Serganova V. - Representation theory (2005)

Department of Mathematics
University of California, Berkeley
Mathematics 252
Representation Theory
Vera Serganova, Fall 2005
My office hours are 10:00-11:30 on Wednesdays and Fridays, in 709 Evans Hall. I can be
reached by telephone at (64)2-2150 and electronic mail at serganov@math. You are welcome to ask
questions by email. Homework assignments and course notes can be found on my web page http://
math.berkeley.edu/ serganov. First homework assignment is due on Friday, September 9
There is no required text for this course. Good references are Fulton, Harris, Representation
Theory,Serre, Linear Representations of Finite Groups, Curtis, Reiner, Representation
Theory of Finite Groups and Associative Algebras,Gabriel, Roiter, Representations of
Finite-dimensional Algebras. I will try to post course notes regularly on the web.
To understand this course you need basic knowledge of Algebra and a good knowledge of Linear
Algebra. In other words you have to know basic facts about groups and rings and you should feel
very comfortable when working with linear operators.
Each Friday I will give you a problem assignment (2-4 problems) on the material of the week
lectures. The homework will be collected the next Friday.
The grade will be computed according to the following proportions: 50% for your homework
and 50% for the take home final. But if you solve all problems in your final (there will be hard ones
in it) you get A for the course.
Course outline
• Representations of groups. Definitions and examples

• Schur’s Lemma. Complete reducibility in case of zero characteristic
• Characters and orthogonality relation
• First examples: abelian groups, dihedral group Dn , S3 , S4 , A5 e.t.c.
• Induced representation. Frobenius reciprocity. Mackey’s criterion
• Representations of associative rings. Density theorem. Semi-simple rings, Wedderburn’s the-
orem. Decomposition of a group algebra
• Representations of non-semisimple rings. Blocks. Injective and projective modules
• Representations of symmetric groups, Young diagrams and Frobenius formula
• Representations of general linear group, Weyl duality and Schur’s polynomials ( if time permits)
• Complex representations of linear groups over finite fields, Hecke algebra
• Compact groups and their representations. Peter-Weyl theorem
• Real representations and representations over subfields of C. Schur indices
• Artin’s and Brauer’s theorems
• Representations of quivers. Definition and examples
• Gabriel’s theorem
• Representations over fields of nonzero characteristic (if time permits)
1
REPRESENTATION THEORY.
LECTURE NOTES
VERA SERGANOVA
1. Some problems involving representation theory

Hungry knights. There are n hungry knights at a round table. Each of them
has a plate with certain amount of food. Instead of eating every minute each knight
takes one half of his neighbors servings. They start at 10 in the evening. What can
you tell about food distribution in the morning?
Solution. Denote by xi the amount of food on the plate of the i-th knight. The
distribution of food at the table can be described by a vector x = (x1, x2 , . . . , xn ) ∈
Rn . Every minute a certain linear operator Φ is applied to a distribution x. Thus, we
have to find lim Φm as m approaches infinity. To find the limit we need to diagonalize
Φ, and the easiest way to do this is to write
T + T −1
Φ= ,
2
where T is the rotation operator:
T (x1 , . . . , xn ) = (xn , x1, . . . , xn−1 ) .
It is easy to see that the eigenvalues of T are the n-th roots of 1. Hence the eigenvalues
k
of Φ are ε +ε
−k
2
, where ε is a primitive root of 1, k = 1, . . . , n. The set of eigenvalues
of Φ is
2πk
cos | k = 1, . . . , n .
n
Let us chose a new basis {v1, . . . , vn } in Cn such that
2πk
Φvk = cos vk .
n
For example, we can put vk = ε−k , ε−2k , . . . , ε−nk .

If n is odd all eigenvalues of Φ except 1 have the absolute value less than 1.
Therefore if x = a1 v1 + · · · + an vn , then
n m
m
X 2πk
lim Φ x = lim ak cos vk = an vn .
m→∞ m→∞ n
k=1
Date: August 30, 2005.

1
2 VERA SERGANOVA
But vn = (1, . . . , 1). Therefore eventually all knights will have the same amount of
food equal to the average x1 +···+x
n
n
.
In case when n is even the situation is different, since there are two eigenvalues
with absolute value 1, they are 1 and −1. Hence as m → ∞,
Φm x → (−1)m an/2vn/2 + an vn .
Recall that vk = (−1, 1, −1, . . . , 1). Thus, eventually food alternates between even
an ±an/2
and odd knights, the amount on each plate is approximately 2
, where
x1 + · · · + xn x1 − x2 + · · · − xn
an = , an/2 = .
n n
Slightly modifying this problem we will have more fun.
Breakfast at Mars. It is well known that marsians have four arms, a standard
family has 6 persons and a breakfast table has a form of a cube with each person
occupying a face on a cube. Do the analog of round table problem for the family of
marsians.
Supper at Venus. They have five arms there, 12 persons in a family and sit on
the faces of a dodecahedron (a regular polyhedron whose faces are pentagons).
Tomography problem. You have a solid in 3-dimensional space of unknown
shape. You can measure the area of every plane cross-section which passes through
the origin. Can you determine the shape of the solid? The answer is yes, if the solid
in question is convex and centrally symmetric with respect to the origin.
In all four problems above the important ingredient is a group of symmetries.
In the first case this is a cyclic group of rotations of the table, in the second one
the group of rotations of a cube, in the Venus problem the group of rotations of
a dodecahedron (can you describe these groups?). Finally, in the last problem the
group of all rotations in R3 appears. In all cases the group acts on a vector space via
linear operators, i.e. as we have a linear representation of a group. The main part of
this course deals with representation of groups.
Linear algebra problems. Every standard course of linear algebra discusses the
problem of classification of all matrices in a complex vector space up to equivalence.
(Here A is equivalent to B if A = XBX −1 for some invertible X.) Indeed, there exists
some basis in Cn , in which A has a canonical Jordan form. The following problem is
less known.
Kronecker problem. Let V and W be finite-dimensional vector spaces over
algebraically closed field k, A and B : V → W be two linear operators. Classify
all pairs (A, B) up to the change of bases in V and W . In other words we have to
classify pairs of matrices up to the following equivalence relation: (A, B) is equivalent
to (C, D) if there are invertible square matrices X and Y such that
C = XAY , D = XBY.
REPRESENTATION THEORY. LECTURE NOTES 3
Theorem 1.1. There exist decompositions

V = V1 ⊕ · · · ⊕ Vk , W = W1 ⊕ · · · ⊕ Wk ,
such that A (Vi ) ⊂ Wi , B (Vi ) ⊂ Wi and for each i ≤ k there exist bases in Vi and Wi
such that the matrices for A and B have one of the following forms (here 1n denotes
the identity matrix of size n, Jn the nilpotent Jordan block of size n ):
A = 10n , B = 01n , n ≥ 0;

A = (1n , 0) , B = (0, 1n ) , n ≥ 0;
A = 1n , B = Jn , n ≥ 1;
A = t1n + Jn , B = 1n , n ≥ 1, t ∈ k.
2. Representations of groups. Definition and examples.

Let k denote a field, V be a vector space over k. By GL (V ) we denote the group
of all invertible linear operators in V . If dim V = n, then GL (V ) is isomorphic to
the group of invertible n × n matrices with entries in k.
A (linear) representation of a group G in V is a homomorphism
ρ : G → GL (V ) .
The dimension of V is called the degree or the dimension of a representation ρ and
it may be infinite. For any s ∈ G we denote by ρs the image of s in GL (V ) and for
any v ∈ V we denote by ρs v the image of v under the action of ρs . The following
properties are obvious:
ρs ρt = ρst , ρ1 = Id , ρ−1
s = ρs−1 , ρs (xv + yw) = xρs v + yρs w.
Examples.
1. Let G = Z with operation +, V = R2, ρn is given by the matrix
1 n
0 1
for n ∈ Z.
2. Permutation representation. Let G = Sn , V = k n . For each s ∈ Sn put

ρs (x1, . . . , xn ) = xs(1), . . . , xs(n) .
3. Trivial representation. For any group G the trivial representation is the homo-
morphism ρ : G → k ∗ such that ρs = 1 for all s ∈ G.
4. Let G be a group and
F (G) = {f : G → k}
be the space of functions on G with values in k. For any s ∈ G, f ∈ F (G) and t ∈ G
let
ρs f (t) = f (ts) .
Then ρ : G → GL (F (G)) is a linear representation.
4 VERA SERGANOVA
5. Regular representation. Recall

P that the group algebra k (G) is the vector space of
all finite linear combinations cg g, cg ∈ k with natural multiplication. The regular
representation R : G → GL (k (G)) is defined in the following way
X X
Rs cg g = cg sg.
Two representations ρ : G → GL (V ) and σ : G → GL (W ) are equivalent (or

isomorphic) if there exists an isomorphism T : V → W such that for all s ∈ G
T ◦ ρs = σs ◦ T.
Example. If G is finite then the representations in examples 4 and 5 are equivalent.
Indeed, define T : F (G) → k (G) by the formula
X
T (f) = f (g) g −1 .
g∈G
Then for any f ∈ F (G) we have

X X X
T (ρs f) = ρs f (g) g −1 = f (gs) g −1 = f (h) sh−1 = Rs (T f) .
g∈G g∈G h∈G
3. Operations with representations

Restriction on a subgroup: Let H be a subgroup of G. For any ρ : G → GL (V )
we denote by ResH ρ the restriction of ρ on H.
Lift. Let p : G → H be a homomorphism of groups. For every representation
ρ : H → GL (V ), ρ ◦ p : G → GL (V ) is also a representation. We often use
this construction in case when H = G/N is a quotient group and p is the natural
projection.
Direct sum. If we have two representations ρ : G → GL (V ) and σ : G → GL (W ),
then we can define ρ ⊕ σ : G → GL (V ⊕ W ) by the formula
(ρ ⊕ σ)s (v, w) = (ρs v, σsw) .
Tensor product. The tensor product of ρ : G → GL (V ) and σ : G → GL (W ) is
defined by the formula
(ρ ⊗ σ)s v ⊗ w = ρs v ⊗ σs w.
Exterior tensor product. Let ρ : G → GL (V ) and σ : H → GL (W ) be representa-
tions of two different groups, then their exterior product ρ⊠σ : G×H → GL (V ⊗ W )
is defined by
(ρ ⊠ σ)(s,t) v ⊗ w = ρs v ⊗ σtw.
If δ : G → G × G is the diagonal embedding, then
ρ ⊗ σ = (ρ ⊠ σ) ◦ δ.
Dual representation. For any representation ρ : G → GL (V ) one can define the

dual representation ρ∗ : G → GL (V ∗) by the formula
< ρ∗s ϕ, v >=< ϕ, ρ−1
s v >
for any v ∈ V, ϕ ∈ V ∗ . Here <, > denotes the natural pairing between V and V ∗ .
More generally, if ρ : G → GL (V ) and σ : G → GL (W ) are two representations,
then one can naturally define the representation τ of G in Homk (V, W ) by the formula
τs ϕ = σs ◦ ϕ ◦ ρ−1
s , s ∈ G, ϕ ∈ Homk (V, W ) .
4. Invariant subspaces and irreducibility

Given a representation ρ : G → GL (V ). A subspace W ⊂ V is called invariant if
ρs (W ) ⊂ W for any s ∈ G. One can define naturally the subrepresentation
ρW : G → GL (W )
and the quotient representation
σ : G → GL (V/W ) .
Example. Let ρ : Sn → GL (k n ) be the permutation representation, then
W = {(x1, . . . , xn ) | x1 = x2 = · · · = xn }
and
W ′ = {(x1 , . . . , xn ) | x1 + x2 + · · · + xn = 0}
are invariant subspaces.
Theorem 4.1. (Maschke) Let G be a finite group and char k do not divide |G|. Let
ρ : G → GL (V ) be a representation and W be an invariant subspace. Then there
exists another invariant subspace W ′ such that V = W ⊕ W ′ .
Proof. Let W ′′ be a subspace (not invariant) such that V = W ⊕ W ′′. Let P : V → V
be the linear operator such that P|W = Id and P (W ′′) = 0. Then P 2 = P . Such
operator is called a projector. Let
1 X
P̄ = ρg ◦ P ◦ ρ−1
g .
|G| g∈G
Check that ρs ◦ P̄ ◦ ρ−1

s = P̄ , and hence ρs ◦ P̄ = P̄ ◦ ρs for any s ∈ G. Check also
that P̄|W = Id and Im P̄ = W . Hence P̄ 2 = P̄ .
Let W ′ = Ker P̄. First, we claim that W ′ is invariant. Indeed, let w ∈ W ′ , then
P̄ (ρs w) = ρs P̄ w = 0, hence ρs w ∈ Ker P̄ = W ′.
Now we prove that V = W ⊕ W ′ . Indeed, W ∩ W ′ = 0, since P̄|W = Id. On the
other hand, for any v ∈ V , we have w = P̄ v ∈ W and w′ = v − P̄ v ∈ W ′ . Thus,
v = w + w′, and therefore V = W + W ′ .
6 VERA SERGANOVA
In the previous example V = W ⊕ W ′ if char k does not divide n. Otherwise,

W ⊂ W ′, and the theorem is not true.
If G is infinite group, theorem is not true. Consider the representation of Z in R2
from Example 1. It has the unique one-dimensional invariant subspace, therefore R2
does not split into a direct sum of two invariant subspaces.
A representation is called irreducible if it does not contain a proper non-zero in-
variant subspace.
Exercise. Show that if char k does not divide n, then the subrepresentation W ′
of the permutation representation is irreducible.
Lemma 4.2. Let G be a finite group, ρ : G → GL (V ) be an irreducible representa-
tion. Then dim V ≤ |G|.
Proof. Take any non-zero v ∈ V , then the set {ρs v}s∈G spans an invariant subspace
which must coincide with V . Hence dim V ≤ |G|.
Example. Let ρ : G → GL (V ). We claim that ρ ⊗ ρ is irreducible if and only if
dim V = 1. Indeed the subspaces S 2 V, Λ2V ⊂ V ⊗ V are invariant and Λ2V = {0}
only in case when dim V = 1.
A representation is called completely reducible if it splits into a direct sum of
irreducible subrepresentations.
Corollary 4.3. Let G be a finite group and k be a field such that char k does not
divide |G|. Then every finite-dimensional representation of G is completely reducible.
Proof. By induction on dim V .
5. Schur’s Lemma
For any two representations ρ : G → GL (V ), σ : G → GL (W ) let
HomG (V, W ) = {T ∈ Homk (V, W ) | σs ◦ T = T ◦ ρs , s ∈ G} .
An operator T ∈ HomG (V, W ) is called an intertwining operator. It is clear that
HomG (V, W ) is a vector space. Moreover, if ρ = σ, then HomG (V, V ) = EndG (V ) is
closed under operation of composition, and therefore it is a k-algebra.
Lemma 5.1. Let T ∈ HomG (V, W ), then Ker T and Im T are invariant subspaces.
Proof. Let v ∈ Ker T , then T (ρs v) = ρs (T v) = 0, hence ρs v ∈ Ker T .
Let w ∈ Im T . Then w = T v for some v ∈ V and ρs w = ρs (T v) = T (ρs v) ∈
Im T .
Corollary 5.2. (Schur’s lemma) Let ρ : G → GL (V ) and σ : G → GL (W ) be
irreducible representations of G, then any T ∈ HomG (V, W ) is either isomorphism
or zero.
Proof. Since both V and W do not have proper invariant subspaces, then either
Im T = W , Ker T = {0} or Im T = {0}.
Corollary 5.3. If ρ : G → GL (V ) is irreducible then EndG (V ) is a division ring. If

the field k is algebraically closed and V is finite-dimensional, then EndG (V ) = k Id.
Proof. The first assertion follows immediately from the Corollary 5.2. To prove the
second, let T ∈ EndG (V ) and T 6= 0. Then T is invertible. Let λ be an eigenvalue of
T and S = T − λId. Since S ∈ EndG (V ) and Ker S 6= {0}, by Corollary 5.2, S = 0.
Thus, T = λ Id.
Corollary 5.4. Let G be an abelian group, ρ : G → GL (V ) be an irreducible finite-
dimensional representation of G over algebraically closed field k. Then dim V = 1.
Irreducible representations of a finite cyclic group over C. Let G be a
cyclic group of order n and g be a generator. By Corollary 5.4 every irreducible
representation of G is one-dimensional. Thus, we have to classify homomorphisms
ρ : G → C∗ . Let ρg = ε. Then clearly ε is an n-th root of 1. Therefore we have
exactly n non-equivalent irreducible representations.
Irreducible representations of a finite abelian group over C. Any finite
abelian group is a direct product G1 × · · · × Gk of cyclic groups. Let gi be a generator
of Gi . Then any irreducible ρ : G → C∗ is determined by its values ρgi = εi , where
|G |
εi i = 1. Hence the number of isomorphism classes of irreducible representations of
G equals |G|.
Remark 5.5. It is not difficult to see that the set of one-dimensional representations
of G is a group with respect to the operation of tensor product. In case when G is
finite and abelian and k is algebraically closed, all irreducible representations are one
dimensional and form a group. We denote this group by G∨ . As easily follows from
above G∨ ∼ = G when k = C, however this isomorphism is not canonical.
Here is another application of Schur’s Lemma.
Theorem 5.6. Let ρ ∼ = ρ1 ⊕ · · · ⊕ ρk ∼
= σ1 ⊕ · · · ⊕ σm , where ρi , σj are irreducible.
Then m = k and there exists s ∈ Sk such that ρj ∼ = σs(j) .
Proof. Let V be the space of a representation ρ. There are two decompositions of V
into the direct sum of irreducible decomposable subspaces
V = V1 ⊕ · · · ⊕ Vk = W1 ⊕ · · · ⊕ Wm .
Let pi : V → Wi be the projection which maps Wj to zero for j 6= i, qj : Vj → V be
the embedding. Then pi ∈ HomG (V, Wi ) and qj ∈ HomG (Vj , W ). The map
m X
X k
F = pi ◦ qj : ⊕kj=1 Vj → ⊕m
i=1 Wi
i=1 j=1
is an isomorphism. There exists i such that pi ◦ q1 6= 0. (Otherwise F (V1 ) = 0

which is impossible.) Put s (1) = i and note that pi ◦ q1 is an isomorphism by Schur’s
Lemma. We continue inductively. For each j there exists i such that pi ◦ qj is an
8 VERA SERGANOVA
isomorphism and i 6= s (r) for any r < j. (Indeed, otherwise F (V1 ⊕ · · · ⊕ Vj ) ⊂

Ws(1) ⊕ · · · ⊕ Ws(j−1) which is impossible because F is an isomorphism.) We put
i = s (j). Thus, we can construct an injective map
s : {1, . . . , k} → {1, . . . , m}
such that ρj ∼
= σs(j) . In particular, k ≤ m. But by exchanging ρi and σj we can prove
that m ≤ k. Hence k = m and s is a permutation.
PROBLEM SET # 1
MATH 252
Due September 9.
1. Classify irreducible representations of Z over C.
2. Classify one-dimensional represenations of Sn over any field k such that char k 6=
2.
3. Let ρ be an irreducible representation of G and σ be an irreducible representation
of H. Is it always true that the exterior tensor product of ρ and σ is an irreducible
representation of G × H?
Date: August 30, 2005.

1
REPRESENTATION THEORY
WEEK2
VERA SERGANOVA
1. Characters
For any finite-dimensional representation ρ : G → GL (V ) its character is a func-
tion χρ : G → k defined by
χρ (s) = tr ρs .
It is easy to see that characters have the following properties
(1) χρ (1) = dim ρ;
= σ, then χρ = χσ ;
(2) if ρ ∼
(3) χρ⊕σ = χρ + χσ ;
(4) χρ⊗σ = χρχσ ;
(5) χρ∗ (s) = χρ (s−1 );
(6) χρ (sts−1) = χρ (t).
Example 1. If R is a regular representation, then χR (s) = 0 for any s 6= 1 and
χR (1) = |G|.
Example 2. Let ρ : G → GL (V ) be a representation. Recall that ρ⊗ρ =sym⊕alt,
where alt:G → GL (Λ2 V ) and sym:G → GL (S 2V ). Let us calculate χsym and χalt.
For each s ∈ G let λ1 , . . . , λn be eigenvalues of ρs taken with multiplicities. Then
the eigenvalues of alts are λi λj for all i < j and the eigenvalues of syms are λi λj for
i ≤ j. Hence
X X
χsym (s) = λi λj , χalt (s) = λi λj ,
i≤j i<j
and therefore X
λ2i = tr ρs2 = χρ s2 .

χsym (s) − χalt (s) =
i
On the other hand by properties (3) and (4)
χsym (s) + χalt (s) = χρ⊗ρ (s) = χ2ρ (s) .
Thus, we get
χ2ρ (s) + χρ (s2) χ2ρ (s) − χρ (s2 )
χsym (s) = , χalt (s) = .
2 2
Date: September 9, 2005.
1
2 VERA SERGANOVA
Starting from this point we assume that G is finite and k is algebraically closed of
characteristic 0.
Introduce the non-degenerate symmetric bilinear form on the space of functions
F (G) by the formula
1 X
f s−1 g (s) .

(f, g) =
|G| s∈G
Theorem 1.1. Let ρ, σ be irreducible. If ρ and σ are not isomorphic, then (χρ , χσ ) =
0. If ρ and σ are isomorphic, then (χρ, χσ ) = 1.
Proof. Let V be the space of the representation ρ and W be the space of σ. Denote
n = dim V , m = dim W . Choose a basis v1, . . . , vn in V , w1, . . . , wm in W . Define
P (i, j) : W → W by
P (i, j) vk = δjk wi .
Lemma 1.2. For any T ∈ Homk (V, W ) let
1 X
T̄ = σs ◦ T ◦ ρs−1 .
|G| s∈G
Then T̄ ∈ HomG (V, W ). If V = W , then tr T = tr T̄ .
Proof. Direct calculations.
For any T ∈ Hom (V, W ) let Tkl denote the corresponding matrix entry. For
example, P (i, j)kl = δik δjl. Then
1 X
P̄ (i, j)kl = (σs )ki (ρs−1 )jl .
|G| s∈G
If σ and ρ are not isomorphic, then by Schur’s Lemma
P̄ (i, j)kl = 0
for all i, j, k, l. In particular, P̄ (i, j)ij = 0 and therefore
m X
n m n
X 1 XXX
P̄ (i, j)ij = (σs )ii (ρs−1 )jj = 0.
i=1 j=1
|G| i=1 j=1 s∈G
But m X
n
X
(σs )ii (ρs−1 )jj = χσ (s) χρ s−1 .

i=1 j=1
Hence
1 X
χσ (s) χρ s−1 = (χρ , χσ ) = 0.

|G| s∈G
= σ. The by Property (2), we may assume ρ = σ. Then by Schur’s Lemma
Let now ρ ∼
P̄ (i, j) = λ Id .
REPRESENTATION THEORY WEEK2 3
Since tr P̄ (i, j) = tr P (i, j) = δij , we have

δij
P̄ (i, j) = Id .
n
Then
n X
n n X
n
X X δij
P̄ (i, j)ij = = 1,
i=1 j=1 i=1 j=1
n
which implies (χρ , χρ ) = 1.
Corollary 1.3. Let ρ = m1ρ1 ⊕ · · · ⊕ ml ρl be decomposition into the sum of irre-
ducibles. Then mi = (χρ , χρi ).
The number mi is called the multiplicity of an irreducible representation ρi in ρ.
Corollary 1.4. A representation ρ is irreducible iff (χρ , χρ ) = 1.
Corollary 1.5. Every irreducible representation ρ appears in the regular represen-
tation with multiplicity dim ρ.
Proof.
1
(χρ , χR) = χρ (1) χR (1) = dim ρ.
|G|

Corollary 1.6. Let ρ1 , . . . , ρl be all (up to isomorphism) irreducible representations
of G and ni = dim ρi . Then
n21 + · · · + n2l = |G|.
Example 3. Let G act on a finite set X and
( )
X
k (X) = bx x | bx ∈ k .
x∈X
Define ρ : G → GL (k (X)) by
X X
ρs bx x = bx s (x) .
x∈X x∈X
It is easy to check that ρ is a representation and

χρ (s) = | {x ∈ X | s (x) = x} |.
Clearly, ρ contains a trivial subrepresentation. To find the multiplicity of a trivial
representation in ρ we have to calculate (1, χρ ).
1 X 1 X X 1 XX 1 X
(1, χρ ) = χρ (s) = 1= 1= |Gx |,
|G| s∈G |G| s∈G |G| x∈X s∈G |G| x∈X
s(x)=x x
4 VERA SERGANOVA
where
Gx = {s ∈ G | s (x) = x} .
|G|
Let X = X1 ∪ · · · ∪ Xm be the disjoint union of orbits. Then |Gx | = |Xi |
for each
x ∈ Xi and therefore
m
1 X X |G|
(1, χρ ) = = m.
|G| i=1 x∈X |Xi |
i
Now let us evaluate (χρ , χρ ).

 2
1 X X 1 X X 1 X
(χρ, χρ ) =  1 = 1= |Gx ∩ Gy |.
|G| s∈G |G| s∈G |G|
s(x)=x s(x)=x,s(y)=y (x,y)∈X×X
Let σ be the representation associated with the action of G on X × X. Then the

last formula implies
(χρ, χρ ) = (1, χσ ) .
Thus, ρ is irreducible iff |X| = 1, and ρ has two irreducible components iff the action
of G on X × X with removed diagonal is transitive or |X| = 2.
Let
C (G) = f ∈ F (G) | f sts−1 = f (t) .

It is easy to check that the restriction of (,) on C (G) is non-degenerate.

Theorem 1.7. The characters of irreducible representations of G form an orthonor-
mal basis in C (G).
Proof. We have to show that if f ∈ C (G) and (f, χρ) = 0 for any irreducible ρ, then
f = 0. The following lemma is straightforward.
Lemma 1.8. Let ρ : G → GL (V ) be a representation, f ∈ C (G) and
1 X
f s−1 ρs .

T =
|G| s∈G
Then T ∈ EndG V and tr T = (f, χρ).
Thus, for any irreducible ρ we have
1 X
f s−1 ρs = 0.

(1.1)
|G| s∈G
But then the same is true for any representation ρ, since any representation is a
direct sum of irreducibles. Apply (?equ1?) for the case when ρ = R is the regular
representation. Then
1 X 1 X
f s−1 Rs 1 = f s−1 s = 0.

|G| s∈G |G| s∈G
Hence f (s−1 ) = 0 for all s ∈ G, i.e. f = 0
REPRESENTATION THEORY WEEK2 5
Corollary 1.9. The number of isomorphism classes of irreducible representations

equal the number of the conjugacy classes.
PROBLEM SET # 2
MATH 252
Due September 16.

In this problem set the field is algebraically closed and has zero characteristic, G
is finite and representations are finite-dimensional.
1. Show that the statement of Problem 3 in the first problem set is correct under
above assumptions.
2. Let ρ : G → GL (V ) be a representation. Show that each irreducible subrepre-
sentation of V has multiplicity 1 iff EndG V is a commutative ring.
3. Let G be the subgroup of quaternions of 8 elements, that contains ±1, ±i, ±j, ±k
with relations i2 = j 2 = k 2 = −1, ij = k, jk = i, ki = j, ij = −ji, ik = −ki,
jk = −kj. Classify irreducible representations of G over C.

1
WEEK 3
VERA SERGANOVA
1. Examples.
Example 1. Let G = S3. There are three conjugacy classes in G, which we
denote by some element in a class:1,(12),(123). Therefore there are three irreducible
representations, denote their characters by χ1, χ2 and χ3. It is not difficult to see
that we have the following table of characters
1 (12) (123)
χ1 1 1 1
χ2 1 −1 1
χ3 2 0 −1
The characters of one-dimensional representations are given in the first and the second
row, the last character χ3 can be obtained by using the identity
(1.1) χperm = χ1 + χ3,
where χperm stands for the character of the permutation representation.
Example 2. Let G = S4 . In this case we have the following character table (in
the first row we write the number of elements in each conjugacy class).
1 6 8 3 6
1 (12) (123) (12) (34) (1234)
χ1 1 1 1 1 1
χ2 1 −1 1 1 −1
χ3 3 1 0 −1 −1
χ4 3 −1 0 −1 1
χ5 2 0 −1 2 0
First two rows are characters of one-dimensional representations. The third can be
obtained again from (1.1), χ4 = χ2χ3 , the corresponding representation is obtained
as the tensor product ρ4 = ρ2 ⊗ ρ3. The last character can be found from the
orthogonality relation. Alternative way to describe ρ5 is to consider S4 /V4 , where
V4 = {1, (12) (34) , (13) (24) , (14) (23)}

1
2 VERA SERGANOVA
is the Klein subgroup. Observe that S4 /V4 ∼

= S3 , and therefore the two-dimensional
representation σ of S3 can be extended to the representation of S4 by the formula
ρ5 = σ ◦ p,
where p : S4 → S3 is the natural projection.
Solution of the marcian problem. Recall that S4 is isomorphic to the group
of rotations of a cube. Hence it acts on the set of faces F , and therefore we have a
representation
ρ : S4 → GL (C (F )) .
It is not difficult to calculate χρ using the formula
χρ (s) = | {x ∈ F | s (x) = x} |.
We obtain
χρ (1) = 6, χρ ((12)) = χρ ((123)) = 0, χρ ((12) (34)) = χρ ((1234)) = 2.
Furthermore,
(χρ , χ1) = (χρ , χ4) = (χρ, χ5) = 1, (χρ , χ2) = (χρ, χ3 ) = 0.
Hence χρ = χ1 + χ4 + χ5, and C (F ) = W1 ⊕ W2 ⊕ W3 the sum of three invariant
subspaces. The intertwining operator T : C (F ) → C (F ) of food redistribution must
be a scalar operator on each Wi by Schur’s Lemma. Note that
( )
X
W1 = ax | a ∈ C ,
x∈F
( )
X
W2 = axx | ax = −axop ,
x∈F
( )
X X
W3 = ax x | ax = 0, ax = axop ,
x∈F
where xop denotes the face opposite to the face x. A simple calculation shows that
T|W1 = Id, T|W2 = 0, T|W3 = − 12 Id. Therefore T n (v) approaches a vector in W1 as
n → ∞, and eventually everybody will have the same amount of food.
Example 3. Now let G = A5 . There are 5 irreducible representation of G over
C. Here is the character table
1 20 15 12 12
1 (123) (12) (34) (12345) (12354)
χ1 1 1 1 1 1
χ2 4 1 0 −1 −1
χ3 5 −1 1 0√ 0√
1+ 5 1− 5
χ4 3 0 −1 2√ 2√
1− 5 1+ 5
χ5 3 0 −1 2 2
REPRESENTATION THEORY. WEEK 3 3
To obtain χ2 use the permutation representation and (1.1) again. Let χsym and
χalt be the characters of the second symmetric and the second exterior powers of ρ2
respectively. Then we obtain
1 (123) (12) (34) (12345) (12354)
χsym 10 1 2 0 0
χalt 6 0 −2 1 1
It is easy to check that

(χsym , χsym) = 3, (χsym , χ1) = (χsym, χ2) = 1.
Therefore
χ3 = χsym − χ1 − χ2
is the character of an irreducible representation.
Furthermore,
(χalt, χalt) = 2, (χalt, χ1) = (χalt, χ2) = (χalt, χ3) = 0.
Therefore χalt = χ4 + χ5 is the sum of two irreducible characters. First we find the
dimensions of ρ4 and ρ5 using
12 + 42 + 52 + n24 + n25 = 60.
We obtain n4 = n5 = 3. The equations
(χ4 , χ1 + χ2) = 0, (χ4, χ3 ) = 0
imply
χ4 ((123)) = 0, χ4 ((12) (34)) = −1.
The same argument shows
χ5 ((123)) = 0, χ5 ((12) (34)) = −1.
Finally denote
x = χ4 ((12345)) , y = χ4 ((12354))
and write down the equation (χ4, χ4) = 1. It is
1
9 + 15 + 12x2 + 12y 2 = 1,

60
or
(1.2) x2 + y 2 = 3.
On the other hand, (χ4 , χ1) = 0, that gives
3 − 15 + 12 (x + y) = 0,
or
(1.3) x + y = 1.
4 VERA SERGANOVA
One can solve (1.2) and (1.3)

√ √
1+ 5 1− 5
x= ,y = .
2 2
The second solution √ √
1− 5 1+ 5
x= ,y =
2 2
will give the last character χ5.
2. Modules
Let R be a ring, usually we assume that 1 ∈ R. An abelian group M is called
a (left) R-module if there is a map α : R × M → M, (we write α (a, m) = am )
satisfying
(1) (ab) m = a (bm);
(2) 1m = m;
(3) (a + b) m = am + bm;
(4) a (m + n) = am + an.
Example 1. If R is a field then any R-module M is a vector space over R.
Example 2. If R = k (G) is a group algebra, then every R-module defines the
representation ρ : → GL (V ) by the formula
ρs v = sv
for any s ∈ G ⊂ k (G), v ∈ V . Conversely, every representation ρ : G → GL (V ) in a
vector space V over k defines on V a k (G)-module structure by
!
X X
as s v = as ρs v.
s∈G s∈G
Thus, representations of G over k are k (G)-modules.

A submodule is a subgroup invariant under R-action. If N ⊂ M is a submodule
then the quotient M/N has the natural R-module structure. A module M is simple
or (irreducible) if any submodule is either zero or M. A sum and an intersection of
submodules is a submodule.
Example 3. If R is an arbitrary ring, then R is a left module over itself, where
the action is given by the left multiplication. Submodules are left ideals.
For any two R-modules M and N one can define an abelian group HomR (M, N)
and a ring EndR (M) in the manner similar to the group case. Schur’s Lemma holds
for simple modules in the following form.
Lemma 2.1. Let M and N be simple modules and ϕ ∈ HomR (M, N), then either
ϕ is an isomorphism or ϕ = 0. For a simple module M, EndR (M) is a division ring.
Recall that for every ring R one defines Rop as the same abelian group with new
multiplication * given by
a ∗ b = ba.
Lemma 2.2. The ring EndR (R) is isomorphic to Rop .
Proof. For each a ∈ R define ϕa ∈ End (R) by the formula
ϕa (x) = xa.
It is easy to check that ϕa ∈ EndR (R) and ϕba = ϕa ◦ ϕb . Hence we constructed a
homomorphism ϕ : Rop → EndR (R). To prove that ϕ is injective let ϕa = ϕb . Then
ϕa (1) = ϕb (1), i.e. a = b. To prove surjectivity of ϕ, note that for any γ ∈ EndR (R)
one has
γ (x) = γ (x1) = xγ (1) .
Therefore γ = ϕγ(1).
Lemma 2.3. Let ρi : G → GL (Vi ), i = 1, . . . , l be pairwise non-isomorphic irre-
ducible representations of a group G over algebraically closed field k, and
V = V1⊕m1 ⊕ · · · ⊕ Vl⊕ml .
Then
= Endk (k m1 ) × · · · × Endk (k ml ) .
EndG (V ) ∼
Proof. First, note that the Schur’s Lemma implies that ϕ Vi⊕mi ⊂ Vi⊕mi for any

ϕ ∈ EndG (V ), i = 1, . . . , l. Hence
⊕m
= EndG V1⊕m1 × · · · × EndG Vl l .

EndG (V ) ∼
Therefore it suffices to prove the following
Lemma 2.4. For an irreducible representation of G in W
EndG W ⊕m ∼ = Endk (k m ) .

Proof. Let pi : W ⊕m → W denotes the projection onto the i-th component and
qj : W → W ⊕m be the embedding of the j-th component. Let ϕ ∈ EndG (W ⊕m ).
Then by Schur’s Lemma pi ◦ ϕ ◦ qj = ϕij Id for some ϕij ∈ k. Therefore we have the
map Φ : EndG (W ⊕m ) → Endk (k m ), (the latter is just the matrix ring) defined by
Φ (ϕ) = (ϕij ). Check yourself that Φ is an isomorphism.

Theorem 2.5. Let k be algebraically closed, char k = 0. Then
= Endk (k n1 ) × · · · × Endk (k nl ) ,
k (G) ∼
where n1 , . . . , nl are dimensions of irreducible representations.
6 VERA SERGANOVA
Proof. By Lemma 2.2

op
= k (G) ∼
Endk(G) (k (G)) ∼ = k (G) ,
op
since k (G) = k (G) via g → g −1 . On the other hand
∼
k (G) = V1⊕n1 ⊕ · · · ⊕ Vl⊕nl ,

since every irreducible ρi : G → GL (Vi ) appears with the multiplicity ni = dim Vi .
Therefore Lemma 2.3 implies theorem.
3. Finitely generated modules and Noetherian rings.

A module M is finitely generated if there exist x1, . . . , xn ∈ M such that M =
Rx1 + · · · + Rxn .
Lemma 3.1. Let
q p
0→N −
→M −
→L→0
be an exact sequence of R-modules. If M is finitely generated, then L is finitely
generated. If M and L are finitely generated, then M is finitely generated.
Proof. First assertion is obvious. For the second let
L = Rx1 + . . . Rxn , N = Ry1 + . . . Rym ,
then M = Rp−1 (x1) + · · · + Rp−1 (xn ) + Rq (y1 ) + · · · + Rq (ym ).
Lemma 3.2. The following conditions on a ring R are equivalent
(1) Every increasing chain of left ideals is finite, in other words for any sequence
I1 ⊂ I2 ⊂ . . . , In = In+1 = In+2 = . . . starting from some n;
(2) Every left ideal is finitely generated R-module.
Proof. (1)⇒ (2). Assume that some left ideal I is not finitely generated. Then there
exists an infinite sequence of xn ∈ I such that
xn+1 ∈
/ Rx1 + . . . Rxn .
But then In = Rx1 + . . . Rxn form an infinite increasing chain of ideals.
(2)⇒ (1). Let I1 ⊂ I2 ⊂ . . . be an increasing chain of ideals . Let I = ∪n In .
Then I = Rx1 + . . . Rxs , where xj ∈ Inj . Let m be maximal among n1 , . . . , ns . Then
I = Im , and therefore the chain is finite.
A ring satisfying the conditions of Lemma 3.2 is called left Noetherian.

Lemma 3.3. Let R be a left Noetherian ring and M be a finitely generated R-
module. Then every submodule of M is finitely generated.
Proof. Let M = Rx1 + . . . Rxn , the there exists a surjective homomorphism p :

R ⊕ · · · ⊕ R → M, such that
p (r1 , . . . , rn ) = r1 s1 + · · · + rn sn .
As follows from the first part of Lemma 3.1, it suffices to prove the statement for a
free module. It can be done by induction using the second part of Lemma 3.1.

Let R be a commutative ring. An element x ∈ R is called integral over Z if

xn + an−1 x + · · · + a0 = 0 for some ai ∈ Z. This condition is equivalent to the
condition that Z [x] ⊂ R is finitely generated Z-module. Complex numbers integral
over C are usually called algebraic integers. Obviously, if a rational number z is
algebraic integer, then z ∈ Z.
Lemma 3.4. The set of integral elements in a commutative ring R is a subring.
Proof. If Z [x] and Z [y] are finitely generated over Z, then Z [x, y] is also finitely
generated. Let s ∈ Z [x, y], then Z [s] is finitely generated since Z is Noetherian ring
and we can apply Lemma 3.3.

4. The center of the group algebra k (G)

We have assumptions char k = 0, k̄ = k, G is a finite group. Let Z (G) denote the
center of k (G). It is obvious that
( )
X
Z (G) = f (s) s | f ∈ C (G) .
s∈G
On the other hand, by Theorem 2.5 we have

k (G) = Endk (k n1 ) × · · · × Endk (k nl ) .
Therefore Z (G) is isomorphic to k l as a commutative ring. Let ei denote the identity
element in Endk (k ni ). Then e1, . . . , el form a basis in Z (G) and
ei ej = δij ei , 1 = e1 + · · · + el .
For an irreducible representation ρj : G → GL (Vj ) we have
(4.1) ρj (ei ) = δij Id .
Lemma 4.1. If χi = χρi , ni = dim Vi , then
ni X
χi s−1 s.

(4.2) ei =
|G|
8 VERA SERGANOVA
Proof. We need to check (4.1). Since ρj (ei ) ∈ EndG (Vj ), we have ρj (ei ) = λId. To
find λ calculate
ni X ni
χi s−1 χj (s) =

tr ρj (ei ) = (χi , χj ) = δij ni .
|G| |G|

Lemma 4.2. Define ωi : Z (G) → k by the formula
X 1 X
ωi as s = as χi (s) .
ni
Then ωi is a homomorphism of rings and
ω = (ω1 , . . . , ωl ) : Z (G) → k l
is an isomorphism.
Proof. Check that ωi (ej ) = δij using again the orthogonality relation.
P
Lemma 4.3. Let u = as s ∈ Z (G). If all as are algebraic integers, then u is
integral over Z.
Proof. Let c ⊂ G be some conjugacy class and let
X
δc = s.
s∈c
If c1 , . . . , cl are disjoint conjugacy classes, then clearly Zδc1 + · · · + Zδcl is a subring in

Z (G). On the other hand, it is clearly a finitely generated Z-module, and therefore
every element of it isPintegral over Z. But then for any set of algebraic integers
b1 , . . . , bl the element bi δci is integral over Z, which proves Lemma.
Theorem 4.4. The dimension of an irreducible representation divides |G|.
Proof.
P For every s ∈ G, χi (s) is an algebraic integer. Therefore by Lemma 4.3
−1
u = s∈G χi (s ) s is integral over Z. Hence ωi (u) is an algebraic integer. But
1 X |G| |G|
ωi (u) = χi (s) χi s−1 = (χi , χi) = .
ni ni ni
|G|
Therefore ni
∈ Z.
Theorem 4.5. Let Z be the center of G. The dimension n of an irreducible repre-
sentation divides |G|
|Z|
.
Proof. Consider
ρm = ρ⊠m : G × · · · × G → GL V ⊗m .

Then Ker ρm contains a subgroup

′
Zm = {(z1, . . . , zm) ∈ Z m | z1z2 . . . zm = 1} .
If ρ is irreducible, then ρm is irreducible, and dim ρm = (dim ρ)m divides |Gm /Zm ′
|=
|G| m |G|
|Z|m−1
. Since this is true for any m, then dim ρ divides |Z| (check yourself using prime
factorization).
PROBLEM SET # 3
MATH 252
Due September 23.

1. Consider the action of the group A5 on the faces of a dodecahedron. Decompose
the corresponding representation of A5 into a sum of irreducibles and solve the Venus
problem by diagonalizing the intertwining operator.
2. Let Dn denote the dihedral group, which is the group of all symmetries of a
regular n-gon. Classify irreducible representations of Dn over C.

1
WEEK 4
VERA SERGANOVA
1. Induced modules
Let B ⊂ A be rings and M be a B-module. Then one can construct induced
module IndA
B M = A ⊗B M as the quotient of a free abelian group with generators
from A × M by relations
(a1 + a2) × m − a1 × m − a2 × m, a × (m1 + m2 ) − a × m1 − a × m2, ab × m − a × bm,
and A acts on A ⊗B M by left multiplication. Note that j : M → A ⊗B M defined by
j (m) = 1 ⊗ m
is a homomorphism of B-modules.
Lemma 1.1. Let N be an A-module, then for ϕ ∈ HomB (M, N) there exists a
unique ψ ∈ HomA (A ⊗B M, N) such that ψ ◦ j = ϕ.
Proof. Clearly, ψ must satisfy the relation
ψ (a ⊗ m) = aψ (1 ⊗ m) = aϕ (m) .
It is trivial to check that ψ is well defined.
Exercise. Prove that for any B-module M there exists a unique A-module satisfying
the conditions of Lemma 1.1.
Corollary 1.2. (Frobenius reciprocity.) For any B-module M and A-module N
there is an isomorphism of abelian groups
HomB (M, N) ∼ = HomA (A ⊗B M, N) .
Example. Let k ⊂ F be a field extension. Then induction IndFk is an exact
functor from the category of vector spaces over k to the category of vector spaces
over F , in the sense that the short exact sequence
0 → V1 → V2 → V3 → 0
becomes an exact sequence
0 → F ⊗k V1 ⊗ → F ⊗k V2 → F ⊗k V3 → 0.
1
2 VERA SERGANOVA
In general, the latter property is not true. It is not difficult to see that induction is
right exact, i.e. an exact sequence of B-modules
M →N →0
induces an exact sequence of A-modules
A ⊗B M → A ⊗B N → 0.
But an exact sequence
0→M →N
is not necessarily exact after induction.
Later we discuss general properties of induction but now we are going to study
induction for the case of groups.
2. Induced representations for groups.

Let H be a subgroup of G and ρ : H → GL (V ) be a representation. Then the
induced representation IndG
H ρ is by definition a k (G)-module
k (G) ⊗k(H) V.
Lemma 2.1. The dimension of IndG H ρ equals the product of dim ρ and the index
[G : H] of H. More precisely, let S is a set of representatives of left cosets, i.e.
a
G= sH,
s∈S
then
(2.1) k (G) ⊗k(H) V = ⊕s∈S s ⊗ V.
For any t ∈ G, s ∈ S there exist unique s′ ∈ S, h ∈ H such that ts = s′h and the
action of t is given by
(2.2) t (s ⊗ v) = s′ ⊗ ρh v.
Proof. Straightforward check.
Lemma 2.2. Let χ = χρ and IndG G
H χ denote the character of IndH ρ. Then
X 1 X
IndG −1 −1

(2.3) H χ (t) = χ s ts = χ s ts .
|H|
−1s∈S,s ts∈H −1 s∈G,s ts∈H
Proof. (2.1) and (2.2) imply
X
IndG
H χ (t) = χ (h) .
s∈S,s′ =s
Since s = s′ implies h = s−1 ts ∈ H, we obtain the formula for the induced character.
Note also that χ (s−1 ts) does not depend on a choice of s in a left coset.
Corollary 2.3. Let H be a normal subgroup in G. Then IndG

H χ (t) = 0 for any
t∈
/ H.
Theorem 2.4. For any ρ : G → GL (V ), σ : H → GL (W ), we have the identity
IndG

(2.4) H χσ , χρ G = (χσ , ResH χρ)H .
Here a subindex indicates the group where we take inner product.

Proof. It follows from Frobenius reciprocity, since
dim HomG IndG

H W, V = dim HomH (W, V ) .

Note that (2.4) can be proved directly from (2.3). Define two maps
ResH : C (G) → C (H) , IndG
H : C (H) → C (G) ,
the former is the restriction on a subgroup, the latter is defined by (2.3). Then for
any f ∈ C (G) , g ∈ C (H)
IndG

(2.5) H g, f G = (g, ResH f)H .
Example 1. Let ρ be a trivial representation of H. Then IndG H ρ is the permutation

representation of G obtained from the natural left action of G on G/H (the set of
left cosets).
Example 2. Let G = S3 , H = A3 , ρ be a non-trivial one dimensional representa-
tion of H (one of two possible). Then
IndG G G
H χρ (1) = 2, IndH χρ (12) = 0, IndH χρ (123) = −1.
Thus, by induction we obtain an irreducible two-dimensional representation of G.

Now consider another subgroup K of G = S3 generated by the transposition (12),
and let σ be the (unique) non-trivial one-dimensional representation of K. Then
IndG G G
K χσ (1) = 3, IndK χσ (12) = −1, IndH χρ (123) = 0.
3. Double cosets and restriction to a subgroup

If K and H are subgroups of G one can define the equivalence relation on G : s ∼ t
iff s ∈ KtH. The equivalence classes are called double cosets. We can choose a set of
representative T ⊂ G such that
a
G= K tH .
s∈T
We define the set of double cosets by K\G/H. One can identify K\G/H with K-
orbits on S = G/H in the obvious way and with G-orbits on G/K × G/H by the
formula
KtH → G (K, tH) .
4 VERA SERGANOVA
def
Example. Let Fq be a field of q elements and G = GL2 (Fq ) = GL F2q . Let B be

the subgroup of upper-triangular matrices in G. Check that |G| = (q 2 − 1) (q 2 − q),

|B| = (q − 1)2 q and therefore [G : B] = q + 1. Identify G/B with the set of lines P1
in F2q and B\G/B with G-orbits on P1 × P1 . Check that G has only two orbits on
P1 × P1 : the diagonal and its complement. Thus, |B\G/B| = 2 and
G = B ∪ BsB,
where
0 1
s=
1 0
`
Theorem 3.1. Let T ⊂ G such that G = s∈T KtH. Then
ResK IndG K s
H ρ = ⊕s∈T IndK∩sHs−1 ρ ,
where
ρsh = ρs−1 hs ,
def
for any h ∈ sHs−1 .

Proof. Let s ∈ T and W s = k (K) (s ⊗ V ). Then by construction, W s is K-invariant
and
k (G) ⊗k(H) V = ⊕s∈T W s .
Thus, we need to check that the representation of K in W s is isomorphic to IndK s
K∩sHs−1 ρ .
We define a homomorphism
α : IndK
K∩sHs−1 V → W
s
by α (t ⊗ v) = ts ⊗ v for any t ∈ K, v ∈ V . It is well defined

α (th ⊗ v − t ⊗ ρsh v) = ths ⊗ v − ts ⊗ ρs−1 hs v = ts s−1 hs ⊗ v − ts ⊗ ρs−1 hs v = 0

and obviously surjective. Injectivity can be proved by counting dimensions.

Example. Let us go back to our example B ⊂ SL2 (Fq ). Theorem 3.1 tells us that
for any representation ρ of B
IndG G ′
B ρ = ρ ⊕ IndH ρ ,
where H = B ∩ sBs−1 is a subgroup of diagonal matrices and

′ a 0 b 0
ρ =ρ
0 b 0 a
Corollary 3.2. If H is a normal subgroup of G, then
ResH IndG s
H ρ = ⊕s∈G/H ρ .
4. Mackey’s criterion
To find IndG G

H χ, IndH χ we can use Frobenius reciprocity and Theorem 3.1.
X
IndG G G
IndH s

H χ, IndH χ G
= ResH IndH χ, χ H
= H∩sHs−1 χ , χ H =
s∈T
X X
= (χs , ResH∩sHs−1 χ)H∩sHs−1 = (χ, χ)H + (χs , ResH∩sHs−1 χ)H∩sHs−1 .
s∈T s∈T \{1}
We call two representation disjoint if they do not have the same irreducible com-
ponent, i.e. their characters are orthogonal.
Theorem 4.1. (Mackey’s criterion) IndG H ρ is irreducible iff ρ is irreducible and ρ
s
−1
and ρ are disjoint representations of H ∩ sHs for any s ∈ T \ {1}.
Proof. Write the condition
IndG G

H χ, IndH χ G
=1
and use the above formula.
Corollary 4.2. Let H be a normal subgroup of G. Then IndG
H ρ is irreducible iff ρ
s
is not isomorphic to ρ for any s ∈ G/H, s ∈

/ H.
Remark 4.3. Note that if H is normal, then G/H acts on the set of representations
of H. In fact, this is a part of the action of the group Aut H of automorphisms of
H on the set of representation of H. Indeed, if ϕ ∈ Aut H and ρ : H → GL (V ) is a
representation, then ρϕ : H → GL (V ) defined by
ρϕt = ρϕ(t),
is a new representation of H.
5. Some examples
Let H be a subgroup of G of index 2. Then H is normal and G = H ∪ sH for
some s ∈ G\H. Suppose that ρ is an irreducible representation of H. There are two
possibilities
(1) ρs is isomorphic to ρ;
(2) ρs is not isomorphic to ρ.
Hence there are two possibilities for IndG
Hρ :
G ′ ′
(1) IndH ρ = σ ⊕ σ , where σ and σ are two non-isomorphic irreducible represen-
tations of G;
(2) IndG H ρ is irreducible.
For instance, let G = S5 , H = A5 and ρ1 , . . . , ρ5 be irreducible representation of
H, where the numeration is from lecture notes week 3. Then for i = 1, 2, 3
IndG
H ρi = σi ⊕ (σi ⊗ sgn) ,
6 VERA SERGANOVA
here sgn denotes the sign representation. Furthermore, IndG G

H ρ4 = IndH ρ5 is ir-
∼
reducible. Thus S5 has two 1, 5, 4-dimensional irreducible representations and one
6-dimensional.
Now let G be a subgroup of GL2 (Fq ) of matrices

a b
0 1
We want to classify irreducible representations of G over C. |G| = q 2 − q, G has the
following conjugacy classes

1 0 1 1 a 0
, , ,
0 1 0 1 0 1
in the last case a 6= 1. Note that the subgroup H of matrices

1 b
0 1
is normal, G/H ∼ = F∗q ∼= Zq−1 . Therefore G has q − 1 one-dimensional representations
which can be lifted from G/H. That leaves one more representation, its dimension
must be q − 1. We hope to obtain it by induction from H. Let σ be a non-trivial
irreducible representation of H (one-dimensional). Then dim IndG H σ = q − 1 as
required. Note that for any previously constructed one-dimensional representation ρ
of G we have
IndG

H σ, ρ G = (σ, ResH ρ)H = 0,
as ResH ρ is trivial. Therefore IndG
H σ is irreducible. The character takes values q − 1,
−1 and 0 on the corresponding conjugacy classes.
Remark 5.1. To find all one-dimensional representation of a group G, find its com-
mutator G′ , which is a subgroup generated by ghg −1 h−1 for all g, h ∈ G. One-
dimensional representations of G are lifted from one-dimensional representations of
G/G′ .
PROBLEM SET # 4
MATH 252
Due September 30.

1. Let H ⊂ K ⊂ G. Show that
IndG K G
K IndH ρ = IndH ρ
∼
for any representation ρ of H.
2. Let G be the group of matrices
 
1 x y
 0 1 z 
0 0 1
where x, y, z are elements of the finite field F5 . Classify irreducible representations
of G over C.

1
WEEK 5
1. Invariant forms
Recall that a bilinear form on a vector space V is a map
B :V ×V →k
satisfying
B (cv, dw) = cdB (v, w) , B (v1 + v2 , w) = B (v1, w)+B (v2 , w) , B (v, w1 + w2 ) = B (v, w1)+B (v, w2) .
One can also think about a bilinear form as a vector in V ∗ ⊗ V ∗ or as a homo-
morphism B : V → V ∗ given by the formula Bv (w) = B (v, w). A bilinear form is
symmetric if B (v, w) = B (w, v) and skew-symmetric if B (v, w) = −B (w, v). Every
bilinear form is a sum B = B + + B − of a symmetric and a skew-symmetric form,
B (v, w) ± B (w, v)
B ± (v, w) = .
2
Such decomposition corresponds to the decomposition
(1.1) V ∗ ⊗ V ∗ = S 2 V ∗ ⊕ Λ2 V ∗ .
A form is non-degenerate if B : V → V ∗ is an isomorphism, in other words B (v, V ) =
0 implies v = 0.
Let ρ : G → GL (V ) be a representation. We say that a bilinear form B on V is
G-invariant if
B (ρs v, ρs w) = B (v, w)
for any v, w ∈ V , s ∈ G.
The following properties of an invariant form are easy to check
(1) If W ⊂ V is an invariant subspace, then W ⊥ = {v ∈ V | B (v, W ) = 0} is
invariant. In particular, Ker B is invariant.
(2) B : V → V ∗ is invariant iff B ∈ HomG (V, V ∗ ).
(3) If B is invariant, then B + and B − are invariant.
Lemma 1.1. Let ρ be an irreducible representation of G, then any bilinear invariant
form is non-degenerate. If k̄ = k, then a bilinear form is unique up to multiplication
on a scalar.
Proof. Follows from (2) and Schur’s lemma.

1
2 REPRESENTATION THEORY WEEK 5
Corollary 1.2. A representation ρ of G admits an invariant form iff χρ (s) = χρ (s−1 )

for any s ∈ G.
Lemma 1.3. If k̄ = k, then an invariant form on an irreducible representation ρ is

either symmetric or skew-symmetric. Let
1 X
χ ρ s2 .

mρ =
|G| s∈G
Then mρ = 1, 0 or −1. If mρ = 0, then ρ does not admit an invariant form. If

mρ = ±1, then mρ admits a symmetric (skew-symmetric) invariant form.
Proof. Recall that ρ ⊗ ρ = ρalt ⊕ ρsym .
1 X χρ (s2 ) + χρ (s2)
(χsym , 1) = ,
|G| s∈G 2
1 X χρ (s2 ) − χρ (s2)
(χalt, 1) = .
|G| 2
s∈G
Note that
1 X
χρ s2 = (χρ , χρ∗ ) .

|G|
s∈G
Therefore
(χρ , χρ∗ ) + mρ (χρ, χρ∗ ) − mρ
(χsym, 1) = , (χalt, 1) =
2 2
If ρ does not have an invariant form, then (χsym, 1) = (χalt, 1) = 0, and χρ∗ 6= χρ ,
hence (χρ , χρ∗ ) = 0. Thus, mρ = 0.
If ρ has a symmetric invariant form, then (χρ , χρ∗ ) = 1 and (χsym, 1) = 1. This
implies mρ = 1. Similarly, if ρ admits a skew-symmetric invariant form, then mρ =
−1.
Let k = C. An irreducible representation is called real if mρ = 1, complex if mρ = 0

and quaternionic if mρ = −1. Since χρ (s−1 ) = χ̄ρ (s), then χρ takes only real values
for real and quaternionic representations. If ρ is complex then χρ (s) ∈/ R at least for
one s ∈ G.
Example. Any irreducible representation of S4 is real. A non-trivial representa-
tion of Z3 is complex. The two-dimensional representation of quaternionic group is
quaternionic.
Exercise. Let |G| be odd. Then any non-trivial irreducible representation of G
over C is complex.
REPRESENTATION THEORY WEEK 5 3
2. Some generalities about field extension

Lemma 2.1. If char k = 0 and G is finite, then a representation ρ : G → GL (V ) is
irreducible iff EndG (V ) is a division ring.
Proof. In one direction it is Schur’s Lemma. In the opposite direction if V is not
irreducible, then V = V1 ⊕ V2 , then the projectors p1 and p2 are intertwiners such
that p1 ◦ p2 = 0.
For any extension F of k and a representation ρ : G → GL (V ) over k we define
by ρF the representation G → GL (F ⊗k V ).
For any representation ρ : G → GL (V ) we denote by V G the subspace of G-
invariants in V , i.e.
V G = {v ∈ V | ρs v = v, ∀s ∈ G} .
Lemma 2.2. (F ⊗k V )G = F ⊗k V G .
Proof. The embedding F ⊗k V G ⊂ (F ⊗k V )G is trivial. On the other hand, V G is
the image of the operator
1 X
p= τs ,
|G| s∈G
in particular dim V G equals the rank of p. Since rank p does not depend on a field,
we have
G
dim F ⊗k V G = dim (F ⊗k V ) .

Corollary 2.3. Let ρ : G → GL (V ) and σ : G → GL (W ) be two representations
over k. Then
HomG (F ⊗k V, F ⊗k W ) = F ⊗ HomG (V, W ) .
In particular,
dimk HomG (V, W ) = dimF HomG (F ⊗k V, F ⊗k W ) .
Proof.
G
HomG (V, W ) = (V ∗ ⊗ W ) .

Corollary 2.4. Even if a field is not algebraically closed
dim HomG (V, W ) = (χρ , χσ ) .
A representation ρ : G → GL (V ) over k is called absolutely irreducible if it remains
irreducible after any extension of k. This is equivalent to (χρ , χρ) = 1. A field is
splitting for a group G if any irreducible representation is absolutely irreducible. It
is not difficult to see that some finite extension of Q is a splitting field for a finite
group G.
3. Representations over R
A bilinear symmetric form B is positive definite if B (v, v) > 0 for any v 6= 0.
Lemma 3.1. Every representation of a finite group over R admits positive-definite
invariant symmetric form. Two invariant symmetric forms on an irreducible repre-
sentation are proportional.
Proof. Let B ′ be any positive definite form. Define
1 X ′
B (v, w) = B (ρs v, ρs w) .
|G|
s∈G
Then B is positive definite and invariant.

Let Q (v, w) be another invariant symmetric form. Then from linear algebra we
know that they can be diagonalized in the same basis. Then for some λ ∈ R,
Ker (Q − λB) 6= 0. Since Ker (Q − λB) is invariant, Q = λB.
Theorem 3.2. Let R ⊂ K be a division ring, finite-dimensional over R. Then R is
isomorphic R, C or H (quaternions).
Proof. If K is a field, then K ∼
= R or C, because C = R̄ and [C : R] = 2. Assume
that K is not commutative. For any x ∈ K\R, R [x] = C. Therefore we have a chain
R ⊂ C ⊂ K. Let f (x) = ixi−1. Obviously f is an automorphism of K and f 2 = id.
Hence K = K + ⊕ K − , where
K ± = {x ∈ K | f (x) = ±x} .
Moreover, K + K + ⊂ K + , K − K − ⊂ K + , K + K − ⊂ K − , K − K + ⊂ K − . If x ∈ K + ,
then C [x] is a finite extension of C. Therefore K + = C. For any nonzero y ∈ K −
the left multiplication on y defines an isomorphism of K + and K − as vector spaces
over R. In particular dimR K − = dimR K + = 2. For any y ∈ K − , x ∈ C, we have
y x̄ = xy, therefore y 2 ∈ R. Moreover, y 2 < 0. (If y 2 > 0, then y 2 = b2 for some
real b and (y − b) (y + b) = 0, which is impossible). Put j = √ y 2 . Then we have
−y
k = ij = −ji, ki = (ij) i = j, K = R [i, j] is isomorphic to H.
Lemma 3.3. Let ρ : G → GL (V ) be an irreducible representation over R, then there
are three possibilities:
(1) EndG (V ) = R and (χρ, χρ ) = 1;
= C and (χρ, χρ ) = 2;
(2) EndG (V ) ∼
= H and (χρ , χρ ) = 4.
(3) EndG (V ) ∼
Proof. Lemma 2.1 and Theorem 3.2 imply that EndG (V ) is isomorphic to R, C or H,
(χρ , χρ ) = 1, 2 or 4 as follows from Corollary 2.4.

4. Relationship between representations over R and over C

Hermitian invariant form. Recall that a Hermitian form satisfies the following
conditions
H (av, bw) = ābH (v, w) , H (w, v) = H̄ (v, w) .
The following Lemma can be proved exactly as Lemma 3.1.
Lemma 4.1. Every representation of a finite group over C admits positive-definite
invariant Hermitian form. Two invariant Hermitian forms on an irreducible repre-
sentation are proportional.
Let ρ : G → GL (V ) be a representation over C. Denote by V R a vector space V
as a vector space over R of double dimension. Denote by ρR the representation of G
in V R . Check that
(4.1) χρR = χρ + χ̄ρ .
Theorem 4.2. Let ρ : G → GL (V ) be an irreducible representation over C.
(1) If ρ can be realized by matrices with real entries, then ρ admits an invariant
symmetric form.

(2) If EndG V R = C, then ρ is complex, i.e. ρ does not admit a bilinear invariant
symmetric form.

(3) If EndG V R = H, then ρ admits an invariant skew-symmetric form.
Proof. (1) follows from Lemma 3.1. For (2) use (4.1). Since (χρ , χρ) = 2 by
Lemma 3.3, then χρ 6= χ̄ρ , and therefore ρ is complex.

Finally let us prove (3). Let j ∈ EndG V R = H, then j (bv) = b̄v for any b ∈ C.
Let H be a positive-definite Hermitian form on V . Then
Q (v, w) = H (jw, jv)
is another invariant positive-definite Hermitian form. By Lemma 4.1 Q = λH for
some λ > 0. Since j 2 = −1, λ2 = 1 and therefore λ = 1. Thus,
H (v, w) = H (jw, jv) .
Set
B (v, w) = H (jv, w) .
Then B is a bilinear invariant form, and
B (w, v) = H (jw, v) = H jv, j 2w = −H (jv, w) = −B (v, w) ,

hence B is skew-symmetric.
Corollary 4.3. Let σ be an irreducible representation of G over R. There are three
possibilities for σ
σ is absolutely irreducible and χσ = χρ for some real representation ρ of G over C;
χσ = χρ + χ̄ρ for some complex representation ρ of G over C;
χσ = 2χρ for some quaternionic representation ρ of G over C.
5. Representations of symmetric group

Let A denote the group algebra Q (Sn ). We will see that Q is a splitting field for
Sn . We realize irreducible representation of Sn as minimal left ideals in A.
Conjugacy classes are enumerated by partitions m1 ≥ · · · ≥ mk > 0, m1 + · · · +
mk = n. To each partition we associate the table of n boxes with rows of length
m1 , . . . , mk , it is called a Young diagram. Young tableau is a Young diagram with
entries 1, . . . , n in boxes. Given a Young tableau λ, we denote by Pλ the subgroup
of permutations preserving rows and by Qλ the subgroup of permutations preserving
columns. Introduce the following elements in A
X X
aλ = p, bλ = (−1)q q, cλ = aλbλ .
p∈Pλ q∈Qλ
The element cλ is called Young symmetrizer.

Theorem 5.1. Vλ = Acλ is a minimal left ideal in A, therefore Vλ is irreducible. Vλ
is isomorphic to Vµ iff the Young tableaux µ and λ have the same Young diagram.
Any irreducible representation of Sn is isomorphic to Vλ for some Young tableau λ.
Note that the last assertion of Theorem follows from the first two, since the number
of Young diagrams equals the number of conjugacy
P classes.
Examples. For partition P (n), cλ =s
aλ = s∈Sn s, and the representation is trivial.
For (1,. . . ,1), cλ = bλ = s∈Sn (−1) s.
Let us consider partition (n − 1, 1). Then
 
X
cλ =  s (1 − (1n)) .
s∈Sn−1
Clearly, aλcλ = cλ, therefore ResSn−1 Vλ contains the trivial representation. Let
V = IndSSnn−1 (triv) .
Note that V is the permutation representation of Sn . By Frobenius reciprocity we
have a homomorphism V → Vλ . Therefore V = Vλ ⊕triv.
Now we will prove Theorem 5.1. First, note that Sn acts on the Young tableaux
of the same shape, and
as(λ) = saλs−1 , bs(λ) = sbλs−1 , cs(λ) = scλ s−1 .
Check yourself the following
Lemma 5.2. If s ∈ Sn , but s ∈
/ Pλ Qλ, then there exists two numbers i and j in the
same row of λ and in the same column of s (λ).
It is clear also that for any p ∈ Pλ , q ∈ Qλ
paλ = aλ p = aλ, qbλ = bλ q = (−1)q bλ, pcλ q = (−1)q cλ.
Lemma 5.3. Let y ∈ A such that for any p ∈ Pλ , q ∈ Qλ

q
pyq = (−1) y.
Then y ∈ Qcλ .
Proof. It is clear that y has a form
X X X
ds (−1)q psq = ds aλsbλ,
s∈Pλ \Sn /Qλ p∈Pλ ,q∈Qλ s∈Pλ \Sn /Qλ
for some ds ∈ Q. We have to show that if s ∈/ Pλ Qλ then aλsbλ = 0. That follows

from Lemma 5.2. There exists (ij) ∈ Pλ ∩ Qs(λ) . Then
aλ sbλs−1 = aλbs(λ) = aλ (ij) (ij) bs(λ) = aλbs(λ) = −aλbs(λ) = 0.

Corollary 5.4. cλ Acλ ⊂ Qcλ .
Lemma 5.5. Let W be a left ideal in a group algebra k (G) (char k = 0). Then
W 2 = 0 implies W = 0.
Proof. Since k (G) is completely reducible k (G) = W ⊕ W ′, where W ′ is another left
ideal. Let y ∈ EndG (k (G)) such that y|W = Id, y (W ′ ) = 0. But we proved that any
y ∈ EndG (k (G)) is a right multiplication on some u ∈ k (G) (see lecture notes 3).
Then we have u2 = u, W = Au, in particular u ∈ W . If W 6= 0, then u 6= 0 and
u2 = u 6= 0. Hence W 2 6= 0.
Corollary 5.6. Acλ is a minimal left ideal.
Proof. Let W ⊂ Acλ be a left ideal. Then either cλ W = Qcλ or cλ W = 0 by
Corollary 5.4. In the former case W = Acλ W = Acλ . In the latter case W 2 ⊂
Acλ W = 0, and W = 0 by Lemma 5.5.

n!
Corollary 5.7. c2λ = nλ cλ , where nλ = dim Vλ
.
Proof. From the proof of Lemma 5.5, cλ = nλ u for some idempotent u ∈ Q (Sn ).
Therefore cλ = nλ u. To find nλ note that trk(G) u = dim Vλ , trk(G) cλ = |Sn | = n!.
Lemma 5.8. Order partitions lexicographically. If λ > µ, then there exists i, j in
the same row of λ and in the same column of µ.
Proof. Check yourself.
Corollary 5.9. If λ < µ, then cλ Acµ = 0.
Proof. Sufficient to check that cλ scµ = 0 for any s ∈ Sn , which is equivalent to
cλ scµ s−1 = cλcs(µ) = 0.
Let (ij) ∈ Qλ ∩ Ps(µ) . Then

cλ (ij) (ij) cs(µ) = cλ cs(µ) = −cλ cs(µ) = 0.

Lemma 5.10. Vλ and Vµ are isomorphic iff λ and µ have the same Young diagram.
Proof. If λ and µ have the same diagram, then λ = s (µ) for some s ∈ Sn and
Acλ = Ascµ s−1 = Acµ s−1 . Assume λ > µ, then cλ Acµ = 0 and cλ Acλ 6= 0. Therefore
Acλ and Acµ are not isomorphic.
Corollary 5.11. If λ and µ have different diagrams, then cλ Acµ = 0.
Proof. If cλ Acµ 6= 0, then Acλ Acµ = Acµ . On the other hand Acλ A has only
components isomorphic to Vλ . Contradiction.
Lemma 5.12. Let ρ : Sn → GL (V ) be an arbitrary representation. Then the
multiplicity of Vλ in V equals the rank of ρ (cλ).
Proof. The rank of cλ is 1 in Vλ and 0 in any Vµ with another Young diagram.
PROBLEM SET # 5
MATH 252
Due October 7.
1. Classify irreducible representations of A4 (even permutations) over R and over
C. What is the splitting field for A4 .
2. Let G be a finite group, r be the number of conjugacy classes in G and s be the
number of conjugacy classes in G preserved by the involution g → g −1 . Prove that
the number of irreducible representations of G over R is equal to r+s
2
.
3. If λ is a Young tableau, then the conjugate tableau λ′ is obtained from λ by
symmetry about diagonal (rows and columns switch). Show that Vλ′ is isomorphic
Vλ ⊗ sgn, where sgn is one-dimensional sign representation. (Hint: you probably have
to show that Q (Sn ) aλ bλ and Q (Sn ) bλaλ are isomorphic).

1
WEEK 6
1. Dimension formulae
For a Young tableau (diagram) λ, λi denotes the number of boxes in the i-th row.
We write α ∈ λ if α is a box of λ. Let |λ| denote the number of boxes in λ and λ − α
denote a diagram which can be obtained from λ by removing one box. For example,
for partition λ = (5, 3, 1) the possible λ − α are (4,3,1), (5,2,1) and (5,3).
Theorem 1.1. ResSn−1 Vλ = ⊕Vλ−α .
Lemma 1.2. Let |µ| = |λ| − 1. If the diagram of µ is different from the diagram of
λ − α for all possible α, then there are i, j either in the same row of λ and in the
same column of µ or in the same row of µ and in the same column of λ.
Proof. Let λ and µ do not satisfy the condition of Lemma. Choose the smallest k
such that λk 6= µk . Assume first, that µk > λk , then by pigeon hole principle one can
find two entries of k-th row of µ in the same column of λ. (We assume that this does
not happen with the first k − 1 rows).
Assume now that λk > µk . Since λ has just one more entry than µ, λk > µk + 1
implies that two entries in the k-th row of λ are in the same column of µ. Therefore
λk = µk + 1, moreover the last entry appears in the first k rows of λ. In this case
we move to the next row, and step by step prove that λi = µi for all i 6= k. Hence
µ = λ − α.
Lemma 1.3. If the diagram of µ is different from the diagram of λ−α for all possible
α, then cµ Acλ = 0.
Proof. Let s ∈ Sn . First assume that there are two entries in the same row of µ and
in the same column of λ. Then the same is true for µ and s (λ) for any s ∈ Sn as we
can see from the proof of Lemma 1.2. Hence for this pair of entries i, j we have
aµ (ij)2 bs(λ) = aµ bs(λ) = −aµ bs(λ) = 0.
Therefore aµ sbλ s−1 = 0 and aµ sbλ = 0 for any s ∈ Sn . That implies aµ Abλ = 0.
Similarly we can prove that if there are two entries in the same column of µ and in
the same row of λ, then bµ Aaλ = 0. Together that implies cµ Acλ = 0.
Corollary 1.4. If a Young diagram µ can not be obtained from λ by removing one
box, then the multiplicity of Vµ in ResSn−1 Vλ is zero.
Date: October 7, 2005.
1
Proof. Follows from Lemma 5.12 (lecture notes 5).

Lemma 1.5. Let µ = λ − α. Then cµ cλ 6= 0. Therefore f : Vµ → Vλ given by
f (x) = xcλ is an injective homomorphism of Sn−1 -modules.
Proof. If we write X X
cµ cλ = us s, c2µ = vs s,
s∈Sn s∈Sn−1
then as one can easily see that us = vs for any s ∈ Sn−1 .
Lemma 1.6. If µ = λ − α, then cµ Acλ ⊂ Qcµ cλ .
Proof. First, prove that if there is no (ij) ∈ Qµ Ps(λ) then s ∈ QλPλ . Thus, bµ saµ 6= 0,
or equivalently bµ as(λ) 6= 0, implies s ∈ QλPλ . Then prove that for any s ∈ QλPλ ,
cµ scλ ∈ Qcµ cλ .
Corollary 1.4, Lemma 1.5 and Lemma 1.6 imply Theorem 1.1.
Corollary 1.7. X
dim Vλ = dim Vλ−α .
Remark 1.8. Every function f (λ) on the set of Young diagrams satisfying
X
(1.1) f (λ) = f (λ − α) , f (1) = 1
coincides with dim Vλ .
S
Corollary 1.9. IndSn+1
n
Vλ = ⊕Vµ , where µ runs the set of all diagrams obtained
from λ by adding one box.
A Young tableau is standard if entries in every row and entries in every column
are in increasing order.
Corollary 1.10. dim Vλ equals the number of all standard tableaux on a diagram λ.
Proof. Check that the number dλ of standard tableaux satisfies (1.1).

For a box α ∈ λ, let hα be the hook diagram containing α, all boxes below α and
all boxes to the right of α. Let
Y
h (λ) = |hα|.
α∈λ
Example. If λ is (3,2,1), then h (λ) = 45.

Lemma 1.11. Let λ = (λ1 , . . . , λk ) and λ̄ = (λ1 + k − 1, λ2 + k − 2, . . . , λk ). Then
λ̄1 ! . . . λ̄k !
(1.2) h (λ) = Q .
i<j λ̄i − λ̄j
Proof. Direct calculation.
Lemma 1.12. Let

Y
V (x1, . . . , xk ) = (xi − xj ) .
i<j
Then
k
X k (k − 1)
(1.3) xi (V (x1, . . . , xk ) − V (x1, . . . , xi − 1, . . . , xk )) = V (x1, . . . , xk ) .
i=1
2
Proof. Since V is a skew symmetric polynomial of x, it is easy to check that (xi − xj )

divides the left hand side of the identity. Since the degree of the left hand side
polynomial is k(k−1)
2
, the same as the degree of V , the LHS polynomial is proportional
to V . The leading coefficient of LHS is the same as of
k
X ∂ k (k − 1)
xi V (x1, . . . , xk ) = V (x1 , . . . , xk ) .
i=1
∂xi 2
That proves the identity.
Lemma 1.13.
n X 1
= .
h (λ) h (λ − α)
Proof. Using (1.2) write
k
V λ̄1 , . . . , λ̄k X V λ̄1 , . . . , λ̄i − 1, . . . , λ̄k
n = .
λ̄1 ! . . . λ̄k ! i=1
λ̄1 ! . . . λ̄ i − 1 ! . . . λk !
This is equivalent to
k
X

(1.4) nV λ̄1 , . . . , λ̄k = λ̄i V λ̄1 , . . . , λ̄i − 1, . . . , λ̄k .
i=1
k(k−1)
Use now that λ̄1 + · · · + λ̄k = n + 2
and apply (1.3) to prove (1.4).

|λ|!
Corollary 1.14. (Hook formula) dim Vλ = h(λ)
.
|λ|!
Proof. Just check that h(λ)
satisfies (1.1).

2. Representations of GLk .
Matrix coefficients. Let ρ : G → GL (V ) be a (finite-dimensional) representa-
tion. For any ϕ ∈ V ∗ , v ∈ V define
fv,ϕ (s) = hϕ, ρs vi .
This function f is called a matrix coefficient.
Let G = GLk = GL Ck and C [G] denote the space of all polynomial functions
on G. We call a representation ρ : G → GL (V ) polynomial if fv,ϕ ∈ C [G] for all
v ∈ V, ϕ ∈ V ∗.
Examples. The standard representation in the space E = Ck is polynomial, but
the dual representation in E ∗ is not.
The whole space C [G] has a natural structure of a representation if we put
Rg f (x) = f (xg) .
Check that the space Cn [G] of homogeneous polynomials of degree n is invariant.
Thus, there is a decomposition
C [G] = ⊕∞
n=0 Cn [G] .
Let ρ : G → GL (V ) be a polynomial representation. For any ϕ ∈ V ∗ define a map

ρ′ϕ : V → C [G] by the formula
ρ′ϕ (v) = fv,ϕ .
Check that this map is an intertwiner, i.e.
ρ′ϕ (ρg v) = Rg ρ′ϕ (v) .
That implies
Lemma 2.1. Every irreducible polynomial representation of G is a subrepresentation
in C [G].
Lemma 2.2. Consider the representation of G in (E ∗ )⊗n ⊗ E ⊗n given by the formula
ρg (ϕ1 ⊗ · · · ⊗ ϕn ⊗ v1 ⊗ · · · ⊗ vn ) = ϕ1 ⊗ · · · ⊗ ϕn ⊗ gv1 ⊗ · · · ⊗ gvn ,
for all ϕi ∈ E ∗, vj ∈ E.
The map π : (E ∗ )⊗n ⊗ E ⊗n → Cn [G] given by
ϕ ⊗ v 7→ fv,ϕ
for each ϕ ∈ (E ∗ )⊗n , v ∈ E ⊗n , is surjective.
Proof. Let e1, . . . , ek be a basis in E and f1 , . . . , fk be the dual basis in E ∗. Then
fj1 ⊗ · · · ⊗ fjn ⊗ ei1 ⊗ · · · ⊗ ein 7→ gi1 j1 . . . gin jn ,
where gij is a matrix entry of a matrix g in the basis e1, . . . , ek . Thus, the monomial
basis of Cn [G] belongs to the image of π.
Remark 2.3. Here I made a mistake during the lecture. To get an isomorphism we
have to consider S n (E ∗ ⊗ E) ⊂ (E ∗ )⊗n ⊗ E ⊗n .
To classify polynomial irreducible representation of G we have to find all irreducible
subrepresentations of E ⊗n . We will do this in the next section.
3. Duality between GLk and Sn

Consider the representation ρ : Sn → GL (E ⊗n ) defined by the formula
ρs (v1 ⊗ · · · ⊗ vn ) = vs(1) ⊗ · · · ⊗ vs(n) ,
and the representation ρ : GLk → GL (E ⊗n ) defined by
ρg (v1 ⊗ · · · ⊗ vn ) = gv1 ⊗ · · · ⊗ gvn .
We see immediately that ρs ◦ ρg = ρg ◦ ρs for any s ∈ Sn , g ∈ GLk . Thus we can
consider ρ as the representation of the direct product Sn × GLk .
Theorem 3.1. Let Γn,k denote the set of all Young diagrams with n boxes such that
the number of rows is not bigger than k. Then
E ⊗n = ⊕λ∈Γn,k Vλ ⊠ Wλ ,
where Vλ is the irreducible representation of Sn associated with λ and Wλ is an
irreducible representation of GLk . Moreover, Wλ and Wµ are not isomorphic if λ 6= µ.
Corollary 3.2. Fill the boxes of λ in some way. Then ρcλ (E ⊗n ) is an invariant
subspace isomorphic to Wλ.
Example. PLet λ = (n) be a one row diagram. Then Vλ is the trivial representation
of Sn , cλ = s∈Sn s, and Wλ = S n (E).
PIf k > n,s and λ = (1, . . .n, 1) (one row), then Vλ is the sign representation, cλ =
s∈Sn (−1) s, and Wλ = Λ (E).
To prove Theorem 3.1 we need the following general statement.
Theorem 3.3. Let ρ : G → GL (V ), σ : K → GL (V ) be two representations in the
same vector space V over algebraically closed F . Let
EndG (V ) = σ (F (K))
and ρ is completely reducible. Then
V = ⊕m
i=1 Vi ⊠ Wi ,
where Vi is an irreducible representation of G, Wi is an irreducible representation of

K. Moreover, Vi is not isomorphic to Vj if i 6= j and similarly, Wi is not isomorphic
to Wj if i 6= j.
Proof. Since ρ is completely reducible, one can write

V = ⊕m
i=1 (Vi ⊗ Wi ) ,
the action of G is trivial on Wi . Then

m
Y
EndG (V ) = EndF (Wi ) .
i=1
Thus, σ : F (K) → EndF (Wi ) is surjective, that implies that each Wi is irreducible
over K and Wi ∼6 Wj if i 6= j.
=
Remark 3.4. In general, we say that G and K satisfying the conditions of Theorem 3.3
form a dual pair. Such situation often happens in representation theory. The simplest
example is an action of G × G in k (G) given by
X X
R(g,h) us s = us gsh−1 .
s∈G s∈G
Lemma 3.5. In the situation of Theorem 3.1 we have

EndSn E ⊗n = ρ (C (GLk )) .

Proof. Let Mk denote the algebra EndC (E), in other words, Mk is the matrix algebra.
First,
EndC E ⊗n = Mk⊗n .

Thus, we are looking at the Sn invariant subalgebra

Sn
EndSn E ⊗n = Mk⊗n = S n (Mk ) .

In other words, EndC (E ⊗n ) is spanned by

X
ms(1) ⊗ · · · ⊗ ms(n)
s∈Sn
for all possible m1, . . . , mn ∈ Mk .

Our next claim is that EndC (E ⊗n ) is the span of m ⊗ · · · ⊗ m for all possible
m ∈ Mk . It follows from the following
Lemma 3.6. For an arbitrary vector space V , S n (V ) is spanned by v n for all v ∈ V .
Proof. Everybody knows the formula
4xy = (x + y)2 − (x − y)2 ,
which proves the statement for n = 2. Less know is the following general formula
X n
2n x1 . . . xn = (−1)i2 +···+in x1 + (−1)i2 x2 + · · · + (−1)in xn .
i2 =0,1,...,in =0,1

Now let U be the span of g ⊗ · · · ⊗ g (g ∈ GLk ) in EndC (E ⊗n ) = S n (Mk ). By

definition, U = ρ (C (GLk )). Note that GLk is a dense subset in Mk , therefore
U is a dense subset in S n (Mk ). But U is a linear subspace in S n (Mk ). Hence
U = S n (Mk ) = EndC (E ⊗n ).
Note that Lemma 3.5 together with Theorem 3.3 imply that
E ⊗n = ⊕λ∈Γ′ Vλ ⊠ Wλ,
for some set Γ′ of Young diagrams with n boxes. It is left to show that Γ′ = Γn,k .
Obviously, Γ′ consists of all diagrams λ for which Wλ = cλ (E ⊗n ) 6= 0. Fill the boxes
of λ in increasing order from 1 to n from left to right starting from the top and
consider cλ defined by this tableaux. An element v = ei1 ⊗ · · · ⊗ ein of the standard
basis in E ⊗n can be represented by the same tableau with entries ei1 , . . . , ein . If λ
has more than k rows, then one can find ej which appears twice in the same column.
Then bλ (v) = 0, and therefore cλ v = 0. Since this holds for any basis vector, we have
cλ (E ⊗n ) = 0. Hence λ ∈
/ Γ′ if λ has more than k rows. On the other hand, if λ has
k or less rows, one can check that

⊗λk
cλ e⊗λ
1
1
⊗ · · · ⊗ e k 6= 0.
Therefore Γ′ = Γn,k . Theorem 3.1 is proven.
PROBLEM SET # 6
MATH 252
Due October 14.

Use notations of lecture notes 5 and 6.
1. Let λ = (n − k, 1, ..., 1) be the hook diagram with n boxes. Show that Vλ is
isomorphic to Λk V , where V is n − 1-dimensional subrepresentation in the standard
permutation representation of Sn .(Hint: Use Lemma 5.12 in Lecture notes 5)
2. Show that the GLk -representation Wλ ⊗ E decomposes into direct sum of Wµ
for all µ ∈ Γn+1,k which can be obtained from λ by adding one box. (Hint: check
when ρcµ ρcλ (E ⊗n+1 ) 6= 0.)

1
WEEK 7
1. Characters of GLk and Sn

A character of an irreducible representation of GLk is a polynomial function con-
stant on every conjugacy class. Since the set of diagonalizable matrices is dense in
GLk , a character is defined by its values on the subgroup of diagonal matrices in GLk .
Thus, one can consider a character as a polynomial function of x1, . . . , xk . Moreover,
a character is a symmetric polynomial of x1, . . . , xk as the matrices diag (x1, . . . , xk )
and diag xs(1) , . . . , xs(k) are conjugate for any s ∈ Sk .
For example, the character of the standard representation in E is equal to x1 +
· · · + xk and the character of E ⊗n is equal to (x1 + · · · + xk )n .
Let λ = (λ1 , . . . , λk ) be such that λ1 ≥ λ2 ≥ · · · ≥ λk ≥ 0. Let Dλ denote the
λ
determinant of the k × k-matrix whose i, j entry equals xi j . It is clear that Dλ
Q a skew-symmetric polynomial of x1 , . . . , xk . If ρ = (k − 1, . . . , 1, 0) then Dρ =
is
i≤j (xi − xj ) is the well known Vandermonde determinant. Let
Dλ+ρ
Sλ = .
Dρ
It is easy to see that Sλ is a symmetric polynomial of x1 , . . . , xk . It is called a Schur
polynomial. The leading monomial of Sλ is the xλ1 . . . xλk k (if one orders monomials
lexicographically) and therefore it is not hard to show that Sλ form a basis in the
ring of symmetric polynomials of x1 , . . . , xk .
Theorem 1.1. The character of Wλ equals to Sλ .
I do not include a proof of this Theorem since it uses beautiful but hard combina-
toric. The proof is much easier in general framework of Lie groups and is included in
261A course.
Exercise. Check that
Q
i<j λ̄i − λ̄j
Y λ̄i − λ̄j
dim Wλ = = ,
i<j
(ρi − ρj ) (k − 1)! (k − 2)! . . . 1!
if λ̄ = λ + ρ.
Now we use Schur-Weyl duality to establish the relation between characters of Sn
and GLk . Recall that the conjugacy classes in Sn are given by partitions of n. Let
C (µ) be the class associated with the partition µ in the natural way. Let ρ denote
1
the representation of Sn × GLk in E ⊗n . Let r be the number of rows in µ. Then one

can see that
(1.1) tr (ρs×g ) = (xµ1 1 + · · · + xµk 1 ) . . . (xµ1 r + · · · + xµk r ) ,
for any s ∈ C (µ) and a diagonal g ∈ GLk . Denote by Pµ the polynomial in the right
hand side of the identity. Let χλ be the character of Vλ . Since
X
tr (ρs×g ) = χλ (s) Sλ (g) ,
λ∈Γn,k
one obtains the following remarkable relation

X
(1.2) Pµ = χλ (s) Sλ .
λ∈Γn,k
2. Representations of compact groups

Let G be a group and a topological space. We say that G is a topological group
if the multiplication map G × G → G and the inverse G → G are continuous maps.
Naturally, G is compact if it is compact topological space.
Examples. The circle
S 1 = {z ∈ C | |z| = 1} .
A torus T n = S 1 × · · · × S 1.
Unitary group
Un = X ∈ GLn | X̄ t X = 1n .

Special unitary group

SUn = {X ∈ Un | det X = 1} .
Orthogonal group
On = x ∈ GLn (R) | X t X = 1n .

Special orthogonal group

SOn = {X ∈ On | det X = 1} .
Theorem 2.1. Let G be a compact group. There exists a unique measure on G such
that Z Z
f (ts) dt = f (t) dt,
G G
R
for any integrable function f on G and any s ∈ G, and G dt = 1.
In the same way there exists a measure d′ t such that
Z Z Z
′
f (st) dt = f (t) d t, d′ t = 1.
G G G
′
Moreover, for a compact group dt = d t.
The measure dt (d′ t) is called right-invariant (left-invariant) measure, or Haar mea-

sure.
We do not give the proof of this theorem in general. However, all examples we
consider are smooth submanifolds in GLk . Thus, to define the invariant measure we
just need to define a volume in the tangent space at identity T1G and then use right
(left) multiplication to define it on the whole group. More precisely, let γ ∈ ΛtopT1∗G.
Then
γs = m∗s (γ) ,
where ms : G → G is the right (left) multiplication on sR and m∗s is the induced map
ΛtopT1∗G → ΛtopTs∗G. After this normalize γ to satisfy G γ = 1.
Consider a vector space over C equipped with topology such that addition and
multiplication by a scalar are continuous. We always assume that a topological
vector space satisfies the following conditions
(1) for any v ∈ V there exist a neighborhood of 0 which does not contain v;
(2) there is a base of convex neighborhoods of zero.
Topological vector spaces satisfying above conditions are called locally convex. We
do not go into the theory of such spaces. All we need is the fact that there is a
non-zero continuous linear functional on a locally convex space.
A representation ρ : G → GL (V ) is continuous if the map G × V → V given by
(s, v) 7→ ρs v is continuous.
Regular representation. Let G be a compact group and L2 (G) be the space of
all complex valued functions on G such that
Z
|f (t) |2 dt
exists. Then L2 (G) is a Hilbert space with respect to Hermitian form

Z
hf, gi = f¯ (t) g (t) dt.
G
2
Moreover, a representation R of G in L (G) given by
Rs f (t) = f (ts)
is continuous and the Hermitian form is G-invariant.
A representation ρ : G → GL (V ) is called topologically irreducible if any invariant
closed subspace of V is either V or 0.
Lemma 2.2. Every irreducible representation of G is isomorphic to a subrepresen-
tation in L2 (G).
Proof. Let ρ : G → GL (V ) be irreducible. Pick a non-zero linear functional ϕ on V
and define the map Φ : V → L2 (G) which sends v to the matrix coefficient fv,ϕ (s) =
hϕ, ρs vi. It is clear that a matrix coefficient is a continuous function on G, therefore
fv,ϕ ∈ L2 (G). Furthermore Φ is a continuous intertwiner and Ker Φ = 0.
Recall that a Hilbert space is a space over C equipped with positive definite Her-
mitian form h, i complete in topology defined by the norm
kvk = hv, vi1/2 .

We need the fact that a Hilbert space has an orthonormal topological basis. A
continuous representation ρ : G → GL (V ) is called unitary if V is a Hilbert space
and
hv, vi = hρg v, ρg vi
for any v ∈ V and g ∈ G. The regular representation of G in L2 (G) is unitary. In
fact, Lemma 2.2 implies
Corollary 2.3. Every topologically irreducible representation of a compact group G
is a subrepresentation in L2 (G).
Lemma 2.4. Every irreducible unitary representation of a compact group G is finite-
dimensional.
Proof. Let ρ : G → GL (V ) be an irreducible unitary representation. Choose v ∈ V ,
kvk = 1. Define an operator T : V → V by the formula
T x = hv, xi v.
One can check easily that T is self-adjoint, i.e.
hx, T yi = hT x, yi .
Let Z
ρg T ρ−1

T̄ x = g x dg.
G
Then T̄ : V → V is an intertwiner and a self-adjoint operator. Furthermore, T̄ is

compact, i.e. if
S = {x ∈ V | kxk = 1} ,
then T̄ (S) is a compact set in V . Every self-adjoint compact operator has an eigen-
vector. To construct an eigen vector
find x ∈ S such that | T̄ x, x | is maximal. Then
T̄ x = λx. Since Ker T̄ − λ Id is an invariant subspace in V , Ker T̄ − λ Id = 0.
Hence T̄ = λId. Note that for any orthonormal system of vectors e1 , . . . , en ∈ V
X X
ei , T̄ ei = hei , T eii ≤ 1,
that implies λn ≤ 1. Hence dim V ≤ λ1 .

Corollary 2.5. Every irreducible continuous representation of a compact group G
is finite-dimensional.
3. Orthogonality relations and Peter-Weyl Theorem

If ρ : G → GL (V ) is a unitary representation. Define a matrix coefficient by the
formula
fv,w (g) = hw, ρg vi .
It is easy to check that
fv,w g −1 = f¯w,v (g)

(3.1)
Theorem 3.1. For an irreducible unitary representation ρ : G → GL (V )
Z
1
hfv,w , fv′ ,w′ i = f¯v,w (g) fv′ ,w′ (g) dg = hv, v ′i hw′ , wi .
G dim ρ
The matrix coefficient of two non-isomorphic representation are orthogonal in L2 (G).
Proof. Define T ∈ EndC (V )
T x = hv, xi v ′
and Z
T̄ = ρg T ρ−1
g dg.
G
As follows from Shur’s lemma, T̄ = λId. Since
tr T̄ = tr T = hv, v ′i ,
we obtain
hv, v ′i
T̄ = .
dim ρ
Hence
1
w′ , T̄ w = hv, v ′i hw′ , wi .
dim ρ
On the other hand,
Z Z
′ ′
v, ρ−1 ′
fw,v g −1 fv′ ,w′ (g) dg =

w , T̄ w = w, g w ρg v dg =
G G
Z
1
= f¯v,w (g) fv′ ,w′ (g) dg =
hfv,w , fv′ ,w′ i .
G dim ρ
In fv,w and fv′ ,w′ are matrix coefficients of two non-isomorphic representation, the
T̄ = 0, and the calculation is even simpler.
Corollary 3.2. Let ρ and σ be two irreducible representations, then hχρ , χσ i = 1 if
ρ is isomorphic to σ and hχρ , χσ i = 0 otherwise.
Theorem 3.3. (Peter-Weyl) Matrix coefficient form a dense set in L2 (G) for a
compact group G.
Proof. We will prove the Theorem under assumption that G ⊂ GL (E), in other
words we assume that G has a faithful finite-dimensional representation. Let M =
EndC (E). The polynomial functions C [M] on M form a dense set in the space
of continuous functions on G (Weierstrass theorem), and continuous functions are
dense in L2 (G). On the other hand, C [M] is spanned by matrix coefficients of
all representations in T (E) = ⊕∞
n=0 E
⊗n
. Hence matrix coefficients are dense in
2
L (G).
Corollary 3.4. The characters of irreducible representations form an orthonormal
basis in the subspace of class function in L2 (G).
Corollary 3.5. Let G be a compact group and R denote the representation of G × G
in L2 (G) given by the formula
Rs,t f (x) = f s−1 xt .

Then
= ⊕ρ∈Gb Vρ ⊠ Vρ∗,
L2 (G) ∼
where G
b denotes the set of isomorphism classes of irreducible unitary representations
of G and the direct sum is in the sense of Hilbert spaces.
Remark 3.6. Note that it follows from the proof of Theorem 3.3, that if E is a faithful
representation of a compact group G, then all other irreducible representations appear
in T (E) as subrepresentations.
4. Examples
Example 1. Let S 1 = {z ∈ C | |z| = 1}, z = eiθ . The invariant measure on S 1
dθ
is 2π The irreducible representations are one dimensional. They are given by the
characters χn : S 1 → C∗, where χn (θ) = einθ . Hence Sb1 = Z and
L2 S 1 = ⊕n∈Z Ceinθ ,

this is well-known fact that every periodic function can be extended in Fourier series.
Example 2. Let G = SU2 . Then G consists of all matrices
a b
,
−b̄ ā
satisfying the relations |a|2 + |b|2 = 1. One also can realize SU2 as the subgroup
of quaternions with norm 1. Thus, topologically SU2 is isomorphic to the three-
dimensional sphere S 3 . To find all irreducible representation of SU2 consider the
polynomial ring C [x, y] with the action of SU2 given by the formula

a b
ρg (x) = ax + by, ρg (y) = −b̄x + āy, if g = .
−b a
Let ρn be the representation of G in the space Cn [x, y] of homogeneous polynomials

of degree n. The monomials xn , xn−1 y, . . . , y n form a basis of Cn [x, y]. Therefore
dim ρn = n + 1. We claim that all ρn are irreducible and that every irreducible
representation of SU2 is isomorphic to ρn . Hence G b = Z+ . We will show this by
checking that the characters χn of ρn form an orthonormal basis in the Hilbert space
of class functions on G.
Note that every unitary matrix is diagonal in some orthonormal basis, therefore
every conjugacy class of SU2 intersects the diagonal subgroup. Moreover, z0 z0 and

(z̄0 0z ) are conjugate. Hence the set of conjugacy classes can be identified with S 1
quotient by the equivalence relation z ∼ z̄. Let z = eiθ , then
z n+1 − z −n−1 sin (n + 1) θ
(4.1) χn (z) = z n + z n−2 + · · · + z −n = = .
z − z −1 sin θ
Now let us calculate the scalar product in the space of class function. It is clear that
the invariant measure dg on G is proportional to the standard volume form on the
three-dimensional sphere induced by the volume form on R4 . Let C (θ) denote the
conjugacy class of all matrices with eigenvalues eiθ , e−iθ . The characteristic polyno-
mial of a matrix from C (θ) equals t2 − 2 cos θt + 1. Thus, we obtain a + ā = 2 cos θ,
or a = cos θ + yi for real y. Hence C (θ) satisfy the equation
|a|2 + |b2 | = cos2 θ + y 2 + |b|2 = 1,
or
y 2 + |b|2 = sin2 θ.
In other words, C (θ) is a two-dimensional sphere of radius sin θ. Hence for a class
function φ on G
1 2π
Z Z
φ (g) dg = φ (θ) sin2 θdθ.
π 0
All class function are even functions on S 1, i.e. they satisfy the condition φ (−θ) =
φ (θ). One can see easily from (4.1) that χn (θ) form an orthonormal basis in the
space of even function on the circle with respect to the Hermitian product
1 2π
Z
hϕ, ηi = ϕ̄ (θ) η (θ) sin2 θdθ.
π 0
Example 3. Let G = SO3 . Recall that SU2 can be realized as the set of quater-
nions with norm 1. Consider the representation γ of SU2 in H defined by the formula
γg (α) = gαg −1 . One can see that the 3-dimensional space Him of pure imaginary
quaternions is invariant and (α, β) = Re αβ̄ is invariant positive definite scalar
product on Him . Therefore ρ defines a homomorphism γ : SU2 → SO3. Check that
Ker γ = {1, −1} and that γ is surjetive. Hence SO3 ∼ = SU2/ {1, −1}. Thus, every
representation of SO3 can be lifted to the representations of SU2 , and a representa-
tion of SU2 factors to the representation of SO3 iff it is trivial on −1. One can check
easily that ρn (−1) = 1 iff n is even. Thus, an irreducible representations of SO3 is
isomorphic to ρ2m and dim ρ2m = 2m + 1. Below we give an independent realization

of irreducible representation of SO3 .
Harmonic analysis on a sphere. Consider the sphere S 2 in R3 defined by the
equation x2 + y 2 + z 2 = 1. It is clear that SO3 acts in the space of complex-valued
functions on S 2 . Introduce differential operators in R3:
−1 2 3 1 2
x + y 2 + z 2 , h = x∂x + y∂y + z∂z + , f = ∂x + ∂y2 + ∂z2 ,

e=
2 2 2
note that e, f, and h commute with the action of SO3 and satisfy the relations
[e, f] = h, [h, e] = 2e, [h, f] = −2f.
Let Pn be the space of homogeneous polynomial of degree n and Hn = Ker f ∩Pn . The
polynomials of Hn are harmonic polynomials since they are annihilated by Laplace
operator. For any ϕ ∈ Pn
3
h (ϕ) = n + ϕ.
2
If ϕ ∈ Hn , then
3
fe (ϕ) = ef (ϕ) − h (ϕ) = − n + ϕ,
2
and by induction

k k−1 k−1 3k
fe (ϕ) = efe (ϕ) − he (ϕ) = − nk + k (k − 1) + ek−1 ϕ.
2
In particular, this implies that
(4.2) fek (Hn ) = ek−1 (Hn ) .
We prove that
(4.3) Pn = Hn ⊕ e (Hn−2 ) ⊕ e2 (Hn−4 ) + . . .
by induction on n. Indeed, by induction assumption
Pn−2 = Hn−2 ⊕ e (Hn−4 ) + . . . ,
then (4.2) implies fe (Pn−2 ) = Pn−2 . Hence Hn ∩ ePn−2 = 0. On the other hand,
f : Pn → Pn−2 is surjective, and therefore dim Hn + dim Pn−2 = dim Pn . Therefore
(4.4) Pn = Hn ⊕ Pn−2 ,
which implies (4.3). Note that after restriction on S 2, the operator e acts as the
multiplication on −1
2
.
Hence (4.3) implies that
C S 2 = ⊕n≥0 Hn .

To calculate the dimension of Hn use (4.4)

(n + 1) (n + 2) n (n − 1)
dim Hn = dim Pn − dim Pn−2 = − = 2n + 1.
2 2
Finally, we claim that the representation of SO3 in Hn is irreducible and isomorphic

to ρ2n . Check that ϕ = (x + iy)n ∈ Hn and the rotation on the angle θ about z−
axis maps ϕ to einθ ϕ. Since this rotation is the image of
eiθ/2 0
,
0 e−iθ/2
under the homomorphism γ : SU2 → SO3 , the statement follows from (4.1).
Recall now the following theorem (Lecture Notes 1).
A convex centrally symmetric solid in R3 is uniquely determined by the areas of
the plane cross-sections through the origin.
A convex solid B can be defined by an even continuous function on S 2. Indeed,
for each unit vector v let
ϕ (v) = sup t2/2 | tv ∈ B .

Define a linear operator T in the space of all even continuous functions on S 2 by the
formula Z 2π
T ϕ (v) = ϕ (w) dθ,
0
where w runs the set of unit vectors orthogonal to v, and θ is the angular parameter
on the circle S 2 ∩ v ⊥ . Check that T ϕ (v) is the area of the cross section by the plane
v ⊥ . We have to prove that T is invertible.
Obviously T commutes with the SO3 -action. Therefore T can be diagonalized.
Moreover, T acts on H2n as the scalar operator λn Id. We have to check that λn 6= 0
2n
for all n. Let ϕ = (x + iy) ∈ H2n . Then ϕ (1, 0, 0) = 1 and
Z 2π Z 2π
2n n
T ϕ (1, 0, 0) = (iy) dθ = (−1) sin2n θdθ,
0 0
2 2
here we take the integral over the circle y + z = 1, and assume y = sin θ, z = cos θ.
Since T ϕ = λn ϕ, we obtain
Z 2π
n
λn = (−1) sin2n θdθ 6= 0.
0
PROBLEM SET # 7
MATH 252
Due October 21.

1. Show that SO4 is isomorphic to the quotient of SU2 × SU2 by the subgroup
generated by (−1, −1). Hint : consider the representation of SU2 × SU2 in the space
of quaternions H by left and right multiplication.
2. Show the following identity for representations of SU2
ρm ⊗ ρn = ρm+n ⊕ ρm+n−2 ⊕ · · · ⊕ ρm−n ,
assuming m ≥ n.

1
WEEK 8
VERA SERGANOVA
1. Representations of SL2 (R)

In this section
G = SL2 (R) = {g ∈ GL2 (R) | det g = 1} .
Let K be the subgroup of matrices

cos θ sin θ
gθ = .
− sin θ cos θ
The group K is a maximal compact subgroup of G, clearly K is isomorphic to S 1. If
ρ : G → GL (V ) is a unitary representation of G in a Hilbert space then then ResK ρ
splits into the sum of 1-dimensional representations of V . In particular, one can find
v ∈ V such that ρgθ (v) = einθ v. Define the matrix coefficient function f : G → C
given by
f (g) = hv, ρg vi .
Then f satisfies the condition
f (ggθ ) = einθ f (g) .
Thus, one can consider f as a section of a linear bundle on the space G/K (if n = 0,
then f is a function). Thus, it is clear that the space G/K is an important geometric
object, where the representations of G are “realized”.
Consider the Lobachevsky plane
H = {z ∈ C | Im z > 0}
2 2
with metric defined by the formula dx y+dy2 and the volume form dxdy y2
. Then G
coincides with the group of rigid motions of H preserving orientation. The action of
G on H is given by the formula
az + b
z 7→ .
cz + d
One can check easily that G acts transitively on H, preserves the metric and volume.
Moreover, the stabilizer of i ∈ H coincides with K. Thus, we identify H with G/K.
The first series of representations we describe is called the representations of dis-
crete series. Those are the representations with matrix coefficients in L2 (G). Let
Date: November 7, 2005.
1
2 VERA SERGANOVA
n/2
H+n be the space of holomorphic densities on H, the expressions ϕ (z) (dz) , where
ϕ (z) is a holomorphic function on H satisfying the condition that
Z
|ϕ|2y n−2 dzdz̄
is finite. Define the representation of G in H+n by

n/2
az + b 1
ρg ϕ (z) (dz) =ϕ (dz)n/2 ,
cz + d (cz + d)n
and Hermitian product on Hn the formula
D E Z
n/2 n/2
(1.1) ϕ (dz) , ψ (dz) = ϕ̄ψy n−2 dzdz̄,
for n > 1. For n = 1 the product is defined by

D E Z ∞
n/2 n/2
(1.2) ϕ (dz) , ψ (dz) = ϕ̄ψdx,
−∞
in this case H+ 2
1 consists of all densities which converge to L -functions on the bound-
ary (real line). Check that this Hermitian product is invariant.
Let us show that Hn is irreducible. It is convenient to consider Poincaré model of
Lobachevsky plane using the conformal map
z−i
w= ,
z+i
that maps H to a unit disk |w| < 1. Then the group G acts on the unit disk by
aw+b
linear-fractional maps w → b̄w+ā for all complex a, b
satisfying |a| − |b| = 1, and K is defined by the condition b = 0. If a = eiθ , then
2 2
dwdw̄
ρgθ (w) = e2iθ w. The invariant volume form is 1− w̄w
.
n/2
It is clear that w (dw) for all k ≥ 0 form an orthogonal basis in H+
k
n , each vector
k n/2
w (dw) is an eigen vector with respect to K, namely

ρgθ wk (dw)n/2 = e(2k+n)iθ wk (dw)n/2 .
It is easy to check now that H+ n is irreducible. Indeed, every invariant closed subspace
V has a topological basis consisting of eigenvectors of K, in other words wk (dw)n/2 for
some positive k must form a topological basis of V . Without loss of generality assume
that V contains (dw)n/2 , then by applying ρg one can get that (bw+a) 1
n (dw)
n/2
, and
1
in Taylor series for (bw+a)n all elements of the basis appear with non-zero coefficients.
That implies wk (dw)n/2 ∈ V for all k ≥ 0, hence V = H+ n.
One can construct another series of representations H−n by considering holomorphic
densities in the lower half-plane Im z < 0.
Principal series. These representations are parameterized by a continuous pa-
rameter s ∈ Ri (s 6= 0). Consider now the action of G on a real line by linear fractional
1+s
7 ax+b
transformations x → cx+d
. Let Ps+ denotes the space of densities ϕ (x) (dx) 2 with
G-action given by

1+s
ax + b 1+s
ρg ϕ (x) (dx) 2 = ϕ |cx + d|−s−1 (dx) 2 .
cx + d
The Hermitian product given by
Z ∞
(1.3) hϕ, ψi = ϕ̄ψdx
−∞
is invariant. The property of invariance justify the choice of weight for the density as
1+s 1+s̄
(dx) 2 (dx) 2 = dx, thus the integration is invariant. To check that the represen-
tation is irreducible one can move the real line to the unit circle as in the example
1+s
of discrete series and then use eikθ (dθ) 2 as an orthonormal basis in Ps+ . Note that
the eigen values of ρgθ in this case are e2kiθ for all integer k.
The second principal series Ps− can be obtained if instead of densities we consider
the pseudo densities which are transformed by the law

1+s
ax + b 1+s
ρg ϕ (x) (dx) 2 = ϕ |cx + d|−s−1 sgn (cx + d) dx 2 .
cx + d
Complementary series. Those are representations which do not appear in the
regular representation L2 (G). They can be realized as the representations in Cs of
1+s
all densities ϕ (x) (dx) 2 for real 0 < s < 1. An invariant Hermitian product is
Z ∞Z ∞
(1.4) hϕ, ψi = ϕ̄ (x) ψ (y) |x − y|s−1dxdy.
−∞ −∞
2. Semisimple modules and density theorem

We assume that R is a unital ring. Recall that an R-module is semi-simple if for
any submodule N ⊂ M there exists a submodule N ′ such that M = N ⊕ N ′ and
R-module M is simple if any submodule of M is either M or 0.
Lemma 2.1. Every submodule and every quotient of a semisimple module is semisim-
ple.
Proof. If Let N be a submodule of a semisimple module M, and let P be a submodule
of N. Since M = P ⊕ P ′ , then there exists an R-invariant projector p : M → P . The
restriction of p to N defines the projector N → P .
Lemma 2.2. Any semisimple R-module contains a simple submodule.
Proof. Let M be semisimple, m ∈ M. Let N be a maximal submodule in Rm which
does not contain m (exists by Zorn’s lemma, take all submodules which do not contain
m ). Then Rm is semisimple and Rm = N ⊕ N ′ . We claim that N ′ is simple. Indeed,
if N ′ is not simple, then it contains a proper submodule P . But m ∈/ P ⊕ N, since
P ⊕ N 6= Rm. That contradicts maximality of N.
4 VERA SERGANOVA
Lemma 2.3. The following conditions on a module M are equivalent

(1) M is semisimple;
P
(2) M = i∈I Mi for some simple submodules Mi ;
(3) M = ⊕j∈J Mj for some simple submodules Mj .
Proof. (1) ⇒ (2) Let {Mi }i∈I be the collection of all simple submodules. Let N =
′ ′
P
i∈I M i , assume that N 6= M, then M = N ⊕ N and N contains a simple submod-
ule. Contradiction. P
To prove (2) ⇒ (3) let J ⊂ I be minimal such that M = j∈J Mj (check that it
exists
P by Zorn’s lemma). By minimality of J for any k ∈ J , Mk does not belong to
j∈J−k Mj . Therefore M = ⊕j∈J Mj .
Finally, let us prove (3) ⇒ (1). Let N ⊂ M be a submodule and S ⊂ J be a
maximal subset such that N ∩ ⊕j∈S Mj = 0. Let M ′ = N ⊕ (⊕j∈S Mj ). We claim that
M ′ = M. Indeed, assume that the statement is false. Then there exists k such that
Mk ∩ M ′ = 0. But then N ∩ ⊕j∈S+k Mj = 0. Contradiction.
Lemma 2.4. Let M be a semisimple module. Then M is simple iff EndR (M) is a
division ring.
Proof. In one direction this is Shur’s lemma. In the opposite direction if M = M1 ⊕
M2 , then the projectors p1 , p2 satisfy p1 p2 = 0 and therefore p1 , p2 are not invertible.

Lemma 2.5. Let EndR (M) = K, EndK (M) = S. Then K b = EndR (M ⊕n ) ∼ =
⊕n ∼
Matn (K) and EndKb (M ) = S, the last isomorphism is given by the diagonal action
s (v1 , . . . , vn ) = (sv1 , . . . , svn ) .
Proof. See similar statement in lecture notes 3.
Theorem 2.6. (Jacobson-Chevalley density theorem). Let M be a semisimple R-
module, K = EndR (M), S = EndK (M). Then for any v1, . . . , vn ∈ M and X ∈ S
there exists r ∈ R such that rvi = Xvi for all i = 1, . . . , n.
Proof. First, let us prove it for n = 1. It suffices to show that Rv is S-invariant.
Indeed, M = Rv ⊕ N, and p be the projector on N with kernel Rv. Then p ∈ K,
hence Ker p is S-invariant.
For arbitrary n, note that M ⊕n is semisimple and use Lemma 2.5. Then for any
X ∈ S, v = (v1 , . . . , vn ) there exists r ∈ R such that
r (v1, . . . , vn ) = X (v1, . . . , vn ) .

Corollary 2.7. Let M be a semisimple R-module, K = EndR (M), and M is finitely
generated over K. Then the natural map R → EndK (M) is surjective.
Corollary 2.8. Let R be an algebra over an algebraically closed field k, and ρ : R →
Endk (V ) be an irreducible finite-dimensional representation. Then ρ is surjective.
Corollary 2.9. Let R, ρ and V be as in previous statement but k is not algebraically

= EndD (V ) for some division ring D containing k.
closed. Then ρ (R) ∼
3. Semisimple rings
A ring R is semi-simple if every R-module is semisimple. For example, a group
algebra k (G) for a finite group G such that char k does not divide |G| is semisimple.
Lemma 3.1. Let R be semisimple, then R is a finite direct sum of its minimal left
ideals.
Proof. R is semisimple over itself. Hence R = ⊕i∈I Li , where Li are minimal left
ideals. But 1 = l1 + · · · + ln , hence I is finite.
Theorem 3.2. (Wedderburn-Artin) Every semisimple ring is isomorphic to Matn1 (D1 )×
· · · × Matnk (Dk ) for some division rings D1 , . . . , Dk .
Proof. Consider the submodules J1 , . . . , Jk , each Ji is the sum of all isomorphic min-
imal left ideals. Then Ji is a two-sided ideal and
R = J1 ⊕ · · · ⊕ Jk ,
where Ji = li⊕ni .
Let Diop
= EndR (li), then
op op
Rop ∼
= EndR (R) = EndR (J1 ) × · · · × EndR (Jk ) = Matn1 (D1 ) × · · · × Matnk (Dk ) .

PROBLEM SET # 8
MATH 252
Due October 28.

1. Classify all irreducible (continuous) representations of O2 (the group of orthog-
onal 2 × 2-matrices).
2. Check that the Hermitian products (1.2) and (1.4) defined in lecture notes 8
are invariant.
+
3. Show that the operator T : Ps+ → P−s defined by the formula
Z ∞
1+s 1−s
s−1
T φ (x) dx 2 = φ (x) |x − y| dx dy 2
−∞
+
is intertwining. Therefore Ps+ is isomorphic to P−s .

1
WEEK 9
1. Jordan-Hölder theorem and indecomposable modules

Let M be a module satisfying ascending and descending chain conditions (ACC
and DCC). In other words every increasing sequence submodules M1 ⊂ M2 ⊂ . . . and
any decreasing sequence M1 ⊃ M2 ⊃ . . . are finite. Then it is easy to see that there
exists a finite sequence
M = M0 ⊃ M1 ⊃ · · · ⊃ Mk = 0
such that Mi /Mi+1 is a simple module. Such a sequence is called a Jordan-Hölder
series. We say that two Jordan Hölder series
M = M0 ⊃ M1 ⊃ · · · ⊃ Mk = 0, M = N0 ⊃ N1 ⊃ · · · ⊃ Nl = 0
are equivalent if k = l and for some permutation s Mi /Mi+1 ∼
= Ns(i) /Ns(i)+1 .
Theorem 1.1. Any two Jordan-Hölder series are equivalent.
Proof. We will prove that if the statement is true for any submodule of M then it
is true for M. (If M is simple, the statement is trivial.) If M1 = N1 , then the
statement is obvious. Otherwise, M1 + N1 = M, hence M/M1 ∼ = N1 / (M1 ∩ N1 ) and
M/N1 = M1 / (M1 ∩ N1). Consider the series
∼
M = M0 ⊃ M1 ⊃ M1 ∩N1 ⊃ K1 ⊃ · · · ⊃ Ks = 0, M = N0 ⊃ N1 ⊃ N1∩M1 ⊃ K1 ⊃ · · · ⊃ Ks = 0.
They are obviously equivalent, and by induction assumption the first series is equiv-
alent to M = M0 ⊃ M1 ⊃ · · · ⊃ Mk = 0, and the second one is equivalent to
M = N0 ⊃ N1 ⊃ · · · ⊃ Nl = {0}. Hence they are equivalent.
Thus, we can define a length l (M) of a module M satisfying ACC and DCC, and
if M is a proper submodule of N, then l (M) < l (N).
A module M is indecomposable if M = M1 ⊕ M2 implies M1 = 0 or M2 = 0.
Lemma 1.2. Let M and N be indecomposable, α ∈ HomR (M, N), β ∈ HomR (N, M)
be such that β ◦ α is an isomorphism. Then α and β are isomorphisms.
Proof. We claim that N = Im α ⊕ Ker β. Indeed, Im α ∩ Ker β = 0 and for any x ∈ N
one can write x = y + z, where y = α ◦ (β ◦ α)−1 ◦ β (x), z = x − y. Then since N is
indecomposable, Im α = N, Ker β = 0 and N ∼ = M.

1
Lemma 1.3. Let M be indecomposable module of finite length and ϕ ∈ EndR (M),
then either ϕ is an isomorphism or ϕ is nilpotent.
Proof. There is n > 0 such that Ker ϕn = Ker ϕn+1 , Im ϕn = Im ϕn+1 . In this case
Ker ϕn ∩ Im ϕn = 0 and hence M ∼= Ker ϕn ⊕ Im ϕn . Either Ker ϕn = 0, Im ϕn = M
or Ker ϕn = M. Hence the lemma.
Lemma 1.4. Let M be as in Lemma 1.3 and ϕ, ϕ1, ϕ2 ∈ EndR (M), ϕ = ϕ1 + ϕ2. If
ϕ is an isomorphism then at least one of ϕ1 , ϕ2 is also an isomorphism.
Proof. Without loss of generality we may assume that ϕ = id. But in this case ϕ1
and ϕ2 commute. If both ϕ1 and ϕ2 are nilpotent, then ϕ1 + ϕ2 is nilpotent, but this
is impossible as ϕ1 + ϕ2 = id.
Corollary 1.5. Let M be as in Lemma 1.3. Let ϕ = ϕ1 + · · · + ϕk ∈ EndR (M). If
ϕ is an isomorphism then ϕi is an isomorphism at least for one i.
It is obvious that if M satisfies ACC and DCC then M has a decomposition
M = M1 ⊕ · · · ⊕ Mk ,
where all Mi are indecomposable.
Theorem 1.6. (Krull-Schmidt) Let M be a module of finite length and
M = M1 ⊕ · · · ⊕ Mk = N1 ⊕ · · · ⊕ Nl
for some indecomposable Mi and Nj . Then k = l and there exists a permutation s
such that Mi ∼
= Ns(j) .
Proof. Let pi : M1 → Ni be the restriction to M1 of the natural projection M → Ni ,
and qj : Nj → M1 be the restriction to Nj of the natural projection M → M1 . Then
obviously q1p1 + · · · + ql pl = id, and by Corollary 1.5 there exists i such that qi pi is
an isomorphism. Lemma 1.2 implies that M1 ∼ = Ni . Now one can easily finish the
proof by induction on k.
2. Some facts from homological algebra

The complex is the graded abelian group C· = ⊕i≥0 Ci . We will assume later that
all Ci are R-modules for some ring R. A differential is an R-morphism of degree −1
such that d2 = 0. Usually we realize C· by the picture
d d
→ · · · → C1 −
− → C0 → 0.
We also consider d of degree 1, in this case the superindex C · and
d d
0 → C0 −
→ C1 −
→ ...
All the proofs are similar for these two cases.
Homology group is Hi (C) = (Ker d ∩ Ci ) /dCi+1 .
Given two complexes (C· , d) and (C·′ , d′). A morphism f : C· → C·′ preserving
grading and satisfying f ◦ d = d′ ◦ f is called a morphism of complexes. A morphism
of complexes induces the morphism f∗ : H· (C) → H· (C ′).
Theorem 2.1. (Long exact sequence). Let
g f
→ C·′ −
0 → C· − → C·′′ → 0
be a short exact sequence, then the long exacts sequence
δ g∗ f∗ δ g∗
− → Hi (C ′) −
→ Hi (C) − → Hi (C ′′) − → ...
→ Hi−1 (C) −
where δ = g −1 ◦ d′ ◦ f −1 , is exact.
Let f, g : C· → C·′ be two morphisms of complexes. We say that f and g are
homotopically equivalent if there exists h : C· → C·′ (+1) (the morphism of degree 1)
such that f − g = h ◦ d + d′ ◦ h.
Lemma 2.2. If f and g are homotopically equivalent then f∗ = g∗ .
Proof. Let φ = f − g, x ∈ Ci and dx = 0. Then
φ (x) = h (dx) + d′ (hx) = d′ (hx) ∈ Im d′ .
Hence f∗ − g∗ = 0.
We say that complexes C· and C·′ are homotopically equivalent if there exist f :
C· → C·′ and g : C·′ → C· such that f ◦ g is homotopically equivalent to idC ′ and g ◦ f
is homotopically equivalent to idC . Lemma 2.2 implies that homotopically equivalent
complexes have isomorphic homology. The following Lemma is straightforward.
Lemma 2.3. If C· and C·′ are homotopically equivalent then the complexes HomR (C·, B)
and HomR (C·′) are also homotopically equivalent.
Note that the differential in HomR (C· , B) has degree 1.
3. Projective modules
An R-module P is projective if for any surjective morphism φ : M → N and any
ψ : P → N there exists f : P → M such that ψ = φ ◦ f.
Example. A free module is projective. Indeed, let {ei }i∈I be the set of free
generators of a free module F , i.e. F = ⊕i∈I Rei . Define f : F → M by f (ei ) =
φ−1 (ψ (ei )).
Lemma 3.1. The following conditions on a module P are equivalent
(1) P is projective;
= P ⊕ P ′;
(2) There exists a free module F such that F ∼
(3) Any exact sequence 0 → N → M → P → 0 splits.
Proof. (1) ⇒ (3) Consider the exact sequence

ϕ ψ
0→N −
→M −
→ P → 0,
then since ψ is surjective, there exists f : P → M such that ψ ◦ f = idP .
(3) ⇒ (2) Every module is a quotient of a free module. Therefore we just have to
apply (3) to the exact sequence
0→N →F →P →0
for a free module F .
(2) ⇒ (1) Let φ : M → N be surjective and ψ : P → N. Choose a free module F
so that F = P ⊕ P ′ . Then extend ψ to F → N in the obvious way and let f : F → M
be such that φ ◦ f = ψ. Then the last identity is true for the restriction of f to P .
A projective resolution of M is a complex P· of projective modules such that

= M. A projective resolution always exists since one
Hi (P· ) = 0 for i > 0 and H0 (P· ) ∼
can easily construct a resolution by free modules. Below we prove the “uniqueness”
statement.
Lemma 3.2. Let P· and P·′ be two projective resolutions of the same module M.
Then there exists a morphism f : P· → P·′ of complexes such that f∗ : H0 (P· ) →
H0 (P·′ ) induces the identity idM . Any two such morphisms f and g are homotopically
equivalent.
Proof. Construct f inductively. Let p : P0 → M and p′ : P0′ → M be the natural
projections, define f : P0 → P0′ so that p′ ◦ f = p. Then
f (Ker p) ⊂ Ker p′ , Ker p = d (P1 ) , Ker p′ = d′ (P1′ ) ,
hence f ◦ d (P1 ) ⊂ d′ (P1′ ), and one can construct f : P1 → P1′ such that f ◦ d = d′ ◦ f.
Proceed in the same manner to construct f : Pi → Pi .
To check the second statement, let ϕ = f − g. Then p′ ◦ ϕ = 0. Hence
ϕ (P0 ) ⊂ Ker p′ = d′ (P1′ ) .
Therefore one can find h : P0 → P1′ such that d′ ◦ h = ϕ. Furthermore,
d′ ◦ h ◦ d = ϕ ◦ d = d′ ◦ ϕ,
hence
(ϕ − h ◦ d) (P1 ) ⊂ P1′ ∩ Ker d′ = d′ (P2′ ) .
Thus one can construct h : P1 → P2′ such that d′ ◦ h = ϕ − h ◦ d. Then proceed
′
inductively to define h : Pi → Pi+1 .
Corollary 3.3. Every two projective resolutions of M are homotopically equivalent.
Let M and N be two modules and P· be a projective resolution of M. Consider

the complex
0 → HomR (P0 , N) → HomR (P1 , N) → . . . ,
where the differential is defined naturally. The cohomology of this complex is denoted
by Ext·R (M, N). Lemma 2.3 implies that Ext·R (M, N) does not depend on a choice
of projective resolution for M. Check that Ext0R (M, N) = HomR (M, N).
Example 1. Let R = C [x] be the polynomial ring. Any simple R-module is
one-dimensional and isomorphic to C [x] / (x − λ). Denote such module by Cλ . A
projective resolution of Cλ is
d
0 → C [x] −
→ C [x] → 0,
where d (1) = x − λ. Let us calculate Ext· (Cλ , Cµ ). Note that HomC[x] (Cµ ) = C,
hence we have the complex
d∗
0 → C −→ C → 0
where d∗ = λ−µ. Hence Ext· (Cλ , Cµ ) = 0 if λ 6= µ and Ext0 (Cλ , Cλ ) = Ext1 (Cλ , Cλ) =
C.
Example 2. Let R = C [x] / (x2 ). Then R has one up to isomorphism simple
module, denote it by C0. A projective resolution for C0 is
d d
... −
→R−
→ R → 0,
where d (1) = x and Exti (C0 , C0) = C for all i ≥ 0.
4. Representations of artinian rings

An artinian ring is a unital ring satisfying the descending chain condition for
left ideals. We will see that an artinian ring is a finite length module over itself.
Therefore R is automatically noetherian. A typical example of an artinian ring is a
finite-dimensional algebra over a field.
Theorem 4.1. Let R be an artinian ring, I ⊂ R be a left ideal. If I is not nilpotent,
then I contains an idempotent.
Proof. Let J be a minimal left ideal, such that J ⊂ I and J is not nilpotent. Then
J 2 = J . Let L be a minimal left ideal such that L ⊂ J and J L 6= 0. Then there
is x ∈ L such that J x 6= 0. But then J x = L by minimality of L. Thus, for some
r ∈ R, rx = x, hence r2 x = rx and (r2 − r) x = 0. Let N = {y ∈ J | yx = 0}. Then
N is a proper left ideal in J and therefore N is nilpotent. Thus, we obtain
r2 ≡ r mod N.
Let n = r2 − r, then
(r + n − 2rn)2 ≡ r2 + 2rn − 4r2 n mod N 2 ,
r2 + 2rn − 4r2 n ≡ r + n − 2rn mod N 2 .
Hence r1 = r + n − 2rn is an idempotent modulo N 2 . Repeating this process several

times we obtain an idempotent.
Corollary 4.2. If an artinian ring does not have nilpotent ideals, then it is semisim-
ple.
Proof. The sum S of all minimal left ideals is semisimple. By DCC S is a finite direct
sum of minimal left ideals. Then S contains an idempotent e, which is the sum of
idempotents in each direct summand. Then R = S ⊕ R(1 − e), however that implies
R = S.
Important notion for a ring is the radical. For an R-module M let
Ann M = {x ∈ R | xM = 0} .
Then the radical rad R is the intersection of Ann M for all simple R-modules M.
Theorem 4.3. If R is artinian then rad R is a maximal nilpotent ideal.
Proof. First, let us show that rad R is nilpotent. Assume the contrary. Then rad R
contains an idempotent e. But then e does not act trivially on a simple quotient of
Re. Contradiction.
Now let us show that any nilpotent ideal N lies in rad R. Let M be a simple
module, then NM 6= M as N is nilpotent. But NM is a submodule of M. Therefore
NM = 0. Hence N ⊂ Ann M for any simple M.
Corollary 4.4. An artinian ring R is semisimple iff rad R = 0.
Corollary 4.5. If R is artinian, then R/ rad R is semisimple.
Corollary 4.6. If R is artinian and M is an R-module, then for the filtration
M ⊃ (rad R) M ⊃ (rad R)2 M ⊃ · · · ⊃ (rad R)k M = 0
all quotients are semisimple. In particular, M always has a simple quotient.
Theorem 4.7. If R is artinian, then it has finite length as a left module over itself.
Proof. Consider the filtration R = R0 ⊃ R1 ⊃ · · · ⊃ Rs = 0 where Ri = (rad R)i .
Then each quotient Ri /Ri+1 is semisimple of finite length. The statement follows.
Let R be an artinian ring. By Krull-Schmidt theorem R (as a left module over
itself) has a decomposition into direct sum of indecomposable submodules R = L1 ⊕
· · · ⊕ Ln . Since EndR (R) = Rop, the projector on each component Li is given by
multiplication on the right by some idempotent ei . Thus, R = Re1 ⊕ · · · ⊕ Ren ,
where ei are idempotents and ei ej = 0 if i 6= j. This decomposition is unique up
to multiplication on some unit on the right. Since Rei is indecomposable, ei can
not be written as a sum of two orthogonal idempotents, such idempotents are called
primitive. Each module Rei is projective.
Lemma 4.8. Let R be artinian, N = rad R and e be a primitive idempotent. Then

Ne is a unique maximal submodule of Re.
Proof. Since Re is indecomposable, every proper left ideal is nilpotent, (otherwise it
has an idempotent and therefore splits as a direct summand in Re). But then this
ideal is in N ∩ Re = Ne.
A projective module P is a projective cover of M if there exists a surjection P → M.
Theorem 4.9. Let R be artinian. Every simple R-modules S has a unique (up to an
isomorphism) indecomposable projective cover isomorphic to Re for some primitive
idempotent e ∈ R. Every indecomposable projective module has a unique (up to an
isomorphism )simple quotient.
Proof. Every simple S is a quotient of R, and therefore it is a quotient of some
indecomposable projective P = Re. Let φ : P → S be the natural projection. For
any indecomposable projective cover P1 of S with surjective morphism φ1 : P1 → S
there exist f : P → P1 and g : P1 → P such that φ = φ1 ◦ f and φ1 = φ ◦ g. Therefore
φ = φ ◦ g ◦ f. Since Ker φ is the unique maximal submodule, g ◦ f (P1 ) = P . In
particular g is surjective. Indecomposability of P1 implies P ∼ = P1 . Thus, every
simple module has a unique indecomposable projective cover.
On the other hand, let P be an indecomposable projective module. Corollary 4.6
implies that P has a simple quotient S. Hence P is isomorphic to the indecomposable
projective cover of S.
Corollary 4.10. Every indecomposable projective module over an artinian ring R is
isomorphic to Re for some primitive idempotent e ∈ R. There is a bijection between
the ismorphism classes of simple R-modules and isomorphism classes of projective
indecomposable R-modules.
Example. Let R = F3 (S3 ). Let r be a 3-cycle and s be a transposition. Then
r − 1, r2 − 1, sr − s and sr2 − s span a maximal nilpotent ideal. Hence R has two (up
to an isomorphism) simple modules L1 and L2 , where L1 is a trivial representation
of S3 and L2 is a sign representation. Choose primitive idempotents e1 = −s − 1 and
e2 = s − 1, then 1 = e1 + e2 . Hence R has two indecomposable projective modules
P1 = Re1 and P2 = Re2 . Note that
Re1 ∼ S2 = IndS3 (sgn) .
= IndS3 (triv) , Re2 ∼ S2
Thus, P1 is just 3-dimensional permutation representation of S3, and P2 is obtained

from P1 by tensoring with sgn. It is easy to see that P1 has a trivial submodule
as well as a trivial quotient, and sgn is isomorphic to the quotient of the maximal
submodule of P1 by the trivial submodule. One can easily get a similar description
for P2 Thus, one has the the following exact sequences
0 → L2 → P2 → P1 → L1 → 0, 0 → L1 → P1 → P2 → L2 → 0,
therefore
· · · → P1 → P2 → P2 → P1 → P1 → P2 → P2 → P1 → 0
is a projective resolution for L1 , and
· · · → P2 → P1 → P1 → P2 → P2 → P1 → P1 → P2 → 0
is a projective resolution for L2 . Now one can calculate Ext between simple modules
Extk (Li , Li ) = 0 if k ≡ 1, 2 mod 4, Extk (Li , Li ) = F3 if k ≡ 0, 3 mod 4,
and if i 6= j, then
Extk (Li , Lj ) = 0 if k ≡ 0, 3 mod 4, Extk (Li , Lj ) = F3 if k ≡ 1, 2 mod 4.
PROBLEM SET # 1
MATH 252
Due November 4.
1. Let R be the algebra of polynomial differential operators. In other words R is
∂
generated by x and ∂x with relation
∂ ∂
x−x = 1.
∂x ∂x
(The algebra R is called the Weyl algebra.) Let M = C [x] have a structure of R-
module in the natural way. Show that EndR (M) = C, M is an iireducible R-module
and the natural map R → EndC (M) is not surjective.
2. Let R be a subalgebra of upper triangular matrices in Matn (C). Classify
simple and indecomposable projective modules over R and evaluate Ext·R (M, N) for
all simple M and N.

1
WEEKS 10 − 11
1. Representations of quivers
I follow here Crawley-Boevey lectures trying to give more details concerning ex-
tensions and exact sequences.
A quiver is an oriented graph. If Q is a quiver, then we denote by Q0 the set of
vertices and by Q1 the set of arrows. Usually we denote by n the number of vertices.
If γ : j ← i is an arrow then i = s (γ), j = t (γ).
Fix an algebraically closed field k. A representation V of a quiver is a collection of
vector spaces {Vi }i∈Q0 and linear maps ργ : Vi → Vj for each arrow γ : i → j. For two
representations of a quiver Q, ρ in V and σ in W define a homomorphism φ : V → W
as a set of linear maps φi : Vi → Wi such that the diagram
ργ
Vj ←− Vi
↓ φj ↓ φi
σγ
Wj ←− Wi
is commutative. We say that two representations V and W are isomorphic if there

is a homomorphism φ ∈ HomQ (V, W ) such that each φi is an isomorphism. One can
define a subrepresentation and a direct sum of representation of Q in the natural way.
A representation is irreducible if it does not have non-trivial proper subrepresentation
and indecomposable if it is not a direct sum of non-trivial subrepresentations.
Example 1.1. Let Q be the quiver • → •. A representation of Q is a pair of vector

spaces V and W and a linear operator ρ : V → W . Let V0 = Ker ρ, V1 is such that
V = V0 ⊕ V1 , W0 = Im ρ, and W1 is such that W = W0 ⊕ W1 . Then V0 → 0, V1 → W0
and 0 → W1 are subrepresentations and ρ is their direct sum. Furthermore, V0 → 0
is the direct sum of dim V0 copies of k → 0, V1 → W0 is the direct sum of dim V1
copies of k → k and finally 0 → W1 is the direct sum of dim W1 copies of 0 → k.
Thus, we see that there are exactly three isomorphism classes of indecomposable
representations of Q, 0 → k, k → k, k → 0. The first and the last one are irreducible,
0 → k is a subrepresentaion of k → k and k → 0 is a quotient of k → k by 0 → k.
Date: December 4, 2005.

1
2 REPRESENTATION THEORY. WEEKS 10 − 11
2. Path algebra
Given a quiver Q. A path p is a sequence γ1 . . . γk of arrows such that s (γi ) =
t (γi+1 ). Put s (p) = s (γk ), t (p) = t (γ1 ). Define a composition p1 p2 of two paths
such that s (p1 ) = t (p2 ) in the obvious way and we set p1 p2 = 0 if s (p1 ) 6= t (p2 ).
Introduce also elements ei for each vertex i ∈ Q0 and define ei ej = δij ei , ei p = p if
i = t (p) and 0 otherwise, pei = p if i = s (p) and 0 otherwise. The path algebra k (Q)
is the set of k-linear combinations of all paths and ei with composition extended by
linearity from ones defined above.
One can easily check the following properties of a path algebra
(1) k (Q) is finite-dimensional iff Q does not have oriented cycles;
(2) If Q is a disjoint union of Q1 and Q2, then k (Q) = k (Q1 ) × k (Q2);
(3) The algebra k (Q) has a natural Z-grading ⊕∞ n=0 k (Q)n defined by deg ei = 0
and the degree of a path p being the length of the path;
(4) Elements ei are primitive idempotents of k (Q), and hence k (Q) ei is an inde-
composable projective k (Q)-module.
The first three properties are trivial, let us check the last one. Suppose ei is not
primitive, then one can find an idempotent ε ∈ k (Q) ei . Let ε = c0ei +c1 p1 +· · ·+ck pk ,
where s (pj ) = i for all j ≤ k. Then ε2 = ε implies c0 = 0 or 1. Let c0 = 0, ε = εl +. . . ,
where deg εl = l and other terms have degree greater than l. But then ε2 starts with
degree greater than 2l, hence ε = 0. If c0 = 1, apply the same argument to the
idempotent (ei − ε).
Given a representation ρ of Q one can construct a k (Q)-module
M
V = Vi , ei Vj = δij IdVj , γv = ργ v if v ∈ Vs(γ) , γ (v) = 0 otherwise.
i∈Q0
For any path p = γ1 . . . γk and v ∈ V put pv = ργ1 ◦ · · · ◦ ργk (v).

On the other hand, every k (Q)-module V defines a representation ρ of Q if one
puts Vi = ei V .
The following theorem is straightforward.
Theorem 2.1. The category of representations of Q and the category of k (Q)-
modules are equivalent.
Lemma 2.2. The radical of k (Q) is spanned by all paths p satisfying the property
that there is no return paths, i.e. back from t (p) to s (p).
Proof. It is easy to see that the paths with no return span a two-sided ideal R. Note
that Rn = 0, where n is the number of vertices. Thus, R ⊂ rad k (Q). On the other
hand, let y ∈/ R and p be a shortest path in decomposition of y which has a return
path. Choose a shortest path s such that τ = sp is an oriented cycle. Consider the
representation of Q which has k in each vertex of τ and 0 in all other vertices. Let
ργ = Id, if γ is included in τ and ργ = 0 otherwise. Let V be the corresponding k (Q)-
module. Then V is simple, sy (V ) 6= 0. Hence y ∈ / rad k (Q). Contradiction.
REPRESENTATION THEORY. WEEKS 10 − 11 3
Example 2.3. If Q has one vertex and n loops then k (Q) is a free associative algebra
with n generators. If Q does not have cycles, then k (Q) is the subalgebra in Matn (k)
generated by elementary matrices Eii for each i ∈ Q0 and Eij for each arrow i → j.
3. Standard resolution
Theorem 3.1. Let Q be a quiver, A = k (Q) and V be an A-module. Then the
sequence
f g
M M
0→ Aej ⊗ Vi −
→ Aei ⊗ Vi −
→ V → 0,
γ=(i→j)∈Q1 i∈Q0
where f (aej ⊗ v) = aej γ ⊗ v − aej ⊗ γv, g (aei ⊗ v) = av for any v ∈ Vi , is exact. It

is a projective resolution.
Proof. First, check that g ◦ f = 0. Indeed,
g (f (aej ⊗ v)) = g (aej γ ⊗ v − aej ⊗ γv) = aej γv − aej γv = 0.
Since V = ⊕eiVi , g is surjective. To check that f is injective, introduce the grading
on A ⊗ V using deg V = 0. By grf denote the homogeneous part of highest degree
for f. Note that the grf increases the degree by one and
grf = ⊕γ∈Q1 fγ , where fγ : Aej ⊗ Vi → Aγ ⊗ Vi is defined by
fγ (aej ⊗ vi ) = aej γ ⊗ vi,
for γ : i → j. One can see from this formula that fγ is injective, therefore grf is
injective and hence f is injective.
To prove that Im f = Ker g note that
aej γ ⊗ v ≡ aej ⊗ γv mod Im f,
L
therefore for any x ∈ i∈Q0 Aei ⊗ Vi
x ≡ x0
mod Im f
L
for some x0 of degree 0. In other words x0 ∈ i∈Q0 kei ⊗ Vi . If g (x) = 0, then
g (x0) = 0, and if g (x0) = 0, then obviously x0 = 0. Hence x ≡ 0 mod Im f.
Theorem 3.1 implies that Ext1 (X, Y ) can be calculated as coker d of the following
complex
d
M M
(3.1) 0→ Homk (Xi , Yi ) −
→ Homk (Xi , Yj ) → 0,
i∈Q0 γ=(i→j)∈Q1
where
(3.2) dφ (x) = φ (γx) − γφ (x)
for any x ∈ Xi , γ = (i → j).
Lemma 3.2. Every ψ ∈ Ext1 (X, Y ) induces a non-split exact sequence

0 → Y → Z → X → 0.
If Ext1 (X, Y ) = 0, then every exacts sequence as above splits.
Proof. Let
0→Y →Z →X →0
be an exact sequence of representations of Q. Then Zi can be identified with Xi ⊕ Yi
for every i. For every arrow γ : i → j the action on Z is defined by
γ (x, y) = (γx, γy + ψγ (x)) ,
for some ψγ ∈ Homk (Xi , Yj ). Thus, ψ can be considered as an element in the second
non-zero term of (3.1). If the exact sequence splits, then there is η ∈ HomQ (X, Z)
such that for each i ∈ Q0, x ∈ Xi
η (x) = (x, φi (x)) ,
for some φi ∈ Homk (Xi , Yi ). Furthermore, η ∈ HomQ (X, Z) iff for each γ : i → j
γ (x, φi (x)) = (γx, γφi (x) + ψγ (x)) = (γx, φj (γx)) ,
which implies
ψγ (x) = φj (γx) − γφi (x) .
In other words, ψ = dφ. Thus, Ext1 (X, Y ) parameterizes the set of non-split exact
sequences
0 → Y → Z → X → 0.

Corollary 3.3. In the category of representations of Q, Exti (X, Y ) = 0 for i ≥ 2.
Corollary 3.4. Let
0→Y →Z →X →0
be a short exact sequence of representations of Q, then
Ext1 (V, Z) → Ext1 (V, X) , Ext1 (Z, V ) → Ext1 (Y, V )
are surjective.
Lemma 3.5. If X and Y are indecomposable and Ext1 (Y, X) = 0, then every non-
zero ϕ ∈ HomQ (X, Y ) is either surjective or injective.
Proof. Use the exact sequences
0 → Ker ϕ → X → Im ϕ → 0,
(3.3) = Y/ Im ϕ → 0.
0 → Im ϕ → Y → S ∼
The exact sequence (3.3) can be considered as an element ψ ∈ Ext1 (S, Im ϕ) by

use of Lemma 3.2. By Corollary 3.3 we have an isomorphism g : Ext1 (S, Im ϕ) ∼
=
Ext1 (S, X). Then g (ψ) induces the exact sequence
0 → X → Z → S → 0,
and this exact sequence together with (3.3) form the following commutative diagram
α
0 → X − → Z → S → 0
β
↓ ↓γ ↓
δ
0 → Im ϕ −
→ Y → S → 0
here β and γ are surjective. We claim that the sequence

α+β γ−δ
0 → X −−→ Z ⊕ Im ϕ −−→ Y → 0
is exact. Indeed, α + β is obviously injective and γ − δ is surjecitve. Finally, dim Z =
dim X + dim S, dim Im ϕ = dim Y − dim S. Therefore,
dim (Z ⊕ Im ϕ) = dim X + dim Y,
and therefore Ker (γ − δ) = Im (α + β).
But Ext1 (Y, X) = 0. Hence the last exact sequence splits, Z ⊕ Im ϕ ∼
= X⊕Yand
by Krull-Schmidt theorem either X = Im ϕ or Y = Im ϕ.
∼ ∼
Intoduce dim X as a vector x = (x1 , . . . , xn ) ∈ Zn where n is the number of vertices
and xi = dim Xi . Define the bilinear form
X X
hx, yi = xi yi − xi yj = dim HomQ (X, Y ) − dim Ext1 (X, Y )
i∈Q0 (i→j)∈Q1
(the equality follows from (3.1)). We also introduce the symmetric form
(x, y) = hx, yi + hy, xi
and the quadratic form
q (x) = hx, xi .
4. Bricks
Here we discuss further properties of finite-dimensional representations of A =
k (Q).
Recall that if X is indecomposable and has finite length, then ϕ ∈ EndQ (X) is
either isomorphism or nilpotent. Since we assumed that k is algebraically closed, ϕ =
λ Id for any invertible ϕ ∈ EndQ (X). A representation X is a brick, if EndQ (X) = k.
If X is a brick, then X is indecomposable. If X is indecomposable and Ext1 (X, X) =
0, then X is a brick due to Lemma 3.5.
Example 4.1. Consider the quiver • → •. Then every indecomposable is a brick.

For the Kronecker quiver • ⇒ • the representation k 2 ⇒αβ k 2 with α = Id, β = (0100 ) is
01
not a brick. Indeed, ϕ = (ϕ1, ϕ2 ) where ϕ1, ϕ2 are matrices (00), belongs to EndQ (X).
Lemma 4.2. Let X be indecomposable and not a brick, then X contains a brick W
such that Ext1 (W, W ) 6= 0.
Proof. Choose ϕ ∈ EndQ (X), ϕ 6= 0 of minimal rank. Since rk ϕ2 < rk ϕ, ϕ2 = 0. Let
Y = Im ϕ, Z = Ker ϕ. Let Z = Z1 ⊕ · · · ⊕ Zp be a sum of indecomposables. Let pi :
Z → Zi be the projection. Choose i so that pi (Y ) 6= 0 and let η = pi ◦ ϕ ∈ EndQ (X)
(well defined since Im ϕ ∈ Ker ϕ). Note that by our assumption rk η = rk ϕ, therefore
pi : Y → Zi is an embedding. Let Yi = pi (Y ). Then Ker η = Z, Im η = Yi .
We claim now that Ext1 (Zi , Zi ) 6= 0. Indeed, Ext1 (Yi , Z) 6= 0 by exact sequence
η
0→Z→X−
→ Yi → 0
and indecomposability of X. Then the induced exact sequence
η
0 → Zi → Xi −
→ Yi → 0
does not split also. (If it splits, then Zi is a direct summand of X, which is impossible).
Therefore Ext1 (Yi , Zi ) 6= 0. But Yi is a submodule of Zi . By Corollary 3.4 we have
the surjection
Ext1 (Zi , Zi ) → Ext1 (Yi , Zi ) .
If Zi is not a brick, we repeat the above construction for Zi e.t.c. Finally, we get
a brick.
Corollary 4.3. Assume that the quadratic form q is positive definite. Then every
indecomposable X is a brick with trivial Ext1 (X, X); moreover, if x = dim X, then
q (x) = 1.
Proof. Assume that X is not a brick, then it contains a brick Y such that Ext1 (Y, Y ) 6=
0. Then
q (Y ) = dim EndQ (Y ) − dim Ext1 (Y, Y ) = 1 − dim Ext1 (Y, Y ) ≤ 0,
but this is impossible. Therefore X is a brick. Now
q (x) = dim EndQ (X) − dim Ext1 (X, X) = 1 − dim Ext1 (X, X) ≥ 0,
hence q (x) = 1 and dim Ext1 (X, X) = 0.
5. Orbits in representation variety

Fix a quiver Q, recall that n denotes the number of vertices. Let x = (x1, . . . , xn ) ∈
Zn≥0 . Define
Y
Rep (x) = Homk (k xi , k xj ) .
(i→j)∈Q1
It is clear that every representation of Q of dimension x is a point in Rep (x). Let

Y
GL k i .

G=
i∈Q0
Then G acts on Rep (x) by the formula gϕij = gi ϕgj−1 , for each arrow i → j. Two
representations of Q are isomorphic iff they belong to the same orbit of G. For a
representation X we denote by OX the corresponding G-orbit in Rep (x).
Note that X X
dim Rep (x) = xi xj , dim G = x2i ,
(i→j)∈Q1 i∈Q0
therefore
(5.1) dim Rep (x) − dim G = −q (x) .
Since G is an affine algebraic group acting on an affine algebraic variety, we can work
in Zariski topology. Then each orbit is open in its closure, if O and O′ are two orbits
and O′ ⊂ Ō, O 6= O′ , then dim O′ < dim O. Finally, we need the formula
dim OX = dim G − dim StabX ,
here StabX stands for the stabilizer of X. Also note that in our case the group G is
connected, therefore each G-orbit is irreducible.
Lemma 5.1. dim StabX = dim AutQ (X) = dim EndQ (X).
Proof. The condition that φ ∈ EndQ (X) is not invertible is given by the polynomial
equations det φi = 0. Since AutQ (X) is not empty, we are done.
Corollary 5.2.
codim OX = dim Rep (x)−dim G+dim StabX = −q (x)+dim EndQ (X) = dim Ext1 (X, X) .
Lemma 5.3. Let Z be a nontrivial extension of Y by X, i.e. there is a non-split
exact sequence
0 → X → Z → Y → 0.
Then OX⊕Y ⊂ ŌZ and OX⊕Y 6= OZ .
Proof. Write each Zi as Xi ⊕ Yi and define giλ |Xi = Id, giλ |Yi = λ Id for any λ =
6 0.
λ
Then obviously X ⊕ Y belongs to the closure of gi (Z). It is left to check that
X ⊕Y ∼ 6 Z. But the sequence is non-split, therefore
=
dim HomQ (Y, Z) < dim HomQ (Y, X ⊕ Y ) .

Corollary 5.4. If OX is closed then X is semisimple.
6. Dynkin and affine graphs

Let Γ be a connected graph with n vertices, then Γ defines a symmetric bilinear
form (·, ·) on Zn X X
(x, y) = 2xi yi − xi yj .
i∈Γ0 (i,j)∈Γ1
If Γ is equipped with orientation then the symmetric form coincides with the intro-
duced earlier symmetric form of the corresponding quiver. The matrix of the form
(·, ·) in the standard basis is called the Cartan matrix of Γ.
2−1

Example 6.1. The Cartan matrix of • − • is −12 .
Theorem 6.2. Given a connected graph Γ, exactly one of the following conditions
holds
(1) The symmetric (·, ·)form is positive definite, then Γ is called Dynkin graph.
(2) The symmetric form (·, ·)is positive semidefinite, there exist δ ∈ Zn>0 such that
(δ, x) = 0 for any x ∈ Zn . The kernel of (·, ·) is Zδ. In this case Γ is called
affine or Euclidean.
(3) There is x ∈ Zn≥0 such that (x, x) < 0. Then Γ is called of indefinite type.
A Dynkin graphs is one of An , Dn , E6, E7 , E8 . An affine graphs is one of Ân , D
bn, E
b6, E
b7 , E
b8 .
Every affine graph is obtained from a Dynkin graph by adding one vertex.
Proof. First, we check that An , Dn , E6 , E7, E8 define a positive definite form using the
Sylvester criterion and the fact that every subgraph of a Dynkin graph is Dynkin.
One can calculate the determinant of the Cartan matrix inductively. It is n + 1 for
An , 4 for Dn , 3 for E6 , 2 for E7 and 1 for E8 . In the same way one can check that
the Cartan matrices of affine graphs have determinant 0 and corank 1. The rows are
linearly dependent with positive coefficients. Any other graph Γ has an affine graph
Γ′ as a subgraph, hence either (δ, δ) < 0 or (2δ + αi , 2δ + αi ) < 0, if αi is the basis
vector corresponding to a vertex i which does not belong to Γ′ but is connected to
some vertex of Γ′ .
(α,α)
A vector α ∈ Zn is called a root if q (α) = 2
≤ 1. It is clear that α1 , . . . , αn are
roots. They are called simple roots.
Lemma 6.3. Let Γ be Dynkin or affine. If α is a root and α = m1 α1 + · · · + mn αn ,
then either all mi ≥ 0 or all mi ≤ 0.
P P
Proof. Let α = β − γ, where β = i∈I miαi , γ = j ∈I / mj αj for some mi , mj ≥ 0,
then q (α) = q (β) + q (γ) − (β, γ). Since Γ is Dynkin or affine, then q (β) ≥ 0,
q (γ) ≥ 0. On the other hand (β, γ) ≤ 0. Since q (α) ≤ 1, only one of three terms
q (β), q (γ), − (β, γ) can be positive, which is possible only if β or γ is zero.
A root α is positive if α = m1α1 + · · · + mn αn , mi ≥ 0 for all i.
A quiver has finite type if there are finitely many isomorphism classes of indecom-
posable representations.
Theorem 6.4. (Gabriel) A connected quiver Q has finite type iff the corresponding
graph is Dynkin. For a Dynkin quiver there exists a bijection between positive roots
and isomorphism classes of indecomposable representations.
Proof. If Q is of finite type, then Rep (x) has finitely many orbits for each x ∈ Zn≥0 .
If Q is not Dynkin, then there exists x ∈ Zn≥0 such that q (x) ≤ 0. If Q has finite
type, then Rep (x) must have an open orbit OX . By Corollary 5.2
(6.1) codim OX = dim EndQ (X) − q (x) > 0.
Contradiction.
Now suppose that Q is Dynkin. Every indecomposable representation X is a brick
with trivial self-extensions by Corollary 4.3. Hence q (x) = 1, i.e. x is a root. By (6.1)
OX is the unique open orbit in Rep (x). What remains is to show that for each root x
there exists an indecomposable representation of dimension x. Indeed, let X be such
that dim OX in Rep (x) is maximal. We claim that X is indecomposable. Indeed,
let X = X1 ⊕ · · · ⊕ Xs be a sum of indecomposable bricks. Then by Lemma 5.3
Ext1 (Xi , Xj ) = 0. Therefore q (x) = s = 1. Hence X is indecomposable.
PROBLEM SET # 10
MATH 252
Due November 14.

1. Let Q be a connected quiver and k (Q) be the path algebra of Q. Show that
the center of k (Q) is isomorphic either to k, or to k [x], and that the latter happens
only in the case when Q is an oriented cycle.
2. Classify indecomposable representations of the quiver:
• → • ← •.
3. Classify indecomposable representations of the quiver:
• → • ← •
↑
•

1
WEEK 12
1. Reflection functors
Let Q be a quiver. We say a vertex i ∈ Q0 is +-admissible if all arrows containing i
have i as a target. If all arrows containing i have i as a source, we call i−-admissible.
By σi (Q) we denote the quiver obtained from Q by inverting all arrows containing i.
Let i be a +-admissible vertex and Q′ = σi (Q). Let us introduce the functor Fi+ :
RepQ → RepQ′ . Let X be a representation of Q. Define X ′ = Fi+ X as follows. If
j 6= i, then Xj′ = Xj . Put Xi′ = Ker h, where
X M
h= ργ : Xj → Xi ,
γ=(j→i)∈Q1
For each γ = (i → j)L ∈ Q define ρ′γ : Xi′ → Xj = Xj′ as the natural projection on
′
the component Xj ∈ Xj .
If i is a −-admissible vertex and Q′ = σi (Q) one can define the functor Fi− :
RepQ → RepQ′ as follows. Let X ′ = Fi− (X), where Xj′ = Xj for i 6= j, and
Xi′ = Coker eh, where X
h=
e ργ : Xi → ⊕Xj ,
γ=(i→j)∈Q1
and for each γ = (j → i) ∈ Q′ define ρ′γ : Xj = Xj′ → Xi by restriction of the

′
projection ⊕Xj → Coker e h to Xj .

Example. Let Q be the quiver 1 → 2, and X is the representation k → 0, then
F1− (X) = 0 and F2+ (X) is k ← k.
It is easy to check that Fi+ is left-exact (maps an injection to an injection) and Fi−
is right exact (maps a surjection to a surjection). Let Li denote the representation
of Q which has k in the vertex i and zero in all other vertices. Then Fi+ (Li ) = 0 and
Fi− (Li ) = 0.
Theorem 1.1. Let X be an indecomposable representation of Q and i be a +-
= Li . Otherwise X ′ = Fi+ (X) is inde-
admissible vertex. Then Fi+ (X) = 0 iff X ∼
composable,
′
X
(1.1) dim Xi = − dim Xi + dim Xj
(j→i)
and Fi− Fi+ (X) = X.

∼

1
If i is −-admissible vertex and X is indecomposable, then Fi− (X) = 0 iff X ∼

= Li .
Otherwise Fi− (X) is indecomposable, the dimension of Fi− (X) can be calculated by
the same formula (1.1) and Fi+ Fi− (X) ∼= X.
Proof. Note that if X ∼
6 Li , then h must be surjective because of indecomposability of
=
X, hence the formula (1.1) holds. Furthermore, we have the following exact sequence
′ e
h h
M
(1.2) 0 → Xi − → Xj −
→ Xi → 0
(j→i)∈Q1
and Fi− X ′ i = Coker e

h∼= Xi . Observe also that eh is injective by definition, hence
′
X is indecomposable. Thus, we proved the first part of the theorem.
For the case of −-admissible vertex, define D: Rep Q → Rep Qop, where Qop is
the quiver with all arrows of Q reversed, D (Xj ) = Xj∗ , D (ργ ) = ρ∗γ , and note that
D ◦ Fi+ = Fi− ◦ D. Since D reverses all maps and change Ker to Coker, the second
statement of the theorem follows immediately.
Note that for an arbitrary X the sequence (1.2) is not exact but e h is injective and
− +
h◦e h = 0. Therefore one can define a natural injection φ : Fi Fi X → X, where
φj = id for all j 6= i and φi coincides with h restricted to Coker e
h. In the similar way
+ −
one can define the natural surjection ψ : X → Fi Fi X if i is a −-admissible vertex.
Finally let Q′ = σi (Q), X be a representation of Q′ and Y be a representation of
Q, X ′ = Fi− X and Y ′ = Fi+ Y . Let η ∈ HomQ (X, Y ′ ), define χ ∈ HomQ′ (X ′ , Y ) by
putting χj = Id for j 6= i and obtaining χi from following commutative diagram
e
h h ′
Xi −
→ ⊕Xj −→ Xi → 0
↓η i ↓⊕ηj ↓χi
e
h h
0 → Yi′ −
→ ⊕Yj′ −
→ Yi
Note χi is uniquely determined by η. In the same way for each χ ∈ HomQ (X ′ , Y )

one can define η ∈ HomQ′ (X, Y ′ ). A routine check now proves the following
Lemma 1.2. Let Q′ = σi (Q), X be a representation of Q′ and Y be a representation
of Q, then
HomQ′ Fi− X, Y ∼ = HomQ X, Fi+ Y .

2. Reflection functors and change of orientation.

Lemma 2.1. Let Γ be a connected graph without cycles, Q and Q′ be two quivers
on the same graph. Then there exists an enumeration of vertices such that Q′ =
σk ◦ · · · ◦ σ1 (Q) and i is a +-admissible vertex for σi−1 ◦ · · · ◦ σ1 (Q).
Proof. It is sufficient to prove the statement for two quivers Q and Q′ different at
one arrow. So let γ ∈ Q1 . After removing γ, Q splits in two connected components;
let Q′′ be the component which contains t (γ). Enumerate vertices of Q′′ in such a
way that if i → j ∈ Q′′1 , then i > j. This is possible since Q′′ does not have cycles.
Check that Q′ = σk ◦ · · · ◦ σ1 (Q) (here k is the number of all vertices in Q′′) and i is
a +-admissible vertex for σi−1 ◦ · · · ◦ σ1 (Q).
Theorem 2.2. Let i be a +-admissible vertex for Q and Q′ = σi (Q). Then Fi+
and Fi− establish a bijection between indecomposable representations of Q (non-
isomorphic to Li ) and indecomposable representations of Q′ (non-isomorphic to Li .)1
Theorem 2.2 follows from 1.1. Together with Lemma 2.1 it allows to change an
orientation on a quiver if the quiver does not have cycles.
3. Weyl group and reflection functors.

Given any graph Γ, one can associate with it a certain linear group, which is called
a Weyl group of Γ. We denote by α1 , . . . , αn vectors in the standard basis of ZΓ0 = Zn ,
αi corresponds to the vertex i. These vectors are called simple roots. For each simple
root αi put
2 (x, αi)
ri (x) = x − αi .
(αi , αi )
One can check that ri preserves the scalar product and ri2 = id. The linear transfor-
mation ri is called a simple reflection. If Γ has no loops, ri also preserves the lattice
generated by simple roots. Hence ri maps roots to roots. If Γ is Dynkin, the scalar
product is positive-definite, and ri is a reflection in the hyperplane orthogonal to αi .
The Weyl group W is a group generated by r1 , . . . , rn . For a Dynkin diagram W is
finite (since the number of roots is finite).
Example. Let Γ = An . Let ε1 , . . . , εn+1 be an orthonormal basis in Rn+1 . Then
one can take the roots of Γ to be εi −εj , simple roots to be ε1 −ε2, ε2 −ε3, . . . , εn −εn+1 ,
ri (εj ) = 0 if j 6= i, i + 1, and ri (εi) = εi+1 . Therefore W is isomorphic to the
permutation group Sn+1 .
One can check by direct calculation, that (1.1) implies
Lemma 3.1. If X is an indecomposable representation of Q and dim X = x 6= αi ,
then dim Fi± X = ri (x).
An element c = r1 . . . rn ∈ W is called a Coxeter transformation. It depends on
the enumeration of simple roots.
Example. In the case Γ = An a Coxeter element is always a cycle of length n + 1.
Lemma 3.2. If c (x) = x, then (x, αi) = 0 for all i. In particular for a Dynkin graph
c (x) = x implies x = 0.
1We will denote by the same letter Li the representations of quivers with different orientation.
Proof. By definition,
2 (αi , x + a1α1 + . . . ai−1 αi−1 )
c (x) = x + a1 α1 + · · · + an αn , ai = − .
(αi , αi )
The condition c (x) = x implies all ai = 0. Hence (x, αi ) = 0 for all i.
4. Coxeter functor.
Let Q be a graph without oriented cycles. We call an enumeration of vertices
admissible if i > j for any arrow i → j. Such an enumeration always exists. One can
easily see that every vertex i is a +-admissible for σi−1 ◦ · · · ◦ σ1 (Q) and −-admissible
for σi+1 ◦ · · · ◦ σn (Q). Furthermore,
Q = σn ◦ σn−1 ◦ · · · ◦ σ1 (Q) = σ1 ◦ · · · ◦ σn (Q) .
Define Coxeter functors
Φ+ = Fn+ ◦ · · · ◦ F2+ ◦ F1+ , Φ− = F1− ◦ F2− ◦ · · · ◦ Fn− .
Lemma 4.1. (1) HomQ (Φ− X, Y ) ∼ = HomQ (X, Φ+ Y );
(2) If X is indecomposable and Φ X 6= 0, then Φ− Φ+ X ∼
+
= X;
(3) If X is indecomposable of dimension x and Φ X 6= 0, then dim Φ+ X = c (x);
+
(4) If Q is Dynkin, then for any indecomposable X there exists k such that
k
(Φ+ ) X = 0.
Proof. (1) follows from Lemma 1.2, (2) follows from Theorem 1.1, (3) follows from
Lemma 3.1. Let us prove (4). Since W is finite, c has finite order h. It is sufficient
to show that for any x there exists k such that ck (x) is not positive. Assume that
this is not true. Then y = x + c (x) + · · · + ch−1 (x) > 0 is c invariant. Contradiction
with Lemma 3.2.

Lemma 4.2. Φ± does not depend on a choice of admissible enumeration.
Proof. Note that if i and j are disjoint and both + (−)-admissible, then Fi+ ◦ Fj+ =
Fj+ ◦ Fi+ Fi− ◦ Fj− = Fj− ◦ Fi− . If a sequence i1, . . . , in gives another admissible

enumeration of vertices, and ik = 1, then 1 is disjoint with i1, . . . , ik−1 , hence
F1+ ◦ Fi+k−1 ◦ · · · ◦ Fi+1 = Fi+k−1 ◦ · · · ◦ Fi+1 ◦ F1+ .
Now proceed by induction. Similarly for Φ− .
In what follows we always assume that an enumeration of vertices is admissible.
Corollary 4.3. Let Q be a Dynkin quiver, X be an indecomposable representation of
dimension x, and k be the minimal number such that ck+1 (x) is not positive. There
exists a unique vertex i such that
k
= Φ− ◦ F1− ◦ · · · ◦ Fi−1
x = c−k r1 . . . ri−1 (αi ) , X ∼ −
(Li ) .
In particular, x is a positive root and for each positive root x, there is a unique (up
to an isomorphism) indecomposable representation of dimension x.
Proof. Follows from Theorem 1.1 and Lemma 3.1.

5. Further properties of Coxeter functors

Here we assume again that Q is a quiver without oriented cycles and the enumera-
tion of vertices is admissible. We discuss the properties of the bilinear form h, i. Since
we plan to change an orientation of Q we use a subindex h, iQ , where it is needed to
avoid ambiguity.
Lemma 5.1. Let i be a +-admissible vertex, Q′ = σi (Q), and h, iQ , h, iQ′ the corre-
sponding bilinear forms. Then
hri (x) , yiQ′ = hx, ri (y)iQ .
Proof. It suffices to check the formula for a subquiver containing i and all its neigh-
bors. Let x′ = ri (x) and y ′ = ri (y). Then
X X
x′i = −xi + xj , yi′ = −yi + yj ,
i6=j i6=j
X X X
hx′, yiQ′ = x′i yi − x′i yj + xj yj = −x′i yi′ + xj yj ,
i6=j i6=j i6=j
X X
hx, y ′iQ = xi yi′ − yi′ xj = −x′iyi′ + xj yj .
i6=j i6=j

Corollary 5.2. For a Coxeter element c we have
c−1 (x) , y = hx, c (y)i .
If Φ+ (Y ) 6= 0, Φ− (X) 6= 0, then
dim Ext1 X, Φ+ (Y ) = dim Ext1 Φ− (X) , Y .

Proof. First statement follows directly from Lemma 5.1. The second statement fol-
lows from the first statement, Lemma 1.2 and the identity
hx, yiQ = dim HomQ (X, Y ) − dim Ext1 (X, Y ) .

Let A = k (Q) be the path algebra. Recall that any indecomposable projective
module is isomorphic to Aei.
Lemma 5.3. Fi+ ◦ · · · ◦ F1+ (Aei) = 0, Fi−1
+
◦ · · · ◦ F1+ (Aei) ∼
= Li .
Proof. One can check by direct calculation that for each component ej Aei, ej Aei = 0
for j > i, and
Fk+ ◦ · · · ◦ F1+ (ej Aei) = ej Aei for k < j, Fj+ ◦ · · · ◦ F1+ (ej Aei ) = 0.

Corollary 5.4. Φ+ (P ) = 0 for any projective module P . For any indecomposable
projective Aei we have
(5.1) Aei = F1− ◦ · · · ◦ Fi−1
−
(Li ) .
Proof. The first statement follows from Lemma 5.3 immediately. For the second use
Theorem 1.1 and Lemma 5.3.

An injective module is a module I such that for any injective homomorphism
i : X → Y and any homomorphism ϕ : X → I, there exists a homomorphism
ψ : Y → I such that ϕ = ψ ◦ i. A module I is injective iff Ext1 (X, I) = 0 for any
X. One can see analogy with projective modules, however in general there is no nice
description of injective (like a summand of a free module).
Exercise. Check that Q is an injective Z-module.
In case when A is a finite-dimensional algebra, injective modules are easy to de-
scribe. Indeed, the functor D : A − mod → mod − A such that D (X) = X ∗
maps left projective modules to right injective and vice versa. Therefore any inde-
composable injective module is isomorphic to (ej A)∗ . Since D ◦ Φ+ = Φ− ◦ D, one
can see easily that Φ− (I) = 0 for any injective module I. Moreover, one can prove
similarly to the projective case that
∗
(ej A) ∼= F + ◦ · · · ◦ F + (Lj ) .
n j+1
∗
Let P (j) = Aej and I (j) = (ej A) and p (j) = dim P (j), i (j) = dim I (j). Then
(5.2) c (p (j)) = rn . . . r1 (p (j)) = rn . . . rj+1 (−αj ) = −i (j) .
Note that Ext1 (Aej , X) = 0 for any X and dim HomQ (Aej , X) = xj . Hence
(5.3) hp (j) , xi = xj .
On the other hand, Ext1 (X, (ej A)∗ ) = 0 and
HomQ (X, (ej A)∗) ∼ = HomQ (ej A, X ∗ ) ,
∗
which implies dim HomQ (X, (ej A) ) = xj . Thus, we obtain
(5.4) hx, i (j)i = xj .
Combine together (5.2), (5.3), (5.4) and get
hp (j) , xi + hx, c (p (j))i = 0.
Since p (1) , . . . , p (n) form a basis, the last equation implies that for arbitrary x and
y
(5.5) hy, xi + hx, c (y)i = 0.
6. Affine root system

Let Γ be an affine Dynkin graph. Then the kernel of bilinear symmetric form in
n
Z is one-dimensional and generated by
δ = a0α0 + a1 δ1 + · · · + an δn .
We assume without loss of generality that the vertex α0 is such that a0 = 1. By
removing 0 from Γ we get a Dynkin graph which we denote by Γ0 . In affine case
roots can be of two kinds: real, if q (α) = 1, or imaginary, q (α) = 0.
Lemma 6.1. Imaginary roots are all proportional to δ, real roots can be written as
α + mδ for some root δ of Γ0 . Every real root can be obtained from a simple root by
the action of the Weyl group W .
Proof. The first statement is obvious, the second follows from the fact that q (α) =
q (α + mδ), hence the projection on the hyperplane generated by α1 , . . . αn maps a
root to a root. To prove the last statement, note that ri maps every positive root
different from αi to a positive root. Let α be a positive real root, α = a0α0 +· · ·+an αn ,
and h (α) = a0 + a1 + · · · + an . Then (α, αi ) > 0 at least for one i. But then
h (ri (α)) < h (α). Thus, one can decrease h (α) by application of simple reflection.
In the end one can get a root of height 1, which is a simple root. Similarly for negative
roots.
7. Kronecker quiver
In this section we use Coxeter functors to classify indecomposable representation of
the quiver Â1 = • ⇒ •. The admissible enumeration of vertices is 1 ⇒ 0, δ = α0 + α1.
Positive real roots are
mα1 + (m + 1) α0 = −α1 + (m + 1) δ, (m + 1) α1 + mα0 = α1 + mδ, m ≥ 0.
The Coxeter element c = r1 r0 satisfies
c (α1 ) = α1 + 2δ, c (δ) = δ.
Let x = mα1 +lδ. If m > 0 then c−s (x) is not positive for sufficiently large s. Hence if
s s
X is indecomposable of dimension x, then (Φ− ) X = 0. If m < 0, then (Φ+ ) X = 0.
Thus if m 6= 0, then as in the case of Dynkin quiver, X can be obtained from some
Li by application of reflection functor. In particular, we obtain that the dimension
of an indecomposable representation is always a root and if this root is real, then
the indecomposable with this dimension is unique up to an isomorphism. Indeed, we
have either
k m ⇒AB k
m+1
,
where A = (1m , 0) , B = (0, 1m ), or

k m+1 ⇒C m
D k ,
where C = At , D = B t.
Classification of indecomposables of dimension mδ is equivalent to classification of
pairs of linear operators (A, B) : k m → k m up to equivalence (A, B) ∼ (P AQ−1 , P BQ−1).
Assume that A is invertible, then one may assume that A = Id, and then classify B up
to conjugation. Indecomposability of the representation implies that B is equivalent
to the Jordan block with some eigenvalue µ. Denote the corresponding representa-
tion by ρµ . If B is invertible, then A is equivalent to a Jordan block. Denote such
representation by σµ . One can see that ρµ is isomorphic to σµ−1 if µ 6= 0. Now let us
prove that at least one of A and B is invertible. Indeed, indecomposability implies
that Ker A ∩ Ker B = 0. Hence A + tB is invertible for some t. Without loss of gen-
erality one can assume that A + tB = id. But then either A or B must be invertible.
Thus, we proved that indecomposable representation of dimension (m, m) = δ are
parameterized by a projective line.
For other affine quivers, the situation is more complicated, as there are real roots
which remain positive under Coxeter transformation. For example consider the quiver
Db4
5
↓
1 → 2 → 4
↑
3
Then c (α1 + α2 + α3 ) = α4 + α2 + α5 , c2 (α1 + α2 + α3) = 0.
PROBLEM SET # 11
MATH 252
Due November 21.

1. Let Rep (a, b, c) be the space of all representations of the quiver
•→•←•
with dimension vector (a, b, c). List all orbits in Rep (a, b, c). Show that there is only
one open orbit. Describe the open orbit OX in terms of decomposition of X into
direct sum of indecomposable representations.
2. Classify indecomposable representations of the quiver An with orientation:
• → • → · · · → •.
You can use Gabriel’s theorem.

1
WEEK 13.
1. Preprojective and preinjective representations.

The goal of these notes is to obtain a classification of indecomposable representa-
tions of affine quivers. It is rather technical and long.
s
A representation X is called preprojective if (Φ+ ) X = 0 for some s, preinjective
s
if (Φ− ) (X) = 0 for some s and regular if it is not preprojective or preinjective.
Example 1.1. For Kronecker quiver all preinjective indecomposable representations
have dimension α1 + mδ, all preprojective representations have dimension −α1 + mδ,
and regular indecomposable representations have dimension mδ.
s
Lemma 1.2. If X is preprojective, then X = (Φ− ) P for some projective P . If X
s
is preinjective, then X = (Φ+ ) I for some injective I.
s s+1
Proof. Suppose (Φ+ ) X 6= 0, and (Φ+ ) X = 0. Then
+
s s
Fi−1 −
= Φ− F1− . . . Fi−1
. . . F1+ Φ+ X = Li , X ∼ (Li )
as in Corollary 4.3 (week 12). Therefore by Corollary 5.4 (week12)
s
X∼= Φ− (Aei ) .
For preinjective similarly.
Corollary 1.3. If X is an indecomposable preprojective or preinjective, then dim X
is a real root. If X and Y are preprojective indecomposable representations of the
same dimension, then X ∼ = Y . An indecomposable preprojective representation is a
brick with trivial self-extensions.
We see from above corollary that preprojective and preinjective indecomposables
can be described precisely in terms of reflection functors in the same way as it was
done for Dynkin quivers. The next lemma allows “to separate” preinjective, prein-
jective and regular indecomposable representations.
Lemma 1.4. If X, Y are indecomposable and X is preprojective, Y is not, then
HomQ (Y, X) = Ext1 (X, Y ) = 0. If X is preinjective, Y is not, then HomQ (X, Y ) =
Ext1 (Y, X) = 0.

1
2 REPRESENTATION THEORY. WEEK 13.
s
Proof. Let X be preprojective. Then X = (Φ− ) P for some projective P . Then
s s
Ext1 Φ− P, Y = Ext1 P, Φ+ Y = 0

by Corollary 5.2 (week12). On the other hand,

s+1 s+1 + s+1
Φ+ = Φ−
X = 0, Y ∼ Φ Y
and
s+1 + s+1 s+1 s+1
HomQ X, Φ− Φ Y = HomQ Φ+ X, Φ+ Y = 0.
For preinjective use duality.
Let now Q be affine and define defect X by
def (X) = hδ, xi = − hx, δi .
We write x ≤ y if y − x ∈ Zn≥0
Lemma 1.5. If x < δ and X is indecomposable of dimension x, then X is a brick,
x is a root and Ext1 (X, X) = 0.
Proof. If X is not a brick, then there is a brick Y ⊂ X such that Ext1 (Y, Y ) =
6 0
(proven week 11). But then q (y) ≤ 0, which is impossible as y < δ. Hence X is a
brick. Since q (x) > 0, we have Ext1 (X, X) = 0 and q (x) = 1.
Lemma 1.6. There is an indecomposable representation of dimension δ.
Proof. Pick an orbit OZ in Rep (δ) of maximal dimension. Then Z is indecompos-
able, because otherwise Z = X1 ⊕ · · · ⊕ Xp , where Xi are as in previous lemma,
Ext1 (Xi , Xj ) = 0 and then q (z) = p.
Lemma 1.7. If X is regular, then there is s such that cs (x) = x.
Proof. One can find s such that cs (x) = x + lδ. But if l 6= 0, then cds (x) < 0 for
some d ∈ Z. This contradicts regularity of X.
Theorem 1.8. Let X be indecomposable.
(1) If X is preprojective, then def (X) < 0;
(2) If X is regular, then def (X) = 0;
(3) If X is preinjective, then def (X) > 0.
Proof. Let X be preprojective, Z as in Lemma 1.6. Then Ext1 (X, Z) = 0 by
s
Lemma 1.4. On the other hand, X = (Φ− ) Aei. Hence
s s
HomQ (X, Z) = HomQ Φ− Aei , Z = HomQ Aei , Φ+ Z 6= 0,

and hx, δi > 0. For preinjective X use duality.

Finally, let X be regular. Assume def (X) 6= 0, say def (X) > 0. Since x is regular
cs (x) = x for some s. Then y = x + c (x) + · · · + cs−1 (x) is c-invariant, therefore
x + c (x) + · · · + cs−1 (x) = mδ by Lemma 3.2 (week 12). But hδ, ci (x)i = hδ, xi > 0
for all i < s, hence hδ, mδi > 0. But hδ, δi = q (δ) = 0. Contradiction.
REPRESENTATION THEORY. WEEK 13. 3
2. Regular representations
In this section we describe indecomposable regular representations.
We say that a representation is regular if it is a direct sum of indecomposable
regular representations.
Theorem 2.1. If X, Y are regular and ϕ ∈ HomQ (X, Y ) then Im ϕ, Ker ϕ, Coker ϕ
are regular. If
0→X →Z →Y →0
is an exact sequence then Z is regular.
Proof. By Lemma 1.4 Im ϕ does not have preinjective summand and preprojective
summand. By the same reason Ker ϕ does not have preinjective summand and def
(Ker ϕ) = def (X)-def (Im ϕ) = 0. Hence Ker ϕ is regular. Similarly, Coker ϕ is
regular.
Finally, suppose Z has a preprojective direct summand Zi . This is impossible by
the long exact sequence
HomQ (X, Zi ) = 0 → HomQ (Z, Zi ) → HomQ (Y, Zi ) = 0.
Similarly, Z could not have preinjective direct summand.
A regular representation X is called regular simple if X has no proper non-trivial
regular subrepresentations. By Theorem 2.1 a regular simple representation is a brick,
hence q (x) ≤ 1 and x is a root.
Example 2.2. For Kronecker quiver a representation k ⇒µλ k is simple for any
(λ, µ) 6= (0, 0). One can see easily from classification of indecomposables that those
are all regular simple representations.
Example 2.3. For the quiver D
b 4 an indecomposable representation
k
↓τ 2
τ1 τ4
→ k2 −
k − → k
τ3
↑
k
is regular simple iff Im τi 6= Im τj for all i 6= j.
Let δ = a0α0 + · · · + an αn . Without loss of generality we may assume that a0 = 1.
Let P = Ae0, p = dim P and R be the indecomposable preprojective representation
of dimension r = p + δ. There are the following identities
hp, δi = hr, δi = 1, hp, ri = 2, hr, pi = 0.
1
Since Ext (P, R) = 0, HomQ (P, R) = k 2 .
Lemma 2.4. Let θ ∈ HomQ (P, R). If θ 6= 0, then θ is injective. Let η ∈ HomQ (R, P ).
If η 6= 0 then η is surjective.
Proof. Both Ker θ and Im θ are preprojective, because P and R are preprojective.
Since −1 = def (P ) = def (Ker θ) + def (Im θ), either Ker θ or Im θ is zero. The
second statement is similar.
Corollary 2.5. HomQ (R, P ) = Ext1 (R, P ) = 0.
Proof. Let η ∈ HomQ (R, P ) be surjective, then η = 0 since P is projective and must
split as a direct summand of R, but R is indecomposable. Hence HomQ (R, P ) = 0.
Since hr, pi = 0, Ext1 (R, P ) = 0.
Corollary 2.6. Let θ ∈ HomQ (P, R), θ 6= 0. Then Zθ = Coker θ is indecomposable
regular.
Proof. Use the sequence
0 → P → R → Zθ → 0.
The long exact sequence
0 = HomQ (R, P ) → HomQ (R, R) → HomQ (R, Zθ ) → Ext1 (R, P ) = 0
implies HomQ (R, Zθ ) = EndQ (R) = k. The long exact sequence
0 → HomQ (Zθ , Zθ ) → HomQ (R, Zθ ) = k → . . .
implies HomQ (Zθ , Zθ ) = k. Hence Zθ is indecomposable. Since def (Zθ ) = 0, Zθ is
regular.
Lemma 2.7. Let X be regular indecomposable and x0 6= 0, where x0 = dim X0 .
Then there exists θ ∈ HomQ (P, R) such that HomQ (Zθ , X) 6= 0.
Proof. First note that
dim HomQ (P, X) = x0 = dim (Q, X)
since hp, xi = hq, xi = x0.
Any θ ∈ HomQ (P, R) defines the linear map
θ∗ : HomQ (R, X) → HomQ (P, X) .
Since dim HomQ (P, R) = 2, one can find θ ∈ HomQ (P, R) such that θ∗ is not in-
vertible. Then there is ϕ ∈ HomQ (R, X) such that θ∗ (ϕ) = ϕ ◦ θ = 0. Then
ϕ (θ (P )) = 0, and ϕ is well defined homomorphism Zθ → X.
Corollary 2.8. Let X be regular simple, then x ≤ δ.
Proof. We already know that x is a root. If x0 6= 0, then HomQ (Zθ , X) 6= 0 for some
θ and therefore X is a quotient of Zθ , hence x ≤ δ. If x0 = 0, then x < δ.
Example 2.9. In case of D b 4 the regular simples have dimensions α1 + α2 + α3 ,
α1 + α2 + α4 , α1 + α2 + α3 , α3 + α2 + α4 , α3 + α2 + α5 , α4 + α2 + α5 or δ. There is
exactly one simple for each real root and one-parameter family for δ.
Our next step is to describe extensions between regular simple representations.
=Y
Lemma 2.10. Let X and Y be two regular simples, then HomQ (X, Y ) = k iff X ∼
1 + 1
and Ext (X, Y ) = k iff Y = Φ X. Otherwise Ext (X, Y ) = HomQ (X, Y ) = 0.
∼
Proof. The statement about Hom is trivial since any nonzero ϕ ∈ HomQ (X, Y ) is an
isomorphism . To prove the statement about Ext1 , use (5.5) from lecture notes week
6 Φ+ X, X, then
12. First, assume that Y ∼=
hx, yi = dim HomQ (X, Y ) − dim Ext1 (X, Y ) ≤ 0,
hy, c (x)i = dim HomQ Y, Φ+ X − dim Ext1 Y, Φ+ X ≤ 0.

Since hx, yi + hy, c (x)i = 0, Ext1 (X, Y ) = 0.

Now assume that X ∼ = Y . If x is a real root, then Ext1 (X, X) = 0, and hx, xi = 1.
Then hx, c (x)i = −1, which implies Ext1 (X, Φ+ X) = k. If x = δ, then
HomQ (X, X) = Ext1 (X, X) = k.

s
Corollary 2.11. If X is regular simple and x < δ, then (Φ+ ) X ∼ = X for some s. If
x = δ, then Φ+ X ∼= X.
s
The minimal number s such that (Φ+ ) X ∼ = X is called the period of X. Regular
simples can be divided in orbits under action of Φ+ .
In the category of regular representations one can define Jordan-Hölder series and
regular length, and regular series is again unique up to permutation.
The following theorem gives a complete description of indecomposable regular rep-
resentations.
Theorem 2.12. Let X be regular indecomposable then there is a unique filtration
(2.1) 0 ⊂ X1 ⊂ X2 ⊂ · · · ⊂ Xr = X
such that Yi ∼
= Xi /Xi−1 is regular simple and
Yi−1 ∼ = Ext1 (Yr , Yr−1 ) ∼
= Φ+ (Yi ) , Ext1 (Yr , Xr−1 ) ∼ = k.
Moreover, Ext1 (Z, Xr−1 ) = 0 for any regular simple Z not isomorphic to Yr .
Proof. We prove Theorem by induction on the regular length of X. Check yourself
case r = 2. If X has length r then it has a filtration (2.1), although it might be not
unique. Assume first that Xr−1 is indecomposable. Then it satisfies all the statements
of Theorem by induction assumption. Consider the exact sequence
0 → Xr−2 → Xr−1 → Yr−1 → 0
and induced long exact sequence
a b c d
HomQ (Yr , Xr−1 ) − → Ext1 (Yr , Xr−2 ) −
→ HomQ (Yr , Yr−1 ) − → Ext1 (Yr , Xr−1 ) −
→ Ext1 (Yr , Yr−1 ) → 0.
We claim that d is an isomorphism. If Ext1 (Yr , Xr−2 ) = 0, then it is trivial. Assume
that Ext1 (Yr , Xr−2 ) 6= 0. By induction assumption Yr ∼
= Φ+ Yr−2 ∼
= Yr−1 . Then a = 0
by uniqueness of filtration for Xr−1 , b must be an isomorphism, c forced to be zero,

and therefore d is an isomorphism.
Now we prove that Xr−1 is indecomposable. Assume the opposite. Then Xr−1 =
Z1 ⊕· · ·⊕Zs where each Zi is indecomposable and satisfies the statement of Theorem.
Assume that Z1 has maximal length among Zi . Then there is a surjective homomor-
phism pi : Z1 → Zi for each i, and this homomorphism induces the isomorphism
pi∗ : Ext1 (Yr , Z1 ) → Ext1 (Yr , Zi ) ∼
= k.
Consider the exact sequence
M
(2.2) 0→ Zi → X → Yr → 0.
It is induced by some ψ ∈ Ext1 (Yr , Zi ). But ψ = ψ1 + · · · + ψs , ψi ∈ Ext1 (Yr , Zi ).

L
Hence each ψi = ci pi∗ (ψ1). Let
n M o
′
Z = (z1, . . . , zs) ∈ Zi | zi = ci pi (z1 ) ,
then one can find X ′ ⊂ X such that

0 → Z ′ → X ′ → Yr → 0,
is a subsequence of (2.2). Then X ′ splits as a summand in X. Contradiction.
Check now that X has a unique regular maximal submodule Xr−1 (that implies
the uniqueness of filtration). Consider the exact sequence
f
0 → Xr−1 → X −
→ Yr → 0.
Let X ′ be another maximal submodule, then f (X ′ ) = Yr and we have an an exact
sequence
0 → Xr−1 ∩ X ′ → X ′ → Yr → 0.
However, the regular length of X ′ is r − 1, hence Xr−1 ∩ X ′ = Xr−2 . Therefore
X/Xr−2 ∼
= Yr ⊕ Yr−1 . But the sequence
0 → Yr−1 → X/Xr−2 → Yr → 0
= Ext1 (Yr , Yr−1 ).
does not split since it is induced by a non-zero element in Ext1 (Yr , Xr−1 ) ∼
Contradiction.
Corollary 2.13. Let Y1 , . . . , Yr be a sequnce of simple regular representations such
= Φ+ Yi . Then there exists a unique up to an isomorphism regular inde-
that Yi−1 ∼
composable X with filtration 0 ⊂ X1 ⊂ X2 ⊂ · · · ⊂ Xr = X such that Xi /Xi−1 ∼ = Yi .
Proof. Construct X inductively using the isomorphism
= Ext1 (Yt+1 , Yt ) ∼
Ext1 (Yt+1 , Xt ) ∼ = k.

As follows from Theorem 2.12 simple regular subquotients of an indecomposable

regular representation belong to one orbit of Φ+ . Thus, each orbit of Φ+ in the set
of simple regular representations defines a family of indecomposables called a tube.
Lemma 2.14. Let X be regular indecomposable, then dim X is a root.
Proof. Lemma follows from Theorem 2.12 and the following fact. Let α, β be real
roots. Then (α, β) = −1 implies α + β is a real root, (α, β) = −2 implies α + β is an
imaginary root.
Lemma 2.15. Every tube contains exactly one indecomposable representation iso-
morphic to Zθ .
s
Proof. Let X be simple regular of period s, i.e. (Φ+ ) X ∼
= X. If x = dim X, then
(2.3) x + c (x) + · · · + cs−1 (x) = mδ.
i
Choose y = ci (x) such that y0 6= 0. Let Y = (Φ+ ) X. By Lemma 2.7 there
exist θ ∈ HomQ (P, R) and a non-zero homomorphism ϕ : Zθ → Y . Then the
indecomposable Zθ has the filtration
0 ⊂ Z1 ⊂ Z2 ⊂ · · · ⊂ Zs = Zθ
such that Zs /Zs−1 ∼
= Y . Then Zθ is in a tube. Moreover, one can see now that m = 1
in (2.3) and therefore Zθ is unique.
3. Indecomposable representations of affine quivers
In the next theorem we summarize our results about affine quivers.
Theorem 3.1. Let Q be an affine quiver, then dimension of every indecomposable
representation of Q is a root. If α is a real root, then there exists exactly one (up
to an isomorphism) indecomposable representation of dimension α. If α = mδ, then
there are infinitely many indecomposable representations of dimension α.
Proof. Let α be the dimension of an indecomposable representation X. If hα, δi = 6 0,
then X is preprojective or preinjective, and α is a real root by Corollary 1.3. If
hα, δi = 0, then X is regular and α is a root by Lemma 2.14. The uniqueness of
X follows from Theorem 2.12. We also have to prove that for each α there is an
indecomposable of dimension α. If hα, δi > 0, choose the minimal i and s such that
s −
ri . . . r1 cs (α) < 0, then put X = (Φ− ) ◦ F1− ◦ · · · ◦ Fi−1 (Li ). The case hα, δi < 0 is
similar. Let hα, δi = 0. Assume first that α < δ. Construct X as an orbit of maximal
dimension in Rep (α). If α = β + mδ, for some β < δ, construct an indecomposable
Y of dimension β, and extend it using the description of a tube.
Example 3.2. Consider the quiver Ân . The indecomposable representations of real
dimension and regular indecomposables of imaginary dimension with period greater
than 1 are enumerated by counterclockwise walks around the quiver ( ignoring the
orientation). A basis {vi } in representation X is enumerated by vertices which appear
in a walk. For each γ put ργ (vi ) = vi+1 if the orientation of γ is counterclockwise

and ργ (vi+1 ) = vi if the orientation of γ is clockwise. The last vector in the walk
goes to 0 if γ is counterclockwise oriented.
If X has imaginary dimension and X is in a tube of period 1, then X can be
described by the following construction. Put Xi ∼ = k m for all i, ργ = Id for all γ
except one arrow σ ( it does not matter which one you choose). Let ρσ be a Jordan
block with non-zero eigenvalue.
Let Q and Q′ two different quivers in the graph Ân . It is not always possible to
obtain Q′ from Q using reflection functors. If Q and Q′ have the same number of
clockwise (hence counterclockwise) arrows, one can obtain Q′ from Q by a chain of
reflections.
Check yourself that if Q has p counterclockwise arrows, then Q has one tube of
period p − 1 and one tube of period n − p − 1.
WEEK 14
1. Applications of quivers
Two rings A and B are Morita equivalent if the categories of A− modules and
B-modules are equivalent. A projective finitely generated A-module P is a projective
generator if any other projective finitely generated A-module is isomorphic to a direct
summand of P ⊕n for some n.
Theorem 1.1. A and B are Morita equivalent iff there exists a projective generator
P in A− mod such that B ∼ = EndA (P ). The functor X 7→ HomA (P, X) establishes
the equivalence between A − mod and B − mod .
For the proof see, for example, Bass “Algebraic K-theory”.
Assume now that C is a finite-dimensional algebra over algebraically closed field k.
Let P1 , . . . , Pn be a set of representatives of isomorphism classes of indecomposable
projective C-modules. Then P = P1 ⊕ · · · ⊕ Pn is a projective generator, and A =
EndC (P ) is Morita equivalent to C.
Example 1.2. Let C be semisimple, then C ∼ = Matm1 (k) × · · · × Matmn (k), and
A∼= k n . Let
C = XY

0Z ∈ Matp+q (k) | X ∈ Matp (k) , Y ∈ Matp,q (k) , Z ∈ Matq (k) .
Then
A = {(xy
0z ) | x, y, z ∈ k} .
Let R be the radical of C. Then each indecomposable projective Pi has the filtration
Pi ⊃ RPi ⊃ R2 Pi ⊃ · · · ⊃ 0 such that Rj Pi /Rj+1 Pi is semisimple for all j. Recall that
Pi /RPi is simple (lecture notes 9), hence HomC (Pi , Pj /RPj ) = 0 if i 6= j. Define the
quiver Q in the following way. Vertices are enumerated by indecomposable projective
modules P1 , . . . , Pn , the number of arrows i → j equals dim HomC (Pi , RPj /R2 Pj ).
We construct a surjective homomorphism φ : k (Q) → A. (This construction is not
canonical). Fist set φ (ei) = IdPi . Let γ1 , . . . , γs be the set of arrows from i to
j, choose a basis η1 , . . . , ηs ∈ HomC (Pi , RPj /R2 Pj ), each ηl can be lifted to ξl ∈
HomC (Pi , RPj ) as Pi is projective. Define φ (γl ) = ξl . Now φ extends in the unique
way to the whole k (Q) since k (Q) is generated by idempotents ei and arrows.
Since φ is surjective, then A ∼ = k (Q) /I for some two-sided ideal I ⊂ k (Q). The
pair Q and an ideal I in k (Q) is called a quiver with relations. The problem of clas-
sification of indecomposable C-modules is equivalent to the problem of classification
1
2 REPRESENTATION THEORY. WEEK 14
of indecomposable representations of Q satisfying relations I. In some cases such

quiver approach is very useful.
Example 1.3. Let k be the algebraic closure of F3 and C = k [S3 ]. In lecture
notes 9 we showed that C has two indecomposable projectives P+ = IndSS32 triv and
P− = IndSS32 sgn. The quiver Q is
α
• ⇐⇒ β•
with relations αβα = 0, βαβ = 0. The quiver itself is Â2 , indecomposable representa-
tions have dimensions (m, m), (m + 1, m) and (m, m + 1). Since we have the precise
description, it is not difficult to see that only six indecomposable representations
satisfy the relations. They are
k ⇐⇒ 0; 0 ⇐⇒ k; k ⇐⇒ k, α = 1, β = 0 or α = 0, β = 1,
k 2 ⇐⇒ k, α = (10) , β = 01 ; k ⇐⇒ k 2 , α = 01 , β = (10) .

The first two representations correspond to irreducible representations triv and sgn,
the last two are projectives. Two representations of dimension (1,1) correspond to
the quotients of P+ and P− by the minimal submodules.
In fact one can apply the quiver approach to any category C which satisfies the
following conditions
(1) All objects have finite length;
(2) Any object has a projective resolution;
(3) For any two objects X, Y , Hom (X, Y ) is a vector space over an algebraically
closed field k.
We do not need the assumption that the number of simple or projective objects is
finite. We illustrate this in the following example.
Example 1.4. Let Λ be the Grassmann algebra with two generators, i.e. Λ = k <
x, y > / (x2, y 2, xy + yx). Consider the Z-grading Λ = Λ0 ⊕ Λ1 ⊕ Λ2, where Λ0 = k,
Λ1 is the span of x and y, Λ2 = kxy. Let C denote
L the category of graded Λ-modules.
In other words, objects are Λ-modules M = i∈Z Mi , such that Λi Mj ⊂ Mi+j and
morphisms preserve the grading. All projective modules are free. An indecomposable
projective module Pi is isomorphic to Λ with shifted grading deg (1) = i. Thus, the
quiver Q has infinitely many vertices enumerated by Z:
α α α
· · · ⇐αβii • ⇐βi+1
i+1 i+2
• ⇐βi+2 i+3
• ⇐βi+3 ...
Here αi+1 , βi+1 ∈ Hom (Pi+1 , Pi ), αi+1 (1) = x, βi+1 (1) = y. Relations are αi αi+1 =
βi βi+1 = 0, αi βi+1 + βi αi+1 = 0.
Let us classify the indecomposable representations of above quiver. Assume first
that, that there exists v ∈ Xi+1 such that αi βi+1 6= 0, Then the subrepresentation
V spanned by v, αi+1 v, βi+1 v, αi βi+1 v splits as a direct summand in X. If X is
indecomposable, then X = V . The corresponding object in C is Pi+1 .
Now assume that αi βi+1 Xi+1 = 0 for any i ∈ Z. That is equivalent to putting the
new relations for Q: every path of length 2 is zero. Consider the subspaces
Wi = Im αi+1 + Im βi+1 ⊂ Xi , Zi+1 = Ker αi+1 ∩ Ker βi+1 ⊂ Xi+1 .
One can find Ui ⊂ Xi and Yi+1 ⊂ Xi+1 such that Xi = Ui ⊕ Wi , Xi+1 = Zi+1 ⊕ Yi+1 .
Check that Wi ⊕ Yi+1 is a subrepresentation, which splits as a direct summand in
X. If X is indecomposable and Wi 6= 0, then X = Wi ⊕ Yi+1 . Thus, we reduced
our problem to Kronecker quiver • ⇐ •! There is the obvious bijection between
indecomposable non-projective objects from C and the pairs (Y, i), where Y is an
indecomposable representation of Kronecker quiver, i ∈ Z (defines the grading).
Remark 1.5. The last example is related to the algebraic geometry as the derived
category of C is equivalent to the derived category of coherent sheaves on P1 .
Remark 1.6. If in the last example we increase the number of generators in Λ, then
the problem becomes wild (definition below).
Let C be a finite-dimensional algebra. We say that C is finitely represented if C has
finitely many indecomposable representations. We call C tame if for each d ⊂ Z>0 ,
there exist a finite set M1 , . . . , Mr of C − k [x] bimodules (free of rank d over k [x])
such that every indecomposable representation of C of dimension d is isomorphic to
Mi ⊗k[x] k [x] / (x − λ) for some i ≤ r, λ ∈ k. Finally, C is wild if there exists a
C − k < x, y > bimodule M such that the functor X 7→ M ⊗k<x,y> X preserves
indecomposability and is faithful. We formulate here without proof the following
results.
Theorem 1.7. Every finite-dimensional algebra over algebraically closed field k is
either finitely represented or tame or wild
Theorem 1.8. Let Q be a connected quiver without oriented cycles. Then k (Q) is
finitely represented iff Q is Dynkin, k (Q) is tame iff Q is affine.
Theorem 1.9. Let Algn be the algebraic variety of all n-dimensional algebras over
k. Then the set of finitely represented algebras is Zariski open in Algn .
2. Frobenius algebras
Let A be a finite-dimensional algebra over k. Recall that we denote by D the
functor mod − A → A − mod , such that D (X) = X ∗ . Recall also that D maps
projective modules to injective and vice versa.
A finite-dimensional A algebra over k is called a Frobenius algebra if D (AA ) is
isomorphic to A, where AA is the right A-module over itself.
Theorem 2.1. The following conditions on A are equivalent
(1) A is a Frobenius algebra;
(2) There exists a non-degenerate bilinear form h·, ·i on A such that hab, ci =
ha, bci;
(3) There exists λ ∈ A∗ such that Ker λ does not contain non-trivial left or right
ideals.
Proof. A form h·, ·i gives an isomorphism µ : A → A∗ by the formula x → h·, xi. The
condition hab, ci = ha, bci is equivalent to µ being a homomorphism of modules. A
linear functional λ can be constructed by λ (x) = h1, xi. Conversely, given λ, one can
define hx, yi = λ (xy). The condition Ker λ does not contains non-trivial one-sided
ideals is equivalent to the condition that the left and right kernels of h·, ·i are zero.
Lemma 2.2. Let A be a Frobenius algebra. An A-module X is projective iff it is
injective.
Proof. A projective module X is a direct summand of a free module, but a free
module is injective as D (AA ) is isomorphic to A. Hence, X is injective. By duality
an injective module is projective.
Example 2.3. A group algebra k (G) is Frobenius. Take
!
X
λ ag g = a1 .
g∈G
The corresponding bilinear form is symmetric.

A Grassmann algebra Λ = k < x1 , . . . , xn > / (xixj + xj xi ) is Frobenius. Put
!
X
λ ci1 ...ik xi1 . . . xik = c12...n .
i1 <···<ik
In a sense Frobenius algebras generalize group algebras. For example, if T ∈

Homk (X, Y ) for two k (G)-modules X and Y then
X
T̄ = gT g −1 ∈ HomG (X, Y ) .
g∈G
This idea of taking average over the group is very important in representation theory.
It has an analog for Frobenius algebras.
Choose a basis e1, . . . , en in a Frobenius algebra A. Let f1, . . . , fn be the dual basis,
i.e.
(2.1) hfi , ej i = δij .
Every a ∈ A can be written
X X
(2.2) a= hfi , ai ei = ha, eii fi .
and
X X X X
(2.3) aei ⊗ fi = hfj , aeii ej ⊗ fi = hfj a, eii ej ⊗ fi =
ej ⊗ fj a.
P
Lemma 2.4. Let X and Y be A-modules, T ∈ Homk (X, Y ). Then T̄ = ei T fi ∈
HomA (X, Y ).
Proof. Direct calculation using (2.2) and (2.3).

Example 2.5. If A = k (G), the dual bases can be chosen as {g}g∈G and {g −1 }g∈G .
Hence T̄ = gT g −1 .
P
In Frobenius algebra one can use the following criterion of projectivity.

Theorem 2.6. An A-module X is injective (hence projective) if there exists T ∈
Endk (X) such that T̄ = Id.
Proof. First, assume the existence of T . We have to show that X is injective, in
other words, for any embedding ε : X → Y there exists π ∈ HomA (Y, X) P such that
π ◦ ε = Id. There exists p ∈ Homk (Y, X) such that p ◦ ε = Id. Put π = ei T pfi .
Then for any x ∈ X we have
X X X
π (ε (x)) = ei T pfi (ε (x)) = eiT (pε (fi x)) = ei T (fi x) = T̄ x = Id .
Here we use fi ε = εfi . By Lemma 2.4 π ∈ HomA (X, Y ).
Now assume that X is injective. Define the map δ : X → A ⊗k X by the formula
X
f (x) = ei ⊗ fi x.
Then f ∈ HomA (X, A ⊗k X) by (2.3). It is obvious that f is injective. Thus, we
may consider X as a submodule of X, moreover X is a direct summand because X is
injective. So we have a projector τ : A ⊗k X → X. Let S ∈ Homk (A ⊗k X, A ⊗k X)
be defined by the formula
S (a ⊗ x) = h1, ai 1 ⊗ x.
Then
X X X
S̄ (a ⊗ x) = eiS (fi a ⊗ x) = h1, fi ai ei ⊗ x = hfi , ai ei ⊗ x = a ⊗ x
due to (2.2). Put T = τ ◦ S ◦ δ. Then T̄ = Id.
3. Relative projective and injective modules in group algebra

Let H be a subgroup of a group G. A k (G)-module X is H-injective if any exact
sequence of k (G)-modules
0 → X → Y → Z → 0,
which splits over k (H), splits over k (G).
In the similar way one defines H-projective module.
Let {g1 , . . . , gr } be a set of representatives in the set of left cosets G/H. For any
k (G)-modules X, Y , and T ∈ HomH (X, Y ) put
X
T̄ = gi T gi−1.
Prove yourself the following
Lemma 3.1. T̄ does not depend on a choice of representatives and T̄ ∈ HomG (X, Y ).
Theorem 3.2. The following conditions on k (G)-module X are equivalent
(1) X is H-injective;
(2) X is a direct summand in IndG
H X;
(3) X is H-projective;
(4) There exists T ∈ EndH (X) such that T̄ = Id.
Proof. This theorem is very similar to Theorem 2.6. To prove 1 ⇒ 2 check that
δ : X → IndG
H X defined by the formula
X
δ (x) = gi ⊗ gi−1 x,
defines an embedding of X. By injectivity X is a direct summand of IndGH X.
G
To prove 3 ⇒ 2 use the projection IndH X → X defined by g ⊗ x 7→ gx.
Now prove 2 ⇒ 4. Define S : IndG G
H X → IndH X by
X
S gi ⊗ xi = 1 ⊗ x1 ,
here we assume that g1 = 1. Check that S ∈ EndH IndG

H X and S̄ = Id. Then
obtain T = τ ◦ S ◦ δ, where τ : IndG
H X → X be the projection such that τ ◦ δ = Id.
Prove yourself 4 ⇒ 1 and 4 ⇒ 3 similarly to the first part of the proof of Theo-
rem 2.6.

The following corollary is important for us. Let p be prime. Recall that if |G| = ps r
with (p, r) = 1, then there exists a subgroup P of order ps . It is called a Sylow
subgroup. Two Sylow p-subgroups are conjugate in G.
Corollary 3.3. Let char k = p and P be a Sylow p-subgroup. Then every k (G)-
module X is P -injective.
Proof. We have to check condition (4) from Theorem 3.2. But r = [G : P ] is invertible
in k. So we can put T = r1 Id.
4. Finitely represented group algebras

Let char k = p, |G| = ps r with (p, r) = 1.
Lemma 4.1. Let H be a cyclic p-group, i.e. |H| = ps . Then there are exactly ps
isomorphism classes of indecomposable representations of H over k, exactly one for
each dimension. More precisely each indecomposable Lm of dimension m ≤ ps is
isomorphic to k (H) / (g − 1)m , where g is a generator of H.
s
= k [α] /αp , where α = g − 1, the corresponding quiver is the
Proof. Since k (H) ∼
s
loop quiver with one relation αp = 0. Hence α is a nilpotent Jordan block of order
≤ ps .
Theorem 4.2. If a Sylow p-subgroup of G is cyclic, then k (G) is finitely represented.

Moreover, the number of indecomposable k (G)-modules is not greater than |G|.
Proof. By Corollary 3.3 every indecomposable k (G)-module is P -injective. There-
fore, any indecomposable X is a direct summand in IndG P Li for some i. Clearly,
the number of such direct summands is finite. Now we will obtain the upper bound
on the number of indecomposable representations. Let X be an indecomposbale
k (G)-module, then by injectivity of X, X is a direct summand in IndG P X. De-
compose X into a direct sum of indecomposable k (P )-modules, then X must be a
direct summand in IndG P Li for some P -indecomposable summand Li of X. Hence
dim X ≥ dim Li = i. So if dim X = i, then X can be realized as a summand in
IndGP (Lj ) for some j ≤ i. To calculate the total number of non-isomorphic inde-
composable k (G)-modules, we can count in each IndG P Li only indecomposable k (G)-
components of dimension ≥ i since others are realized in IndG
P Lj for j < i. Since there
is no more than r such components for each i, the total number of non-isomorphic
indecomposable k (G)-modules is not greater than ps r = |G|.
Lemma 4.3. If P is a non-cyclic p-group, then P contains a normal subgroup N
= Zp × Zp .
such that P/N ∼
Proof. If P is abelian, the statement follows from the classification of finite abelian
groups. If P is not abelian, then P has a non-trivial center Z, and P/Z is not cyclic.
The statement follows by induction on |P |.
Lemma 4.4. The group S = Zp × Zp has an indecomposable representation of
dimension n for each n ∈ Z≥0 .
Proof. Let g and h be two generators of S, α = g − 1, β = h − 1. Then A =
k (S) / (α2 , β 2, αβ, βα) is the subalgebra of k (Q) for Kronecker quiver Q. In par-
ticular, one can see easily that every indecomposable representation of Q remains
indecomposable after restriction to A. This implies the Lemma.
Theorem 4.5. If a p-Sylow subgroup of G is not cyclic, then G has an indecompos-
able representation of arbitrary high dimension.
Proof. By Lemma 4.3 and Lemma 4.4, P has an indecomposable representation Y
of dimension n for any positive integer n. Decompose IndG
P Y into direct sum of
indecomposable k (G)-modules. At least one component X contains Y as an inde-
composable k (P ) component. Hence dim X ≥ n.
Corollary 4.6. The group algebra k (G) is finitely represented over a field of char-
acteristic p iff a Sylow p-subgroup of G is cyclic.
FINAL EXAM
MATH 252
Choose and solve one problem from the list below. The exam is due December 14.
Problem 1. Let PSL2 (Fq ) be the quotient of SL2 (Fq ) by the center. Compute
the table of irreducible characters for the group PSL2 (Fq ) over C.
Problem 2. Let k be algebraically closed of characteristic 0, G be a finite group,
K be a subgroup. The Hecke algebra H (G, K) is the subalgebra of the group algebra
k (G)
H (G, K) = {u ∈ k (G) |h1 uh2 = u for any h1, h2 ∈ H} .
(a) Show that dim H (G, K) equals the number of double cosets K\G/K.
(b) Show that H (G, K) is commutative iff the multiplicity of any irreducible rep-
resentation of G in IndGK (triv) is not greater than 1.
(c) Prove that H (G, K) is commutative for G = Sp+q , K = Sp × Sq .
(d) Describe H (G, K) for G = GL3 (Fq ), K being the subgroup of upper triangular
matrices.
Problem 3. Let G be the group of symmetries of an n-dimensional cube in Rn .
(a) Classify irreducible representation of G. Hint: show that G is a semi-direct
product of Sn and the normal subgroup H, isomorphic to Z2n , and use induction from
H.
(b) Let ρ be the permutation representation of G induced by the action of G on the
set of vertices of an n-dimensional cube. Decompose ρ into the sum of irreducibles.
Problem 4.
(a) Let R be a ring, S be a subring. Show that if P is a projective S-module, then
IndRS P = R ⊗S P is a projective R-module.
(b) Let p be a prime number. Describe projective indecomposable and irreducible
representations of Sp over Fp . Hint: use induction from Sp−1 .
(c) For p = 5 find dimensions of all irreducible representations.
Problem 5. Show that the Coxeter functor Φ+ coincides with the functor τ defined
in Crawley-Boevey lectures (page 22), i.e. Φ+ (X) =DExt1 (X, k (Q)) for any X.
Problem 6. Choose your favorite affine quiver different from Ân (if you choose
Dn do it for all n) and your favorite orientation. Describe all indecomposable repre-
b
sentations, regular simple representations and tubes.

1
SOLUTIONS OF SELECTED HOMEWORK PROBLEMS
MATH 252
Problem. Let G be the group of matrices

 
1 x y
0 1 z 
0 0 1
where x, y, z are elements of the finite field F5 . Classify irreducible representations
of G over C.
Solution. There are 5 conjugacy classes with one element
 
1 0 y
0 1 0  ,
0 0 1
for any y ∈ F5 and 24 conjugacy classes, each has one representative
 
1 x 0
0 1 z  ,
0 0 1
for some x, y ∈ F5 such that x 6= 0 or y 6= 0. Let H = [G, G]. Then H coincides with
the center of G and consists of matrices
 
1 0 y
0 1 0 .
0 0 1
Ther is 25 one-dimensional representations, obtained from the representation of
G/H = Z5 × Z5 . The remaining four representations have dimension 5, and can
be obtained by induction from the subgroup K of matrices
 
1 x y
Mx,y = 0 1 0  .
0 0 1
Let u ∈ F∗5 and χu (Mx,y ) = e2πuyi/5 . Then ρu = IndG 6 ρv if
K χu has dimension 5, ρu =
∼
v 6= u, since the action of the center is different. Finally
X
hχρu , χρu iG = hResK χρu , χu iK = χtu , χu K = 1,
t∈F5

1
2 SOLUTIONS OF SELECTED HOMEWORK PROBLEMS MATH 252
where χtu (Mx,y ) = χu (Mx,y+xt ). That proves irreducibility of each ρu . Alternatively,

one can prove that ρu is irreducible by assuming the contrary. Then a subrepresen-
tation must have dimension 1 (divides the order of the group), but this is impossible
since ρu (H) 6= 1.
Problem. Let G be a finite group, r be the number of conjugacy classes in G
and s be the number of conjugacy classes in G preserved by the involution g → g −1 .
Prove that the number of irreducible representations of G over R is equal to r+s 2
.
Solution. Let χ be an irreducible character of G over C. If χ (g) ∈ R for all g ∈ G,
then there is one irreducible representation of G over R with character χ (real) or
2χ (quaternionic). If χ (g) ∈ / R at least for one g, then the pair χ and χ̄ produce
one irreducible representation of G over R (complex) with character χ + χ̄. Hence
if m is the number of irreducible representations of G over R and p is the number
of irreducible characters ϕ such that ϕ (g) ∈ R for all g, then m = p + r−p 2
= r+p2
.
Define the linear operator T on the space of class functions by the formula T ϕ (g) =
ϕ (g) + ϕ (g −1 ). Then rk T = s + r−s
2
= r+s
2
. On the other hand, if ϕ is an irreducible
−1
character, then ϕ̄ (g) = ϕ (g ), hence T (ϕ) = ϕ + ϕ̄. Since irreducible characters
form a basis in the space of class function, one obtains rk T = m.
Problem. Let R be the algebra of polynomial differential operators. In other
∂
words R is generated by x and ∂x with relation
∂ ∂
x−x = 1.
∂x ∂x
(The algebra R is called the Weyl algebra.) Let M = C [x] have a structure of R-
module in the natural way. Show that EndR (M) = C, M is an irreducible R-module
and the natural map R → EndC (M) is not surjective.
Solution. Note that 1 generates M and if f ∈ EndR (M) then f (p) = pf (1)
∂
for any p ∈ M. But ∂x (f (1)) = 0. Hence f (1) = c for some c ∈ C. Therefore
EndR (M) = C. On the other hand, every submodule of M contains 1, therefore
M is irreducible. Finally, note that every d ∈ R has a finite-dimensional kernel.
Therefore EndC (M) 6= R.
Problem. Let R be the subalgebra of upper triangular matrices in Matn (C). Clas-
sify simple and indecomposable projective modules over R and evaluate Ext·R (M, N)
for all simple M and N.
Solution. Let Eij denote the elementary matrix with 1 in one place. Then
primitive idempotents are Eii , i = 1, . . . , n. Indecomposable projectives are Pi =
REii . Note that Pi is isomorphic to the maximal submodule of Pi+1 . Hence simple
modules are Si = Pi /Pi−1 , if we put P0 = 0. Thus, the complex
0 → Pi−1 → Pi → 0
is a projective resolution of Si . Hence Extk (Si , Sj ) = 0 if k > 1. Now use that

= C and HomR (Pi , Sj ) = 0 if i 6= j, because each Pi has a unique
HomR (Pi , Si ) ∼
SOLUTIONS OF SELECTED HOMEWORK PROBLEMS MATH 252 3
simple quotient isomorphic to Si . Thus, we obtain HomR (Si , Sj ) = 0 if i 6= j,

HomR (Si , Si ) = C, Ext1 (Si , Sj ) = 0 if i 6= j + 1, Ext1 (Si+1 , Si ) = C.
Problem. Let Q be a connected quiver and k (Q) be the path algebra of Q. Show
that the center of k (Q) is isomorphic either to k, or to k [x], and that the latter
happens only in the case when Q is an oriented cycle.
Solution. Let c be an element of the center of k (Q). Without loss of generality
we may assume that c is a linear combination of paths of the same length. Assume
that there is an element of the P center c of non-zero degree (recall that degree is the
length of a path). Write c = cij , where cij = ei cej .
First, we claim that cij = 0 if i 6= j. Indeed, if cij 6= 0, then ei c = cei implies
ei cij ∈ k (Q) ei , which is impossible.
Next, we claim that if cii 6= 0 for one i, then cjj 6= 0 for all j. Indeed, assume
the opposite, then, since Q is connected, there exists γ = i → j such that either
cii = 0, cjj 6= 0 or cjj = 0, cii 6= 0. In the former case ej γcei = γcii = 0 and
ej cγei = cjj γ 6= 0, which contradicts γc = cγ. Similarly, in the latter case ej γcei 6= 0,
ej cγei = 0. Contradiction.
Finally, let γ and δ ∈ Q1 and s (γ) = s (δ) = i. Then cγ = γc = γcii implies
cii ∈ k (Q) γ. By the same reason cii ∈ k (Q) δ, which implies γ = δ. In the same
way, if t (γ) = t (δ), then γ = δ. Thus, if there is a central c such that deg c > 0, then
Q is one oriented cycle.
Assume first, that QPis not an oriented cycle. The any central element c has degree
0, and therefore c = bi ei . If γ = i → j, then cγ = γc implies bi = bj . But Q is
connected, hence b1 = · · · = bn . That proves that the center of k (Q) is isomorphic
to k.
Let k (Q) be one oriented cycle of length n. Since we already proved that a central
element P c is a combination of cycles, n divides deg c. A central element of degree sn
equals b sn-cycles, and hence the center is isomorphic to k [z], where z is the sum
of all n-cycles.
Problem. Let Rep (a, b, c) be the space of all representations of the quiver
•→•←•
with dimension vector (a, b, c). List all orbits in Rep (a, b, c). Show that there is only
one open orbit. Describe the open orbit OX in terms of decomposition of X into
direct sum of indecomposable representations.
Solution. A point Rep (a, b, c) is a pair of linear operators P : k a → k b and
Q : k c → k b . An orbit is determined by three numbers, p = rk P , q = rk Q and
r = dim (Im P ∩ Im Q), and we have p ≤ min (a, b), q ≤ min (b, c), r ≤ min (p, q).
Positive roots corresponding to indecomposable modules are
α1 = (1, 0, 0) , α2 = (0, 1, 0) , α3 = (0, 0, 1) , β1 = α1 +α2, β2 = α2 +α3 , γ = α1 +α2+α3 ,
and the decomposition of (P, Q) into the sum of indecomposables is
(a − p) α1 + (b − p − q + r) α2 + (c − q) α3 + (p − r) β1 + (q − r) β2 + rγ.
4 SOLUTIONS OF SELECTED HOMEWORK PROBLEMS MATH 252
By Xν we denote the indecomposable representation of dimension ν. Check that

indecomposable projectives are
Xβ1 = Ae1, Xα2 = Ae2, Xβ2 = Ae3,
and the projective resolutions of Xα1 , Xα3 and Xγ are
0 → Xα2 → Xβ1 → 0, 0 → Xα2 → Xβ2 → 0,
0 → Xα2 → Xβ1 ⊕ Xβ2 → 0.
Therefore,
Ext1 (Xα1 , Xα2 ) = Ext1 (Xα3 , Xα2 ) = Ext1 (Xα1 , Xβ2 ) = Ext1 (Xα3 , Xβ1 ) = Ext1 (Xγ , Xα2 ) = k,
all other Ext1 are trivial.
To determine the open orbit we find possible triples of positive roots without
mutual extensions.
{α1 , β1, γ} , {α3 , β2, γ} , {α1, α3 , γ} , {β1, β2, γ} , {β1, β2, α2 } .
The open orbit in Rep (a, b, c) is a combination of one of these triples, here x =
(a, b, c) :
(1) If a ≥ b ≥ c, then x = (a − b) α1 + (b − c) β1 + cγ;
(2) If a ≤ b ≤ c, then x = (c − b) α3 + (b − a) β1 + aγ;
(3) If a, c ≥ b, then x = (a − b) α1 + (c − b) α3 + bγ;
(4) If a, c ≤ b, a + c ≥ b, then x = (b − c) β1 + (b − a) β2 + (a + c − b) γ;
(5) If a, c, a + c ≤ b, then x = aβ1 + cβ2 + (b − a − c) α2 .

Serganova V. - Representation theory (2005)

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Serganova V. - Representation theory (2005)

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Serganova V. - Representation theory (2005)

Uploaded by

Copyright:

Available Formats

Department of Mathematics

University of California, Berkeley

• Representations of groups. Definitions and examples

1. Some problems involving representation theory

Date: August 30, 2005.

Theorem 1.1. There exist decompositions

2. Representations of groups. Definition and examples.

5. Regular representation. Recall

Two representations ρ : G → GL (V ) and σ : G → GL (W ) are equivalent (or

Then for any f ∈ F (G) we have

3. Operations with representations

Dual representation. For any representation ρ : G → GL (V ) one can define the

4. Invariant subspaces and irreducibility

Check that ρs ◦ P̄ ◦ ρ−1

In the previous example V = W ⊕ W ′ if char k does not divide n. Otherwise,

Corollary 5.3. If ρ : G → GL (V ) is irreducible then EndG (V ) is a division ring. If

is an isomorphism. There exists i such that pi ◦ q1 6= 0. (Otherwise F (V1 ) = 0

isomorphism and i 6= s (r) for any r < j. (Indeed, otherwise F (V1 ⊕ · · · ⊕ Vj ) ⊂

Date: August 30, 2005.

Since tr P̄ (i, j) = tr P (i, j) = δij , we have

It is easy to check that ρ is a representation and

Now let us evaluate (χρ , χρ ).

Let σ be the representation associated with the action of G on X × X. Then the

It is easy to check that the restriction of (,) on C (G) is non-degenerate.

Corollary 1.9. The number of isomorphism classes of irreducible representations

Due September 16.

Date: September 8, 2005.

Date: September 16, 2005.

is the Klein subgroup. Observe that S4 /V4 ∼

It is easy to check that

One can solve (1.2) and (1.3)

Thus, representations of G over k are k (G)-modules.

Proof. By Lemma 2.2

k (G) = V1⊕n1 ⊕ · · · ⊕ Vl⊕nl ,

3. Finitely generated modules and Noetherian rings.

A ring satisfying the conditions of Lemma 3.2 is called left Noetherian.

Proof. Let M = Rx1 + . . . Rxn , the there exists a surjective homomorphism p :

Let R be a commutative ring. An element x ∈ R is called integral over Z if

4. The center of the group algebra k (G)

On the other hand, by Theorem 2.5 we have

If c1 , . . . , cl are disjoint conjugacy classes, then clearly Zδc1 + · · · + Zδcl is a subring in

Then Ker ρm contains a subgroup

Due September 23.

Date: September 15, 2005.

2. Induced representations for groups.

Corollary 2.3. Let H be a normal subgroup in G. Then IndG

Here a subindex indicates the group where we take inner product.

Example 1. Let ρ be a trivial representation of H. Then IndG H ρ is the permutation

Thus, by induction we obtain an irreducible two-dimensional representation of G.

3. Double cosets and restriction to a subgroup

the subgroup of upper-triangular matrices in G. Check that |G| = (q 2 − 1) (q 2 − q),

for any h ∈ sHs−1 .

by α (t ⊗ v) = ts ⊗ v for any t ∈ K, v ∈ V . It is well defined

and obviously surjective. Injectivity can be proved by counting dimensions.

where H = B ∩ sBs−1 is a subgroup of diagonal matrices and

is not isomorphic to ρ for any s ∈ G/H, s ∈

here sgn denotes the sign representation. Furthermore, IndG G

Due September 30.

Date: September 23, 2005.