REPRESENTATION THEORY EXERCISES AND SOLUTIONS

1. Due on 06th March

Definition 1.1. A (continuous) quotient representation of a (continuous) representation
(ρ, V ) of a group G is a representation π of G on a quotient space V /W , where W is G
invariant, such that
π(g).(v + W ) = ρ(g).v + W, g ∈ G, v ∈ V.
Definition 1.2. A coinvariant representation of a representation (ρ, V ) of a group G is a
quotient representation (ρG , VG ) as above, where VG = V /W , and W is the subspace generated
by {g.v − v | g ∈ G, v ∈ V }.
1.1. Exercise.
(1) Let π be the representation of R on R3 given by the map
 
1 t
t 7→  1 t  .
1
Compute πG .
(2) Let V be a Hilbert space and W ⊆ V . Then V /W ∼ = W ⊥.
(3) Let π be a unitary representation then πG ∼
= πG.
(4) Produce a counter-example of (3) in non-unitary case.
1.2. Solution.
(1) Let v = (v1 , v2 , v3 )t . If π(t)v − v = t(v2 , v3 , 0)t . Clearly the subspace generated by
such vectors is R2 . Hence πG = V /R2 ∼ = R which can be realized as R(0, 0, 1)t .
⊥
(2) As V = W ⊕ W there exists a projection map
Pr : V → W.
This map is linear and surjective as W ⊆ V is mapped identically. Kernel of this
map is W ⊥ . Hence the claim follows from the first isomorphism theorem and the fact
(in a Hilbert space) that W ⊥⊥ = W .
(3) First we note that if W be the subspace generated by {gv − v | g ∈ G, v ∈ V } then
W ⊥ = V G . To see that, if w ∈ W ⊥ then
hw, gv − vi = 0 ⇐⇒ hw − g −1 w, vi = 0,
for all g ∈ G and v ∈ V , using unitarity. Using non-degeneracy of the inner product
we conclude that gw = w for g ∈ G, so w ∈ V G . So using (2) we have a vector space
isomorphism induced by the projection map Pr : V → V G . It is enough to prove
1
2 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

that this map commutes with corresponding representations. To see that, note that
V = V G ⊕ W as a Hilbert space direct sum. Then
π G (g) ◦ Pr(v + W ) = π G (g)(v) = v = π(g)v = Pr(π(g)v + W ) = Pr ◦πG (g)(v + W ).
This completes the proof.

1.3. Exercise. Let Sn be the symmetric group of n letters for n ≥ 1. Suppose that ω(σ) is
the number of disjoint cycles in the decomposition of a permutation σ ∈ Sn . Prove that for
z∈C
n  
1 X ω(σ) Y z−1
z = 1+ .
n! σ∈S j=1
j
n

1.4. Solution. As both sides of the equation are polynomials it is enought o prove the
formula for z = m ∈ N. Let V = Cm and W = V ⊗n . Let Sn acts on W as follows: Fix a
basis {v1 . . . , vm } of V . An element σ ∈ Sn acts on a basis vector of W by
σ(vi1 ⊗ . . . ⊗ vin ) := viσ(1) ⊗ . . . ⊗ viσ(n) .
for 1 ≤ ij ≤ m, 1 ≤ j ≤ n. We call this representation π. Therefore each π(σ) is a
permutation matrix. So the character of π
χπ (σ) := Tr(π(σ))
is the number of fixed basis vectors by Sn , i.e. the number of n-tuple (i1 , . . . , in ) which are
invariant by the action σ. Note that for any fixed choice of such tuple the function j 7→ ij
will be fixed by σ along the cycles of it. But for the indices we can arbitrarily choose ij in
{1, . . . , m}. So total number of fixed vector is mω(σ) . Now the left hand side of our desired
expression
1 X ω(σ) 1 X
m = χπ (g) = dim(W Sn ),
n! σ∈S |Sn | g∈S
n n

from character theory of the finite groups. Also, as any representation of a finite group is
unatry (unitarizable), using previous question we have
W Sn ∼= WS = V ⊗n /hσ(vi ⊗ . . . ⊗ vi ) = (vi ⊗ . . . ⊗ vi )i = Symn V.
n 1 n 1 n

It is a well-known easy to check fact that

  Y n  
n n+m−1 m 1
dim(Sym V ) = = 1+ − .
m−1 j=1
j j

This concludes the proof.

Definition 1.3. A continuous linear operator T on a Hilbert space H to itself is called
compact if there exists a sequence {Tn } of continuous linear operators with dim Im(Tn ) < ∞
such that Tn → T in the operator norm topology.
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 3

1.5. Exercise. Let T be a compact normal operator on a Hilbert space H. Using the
Spectral Theorem stated in the class prove that there is an at most countable set S ⊆ C
such that the eigenspace ker(T − λ) is non-zero iff λ ∈ S and is finite dimensional if λ 6= 0.
Also prove that the Hilbert space orthogonal direct sum decomposition
H=⊕
b λ∈S ker(T − λ).
In particular, if T 6= 0, S is not singleton.
1.6. Solution. Let T 6= 0 be a compact operator (if T = 0 then whole statement is trivial).
Then (equivalently),
{T v | kvk= 1} is compact.
First we prove that eigenvalues of T is upper and lower bounded and any nonzero eigenspace
is finite dimensional. First claim can be restated as, for and  > 0 there are only finitely
many eigenvalues whose absolute values are larger than . If not, we may choose {vi } a set
of orthonormal vectors such that vi is λi -eigenvector with λi |≥ . Then {T vi } is a infinite
set which has no limit point, as,
kλi vi − λj vj k2 = |λi |2 +|λj |≥ 22 ,
contradicting compactness of T . A similar argument also shows that each 0 6= λ-eigenspace
is finite dimensional. Hence, number of eigenvalues is at most countable as set of eigenvalues
is,
∪∞n=1 {λ is T eigenvalue | |λ|> 1/n}.
So it is now enough to prove the spectral decomposition of H. First we need to show that
T has an eigenvalue which equals to
λ := sup{|hT v, vi|| kvk= 1}.
(exists as T is compact). There exists vn ∈ H with kvn k= 1 such that
|hT vn , vn i|→ λ.
As T is compact we pass through a subsequence to get w := lim T vn . Now as
kT vn k≥ hT vn , vn i → λ
we have that kwk≥ λ > 0. Now using a well known fact of a bounded linear operator on a
Hilbert space that
sup{kT vk| kvk= 1} =: kT k= λ := sup{|hT v, vi|| kvk= 1},
and the fact that hT vn , vn i ∈ R as T is self-adjoint, we conclude that
kT vn − λvn k2 = kT vn k2 −λ2 − 2λhT vn , vn i ≤ 2λ2 − 2λhT vn , vn i → 0.
But, as T vn → w from above we conclude that λvn → w. Because T is bounded linear we
also have that λT vn → T w. In other words, T w = λw. Now a usual argument using Zorn’s
lemma after finding maximal sum of eigensubspace completes the proof.
4 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

1.7. Exercise. Prove that any sub-representation of T := R/Z on L2 (T) is of the form
EX = ⊕n∈X Cen ,
for some subset X ⊆ Z. en (z) = exp(2πinz).
1.8. Solution. Let us call
FY := ⊕n∈Y Cen
for some Y ⊂ Z. For a sub-representation E of π (as right regular representation) on L2 (T)
let us define
X := {n ∈ Z | fˆ(n) 6= 0 for some f ∈ E}.
Now E ⊂ FX is trivial to show; because if 0 6= f ∈ E then from the Fourier expansion of f
for all n such that fˆ(n) 6= 0 the character en ∈ FX . Now we show that FX ⊂ E. It is enough
to show that if n ∈ X then en ∈ E. Let n ∈ X so by the definition of X there exists f ∈ E
such that fˆ(n) 6= 0. Let us consider the L2 (T)-valued function
Z
φ : x 7→ π(t)f (x)en (−t)dt.
T

By a change of variable in the integral we can see that φ(x) = en (x)fˆ(n). As fˆ(n) 6= 0 by
assumption it is enough to show that φ ∈ E to show that en ∈ E. But φ(x) can be written
as a limit, from the theory of Riemann integral, as
Z 1 m
1 X
φ(x) := π(t)f (x)en (−t)dt = lim ρ(i/m)f (x)en (i/m).
0 m→∞ m
i=0
But we note that each summand in the RHS is a translate of f ∈ E hence belongs to E.
Therefore, their sum and the limit of the sum also belong to E as E is a closed subspace,
completing the proof.
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 5

2. Due on 13th March

Definition 2.1. Given finite dimensional representations (πi , Vi ) of Gi for i = 1, 2 we define
the external tensor product representation π1  π2 of the group G1 × G2 on V1 ⊗ V2 by

π1  π2 (g1 , g2 )(v1 ⊗ v2 ) = π1 (g1 )v1 ⊗ π2 (g2 )v2 .

2.1. Exercise. Check that π1  π2 is indeed a representation and is unitary if πi ’s are so.

2.2. Solution. Checking that the tensor product a representation is easy. To check unitarity
one notes that,

hπ1  π2 (g1 , g2 )v1 ⊗ v2 , π1  π2 (g1 , g2 )w1 ⊗ w2 i

= hπ1 (g1 )v1 ⊗ π2 (g2 )v2 , π1 (g1 )w1 ⊗ π2 (g2 )w2 i
= hπ1 (g1 )v1 , π1 (g1 )w1 ihπ2 (g2 )v2 , π2 (g2 )w2 i
= hv1 ⊗ v2 , w1 ⊗ w2 i.

2.3. Exercise. Let π be a finite dimensional representation of a finite group G with the
character χπ . Prove that
1 X
π is irreducible ⇐⇒ |χπ (g)|2 = 1.
|G| g∈G

2.4. Solution. Being a finite dimensional representation π is semisimple so that,

X
χπ = nπ (σ)χσ ,
σ∈G
b

where nπ (σ) = dim HomG (π, σ). So

1 X X 1 X X
|χπ (g)|2 = nπ (σ)nπ (τ ) χσ (g)χτ (g) = nπ (σ)2 .
|G| g∈G |G| g∈G
σ,τ ∈Ĝ σ∈Ĝ

As each nπ (σ) ∈ N the LHS will be 1 iff nπ (σ) = 1 for some σ ∈ Ĝ and nπ (τ ) = 0 if σ 6= τ ∈ Ĝ
which is equivalent of saying that π is isomorphic to the irreducible representation σ.

2.5. Exercise. Let G

b be the isomorphism class of unitary irreducible representation of a
group G. Show that the map
(π1 , π2 ) 7→ π1  π2
b1 × G
induces a bijection from G 1 × G2 .
b2 to G\
6 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

2.6. Solution. First we show that the map is well-defined. It amounts to show that when-
ever πi ’s are irreducible then π1  π2 is also irreducible. Using the previous exercise it is
enough to show that
1 X
|χπ1 π2 (g)|2 = 1.
|G| g∈G ×G
1 2

But the above sum can be written as

1 X X 1 X 1 X
|χπ1 (g1 )|2 |χπ2 (g2 )|2 = |χπ1 (g1 )|2 |χπ2 (g2 )|2 ,
|G1 ||G2 | g ∈G g ∈G |G1 | g ∈G |G2 | g ∈G
1 1 2 2 1 1 2 2

which is 1 as πi ’s are irreducible. Now we will show that the map is injective. It is enough
to prove that
χπ1 π2 = χπ10 π20 =⇒ χπi = χπi0 , i = 1, 2.
The left side of the above implies that

χπ1 (g1 )χπ2 (g2 ) = χπ10 (g1 )χπ20 (g2 ), ∀(g1 , g2 ) ∈ G1 × G2 .

Plugging gi = ei ∈ Gi and using the fact that χπ (e) = dim(V ) we conclude that χπi = ci χπi0
for some nonzero constants ci for i = 1, 2. From the irreducibility criterion of πi and of πi0
one proves that ci = 1, which proves injectivity. Surjectivity follows from the fact that

|G|=
b |CG | and CG×H = CG × CH ,

where CG is the conjugacy classes of G.

Definition 2.2. Given a Banach space E, the dual Banach space E 0 of E is the space of all
continuous linear maps E → C with the norm

kλkE 0 := sup |λ(v)|.

kvk=1

Given a representation π of a topological group G on E, i.e. a homomorphism π : G →

BGL(E) we define the (topological) contragredient representation π̃ of G on E 0 by the formula

hπ̃(g)λ, vi = hλ, π(g −1 )vi.

2.7. Exercise. Prove that if π is a continuous representation of a topological group G on a

Banach space E with
supkπ(g)kL(E) < ∞,
g∈G

then the contragredient of π is a continuous representation.

 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 7

2.8. Solution. As π(g) is bounded for all g ∈ G and for any g, g 0 ∈ G and λ, λ0 ∈ E 0 one
has
π̃(g)λ − π̃(g 0 )λ0 = π̃(λ − π̃(g −1 g 0 )λ + π̃(g −1 g 0 )(λ − λ0 ).
it is enough to prove the continuity at the pair (1, 0) ∈ G × E 0 . But we have
kπ̃(g)λk= sup |hλ, π(g 1− vi|≤ supkπ(g)kkλk.
kvk=1 g∈G

Letting λ enough close to 0 for arbitrary g we get the required continuity.

2.9. Exercise. Let π be a unitary representation of G on a Hilbert space H. Show that the
functions
g 7→ hπ(g)v, wi,
for v, w ∈ H, are matrix coefficients of π̃.

2.10. Solution. First we note that for any Hilbert space H there is a canonical isomorphism
between the complex conjugate H̄ and the dual H 0 through the map
v 7→ {λv : w 7→ hw, vi},
by Riesz representation theorem. Also by definition
π̃(g)λ(v) = λ(π(g −1 )v) =⇒ π̃(g)λv (w) = hw, π(g)vi = λπ(g)v (w),
using unitarity of π. Now from the fact that

hλv , λw iH 0 = hv, wiH̄ = hv, wiH

we get that all matrix coefficient of π̃ are of the form

hπ̃(g)λv , λw i = hπ(g)v, wi,

for v, w ∈ H and g ∈ G.

2.11. Exercise. Let G be a group and H a subgroup. Let π be a finite-dimensional repre-

sentation of G on a vector space V . Assume that the restriction of π to H is a direct sum
of pairwise non- isomorphic irreducible representations Vi , 1 ≤ i ≤ k, of H. Show that any
subrepresentation of π is of the form
M
W = Vi
i∈I

for some subset I of {1, . . . , k}.

8 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

2.12. Solution. Let W be a G-stable subspace of V . As as a H-representation

k
M
π |H = V = Vi ,
i=1

with Vi pairwise non-isomorphic irreducible representation of H,

M
W ⊆ Vi .
i∈I

But Vi ’s are, being pairwise non isomorphic, isotypic components. So any irreducible sub-
space of W is an irreducible subspace of V , hence is equal to some Vi . So
Vi ⊆ W, ∀i ∈ I.
Thus M
W = Vi .
i∈I

2.13. Exercise. Let Vm be the subspace of degree m homogeneous polynomials in C[x, y].
We consider the representation πm of SL2 (C) on Vm by
πm (g)f : (x, y) 7→ f ((x, y)g).
Prove that for each m ≥ 0 the representation πm is irreducible.
Hint: consider the restriction of πm to the subgroup T of diagonal matrices, and its action
on a natural basis of Vm to decompose it into irreducible representations.

2.14. Solution. We fix a basis of Vm to be

{ei }m
i=0 , ei := xi y m−i .
We note that the diagonal subgroup A of SL2 (C), namely
   
λ ×
A := a(λ) := |λ∈C
λ−1
acts on ei by
a(λ)ei = λ2i−m ei .
Hence A stabilizes each one dimensional subspace Cei . Therefore, as an A-representation
m
M
Vm = Cei .
i=0

Because each Cei is one dimensional they are trivially irreducible. Hence, applying previous
question we can conclude that if 0 6= V ⊆ Vm is any SL2 (C)-stable irreducible subspace then
M
V = Cei , I ⊆ {0, . . . , m}.
i∈I
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 9

As V nonzero there exists ei ∈ V for some 0 ≤ i ≤ m. Also as V is SL2 (C)-stable

   
1 1 1
e, e ∈ V.
1 i 1 1 i
But this implies that
xi (x + y)m−i , (x + y)i y m−i ∈ V.
Now expanding by the binomial theorem we see that
m  
i m−i
X m−i
x (x + y) = ej ∈ V.
j=i
j − i
From the irreducible decomposition of V we conclude that j ∈ I, ∀j ≥ i. Similarly, expanding
the other monomial we can conclude j ∈ I, ∀j ≤ i. Hence I = {0, . . . , m} or V = Vm . From
irreducibility assumption of V we conclude that Vm is irreducible.
10 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

3. Due on 27th March

G is a compact group unless mentioned otherwise.

3.1. Exercise. The regular representation of G on L2 (G) decomposes as a Hilbert space

direct sum
M
L2 (G) = M (π),
d
π
with notations from the lecture. Let
M
A := M (π),
π

be the algebraic direct sum of M (π)’s. Prove that TFAE

(1) f ∈ L2 (G) is in A.
(2) span(λG (g)f ) is finite dimensional.
(3) span(ρG (g)f ) is finite dimensional.
(4) span(νG (g)f ) is finite dimensional.

3.2. Solution. We will show that (1) ⇐⇒ (2). Rests will follow similarly. First of all, A
is set of finite linear combinations of matrix coefficients. If f ∈ A then f can be written as
n
X
f= ci f i , fi ∈ M (πi ).
i=1
Ln
Clearly, any λ(g)f will be in i=1 M (πi ), which is finite dimensional as each M (πi ) is. Hence
(2). Conversely, if span(λ(g)f ) is finite dimensional, hence closed, being G-stable, should
decompose into finitely many irreducible representations. So f = λ(e)f is contained in a
finite sum of M (π)’s. Hence (1).

3.3. Exercise. Let G be the circle group S 1 = R/Z. Identify all the irreducible representa-
tions of G. Also compute λG (f ) for f ∈ L1 (R/Z).

3.4. Solution. We have already discussed all sub-representations of regular representation

of the circle group. As f ∈ L1 (R/Z) it has a Fourier expansion, namely,
X Z
f (x) = ˆ ˆ
f (n)e(nx), f (n) = f (x)e(−nx)dx.
n∈Z R/Z

Then by definition, for φ ∈ L2 (R/Z)

Z Z
λ(f )φ(g) = f (x)λ(x)φ(g)dx = f (x)φ(g − x)dx
R/Z R/Z
Z X X
= fˆ(n)e(n(g − x))φ(x)dx = fˆ(n)φ̂(n)e(ng).
R/Z n∈Z n∈Z
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 11

Definition 3.1. Let ρ be a unitary representation of G and π ∈ G.b The π-isotypic component
of ρ is the closure of the sum of all sub-representations which are isomorphic to π.
3.5. Exercise. Prove that the projector on the π-isotypic component of ρ is ρ(dim(π)χπ ).

3.6. Solution. We know that any unitary representation of G is Hilbert direct sum of irre-
ducible representations. We first claim that
Z
φπ = ρ(dim(π)χπ ) := dim(π) χπ (g)ρ(g)dµ(g)
G

is an intertwiner. To see this, we first note that χπ being a character of a representation is

a class function i.e. conjugacy invariant. Now we note that for any unitary representation ρ
the operator ρ(f ) is an intertwiner with where f is an L1 class function:
Z
ρ(f ) ◦ ρ(g)v = f (h)ρ(h)ρ(g)vdµ(h)
G
Z
= f (hg −1 )ρ(h)vdµ(h)
G
Z
= f (g −1 h)ρ(h)vdµ(h)
ZG
= ρ(g)f (h)ρ(h)vdµ(h) = ρ(g) ◦ ρ(f )v.
G

In particular, φπ acts on the irreducible components of ρ by a constant, by Schur’s lemma.

If τ is an irreducible component the constant by which φπ acts on it is
Z
−1 dim(π)
dim(τ ) Tr(φπ ) = χπ (g)χτ (g)dµ(g).
dim(τ ) G
By orthonormality of characters the above integral vanishes for all τ  π and is identity if
τ ∼
= π. So φπ is identity on π-isotypic component and zero on other isotypic components,
hence φπ is the projector.

3.7. Exercise. Let ρ : G ,→ Un (C) be a faithful representation. Prove that for all π ∈ G
b
there exist a, b ≥ 0 such that
π ,→ ρ⊗a ⊗ ρ̌⊗b .
Hint: Use Stone-Weierstrass theorem.

3.8. Solution. First we claim that given a set S ⊆ G, b S=G b if and only if the linear span
2
M of matrix coefficients of ρ ∈ S is dense in L (G). This is easy to see. As if M is not
dense then M̄ ⊥ , being nonzero sub-representation of the regular representation, will contain
an irreducible representation which will be outside of S. On the other hand, if an irreducible
representation π is not in S then M will automatically be not dense. Now let S be the set of
12 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

irreducible sub-representations of ρ⊗a ⊗ ρ̌⊗b for a, b ≥ 0. Thus, using the claim, it is enough
to prove that
M := { matrix coefficients of ρ⊗a ⊗ ρ̌⊗b for a, b ≥ 0 }
is dense in C(G) in the supnorm topology. But this is clearly true from the Stone-Weierstrass
theorem.
3.9. Exercise. Let ρ : SUn (C) ,→ Un (C) be a representation of SUn (C). Prove that
1 ,→ ρ⊗m ⇐⇒ n | m,
where 1 is the trivial representation.
3.10. Solution. First we assume that 1 ,→ ρ⊗m . This means that there exists a nonzero
invariant vector v, i.e. ρ⊗m (g)v = v. We write v has linear combination of standard basis
elements: X
v= ci1 ,...,im ei1 ⊗ . . . ⊗ eim .
As v is fixed by SU(n) it is fixed by its center which is isomorphic to Zn . In fact, the center
is
{λI | λn = 1}.
From the fact that λI.v = v we claim that
λI.ei1 ⊗ . . . ⊗ eim = λm ei1 ⊗ . . . ⊗ eim = ei1 ⊗ . . . ⊗ eim .
Thus λm = 1. Choosing λ to be primitive we conclude that n | m. Conversely, if n | m we
need to show that ρm contains an invariant vector. It is enough to show that 1 ,→ ρ⊗n , as
then
1 ,→ (ρ⊗n )⊗m/n ∼
= ρ⊗m .
First we note that,
ρm ∼= ρ̌m ∼
= Hom((Cn )⊗n , C).
Considering the determinant a multilinear map, i.e. det ∈ Hom((Cn )⊗n , C), we choose the
image of it in v ∈ ρ⊗n . We show that v is a nonzero invariant vector. v is nonzero as det is
nonzero map. To show v is invariant we show that det is invariant. In fact,
g. det(v1 ⊗. . .⊗vn ) = det(g −1 v1 ⊗. . .⊗g −1 vn ) = det(g −1 ) det(v1 ⊗. . .⊗vn ) = det(v1 ⊗. . .⊗vn ),
concluding the proof.
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 13

4. Due on 10th April

Definition 4.1. A topological group G is called a complex Lie group if if G is complex
analytic manifold with a group structure such that inverse and multiplication operations of
G are holomorphic.
4.1. Exercise.
(1) Show that the groups Cn , GLn (C), SLn (C), Sp2n (C) are complex Lie groups.
(2) Show that SU2 (C) is not a complex Lie group. Give other examples of Lie groups
which are not complex Lie groups.
(3) Let G be a complex Lie group. Then show that Lie(G) has a natural structure of a
complex Lie algebra.
f
(4) If G1 −
→ G2 is a homomorphism of complex Lie groups then show that df is C-linear
morphism of respective Lie algebras.
(5) Show that sl2 (C), as a real Lie algebra, has irreducible representations sl2 (C) →
gln (R) which are not isomorphic to any of ρn (notation as in lecture).

4.2. Solution.
(1) This is direct. Each of the groups is defined by a zero set of some holomorphic
functions on CN for some N hence a complex analytic manifold. Showing that mul-
tiplication and inverse operations are holomorphic is easy.
(2) This is clear from the real dimension of SU2 (C). If it is a complex Lie group then
the real dimension would be even. But dimension of SU2 (C) is 3.
(3) Standard.
(4) As f is a group morphism df is C-linear. We will show that df respects the Lie
bracket. Note that for any X, Y ∈ g we have
df ([X, Y ]1 ) = df (∂t=0 ∂s=0 (exp(tX), exp(sY )))
= ∂t=0 df (∂s=0 (exp(tX), exp(sY )))
= ∂t=0 ∂s=0 f (exp(tX), exp(sY ))
= ∂t=0 ∂s=0 (f (exp(tX)), f (exp(sY )))
= ∂t=0 ∂s=0 (exp(tdf (X)), exp(sdf (Y )))
= [df (X), df (Y )]2
This confirms that df is a Lie algebra morphism.
(5) Consider the vector space Vm,n of polynomials in variable (z1 , z2 , z¯1 , z¯2 ) which are
homogeneous of degree m in z = (z1 , z2 ) and homogeneous of degree n in z̄ = (z¯1 , z¯2 ),
with SL2 (C) action
πm,n (g)P (z, z̄) = P (zg, z̄ḡ).
Consequently, we get sl2 (C) representation by differentiating above representation.
Clearly this representation is not C-linear but R-linear unless either m or n is zero.
14 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

Hence it is not isomorphic to any of the ρn . This representations are irreducible

which can be proved similarly as the case of sl2 (C).
4.3. Exercise. Let G = SL3 (C) and g = sl3 (C) Let T be the diagonal torus in G. Consider
the adjoint representation
ρ := AdG : G → GL(g).
Construct the weight diagram of ρ with respect to T and find the weight spaces.
4.4. Solution. We will work over the Lie algebra. The Lie algebra t of the diagonal torus
can be parametrized as
t := {x = diag(x1 , x2 , x3 ) | x1 + x2 + x3 = 0, xi ∈ C}.
Let ρ be the adjoint representation of g on gl(g). Eij denotes the standard elementary
matrices. First we note that
ad(x)Eij = (xi − xj )Eij .
So weights are of the form Li − Lj where Li ∈ t∗ given by Li (x) = xi for x ∈ t. Hence the
weight diagram form a Hexagon.

So weight spaces are gLi −Lj for i 6= j and i, j = 1, 2, 3. Here

gλ = {Y ∈ g | ad(x)Y = λ(x)E}.
Consequently we have the weight space decomposition
g = t ⊕ ⊕1≤i6=j≤3 gLi −Lj .
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 15

4.5. Exercise. Let ( )

8
1X
R := {±ei ± ej , i 6= j} ∪ ±ei .
2 i=1
(in the first set, signs are chosen independently; in the second, even number of signs should
be pluses). Prove that R is a reduced root system with Dynkin diagram of E8 .
4.6. Solution. Clearly R spans R8 . Then we check every nαβ = n(α, β) is an integer.
We check case by case. Let p = 12 8i=1 ±ei with the prescribed property (with abuse of
P
notation).
n(±ei ± ej , ±ek ± el ) = ±δik ± δil ± δjk ± δjl ∈ {−2, −1, 0, 1, 2}
is clearly an integer. Also
X X
(p, p) = 1/4( ±ei , ±ei ) = 2
and X
(±ei , ±ej , ±ei ) ∈ {−2, 0, 2}
imply that n(±ei ± ej , p) is an integer, in fact ∈ {−1, 0, 1}. Lastly we see that if p, q are two
different roots of p type then
n(p, q) = 1/4(number of positive signs − number of negative signs) ∈ {−2, −1, 0, 1, 2}
which is also integer. Now we need to show that for any 2 roots α and β
sα (β) = β − n(β, α)α
is also a root. Again we check case by case. When α, β of the form ±ei , ±ej result follows
from n(α, β). When α = ±ei ± ej and β = p then description of p gives the result. Other
cases follow similarly. Lastly we check that only multiple of a root which is also a root is
±1.
16 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

5. Due on 10th May

Fix G = SUn (C).

5.1. Exercise. Let π be the representation of G on the space VN of polynomials of n complex

variables homogeneous of degree N , given by

π(g)P (z) = P (zg).

(1) Find all weights, the highest weight, and the dimension of the representation.
(2) Use the Weyl dimension formula to prove that π is irreducible.

5.2. Solution. First we note that the complexified Lie algebra of G is g = sln (C). One can
construct an explicit basis of VN :

B = {z1k1 . . . znkn | k1 + . . . kn = N }.

Let ei ∈ h0 such that

ei (H) = Hi , H = diag(H1 , . . . , Hn ).
Pn−1
Clearly en = − i=1 ei . Let dπ be the representation of the Lie algebra. We will look at
the action on the basis vectors.

dπ(Hi ) = ∂t=0 π(exp(tHi ))z1k1 . . . znkn = ki z1k1 . . . znkn .

Thus weights are

Xn n
X
{ ki ei | ki = N, ki ≥ 0}.
i=1 i=1

The highest weight is clearly N e1 as ei > ej if i <
j. As a complex vector space VN is

isomorphic to SymN (Cn ), hence dimension is N +n−1
N
.
By the Highest weight theorem we know that there exists an irreducible sub representation
VN e1 of VN corresponding to the highest weight N e1 . So to prove that VN is irreducible it is
enough to show that VN e1 has same dimension as VN . We will now calculate dimension of
VN e1 using the Weyl dimension formula.
We choose a positive system by

R+ = {ei − ej | i < j}.

Therefore the half sum of the positive roots is

n
X n + 1 − 2i
ρ= ei .
i=1
2
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 17

By the Weyl dimension formula we get that

Y hN e1 + ρ, αi
dim(VN e1 ) =
+
hρ, αi
α∈R
Y hN e1 , ei − ej i + h n n+1−2i ei , ei − ej i
P
= Pn n+1−2ii=1 2
i<j
h i=1 2
ei , ei − ej i
n  
Y N +j−i N +n−1
= = .
j=2
j − i n − 1

5.3. Exercise. Let 1 ≤ k ≤ n. Consider the representation V = Λk (Cn ) of G.

(1) Describe all weights of V and describe corresponding weight subspaces.
(2) Find the highest weight of V .
(3) Prove that V is irreducible.
5.4. Solution. Consider the standard basis of V :
B = {vi1 ⊗ . . . vik | 1 ≤ i1 < · · · < ik ≤ n}.
We will claim that
{ei1 + · · · + eik | 1 ≤ i1 < · · · < ik ≤ n}
is the set of weight vectors. To see that we take Hi ∈ h in usual notation and note that
dπ(H)vi1 ⊗ . . . vik = (ei1 + · · · + eik )(H)vi1 ⊗ . . . vik .
Clearly the highest weight is
ω := e1 + · · · + ek .
n


Dimension of V is .
k
Now we will compute the dimension of the irreducible representation Vω of the highest
weight vector by the Weyl dimension formula. As similar to the previous exercise it is
Y hω + ρ, αi
dim(Vω ) =
hρ, αi
α∈R+
Y h k el , ei − ej i + h n n+1−2i ei , ei − ej i
P P
l=1
= Pn n+1−2i i=1 2
i<j
h i=1 2 ei , ei − ej i
k n  
Y Y 1+j−i n
= = .
i=1 j=k+1
j − i k
Using a same kind of argument as the previous exercise we conclude.
5.5. Exercise. Let g be the complex simple Lie algebra of type G2 . Use the Weyl dimension
formula to find dimensions of the irreducible representations of g. Check that there is a
unique one of dimension 7.
18 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

5.6. Solution. Let α1 be the short root and α2 be the long root. Then the fundamental
weights are
ω1 = 2α1 + α2 , ω2 = 3α1 + 2α2 .
Let Lλ be an irreducible representation with highest weights λ = aω1 + bω2 for a, b ∈ Z≥0 .
We recall the following.
(ω1 , ω1 ) = 1, (ω1 , ω2 ) = 3/2, and (ω2 , ω2 ) = 3.
The positive roots are
ω1 , ω2 , 2ω1 − ω2 , 3ω1 − ω2 , ω2 − ω1 , and 2ω2 − 3ω1 .
Therefore the half sum of the positive root ρ = ω1 + ω2 . After an extensive calculation one
can find that
(a + 1)(a + b + 2)(2a + 3b + 5)(a + 2b + 3)(a + bb + 4)(b + 1)
dim Lλ = .
120
For a = 1, b = 0 and a = 0, b = 1 we get the representations of dimensions 7 and 14
respectively. For any other values of (a, b) it will clearly have larger dimension than 7. So
there exists a unique 7 dimensional representation which is the standard representation C7 .
5.7. Exercise. Let π : G → GL(V ) be an irreducible representation of a compact connected
Lie group G.
(1) What are the weights of the dual representations?
(2) Prove that an irreducible representation is self dual if and only if the weights are
symmetric.
(3) Prove that the adjoint representation is self-dual.
(4) What are the duals of the representation associated to the fundamental weights of G
5.8. Solution. Weights of dual of a representations are negative of the original. To see this,
let e be a weight of V . SO dπ(H)v = e(H)v. Therefore,
˜
dπ(H)λ(v) = ∂t=0 (λ(exp(−tH)v) = −e(H)λ(v),
which confirms the claim.
Clearly if V ∼
= V ∗ then ∀α weight of V it is a weight of V ∗ , hence −α is also a weight of
V . So weights are symmetric. Conversely, let λ be the highest weight of V then −λ is also
a weight of V because of symmetry. So λ is the highest weight of V ∗ . So using the highest
weight theorem V ∼ = V ∗.
Using the previous, as the adjoint representation is irreducible it is enough to show that
roots are symmetric. Which is also the case.
 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 19

6. Due on 29th May

6.1. Exercise. Prove that every matrix coefficient of the regular representation of a locally
compact group G vanishes at infinity. Prove that if G is non-compact then the regular
representation has no finite dimensional sub-representation.

6.2. Solution. Proof of this similar to the proof of the Riemann-Lebesgue lemma. First we
see that if f, g ∈ Cc (G) then the matrix coefficient is
Z Z
−1
φ(x) := hλG (t)f, gi = f (t x)g(x)dx = f (x)g(tx)dx.
G G

Clearly support of φ is in supp(g)supp(f )−1 which is compact. So φ vanishes at infinity. The

general case follows from the fact that Cc (G) is dense in L2 (G). Let π be a finite dimensional
representation of G inside λG . Let {vi }i be a orthonormal basis π. As π is unitary for every
g the matrix π(g) is unitary. Therefore, for every i the sum
X
hπ(g)vi , vj i = 1.
j

Hence, at least one of the matrix coefficients among hπ(g)vi , vj ii,j does not decay at infinity.
This contradicts the presvious result.

6.3. Exercise. Let G be discrete group, and φ = δe be the Dirac function at the identity.
Prove that φ is a function of positive type on G and the GNS-representation πφ associated
to φ is the regular representation of G.

6.4. Solution. To see that φ is a function of positive type we take finitely many ci ∈ C and
gi ∈ G, with say, hk are distinct. Then
X X X XX
ci c¯j φ(gj−1 gi ) = cki c¯kj = | cki |2 ≥ 0.
i,j k gki ,gkj =hk k i

Now as G is discrete we note that φ cyclically generate λG because L2 (G) is closed linear
span of {δg }g∈G and δg = λG (g)δe . So from the GNS theorem it is enough to show that the
matrix coefficient of δe is φ. We see that,
X
hλG (x)δe , δe i = δx (g)δe (g) = δx (e) = φ(x),
g∈G

which completes the proof.

6.5. Exercise. Let π be a unitary representation of a locally compact group G and v ∈ π.

Prove that for the positive type function φ(g) = hπ(g)v, viπ the GNS-representation πφ is
contained in π.
20 REPRESENTATION THEORY EXERCISES AND SOLUTIONS

6.6. Solution. Consider the subrepresentation πξ of π which is cyclically generated from ξ.

Clearly the function of positive type associated with πξ is the function
hπξ (g)ξ, ξi = hπ(g)ξ, ξi = φ(g).
So from the uniqueness of the GNS construction we get that
πφ ∼
= πξ ,→ π.

6.7. Exercise. Show that Raikov’s theorem is not true if P1 (G) is replaced by P≤1 (G).

6.8. Solution. Consider the group R and functions

χn : t 7→ eint
on R. These functions belong to P≤1 (R): for a finite set of cj ∈ C and tj ∈ R we note that
 2
X X X 
cj c¯k χn (tj − tk ) = cj c¯k eintj eintk =  cj eintj  ≥ 0,
 
 
j,k j,k i

and χn (0) = 1. χn converges to the function 0 ∈ P≤1 (R) which follows from the statndard
Riemann-Lebesgue lemma. But χn does uniformly converge to the zero function. Hence
Raikov’s theorem does not hold.

6.9. Exercise. Let φ be a continuous function of positive type on a topological group G.

Show that
H := {x ∈ G | φ(x) = φ(e)}
is a closed subgroup of G and that φ is constant on every double coset HxH, x ∈ G.

6.10. Solution. From the GNS-construction there exists (π, v) such that
φ(g) = hπ(x)v, v).
Let x, y ∈ H. Then,
φ(x−1 ) = hπ(x−1 v, vi = hπ(x)v, vi = φ(e).
Also
φ(xy) = hπ(y)v, π(x−1 )vi = hv, vi−1 hπ(y)v, vihπ(x)v, vi = hv, vi = φ(e).
So H is a group. H is closed because it is level set of a continuous function.
Now we see that for h1 , h2 ∈ H and x ∈ G
φ(h1 xh2 ) = hπ(x)π(h2 )v, π(h−1
1 )vi = hπ(x)v, vi = φ(x),

this proves the second assertion.

 REPRESENTATION THEORY EXERCISES AND SOLUTIONS 21

6.11. Exercise. Let H be an open subgroup of a topological group G and φ be a function

of positive type on H. Show that the trivial extension of φ to G, i.e.,
(
φ(g), if g ∈ H
g 7→
0, if g ∈/H
is of positive type. In fact, the characteristic function χH of H is a function of positive type
on G.
6.12. Solution. Let ci ∈ C and gi ∈ G for i = 1, . . . , N be a finite collection of points. We
order {gi } in following way. Let n1 be such an positive integer such that g1−1 gi ∈ H if i ≤ n,
otherwise not. Among the rest of gi we choose an integer n2 such that gn−1 g ∈ H for all
1 +1 i
n1 + 1 ≤ i ≤ n1 + n2 , otherwise not. We continue this process till it ends, say, k times, i.e.
n1 + · · · + nk = N . Now we see that
gi−1 gj ∈ H ⇐⇒ n1 + · · · + nl + 1 ≤ i, j ≤ n1 + · · · + nl + nl+1 , for some 0 ≤ l ≤ k − 1.
So
N k−1 n1 +···+nl+1
X X X
ci c¯j φ(gi−1 gj ) = ci c¯j φi (gi−1 gj ) ≥ 0.
i.j=1 l=0 i,j=n1 +···+nl +1
This proves that φ is of positive type. Now χH is a function of positive type on H as it is
identically one there which proves the second part.
6.13. Exercise. Let G be locally compact abelian group with a finite non-negative measure
µ on it. Define the Fourier transform
Z
µ̂ : Ĝ → C χ 7→ χ(g)dµ(g).
G

Prove that µ̂ is a function of positive type on Ĝ Also show that for any φ of positive type
on Ĝ there exists a finite Borel measure µ on G such that µ̂ = φ.
6.14. Solution. We note that for any finite set of ci ∈ C and χi ∈ Ĝ
Z Z  2
X X X 
−1
ci c¯j µ̂(χj χi ) = ci c¯j χj χi (g)dg = c¯i χi (g) dg ≥ 0
 

i,j i,j G G 
i

which proves the first assertion. For the second, which is known as the Bochner’s theorem
can be found in Theorem D.2.2 in Bekka-Harpe-Vallette.