Some compact Lie groups
and their representations
In the previous two chapters we got acquainted with the representation theory
of ﬁnite groups. In this chapter we will treat some of the representation theory
of compact Lie groups, without going into the details of deﬁning a Lie group and
showing, for example, the existence of an invariant measure. A good reference,
on which in fact most of this chapter is based, is [1]. Some general, less detailed
remarks can be found in [9].
6.1 Some examples of Lie groups
One can think of a Lie group G as a smooth surface embedded in some Euclidean
space. Moreover, G carries the structure of a group in such a way that the
multiplication map (g, h) → gh, GG → G and the inverse map g → g
−1
, G →
G are smooth. Examples, some of which we already saw, are:
1. GL(n, R),
2. SL(n, R),
3. O(n, R),
4. SO(n, R),
5. U(n) := ¦g ∈ M
n
(C) [ g
∗
g = I¦, where g
∗
stands for the conjugate
transpose of g, i.e., g
∗
= (g
ji
)
ij
if g = (g
ij
)
ij
.
6. SU(n) := ¦g ∈ U(n) [ det(g) = 1¦.
The latter two groups are called the unitary group and the special unitary group.
A more intrinsic deﬁnition is the following: let (., .) be a Hermitian inner product
on a ﬁnitedimensional complex vector space V . Then the corresponding unitary
81
82CHAPTER 6. SOME COMPACT LIE GROUPS ANDTHEIR REPRESENTATIONS
group is the group of all linear maps on V with (gv, gw) = (v, w) for all g ∈ G
and v, w ∈ V . Note that each of the real (complex) matrix groups above is
the zero set of some smooth map F from M
m
(R) (M
m
(C)) to some Euclidean
space R
d
such that the rank of the Jacobian of F at every point of the zero
set of F is d. If, in this situation, the zero set of F is a group with respect to
matrix multiplication—as it is in the above cases—then it is automatically a
Lie group. Such Lie groups are called linear because they are given by a linear
representation. In the present chapter, this class of Lie groups suﬃces for our
needs.
Exercise 6.1.1. How can GL(n, R) be constructed from M
n+1
(R) via the above
construction? What about (R, +)?
Exercise 6.1.2. Show that O(n, R)
∼
= SO(n, R) (Z/2Z). Here means that
the subgroup at the left is a normal subgroup and the subgroup at the right is
a complementary subgroup in the sense that their product is the whole group
and their intersection is the trivial group. Such a product is called semidirect.
If n is odd, then it is a direct product.
Exercise 6.1.3. Show that, in SU(2), any element is conjugate to a diagonal
matrix of the form
e(t) :=
e
it
0
0 e
−it
for certain t ∈ R.
The theory of Lie groups is a blend of group theory and topology (surfaces,
continuity, diﬀerentiability, etc.), as might be expected. For example, one can
ask whether a given Lie group is (pathwise) connected, or whether it is compact.
A typical lemma from the theory is the statement that the connected component
of the identity element is a normal subgroup.
Exercise 6.1.4. Show that O(n, R) is not connected.
Proposition 6.1.5. The Lie group SO(n, R) is connected.
Proof. We will connect an arbitrary matrix g = (a
ij
)
ij
in SO(n, R) to the iden
tity via a continuous and piecewise smooth path in SO(n, R). The ﬁrst step is
to connect it to a diagonal matrix; to this end, suppose ﬁrst that there exists a
position (i, j) such that i > j and a
ij
= 0. Then one can choose this position
such that, in addition, a
i
j
= 0 for all (i
, j
) with i
> j
and either j
< j or
j
= j and i
> i. Schematically, g has the following form:
g =
¸
¸
¸
¸
¸
T ∗ ∗ ∗ ∗
0 a
jj
∗ a
ji
∗
0 ∗ ∗ ∗ ∗
0 a
ij
∗ a
ii
∗
0 0 ∗ ∗ ∗
¸
,
6.1. SOME EXAMPLES OF LIE GROUPS 83
where T is a (j −1) (j −1) upper triangular matrix. Now let r(t), t ∈ R be
the real n n matrix given by
r(t)
i
j
=
cos(t), if i
= j
= j or i
= j
= i,
sin(t), if i
= i and j
= j,
−sin(t), if i
= j and j
= i,
1, if i = i
= j
= j, and
0 otherwise.
Then r(t) has the following block form:
r(t) =
¸
¸
¸
¸
¸
I 0 0 0 0
0 cos(t) 0 −sin(t) 0
0 0 I 0 0
0 sin(t) 0 cos(t) 0
0 0 0 0 I
¸
;
note that r(t) ∈ SO(n, R) for all t ∈ R. Now we have
c(t) := r(t)g =
¸
¸
¸
¸
¸
T ∗ ∗ ∗ ∗
0 a
jj
cos(t) −a
ij
sin(t) ∗ a
ji
cos(t) −a
ii
sin(t) ∗
0 ∗ ∗ ∗ ∗
0 a
jj
sin(t) +a
ij
cos(t) ∗ a
ji
sin(t) +a
ii
cos(t) ∗
0 0 ∗ ∗ ∗
¸
Now R(sin(t), cos(t)) runs through all 1dimensional subspaces of R
2
as t runs
from 0 to π, so that there exists a t
0
∈ [0, π) for which (sin(t
0
), cos(t
0
)) is
perpendicular to (a
jj
, a
ij
) in the standard inner product; we ﬁnd that c(t
0
)
ij
=
0, in addition to c(t
0
)
i
j
= 0 for all (i
, j
) with i
> j
and either j
< j or j
= j
and i
> i. Moreover, the path c(t), t ∈ [0, t
0
] lies in SO(n, R) and connects g
to c(t
0
).
Repeating the above construction, we ﬁnd a path connecting g to an upper
triangular matrix h. Being an element of O(n, R), h is in fact diagonal and
has only entries ±1 on the diagonal. Moreover, as det(h) = 1, the number of
−1s is even, and they can be partitioned into pairs. Each such pair can be
smoothly transformed to 1s by multiplication from the left with an appropriate
r(t), t ∈ [0, π] as above.
This proposition implies that SO(n, R) is the connected component of the
identity in O(n, R). The following exercise will be used later on, for the repre
sentation theory of SO(3, R).
Exercise 6.1.6. Show that SU(2) is connected. Hint: ﬁnd a continuous path
from any given element of SU(2) to some e(t) as in Exercise 6.1.3.
84CHAPTER 6. SOME COMPACT LIE GROUPS ANDTHEIR REPRESENTATIONS
6.2 Representation theory of compact Lie groups
Compact Lie groups (i.e., with the property that their embedding in Euclidean
space is closed and bounded) are particularly nice, because much of the repre
sentation theory of ﬁnite groups works just as well for these. The main tool is
the socalled Haar measure, which enables us to take ‘the average over a group’
just as we did in ﬁnite group theory. Let us state the existence and uniqueness
of this measure in a theorem.
Theorem 6.2.1. Let G be a compact real Lie group, and denote by C(G) the
linear space of all continuous functions G → R on G. There is a unique map
f →
G
f(g)dg, C(G) →R which is
1. linear, i.e., for all e, f ∈ C(G) and α ∈ R, we have
G
(e+f)(g)dg =
G
f(g)dg+
G
e(g)dg and
G
(αf)(g)dg = α
G
f(g)dg,
2. monotonous, i.e., if e(g) ≤ f(g) for all g ∈ G, then also
G
e(g)dg ≤
G
f(g)dg,
3. leftinvariant, i.e., for the action of G induced on C(G) by the left multi
plication, we have for all h ∈ G and f ∈ C(G) that
G
h f(g)dg =
G
f(g)dg,
where h f is the function g → f(h
−1
g) (h ∈ G).
4. and normalized, i.e.,
G
1dg = 1.
This linear map is called the invariant (Haar)integral.
Exercise 6.2.2. A ﬁnite group is a compact Lie group. What is the invariant
integral?
In order to ﬁnd an invariant Haar integral on the compact Lie group T =
¦e(t) [ t ∈ R¦, where e(t) is deﬁned as in Exercise 6.1.3, we search for invariant
measures on e(t), that is, diﬀerential forms invariant under left multiplication.
In general, given a representation r : G → GL(V ), we can make invariant
measures on G by taking certain entries of the matrix r(g)
−1
dr(g). For the
group T and the natural representation, this is e(t)
−1
de(t). The 1, 1 entry of
this matrix is e
−it
de
it
= idt. Since
2π
0
idt = 2πi, the invariant Haarintegral,
viewed as a map assigning a real number to a function f : R/2πZ →R, is
1
2π
2π
0
f(t)dt.
6.2. REPRESENTATION THEORY OF COMPACT LIE GROUPS 85
A similar but more laborious computation yields that the invariant Haar mea
sure for SU
2
(C) is
1
2π
2
sin
2
θ sin ψdθdψdφ,
where
cos θ −i sin θ cos ψ −sin θ sin ψe
iφ
sin θ sin ψe
−iφ
cos θ +i sin θ cos ψ
is a general element of SU
2
(C).
The result of the following exercise will be used in proving that certain
irreducible representations of SU
2
(C) exhaust all irreducible representations.
Exercise 6.2.3. Show that the map sending a class function f : SU(2) → C
to (f ◦ e) : R → C, where e is the oneparameter subgroup of SU(2) deﬁned
in Exercise 6.1.3, is a linear bijection between the space H of continuous class
functions on SU(2) and the linear space of all even, 2πperiodic continuous
functions R → C. Moreover, the norm corresponding to the invariant inner
product (6.1) below corresponds, under this linear bijection, to the normalized
L
2
inner product on the space of continuous even 2πperiodic functions on R.
One could deﬁne representations of compact Lie groups in Hilbert spaces,
but here we shall restrict our attention to ﬁnitedimensional representations.
Also, we shall only consider complex representations.
Deﬁnition 6.2.4. A ﬁnitedimensional complex representation of the Lie group
G is a smooth homomorphism G → GL(V ), where V is a ﬁnitedimensional
complex vector space.
In fact, one can prove that continuous homomorphisms from a Lie group G
are automatically smooth.
For ﬁnitedimensional complex representations of compact Lie groups, al
most all theorems that we proved in the preceding chapters for ﬁnite groups
hold, be it that at some places a sum must be replaced by an integral. To give
an idea how things work, one can try the following exercise.
Exercise 6.2.5. Let r : G → GL(V ) be a ﬁnitedimensional representation of
the compact Lie group G, and let W ⊂ V be an invariant subspace. Show that
W has an invariant complement in V .
We will study the representations of the groups SU(2) and SO(3, R), so it is
good to verify that these are indeed compact Lie groups.
Exercise 6.2.6. In this exercise we prove that U(n), SU(n), O(n, R) and
SO(n, R) are compact Lie groups. To this end, endow the space M
n
(C) with
the Hermitian inner product given by
((a
ij
)
ij
, (b
ij
)
ij
) :=
¸
i,j
a
ij
b
ij
.
1. Show that (a, b) = tr(ab
∗
) for all a, b ∈ M
n
(C).
86CHAPTER 6. SOME COMPACT LIE GROUPS ANDTHEIR REPRESENTATIONS
2. Show that U(n) is bounded with respect to the norm corresponding to
this inner product.
3. Show that U(n) is the zero set of a smooth map F : M
n
(C) →R
n
2
, whose
Jacobian has rank n
2
at each element of U(n).
Thus, U(n) is a Lie group by the construction of the beginning of this chapter,
and compact as it is a closed and bounded set in M
n
(C). Find similar arguments
for the other groups above. Show, on the contrary, that the group O(n, C) of
complex matrices g satisfying g
T
g = Iis not compact for n ≥ 2.
A ﬁnal remark concerns the compact analogon of the inner product on the
space of complexvalued functions on a ﬁnite group: in the case of a compact
Lie group, one should consider the space of Hilbert space of continuous class
functions with Hermitian inner product deﬁned by
(f[h) :=
G
f(g)h(g)dg. (6.1)
Here, the irreducible characters form an orthonormal Hilbert basis in the space
of continuous class functions.
6.3 The 2dimensional unitary group
This section is based on section II.5 of [1].
Let r : SU(2) → GL(V ) be the standard representation of SU(2) on V = C
2
,
the action being given by matrixcolumn multiplication (the elements of V are
understood to be columns). As we have seen before, this action induces linear
actions on P(V )
d
, the space of homogeneous polynomials of degree d deﬁned on
V , by letting
(gP)(ξ) = P(g
−1
ξ) for ξ ∈ V, P ∈ P(V )
d
, g ∈ SU(2).
Let r
d
: SU(2) → GL(P(V )
d
) be the corresponding representation and χ
d
its
character.
We are ready to state the main theorem of this section.
Theorem 6.3.1. The representations r
d
are irreducible for all d = 1, 2, . . ..
Proof of theorem 6.3.1. We will use Exercise 4.1.30. Let A ∈ End(P(V )
d
) be a
linear map that commutes with all r
d
(g) for g ∈ SU(2).
The space P(V )
d
is spanned by the polynomials P
k
: ξ → ξ
k
1
ξ
d−k
2
for k =
0, . . . , d, and is therefore (d + 1)dimensional.
Consider the diagonal subgroup of SU(2). It consists of all matrices of the
form
g
a
:=
a 0
0 a
−1
6.3. THE 2DIMENSIONAL UNITARY GROUP 87
Note that the element a ∈ C must necessarily have norm 1. We compute
(g
a
P
k
)(ξ) = P
k
(g 1
a
ξ)
= (ξ
1
/a)
k
(aξ
2
)
d−k
= a
d−2k
P
k
(ξ).
Hence we conclude that P
k
is an eigenfunction of r
d
(g
a
) with eigenvalue a
d−2k
.
Choose a
0
such that none of these eigenvalues coincide. If A commutes with
r
d
(g
a0
), then the eigenspaces of the latter map are invariant under A. Hence,
as these eigenspaces are all onedimensional, there are complex numbers c
k
for
k = 0, . . . , d such that
AP
k
= c
k
P
k
.
It remains to show that all c
k
are equal. For this purpose, consider the real
rotations s
t
deﬁned by
s
t
:=
cos t −sin t
sin t cos t
and compute
(s
t
P
n
)(ξ) = P
n
(s
−t
ξ)
= (ξ
1
cos t +ξ
2
sin t)
n
=
¸
k
n
k
cos
k
t sin
n−k
t P
k
(ξ)
Applying A to the result, we get
A(s
t
P
n
) =
¸
k
n
k
cos
k
t sin
n−k
t c
k
P
k
.
First applying A and then r
d
(s
t
), and using the fact that the two commute, we
ﬁnd that also
A(s
t
P
n
) =
¸
k
n
k
cos
k
t sin
n−k
t c
n
P
k
.
Taking t such that cos t sin t = 0, it follows from the fact that the P
k
are linearly
independent that c
n
= c
k
for all k = 1, . . . , n, and hence that A = c
n
I.
We wish to prove that these are all irreducible characters. To this end we
shall use some Fourier theory, and the result of Exercise 6.2.3.
Theorem 6.3.2. The representations r
d
are, up to equivalence, all irreducible
representations of the group SU(2).
Proof. In view of Exercise 6.2.3, we can identify H with the space of even 2π
periodic complexvalued functions on R. The matrix of r
d
(e(t)) with respect to
the basis P
0
, . . . , P
d
is diagonal, and its trace is
κ
d
(t) =
d
¸
k=0
e
i(d−2k)t
.
88CHAPTER 6. SOME COMPACT LIE GROUPS ANDTHEIR REPRESENTATIONS
It is readily veriﬁed that κ
0
(t) = 1 (this corresponds to the trivial representation
of the group), and that κ
1
(t) = 2 cos t. Indeed, one has κ
d+2
(t) = κ
d
(t) +
2 cos((d + 2)t). Hence, the functions κ
d
for d = 0, 1, 2, . . . generate the same
space as the functions cos dt. From Fourier theory we know that this subspace
is dense in the Hilbert space of continuous even 2πperiodic functions. This
implies, using the isomorphism from Exercise 6.2.3, that the characters χ
d
span
a dense subspace in the Hilbert space of continuous class functions on SU(2).
Any irreducible character diﬀerent from all r
d
would have inner product 0 with
all r
d
, and hence be zero. This concludes the proof that the r
d
exhaust the set
of irreducible characters of SU(2).
6.4 The 3dimensional orthogonal group
We will obtain the irreducible representations of SO(3, R) from those of SU(2)
by means of the following lemma, which appeared earlier as Exercise 5.3.11.
Lemma 6.4.1. Let G and H be groups, φ : G → H a surjective homomorphism.
Then the (irreducible) representations of H are in bijective correspondence with
the (irreducible) representations r of G with ker r ⊇ ker φ.
Thus, we wish to ﬁnd a surjective homomorphism of SU(2) onto SO(3) and
to determine its kernel. Recall the homomorphism φ : SL(2, C) → L of Example
2.2.9. Here, L is the group of Lorentznorm preserving linear maps from M to
itself, where M is the Rlinear span of the following four complex 22matrices.
e
0
= I, e
1
=
0 1
1 0
, e
2
=
0 −i
i 0
, e
3
=
1 0
0 −1
.
It can be checked that in this way
det(x
0
e
0
+x
1
e
1
+x
2
e
2
+x
3
e
3
) = x
2
0
−x
2
1
−x
2
2
−x
2
3
.
So L is the group
¦B ∈ GL(M) [ det(Bx) = det(x), for all x ∈ M¦.
The homomorphism φ was deﬁned by
φ(A)(x) = AxA
∗
, for all x ∈ M, A ∈ SL(2, C).
From now on, φ will denote the restriction of this homomorphism to SU(2). We
claim that the image is the group
G := ¦B ∈ L [ Be
0
= e
0
and det B = 1¦,
which is, by its action on the invariant 3space e
⊥
0
, easily seen to be isomorphic
to SO(3, R).
First of all φ(SU(2)) ⊆ G. Indeed, if A ∈ SU(2), then A
∗
= A
−1
. Therefore,
φ(A)(e
0
) = e
0
, since e
o
is just the 2 2 identity matrix. Also, the function
6.4. THE 3DIMENSIONAL ORTHOGONAL GROUP 89
det ◦φ is a continuous function SU(2) → R, which can apparently take only
values in ¦−1, 1¦. As SU(2) is connected, and det(φ(I)) = 1, the determinant
of any element in the image φ(SU(2)) is 1.
Now let us prove that G ⊆ φ(SU(2)). We follow the approach given in [4].
First we prove the following two lemmas.
Lemma 6.4.2. Let F be a subgroup of SO(3, R) with the following two proper
ties.
1. F acts transitively on the unit sphere.
2. There is an axis such that F contains all rotations around that axis.
Then F = SO(3, R).
Exercise 6.4.3. Prove this lemma.
Lemma 6.4.4. For every nonzero x ∈ e
⊥
0
there is an A ∈ SU(2) such that
φ(A)x = ce
3
for some c ∈ R and c > 0. Here e
⊥
0
denotes the orthogonal
complement of e
0
with respect to the Lorentz inner product, i.e.,
e
⊥
0
= ¦x
1
e
1
+x
2
e
2
+x
3
e
3
[ x
i
∈ R¦ ⊆ M.
Proof. Any such x is in fact a Hermitian matrix with trace 0. As such it has
two real eigenvalues c, −c, where c > 0, (note that x is nonzero). Let two
(necessarily orthogonal) eigencolumns of norm 1 corresponding to c and −c
be the ﬁrst and second column of a matrix B, respectively. Then the usual
coordinate transformation rule gives
x = B
−1
(cH)B.
We have that B ∈ U(2). Now take A = dB where d
2
= 1/(det B); then A is
still unitary, since [d[ = 1. Furthermore, we have
AxA
−1
= (dB)x(dB)
−1
= dd
−1
BxB
−1
= cH,
and det A = d
2
det B = 1.
Now let us continue with the proof that F := φ(SU(2)) coincides with G.
The group F acts on the unit sphere S in e
⊥
0
deﬁned by x
2
1
+ x
2
2
+ x
2
3
, and by
Lemma 6.4.4 this action is transitive, since any vector in the sphere is mapped
to e
3
by some group element. Hence, to apply Lemma 6.4.2 to F, we need only
check that F contains all rotations around some axis.
We leave this as an exercise to the reader:
Exercise 6.4.5. Show that φ(e(t)) is the rotation around 'e
3
` through an angle
of 2t. Here e : R → SU(2) is the oneparameter group deﬁned earlier.
We conclude that φ : SU(2) → G is a surjective homomorphism. Its kernel
contains ¦I, −I¦, as is easily checked, and the following exercise shows that this
is the whole kernel.
90CHAPTER 6. SOME COMPACT LIE GROUPS ANDTHEIR REPRESENTATIONS
Exercise 6.4.6. Show that ker φ = ¦I, −I¦. Hint: the elements of this kernel
must commute with all of M, hence in particular with e
1
, e
2
, and e
3
.
Finally we can apply lemma 6.4.1: the irreducible representations of SO(3)
correspond to those of SU(2) for which −I is mapped to the identity transfor
mation. More precisely, we wish to know for which d we have r
d
(−I) = I. For
this to be the case, it is necessary and suﬃcient that
((−I)P)(ξ) = P(ξ) for all ξ ∈ V, P ∈ P(V )
d
.
This holds if and only if d is even. Hence the irreducible representations are
W
d
:= P(V )
2d
.
Note that the dimension of W
d
is 2d + 1.
In the following example, we will see another interpretation of the irreducible
representations of SO(3, R).
Example 6.4.7. Consider P(V )
d
the space of homogeneous polynomials of
degree d deﬁned on the vector space V which will be of dimension three here. As
we have seen before, the group GL(3, R) acts on this space P(V )
d
by (Ap)(x) =
p(A
−1
x), for all A ∈ GL(3, R), p ∈ P(V )
d
and x ∈ R
3
. So does the subgroup
SO(3).
Take the subspace U of P(V )
2
generated by the polynomial x
2
1
+ x
2
2
+ x
2
3
.
This is an invariant subspace for the action of SO(3). Hence P(V )
2
is not
irreducible. Actually the spaces P(V )
d
are not irreducible for all d ≥ 2.
To ﬁnd SO(3) invariant subspaces of P(V )
d
, we need to deﬁne the Laplace
operator ∆ on R
3
: ∆ =
∂
2
∂x
2
1
+
∂
2
∂x
2
2
+
∂
2
∂x
2
3
. The space of harmonic polynomials
h
d
is the vector space:
h
d
= ¦p ∈ P(V )
d
[ ∆p = 0¦.
As was noticed in Example 4.1.17, the dimension of P(V )
d
is dimP(V )
d
=
d+2
d
.
We now want to prove that the dimension of h
d
is 2d + 1. A homogeneous
polynomial p of degree d can be written as a sum of homogeneous polynomials
p
k
in x
2
and x
3
with power of x
1
’s as coeﬃcients:
p(x
1
, x
2
, x
3
) = p
0
(x
2
, x
3
) +x
1
p
1
(x
2
, x
3
) +
x
2
1
2
p
2
(x
2
, x
3
) + +
x
d
1
d!
p
d
(x
2
, x
3
)
=
d
¸
k=0
x
k
1
k!
p
k
(x
2
, x
3
)
where the p
k
’s have degree d −k. If one applies the Laplace operator to p (you
see why we took factorials):
∆(p) =
d−2
¸
k=0
x
k
1
k!
p
k+2
(x
2
, x
3
) +
d
¸
k=0
x
k
1
k!
∂
2
∂x
2
2
+
∂
2
∂x
2
3
p
k
(x
2
, x
3
)
6.4. THE 3DIMENSIONAL ORTHOGONAL GROUP 91
Hence an element p ∈ P(V )
d
is in h
d
if and only if for all k = 0, . . . , d −2:
p
k+2
(x
2
, x
3
) = −
∂
2
∂x
2
2
+
∂
2
∂x
2
3
p
k
(x
2
, x
3
)
This formula implies that p in h
d
depends only on the ﬁrst two homogeneous
polynomials p
0
and p
1
. Therefore we conclude that the dimension of h
d
is
precisely the number of homogeneous polynomials in two variables of degree d
and d −1. That is, dimh
d
= 2d + 1.
Of course, the degree of h
d
reminds us of the irreducible representations W
d
of SO(3) we met before. Indeed, we have the following.
Theorem 6.4.8. The space h
d
of harmonic polynomials of degree d is an irre
ducible representation of SO(3).
To see that what we just stated is true, we should ﬁrst check that h
d
is a
representation of SO(3). We state here the following fact without proof.
Fact: The action of the Laplace operator on P(V )
d
commutes with the
action of SO(3).
From this we conclude that h
d
is a representation of SO(3).
To prove that h
d
is irreducible, we claim that h
d
= W
d
. Set p a harmonic
polynomial in h
d
, p(x
1
, x
2
, x
3
) = (x
2
+ ix
3
)
d
. Let R(t) be the matrix of SO(3)
deﬁned by
R(t) =
¸
1 0 0
0 cos t −sin t
0 sin t cos t
¸
Apply R(t) to p. That gives us R(t)p = exp(−idt)p. So p generates an
invariant subspace W of h
d
under the action of R(t).
Any element in SO(3) is conjugate to R(t). So in order to ﬁnd the character
of W
d
on R(t) it suﬃces to compute it on R(t). That is just the value of
the character χ
2d
of r
d
(the irreducible representation of SU(2)) at e
t/2
, i.e.,
¸
2d
k=0
e
i(d−k)t
.
By the decomposition of h
d
in irreducible W
s
’s, we know that its character
is a linear combination of exp(ist). So what we saw above about the action of
R(t) on W leads us to conclude that h
d
= W
d
.
92CHAPTER 6. SOME COMPACT LIE GROUPS ANDTHEIR REPRESENTATIONS
Bibliography
[1] T. Br¨ocker and T. tom Dieck. Representations of Compact Lie Groups.
Springer Verlag, New York, 1995.
[2] F. Albert Cotton. Chemical applications of group theory. Wiley
Interscience, Chichester, third edition, 1990.
[3] J.D.Dixon and B. Mortimer Permutation Groups. Springer Verlag, New
York, 1991.
[4] E.Vinberg. Linear Representations of Groups. Birkh¨auser, Basel, 1989.
[5] William Fulton. Young tableaux, volume 35 of London Mathematical Society
Student Texts. Cambridge university press, 1997.
[6] Juergen Hinze, editor. The permutation group in physics and chemistry,
volume 12 of Lecture notes in chemistry. Springer Verlag, Berlin, 1979.
[7] H.H. Jaﬀe and Milton Orchin. Symmetry in chemistry. Wiley, London,
1965.
[8] A. Nussbaum. Applied group theory; for chemists, physicists and engineers.
PrenticeHall electrical engineering series. PrenticeHall, Englewood Cliﬀs,
1971.
[9] J.P. Serre. Linear representations of ﬁnite groups. Springer Verlag, New
York, 1977.
[10] S. Sternberg. Group theory and physics. Cambridge University Press, Cam
bridge, 1994.
[11] Hermann Weyl. Symmetry. Princeton science library. Princeton University
Press, Princeton, 1989.
93
2
Contents
1 Introduction 1.1 Symmetry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Basic notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Permutation Groups 2.1 Cosets and Lagrange . . 2.2 Quotient groups and the 2.3 Loyd’s puzzle . . . . . . 2.4 Action of groups on sets 5 5 12 17 17 18 22 28 33 33 35 40 42
. . . . . . . . . homomorphism . . . . . . . . . . . . . . . . . .
. . . . . theorem . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
3 Symmetry Groups in Euclidean Space 3.1 Motivation . . . . . . . . . . . . . . . 3.2 Isometries of nspace . . . . . . . . . . 3.3 The Finite Subgroups of O(2, R) . . . 3.4 The Finite Subgroups of O(3, R) . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
4 Representation Theory 51 4.1 Linear representations of groups . . . . . . . . . . . . . . . . . . 51 4.2 Decomposing Displacements . . . . . . . . . . . . . . . . . . . . . 59 5 Character Tables 5.1 Characters . . . . . . . . . . . . . . . . 5.2 Orthogonality of irreducible characters 5.3 Character tables . . . . . . . . . . . . 5.4 Application: symmetry of vibrations . 5.5 Notes . . . . . . . . . . . . . . . . . . 63 63 64 72 77 80 81 81 84 86 88
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
6 Some compact Lie groups and their representations 6.1 Some examples of Lie groups . . . . . . . . . . . . . . 6.2 Representation theory of compact Lie groups . . . . . 6.3 The 2dimensional unitary group . . . . . . . . . . . . 6.4 The 3dimensional orthogonal group . . . . . . . . . .
. . . .
. . . .
. . . .
. . . .
. . . .
. . . .
3
4 CONTENTS .
we think of an object’s symmetry as transformations of space that map the object ‘into itself’. consider once more the chair in Figure 1. But rather than changing viewpoint ourselves. the chair in Figure 1.1.1 Symmetry Group theory is an abstraction of symmetry Symmetry is the notion that an object of study may look the same from diﬀerent points of view. For instance. Consider the transformation r of threedimensional space that maps each point p to the point p constructed as follows: Figure 1. 5 .Chapter 1 Introduction 1.1: Bilateral and rotational symmetry. In this picture we see a plane V cutting the chair into two parts.1 looks the same as its reﬂection in a mirror that would be placed in front of it. and our view on the wheel depicted next to the chair doesn’t change if we rotate our point of view over π/6 around the shaft. What do we mean when we say that an object is symmetric? To answer this question. From: [8].
This type of symmetry is called rotational symmetry. the requirement that the object be symmetric may reduce drawings to half the chair. The map r is called the reﬂection in V . but of a diﬀerent type: the rotation s around the wheel’s axis m over 2π/6 moves all points occupied by the wheel to other points occupied by it. 4. For this reason we call s a symmetry of the wheel. V ) = d(p. Next. V ) and p = p . • when studying functions on the plane that respect rotational symmetry around a ‘centre’ p in the plane. or say that r is a symmetry (transformation) of the chair. Denoting the distance of a point x to V by d(x. 3. Symmetry considerations lead to eﬃciency in the study of symmetric objects. 1. Symmetry registers regularity. Transforming any point in space to its reﬂection in the mirror W and next rotating it over an angle of π around the axis V ∩W . but then on a transformed chair (transformed by means of r). For example. and thus records beauty. gives the same result as the reﬂection r in V . INTRODUCTION let l be the line through p and perpendicular to V . but because of the choice of V it equals the old space occupied by the chair. For this reason. was ‘neutralized’ to the original look. we can reduce the number of arguments from the (usual) two for an arbitrary function on the plane to one. But it does more than that. Usually. p is the unique point on l with d(p . The wheel also has reﬂection symmetry: the reﬂection t in the plane W perpendicular to m and cutting the wheel into two thinner wheels of the same size also maps the wheel to itself. It also has a symmetry. A symmetry of an object or ﬁgure in space is a transformation of space that maps the object to itself. for k = 1. For instance. V ). but this subtlety should be remembered. . respectively. The result is a new set of points. one could argue that the chair is not symmetric under reﬂection in V . if the left legs of the chair are made of wood and the right ones of iron. Symmetry then is the phenomenon that the image chair is indistinguishable from the original chair. and not the materials involved. • when designing a chair. consider the wheel depicted in Figure 1. This is an illustration of what we observed about symmetry: the change of view on the chair from the original chair to its mirror image. Now apply r to all points in the space occupied by the chair. we call the chair invariant under r. we shall speak about symmetries of sets of points. We only considered the space occupied by the chair and the wheel.1. the distance to p. Note that the rotations around m over the angles 2kπ/6 are symmetries of the wheel as well. 5.6 CHAPTER 1. observing symmetry is useful. Symmetry conditions A few more observations can be made regarding the previous cases. 2.
then so is their composition. Groups The transformations under which a given object is invariant. if both translations are symmetries of some object. it is rotation about m over −2π/6. There is one trivial transformation which is a symmetry of any object. In the case of s. let us consider some more examples. the hexagon has at least . In the case of s. as we shall see. ‘group’ is a more general concept. i. SYMMETRY 7 O Figure 1. rotation r around 0 over 2π/6 is one of them. Group theory was inspired by these types of group. sending each point to itself. Two transformations can be composed to obtain another transformation. and s2 for ss. is a twodimensional object. t ◦ s (ﬁrst apply s. we see that s ◦ s (s applied twice) is a symmetry of the wheel as well. namely the identity. see below). Also.1. form a group. 4. In the case of r. Usually. To get a feeling for groups.2: The Hexagon 2. There are lots of transformations of the plane leaving it invariant. Planar groups The hexagon. 10π/6. then t) is a symmetry of the wheel. It is usually denoted by e (but also by 0 or 1. we will leave out the composition sign ◦ and write ts instead of t ◦ s. 3. or. For example.e.1. as is reﬂection s in the vertical line l through the barycentre. they have an inverse transformation.2. as depicted in Figure 1. lying in the plane. it is r itself. Therefore.. which amounts to the same. it is rotation over the angle 2 · (2π/6). However. Symmetry transformations are bijective.
8
CHAPTER 1. INTRODUCTION
Figure 1.3: A frieze pattern. From: [11].
the following symmetries: e, r, r2 , r3 , r4 , r5 , s, sr, sr2 , sr3 , sr4 , sr5 . In fact, these are all symmetries of the hexagon. Thus, each is a (multiple) product of r and s. These two elements are said to generate the group of symmetries. Exercise 1.1.1. Verify that the elements listed are indeed distinct and that r6 = s2 = e, and srs = r5 . Exercise 1.1.2. Show that sr is a reﬂection. In which line? Figure 1.3 shows the palmette motif, which is very frequently used as a decoration on the upper border of wallpaper. Imagine this pattern to be repeated inﬁnitely to the left and right. Then the translation t over the vector pointing from the top of one palmette to the top of the second palmette to its right is a symmetry of the pattern, and so is the reﬂection r in the vertical line l through the heart of a given palmette. With these two, all symmetries of the frieze pattern can be listed. They are: e, t, t2 , t3 , . . . and r, rt, rt2 , rt3 , . . . . We ﬁnd that the symmetry group of the pattern is inﬁnite. However, when ignoring the translational part of the symmetries, we are left with essentially only two symmetries: e and r. Exercise 1.1.3. Find frieze patterns that have essentially diﬀerent symmetry groups from the one of Figure 1.3. A pattern with a more complicated symmetry group is the wallpaper pattern of Figure 1.4, which Escher designed for ‘Ontmoeting’ (Dutch for ‘Encounter’). It should be thought of as inﬁnitely stretched out vertically, as well as horizontally. The translations s and t over the vectors a (vertical from one nose
1.1. SYMMETRY
9
Figure 1.4: Wallpaper pattern for Ontmoeting. From: M.C.Escher, 1944.
to the next one straight above it) and b (horizontal, from one nose to the next one to its right) leave the pattern invariant, but this is not what makes the picture so special. It has another symmetry, namely a glidereﬂection g, which 1 is described as: ﬁrst apply translation over 2 a, and then reﬂect in the vertical line l equidistant to a left oriented and a neighbouring right oriented nose. Note that g 2 = s. Exercise 1.1.4. List all symmetries of the pattern in Figure 1.4.
Space groups
Having seen symmetries of some 2dimensional ﬁgures, we go over to 3dimensional ones. The cube in Figure 1.5 has lots of symmetries. Its symmetry group is generated by the rotations r3 , r4 , r4 whose axes are the coordinate axes, and the reﬂection t through the horizontal plane dividing the cube in two equal parts. Exercise 1.1.5. What is the size of the symmetry group of the cube? Closely related is the methane molecule, depicted in Figure 1.6. Not all of the symmetries of the cube are symmetries of the molecule; in fact, only those that map the hydrogen atoms to other such atoms are; these atoms form a regular tetrahedron. Exercise 1.1.6. Which of the symmetries of the cube are symmetries of the methane molecule? Finally consider the cubic grid in Figure 1.7. Considering the black balls to be natrium atoms, and the white ones to be chlorine atoms, it can be seen as the structure of a saltcrystal. It has translational symmetry in three perpendicular
10
CHAPTER 1. INTRODUCTION
Figure 1.5: The cube.
Figure 1.6: The methane molecule. From: [7].
This exercise concerns symmetries of onedimensional ﬁgures. s0 }.7: Structure of a crystal directions along vectors of the same size. 1. y ∈ R is x − y.7. y ∈ R. SYMMETRY 11 Figure 1. and therefore we may say that the symmetry group of salt crystals is generated by that of the regular octahedron and three translations. Considere therefore the real line R with the Euclidean metric. Show that any isometry of R is either of the form sa : x → 2a−x (reﬂection in a) or of the form ta : x → x + a (translation over a). Such a symmetry must permute the 6 chlorine atoms closest to the given natrium atom. 3. These atoms form a regular octahedron. To understand its further symmetry. Show that if such a group G contains s0 . Prove that {e} and {e. .1. then G = {ta  a ∈ Z} ∪ {t a  a ∈ Z}. we restrict our attention to groups G whose translational part {a ∈ R  ta ∈ G} equals Z. 2.1. 2 4. consider all those symmetries of the crystal that leave a given natrium atom ﬁxed. A map g from R to itself is called an isometry if g(x) − g(y) = x − y for all x. as depicted in Figure 1. where e denotes the identity map on R. in which the distance between x. Conclude that there are only two ‘essentially diﬀerent’ groups describing the symmetries of discrete subsets of R having translational symmetry. Exercise 1.1.8. To describe symmetries of discrete subsets of R. are the only ﬁnite groups of isometries of R.
h. we are ready for some abstract deﬁnitions that should reﬂect the experiences we had in those examples.1.1. it may be inﬁnite. h) of elements of G to another element g · h of G. A group is a set G. Each element of G has an inverse. h ∈ G).7.1. i.12 CHAPTER 1.e. k ∈ G we have g · (h · k) = (g · h) · k. which maps an ordered pair (g. then G is called an Abelian group.e.2. 1] in a manner similar to that of Exercise 1. Exercise 1. .. i. an element e that satisﬁes e · g = g · e = g for all g ∈ G. G contains an identity element.2 Basic notions Formal deﬁnition of a group Having seen some examples of groups.1. 3. Compare the result with your answer to Exercise 1.3. and often denoted by G. 1. Analyse the symmetry groups of discrete subsets of the strip R × [−1.. albeit from the narrow point of view of symmetry groups of ﬁgures in the plane or in threedimensional space. 1.8: The octahedron. The cardinality of G is called the order of the group. This element is denoted by g −1 . 2. Deﬁnition 1. together with an operation ·. INTRODUCTION Figure 1.. satisfying the following axioms. If the operation is not only associative but commutative as well (meaning g · h = h · g for all g.8. The operation is associative. for each g ∈ G there is an h ∈ G such that g · h = h · g = e. i. for all g.e.
if the operation is written as +. 1)}. All sets of transformations found in §1. Deﬁnition 1. 2. The set of all linear transformations of V . The group is often denoted by Sym(X).3. They do not form a group with respect to ordinary multiplication. Also Q and Z form additive groups. (3. and (g n )−1 = (g −1 )n . However. in each case.2. together with composition of maps. Composition is again the group operation. (1.. ·). Lemma 1. 5. as 0 does not have an inverse. often denoted by GL(V ).2. 2. 3. then the groups Sym(X) and Sym(Y ) are essentially the same. the notation g −1 for the inverse of g seems to indicate the uniqueness of that inverse. Some of them are Abelian. Its element are also called permutations.. Let us make notation easier. · g. prove this. .. (1. 2. then Sym(X) = n!. Let X be a set. we have Sym(X) = {e.2. and some aren’t. the composition ◦ serves as operation ·. n Our ﬁrst lemma concerns the inverse of a product. Again Q \ {0} is a multiplicative group. .1 form groups.2. The real numbers R form a group with respect to addition +. 3). 2). The bijections of X form a group. h in a group G we have (gh)−1 = h−1 g −1 . Prove that a group G cannot have more than one identity. For the latter we also write g −n . For two elements g. say X = n.2. like those of the cube. the representative Sym({1. (1. Let us give some examples of groups. If X is ﬁnite. is a group. Also. Usually. For X = {1. This notation is explained in Section 2.3. For ﬁnite sets of cardinality n. Z cannot be turned into a multiplicative group. the unit element being 1. By ﬁxing a basis of V . For g in a group G we write g n for the product g · g · . Example 1. Let V be a vector space. The group inverse coincides with the ‘functional’ inverse.4. the unit element being 0. Leaving out 0 we do obtain an Abelian group (R \ {0}. 3). like the the chair.. its element can be written as invertible matrices.2.1. 3). 1. 2.. it is never left out.5. However. 3}. BASIC NOTIONS 13 Exercise 1. 4. (2. the only invertible elements being {±1}. we leave out the operation · if it does not cause confusion. n}) of this class of groups is denoted by Sym(n) or Sn . If two sets X and Y have the same cardinality.
. namely the symmetry group of the tetrahedron. Lemma 1. ﬁrst take oﬀ the coat and the shirt comes last.2. H is nonempty. H is closed under taking inverses. does not work here: h could stand for g −1 h as well as for hg −1 . Prove that a subgroup of a group contains the group’s identity. . With the methane molecule in Figure 1. Deﬁnition 1. For example. it is not immaterial in which order you perform the operations.2. The intersection of a family of subgroups of a group G is a subgroup of G. Prove that for a subset H of a ﬁnite group G to be a subgroup. Show that the notation for fractions which is usual for integers.’ Exercise 1. Deﬁnition 1.7. 1.2.14 CHAPTER 1. Instead of {g1 . gn are called generators of this group. . gn } we also write g1 . and the latter g two may diﬀer. gn . We found an example of a subgroup. for a subset S of G the intersection of all groups containing S is a subgroup.10.8. for all h ∈ H.11.9.. and when in dressing you start with the shirt and end up with the coat.2. h ∈ H we have gh ∈ H. Generation The following lemma is needed to formalize the concept of a subgroup generated by some elements.9 is generated by a rotation around its center over 2π/3. as Weyl points out in [11]: ‘When you dress. . H is closed under the operation. . Prove that for a subset H of G to be a subgroup of G it is necessary and suﬃcient that H be nonempty and that gh−1 ∈ H for all g. then in undressing you observe the opposite order. .2. As a consequence of Lemma 1. some of the symmetries of the cube form a smaller group.6. Z is generated by 1. and the symmetry group of the triquetrum depicted in Figure 1. . It is denoted by S and called the subgroup generated by S. . for all g. it is necessary and suﬃcient that H be nonempty and closed under the operation of G. . i. A subgroup of a group G is a subset H of G satisfying the following conditions.2. Exercise 1.10. 3. Some groups are generated by a single element. . The elements g1 . This means that. 2. .6. .2. INTRODUCTION This is very natural. h ∈ H.e. Exercise 1. we have h−1 ∈ H. .
two integers are identiﬁed whenever their diﬀerence is a multiple of n.9) is ‘the same’ as Z/3Z. In fact. which is denoted by Z/nZ for reasons that will become clear later.1) (0. Deﬁnition 1. Consider the ﬂat torus T . p2 + q2 ). has integral entries. Deﬁnition 1.2.2. Determine which elements of T have ﬁnite order. (Thus. their sum corresponds to the point r on S with the property that r − (p1 + q1 . 1] × [0. addition modulo n being the operation.2. (0.10: The ﬂat torus. This set forms a group. 1) × [0.0) Figure 1. as all such groups are the same in a sense to be made precise later. . one could say that it is the cyclic group of this order. 1] with opposite edges glued together.15. Exercise 1.0) (1. If there exists an element g ∈ G for which G = g . The order of Z/nZ is n. Exercise 1.1) (1. In this set. then G is called a cyclic group.12. such that the arrows in Figure 1.) The torus has the following group structure: given two points p = (p1 . BASIC NOTIONS 15 Figure 1. Thus the symmetry group of the triquetrum (Figure 1.10 match. The order of g is called the order of g.13.14. It may be inﬁnite. q2 ) on the torus. and it is a cyclic group. Consider the integers modulo n. It can be seen as the square S = [0.2. the set is really in bijection with [0. Prove that a cyclic group is Abelian. the latter term of which may not be in S. 1). where n is a positive integer.1. Let g be an element of a group G. p2 ) and q = (q1 . and which elements have inﬁnite order.2.9: The triquetrum.
16 CHAPTER 1. INTRODUCTION .
Indeed.Chapter 2 Permutation Groups The realization of a group by means of symmetries (more generally. where (12)H = {(12). (13). 1. we write ST for the set {st  s ∈ S. 17 . the additive group. Deﬁnition 2. and H = m = mZ. left cosets coincide with right cosets. . t ∈ T }. (23)}. this is sometimes done.1. the symmetry group of an equilateral triangle (check). . We shall only use G/H in these notes. Then the set of left cosets is {H. Take the subgroup H = (123) . but this meaning of \ should not be confused with the usual settheoretic one. the subgroup generated by m. For s ∈ G and T ⊂ G we also write sT for {s}T and T s for T {s}. For subsets S and T of a group G. It would be logical to denote the set of right cosets of H in G by H \ G. 2 + H.2. and those of the form Hg are called right cosets of H in G. (123). 1 + H. (12). m − 1 + H}.3. For a subgroup H of G the sets of the form gH with g ∈ G are called left cosets of H in G.1.1 Cosets and Lagrange Recall the notion of subgroup from Deﬁnition 1. (13).2.1. .7. Let us ﬁrst focus on the latter aspect. G = Z. (132)}. G = S3 = {(1). . for some m ∈ Z. (12)H}. 2. The set of left cosets is {H.1. Example 2. Since x + H = H + x. so left and right cosets coincide. Deﬁnition 2. Observe that (12)H = H(12) = G \ H. The set of all left cosets of H in G is denoted by G/H. (23). bijections of a set) is closely related to the structure of its subgroups. 2.
1.1. Show that (13) (12) is not a right coset of (12) in S3 . G = Rn . 2. Deﬁnition 2. Also. Then (13)H = {(13). Next. as the product of two elements of H is automatically understood to denote the product in H and not in G. This enables us to leave out the operation symbol throughout. then H divides G. It is easy to verify that left cosets are equivalence classes of the equivalence relation g ∼ h ⇔ g −1 h ∈ H. If the number of left cosets of a subgroup H in a group G is ﬁnite. (123)} and H(13) = {(13).1. If H and K are subgroups of a ﬁnite group G. PERMUTATION GROUPS 3.1.1. this number is called the index of H in G and denoted by [G : H]. Take for instance G = S3 and H the cyclic subgroup (12) . we will see the functions between two groups that preserve the structure of groups. Lemma 2. A similar statement holds for right cosets. m ] = m and Theorem 2. Exercise 2.18 CHAPTER 2. In the previous examples.4. The left cosets of a subgroup H in the group G form a partition of the group G.5. Exercise 2.6. Corollary 2. and that (12) (123) is a right coset of (123) in S3 .9. then the order of g divides that of G. Prove that HK = H∩K . we shall always make it clear to which of the two groups each element belongs.2 Quotient groups and the homomorphism theorem If two groups G and H and some of their elements are in the picture. For g ∈ G. Then the left cosets are the aﬃne subspaces parallel to H.1. (132)}.8 (Lagrange’s Theorem). Example 2. If G is a ﬁnite group and H a subgroup. the map h → gh is a bijection between H and the left coset gH.10. 4. . and similarly for right cosets. we have [Z : [S3 : (123) ] = 2 . we shall use e for the unit element of both groups. A left coset need not be a right coset. So we have the following lemma. the (additive) translation group of a real vector space and H is a linear subspace of G. then H ∩ K HK is also a subgroup of G.7. If G is a ﬁnite group and g ∈ G.1.
that both groups are isomorphic. and denoted by im(φ). A homomorphism from G to H is a map φ : G → H satisfying φ(g1 g2 ) = φ(g1 )φ(g2 ) for all g1 . the set Inn(G) = {φg  g ∈ G} is a subgroup of Aut(G). and φ(eG ) = eH . Example 2. then φ is called an isomorphism. x → x + mZ. Exercise 2.2. Then φ is injective if and only if ker φ = {e}. respectively. and it is clear that ∼ = = is an equivalence relation. Prove that ker(φ) and im(φ) are subgroups of G and H. Here φg denotes conjugation with g. Let φ be a group homomorphism from G to H. is a homomorphism of additive groups. Aut(G) is a group with respect to composition as maps. Let φ : G → H be a homomorphism.1. it is called outer otherwise.2. .6. QUOTIENT GROUPS AND THE HOMOMORPHISM THEOREM 19 Deﬁnition 2.3. We claimed earlier that the symmetry group of the triquetrum (Figure 1. The map Z → Z/mZ.7.5. the map g → det(g) is a homomorphism G → C∗ . By Aut(G) we denote the set of all automorphisms of G. then G is called isomorphic to H. Deﬁnition 2.2. Its kernel is the group SL(V ). If there exists an isomorphism from a group G to a group H. Then φg is an element of Aut(G). Exercise 2. The precise formulation is. If. For a group homomorphism φ : G → H the subset {g ∈ G  φ(g) = e} of G is called the kernel of φ.8.4.2. Example 2. For G = GL(V ). An automorphism of G is called an inner automorphism if it is of the form φg form some g.2. called conjugation with g. and denoted by ker(φ). Example 2. An automorphism of G is an isomorphism from G to itself. For g ∈ G.2. Show that for every g ∈ G we have g ker(φ)g −1 ⊆ ker(φ).2. Let G and H be groups. Prove: 1. Exercise 2. The set φ(G) = {φ(g)  g ∈ G} ⊆ H is called the image of φ. φ is a bijection.2.2.9) is ‘the same as’ Z/3Z. g2 ∈ G. Its kernel is mZ. the group of invertible linear transformations of a complex vector space V . we deﬁne φg : G → G by φg (h) = ghg −1 . this fact is denoted by G ∼ H. in addition.2. 2.2.
2. x0 − x3 so that x = det(ψ(x)). C).2.10. ϕ(A) preserves the metric due to the multiplicative properties of the determinant: ϕ(A)(x) = det(Aψ(x)A∗ ) = det(A) det(ψ(x)) det(A∗ ) = det(ψ(x)) = x. If H is normal. Is (Z/2Z) ⊕ (Z/2Z) isomorphic to Z/4Z? Exercise 2. The map G → G/H. As we have seen before.12. Exercise 2. z. as follows: (gH) · (kH) = (gk)H. y = z. 0 1 2 3 To x ∈ R4 we associate a 2 × 2matrix ψ(x) as follows: ψ(x) = x0 + x3 x1 + ix2 x1 − ix2 . ¯ ¯ ¯ and one checks that this space is invariant under M → AM A∗ for any A ∈ GL(2.9. g → gH is a homomorphism G → G/H. Take the group Z/4Z. Now. if gHg −1 ⊂ H for all g ∈ G. u ∈ C and x = x. PERMUTATION GROUPS Example 2. or. y.11.20 CHAPTER 2. which is the group of all linear transformations of the vector space R4 that preserve the Lorentz metric x := x2 − x2 − x2 − x2 . C) and the Lorentz group L. Determine the subgroups of (Z/2Z) ⊕ (Z/2Z).2. Deﬁnition 2. There is an interesting homomorphism between SL(2. . then we can deﬁne an operation on the set G/H of all left cosets of H in H.9 satisﬁes ϕ(AB) = ϕ(A)ϕ(B). Here A∗ = (¯ji )ij if A = (aij )ij . Exercise 2. the subgroup (1234) of S4 and the multiplicative subgroup generated by i in C. Also.2. This operation turns G/H into a group. we say that H is normal. C). the map ϕ : SL(2. C) → L.2. When that happens. and thus ﬁnish the proof that ϕ is a homomorphism SL(2.2. a Indeed. for A ∈ SL(2. We denote it by H G. Verify that they are all isomorphic. it is not always the case that left cosets and right cosets of a subgroup H are the same. C) → L given by ϕ(A)(x) = ψ −1 (Aψ(x)A∗ ) is a group homomorphism. A subgroup H of G is called normal if every left coset of H in G is a right coset of H in G. equivalently. u = u . Show that ϕ in Example 2. called the quotient group of G by H. ψ is a linear isomorphism from R4 onto ψ(R4 ) = x y z u x.13.
The kernel of a group homomorphism φ : G → H is a normal subgroup of G. Deﬁnition 2. K normal subgroups of G such that H ⊇ K. It is also surjective onto its image.2. QUOTIENT GROUPS AND THE HOMOMORPHISM THEOREM 21 Exercise 2. hence it is an isomorphism. The subgroup (12)(34).18.5 for the proof. Example 2.13) Verify that the operation on G/H is welldeﬁned.2. Let H be a normal subgroup of the group G (cf.19.2. 2. Prove that Inn(G) (cf. then φ(g) = e.4) is a normal subgroup of Aut(G) (cf. Write an element of G as a v 0 1 with a ∈ SO(3) and v ∈ R3 .e. π is a homomorphism: π((gN )(kN )) = π((gk)N ) = φ(gk) = φ(g)φ(k) = π(gN )π(kN ).2. i. Let G be a group.17. Consider the subgroup N of all translations.2. Indeed. Theorem 2. does not depend on the choice of representatives g and k.3 shows that π is injective. Then G/ ker φ ∼ = im φ. Note that one really needs normality of H. Let G be the group of Euclidean motions in R3 (see Chapter 3). Exercise 2. Exercise 2. See Exercise 2. it follows that ker(π) = {eN }. = v . thus.14. if π(gN ) = e. and H. Finally. H/K a normal subgroup of G/K. Prove that any subgroup of index 2 is normal.2.2. This presentation is convenient as the operation in G is now given by the multiplication of 4 × 4 matrices. Let φ : G → H be a group homomorphism. We claim that π : G/N → im φ. Lemma 2. hence φ(g ) = φ(g).2. 1.2. Also.. Verify that G/H does indeed become a group with this operation. Theorem 2.2.20 (First Isomorphism Theorem).2. Example 2.15. Let N := ker φ. Then K is a normal sugroup of H.16. gN → φ(g) is an isomorphism.1). it is welldeﬁned: if g N = gN . hence gN = eN . and G/H ∼ (G/K)/(H/K).2. then g −1 g ∈ N so that φ(g −1 g ) = e. Exercise 2. (13)(24) is normal in S4 . Proof.2. N consists of all matrices I 0 Then N is a normal subgroup of G. 1 .2. and Exercise 2.
2. K is normal in H. H/K is normal in G/K. What is wrong? To answer this question. the puzzle is in the state of Figure 2. Prove the second Isomorphism Theorem. for your friend is watching you and you don’t seem to be able to transform the puzzle back into its initial state. so a fortiori this holds for g ∈ H. sighing that it is too diﬃcult for him. You leave the room to make some coﬀee and when you come back your friend returns the puzzle to you. 2. K a normal subgroup of HK. As time passes. Conﬁdently you start moving around the plastic squares. for readers unfamiliar with the puzzle a possible sequence of moves is shown in Figure 2. Consider the map π : G → (G/K)/(H/K).2. By Theorem 2. and asking you to show him how to solve it. G/H ∼ (G/K)/(H/K) as claimed. At this time.3.22. Then HK is a subgroup of G. . we shall look at some basic properties of permutations.2. you start getting more and more nervous. Hence.3 Loyd’s puzzle A MessedUp Puzzle Suppose that a friend borrows your 15puzzle just after you solved it. and K a normal subgroup of G. H ∩ K is a normal subgroup of H. g → (gK)(H/K). Let G be a group.1. For any g ∈ G. H a subgroup of G. Exercise 2. PERMUTATION GROUPS A E I M B F J N C G K O D H L Figure 2.19. we have gKg −1 ⊆ K. so that it is in the initial state depicted in Figure 2. For any g ∈ G and h ∈ H. Hence.1: Initial state of Sam Loyd’s 15puzzle.21 (Second Isomorphism Theorem).22 CHAPTER 2. = Theorem 2. we have (gK)(hK)(gK)−1 = (ghg −1 )K ∈ H/K as H is normal in G. and its kernel equals {g ∈ G  gK ∈ H/K} = H.2. and HK/K ∼ = H/(H ∩ K). It is clearly a surjective homomorphism. Proof.
A E I M B F J O C G K N D H L A E I M B F J O C G K N D H A E I B F J O C G D H K L M N L Figure 2. LOYD’S PUZZLE 23 A E I M B F J O C G K N D H L Figure 2. .3.2.2: Messed up state of the 15puzzle.3: Possible moves of the 15puzzle.
The image of i is the number that appears below i in the matrix. consider the permutation π from the previous example. we get the picture in Figure 2. . because each ﬁxes the elements occurring in the other. in which both rows contain the numbers 1. . A permutation of this form is called a 3cycle. 5). In this notation. 5. 6) and the 3cycle (3. Hence Sn can be seen as a matrix group. . π is the product of the 2cycle (2. 3. π(5) = 3 and π(3) equals 1 again. With it we associate the 5 × 5matrix Aπ = (aij ) with aij = 1 if π(j) = i. 5. 5. n} → {1. . . a permutation is represented by a diagram as follows: draw two rows of dots labeled 1. 3). n} one writes a 2 × n matrix. n. . .4. i. and this fact makes them commute. . . 3) = (5. . We ﬁnd that π = (1. Then draw a line from vertex i in the upper row to vertex π(i) in the lower. It can be checked that Aπσ = Aπ Aσ . In our example. Tworow notation: for a permutation π : {1. The permutation that satisﬁes these equations and that ﬁxes the other elements is denoted by (1. These are called disjoint. PERMUTATION GROUPS Permutations and their Signs Recall the deﬁnition of a permutation and the symmetric groups Sn . otherwise. .24 CHAPTER 2.e. We have π(1) = 5. three of which we shall discuss here. An example is 1 2 3 4 5 π= ∈ S5 . multiplication being the ordinary matrix multiplication. 4) = (2. 5. 1. 1) = (3. Diagram notation: in this notation. it is usual not to write down the 1cycles. . but this abuse of notation turns out to be convenient rather than confusing. 5 4 1 2 3 Disjoint cycle notation: To explain this notation. one under the other. . and this factorization is unique up to changes in the order. 1). But π does not ﬁx 2 or 4. 3)(2. 0.. Permutations can be represented in several ways. . it is not always clear to which Sn the permutation belongs. 4)(1. Permutation matrix notation: consider once again the permutation π ∈ S5 . π(2) = 4 and π(4) = 2. n. . Leaving these out. indeed. Any permutation can be written as a product of disjoint cycles. . . where we take the normal matrix product in the righthand side.
3. in the diagram notation. in terms of the lengths of the cycles.2. Related to permutation matrices we have the following important deﬁnition.3. 3). In the former case. . given in disjoint cycle notation.3. in the latter an odd permutation.4: The diagram notation.3. 5. 6)(1.2. Can you give a smaller set of 2cycles that generates the whole group? How many do you need? Exercise 2. Exercise 2. It is the kernel of the homomorphism sgn : Sn → {±1} of Exercise 2. For π ∈ Sn we deﬁne the sign of π by sgn π := det Aπ . For example. The sign of a permutation π is either 1 or −1. Give and prove a general formula for the order of a permutation. called the alternating group and denoted by An . Prove that the sign of a kcycle equals (−1)k+1 .3. Exercise 2. Exercise 2.5.3. Show that the 2cycles in Sn generate that whole group. Show that sgn : Sn → {±1} is a group homomorphism.6. The sign of a permutation can be interpreted in many ways. The even permutations in Sn form a subgroup. LOYD’S PUZZLE 25 1 2 3 4 5 1 2 3 4 5 Figure 2.3.3. Deﬁnition 2.4. π is called an even permutation. Determine the order of a kcycle.1. Determine the order of π = (2. Exercise 2.3.4. the number of intersections of two lines is even if and only if the permutation is even. Prove that the sign of a permutation equals (−1)N where N is the number of crossings of the lines drawn in the diagram notation of the permutation.
. . . we have t−1 ∈ T (t(p)). ∗}. 16}. . Formally. . . (1. 3). for j = 1. . after applying the ﬁrst j moves to s. . For p ∈ {1.1 and 2. transformations change the state of the puzzle. . M (2) = {(1. (2. B. Here. (6.5: Numbering of the positions. ◦ m1 (p)) for all j ≥ 0. 6)}. PERMUTATION GROUPS 1 5 9 13 2 6 10 14 3 7 11 15 4 8 12 16 Figure 2. where mj+1 ∈ M (mj ◦ mj−1 ◦ .5. then u ◦ t ∈ T (p). and it is deﬁned as the set of all products (in S16 ) of the form mk mk−1 . The set of all possible transformations starting in a state with the empty square in position p is denoted by T (p). No Return Possible Let us return to Sam Loyd’s 15puzzle. 10)}. (5. if t ∈ T (p) and u ∈ T (t(p)). Finally. 7). . s(p) is the square occupying position p in the state s and ∗ stands for the empty square. The positions are numbered as in Figure 2. m1 . 6). O. . a state can be described as a function s : {1. The Figures 2.26 CHAPTER 2. in taking k = 0. 16} → {A. This reﬂects the condition that mj+1 should be a possible move. .2 depict states of the puzzle. . Also. . we denote by M (p) ⊆ S16 the set of possible moves in a state s with s(p) = ∗. 2). . we see that the identity is a possible transformation in any state. It can be described in terms of states and moves. . . A move m ∈ M (p) should be interpreted as moving the square from position j to position m(j). 6). and M (6) = {(2. Moves can be combined to form more general transformations. 5)}. 2). more generally. For example M (1) = {(1. In particular. . 16. (6. . Let s ∈ S with s(p) = ∗ and let t ∈ T (p). (2. . Moves and. Then t changes the state s to the .
and we shall presently make clear why.7. . Now that we captured Sam Loyd’s 15puzzle in a comfortable notation. i. which is deﬁned by (t · s)(q) = s(t−1 (q)). That is.5.1) whenever the lefthand side makes sense. and then πt0 also equals (14.e. 15.2. for j = 1.e. 9.3. 8. 10). 6). Note that this is independent of the choice of s. LOYD’S PUZZLE 27 Figure 2.6. order the nonempty squares according to the path in Figure 2.2 into the initial state s0 of Figure 2. state t · s. Now consider any position p. considered as a permutation in S15 . 15). In words. so the move equals (6. we can formalize our question as follows: is there a t0 ∈ T (16) that transforms the messed up state s1 of Figure 2. It interchanges the squares in position 6 and 10 in Figure 2. We have that πu◦t = πu ◦ πt (2. that satisﬁes t0 · s1 = s0 ? The answer is no. the jth nonempty square one encounters when following this path in a particular state s is called the jth square in s. A moment’s thought leads to the . . . 15). 7. This is an odd permutation. The corresponding permutation πm is the cycle (10... We see that the t0 we would like to ﬁnd equals (14.6: Ordering of the nonempty squares in each state. i. . the square at position q after applying t is the same as the square at position t−1 (q) before that transformation. In each state. Now consider the move m depicted in Figure 2. any t ∈ T (p) and some s ∈ S with s(p) = ∗. Then we deﬁne a permutation πt ∈ S15 by πt (i) = j ⇔ the ith square in state s is the jth square in state t · s.1. whenever u ∈ T (t(p)).
3). PERMUTATION GROUPS 6 7 10 8 9 Figure 2. m)) = α(gh. such a transformation cannot possibly be (14.4. m) for all m ∈ M and g. α(e. Prove that the set {t ∈ T (16)  t(16) = 16} is a subgroup of A16 .7. Let G be a group and M a set. α(g. You conclude that your friend must have cheated while you were away making coﬀee. Exercise 2.3.1. As a consequence of this and equation (2. Use this to give another argument why your friend must have cheated. their ‘acting’ on the puzzle states reﬂects an important phenomenon in group theory. Deﬁnition 2. Show that the set {πt  t ∈ T (p) for some p} forms a subgroup of S15 . conclusion that πm is an odd cycle for any move m. and 2. m) = m for all m ∈ M . α(h.4 Action of groups on sets Although the transformations in the 15puzzle do not quite form a group. Therefore. 2. hence an even permutation. 15). h ∈ G.8. namely that of an action. by lifting some of the squares out of the game and putting them back in a diﬀerent way. Exercise 2.7: Moves correspond to odd cycles. . a sequence of such moves will also be mapped to an even permutation by π. An action of G on M is a map α : G × M → M satisfying 1.3.28 CHAPTER 2.
2. 3. and has kernel a group of order 4. The linear action of SL(2. Consider G = S4 and put a = {12. y) for α ∈ R in xdirection and all inﬂations (x. 4} into two parts of size two. 34}. R) acts on R3 by matrixvector multiplication. m). This acts on the solutions of the diﬀerential equation d ( )2 (y) = −y. c d cx + d Note.4. π(j)}. (Write down its nontrivial elements!) Example 2. 2}. and similarly on the set {a. g) = hgh−1 is an action. Let G be a group.2. for subsets S ⊂ G and T ⊂ M we write ST for {α(g. . generated by all translations (x. j}) = {π(i). that this aﬃne part is not invariant under the group. hence induces an action on P1 (C). . There is a natural action (as above) on the set of pairs. 3. . b. we leave α out and write gm for α(g. y) → (x.4. 23}. . Then G acts on itself by conjugation. For {g}T we write gT and for S{m} we write Sm. 2. c = {14. ACTION OF GROUPS ON SETS 29 If the action α is obvious from the context. m) = φ(g)m. It is surjective. C) on C2 by left multiplication permutes the ﬁbers of π. and on the set of all twoelement subsets of {1. the map α : G × G → G deﬁned by α(h. Conversely. 2. 24}. . dx Example 2. however. Here 12 stands for {1. c}). etc. 1. Also.4.6.3. given a homomorphism φ : G → Sym(M ). Example 2. n} by π · i = π(i). we may normalize it to 1. m)). we ﬁnd a homomorphism S4 → Sym({a. In other words.4. . b. m ∈ T }. an action on M is nothing but a homomorphism G → Sym(M ).4. Denote by π : C2 \ {0} → P1 (C) the map sending a pair of homogeneous coordinates to the corresponding point. . we have an action α : G × M → M given by α(g. βy) for β ∈ R∗ . The group GL(3. m)  g ∈ S.4. The group Sn acts on {1.4. n} by π({i. c} of all partitions of {1. . Let us give some examples of a group acting on a set. and the action on the ﬁrst coordinate x is given by ax + b a b x= . that is. On the aﬃne part of P1 (C) where the second homogeneous coordinate is nonzero. we have a homomorphism G → Sym(M ) given by g → (m → α(g. . The group SL(2. Given an action α. Lemma 2. C) acts on the complex projective line P1 (C) as follows. b = {13.5. Remark 2. Thus. Consider the group of all motions of the plane. y) → (x + α.
·) : M → M is a homomorphism. g)) = h(kgk −1 )h−1 = (hk)g(hk)−1 = α(hk. Let G be a group acting on M and let m ∈ M . Let G be a group acting on M .1. α(e. the map sending g ∈ G to the map α(g. First. Then Gm is a subgroup of G. Lagrange’s theorem has an important consequence for problems. the elements of S are called symmetries of T . The following lemma allows us to partition a set M on which a group G acts into orbits.4. A subset T ⊆ M is called invariant under S ⊆ G if ST ⊆ T . Let G be a group acting on a set M . To deﬁne symmetry in a general context. we look at sets that are mapped to themselves by a given group.4. . α(k. Gm = M for any m ∈ M . Deﬁnition 2. but need not be an isomorphism.4.10. α(h. We now want to make clear how the action of groups can help in classiﬁcation and counting problems. More precisely.4. g · m → gGm is well deﬁned and bijective. and the map f : Gm → G/Gm . One must distinguish the elements of G from their actions on M . The set Gm is called the orbit of m under G.12.4. Deﬁnition 2. if G is ﬁnite.4. In this case.7. We check the two conditions of Deﬁnition 2. Lemma 2.30 CHAPTER 2.8. Let G be a group acting on a set M and let m ∈ M . and we shall do so by means of the following problem. g).9. stated diﬀerently. An action of a group G on a set M is said to be transitive if it has only one orbit. As a consequence. in which one has to count the number of symmetries of a given set. and second. g) = ege−1 = g for all g ∈ G. then G = GmGm . Then the sets Gm = {gm  g ∈ G} for m ∈ M partition the set M .11.4. Theorem 2. Deﬁnition 2. Then the set Gm := {g ∈ G  gm = m} is called the stabilizer of m. PERMUTATION GROUPS Proof. or. Remark 2. It can be formulated as follows.
A simple graph is a function f: X 2 → {0. say C4 H10 (see [6]). Formally. . and the ﬁrst step in solving it is to formalize the meanings of the words graph and nonisomorphic. y}. The set of all graphs on X is denoted by Graphs(X). 1} −− π·f This diagram tells us that e ∈ X is an edge in the graph f if and only if π · e 2 is an edge in the graph π · f . this means that the pair is an edge in the graph. the question is formulated rather vaguely. to make them exactly the same one has to relabel the vertices. question has strong analogue in counting all isomers with a given chemical formula. this leads to an action of Sym(X) on Graphs(X) in the following way. 1} −− π· Id X 2 X 2 f − − → {0.8 are essentially the same.2. Of course. we deﬁne π · f by the commutativity of the following diagram. The ﬁrst two graphs in Figure 2. In formula. graphs are associated with a picture like the ones in Figure 2. but seemingly merely theoretical. y} = {π(x). What is the number of nonisomorphic simple graphs with vertex set X? This natural. − − → {0. 1}. Such a relabeling is in fact a permutation of X. π · f (x) = f (π −1 x). ACTION OF GROUPS ON SETS 31 1 2 1 2 1 3 2 3 4 3 4 4 5 Figure 2. Given is a ﬁnite set X with X = v. two of which are isomorphic. π(y)}.4.8: Three graphs. Furthermore. Often. Here X stands for the set of all unordered pairs from X. If the function f 2 takes the value 1 on such a pair {x.8. First of all Sym(X) acts on X by 2 π · {x.
Now our original question can be rephrased very elegantly as follows.16. Consider the permutation π = (1.14. 4)) = 2 and o((12)(34)) = 4. Exercise 2. the number of nonisomorphic graphs on 4 vertices equals (1 · 26 + 6 · 24 + 8 · 22 + 6 · 22 + 3 · 24 )/24 = 11. For v = 4. 5}. o((123)) = 2.13.4. The function f is ﬁxed by π if and only if f is constant on the orbits of π in X . 4.4.4. Lemma 2. Try to ﬁnd a general formula in terms of the cycle lengths of π and the number v.15. we have o(e) = 6. the conjugacy clases of each of these permutations have sizes 1. Calculate the number o(π). 6. Observe that o(π) only depends on the cycle type. 3.32 CHAPTER 2. Example 2. the number of f ∈ Graphs(X) that are ﬁxed by π. g∈G Here ﬁx(g) denotes the number of m ∈ M with g · m = m. Hence. we ﬁnd that the number 2 of nonisomorphic graphs on X equals 1 v! where v = X.4. o((12)) = 4. How many orbits does the action of Sym(X) have on Graphs(X)? The following lemma answers this question in general. 2. 8. π∈Sym(X) . Can you ﬁnd all? 2o(π) . Denoting the number of orbits of π on X by o(π). respectively. 5) on {1.4. 2)(3. 2. Let G be a ﬁnite group acting on a ﬁnite set M . o((1. To apply this lemma to counting graphs.13 (CauchyFrobenius). we need to know for some π ∈ Sym(X). Then the number of orbits of G on M equals 1 G ﬁx(g). 4. Prove Lemma 2. 3. 6 and 3. PERMUTATION GROUPS Graphs that can be obtained from each other by the action of some π ∈ Sym(X) are called isomorphic. This reﬂects 2 the condition that π should map edges to edges and nonedges to nonedges. Exercise 2.
when given a two. as can be seen using the rotations of order 3 around axes through the Catom and an Hatom. Consider one of the other H atoms. respectively. we consider an example that motivates the development of this theory. 3.1 Motivation In low dimensions.e.Chapter 3 Symmetry Groups in Euclidean Space This chapter deals with groups of isometries of ndimensional Euclidean space. preserve a 2. Its orbit Ga consists of all four Hatoms. Applying Lagrange’s theorem we ﬁnd G = 4Ga .or 3dimensional lattice. say b. however. let us compute G. As a start. The hydrogen atoms are arranged in a regular tetrahedron with the Catom in the center. In these two cases.1. we try to ﬁnd all rotation axes: 33 . and the stabilizer (Ga )b consists of the identity and the reﬂection in the plane through a. and then we restrict our attention to n = 2 or 3. i. Its orbit under the stabilizer Ga has cardinality 3. Consider a ﬁxed Hatom a.or threedimensional object. suppose that we want to describe the symmetry group G of the CH4 molecule depicted in Figure 3. we classify the ﬁnite groups of isometries. and point out which of these are crystallographic point groups. As an example. so that we ﬁnd G = GaGa  = GaGa b(Ga )b  = 4 · 3 · 2 = 24. there are few ﬁnite groups of isometries and this enables us. To gain more information on the group G. Only half of these isometries are orientationpreserving. We set up a framework of results that are valid for general n. First.. The stabilizer of a consists of symmetries of the remaining 4 atoms. b and the Catom. to determine its symmetry group simply by enumerating all axes of rotation and investigating if the object admits improper isometries such as inversion and reﬂections.
SYMMETRY GROUPS IN EUCLIDEAN SPACE H C H H H Figure 3. . but does contain other improper isometries. 2. Figure 3. Later on. There are three twofold axes.34 CHAPTER 3. and compose it with the inversion i. namely those through an Hatom and the center of the opposite side of the tetrahedron.2: A regular tetrahedron inside a cube. There are 4 threefold axes. it is denoted by T . and the improper isometries of the tetrahedron may be obtained as follows: take any proper isometry of the cube which is not a isometry of the tetrahedron. but their realizations in terms of isometries are diﬀerent in the sense that there is no isometry mapping one to the other. we may conclude that G is T ∪ i(W \ T ). a group that can be described by Figure 3. we shall see that this information determines the orientation preserving part of G uniquely. Half of the proper isometries of the cube are isometries of the tetrahedron. From the fact that G does not contain the inversion. This point will also be clariﬁed later.1: The CH4 molecule. Note that T ∪ i(W \) and W are isomorphic as abstract groups. 1. Here we see a regular tetrahedron inside a cube. namely those through the centers of two perpendicular sides.2.
hence zero. Lemma 3. 0)2 + d(ry. so that rx + ry − r(x + y) is perpendicular to all of Rn . where we use that r is an isometry ﬁxing 0. ry) = (r−1 z.2. This proves the second statement. ry)2 + d(rx. Lemma 3. rx + ry) = (z.1. y) = x − y = (x − y.2. Consider the reﬂection in a plane through two Hatoms and the Catom. the group G is called a crystallographic point group.2 Isometries of nspace d(x.3. How can it be written as the composition of the inversion and a proper isometry of the cube? The ﬁnite symmetry group of the tetrahedron and the translations along each of three mutually orthogonal edges of the cube in Figure 3. For a ﬁxed a ∈ Rn . ISOMETRIES OF N SPACE 35 Exercise 3. so t−1 g is an isometry ﬁxing 0. g(y)) = d(x. y) for all x. For any g ∈ AO(n.2. The following deﬁnition and lemmas will explain the name of this group. Similarly.2 generate the symmetry group of a crystallike structure. Deﬁnition 3. Deﬁnition 3. which we denote by AO(n. there exists a unique pair (a. 0)2 )/2 = (x.2.2. ry) = = (−d(rx. and it a a is clear that a = g(0) is the only value for which this is the case. rx) + (z. x + y) = (z. y)2 + d(x. if we take the ﬁrst statement for granted. 0)2 )/2 (−d(x. one ﬁnds r(αx) = αrx for α ∈ R. For any z ∈ Rn . Proof. 3. r leaves the inner product invariant. Proof. Let x. R) ﬁxes 0. For this reason. such that g = ta r. r(x + y)). An isometry is a map g : Rn → Rn such that d(g(x). R). This proves the ﬁrst statement. y ∈ Rn . r) where a ∈ Rn and r ∈ AO(n. Then r is an orthogonal Rlinear map.4. Then t−1 g(0) = 0. 0)2 + d(y.1. y ∈ Rn . Let r ∈ AO(n. the isometry ta : x → x + a is called the translation over a. and compute (rx. R).1. y).3. x − y)1/2 = ( i Consider the ndimensional Euclidean space Rn with the distance function (xi − yi )2 )1/2 .2. The set of all isometries is a subgroup of Sym(Rn ). we have (z. Hence. R) be an isometry ﬁxing 0. . Set a := g(0).
5. Deﬁnition 3. so it suﬃces to prove the assertion for r with r0 = 0.2. Its kernel is obviously T (G). Now the proof of Lemma 3.2.2. (The assertion is used in the second part of Deﬁnition 3.4 show that R is welldeﬁned. We are interested in certain groups that act discretely.8.2. v).3 and 3. and let a2 ∈ T (G). Let r ∈ R(G) and a1 ∈ Rn such that ta1 r1 ∈ G. the map R : G → O(n. and the dimension of its Rlinear span is called its rank. Proof. Set v = r(λx + µy) − λr(x) − µr(y) and consider (v.) Hint: If r is an isometry. and check that ta1 r1 ta2 r2 = ta1 +r1 a2 r1 r2 . which we shall identify with the group of all translations in G. This proves that R is a homomorphism. R). and a group homomorphism. Deﬁnition 3. For a group G. Lemma 3. r2 ∈ O(n. for x ∈ Rn . The group of all orthogonal linear maps Rn → Rn is denoted by O(n. one can prove the following. For a subgroup G ⊆ AO(n. Using the linearity of isometries ﬁxing 0. Proposition 3.2.2. hence the name aﬃne orthogonal transformation.7. R). Let r1 . Prove that an isometry is bijective. Then. followed by a translation. a2 ∈ Rn . and this set is invariant under R(G). Using bilinearity of the inner product and orthogonality of r.4 uses r−1 . Its kernel is T (G). An isometry is thus an orthogonal linear map.1.2. A subset V ⊆ Rn is called discrete if for every point p ∈ V there is an open neighbourhood U ⊆ Rn such that U ∩ V = {p}.6. R) and a1 . v) = 0. then t−r0 r is an isometry ﬁxing 0. so we need another proof. A discrete subgroup of Rn is called a lattice. −1 −1 −1 ta1 r1 ta2 r1 t−a1 x = ta1 r1 ta2 (r1 x − r1 a1 ) = ta1 (x − a1 + r1 a2 ) = tr1 a2 x. this is an additive subgroup of Rn . R) deﬁned by g = ta R(g) for some a ∈ Rn is welldeﬁned. . This shows shows that T (G) is invariant under R(G). it can be shown that (v.2. SYMMETRY GROUPS IN EUCLIDEAN SPACE Exercise 3. we write T (G) := {a ∈ Rn  ta ∈ G}.36 CHAPTER 3.
let m be the minimum of the continuous function x→ xi ai i on the unit sphere {x ∈ Rk  x = 1}. hence a lattice. vk ∈ L whose images under π are linearly independent. For v ∈ L we can write k π(v) = i=2 ci π(vi ) . Conversely. . For any nonzero x ∈ Zk . there is an > 0 such that v ≥ for all v ∈ L \ {0}. First let us show that L is a closed subset of Rn . In particular. . This proves that π(L) is discrete. . Clearly. we ﬁnd xi ai = x i i xi ai ≥ x m ≥ m. Let a1 . . Now v1 . Proposition 3. As 0 is an isolated point.3. Let L ⊆ Rn be a lattice. ISOMETRIES OF N SPACE Lemma 3. Now suppose that every lattice of rank k − 1 is of that form. Then the set L= Zai i is a lattice. . After subtracting an that v − π(v) ≤ v1 /2.2. a lattice of rank 0 is of the desired form. we proceed by induction on the rank of the lattice. and such that k π(L) = i=2 Zπ(vi ).2. 37 Proof. m > 0. and let L ⊆ Rn be a lattice of rank k.9. But then v − w ≥ for all distinct v. . Let ⊥ π : Rn → v1 denote the orthogonal projection along v1 .2. . To prove that an additive subgroup of Rn is a lattice. vk are linearly independent. ak ∈ Rn be linearly independent over R. we ﬁnd π(v) ≥ 3/2 v1 .10. . any sequence (vn )n in L with vn − vn+1 converging to 0. x proving the ﬁrst statement. is eventually constant. Being discrete and closed. it suﬃces to prove that 0 is an isolated point. Then π(L) is an ⊥ additive subgroup of v1 . Using integer multiple of v1 from v. Then L is a closed set. . . L has only ﬁnitely many points in any compact set. As the ai are linearly independent. so there exist v2 . w ∈ L. . . It has rank k − 1. Hence. Let v ∈ L be such that π(v) = 0. we may assume √ v ≥ v1 and Pythagoras. Proof. For L as in the lemma. every lattice is of this form. For the converse. one can choose a v1 ∈ L of minimal nonzero norm.
k L= i=1 Zvi . x = 1}. the permutation representation of G on the ﬁnite n set X := i=1 Gai is faithful. As L is discrete. SYMMETRY GROUPS IN EUCLIDEAN SPACE k with ci ∈ Z. Then R(G) is a crystallographic point group.15. A subgroup G ⊆ AO(n. By Proposition 3. . Corollary 3. then T (G) does so. for each i. we have T (G) = {0}. .14. . this does not hold for R(G).2. R) is a group that acts discretely.38 CHAPTER 3.2. and that R(G) = {rn  n ∈ Z}. Example 3. Proof. as the following example shows. Proof. and by minimality of v the scalar is an integer. Then v− i=2 ci vk ∈ L is a scalar multiple of v1 .2. is called a crystallographic group in n dimensions. R) leaving invariant a lattice of rank n. an ∈ L be linearly independent. the Gorbits Gai are ﬁnite. Hence.2. If G ⊆ AO(n. . Let a1 . Let α ∈ R be an irrational O(3. As r has inﬁnite order. In general. . R) for which T (G) is a lattice of rank n. Hence.6. 1 Consider the group G generated by g = te3 r. As the ai span Rn . R) leaving invariant a lattice L of rank n. 0)T is an inﬁnite set in the circle {x ∈ R3  x3 = 0. T (G) is invariant under R(G).2. A subgroup G ⊆ AO(n. it contains only ﬁnitely many vectors of norm ai .11.13. Theorem 3. is called a crystallographic point group in n dimensions. Hence. and R(G) does not act discretely: the R(G)orbit of (1. and deﬁne r ∈ 0 0 . Any crystallographic point group is ﬁnite.12. Now apply Theorem 3. too. This phenomenon does not occur if we require T (G) to have the maximal possible rank. Let G be a subgroup of O(n. 0. From g n = tne3 rn it is clear that G acts discretely. R) is said to act discretely if all its orbits in Rn are discrete. R) by cos α sin α − sin α cos α r= 0 0 multiple of π.2.2. R) be a crystallographic group. Let G ⊆ AO(n. Deﬁnition 3. A group G ⊆ O(n. Then G is ﬁnite.13. G is embedded into the ﬁnite group Sym(X). Deﬁnition 3.
An isometry g ∈ AO(n. R) ﬁxes a point in Rn . the proper ones form a subgroup of index 2.3) are isomorphic as abstract groups.3. det(ra ) = −1. it belongs to O(n.19. R) is called proper or orientation preserving if det r = 1.2. 5. i. Exercise 3. Deﬁnition 3.e. R): ﬁrst classify all ﬁnite subgroups of O(n.21.16. 4. Exercise 3. Show that the symmetry group G of Figure 1. 2. the groups G and hGh−1 for some h ∈ AO(n.4 is not the semidirect product of R(G) and T (G). ﬁxes each vector in the hyperplane a⊥ .2. Such an element is called a reﬂection. a)a. Finally. Show that any ﬁnite subgroup G ⊆ AO(n. Otherwise. up to the point of determining the crystallographic point groups. Exercise 3. a = 1). it is called improper. L) of a crystallographic point group R and a lattice L of rank n left invariant by R. ISOMETRIES OF N SPACE 39 This corollary allows for the following approach to the classiﬁcation of crystallographic groups in O(n.2. R). This is a ﬁner equivalence relation than isomorphism.2. then investigate for each of them whether it leaves invariant a lattice of rank n. 3.. R). The same holds for the groups W and T ∪ i(W \ T ) (see Section 3. in any subgroup G ⊆ AO(n. ra has order 2.2. The determinant of an orthogonal map turns out to be a useful tool there. ‘classiﬁcation’ means ‘up to equivalence under AO(n. R) are considered the same. for each pair (R.17. R). The set of all proper isometries in O(n.18. In the next sections.20. R). Among them is always the semidirect product L R: the subgroup of AO(n.2. Exercise 3. An isometry r ∈ O(n.4). Example 3. R)’. where a ∈ Rn has norm 1 (that is. . we shall carry out this classiﬁcation in the cases n = 2 and 3. there may be several nonequivalent groups G with T (G) = L and R(G) = R. as is shown in the following example. R) containing improper isometries. ra (a) = −a.2. Consider the following map: ra : v → v − 2(v. R) is called proper if R(g) is proper. R) generated by L and R. Prove that. In this setting. Hint: consider the ‘average’ of an arbitrary orbit. Prove that 1. R) is denoted by SO(n. but not equivalent under AO(2. The groups C2 and D1 (see Section 3.
By minimality of φ. gy) g∈G for x. denoted by Cn . In two dimensions. 2. and deﬁne x. The dihedral group of order 2n.22. y := 1 G (gx. There are two possibilities: 1. and equal to Cn . and generated by r2π/n . Let φ > 0 be minimal such that rφ ∈ G. If θ is any other angle such that rθ ∈ G. By Lemma 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE Exercise 3. Let G be a ﬁnite subgroup of the group GL(n.40 CHAPTER 3. σ = 0. R) such that B T B = A. R). these are the only two classes. any element of this group equals some rφ .3 The Finite Subgroups of O(2. as well as in art and many other areas. as we shall now prove. R) is a reﬂection in a line. R).2. . 3.) denote the standard inner product. Hint: let (.1. In fact. Hence G is generated by r2π/n . The linear map rφ : R2 → R2 with matrix cos φ − sin φ sin φ cos φ with respect to the standard basis is an element of SO(2. Let us describe two classes of ﬁnite subgroups of O(2. Show that G leaves invariant a positive deﬁnite symmetric bilinear form. y ∈ Rn .3. Let G be a ﬁnite subgroup of O(2. Show that BGB −1 ⊆ O(n. In order to describe these symmetries.. is symmetric and positive deﬁnite. R).20. One may think of ﬂowers (seen from above). frieze patterns on picture frames. As G is ﬁnite. many plane ﬁgures with certain symmetries exist. R) In nature. and generated by r2π/n and the reﬂection in a ﬁxed line. R).. The cyclic group of order n. R). we investigate the ﬁnite subgroups of O(2. Any improper isometry in O(2. ej . 1. Lemma 3. called rotation around 0 over φ. Conversely.1. R). . The proof of this lemma is straightforward. Show that . then θ = kφ + σ for some integer k and angle σ ≥ 0 and < φ. we have two diﬀerent types of isometries. . and B ∈ GL(n. Escher’s plane tilings and decorative wall papers.2. φ = 2π/n for some integer n > 0. and that it is hence conjugate to a subgroup of O(n. it consists of rotations rφ . see Exercise 3. denoted by Dn .3. G ⊆ SO(2. R). R) of all invertible n × nmatrices. Let A be the matrix with entries aij = ei .
3.3. THE FINITE SUBGROUPS OF O(2, R)
41
2. G contains a reﬂection g in a line l. Then G ∩ SO(2, R) is a subgroup of index 2 in G, and by the ﬁrst case generated by r2π/n for some integer n > 0. Now G is generated by the g and r2π/n , and conjugate to Dn . We have thus proved the following theorem. Theorem 3.3.2. Let G be a ﬁnite subgroup of O(2, R). Then G is conjugate to one of the groups Cn or Dn for some integer n ≥ 1. Exercise 3.3.3. Show that O(n, R) is generated by reﬂections. Hint: show ﬁrst that it is generated by elements that ﬁx an (n − 2)dimensional subspace of Rn pointwise, and that O(2, R) is generated by reﬂections. Using this classiﬁcation, we can ﬁnd which of the ﬁnite groups are crystallographic point groups. Theorem 3.3.4. The following groups are crystallographic point groups: C1 , C2 , C3 , C4 , C6 , D1 , D2 , D3 , D4 , D6 . Moreover, any crystallographic point group G in two dimensions is conjugate to one of the above. Proof. For the ﬁrst part, one only needs to construct invariant lattices for D4 and D6 (see 3.3.5), as the other groups are contained in one of these see. For the second statement, suppose that G is a crystallographic point group. According to Theorem 3.2.13, G must be ﬁnite. The lattice L is invariant under the subgroup K of O(2, R) generated by G ∪ {i}, where i denotes the inversion. As the inversion commutes with all elements of G, K is ﬁnite (either equal to G or twice as large). Therefore, according to Theorem 3.3.2, K is conjugate to Cn or Dn for some n; as K contains the inversion, n must be even. So either n ≤ 6, in which case G is of one of the types mentioned in the theorem, or n ≥ 8. Suppose that the latter were the case and choose a vector c in the lattice of shortest possible length. Now r2π/n c ∈ L and c − r2π/n c ∈ L. A little calculus shows that the latter vector is shorter than c because the rotation angle too small (for n = 8, this can be seen from Figure 3.3), and we arrive at a contradiction.
Exercise 3.3.5. Show that D4 and D6 are crystallographic point groups. Note that we have only classiﬁed all crystallographic point groups, and not the crystallographic groups. There are 17 of the latter, and only 10 of the former; see [2].
42
CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE
chc hc c
Figure 3.3: Why n cannot not be 8.
3.4
The Finite Subgroups of O(3, R)
In this section, we shall repeat the classiﬁcation of §3.3, but now for three dimensions. Lemma 3.4.1. Any element g ∈ SO(3, R) ﬁxes a line, i.e., with respect to some orthonormal basis g has the matrix 1 0 0 M ,
where M is the matrix of an element of SO(2, R). Proof. Let λ1 , λ2 , λ3 be the eigenvalues of g. As g leaves the norm invariant, all have norm 1. As g is proper, we have λ1 λ2 λ3 = 1. Now either all are real, in which case at least one of them is 1, or we may assume that λ1 is real and λ3 = λ2 . But then λ2 λ3 = λ2 2 = 1, so λ1 = 1. An element of SO(3, R) is called a rotation, the line ﬁxed by it its axis and the angle corresponding to M its rotation angle. We present some classes of ﬁnite subgroups of SO(3, R). 1. Fix an axis l and consider the proper rotations around l over the angles 2kπ/n for k = 0, . . . , n − 1. They form a group, denoted by Cn , due to its analogy to the class denoted by Cn in the twodimensional setting. 2. Start with the group Cn of rotations around l and choose a second axis m perpendicular to l. Consider the group generated by Cn and the rotation around m over π. It is twice as large as Cn , and denoted by Dn . Note that it contains only proper isometries. The proper symmetries of an nprism form a group of type Dn . Note that D1 is conjugate to C2 .
3.4. THE FINITE SUBGROUPS OF O(3, R)
43
3. The ﬁve platonic solids (tetrahedron, cube, octahedron, dodecahedron and icosahedron) with vertices on the unit sphere yield examples of ﬁnite subgroups of SO(3, R). However, only three distinct classes arise: • T , the group of all proper rotations leaving a tetrahedron invariant. • W , the group of all proper rotations that are symmetries of a cube. • P , the group of all proper rotations leaving a dodecahedron invariant. The reason is that the octahedron and the cube are polar ﬁgures: it is easily seen that the proper symmetries of a cube are exactly those of the octahedron which has its vertices in the middles of the faces of the cube. A similar statement holds for the dodecahedron and the icosahedron. Theorem 3.4.2. Any ﬁnite subgroup of SO(3, R) is conjugate to one of the following. Cn Dn T, W, P. The following proof is due to Euler. Proof. Let G be a ﬁnite subgroup of SO(3, R). Consider the set S deﬁned by S := {p ∈ R3  p = 1, there exists g ∈ G such that g = e and g(p) = p}. Count the number N of pairs (g, p), with e = g ∈ G and p on the unit sphere, that satisfy g(p) = p. On one hand, each g = e ﬁxes exactly two antipodal points, hence this number is N = 2(G − 1). On the other hand, for each p ∈ S, the number of g = e ﬁxing p equals Gp  − 1. Hence N= (Gp  − 1).
p∈S
for n ≥ 1, for n ≥ 2,
If gp = p and g = e, and h ∈ G, then (hgh−1 )hp = hp and hgh−1 = e, so S is invariant under G. Partition S into orbits of G. The value of Gp  is constant on such an orbit o, namely G/o, due to Lagrange’s theorem. Hence we ﬁnd 2(G − 1) =
o
o(G/o − 1),
where the sum is taken over all orbits of G in S. Hence, dividing both sides by G, we get o 2 = (1 − ). 2− G G o
4. 6. #and s equals (sum i: (11/a[i])). hence each ao is at least 2. On the other hand. fi. n 4. b:=a[Length(a)].1/2+2/3). each orbit in S has size at most G/2. 6 8. 4 ] G:12 12. Hence.x>g/x). Consider the following function in GAP. SYMMETRY GROUPS IN EUCLIDEAN SPACE Order n 2n 12 24 60 Number of orbits on S 2 3 3 3 3 Their sizes 1. od. if we set ao := G/o for each orbit o on S.44 CHAPTER 3. end.[b]).1. 6 × 2 6 × 5. while extend(Concatenation(a.3 leads to Table 3. 12 12. Explain why the function call extend([2. local b. 4/5 ] G:12/5 6. n × 2 4 × 3. This procedure ‘extends’ a in all #possible ways. 15 × 2 Group type Cn Dn T W P Table 3. That is.g.4. when called as extend([2. and prints Orbit Orbit Orbit Orbit sizes:[ sizes:[ sizes:[ sizes:[ 6/5.3.1: The ﬁnite groups of proper rotations in space. 1 2. it remains to check that this data determines the group up to conjugation. if s >= 2 then return false. 3 × 4. 6 ] G:24 30. 20. Exercise 3. 10 × 3. else g:=2/(2s). 12 ] G:60 .o. then we have (1 − 1/ao ) < 2. Now Exercise 3.s+(11/b)) do b:=b+1. under the condition that this sum remains smaller than #2.3].2]."\n"). o As each s ∈ S has nontrivial stabilizer.1) results in an endless loop. 4.4. it provides the numerical data for it. o:=List(a. 30 rotation axes 1×n 1 × n.s) #Precondition: a is a nonempty nondecreasing list of integers >=2.g. 20. n. 3 × 2 4 × 3. extend:=function(a. the function does halt. return true. 8. Print("Orbit sizes:"." G:".o.
The points in the diﬀerent orbits are given diﬀerent shades of grey.3. for n ≥ 1. R) containing improper rotations. for n ≥ 2. R) has index 2 in G. Dn ∪ iDn . THE FINITE SUBGROUPS OF O(3. W ∪ iW. R) are ﬁnite subgroups. Cn ∪ i(Dn \ Cn ). R) in which H has index 2. then this equation deﬁnes a ﬁnite subgroup of O(3. Dn . P ∪ iP. then we have a disjoint union G = H ∪ iH. If G contains the inversion i.4. if G . Conversely. Cn ∪ i(C2n \ Cn ). R).4.4: The orbits of W on the points on rotation axes. Make similar pictures for the group types Cn . Note that this is isomorphic to H × C2 as an abstract group. Then the set G := H ∪ i(G \ H) is a subgroup of SO(3. R). R) is conjugate to one of the following: Cn . P. Dn ∪ i(D2n \ Dn ). T. Describe the symmetry axes for the icosahedron. but not equivalent to it. and H has index 2 in G . T ∪ iT.5.4.2. Dn and T . Compare this with Table 3. Exercise 3. H ⊆ SO(3. which is isomorphic to G as an abstract group. Let G be a ﬁnite subgroup of O(3.1. they form the set S in the proof of Theorem 3. Its group of proper symmetries contains nontrivial rotation axes through the points drawn in that picture.4. and one can recover G by G = H ∪ i(G \ H). Now let us try to include improper rotations. W. This enables us to classify all ﬁnite subgroups of O(3. Any ﬁnite subgroup of O(3.4. Theorem 3. .4 we see a cube with vertices on the unit sphere. R) 45 Figure 3. Cn ∪ iCn . T ∪ i(W \ T ). In Figure 3. Then H := G ∩ SO(3. Now assume that G does not contain i.
1 and the list in Theorem 3. 3. 2. Exercise 3. Check whether G contains any improper rotations. Then. Consider the full symmetry group of the tetrahedron.4. Suppose that G contains improper rotations.5. Determine the symmetry group of the N4 P4 molecule. They determine the subgroup H of G of all proper rotations. 5.46 CHAPTER 3.4. 1.6. Consider the molecule depicted in Figure 3.10. for this reason the second list starts with n = 2.4.4. Note that D1 and C2 describe the same class.7. it is of type Cn ∪i(C2n \Cn ). Exercise 3.1. Suppose that H = Cn and that G contains improper rotations. depicted in Figure 3. based on Table 3.9.4. and look at the lines between antipodal vertices of the cube. If not. This can be seen as follows: consider the group of proper symmetries of a cube with center 0. Otherwise. 4. If so. The last group in the list is made possible by the fact that a group of type W has a subgroup of type T of index 2. .8. one wants to determine the (ﬁnite) symmetry group of a given object. Exercise 3. then G = H. Check whether i ∈ G. SYMMETRY GROUPS IN EUCLIDEAN SPACE We do not give all details of the proof.4. these proper rotations form the symmetry group of a regular tetrahedron with center 0 and points on those axes. Identify the class of its symmetry group in the list of Theorem 3. 2. Determine the symmetry group of the CH4 molecule depicted in Figure 3. Which class in the list of Theorem 3. it is of type Cn ∪ i(Dn \ Cn ). 1.4. then G = H ∪ iH.5. including its improper rotations. according to the list of Theorem 3. unless H is of type Cn .5 does it belong to? Exercise 3. Then if G is cyclic.5.4. List all rotation axes and their orders. but restrict ourselves to the following remarks. but not the inversion.7.6. Suppose that G contains improper rotations. there is only one possibility for G. but not the inversion. In applications. The following procedure. they are axes of rotations of order 3 leaving the cube invariant. Consider the ethane molecule in staggered conﬁguration as depicted in Figure 3.4. What is the type of its symmetry group? Exercise 3.5. can then be helpful.
3. H C H H C H Figure 3.6: The ethane molecule in staggered conﬁguration.molecule. H H . THE FINITE SUBGROUPS OF O(3.4.5: The NH3 . R) 47 N H H H Figure 3.
4. For the sake of compatibility with other texts on crystallographic groups. in our notation. Dm ∪ iDm Cm ∪ i(C2m \ Cm ) Dm ∪ i(D2m \ Dm ) Cm ∪ i(Dm \ Cm ) T. Theorem 3. 3. The proof runs like that of Theorem 3. 2.4.7: The ethane molecule in staggered conﬁguration.11.4. 2. 6 m = 2. There is only one exception to this rule: a trigonal lattice is really hexagonal.48 CHAPTER 3. It is organized as follows: in the ﬁrst column. R) are crystallographic point groups. 3 m = 2. 3 m = 2. see [2]. especially texts on the applications in physics and chemistry. W ∪ iW. W. The last row for each lattice is the full symmetry group of that lattice. 4. there are 230 mutually nonequivalent crystallographic groups in 3 dimensions. Then G is conjugate to one of the following groups: Cm . T ∪ iT. 4. while there are only 32 crystallographic point groups. 4. one ﬁnds lattices classiﬁed up to their symmetry groups. Indeed. we include table 3. Cm ∪ iCm Dm . The second column contains the groups leaving that lattice invariant (but no lattice with a smaller symmetry group). 6 . a crystallographic point group may leave several lattices invariant. Finally. Let G be a crystallographic point group in 3 dimensions. T ∪ i(W \ T ) Like in the twodimensional case. The third and fourth column contain two commonly used sets of symbols for them. m = 1. 3. the question arises which of the ﬁnite subgroups of O(3. 3. 6 m = 1. Most of the above examples are taken from [2].3. SYMMETRY GROUPS IN EUCLIDEAN SPACE P P N P N N P N Figure 3.
2: A Dictionary of Crystallographic Group Names.4. . R) 49 Type of Lattice Triclinic Monoclinic Orthorhombic Trigonal Tetragonal Hexagonal Isometric Our notation C1 C1 ∪ iC1 C2 C1 ∪ i(C2 \ C1 ) C2 ∪ iC2 D2 C2 ∪ i(D2 \ C2 ) D2 ∪ iD2 C3 D3 C3 ∪ iC3 C3 ∪ i(D3 \ C3 ) D3 ∪ iD3 C4 D4 C4 ∪ iC4 C2 ∪ i(C4 \ C2 ) C4 ∪ i(D4 \ C4 ) D4 ∪ i(D4 \ D2 ) D4 ∪ iD4 C6 D6 C6 ∪ iC6 C3 ∪ i(C6 \ C3 ) C6 ∪ i(D6 \ C6 ) D3 ∪ i(D6 \ D3 ) D6 ∪ iD6 T W T ∪ iT T ∪ i(W \ T ) W ∪ iW Schoenﬂies C1 Ci = S2 C2 Cs = C1h C2h D2 = V C2v D2h = Vh C3 D3 C3i = S6 C3v D3d C4 D4 C4h S4 C4v D2d = Vd D4h C6 D6 C6h C3h C6v D3h D6h T O Th Td Wh International 1 1 2 m 2/m 222 mm2 mmm 3 32 3 3m 3m 4 422 4/m 4 4mm 42m 4/m m m 6 622 6/m 6 6mm 6m2 6/m m m 23 432 m3 43m m3m Table 3.3. THE FINITE SUBGROUPS OF O(3.
50 CHAPTER 3. SYMMETRY GROUPS IN EUCLIDEAN SPACE .
then we often write gv instead of ρ(g)v.1.1.1. A (linear) representation of a group G on a vector space V is a homomomorphism ρ : G → GL(V ). In Chapter 2. Deﬁnition 4. we encountered permutation representations. Also. If ρ : G → GL(V ) is a representation of G on V . it is called faithful. The notion We introduce the concept of a linear representation of G. 51 . To experience the use of matrix representations of groups. and similarly for y and z.4. K) is the group of invertible matrices with entries in K. Note that (g. such a map is called matrix representation of G. The dimension of the representation is by deﬁnition the dimension of V . Here ρx stands for rotation over π about the xaxis. and V is called a Gmodule. If ρ is injective. v) → ρ(g)v is indeed an action of G on V in the sense of Deﬁnition 2. Exercise 4.Chapter 4 Representation Theory 4. represent the rotations by matrices and compute the product.2. G is said to act linearly on V . in the present chapter we will start the theory of linear representations. try to guess which orthogonal linear map is the product ρx ρy ρz . if no confusion is possible. we can view a representation as a homomorphism G → GL(n. G denotes a group and K a ﬁeld of characteristic zero.1 Linear representations of groups An abstract group can be represented in various ways. After a choice of basis of V . K).1. All vector spaces will be over K. To check your guess. In this chapter. where GL(n.
ck → (ekl2πi/n ). Then π gives rise to a linear representation Mπ : G → GL(V ). Note that rl is faithful if and only if n and l are coprime.1. K) is a group homomorphism. Recall from 3. It contains the cyclic group Cn of order n with generator c as a subgroup of index 2 and has a reﬂection a (of order 2) with aca = c−1 . . . j ∈ {1. 0 0 1 and (1. . Example 4. extended linearly to all of V . 2. we associate the 5 × 5 matrix Mπ with Mπ ei = eπ(i) . egn }.1. .2) = 1 0 0 . Prove that the correspondence given above between S3 and GL(n. An important example of a linear representation of a ﬁnite group G is the regular representation.1. that is.3 the dihedral group Dn of order 2n. 2) can be represented by the matrix 0 1 0 M(1.1. 2.5. REPRESENTATION THEORY Example 4. Deﬁne ρ : G → GL(V ) by ρ(gi )(egj ) = egi gj .4. 2)(1. R) determined by ck → cos(2kπ/n) − sin(2kπ/n) .3) = 0 0 Exercise 4. .2.2) M(1.3) the permutation matrix presentation of the symmetric group S5 given in Chapter 2: to each permutation π ∈ S5 . n}).8. Let π : G → Sn be a permutation representation.3. 1 0 We know that (1. 3) = (2. . a matrix representation of S3 .1. . . .3) 0 = 1 0 0 0 1 1 0 . 3) can be represented by M(1. . Let G be the cyclic group Cn of order n with generator c. Recall (2. . . The group Dn has a twodimensional real representation r : Dn → GL(2. The above examples can be summarized in a more general construction. .1. . n). gn } and take V to be the vector space of dimension n and basis {eg1 . This is a matrix representation of S5 . Example 4. 0 Example 4. Example 4. in the following way: Mπ (g)ej = eπ(g)j (g ∈ G. Let G = {g1 . It has onedimensional complex representations rl (l ∈ Z) deﬁned by rl : Cn → GL(1.2.6. 3) and this well: 1 M(1. sin(2kπ/n) cos(2kπ/n) a→ 0 1 1 . C). The permutation (1.52 CHAPTER 4. The same thing works for S3 .7. . 0 can be seen from the matrices as 0 0 0 1 . where V is a complex vector space with (formal) basis ei (i = 1. . .
4 is Mπ .3. Show that ρa ∼ ρb if and only if ab = 0. f.1. f : V → K).4. Let r : G → GL(V ) be a linear representation of G on V . Consider the linear representations rl of Cn deﬁned in Example 4. In this case. W ).1. for n = 5. Let R[x]≤3 be the linear space of all polynomials in x of degree ≤ 3. Exercise 4.13. It is straightforward to prove that this really deﬁnes an equivalence relation ∼ on representations. T is an isomorphism of Gmodules. p ∈ R[x]≤3 ).1.1. we saw that V ∗ is a invariant under the action of G on F (V ). Note that F (V ) has the dual space V ∗ of Klinear functions as an invariant subspace. Then r induces the following representation r∗ : G → GL(F (V )): r∗ (g)(f )x = f (r(g)−1 x) (x ∈ V.1. LINEAR REPRESENTATIONS OF GROUPS 53 The regular representation of Example 4. In Example 4. of all maps V → K with pointwise addition (f + g)x = f (x) + g(x) and pointwise scalar multiplication (λf )x = λf (x) (where λ ∈ K.1. A linear map T : V → W is said to intertwine r and s if s(g)T = T r(g) for all g ∈ G. and write (gf )(x) = f (g −1 x) instead.1. where π is the natural permutation representation of Sn (actually. Find a corresponding matrix representation.1. T is called a homomorphism of Gmodules.12. x ∈ V . The matrix representation of Example 4.9. we usually unwind the heavy notation by dropping r wherever possible.1.11.1. Consider the linear representations ρa of Exercise 4.9. This gives rise to the following deﬁnition. we are mainly interested in representations up to isomorphism. Deﬁnition 4. In this case. Let r and s be representations of a group G on vector spaces V and W . For which pairs (l. Exercise 4. 3. Equivalence and Subrepresentations Just as with groups. For a ∈ R we consider the Rlinear representation ρa : R → GL(R[x]≤3 ) deﬁned as ρa (r)p = p(x − ar) (r ∈ R. respectively. The representations r and s are called equivalent if there exists a linear isomorphism T : V → W intertwining them. An important class of constructions of new representations from known ones arises from function space constructions. Consider the vector space F (V ). respectively).10. f ∈ F (V ). The space of all homomorphisms of Gmodules is denoted by HomG (V. Example 4. m) are the representations rl and rm equivalent? Exercise 4. As done before.1. . g ∈ G).7 is Mπ with π : G → Sym(G) the left regular permutation representation of G.10.
14. suppose that n = 2. and consider an element A ∈ GL(V ) whose matrix is a b c d with respect to the basis e1 . . We have A−1 = (ad − bc)−1 so Ax1 (e1 ) = x1 (A−1 e1 ) = Ax1 (e2 ) = x1 (A−1 e2 ) = Ax2 (e1 ) = x2 (A−1 e1 ) = Ax2 (e2 ) = x2 (A−1 e2 ) = (ad − bc)−1 x1 (de1 − ce2 ) = (ad − bc)−1 d (ad − bc)−1 x1 (−be1 + ae2 ) = −(ad − bc)−1 b (ad − bc)−1 x2 (de1 − ce2 ) = −(ad − bc)−1 c (ad − bc)−1 x2 (−be1 + ae2 ) = (ad − bc)−1 a d −b .e. We compute the matrix of A on P (V )1 with respect to x1 . . xn of V ∗ . we leave it out and call W a submodule of V . the set of all elements of F (V ) (see Example 4. Suppose that T : V → W is a Gmodule homomorphism of G on V and W . Example 4. REPRESENTATION THEORY Deﬁnition 4. We ﬁnd that P (V )0 is the onedimensional space of constant functions. . and W is called a subGmodule of V . rW : G → GL(W ) given by rW (g) = r(g)W . is called a subrepresentation of r.54 CHAPTER 4. and let ρ : G → GL(V ) be a representation of a group g. 0) is an invariant subspace. and P (V )1 = V ∗ has dimension n := dim V . −c a . Let r be a representation of a group G on V . In this case.9) that can be written as a Klinear combination of products of elements of V ∗ . . deﬁned by P (V )d = {f ∈ P (V )  f (λx) = λd f (x) for all x ∈ V and λ ∈ K}. Let P (V ) be the set of all polynomial functions on V . . The subspace P (V ) of F (V ) is invariant under G.16. Then ker T is a Ginvariant subspace of V and im T an invariant subspace of W . an ) ∈ Nn satisﬁes a1 + . This implies dim P (V )d = n+d−1 .1. .17. and so are its homogeneous components P (V )d .. x2 . x2 . . .1. Choosing a basis x1 . d prove this! To make all this more explicit. Consider the twodimensional matrix representation ρ : R → GL(R2 ) deﬁned by 1 t Mt = . . e2 dual to x1 . Example 4. .1.1. Let V be a ﬁnitedimensional vector space over K.1. 0 1 Then it is not hard to see that the 1 dimensional subspace spanned by (1. Example 4. If G is clear from the context. + an = d. we ﬁnd that P (V )d is precisely the Klinear span of monomials xa1 · · · xan where n 1 (a1 . i.15. A subspace W is invariant if r(g)W ⊂ W for all g ∈ G.
consider πρ(h) = 1 G ρ(g)π ρ(g −1 h). Let W be any vector space complement of U in V .19. g∈G . K) on K3 sending A to the 3 × 3 matrix just computed. Find the invariant subspaces of the matrix representation ρ : R → GL(R2 ) deﬁned by Mt = et 0 et − 1 . Now 1 2 Ax2 1 A(x1 x2 ) Ax2 2 = = = = (ad − bc)−2 (dx1 − bx2 )2 = (ad − bc)−2 (d2 x2 − 2bdx1 x2 + b2 x2 ) 1 2 (Ax1 )(Ax2 ) = (ad − bc)−2 (dx1 − bx2 )(−cx1 + ax2 ) = (ad − bc)−2 (−cdx2 + (ad + bc)x1 x2 − abx2 ) 1 2 2 (Ax2 ) = (ad − bc)−2 (−cx1 + ax2 )2 = (ad − bc)−2 (c2 x2 − 2acx1 x2 + a2 x2 ). LINEAR REPRESENTATIONS OF GROUPS 55 from which we conclude Ax1 = (ad − bc)−1 (dx1 − bx2 ) and Ax2 = (ad − bc)−1 (−cx1 + ax2 ).1. Deﬁne a new map π := 1 G ρ(g)π ρ(g −1 ) g∈G on V . we have found a matrix representation of GL(2. Can you determine its kernel? Exercise 4.1. Proof.1. Hence A has matrix (ad − bc)−1 d −b −c a on the basis x1 . 1 2 (Ax1 )2 so the matrix sought for is d2 −2 −2bd (ad − bc) b2 −cd c2 ad + bc −2ac . Then there exists a subGmodule W of V such that V = U ⊕ W . and that ρ is a representation of G on a vector space V . 1 Lemma 4.18. Note that this is the tranposed inverse of A. Suppose that G is ﬁnite. and let π be the projection of V onto U with kernel W . Finally. x2 . let us compute the matrix of A on P (V )2 with respect to the basis x2 .4. To see that it does indeed commute with ρ(h). We claim that π is a projection commuting with all ρ(h). x1 x2 . x2 of P (V )1 . Let U be a subGmodule of V . h ∈ G. −ab a2 In conclusion.
both are zero on W and the identity on U .. Now note that im(π) = im(π ) = U . ρ(g)(w)).56 CHAPTER 4.) from an arbitrary one (. the above argument gives an alternative proof of the lemma in the case where V is a ﬁnitedimensional complex vector space. .22 for positive deﬁnite inner products: (vw) = g∈G (ρ(g)(v).) satisfying (ρ(g)(v). then the orthogonal complement U ⊥ of U in V is an invariant subspace of V complementary to U . Now if U is a Ginvariant subspace of V .. ρ(g)(w)) = (v. ρ(g)ππ ρ(g −1 ) g∈G ρ(g)π ρ(g −1 ) g∈G where the second equality follows from the fact that π commutes with each ρ(g). Compute π2 = π 1 G ρ(g)π ρ(g −1 ) g∈G 1 = G 1 = G = π. Thus.2. one can always build an invariant Hermitian inner product (. Suppose that V is endowed with a Hermitian inner product (. This implies that V = im(π)⊕ker(π). indeed. . so that we may take W = ker(π) to conclude the proof of the lemma. we ﬁrst note that ππ = π . as claimed. REPRESENTATION THEORY by the dummy transformation g → hg this equals 1 G = ρ(hg)π ρ((hg)−1 h) g∈G 1 G ρ(h)ρ(g)π ρ(g −1 h−1 h) g∈G = ρ(h) 1 G ρ(g)π ρ(g −1 ) g∈G = ρ(h)π. Remark 4. where both im(π) and ker(π) are subGmodules of V by Exercise 4.15. It is important note that. To verify that π is a projection. w ∈ V. Let ρ be a linear representation of G on a ﬁnitedimensional complex vector space V .1. g ∈ G.1. w) ∀v. .) as was done in Exercise 3. if G is ﬁnite.. This concludes the proof that π is a projection.20.
For ﬁnite groups.10 and 4. Exercise 4. ck → cos(2kπ/n) − sin(2kπ/n) . each irreducible representation is completely reducible. Deﬁnition 4.18 gives an example of a reducible completely reducible representation. irreducible representations are the building blocks of all representations. Show that in Exercises 4. As both dim U and dim W are smaller than n. Exercise 4. .24.19. Suppose that the stament holds for all representations of dimension smaller than n.25.1. Exercise 4. Theorem 4. LINEAR REPRESENTATIONS OF GROUPS 57 Exercise 4. . . then V is called completely reducible. sin(2kπ/n) cos(2kπ/n) Show that it is irreducible for n > 2.18.26.21. and let V be an ndimensional representation.1. there are invariant subspaces that have no invariant complements.23.1. Wl . This terminology is also used for the corresponding representations. The representation of R in Example 4. Proof. .16 is reducible. ⊕ Wl . It has a twodimensional real representation deﬁned by r : Cn → GL(2.27 we shall see that not every reducible representation is completely reducible. Proceed by induction. . A Gmodule V is called reducible if it has subGmodules other than 0 and V . For a ﬁnite group G. and ﬁnd that U is the direct sum of irreducible subGmodules U1 . Exercise 4. Remark 4.1. . Uk of U and W is the direct sum of irreducible subGmodules W1 .1.25 to hold. If we replace R by C. Show that any irreducible representation of an Abelian group G over an algebraically closed ﬁeld is onedimensional. .1. In Exercise 4.1. ⊕ Uk ⊕ W1 ⊕ . U has a invariant complementary subGmodule W . This shows that we need at least some condition on G (like ﬁniteness) for Theorem 4. but it is not the direct sum of irreducibles. . Consider the cyclic group Cn of order n with generator c. we may apply the induction hypothesis. it is called irreducible otherwise. . R).1.1.1.1.1. then we are done. But then V = U1 ⊕ . . By Lemma 4. every ﬁnitedimensional Gmodule is completely reducible. Show that this representation is completely reducible. we obtain a complex representation.1. . By definition. Otherwise there exists a subGmodule U with 0 U V . If V is the direct sum of irreducible subGmodules. .22.4.21 also gives examples of this phenomenon. .1. If V is irreducible.1.
Exercise 4. 2. representation theory of a group G over a ﬁeld of characteristic p tends to be much harder if p divides G than otherwise. Suppose V is a vector space of dimension n with a subspace W of dimension k where 1 < k < n.1. ker(T − λI) equals V . Exercise 4. 0)T is invariant. Exercise 4. Then the embedding r : G → GL(V ) is clearly a faithful reducible representation of G on V .28 (Schur’s Lemma). An important result about irreducible representations known as Schur’s Lemma. Lemma 4. but it has no invariant complement.31. The onedimensional subspace spanned by (1. . Let T : V → V be a homomorphism of Gmodules. In other words.58 CHAPTER 4. Conclude that the converse of Schur’s lemma does not necessarily hold if G is not completely reducible.1. In general. and let V be an irreducible ﬁnitedimensional Gmodule over K.1. Prove that the following converse of Schur’s lemma holds: Let r : G → GL(V ) be a completely reducible representation of a group G. Proof. g ∈ G. Suppose that K is algebraically closed. Let p be a prime.1. The map Cp → GL2 (Z/pZ) deﬁned by ck → 1 0 k 1 is a matrix representation of Cp over Z/pZ. Then any homomorphism of Gmodules V → V is a scalar. For r to be irreducible. and let λ ∈ K be an eigenvalue of T . and consider the cyclic group Cp with generator c.1. Compute the set of all 2 × 2 matrices commuting with Cn in one of the 2dimensional complex representations given in Example 4.1. Conclude that the representation is reducible.6. and it is nonzero as λ is an eigenvalue of T . cf. 3. Then T − λI is also a homomorphism of Gmodules. a matrix commuting with all matrices of an irreducible matrix representation is a multiple of the identity matrix.1.15. 1.30.27. its kernel is a subGmodule of V .1. REPRESENTATION THEORY Exercise 4. and let G be the group of elements in GL(V ) that leave W invariant. Establish that r is not completely reducible. it suﬃces that any linear map A : V → V that commutes with all r(g).29. As V is irreducible. Show that the only Gmodule homomorphisms A : V → V are scalar multiplications.30. is a scalar. Exercise 4. By Example 4.
which we denote by Γ. It√ depicts three point masses (labelled with √ P = {1. 3 1 2 Figure 4.2: A displacement. A displacement of the system is a function f : P → R2 which attaches to each point mass a velocity vector. 1 3).1. The set of all displacements is an Rlinear space. there are 6 degrees of freedom..e. The system can be thought of as vibrating (in two dimensions) about its equilibrium state. i. 3}) at positions (− 1 .2. To be precise. DECOMPOSING DISPLACEMENTS 59 3 1 2 Figure 4.4. (− 1 . and these masses are connected with springs of equal lengths. 4.2. (0. Displacements can be depicted as in Figure 4. − 1 3). We would like to decompose Γ into subspaces in a manner that is consistent with the symmetry group D3 that acts on the system.2 Decomposing Displacements Consider the system of Figure 4. 1) in twodimensional 2 2 2 2 space E = R2 . we . 2.1: A system with 3 point masses and springs. Its dimension is clearly 3 × 2 = 6. in which the point masses form an equilateral triangle.
we shall try to solve this problem ‘by hand’. Theorem 4. the reader will appreciate the elegant character arguments which are to be given in Example 5. without using too much representation theory. 2. 2) corresponds to reﬂection in the line bisecting the edge 12 and meeting point 3. we can write each displacement in a unique way as a purely ‘radial’ displacement and a purely ‘tangental’ displacement (see Figure 4. g ∈ G. such as the one depicted in Figure 4.23.3: The action of (1. 2. To be precise: this space consists of all radial displacements f for which (1. consisting rad of displacements with the ‘same’ velocity vector in all three directions. REPRESENTATION THEORY 3 (1. 2) → −1 0 0 1 and (1.60 CHAPTER 4. respectively.5.2. Here.3) 3 1 2 1 2 Figure 4. have a permutation representation π : D3 → Sym(P ).4). and the spaces Γrad and Γtan of radial and tangental displacements. the permutation (1. and we shall identify D3 with Sym(P ) = S3 . −1 The space Γ becomes a D3 module by setting (gf )(i) = gf (g −1 i). The submodule Γrad contains a onedimensional submodule Γ1 .25 tells us that this is possible. f ∈ Γ. In this way. For example. 3) on a displacement. Our goal can now be formalized as follows: try to write Γ as a direct sum of irreducible submodules. i)(f (1)) = f (i).1. are threedimensional submodules of Γ. The corresponding linear representation D3 → GL(E) is given by (1. The group D3 also acts linearly on E. The map is in fact an isomorphism. . 3) → 1 2 −1 √ − 3 √ 3 . i ∈ P. As a ﬁrst step.2.
3 1 2 Figure 4.2. DECOMPOSING DISPLACEMENTS 61 3 3 1 2 1 2 Figure 4.5: A special radial displacement. .4.4: A radial (left) and a tangental (right) displacement.
and looking at rad the pictures we ﬁnd that the space Γ2 of radial displacements spanned by the rad two in Figure 4. Notes This section is based on a similar (but more diﬃcult) example in [10]. 2)f (1) + f (1) = 0. However.2.6 is also invariant. but this is not the case.2. The space Γ1 must have an invariant complement in Γrad . If your interest is aroused by this example. REPRESENTATION THEORY 3 3 1 2 1 2 Figure 4. you are encouraged to read the pages in that book concerned with the displacemements of a tetrahedral molecule. The representation Γtan can be decomposed in a similar fashion.62 CHAPTER 4. instead of going into details rightaway.23. 3)f (1) + (1. Note that this space consists of all radial displacements for which (1.23.6: Two more special radial displacements. we will ﬁrst treat more representation theory. as we shall see in Example 5. One may wonder if the twodimensional module Γ2 can be decomposed rad further. . and come back to this application in Example 5.
Show that in the case of a permutation representation π : G → Sn . If r is irreducible. The character of r is the function χr deﬁned on G by χr (g) = tr(r(g)) (g ∈ G). we have: 63 . Hence. one has tr(AB) = tr(BA). 5. and all its eigenvalues are nth roots of unity.4.3. then the character χr is also called irreducible. e. the character of r on g is the sum of the eigenvalues of r(g) counted with multiplicities. Deﬁnition 5.1.1 Characters An extremely useful notion regarding matrices is the trace.1. then ρ(g)n = I for all g ∈ G.2.1.5. and ρ : G → GL(V ) is a ﬁnitedimensional representation. over the algebraic closure of K. If G has ﬁnite order n. The trace of a matrix A = (aij )n×n is tr(A) = a11 + · · · + ann . For any complex representation r : G → GL(V ) of a ﬁnite group.Chapter 5 Character Tables In this chapter. the number tr(Mπ (g)) (see Example 4. each ρ(g) is diagonalizable. We next give some properties of complex characters. Prove that for two matrices A and B.1.1.1. Let r be a representation of a group G on a ﬁnitedimensional vector space V . Exercise 5. C.g.1.8) is equal to the number of ﬁxed points of the permutation π(g). Remark 5. If K = C. Exercise 5. G denotes a ﬁnite group and K an algebraically closed ﬁeld of characteristic 0. Thus. then all eigenvalues have absolute value 1. Lemma 5.
6. . that is ﬁxed by the linear action of G on CG deﬁned by (g · f )(h) := f (ghg −1 ).2. equivalently.1. χr (e) = tr(I) = dim V . Theorem 5. . χr (g −1 ) = χr (g). as we have seen in Example 4. The Hermitian inner product is given by 1 χ(g)ψ(g) (χ  ψ) = G g∈G for χ.1. λn of r(g) have absolute value 1. But. Let V1 . .1. The space of all class functions is denoted by H. 5.2 Orthogonality of irreducible characters The following theorem is very important in the representation theory of ﬁnite groups. . with respect to the Hermitian inner product (·  ·). CHAPTER 5.1. The last part states that characters are constant on conjugacy classes of G. .1. χ1 . which in this case is a representation (often called the natural representation).17. As for the second part. . The trace of an arbitrary element A= a b c d is a+d. .2. χr (ghg −1 ) = χr (h) for all g. χs . Then the character of the representation V1 ⊕ · · · ⊕ Vs is the sum of characters χ1 + · · · + χs . or. Example 5. and (a2 + ad + bc + d2 )/(ad − bc)2 on P (V )2 . K). . ψ ∈ H. Take G = GL(2. . The last part follows directly from Exercise 5. We need a deﬁnition to make clear what this theorem states. there are also linear actions on P (V )1 and on P (V )2 . . The ﬁrst part is easy. . Vs be representations of G with characters respectively.17. .64 1.7. We read oﬀ the corresponding characters as traces of the matrices found in 4. Let G be a ﬁnite group. A class function on a group G is a function f : G → C that is constant on conjugacy classes. They are (a + d)/(ad − bc) on P (V )1 . 2. This is the character of the identity. Deﬁnition 5. Therefore: χr (g −1 ) = tr(r(g −1 )) = tr(r(g)−1 ) = λ−1 = i λi = tr(r(g)) = χr (g).1. Lemma 5. Proof.2. CHARACTER TABLES 3. h ∈ G. .1. we have seen that all the eigenvalues λ1 . The irreducible complex characters of G form an orthonormal basis for the space of class functions on G.
5. Let V and W be vector spaces over K. we introduce the tensor product of two representations.2.2. and v ⊗ (λw) − λ(v ⊗ w). ⊗ vπ(d)  π ∈ Sd } is denoted by S d (V ). . ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 65 We have noticed earlier that characters are class functions. Exercise 5. For our ﬁrst purpose. We divide the proof of Theorem 5.1 into two parts: ﬁrst. Exercise 5. .1.17). the direct sum of all S d (V ) for d ∈ N forms an associative algebra. mn ∈ N.6. Just like T (V ). Then V ⊗ W . . n} is a basis of V ⊗ W . . is the quotient of the free vector space over K with basis {v ⊗ w  v ∈ V. w. (λv) ⊗ w − λ(v ⊗ w). . respectively. . . . this algebra is called the tensor algebra on V and denoted by T (V ). Then the quotient of T d (V ) by the subspace spanned by the set {v1 ⊗ v2 ⊗ . . . then the spaces (U ⊗ V ) ⊗ W and U ⊗ (V ⊗ W ) are canonically isomorphic. wn are bases of V and W . The direct sum of all T d (V ) for d ∈ N (including 0. We will identify them. . . . Deﬁnition 5.2. then {vi ⊗ wj  i = 1. called the symmetric algebra on V and denoted by S(V ). . w2 ∈ W and λ ∈ K): (v1 + v2 ) ⊗ w − v1 ⊗ w − v2 ⊗ w. v1 . V. w1 . w ∈ W } by its subspace spanned by all elements of the following forms (for v. .2. . .2. Prove that S d (V ∗ ) is canonically isomorphic to P (V )d (see Example 4.4. Show that if v1 . . v ⊗ (w1 + w2 ) − v ⊗ w1 − v ⊗ w2 . + mn = d} is a basis for S d (V ). It is not hard to see that if U. vn is a basis of V . m1 + . Let V be a vector space over K and let d be a natural number. vm and w1 . j = 1. . . m. . . . . and called the dth symmetric power of V . Show that if v1 . . . and then their being a full basis of the space of class functions. and to stress this fact we usually leave out the operator ⊗ from expressions in S(V ). and write T d (V ) for the dfold tensor product of V with itself. called the tensor product of V and W . we will show the orthonormality of the irreducible characters. v2 ∈ V . leave out parentheses. ⊗ vd − vπ(1) ⊗ vπ(2) ⊗ . . Deﬁnition 5. .5.2. then the set m m {v1 1 · · · vn n  m1 . . . This algebra is commutative. . for which T d (V ) is isomorphic to K) forms an associative algebra with respect to the operator ⊗.3. and W are vector spaces over K.
7.j aij. Deﬁnition 5. the lefthand side equals (r(g)vk ) ⊗ (s(g)wl ) = ( i rik vi ) ⊗ ( j sjl wj ) = i.2) Comparing the righthand sides of equations (5. and consider the linear isomorphism σ : V ⊗ V → V ⊗ V determined by σ(x ⊗ y) = y ⊗ x for x. This representation is denoted by r ⊗ s and called the tensor product of r and s. (5. for a ﬁxed g ∈ G. Let G.66 CHAPTER 5.2.j rik sjl vi ⊗ wj .j aij. Then we write V ∗ for the dual of V (in the notation of Example 4. s. with eigenvalues 1 and −1. the matrix A = (aij. V. Show that σ has precisely two eigenspaces. this is P (V )1 ).kl ) of (r ⊗ s)(g) with respect to the basis vi ⊗ wj of V ⊗ W .7. Consider. Let r : G → GL(V ) and s : G → GL(W ) be linear representations of G. Another important ingredient in our proof of Theorem 5. Lemma 5.2) we obtain aij. . y ∈ V .8. . Fix bases v1 . .2. Proof. Exercise 5. and W be as in Deﬁnition 5.kl vi ⊗ wj . respectively. Let r : G → GL(V ) be a representation of the group G.9. r. wn of V and W .2. CHARACTER TABLES Deﬁnition 5. and therefore χs⊗r (g) = i. and r∗ for the corresponding representation. extended linearly to all of V ⊗ W .kl = rik sjl .1) and (5. . .2. Then we deﬁne a linear representation r ⊗ s : G → GL(V ⊗ W ) of G by (r ⊗ s)(g)(v ⊗ w) := (r(g)v) ⊗ (s(g)w). concluding the proof. by the deﬁnition of the tensor representation. given by (r∗ (g)f )(v) := f (r(g)−1 v). We have (r ⊗ s)(g)(vk ⊗ wl ) = i. .2. . (5.j rii sjj = χr (g)χs (g).2.17.1 is the socalled dual representation. Here G acts on V ⊗ V via the tensor product representation of the linear representation of G on V (given by the Gmodule structure of V ) with itself. vm and w1 . Then the character of r ⊗ s satisﬁes χr⊗s = χr χs . . .1.1) On the other hand.10. and that both spaces are Ginvariant. Let V be a Gmodule.ij = i.
s.7 and t as in Lemma 5. of which the tensor product is a very useful tool in its own right.12 is really an intertwining map. v ∈ V. W ).12. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS Exercise 5. W. Proof. assume that both representations r and s are irreducible. so we leave it as an exercise for the reader. W ) by P = The trace of P equals 1 G χt (g) = g∈G 1 G t(g).1. f ∈ V ∗ . the space of all Clinear maps from V to W by t(g)H := s(g)Hr(g)−1 . G be as in Deﬁnition 5.2.2. . It is a matter of elementary linear algebra to see that T is a linear isomorphism. Let r. It can be written in a particularly nice form. This representation is equivalent to s ⊗ r∗ . using tensor products.1. We wish to compute the inner product (χs  χr ). Prove that the map T deﬁned in the proof of Lemma 5. Deﬁne a representation t of G in Hom(V. we proceed with the ﬁrst part of the proof of this section’s main theorem. Exercise 5. ∗ 67 Lemma 5.2. w ∈ W.2.2. W ) by T (w ⊗ f )(v) = f (v) · w. The best way to see this is to do the computation yourself. V. Having completed all preparations.5. H ∈ Hom(V. W. Let r.11. and a basis (wi )i of W . One ﬁnal lemma before proceeding with the proof of Theorem 5. G be as in Deﬁnition 5.2. Deﬁne the linear map P : Hom(V.2. V.2. and zeroes elsewhere) with respect to the bases (vj )j and (wi )i .7. Show that χr = χr .2. Deﬁne T : W ⊗ V ∗ → Hom(V.2. ﬁrst part. We claim that it intertwines the representations t and s ⊗ r∗ . and extend it linearly. s.12. W ) → Hom(V. and moreover.13. If we choose a basis (vj )j of V with dual basis (v j )j of V ∗ . Proof of Theorem 5. then the map T sends wi ⊗v j to the linear map V → W having the matrix Eij (the matrix with 1 on position (i. g∈G 1 G χs (g)χr (g) g∈G ∗ 1 = G s χs (g)χr (g)) g∈G = (χ  χr ). W ). j).
As we only wish to compute the inner product of their characters.2. suppose that r and s are equivalent. W ). s(g)s(g −1 )H g so that im(P ) = HomG (V. and compute (P H)(g0 v) = 1 G s(g)Hr(g −1 g0 )v g −1 s(g0 g)Hr(g −1 g0 g0 )v g 1 = G = s(g0 ) 1 G s(g)Hr(g −1 )v g = g0 (P Hv).2. a fortiori. W ). This proves the claim. suppose that r and s are not equivalent. W ) is zerodimensional.8 and Exercise 5. Alternatively. we may assume that they are . To see that im(P ) is contained in HomG (V. so that P and.2.12. W ) onto the space HomG (V. s(h)Hr(h−1 ) h g where in third equality the dummy h is replaced by g −1 h. 5. W ). let H ∈ Hom(V. We now distinguish two cases.68 CHAPTER 5. To show that it is indeed a projection. This proves that P is a projection. the space HomG (V. Finally. CHARACTER TABLES Here we used Lemmas 5. where in the second equality. W ). Then. let g0 ∈ G. its trace are 0. by Schur’s lemma. W ). First. W ). In this case (χs  χr ) = 0. the dummy g is replaced by g0 g. then PH = 1 G s(g)Hr(g −1 ) g 1 = G = H. H ∈ Hom(V. and compute P 2H = = 1 G 1 G 1 G 1 G 1 G PH g s(g)s(h)Hr(h−1 )r(g −1 ) h g s(gh)Hr((gh)−1 ) h g 1 = G 1 = G = P H. if H ∈ HomG (V. We claim that P is a projection from Hom(V. and v ∈ V .11.
Proof. The number of nonisomorphic irreducible representations of G is ﬁnite. and write V = V1 ⊕ · · · ⊕ Vt . The dimension of H is the number of conjugacy classes of G. Let r be a complex representation of G on a space V . where the Vi are irreducible submodules. Two representations of a ﬁnite group are equivalent if and only if they have the same character. Proposition 5.5.18. If χi is the character of Vi then the character χr of V can be written as χr = χ1 + · · · + χt .e G (g ∈ G). Since P I = I. The character of the regular representation ρ of G satisﬁes ρ(g) = δg. . Lemma 5. Proof.2. This proves that the characters are an orthonormal system in the space of class functions on G.2. Let s be an irreducible representation of G on a space W . Corollary 5.14. the transformation ρ(e) ﬁxes each element g of G. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 69 equal: W = V and s = r. the image of P is the linear span of the identity map I on V . Then the number of i for which Vi is isomorphic to W is equal to (χr χs ). If g = e. then eh = egh for each h ∈ G.2. In particular. The only contribution to the trace of P is therefore the element in the upper left corner. this number does not depend on the decomposition in irreducibles. Hence (χr  χr ) = 1. and zero otherwise. with V the vector space of dimension G and basis {eg  g ∈ G}.17. Taking any basis of Hom(V. that element is 1. Since eg = g. Proof.2. tr(ρ(e)) = n = G.16. Therefore. Lemma 5.15.2. Hence (χr χs ) = (χ1 χs ) + · · · + (χt χs ). Recall that the regular representation ρ : G → GL(V ) of the group G is deﬁned by ρ(g)(eh ) = egh . The irreducible characters form an orthonormal. whence independent. and by the above (χi χs ) is one if Vi ∼ W . Every irreducible representation W of G is a subrepresentation of the regular representation with multiplicity dim W . = Corollary 5. This implies that tr(ρ(g)) = 0.2. we see that the matrix of P with respect to that basis has only nonzero elements on the ﬁrst row. V ) that starts with I. In this case. set in this ﬁnitedimensional space by the above.
Then Corollary 5. ⊕ rs ⊕ · · · ⊕ rs .20. and let f ∈ H. Then the linear map H := equals 1 (f  χr )I. . . . . Then G = n2 i where ni is the dimension of the representation ri .19. Suppose that r : G → GL(V ) is an irreducible ndimensional complex representation of G.2.19 can be used to decide whether we have already ﬁnd all. Moreover.1.15. . ⊕ r1 ⊕ .2. as r is irreducible.17 we have (χρ  χs ) = 1 G χρ (h)χs (h) = h∈G 1 Gχs (e) = dim(W ). each of these representations appears ni times in the decomposition. it equals 1 G Consequently. n Proof. .2. note that H veriﬁes r(g)H = Hr(g) for all g ∈ G. rs be the irreducible representations of G. . i Suppose that we manage to construct some nonisomorphic irreducible representations of G. CHARACTER TABLES Proof.2. From Proposition 5. Therefore G = χρ (e) = s i=1 ni χri (e) = s i=1 n2 . r2 . g∈G 1 G f (g)r(g) g∈G . we still need to show that the irreducible characters span H. Its trace equals nλ. On the other hand. G and the result follows from Lemma 5. We saw in the lemma that the regular representation ρ can be written as a sum of all irreducible representations ri of G. f (g)χr (g) = (f  χr ). First of all. Let r1 . Hence. Corollary 5. H must be multiplication by a scalar λI. . So n1 times ns times ρ = r1 ⊕ .2. λ = (f  χr )/n.2.70 CHAPTER 5. Proof. In order to ﬁnish the proof of Theorem 5. Lemma 5.
Hence. Theorem 5. .1. and let s : D3 → GL(V ) be the linear representation corresponding to the permutation representation of D3 on {1.2 the D3 module Γ of displacements.2.2. suppose that ni = 1 for all i. all conjugacy classes consist of a single element of G.1.21. According to Lemma 5.24. But. Recall from Section 4. Proof. Half of the following proposition is Exercise 4. G is Abelian.2.20. it suﬃces to prove that the orthoplement of their linear span in H is zero. Hence by Corollary 5. and χs ((1. Example 5. writing an arbitrary representation as a direct sum of irreducible ones. G G g∈G g∈G Thus. Then G = s and so there are G distinct conjugacy classes of G. 3} (see Example 4. Then by the theorem above.8). If G is Abelian. second part.2.2. This settles one implication. (12) and (123). The number of irreducible characters equals the number of conjugacy classes of G. 2)) = 1. Then χs ((1)) = 3. The group G is Abelian if and only if all of its irreducible representations are one dimensional.19. there are G irreducible representations of G.2. 3)) = 0. Let V = R3 . As for the converse. χs ((1. for H to be zero.1. 2.5. with representatives (1).2. ORTHOGONALITY OF IRREDUCIBLE CHARACTERS 71 Proof of Theorem 5.21. The dimension of H equals the number of conjugacy classes of G. let f ∈ H be orthogonal to all irreducible characters. Denote by s the number of irreducible characters of G and denote their dimensions. and the irreducible characters form a basis of this linear space. note that D3 has three conjugacy classes. First of all. Let us apply this knowledge to the regular representation. 2.23. all ni ’s are equal to 1. Let us employ character theory to ﬁnd the decomposition of Γ into irreducible modules. Therefore. Proposition 5. This concludes the proof.2. we see that this linear map is zero for any representation r. f must be identically zero. Hence D3 has three distinct irreducible characters by Theorem 5.22. and can be done without character theory. as before by ni . the linear map H= 1 G f (g)r(g) g∈G is zero for all irreducible representations r. In order to prove that the irreducible characters form a basis of H. Proof. Compute 1 1 Heh = f (g)r(g)eh = f (g)egh .
that is. 2.8 shows that χr = χs χr2 .12. χr2 ((1. Lemma 5.3 Character tables Let χ1 . as can be read of from the traces of the matrices on page 4. χr3 = sgn . χs and columns labelled by g1 . We conclude that the character χr of D3 on Γ decomposes into χr1 + 2χr2 + χr3 . E) is isomorphic to V ∗ ⊗E.2. under each gi we write the size of Ci .2. . 3)) = −1. j)entry of the table is χi (gj ). so they are irreducible. An element of Γ can be extended in a unique way to a linear map V → E. . gs . (23)} consisting of all transpositions and C3 = {(123). so that (χr2 )2 − χr1 − χr2 is the character of a third irreducible representation. . C2 . and χr2 ((1. and χr ((1. By convention. Construct a table with rows labelled by χ1 . and also to take χ1 to be the trivial representation. Cs be the distinct conjugacy classes of G. of dimension 1. we know that there are three irreducible representations. χr ((1)) = 6. . . sgn . 3)) = 0. Hence. .2. Recall this is the one that assigns to each cycle its sign . By Theorem 5. so that V is reducible. . First notice that there are three conjugacy classes: C1 = e. By Theorem 5. Then χr1 ((1)) = 1.2. r2 exhaust all irreducible characters of D3 . The (i. We have (χr1 (χr2 )2 ) = 1 and also (χr2 (χr2 )2 ) = 1. . (132)} consisting of all 3cycles. so that the ﬁrst row contains only ones. χ2 . so that we may identify Γ with the space HomK (V. and χ(r3 ) (123) = 1.21. with representatives g1 . . we can decompose χr = (χr1 + χr2 )χr2 = χr1 χr2 + (χr2 )2 = χr2 + (χr2 )2 . the characters r1 . The representation of D3 on E is equivalent to r2 . . 3)) = 1 and χr2 ((1)) = 2. . Its character equals χr3 ((1)) = 1.72 CHAPTER 5. χ1 is the trivial character. CHARACTER TABLES We have (χs χs ) = (9 + 3 ∗ 1)/6 = 2. . 1) ∈ V is invariant. respectively. By Lemma 5. 1. C2 = {(12). χr3 ((12)) = −1. χr1 ((1. say r3 . It is common to take C1 = {e} = {g1 }. . . This table is called the character table of G. . . χs be the characters of the distinct irreducible representations of G. χr ((1. .21.1. Indeed. The character (χr2 )2 has squared norm 3. 5. 2. For the second one we take the sign representation. Let r1 be the representation of D3 on this line. gs . let r : D3 → V ∗ ⊗E be the corresponding representation. E) of Klinear maps from V to E. . . Both χr1 and χr2 have squared norm 1. and let r2 be the representation of D3 on an invariant complement. the line spanned by (1. g2 . so that the ﬁrst column contains the dimensions of the respective characters. 2)) = 0. and χr1 ((1. (13). Let us check the character table of S3 . First. Let C1 . 2. Example 5. χ2 . 2)) = 1. so it is reducible. HomK (V.2.3. . 2)) = 0.
Finally for the third representation. we ﬁnd (χχ) = 1 n!  ﬁx(σ)2 . Since it is a 1dimensional representation. .3. to this eﬀect. k where F (m) denotes the number of ﬁxedpointfree permutations in Sm . n}. Proof. This number is clearly equal to n F (n − k). it suﬃces to prove that (χχ) = 2. Denoting by ﬁx(σ) the set of ﬁxed points of σ ∈ Sn on {1.2) We have found the complete character table. Its orthogonal complement is also an invariant subspace. The permutation representation π of S3 on C3 has an invariant subspace U spanned by (1. i! . To ﬁnd its character.8). we proceed as follows. The latter representation has dimension 2 and is irreducible. it is equal to its character. But U is just the trivial representation and χπ is equal to 3 on C1 . . S3 trivial U sign standard V (1) 1 1 1 2 (12) 3 1 −1 0 (123) 2 1 1 −1 Proposition 5. We wish to show that χ − χ1 is irreducible.2. The standard representation of Sn is irreducible for all n > 1. z3 ) ∈ C3 such that z1 + z2 + z3 = 0. say V . 3cycles are even). 1. . .5. z2 . we have χ(σ) =  ﬁx(σ). . it is called the standard representation of S3 . σ∈Sn To evaluate the righthand side of this equation we count the number of permutations having precisely k ﬁxed points (0 ≤ k ≤ n). A standard inclusionexclusion argument shows that m F (m) = i=0 (−1)i m! . and let χ1 be the trivial character of Sn . . . it consists of all (z1 .3. n} (see Example 4. 1 on C2 and 0 on C3 (why? see Exercise 5. we observe that χV = χπ − χU . Hence. 1).1.1. Let χ be the character of the ndimensional linear representation corresponding to the permutation representation of Sn on {1. CHARACTER TABLES 73 (2cycles are odd. .
. we note that for n ≥ 3 we have: n Mn − Mn−1 = i=0 (n − i)2 (−1)i 1 . by Theorem 5. so that the character table of Cn is as follows.. CHARACTER TABLES Combining the above considerations we ﬁnd (χχ) = 1 n! n n k2 k=0 n k n−k n−k (−1)i i=0 (n − k)! i! = k=0 k2 1 k! i=0 (−1)i i! =: Mn .3... (n − i)!i! and it is not hard to prove that the righthand side is zero for n ≥ 3. Consider the cyclic group Cn with generator c. the 2cycles. and let ζ ∈ C be a primitive nth root of unity. . . cn−1 1 1 ζ n−1 ζ 2(n−1) · · · (n−1)2 ζ Example 5..... It is readily seen that M2 = 2. To compute the character table of S4 .3.. .4. S4 χ1 (trivial) χ2 (sign) χ4 (standard) (1) 1 1 1 3 (12) 6 1 −1 1 (123) (1234) (12)(34) 8 6 3 1 1 1 1 −1 1 0 −1 −1 ...21 there are no other characters.. They are listed in the following partial character table. . . S4 has ﬁve distinct irreducible characters.21.3. . ﬁve in total. By Theorem 5. . Hence.3. for j = 0. the sign and the standard character.. They are the identity. Example 5. the 3cycles. indeed. and irreducible as they are onedimensional.2. To prove that Mn = 2 for all n ≥ 2. the map χj : Cn → C by χj (ck ) := ζ jk . the 4cycles and the product of two distinct 2cycles. Deﬁne.. see Proposition 5. we ﬁrst count the number of conjugacy classes. .2). namely the trivial character. . n − 1.74 CHAPTER 5.. . Cn e 1 χ1 1 χ2 1 χ3 1 · · · · · · χn 1 c 1 1 ζ ζ2 · · · ζ n−1 c2 1 1 ζ2 ζ4 · · · ζ 2(n−1) .. It has three of them in common with S3 (or.2... . Then the χj are characters of Cn .. with any symmetric group.
Hence. that is. the complex number r((234)) should have multiplicative order 3. 2. Compute the number of elements in each conjugacy class of S4 . 2.5.6. Consider the subgroup T of the orthogonal group SO3 (R) consisting of the proper rotations of the tetrahedron. Exercise 5. the squares of the dimensions of the irreducible characters sum up to S4  = 24. .9.3.19. We wish to compute its character table. The last line can be obtained from the orthonormality conditions. (234). hence it is either ω = e2πi/3 or ω 2 .3.3. 4}. The remaining 2dimensional character χ3 can be found using the orthogonality relations among characters. CHARACTER TABLES 75 By Lemma 5. which is irreducible as it has norm 1. A ∩ B = ∅}. 3. Construct the character table of D4 . apart from the trivial character. Their turns out to be only one solution. and their cardinalities are 1. 4} into two sets of size 2. 4 and 3. respectively. of representations with dimensions m1 ≤ m2 ≤ m3 ≤ m4 . there must be 4 irreducible characters. Let r be one of them. there should be two more characters of degree 1.7.8. A ∪ B = {1. The image of (243) is ﬁxed in both cases. this explains the character χ3 . B}  A = B = 2. 4. A more or less explicit description of the irreducible representations of Sn is known for general n.3. namely m1 = m2 = m3 = 1 and m4 = 3. The sum of their squares should equal T  = 12.3. It involves the beautiful combinatorics of Young tableaux.5. as χr should be orthogonal to the trivial representation. (243) and (13)(24) (make a picture that corresponds to this numbering). a good textbook on which is Fulton’s book [5].2. Consider the permutation representation of S4 on partitions of {1. 3. The conjugacy classes are those of (1). We conclude that the two remaining irreducible characters have dimensions 2 and 3.1. Compute the character table of the Klein group (Z/2Z) ⊕ (Z/2Z). and we only have 1 + 1 + 9 = 11 so far. and so is the image of (13)(24). Exercise 5. Exercise 5. This explains the ﬁrst three lines in Table 5. We then ﬁnd the following character table of S4 . The latter is the product χ5 := χ2 χ4 . As r is nothing but a homomorphism from T to C∗ . Thus. the action of S4 on the set {{A. S4 χ1 χ2 χ3 χ4 χ5 = χ2 χ4 (1) 1 1 1 2 3 3 (12) 6 1 −1 0 1 −1 (123) (1234) (12)(34) 8 6 3 1 1 1 1 −1 1 −1 0 2 0 −1 −1 0 1 −1 Exercise 5. Example 5.3. Show that the character of the corresponding linear representation is χ1 + χ3 .
3. 2 What is the character of G on S 3 (V )? Use these results to ﬁnd new characters of A5 from the irreducible 4dimensional character occurring in the standard permutation representation of degree 5.12. Consider the alternating group A5 on 5 letters. s χi (g)χi (h) = CG (h)δg. Exercise 5. Exercise 5. h ∈ G. and χ1 . Let G be a group and let H be a normal subgroup of G. Determine the sizes of the corresponding conjugacy classes.3.3. . (Hint: if there were one more. .h . Let χ be the character of a complex linear representation of G on V . Show that A5 has only one irreducible character of degree 1. which has the Klein 4group as a normal subgroup. Show that the character of G on the symmetric square S 2 (V ) of V equals χ(g)2 + χ(g 2 ) g→ . Let r be a representation of G with the property that H ⊆ ker r. any representation of G/H can be lifted to one on G.11. Show that r induces a representation of G/H. Prove the column orthogonality of the character table: for g.) . Find a representation corresponding to the last line of Table 5.3. Exercise 5.3. 3. 1. χs a complete set of irreducible characters of G. Write down representatives for each of its conjugacy classes.13. Exercise 5.76 CHAPTER 5. .10. Apply this principle to the group T . 2.1. i=1 This formula is also useful for the construction of character tables. Conversely. CHARACTER TABLES T χ1 χ2 χ3 χ4 (1) 1 1 1 1 3 (234) 4 1 ω ω2 0 (243) 4 1 ω2 ω 0 (13)(24) 3 1 1 1 −1 Table 5. .14.1: The character table of the tetrahedral group Exercise 5. then there would have to be a normal subgroup N such that A5 /N is abelian.
and compute or look up its character table.2. corresponding to irreducible submodules of the representation of the symmetry group of the molecule on the space of all displacements. 5. and the intensity of the light leaving the substance is measured. which allows to observe all irreducibles that occur in the symmetric square of the aforementioned natural representation. 6. only those irreducibles occurring also in the natural 3dimensional representation do actually contribute a peak in the spectrum. They have identical character tables.15. giving rise to a (downward) peak in the intensity (plotted as a function of the frequency). If the radiation excites the molecules from the ground state into vibration. This distinguishes IR from Raman spectroscopy. to predict the number of peaks using either of the above methods. 2) gives a nonequivalent representation and determine its character. Hence. In this section. we describe two of them: infrared (IR) spectroscopy and Raman spectroscopy. not even all irreducible modules in the module of vibrations can be observed by IR spectroscopy. As we have seen in Section 4. The rotation group of the icosahedron is A5 ..3. and compare it with the character of the restriction of the standard representation of S5 to A5 .5. 7. there are precisely two groups of order 8. Show that this character splits into the trivial character and an irreducible 5dimensional character. Exercise 5. 1. 2. Exhibit these groups and their character tables. Write down its character. we proceed as follows. the (inﬁnitesimal) isometries of threedimensional space do not contribute.4. As with Raman spectroscopy—another technique whose physical details we do not discuss here—only actual vibrations may contribute a peak to the intensity spectrum. as well as its character χiso on the 6dimensional . Compute this character. This gives an irreducible representation ρ of A5 of degree 3. for a given molecule with n molecules.4 Application: symmetry of vibrations There exist many spectroscopic techniques in chemistry that can be used to measure symmetry properties of molecules of a given substance. Up to isomorphism.13 to ρ to ﬁnd a 6dimensional character. Apply the symmetric square formula of Exercise 5. However. the substance is exposed to radiation of various frequencies. APPLICATION: SYMMETRY OF VIBRATIONS 77 4. due to technical properties of this technique. Verify that the composition of ρ with conjugation by (1. In the former method. there are several essentially diﬀerent vibrations.e.3. then it is absorbed. Determine the symmetry group G of the molecule at hand. Determine the character χdis of G on the 3ndimensional module of displacements of the molecule. i. 5. Conclude from the orthogonality of the character table that there is one more irreducible character.
1. CHARACTER TABLES Cl Cl P Cl Cl Cl Figure 5. Compute the character χ1 of G on R3 . it is common to use the Schoenﬂies notation (see Table 3. 1.78 CHAPTER 5. and S3 is the composition of a C3 and a σh ). Example 5. respectively.1: The PCl5 molecule submodule of (inﬁnitesimal) isometries of R3 .2). and we will adhere to this tradition in the present example. The symmetry group of this molecule is D3h . Then the number of peaks in IR spectroscopy is the number of irreducible characters in χvib (counted with multiplicity) that also occur in χ1 . 3. 4. and and the number of peaks in Raman spectroscopy is the number of irreducible characters in χvib (counted with multiplicity) that also occur in χ2 . Ci stands for a rotation of order i. σh and σv are reﬂections in a horizontal and vertical plane. In these chemical applications. We want to predict the number of peaks in IR and Raman spectroscopy when applied to the molecule PCl5 of Figure 5. Then χvib := χdis − χiso is the character of G on vibrations of the molecule. . whose character table is as follows (E is the identity.1. and the character χ2 of G on the symmetric square of R3 .4.
Using Raman spectroscopy. Having only real values.23.5. We conclude that χvib = 2A1 + 3E + 2A2 + E . and the character χ4 on rotations turns out to be as follows. 4. χ3 is selfdual.2. D3h χ4 E 3 C3 0 3C2 1 σh 1 S3 2 σv 1 We ﬁnd that χ4 = A2 + E . Let χ1 denote the character of D3h in its natural representation on R3 .4.1. D3h χ1 χ3 E 3 6 C3 0 3 3C2 1 2 σh 1 4 S3 2 1 σv 1 4 As in Example 5.13 we ﬁnd the following values for its symmetric square χ2 . while χ1 = E + A2 .3. we conclude that only vibrations with characters E or A2 are observable with IR spectroscopy. 3. E and E can be observed. so that is the predicted number of peaks when using this technique. D3h χdis E 18 C3 0 3C2 2 σh 4 S3 2 σv 4 This character can be decomposed into irreducible ones as follows: χdis = 2A1 + A2 + 4E + 3A2 + 2E The character χiso is the sum of the characters of D3h on translations and on rotations. The character of D3h on translations equals χ3 = E + A2 . D3h χ2 E 6 C3 0 3C2 2 σh 2 S3 2 σv 2 We ﬁnd that χ2 = 2A1 + E + E . . the number of such irreducible characters in χvib is 5.8). We have already computed χ1 . so that we ﬁnd the following values for χdis . From the above. and let χ3 denote the character of D3h corresponding to its permutation representation on the 6 molecules of PCl5 (see Example 4. this leads to 6 peaks. APPLICATION: SYMMETRY OF VIBRATIONS D3h A1 A2 E A1 A2 E E 1 1 1 2 1 1 2 C3 2 1 1 1 1 1 1 3C2 3 1 1 0 1 1 0 σh 1 1 1 2 1 1 2 S3 2 1 1 1 1 1 1 σv 3 1 1 0 1 1 0 79 2. Their values are easily calculated to be as follows. only vibrations with characters A1 . Using Exercise 5. the character of the module of displacements equals the dual of χ3 times χ1 .
Is it possible to deduce the symmetry of Ni(CO)4 (tetrahedron or square plane) from the IR and/or Raman spectra of the COstretches? Exercise 5. Which spaces fuse into irreducibles for the group of all (proper or improper) symmetries of T (of order 48)? 5.4.5 Notes Most of the content of this chapter can be found in Serre’s book [9].2. CHARACTER TABLES N Pd Cl N Cl cis−PdCl (NH ) 2 3 2 trans−PdCl (NH ) 2 3 2 Figure 5.2: Two conﬁgurations of PdCl2 (NH3 )2 The following exercises shows how spectroscopic techniques may or may not be used to recognize molecules.4.80 N Pd Cl Cl N CHAPTER 5. The last section is entirely based on Theo Beelen’s lectures in the course for which these notes were prepared.4. Let T be the octahedron whose 6 vertices are joined by strings. Determine the decomposition of the 18dimensional space of vibrations of irreducibles of the group of rotational symmetries of T (of order 24). Exercise 5.2. The square planar coordination compound PdCl2 (NH3 )2 can be prepared with two diﬀerent conﬁgurations. Is it possible to discern the cis. .and the transconﬁgurations using information concerning Pd − Cl stretch vibrations in IR spectroscopy? Exercise 5. Interpret the result.3. see Figure 5.4.
on which in fact most of this chapter is based. g ∗ = (g ji )ij if g = (gij )ij . U(n) := {g ∈ Mn (C)  g ∗ g = I}. GL(n. 3.Chapter 6 Some compact Lie groups and their representations In the previous two chapters we got acquainted with the representation theory of ﬁnite groups. the existence of an invariant measure.e.1 Some examples of Lie groups One can think of a Lie group G as a smooth surface embedded in some Euclidean space. 6. i.. 5. Some general. R). O(n. without going into the details of deﬁning a Lie group and showing. Examples. G carries the structure of a group in such a way that the multiplication map (g. 4. SO(n. R). In this chapter we will treat some of the representation theory of compact Lie groups. less detailed remarks can be found in [9].. Then the corresponding unitary 81 . A good reference. SU(n) := {g ∈ U(n)  det(g) = 1}. SL(n. are: 1. G×G → G and the inverse map g → g −1 . Moreover. for example. G → G are smooth. R). . h) → g ·h. where g ∗ stands for the conjugate transpose of g. R). 6. is [1]. A more intrinsic deﬁnition is the following: let (.) be a Hermitian inner product on a ﬁnitedimensional complex vector space V . 2. some of which we already saw. The latter two groups are called the unitary group and the special unitary group.
1. For example. The Lie group SO(n.1. Show that O(n. Schematically.82CHAPTER 6. +)? Exercise 6. R) is not connected.). R) be constructed from Mn+1 (R) via the above construction? What about (R. Note that each of the real (complex) matrix groups above is the zero set of some smooth map F from Mm (R) (Mm (C)) to some Euclidean space Rd such that the rank of the Jacobian of F at every point of the zero set of F is d. ai j = 0 for all (i . In the present chapter. Proof. w) for all g ∈ G and v. etc. R) to the identity via a continuous and piecewise smooth path in SO(n. then it is a direct product. in this situation.3. suppose ﬁrst that there exists a position (i. Then one can choose this position such that.5. j) such that i > j and aij = 0. Proposition 6. The theory of Lie groups is a blend of group theory and topology (surfaces. Show that.1. R). the zero set of F is a group with respect to matrix multiplication—as it is in the above cases—then it is automatically a Lie group. If. Such a product is called semidirect. continuity. ∗ ∗ . R) is connected.4. in SU(2). The ﬁrst step is to connect it to a diagonal matrix.1. Exercise 6. R) (Z/2Z). Show that O(n. gw) = (v. j ) with i > j and either j < j or j = j and i > i. or whether it is compact. to this end. Such Lie groups are called linear because they are given by a linear representation. diﬀerentiability. How can GL(n. any element is conjugate to a diagonal matrix of the form e(t) := eit 0 0 e−it for certain t ∈ R. as might be expected.1. SOME COMPACT LIE GROUPS AND THEIR REPRESENTATIONS group is the group of all linear maps on V with (gv. We will connect an arbitrary matrix g = (aij )ij in SO(n. in addition. Exercise 6. A typical lemma from the theory is the statement that the connected component of the identity element is a normal subgroup.1. R) ∼ SO(n. Here means that = the subgroup at the left is a normal subgroup and the subgroup at the right is a complementary subgroup in the sense that their product is the whole group and their intersection is the trivial group. g has the following form: T 0 g = 0 0 0 ∗ ajj ∗ aij 0 ∗ ∗ ∗ aji ∗ ∗ ∗ aii ∗ ∗ ∗ ∗ ∗ .2. If n is odd. this class of Lie groups suﬃces for our needs. one can ask whether a given Lie group is (pathwise) connected. Exercise 6. w ∈ V .
t0 ] lies in SO(n. cos(t0 )) is perpendicular to (ajj . t ∈ R be the real n × n matrix given by cos(t). Being an element of O(n. cos(t) 0 0 I note that r(t) ∈ SO(n.6. and 0 otherwise. r(t)i j Then r(t) has the following block form: I 0 r(t) = 0 0 0 0 0 cos(t) 0 0 I sin(t) 0 0 0 0 0 − sin(t) 0 0 0 . so that there exists a t0 ∈ [0. if i = j and j = i. we ﬁnd a path connecting g to an upper triangular matrix h. The following exercise will be used later on. Each such pair can be smoothly transformed to 1s by multiplication from the left with an appropriate r(t). 1. j ) with i > j and either j < j or j = j and i > i.1. Now let r(t). the number of −1s is even. R). R) for all t ∈ R.6.1. . Now we have T 0 c(t) := r(t)g = 0 0 0 ∗ ajj cos(t) − aij sin(t) ∗ ajj sin(t) + aij cos(t) 0 ∗ ∗ ∗ ∗ ∗ ∗ aji cos(t) − aii sin(t) ∗ aji sin(t) + aii cos(t) ∗ ∗ ∗ ∗ ∗ ∗ Now R(sin(t). This proposition implies that SO(n. Show that SU(2) is connected. Moreover. in addition to c(t0 )i j = 0 for all (i . for the representation theory of SO(3. Repeating the above construction. sin(t). R). π] as above. cos(t)) runs through all 1dimensional subspaces of R2 as t runs from 0 to π. if i = i = j = j. R). π) for which (sin(t0 ). Hint: ﬁnd a continuous path from any given element of SU(2) to some e(t) as in Exercise 6. if i = j = j or i = j = i. SOME EXAMPLES OF LIE GROUPS 83 where T is a (j − 1) × (j − 1) upper triangular matrix. R) and connects g to c(t0 ). the path c(t). h is in fact diagonal and has only entries ±1 on the diagonal.3. as det(h) = 1.1. we ﬁnd that c(t0 )ij = 0. if i = i and j = j. R) is the connected component of the identity in O(n. aij ) in the standard inner product. = − sin(t). t ∈ [0. Exercise 6. Moreover. t ∈ [0. and they can be partitioned into pairs.
. because much of the representation theory of ﬁnite groups works just as well for these. we have for all h ∈ G and f ∈ C(G) that h · f (g)dg = G G f (g)dg. Let us state the existence and uniqueness of this measure in a theorem. A ﬁnite group is a compact Lie group. and denote by C(G) the linear space of all continuous functions G → R on G. For the group T and the natural representation. The 1.e.1. In general. where h · f is the function g → f (h−1 g) (h ∈ G). C(G) → R which is 1. i. 3. we have (e+f )(g)dg = G G f (g)dg + G e(g)dg and G (αf )(g)dg = α G f (g)dg.. where e(t) is deﬁned as in Exercise 6. we can make invariant measures on G by taking certain entries of the matrix r(g)−1 dr(g).2. which enables us to take ‘the average over a group’ just as we did in ﬁnite group theory.2. that is. and normalized. 4.e. leftinvariant. There is a unique map f → G f (g)dg. linear. the invariant Haarintegral. we search for invariant measures on e(t). this is e(t)−1 de(t). Since 0 idt = 2πi. The main tool is the socalled Haar measure. for all e.84CHAPTER 6. i.. 2. Exercise 6. 1 entry of 2π this matrix is e−it deit = idt. 0 . G This linear map is called the invariant (Haar)integral. Let G be a compact real Lie group. Theorem 6.2. if e(g) ≤ f (g) for all g ∈ G..e.1. with the property that their embedding in Euclidean space is closed and bounded) are particularly nice. then also e(g)dg ≤ G G f (g)dg.e.e. given a representation r : G → GL(V ). SOME COMPACT LIE GROUPS AND THEIR REPRESENTATIONS 6. for the action of G induced on C(G) by the left multiplication. f ∈ C(G) and α ∈ R. i. viewed as a map assigning a real number to a function f : R/2πZ → R.2 Representation theory of compact Lie groups Compact Lie groups (i. is 1 2π 2π f (t)dt. i. monotonous..3. What is the invariant integral? In order to ﬁnd an invariant Haar integral on the compact Lie group T = {e(t)  t ∈ R}. diﬀerential forms invariant under left multiplication. 1dg = 1.
be it that at some places a sum must be replaced by an integral.3. where V is a ﬁnitedimensional complex vector space. and let W ⊂ V be an invariant subspace.5. We will study the representations of the groups SU(2) and SO(3.j aij bij . (bij )ij ) := i. To this end. where e is the oneparameter subgroup of SU(2) deﬁned in Exercise 6. is a linear bijection between the space H of continuous class functions on SU(2) and the linear space of all even.3. b ∈ Mn (C). the norm corresponding to the invariant inner product (6. almost all theorems that we proved in the preceding chapters for ﬁnite groups hold. In this exercise we prove that U(n).1.2.2. R) and SO(n. O(n. to the normalized L2 inner product on the space of continuous even 2πperiodic functions on R.2. 2πperiodic continuous functions R → C. R). 1. but here we shall restrict our attention to ﬁnitedimensional representations. Show that (a. Show that W has an invariant complement in V . SU(n). REPRESENTATION THEORY OF COMPACT LIE GROUPS 85 A similar but more laborious computation yields that the invariant Haar measure for SU2 (C) is 1 sin2 θ sin ψdθdψdφ. b) = tr(ab∗ ) for all a. . To give an idea how things work. The result of the following exercise will be used in proving that certain irreducible representations of SU2 (C) exhaust all irreducible representations. Also. Exercise 6. one can prove that continuous homomorphisms from a Lie group G are automatically smooth. R) are compact Lie groups. One could deﬁne representations of compact Lie groups in Hilbert spaces. so it is good to verify that these are indeed compact Lie groups. Exercise 6. For ﬁnitedimensional complex representations of compact Lie groups. Exercise 6. under this linear bijection.6. Show that the map sending a class function f : SU(2) → C to (f ◦ e) : R → C.6.1) below corresponds.4. Deﬁnition 6. endow the space Mn (C) with the Hermitian inner product given by ((aij )ij .2. Moreover. A ﬁnitedimensional complex representation of the Lie group G is a smooth homomorphism G → GL(V ). 2π 2 where cos θ − i sin θ cos ψ − sin θ sin ψeiφ −iφ sin θ sin ψe cos θ + i sin θ cos ψ is a general element of SU2 (C). In fact. we shall only consider complex representations.2. one can try the following exercise. Let r : G → GL(V ) be a ﬁnitedimensional representation of the compact Lie group G.
.5 of [1]. whose Jacobian has rank n2 at each element of U (n). .3. and is therefore (d + 1)dimensional. one should consider the space of Hilbert space of continuous class functions with Hermitian inner product deﬁned by (f h) := G 2 f (g)h(g)dg. A ﬁnal remark concerns the compact analogon of the inner product on the space of complexvalued functions on a ﬁnite group: in the case of a compact Lie group. Find similar arguments for the other groups above. d. by letting (gP )(ξ) = P (g −1 ξ) for ξ ∈ V. We will use Exercise 4. U(n) is a Lie group by the construction of the beginning of this chapter. As we have seen before. Show that U(n) is bounded with respect to the norm corresponding to this inner product. Show. the space of homogeneous polynomials of degree d deﬁned on V .30. that the group O(n. Theorem 6. and compact as it is a closed and bounded set in Mn (C). on the contrary.1) Here.1. .3. 2. Proof of theorem 6. SOME COMPACT LIE GROUPS AND THEIR REPRESENTATIONS 2. 3. . the action being given by matrixcolumn multiplication (the elements of V are understood to be columns).1. this action induces linear actions on P (V )d . . 6. . It consists of all matrices of the form a 0 ga := 0 a−1 . (6. P ∈ P (V )d . The representations rd are irreducible for all d = 1. k d−k The space P (V )d is spanned by the polynomials Pk : ξ → ξ1 ξ2 for k = 0. We are ready to state the main theorem of this section.86CHAPTER 6. the irreducible characters form an orthonormal Hilbert basis in the space of continuous class functions. g ∈ SU(2). Show that U(n) is the zero set of a smooth map F : Mn (C) → Rn . ..1. Let r : SU(2) → GL(V ) be the standard representation of SU(2) on V = C2 . Thus. C) of complex matrices g satisfying g T g = Iis not compact for n ≥ 2.3 The 2dimensional unitary group This section is based on section II. Consider the diagonal subgroup of SU(2). Let rd : SU(2) → GL(P (V )d ) be the corresponding representation and χd its character. Let A ∈ End(P (V )d ) be a linear map that commutes with all rd (g) for g ∈ SU(2).
A(st Pn ) = k k Taking t such that cos t sin t = 0. and hence that A = cn I. . . . . we get A(st Pn ) = k n cosk t · sinn−k t · ck Pk .3. . . . then the eigenspaces of the latter map are invariant under A. all irreducible representations of the group SU(2). . We compute 1 (ga Pk )(ξ) = Pk (g a ξ) 87 = (ξ1 /a)k (aξ2 )d−k = ad−2k Pk (ξ).3. and its trace is d κd (t) = k=0 ei(d−2k)t . and the result of Exercise 6. consider the real rotations st deﬁned by cos t − sin t st := sin t cos t and compute (st Pn )(ξ) = Pn (s−t ξ) = (ξ1 cos t + ξ2 sin t)n = k n cosk t · sinn−k t · Pk (ξ) k Applying A to the result. THE 2DIMENSIONAL UNITARY GROUP Note that the element a ∈ C must necessarily have norm 1. Theorem 6. it follows from the fact that the Pk are linearly independent that cn = ck for all k = 1. n.3. there are complex numbers ck for k = 0. . Choose a0 such that none of these eigenvalues coincide. We wish to prove that these are all irreducible characters. k First applying A and then rd (st ).2.2. Proof. Pd is diagonal. For this purpose. . as these eigenspaces are all onedimensional. we ﬁnd that also n cosk t · sinn−k t · cn Pk . The representations rd are. . Hence we conclude that Pk is an eigenfunction of rd (ga ) with eigenvalue ad−2k . It remains to show that all ck are equal. up to equivalence. and using the fact that the two commute. .3. The matrix of rd (e(t)) with respect to the basis P0 . we can identify H with the space of even 2πperiodic complexvalued functions on R. To this end we shall use some Fourier theory.2. In view of Exercise 6. d such that APk = ck Pk . . Hence.6. If A commutes with rd (ga0 ).
11.1. Recall the homomorphism φ : SL(2. 1. e1 = 0 1 1 . one has κd+2 (t) = κd (t) + 2 cos((d + 2)t). using the isomorphism from Exercise 6. easily seen to be isomorphic 0 to SO(3.3.2. φ will denote the restriction of this homomorphism to SU(2). that the characters χd span a dense subspace in the Hilbert space of continuous class functions on SU(2). for all x ∈ M. e0 = I. 6. L is the group of Lorentznorm preserving linear maps from M to itself. φ : G → H a surjective homomorphism. A ∈ SL(2.3. 0 1 2 3 So L is the group {B ∈ GL(M )  det(Bx) = det(x). R) from those of SU(2) by means of the following lemma.9. Let G and H be groups. From now on. SOME COMPACT LIE GROUPS AND THEIR REPRESENTATIONS It is readily veriﬁed that κ0 (t) = 1 (this corresponds to the trivial representation of the group). and that κ1 (t) = 2 cos t. We claim that the image is the group G := {B ∈ L  Be0 = e0 and det B = 1}. Here. Indeed. Also. C) → L of Example 2. Hence. This implies. The homomorphism φ was deﬁned by φ(A)(x) = AxA∗ . From Fourier theory we know that this subspace is dense in the Hilbert space of continuous even 2πperiodic functions. for all x ∈ M }. This concludes the proof that the rd exhaust the set of irreducible characters of SU(2). the functions κd for d = 0. we wish to ﬁnd a surjective homomorphism of SU(2) onto SO(3) and to determine its kernel. since eo is just the 2 × 2 identity matrix. e2 = 0 0 i −i . e3 = 0 1 0 0 . by its action on the invariant 3space e⊥ . First of all φ(SU(2)) ⊆ G. where M is the Rlinear span of the following four complex 2×2matrices. .4. the function . Lemma 6. which is.4 The 3dimensional orthogonal group We will obtain the irreducible representations of SO(3. Then the (irreducible) representations of H are in bijective correspondence with the (irreducible) representations r of G with ker r ⊇ ker φ. generate the same space as the functions cos dt. 2.2.88CHAPTER 6. which appeared earlier as Exercise 5. then A∗ = A−1 . . Thus. . Therefore. if A ∈ SU(2). φ(A)(e0 ) = e0 . R). Any irreducible character diﬀerent from all rd would have inner product 0 with all rd . −1 It can be checked that in this way det(x0 e0 + x1 e1 + x2 e2 + x3 e3 ) = x2 − x2 − x2 − x2 . Indeed. and hence be zero. C).
(note that x is nonzero).3. Lemma 6. Furthermore.2 to F . We have that B ∈ U(2). 0 Proof. Now let us continue with the proof that F := φ(SU(2)) coincides with G.4. Now let us prove that G ⊆ φ(SU(2)).4. Lemma 6.4. −I}. Then the usual coordinate transformation rule gives x = B −1 (cH)B. Let two (necessarily orthogonal) eigencolumns of norm 1 corresponding to c and −c be the ﬁrst and second column of a matrix B. and det(φ(I)) = 1. 1. the determinant of any element in the image φ(SU(2)) is 1. Hence. We leave this as an exercise to the reader: Exercise 6. which can apparently take only values in {−1.4.4. to apply Lemma 6. Prove this lemma. and the following exercise shows that this is the whole kernel.6. As SU(2) is connected.5. THE 3DIMENSIONAL ORTHOGONAL GROUP 89 det ◦φ is a continuous function SU(2) → R..4. First we prove the following two lemmas. since any vector in the sphere is mapped to e3 by some group element. where c > 0.4 this action is transitive. 1}. Now take A = dB where d2 = 1/(det B). Let F be a subgroup of SO(3. Exercise 6. i. We conclude that φ : SU(2) → G is a surjective homomorphism. R) with the following two properties. 2. . Show that φ(e(t)) is the rotation around e3 through an angle of 2t. we need only check that F contains all rotations around some axis. since d = 1.4. respectively. Any such x is in fact a Hermitian matrix with trace 0. e⊥ = {x1 e1 + x2 e2 + x3 e3  xi ∈ R} ⊆ M. There is an axis such that F contains all rotations around that axis. R). Here e : R → SU(2) is the oneparameter group deﬁned earlier. then A is still unitary. F acts transitively on the unit sphere. and det A = d2 det B = 1.e. −c.4. We follow the approach given in [4]. we have AxA−1 = (dB)x(dB)−1 = dd−1 BxB −1 = cH. Then F = SO(3. as is easily checked. Its kernel contains {I. For every nonzero x ∈ e⊥ there is an A ∈ SU(2) such that 0 φ(A)x = ce3 for some c ∈ R and c > 0. and by 0 1 2 3 Lemma 6. The group F acts on the unit sphere S in e⊥ deﬁned by x2 + x2 + x2 . As such it has two real eigenvalues c. Here e⊥ denotes the orthogonal 0 complement of e0 with respect to the Lorentz inner product.2.
it is necessary and suﬃcient that ((−I)P )(ξ) = P (ξ) for all ξ ∈ V.invariant subspaces of P (V )d . Hence P (V )2 is not irreducible. More precisely. we wish to know for which d we have rd (−I) = I. the group GL(3. Example 6. Hence the irreducible representations are Wd := P (V )2d . As we have seen before. Hint: the elements of this kernel must commute with all of M . As was noticed in Example 4.1: the irreducible representations of SO(3) correspond to those of SU(2) for which −I is mapped to the identity transformation. Consider P (V )d the space of homogeneous polynomials of degree d deﬁned on the vector space V which will be of dimension three here. SOME COMPACT LIE GROUPS AND THEIR REPRESENTATIONS Exercise 6.4.4. x3 ) 2 d! = k=0 xk 1 pk (x2 . p ∈ P (V )d and x ∈ R3 . In the following example. x3 ) = p0 (x2 . Actually the spaces P (V )d are not irreducible for all d ≥ 2. x2 . We now want to prove that the dimension of hd is 2d + 1. To ﬁnd SO(3). The space of harmonic polynomials 1 2 3 hd is the vector space: hd = {p ∈ P (V )d  ∆p = 0}. Take the subspace U of P (V )2 generated by the polynomial x2 + x2 + x2 . 1 2 3 This is an invariant subspace for the action of SO(3). x3 ) + x1 p1 (x2 . −I}.1. x3 ) + d xd x2 1 p2 (x2 . R) acts on this space P (V )d by (A·p)(x) = p(A−1 x).6. hence in particular with e1 . and e3 . x3 ) + k! d k=0 xk 1 k! ∂2 ∂2 2 + ∂x2 ∂x2 3 pk (x2 .7. A homogeneous polynomial p of degree d can be written as a sum of homogeneous polynomials pk in x2 and x3 with power of x1 ’s as coeﬃcients: d+2 d p(x1 . the dimension of P (V )d is dim P (V )d = . For this to be the case. e2 . Show that ker φ = {I. If one applies the Laplace operator to p (you see why we took factorials): d−2 ∆(p) = k=0 xk 1 pk+2 (x2 .4. we need to deﬁne the Laplace ∂2 ∂2 ∂2 operator ∆ on R3 : ∆ = ∂x2 + ∂x2 + ∂x2 . So does the subgroup SO(3). for all A ∈ GL(3.17. R). P ∈ P (V )d . we will see another interpretation of the irreducible representations of SO(3. R).90CHAPTER 6. This holds if and only if d is even. x3 ) + · · · + 1 pd (x2 . x3 ) k! where the pk ’s have degree d − k. Finally we can apply lemma 6. Note that the dimension of Wd is 2d + 1. x3 ) .
8. From this we conclude that hd is a representation of SO(3). Set p a harmonic polynomial in hd . So p generates an invariant subspace W of hd under the action of R(t). x3 ) = (x2 + ix3 )d . That is. So what we saw above about the action of R(t) on W leads us to conclude that hd = Wd . Indeed.6. x3 ) 91 This formula implies that p in hd depends only on the ﬁrst two homogeneous polynomials p0 and p1 . . Let R(t) be the matrix of SO(3) deﬁned by 1 0 0 R(t) = 0 cos t − sin t 0 sin t cos t Apply R(t) to p. So in order to ﬁnd the character of Wd on R(t) it suﬃces to compute it on R(t). Any element in SO(3) is conjugate to R(t). The space hd of harmonic polynomials of degree d is an irreducible representation of SO(3). x2 . . . the degree of hd reminds us of the irreducible representations Wd of SO(3) we met before. we know that its character is a linear combination of exp(ist). 2d i(d−k)t . we claim that hd = Wd . Fact: The action of the Laplace operator on P (V )d commutes with the action of SO(3). we have the following.4. We state here the following fact without proof. To see that what we just stated is true.. THE 3DIMENSIONAL ORTHOGONAL GROUP Hence an element p ∈ P (V )d is in hd if and only if for all k = 0. dim hd = 2d + 1. x3 ) = − ∂2 ∂2 + 2 ∂x2 ∂x3 2 pk (x2 . Therefore we conclude that the dimension of hd is precisely the number of homogeneous polynomials in two variables of degree d and d − 1. That gives us R(t)p = exp(−idt)p. To prove that hd is irreducible. . i. we should ﬁrst check that hd is a representation of SO(3). Theorem 6. Of course.4. That is just the value of the character χ2d of rd (the irreducible representation of SU(2)) at et/2 . . d − 2: pk+2 (x2 . p(x1 .e. k=0 e By the decomposition of hd in irreducible Ws ’s.
92CHAPTER 6. SOME COMPACT LIE GROUPS AND THEIR REPRESENTATIONS .
Wiley [3] J. Applied group theory. physicists and engineers. Wiley. Interscience. 1991. London. [7] H. PrenticeHall.P. Serre. Princeton University Press. PrenticeHall electrical engineering series. Princeton science library. Cambridge University Press. third edition. Springer Verlag. 1994. [2] F. Birkh¨user. Englewood Cliﬀs. [6] Juergen Hinze.Vinberg. [8] A. volume 35 of London Mathematical Society Student Texts. [11] Hermann Weyl. tom Dieck. [4] E. Chichester. editor. New York. Nussbaum. Basel. New York.Bibliography [1] T. Jaﬀe and Milton Orchin. Cambridge university press. Berlin. Young tableaux. 1990. [9] J. volume 12 of Lecture notes in chemistry. a [5] William Fulton. Linear Representations of Groups. Sternberg. 1995. 1971. 1965. 1977. 1979. Representations of Compact Lie Groups. Group theory and physics. [10] S. 1989. Linear representations of ﬁnite groups. Chemical applications of group theory. Princeton. for chemists. The permutation group in physics and chemistry.H. Albert Cotton. New York.Dixon and B. Symmetry in chemistry. Br¨cker and T.D. Cambridge. o Springer Verlag. 1997. Springer Verlag. Symmetry. Springer Verlag. 1989. Mortimer Permutation Groups. 93 .
This action might not be possible to undo. Are you sure you want to continue?