UNDERSTANDING IRREDUCIBLE REPRESENTATIONS

ALEX ROSENFELD Abstract. This paper acts as an introduction to representation theory. In addition to dening representations of Lie algebras and of nite groups, this paper will go through two standard examples, the Lie algebra sl2 C and the nite group Sn , to illustrate techniques for their construction and use. In addition, those examples will develop some important theory behind representations of Lie algebras and nite groups.

Contents 1. Introduction 2. The Irreducible Representations of sl2 C 3. The Irreducible Representations of Sn 4. Characters of Representations of Finite Groups References 1 2 5 8 9

1. Introduction Groups are structures that are dened by very simple rules, but are varied in description and activity. On the other hand, vector spaces are very easily described in very plain terms. Thus, to get more information about groups, or other mysterious structures like Lie algebras, we consider them as vector spaces. We do this with something called a representation. Representations take the operation of a group or Lie algebra or other complicated structure into an action on a vector space. To get a good grasp of how these representations work, this paper will examine the Lie algebra sl2 C and the nite group Sn . sl2 C is a Lie algebra whose representations are very easy to describe. Even more than this, the development of the representations of sl2 C will act as an entryway into understanding the representations of many other Lie algebras, such as sln C and gln C. Sn does not generalize as well as sl2 C. The techniques used to discover the irreducible representations of Sn are very specic to Sn . However, what is lost in generalizability is gained in understandability as many questions about Sn can be answered relatively simply. For example, each irreducible representation corresponds with a conjugacy class of Sn . This correspondence allows a very easy construction of the irreducible representations from the well known conjugacy classes. Even though it is hard to describe the irreducible representations of a nite group in general, breaking down a representation into those representations is a breeze with characters. Characters are very simple, very powerful tools that in some sense
Date: DEADLINE AUGUST 22, 2008.
1

ALEX ROSENFELD

encode a representation as something similar to a characteristic function. This code uniquely identies a representation and can be used to decompose a representation as a direct sum of irreducible representations. Note that these techniques work for every nite group and so will be used to decompose Sn . Although representation theory is ubiquitous in mathematics, we are especially interested in the eld because it is being used in complexity theory for an attempt to prove P = N P . However, that topic is outside the scope of this paper and so will not be discussed.

2. The Irreducible Representations of sl2 C The following are some basic denitions needed for describing the representations of Lie algebras. Denition 2.1. A Lie algebra is a vector space g equipped with a bilinear, skewsymmetric map [, ] : g g g, called the Lie bracket, satisfying the Jacobi identity: [x, [y, z]] + [y, [z, x]] + [z, [x, y]] = 0 for all x, y, and z in g. Example 2.2. If V is a vector space, then the endomorphisms of V , labeled gl(V ), forms a Lie algebra with bracket [x, y] := x y y x for all x and y in g. Denition 2.3. For Lie algebras g and h a homomorphism from g to h is a linear map : g h where ([x, y]) = [(x), (y)]. Denition 2.4. A representation of a Lie algebra g on a nite-dimensional complex vector space V is a homomorphism : g gl(V ). Given a representation , we can dene an action of g on V by X(v) = (X)(v) Denition 2.5. A representation is irreducible if there is no proper, nontrivial subspace of V that is invariant under g. A very important starting point in the description of representations of Lie algebras is sl2 C. The description of the representations on many dierent Lie algebras is analogous to the process used below, especially sln (C) and gln (C). Denition 2.6. sl2 C is the Lie algebra {A M2 (C)|tr(A) = 0} with bracket [X, Y ] = XY Y X. A useful basis of sl2 C as a vector space is given by H= 1 0 ,F = 0 1 0 0 1 , and B = 0 0 1 0 . 0

It is clear that these form a basis of sl2 C. By simple computation, we have [H, F ] = 2F , [H, B] = 2B, and [F, B] = H. So far we have dened representations and sl2 C. Now, we need to show such representations exist for sl2 C.

UNDERSTANDING IRREDUCIBLE REPRESENTATIONS

Example 2.7. Let n N. Let Wn be an n-dimensional complex vector space. We dene a representation : sl2 C gl(Wn ) by n1 n3 n5 .. (H) = , . n + 5 n + 3 n + 1 0 1 0 1 0 0 1 0 1 1 0 .. .. .. .. (F ) = , and (B) = . . . . . 0 1 1 0 0 1 1 0 0 1 0 It is easily checked that (H)(F ) (F )(H) = 2(F ), (H)(B) (B)(H) = 2(B), and (F )(B) (B)(F ) = (H) showing that is a representation. Also, note that, by our construction, there exists a representation for every dimension. And it turns out that Wn is the canonical representation of sl2 C as follows from the following theorem. Theorem 2.8. Every n-dimensional irreducible representation of sl2 C is isomorphic to Wn . Proof. Let be an irreducible representation of sl2 C on a nite-dimensional complex vector space V . Since C is an algebraically closed eld, (H) gl(V ) has n eigenvalues (counting multiplicities). Let C be such an eigenvalue with eigenvector v, i.e., H(v) = v. The bracket relations between H, B, and F determine how B and F act on v. For example, H(F (v)) = [H, F ](v) + F (H(v)) = 2F (v) + F (v) = ( + 2)F (v). Similarly, H(B(v)) = ( 2)B(v). Let V = {w gl(V ) : H(w) = w} be the eigenspace of . The above facts can be restated as: (F ) maps V into V+2 , (B) maps V into V2 , and, clearly, (H) maps V into V Let z be an eigenvalue of (H) such that z + 2 is not an eigenvalue of (H) (such a z must exist since V is nite dimensional) and let v be a corresponding eigenvector. Thus, F (v) Vz+2 = (0) and so F (v) = 0. Let W = Span{v, B(v), B 2 (v), . . . }. Lemma 2.9. W = V . Proof of lemma. Since is irreducible by assumption, it suces to show that W is invariant under the action of sl2 C. Since sl2 C is generated by H, F , and B, it suces to show that W is invariant under the action of these three elements. To show that W is invariant under H, consider B m (v). By induction, H(B m (v)) = (z 2m)B m (v) which is clearly in W .

ALEX ROSENFELD

To show that W is invariant under F , recall that F (v) = 0. Thus, F (v) is contained in W . For m > 0, F (B m (v)) = [F, B](B m1 (v) + B(F (B m1 (v))). Thus, by induction, F (B m (v)) = m(z m + 1)B m1 (v), which is in W . Finally, it is clear from the denition of W that it is invariant under B. Thus, we have proved that W = V , which concludes the proof of the lemma. Let n = min{m : B m (v) = 0}. We know F (B n (v)) = F (0) = 0 and, by a simple induction argument, F (B n (v)) = n(z n+1)B n1 (v). Thus, n(z n+1)B n1 (v) = 0. Since n > 0 and B n1 (v) = 0, z = n 1. Thus, z, our chosen eigenvalue, is an integer. Since B m (v) = 0 for all m n, W = Span({v, B(v), ..., B n1 (v)}). Since H(B m (v)) = (z 2m)B m (v), each of the vectors B m (v) is in a dierent eigenspace of (H) and thus they are linearly independent. Therefore, {v, B(v), ..., B n1 (v)} is a basis for V . In particular, n is the dimension of V . Thus, V = Vz Vz+2 ... Vz2 Vz . Since z = n 1, we have V = Vn+1 Vn+3 ... Vn3 Vn1 , which is clearly isomorphic to the representation Wn . Note that Wn = Vn+1 Vn+3 ... Vn3 Vn1 . Sometimes it is more convenient to refer to an irreducible representation by its largest eigenvalue rather than its dimension, so we dene V () by V () := W+1 = V V+2 ... V2 V . As mentioned earlier, irreducible representations are the building blocks of representations. The nice thing about sl2 C is that it is easy to determine the irreducible subrepresentations of any representation. Let be any representation. Find the eigenvalues of (H) and their multiplicities. We take the largest eigenvalue, say n, and note that V (n) is a part of our vector space. We then remove 1 from the multiplicities of each eigenvalue of V (n) and repeat the process until all the multiplicities are 0. Those V (n) are the irreducible representations of . For example, let the following diagram be the eigenvalues and multiplicities of (H). multiplicities 1 eigenvalues 3 0 2 3 1 3 0 1 1 0 1 2 3

The largest eigenvalue is 3, so we know that V (3) is part of our representation, so remove the eigenvalues of V (3) , 3, 1, 1, and 3, and get: multiplicities 0 eigenvalues 3 0 2 2 1 2 0 0 1 0 1 2 3

The next highest is 1, so we know V (1) is a part of our representation. We remove the eigenvalues of V (1) to get: multiplicities 0 eigenvalues 3 0 2 1 1 1 0 0 1 0 1 2 3

We repeat the process to nd that again V (1) is a part of our representation and are left with:

UNDERSTANDING IRREDUCIBLE REPRESENTATIONS

All thats left to remove is the one where the eigenvalue is 0, so V (0) is also part of our representation. Thus, V (3) , V (1) , V (1) , and V (0) are the irreducible representations in . It wont be proven in this paper, but, in general, a representation is equal to the direct sum of its irreducible representations. Thus, the space associated to is V (3) V (1) V (1) V (0) . 3. The Irreducible Representations of Sn In addition to Lie algebras, representations are often used to characterize nite groups. Like Lie algebras, nite groups also have irreducible representations and, as in the case of sl2 C, each nite-dimensional representation can be split as a direct sum of a nite number of irreducible representations. Denition 3.1. A representation of a nite group G on a nite-dimensional vector space V is a homomorphism from G to GL(V ) = {A Mn (C) : det(A) = 0}.

multiplicities 0 eigenvalues 3

0 2

0 1

1 0 0 0 0 1 2 3

Denition 3.2. A representation is irreducible if there is no proper, nontrivial subspace of V that is invariant under the action of G. Both denitions are very similar to those used for Lie algebras. The idea is the same: a representation is still an action on a vector space and the irreducible representations form the building blocks of arbitrary representations. Proposition 3.3. The number of irreducible representations for a nite group is equal to the number of conjugacy classes. Example 3.4. Since Sn has at least three conjugacy classes, it has at least three irreducible representations: One is called the trivial representation which is on C and acts by (v) = v for Sn and v C. Another one is called the alternating representation which is also on C, but acts by (v) = sign()v for Sn and v C. A third irreducible representation of Sn is called the standard representation and it is on V = {(z1 , z2 , z3 ) C3 : z1 + z2 + z3 = 0} acting by ((z1 , z2 , z3 )) = (z(1) , z(2) , z(3) ) for Sn and (z1 , z2 , z3 ) C3 . Note that for n > 3 there are more irreducible representations than just these three. We will focus on Sn , the symmetric group on n elements, and its irreducible representations. So, the question becomes: how many conjugacy classes of Sn are there? It turns out that the number of conjugacy classes of Sn is the number of ways of writing n as a sum of a sequence of n descending, nonnegative numbers, which is elaborated in the following proposition. Denition 3.5. A partition of n N is an n-tuple (1 , 2 , ..., n1 , n ) Z such that n = 1 + 2 + ... + n1 + n and 1 2 ... n1 n 0. Proposition 3.6. The conjugacy classes of Sn correspond to partitions of n.

ALEX ROSENFELD

Proof. The conjugacy classes of Sn are uniquely determined by the cycle type of their elements, and each class contains only one cycle type. Thus, the conjugacy classes correspond to a partition of n into the cycle type of the class. In fact, for Sn the irreducible representations are naturally in bijective correspondence with the conjugacy classes; we will demonstrate this correspondence in the remainder of this section. Young diagrams are a fundamental tool for keeping track of combinatorial data related to partitions, and are especially relevant to the representation theory of the symmetric groups. The Young diagram corresponding to the partition (1 , . . . , n ) of n is an arrangement of boxestypically aligned to the top and to the leftwith the 1 boxes in the rst row, 2 boxes in the second row, and so on. For example, the Young diagram for a partition = (3, 3, 2, 1, 0, 0, 0, 0, 0) is the following: 1 2 3 4 5 6 7 8 9 Note: Parts of size 0 are not displayed in a Young diagram. In order to reference certain boxes, the Young diagram has its boxes labeled with numbers 1 through n in reading order. Example 3.7. For S3 ,
1 2 3

is the Young diagram of the partition (3, 0, 0),

1 2 3

is

the Young diagram of the partition (1, 1, 1), and partition (2, 1, 0).

is the Young diagram of the

Numbering the boxes allows us to dene an action of Sn . Let R = {g Sn : g preserves each row} and let C = {g Sn : g preserves each column}. For example, if = (3, 3, 2, 1, 0, 0, 0, 0, 0), (13) R because 1 and 3 are both in the rst row but (13) C because 1 is in the rst column and 3 is in the third column, thereby / switching the columns of those two numbers. Example 3.8. For the Young diagram 1 2 3 , every permutation preserves the rows. Thus, R(3,0,0) = Sn = {1, (12), (13), (23), (123), (321)}. However, only the identity preserves the columns, so C(3,0,0) = {1}. For the Young diagram , only the identity preserves the rows, so R(1,1,1) = {1}. On the other hand, every permutation preserves the column, so C(1,1,1) = Sn = {1, (12), (13), (23), (123), (321)}.
1 2 1 2 3

For the Young diagram 3 , only the identity and (12) preserve the rows, so R(2,1,0) = {1, (12)}. The identity and (13) preserve the column, so C(2,1,0) = {1, (13)}. In order to use R and C to determine the irreducible representations of Sn , its important to know the space of those representations. To that end, let CSn be the group algebra of Sn , i.e., the vector space with basis {eg : g Sn } with multiplication given by eg eh = egh . CSn is part of the vector space used in the construction of the irreducible representations from R and C . Given a partition , we dene a , b CSn by a = gR eg and b = sign(g)eg . gC

UNDERSTANDING IRREDUCIBLE REPRESENTATIONS

Example 3.9. The R and C for the Young diagrams of S3 are: R(3,0,0) = {1, (12), (13), (23), (123), (321)}, so a(3,0,0) = e1 + e(12) + e(13) + e(32) + e(321) + e(123) . C(3,0,0) = {1}, so b(3,0,0) = e1 . R(1,1,1) = {1}, so a(1,1,1) = e1 . C(1,1,1) = {1, (12), (13), (23), (123), (321)}, so b(1,1,1) = e1 e(12) e(13) e(23) + e(321) + e(123) . R(2,1,0) = {1, (12)}, so a(2,1,0) = e1 + e(12) . C(2,1,0) = {1, (13)}, so b(2,1,0) = e1 e(13) .

Denition 3.10. The Young symmetrizer c for a partition is dened to be c = a b .

Example 3.11. The Young symmetrizers of the Young diagrams of S3 are: c(3,0,0) = a(3,0,0) b(3,0,0) = (e1 + e(12) + e(13) + e(32) + e(321) + e(123) )e1 = e1 + e(12) + e(13) + e(32) + e(321) + e(123) . Similarly, c(1,1,1) = a(1,1,1) b(1,1,1) = e1 (e1 e(12) e(13) e(23) + e(321) + e(123) ) = e1 e(12) e(13) e(23) + e(321) + e(123) . For (2, 1, 0), c(2,1,0) = a(2,1,0) b(2,1,0) = (e1 + e(12) )(e1 e(13) ) = e1 + e(12) e(13) e(123) . As the following theorem states (proof omitted), these Young symmetrizers determine the irreducible representations of Sn . Theorem 3.12. Let CSn c = {z c : z CSn }. The representation of Sn on CSn c dened by ()(z) := e z for Sn and z CSn c is irreducible. In addition, every irreducible representation of Sn is of this form for some partition . Example 3.13. Let be in S3 . For the partition (3, 0, 0), c(3,0,0) = e1 + e(12) + e(13) + e(32) + e(321) + e(123) = a(3,0,0) . Since a(3,0,0) is the same if you rearrange the terms, CSn c(3,0,0) = Cc(3,0,0) . Thus, for C, (c(3,0,0) ) = (e1 +e(12) +e(13) +e(32) +e(321) +e(123) ) = (e1 +e(12) +e(13) +e(32) +e(321) +e(123) ) = c(3,0,0) . In other words, for v CSn c , (v) = v. Thus, this is the trivial representation. By similar computation, CSn c(1,1,1) = Cc(1,1,1) . Thus, for v CSn c , (v) = sign()v making this the alternating representation. The partition (2,1,0) is not as simple to analyze as the other representations, so a dierent approach is useful. Consider the vectors v1 = e1 + e(12) e(13) e(132) , v2 = e23 + e(132) e(123) e(12) , and v3 = e13 + e(123) e(1) e(23) . It is easily veried that v1 = (13)c(2,1,0) , v2 = (23)c(2,1,0) , v3 = c(2,1,0) , and that these three vectors span CSn c(2,1,0) . Through simple computation, it is easy to see that (v1 ) = v(1) , (v2 ) = v(2) , and (v3 ) = v(3) . Since v1 + v2 + v3 = 0, it follows that this representation is isomorphic to the standard representation. More than the above, we also can nd a nice formula for the dimension of an irreducible representation. To each box in the Young diagram of , we can associate a number called the hook length. The hook length for a box is calculated by taking the number of boxes below that box plus the number of boxes right of the box plus 1. The stars in the following gure give an example of a hook: o o o o o

ALEX ROSENFELD

For example, if = (3, 3, 2, 1, 0, 0, 0, 0, 0) and if we label each box with the hook length, we have the following gure: 6 4 2 5 3 1 3 1 1 Proposition 3.14. Let h be the product of all hook lengths in the Young diagram. The dimension of V is d! . h Thus, the dimension of an irreducible representation of Sn can be easily determined from the corresponding Young diagram. 4. Characters of Representations of Finite Groups It was noted in the previous section that there are the same number of irreducible representations as conjugacy classes. Characters are functions, which are constant on conjugacy classes, that correspond to a representation. These functions are analogous to the process used for Lie algebras since representations of nite groups are decomposed by characters as representations of Lie algebras are decomposed by eigenvalue. Denition 4.1. Let be a representation of a nite group G on a nite-dimensional complex vector space V . The character of V , V , is dened to be V (g) = tr((g)) for g G. Remark 4.2. Since trace is constant under conjugation, V (hgh1 ) = V (g). That is, characters are class functions.

The interesting thing about characters is that they allow an easy formula to describe how many of an irreducible representation are in an arbitrary representation. The following statements will establish a formula. Denition 4.3. Let (, ) be the Hermitian inner product on the vector space of 1 class functions from G to C dened by (, ) = |G| gG (g)(g).

Theorem 4.4. The characters of irreducible representations of a nite group G are orthonormal under this inner product.

Although it will not be proven here, any representation of a nite group G can be written as a direct sum of irreducible representations. Therefore, the following proposition is a formula for the character of a direct sum in terms of the characters of its summands. Proposition 4.5. V W = V + W . Corollary 4.6. For any representation V and irreducible representation W of a nite group G, (V , W ) is equal to the multiplicity of W in V. Proof. Proof of Corollary 4.6. Let V1 , . . . , Vn be the irreducible representations of G. Since a representation can be written as a sum of irreducible representations, a V = V1a1 Vn n where ai is the multiplicity of Vi in V . Thus, V = a1 V1 + + an Vn . Therefore, (V , W ) = a1 (V1 , W ) + + an (Vn , W ). Since characters of irreducible representations are orthonormal under this product, (Vi , W ) = 1 if Vi = W and (Vi , W ) = 0 otherwise. Thus, (V , W ) = ai , which is the multiplicity of Vi = W in V.

UNDERSTANDING IRREDUCIBLE REPRESENTATIONS

Therefore, any representation can be decomposed uniquely, up to isomorphism, into irreducible representations by studying the characters of the representation. However, knowing what these specic characters look like still needs to be addressed. In general, characters for representations of groups are hard to describe. However, one of the special things about Sn is that the characters follow a specic, though not necessarily nice formula as shown by the following theorem. Theorem 4.7. If is an element of Sn and = (1 , . . . , n ) is a partition for the irreducible representation V , then V () is the coecient of x1 x2 . . . xn in n 1 2 n j j ij where ij is the number of j-cycles in . i<j (xi xj ) j=1 (x1 + . . . xn )

Example 4.8. Using the above formula, it is easy to describe the characters of the irreducible representations. Let 1 be the identity of S3 , let 2 be a representative of the conjugacy class of elements of S3 that have a 2-cycle, and let 3 be a representative of the conjugacy class of elements of S3 that have a 3-cycle. The characters of the irreducible representations are as follows: Let T be the trivial representation. From the formula, T (1 ) = 1, T (2 ) = 1, and T (3 ) = 1. Let A be the alternating representation. From the formula, A (1 ) = 1, A (2 ) = 1, and A (3 ) = 1. Let S be the standard representation. From the formula, S (1 ) = 2, S (2 ) = 0, and S (3 ) = 1. As an example of how to decompose a given representation into irreducible representations, let V be a representation with characters V (1 ) = 5, V (2 ) = 1, and V (3 ) = 1. The following comes from the use of the corollary above. 1 (V (1 )T (1 ) + 3(V (2 )T (2 )) + 2(V (3 )T (3 ))) = 3! 1 (5 1 + 3 1 1 + 2 (1) 1) = 1 6 Therefore, V contains one copy of the trivial representation. (V , T ) =

1 (V (1 )A (1 ) + 3(V (2 )A (2 )) + 2(V (3 )A (3 ))) = 3! 1 (5 1 + 3 1 (1) + 2 (1) 1) = 0 6 Therefore, V does not contain any copies of the alternating representation. (V , A ) = 1 (V (1 )S (1 ) + 3(V (2 )S (2 )) + 2(V (3 )S (3 ))) = 3! 1 (5 2 + 3 1 0 + 2 (1) (1)) = 2 6 Therefore, V contains two copies of the standard representation. Therefore, V = T S S. (V , S ) =

This shows how through characters, it is easy to decompose representations into irreducible representations. References
[1] W. Fulton and J. Harris. Representation Theory: A First Course. Springer-Verlag. 1991.