Royal Physicist: 18.3 Schur 'S Lemmas and Grand Orthogonality Theorem (GOT)
Royal Physicist: 18.3 Schur 'S Lemmas and Grand Orthogonality Theorem (GOT)
Royal Physicist: 18.3 Schur 'S Lemmas and Grand Orthogonality Theorem (GOT)
Pivotal notions of the representation theory of finite groups rest upon Schur’s
lemmas (first lemma and second lemma).
Schur’s First Lemma [1, 2] Let D and D e be two irreducible representations of ℊ.
e be m and n, respectively. Suppose that
Let dimensions of representations of D and D
8
with g 2 ℊ the following relation holds:
e ðgÞ,
DðgÞM ¼ M D ð18:41Þ
t
e are inequivalent. In Case (ii), on the
is
In Case (i), the representations of D and D
other hand, D and D e are equivalent.
Proof ic
ys
(a) First, suppose that m > n. Let B ¼ fψ 1 , ψ 2 , , ψ m g be a basis set of the
representation space related to D. Then we have
Ph
Xm
gð ψ ν Þ ¼ μ¼1
ψ μ Dμν ðgÞ ð1 ν mÞ:
Xm
ϕν ¼ ψ μ M μν ð1 ν nÞ: ð18:42Þ
oy
μ¼1
Xm Xm hXm i
gð ϕ ν Þ ¼ μ¼1
g ψ μ M μν ¼ μ¼1 λ¼1
ψ λ D λμ ð g Þ M μν
Xm hXm i Xm hXn i
¼ ψ λ D λμ ð g ÞM μν ¼ ψ λ M λμ
e μν ðgÞ
D
λ¼1 μ¼1 λ¼1 μ¼1
Xn hXm i Xn
¼ ψ M D e μν ðgÞ ¼ ϕ D e ðgÞ:
μ¼1 λ¼1 λ λμ μ¼1 μ μν
ð18:43Þ
Page 1 of 6
As n < m by supposition, this implies that as in (18.35), D would be decomposed
into representations whose dimension is smaller than m. It is in contradiction to
the supposition that D is irreducible. To avoid this contradiction, we must have
M ¼ 0. This makes (18.41) trivially hold.
(b) Next we consider a case of m < n. Taking a complex conjugate of (18.41) and
exchanging both sides, we get
e ðgÞ{ M { ¼ M { DðgÞ{ :
D ð18:44Þ
Therefore, we get
t
D g1 ¼ DðgÞ1 ¼ DðgÞ{ :
is
Similarly, we have
ic
ys
e g1 ¼ D
D e ð gÞ { :
e ðgÞ1 ¼ D
Ph
in parallel with (18.41) where m > n with a (m, n) matrix M. Figure 18.2
graphically shows the magnitude relationship in representation dimensions
R
between representation matrices and M. Thus, in parallel with the above argu-
ment of (a), we exclude the case where m < n as well.
(c) We consider the third case of m ¼ n. Similarly as before we make a linear
combination of vectors contained in the basis set B ¼ fψ 1 , ψ 2 , , ψ m g such that
Xm
ϕν ¼ μ¼1
ψ μ M μν ð1 ν mÞ, ð18:47Þ
where M is a (m, m) square matrix. If detM ¼ 0, then ϕ1, ϕ2, , and ϕm are
linearly dependent (see Sect. 11.4). With the number of linearly independent
vectors p, we have p < m accordingly. As in (18.43) again, this implies that we
would have obtained a representation of a smaller dimension p for D e , in
e
contradiction to the supposition that D is irreducible. To avoid this contradiction,
we must have detM 6¼ 0. These complete the proof. ∎
Page 2 of 6
(a):
( ) × = ( )
×
(m, m) (m, n) (m, n) (n, n)
(b): <
( ) × = × ( )
(n, n) (n, m) (n, m) (m, m)
t
is
Fig. 18.2 Magnitude relationship between dimensions of representation matrices and M. (a) m > n.
(b) m < n. The diagram is based on (18.41) and (18.46) of the text
ic
ys
Schur’s Second Lemma [1, 2] Let D be a representation of ℊ. Suppose that with
8
g 2 ℊ we have
Ph
M ¼ cE: ð18:50Þ
∎
Schur’s lemmas lead to important orthogonality theorem that plays a fundamental
role in many scientific fields. The orthogonality theorem includes that of matrices
and their traces (or characters).
Theorem 18.3: Grand Orthogonality Theorem (GOT) [5] Let D(1), D(2), be
all inequivalent irreducible representations of a group ℊ ¼ {g1, g2, , gn} of order
n. Let D(α) and D(β) be two irreducible representations chosen from among D(1), D(2),
. Then, regarding their matrix representations, we have the following relationship:
Page 3 of 6
X ð αÞ ðβÞ n
Dij ðgÞ Dkl ðgÞ ¼ δ δ δ , ð18:51Þ
g dα αβ ik jl
where Σg means that the summation should be taken over all n group elements; dα
denotes a dimension of the representation D(α). The symbol δαβ means that δαβ ¼ 1
when D(α) and D(β) are equivalent and that δαβ ¼ 0 when D(α) and D(β) are
inequivalent.
Proof First we prove the case where D(α) ¼ D(β). For the sake of simple expression
we omit a superscript and denote D(α) simply by D. Let us construct a matrix A such
that
X
A¼ g
DðgÞXD g1 , ð18:52Þ
t
X 1 X 1 0 1
is
Dðg0 ÞA ¼ g
D ð g 0
ÞD ð g ÞXD g ¼ g
Dð g 0
ÞD ð g ÞXD g D g Dðg0 Þ
X
¼
X
g
Dðg0 ÞDðgÞXD g1 g0
h
0 1
i
1
ic
Dðg0 Þ
ys
0
¼ g
D ð g g ÞXD ð g g Þ Dðg0 Þ:
ð18:53Þ
Ph
Thanks to the rearrangement theorem, for fixed g0 the element g0g runs through all
the group elements as g does so. Therefore, we have
X h i X
al
0 0 1
g
Dð g g ÞXD ð g g Þ ¼ g
DðgÞXD g1 ¼ A: ð18:54Þ
oy
Thus,
A ¼ λE: ð18:56Þ
ðlÞ
The value of a constant λ depends upon the choice of X. Let X be δi δjðmÞ where all
the matrix elements are zero except for the (l, m)-component that takes 1 (Sects. 12.5
and 12.6). Thus from (18.53) we have
X X
g,p,q
Dip ðgÞδðplÞ δqðmÞ Dqj g1 ¼ D ðgÞDmj g1 ¼ λlm δij ,
g il
ð18:57Þ
Page 4 of 6
X
g
Dil ðgÞDjm ðgÞ ¼ λlm δij : ð18:58Þ
X X
X
1 X 1 1
g i
Dil ðg ÞD mi g ¼ g
D g Dð g Þ ¼ g
D g g
ml ml
X X
¼ g
½DðeÞml ¼ δ
g ml
¼ nδml , ð18:59Þ
t
is
where d is equal to a dimension of D. From (18.59) and (18.60), we get
n
λlm d ¼ nδlm or ic
λlm ¼ δlm :
d
ð18:61Þ
ys
Therefore, from (18.58)
X n
Ph
ðαÞ ðαÞ n
Dil ðgÞDjm ðgÞ ¼ δ δ , ð18:63Þ
g dα lm ij
oy
sentations. Let D(α) and D(β) be such representations with dimensions dα and dβ,
respectively. Let us construct a matrix B such that
X
B¼ g
DðαÞ ðgÞXDðβÞ g1 , ð18:64Þ
Page 5 of 6
X
DðαÞ ðg0 ÞB ¼ g
D ðαÞ 0
ð g ÞD ðαÞ
ð g ÞXD ðβÞ 1
g
X
ð αÞ 0 ð αÞ
ðβÞ 0 1 ðβÞ 0
ðβÞ 1
¼ g
D ð g ÞD ð g ÞXD g D g D ðg Þ ð18:65Þ
X h i
1
¼ g
DðαÞ ðg0 gÞXDðβÞ ðg0 gÞ DðβÞ ðg0 Þ ¼ BDðβÞ ðg0 Þ:
B ¼ 0: ð18:66Þ
ðlÞ
Putting X ¼ δi δjðmÞ as before and rewriting (18.64), we get
X ðαÞ ðβÞ
g
Dil ðgÞDjm ðgÞ ¼ 0: ð18:67Þ
t
Combining (18.63) and (18.67), we get (18.51). These procedures complete the
is
proof. ∎
ic
ys
Ph
al
oy
R
Page 6 of 6