7. Representations of Finite Groups
7.3 Characters
Let T be a representation of a group G over a field K and M the corresponding KG-module. Then the trace TrM(a) with respect to the module M is defined for every element a E KG (Sect. 6.3); it is the trace of the matrix T(a) (in any basis). In particular, for every element x E G, we get the field element X(x)
=
TrM(x). The function X : G -+ K is called the character of the representation T. If T is irreducible, then X is called an irreducible character.The character of the regular representation is called the regular character and is denoted by Xreg •
Proposition 7.3.1.
() { n for x = 1,
Xreg x = 0 for x
¥=
1.The proof is obvious.
Proposition 7.3.2. For every character, X(gxg- l )
=
X(x). In other words, a character is constant on each conjugacy class.Proof. For every representation T, T(gxg- l ) = T(g)T(x)T(g)-l, and the similar matrices T(x) and T(gxg- l ) have the same trace. 0
Observe also that, as an immediate consequence of Corollary 2.6.3, we get the following theorem.
Theorem 7.3.3. Let K be a field of characteristic O. Then every representa- tion is determined uniquely by its character, i. e. equality of characters implies similarity of representations.
Now, let the field K be algebraically closed and Xl, X2, ... ,Xs be all the irreducible characters of the group G over the field K. Denote by Xii the element Xi(9i), where gi E Ci (Cl, C2 , ••• , Cs are the conjugacy classes of the group G). The square matrix X
=
(Xii) is called the character table of the group G over the field K. Let us remark that KG ~ EEl s diMi, where Mi is the module of the ith irreducible representation and di=l i = [Mi : Kj; hence,s Xreg
= E
diXi.i=l
As we have already pointed out, the elements Ci =
E
x, i = 1,2, ... s,xEGj
form a basis of the center C of the group algebra KG. On the other hand, C ~ KS and therefore, if 1 = el
+
e2+ ... +
es is a decomposition of the iden- tity of the algebra C, then {el' e2, ... , es } is also a basis of C. Consequently, there are elements Qij and fiii in the field K such that Ci =E
s Qijej andi=l
7.3 Characters 121
s
ei =
I:
f3ijCj; and thus the matrices A = (Qij) and B = (f3ij) are reciprocal.It turns out that the coefficients j=1 Qij and f3ij are closely related to the character table.
Proposition 7.3.4. Denote by di the dimension of the irreducible represen- tation with character Xi and hj the number of elements in the class Cj. Then
di -1
f3ij
=
-Xi(gj ), where gj E Cj.n
Proof. Observe that the element ej acts on the jth irreducible representation as identity, while the elements ek (k =I-j) act on it trivially. Therefore Xj(ek) = 0 for k =I-j and Xj(ej) = dj. From here,
8 8
Xj(Ci) = Xj(I>ikek) = LQikXj(ek) = djQij.
k=1 k=1
On the other hand, Xj(c;) = hiXji and the formula for Qij follows.
In order to compute f3ij, we use the fact that Xreg =
I:
s diXi. Ob- serve that Xreg(Ckg)=
0 if g-11.
Ck and Xreg(Ckg)=
n if g-1 i=1 E Ck (this follows from Corollary 7.3.1). Therefore, if gj E Cj, then Xreg(ei9;1) = Xreg (t
f3ikCk9;1)=
nf3ij . On the other hand, Xreg( eig;1)= t
dkXk(eig;1)k=1 k=1
= diXi(9;1) because Xk( eig;1) = 0 for k =I- i and Xi( eig;1) = Xi(g;1). The
formula for f3ij follows. 0
Taking into account that the matrices A and B are reciprocal, we obtain immediately the following "orthogonality relations" for characters.
Theorem 7.3.5.
for i =I-j, fori=ji for i =I-j, for i = j.
Corollary 7.3.6. A representation T of a group G over an algebraically closed field of characteristic 0 is irred1lcible if and only if its character X satisfies
122 7. Representations of Finite Groups
Proof. Decompose the representation T into a direct sum of irreducible repre-
s
sentations. Correspondingly, the character X can be expressed as X =
I:
miXi ,where XI, X2,· .. ,Xs are irreducible characters. But then i=1
and this sum is equal to 1 if and only if X
=
Xi for some i, i. e., in view of Theorem 7.3.3, if T is an irreducible representation. D If ]{ = <C is the field of complex numbers, then the orthogonality relations can be given a slightly different form. To that end, we introduce the following lemma.Lemma 7.3.7. If X is the character of a d-dimensional representation of a group G over the field of complex numbers, then, for every 9 E G, X(g) is a sum of d n-th roots of unity and X(g-I) = x(g ), where as usual,
z
is the complex conjugate of ihe number z.Proof. Since gn
=
e, we get (T(g)r=
E for every element 9 E G. Since the polynomial xn - 1 has no multiple roots, it follows that the matrix T(g) is similar to the diagonal matrix(
CI C2
0)
T(g) '"
o
Cd whereci
= 1.From here, X(g) = CI
+
C2+ ... +
Cd ando o
This results in
( -I) -1
+
-I+ +
-I - -+ - -()
X 9 = cI c2 . . . Cd = CI
+
C2+ . . .
cd = X 9 .D In particular, Xi(gjl)
rem 7.3.5 take the form
Xij and the orthogonality relations of Theo-
1
s{O
;: L
hkXikXjk = 11
k=1 8{O
;: L
XkiXkj = Ilh i k=17.4 Algebraic Integers
7.4 Algebraic Integers 123 for i =1= j,
for i
=
j;for i =1= j, for i
=
j.In this section we shall need some properties of algebraic integers. Recall that an algebraic integer is, by definition, a (complex) root of an equation xm
+
alXm-1+ ... +
am = 0 with integral coefficients ai.Proposition 7.4.1. A rational number which is an algebraic integer zs an integer.
Proof. Let z be a root of an equation Xffi
+
alXm-1+ ... +
am = 0 with integers ai and z = pi q with relatively prime integers p and q>
1. Passing to a common denominator, we get pm = - a1Qpm-l - azqZpm-Z - ... - amqm.This is impossible because p and q are relatively prime. 0 The following lemma provides a convenient criterion for a number z to be an algebraic integer.
Lemma 7.4.2. In order that z be an algebraic integer, it is necessary and suffi- cient that there exist complex numbers Yl, Y2, ... ,Yt such that ZYi =
L
t aijYj , where all aij are integers and not all Yi are zero. j=1Proof. If z is a root of an integral equation xm
+
alXm-1+ ... +
am = 0, then we may take, trivially, Yl = 1, Y2 = Z, ... , Ym = zm-l.Conversely, let Yl, Yz , ... ,Yt have the required property. Denote by A the matrix (aij) and by Y the column vector whose coordinates are Yl, Y2, ... , Yt . Then (zE - A)Y
=
0 and thus det (zE - A)=
O. However, the determinant det (zE - A) = zt +alzt-1+ ... +
at, where ai are integral linear combinations of products of elements of the matrix A and thus integers. We conclude thatz is an aigebraic integer. 0
Corollary 7.4.3. The sum and product of algebraic integers are algebraic integers. In other words, the algebraic integers form a ring.
t
Proof. Let Yl, Yz, .. ·, Yt be complex numbers such that ZYi =
L
aijYj (withr j=1
integers aij) and y~, y~, ... ,y~ such that Zl yi =
L
a~j yj (with integers a~j)' j=1124 7. Representations of Finite Groups
Then one can see easily that the numbers {Yiyj
I
i = 1,2, ... , t; J 1,2, ... ,r} satisfy similar conditions for the numbers z+
Zl and ZZI. 0Since the roots of unity are obviously algebraic integers, we obtain the following corollary of Lemma 7.3.7.
Corollary 7.4.4. If X is a character of a group G over the field of complex numbers, then X(x) is an algebraic integer for every x E G.
We shall now employ the notation of the previous section. In particular, let X =
(X
ij) be the character table of a group G over the field of complex numbers.Theorem 7.4.5. All numbers D:ij =
d:
hi Xji are algebraic integers.J
Proof. Note that, for all i and j, CiCj is an element of the center of the algebra {;G. On the other hand, CiCj is an integral linear combination of the elements of the group G. It follows that CiCj =
L:
rijkCk , where rijk are integers. Besides,k
CiCj =
(2::
D:ipep)(2::
D:jqeq) =2::
D:;pD:jpepp q p
and Ck =
L:
D:kpep ; thus CiCj =L:
rijkD:kpep and D:ipD:jp =L:
rijkD:kp' Writingp k,p k
Z
=
D:ip, Yj=
D:jp (for a fixed p), we can apply Lemma 7.4.2 and concludethat D:ip is an algebraic integer. 0
Corollary 7.4.6. The dimensions di of irreducible complex representations divide the order of the group.
Proof. Rewrite the list of the orthogonal relations of Theorem 7.3.5 to the form
~
hkXik ,'( -1) _~
L d. X, gk - d'
k=l I ,
S· lnce -d-'-hkXik
,
= D:ki an d Xi gk (-1) are age ratc mtegers, a so t e num er l b " 1 h b -d' n.,
ISan algebraic integer. As a rational number, it must be an integer, as required.
o