STUDENT ARTICLE

4 Young Tableaux and the Representations of the Symmetric Group

Yufei Zhao Massachusetts Institute of Technology 10 Cambridge, MA 02139 [email protected] Abstract We explore an intimate connection between Young tableaux and representations of the symmetric group. We describe the construction of Specht modules which are irreducible representations of Sn , and also highlight some interesting results such as the branching rule and Youngs rule. Some knowledge of basic representation theory is assumed.

4.1

Introduction

In this article, we explore a connection between representations of the symmetric group Sn and combinatorial objects called Young tableaux. We dene Young tableaux in Section 4.2, but for now, it sufces to say that they are llings of a certain conguration of boxes with entries from {1, 2, . . . , n}, an example of which is shown below. 1 2 4 3 5 6 7 8 9 So how are representations of Sn related to Young tableau? It turns out that there is a very elegant description of irreducible representations of Sn through Young tableaux. Let us have a glimpse of the results. Recall that there are three irreducible representations of S3 . It turns out that they can be described using the set of Young diagrams with three boxes. The correspondence is illustrated below.

trivial representation

sign representation

standard representation

It is true in general that the irreducible representations of Sn can be described using Young diagrams of n boxes! Furthermore, we can describe a basis of each irreducible representation using standard Young tableaux, which are numberings of the boxes of a Young diagram with 1, 2, . . . , n such that the rows and columns are all increasing. For instance, the bases of the standard representation of S3 correspond to the following two standard Young tableaux: 1 2 3 1 3 2

Yufei Zhao, Massachusetts Institute of Technology 10, is a mathematics and computer science major. His favorite mathematical area is combinatorics, but he also enjoys algebra and number theory. He is regularly involved with the training of the Canadian team for the International Math Olympiad.

33

34

T HE H ARVARD C OLLEGE M ATHEMATICS R EVIEW 2.2

The dimension of the irreducible representations can be easily computed from its Young diagram through a result known as the hook-length formula, as we explain in Section 4.4. There are many other surprising connections between Young tableaux and representations of Sn , one of which is the following. Suppose we have an irreducible representation in Sn and we want to nd its induced representation in Sn+1 . It turns out that the induced representation is simply the direct sum of all the representations corresponding to the Young diagrams obtained by adding a new square to the original Young diagram! For instance, the induced representation of the standard representation from S3 to S4 is simply
4 IndS S3

Similarly, the restricted representation can be found by removing a square from the Young diagram: ResS2 = .

In this paper, we describe the connection between Young tableaux and representations of Sn . The goal is to attract readers to the subject by showing a selection of very elegant and surprising results. Most proofs are omitted, but those who are interested may nd them in [Fu], [FH], or [Sa]. We assume familiarity with the basics of group representations, including irreducible representations and characters. Induced representations are used in Section 4.5. For references on group representations, see [FH], [Sa], or [Se]. In Section 4.2, we introduce Young diagrams and Young tableaux. In Section 4.3, we introduce tabloids and use them to construct a representation of Sn known as the permutation module M . However, permutation modules are generally reducible. In Section 4.4, we construct irreducible representations of Sn known as Specht modules S . Specht modules S correspond bijectively to Young diagrams and they form a complete list of irreducible representations. In Section 4.5, we discuss the Young lattice and the branching rule, which are used to determine the induced and restricted representations of S . Finally, in Section 4.6, we introduce Kostka numbers and state a result concerning the decomposition of permutation modules into the irreducible Specht modules.

4.2

Young Tableaux

First we need to settle some denitions and notations regarding partitions and Young diagrams. Denition 1. A partition of a positive integer n is a sequence of positive integers = (1 , 2 , , l ) satisfying 1 2 l > 0 and n = 1 + 2 + + l . We write n to denote that is a partition of n. For instance, the number 4 has ve partitions: (4), (3, 1), (2, 2), (2, 1, 1), (1, 1, 1, 1). We can also represent partitions pictorially using Young diagrams as follows. Denition 2. A Young diagram is a nite collection of boxes arranged in left-justied rows, with the row sizes weakly decreasing.1 The Young diagram associated to the partition = (1 , 2 , , l ) is the one that has l rows, and i boxes on the ith row. For instance, the Young diagrams corresponding to the partitions of 4 are

(4)

(3, 1)

(2, 2)

(2, 1, 1)

(1, 1, 1, 1)

Since there is a clear one-to-one correspondence between partitions and Young diagrams, we use the two terms interchangeably, and we will use Greek letters and to denote them. A Young tableau is obtained by lling the boxes of a Young diagram with numbers.
1 The notation used here is known as the English notation. Most Francophones, however, use the French notation, which is the upside-down form of the English notation. E.g. (3, 1) as .

Y. Z HAO YOUNG TABLEAUX AND THE R EPRESENTATIONS OF THE S YMMETRIC G ROUP 35 Denition 3. Suppose n. A (Young) tableau t of shape , is obtained by lling in the boxes of a Young diagram of with 1, 2, . . . , n, with each number occurring exactly once. In this case, we say that t is a -tableau. For instance, here are all the tableaux corresponding to the partition (2, 1): 1 2 3 2 1 3 1 3 2 3 1 2 2 3 1 3 2 1

Denition 4. A standard (Young) tableau is a Young tableaux whose the entries are increasing across each row and each column. The only standard tableaux for (2, 1) are 1 2 3 Here is another example of a standard tableau: 1 2 4 3 5 6 . 7 8 9 The denitions that we use here are taken from [Sa], however, other authors have different conventions. For instance, in [Fu], a Young tableau is a lling which is weakly increasing across each row and strictly increasing down each column, but may have repeated entries. We call such tableaux semistandard and we use them in Section 4.6. Before we move on, let us recall some basic facts about permutations. Every permutation Sn has a decomposition into disjoint cycles. For instance (123)(45) denotes the permutation that sends 1 2 3 1 and swaps 4 and 5 (if n > 5, then by convention the other elements are xed by ). The cycle type of is the partition whose parts are the lengths of the cycles in the decomposition. So (123)(45) S5 has cycle type (3, 2). It is a basic result that two elements of Sn are conjugates if and only if they have the same cycle type. The easiest way to see this is to consider conjugation as simply a relabeling of the elements when the permutation is written in cycle notation. Indeed, if = (a1 a2 . . . ak )(b1 b2 . . . bl ) , and sends x to x , then 1 = (a1 a2 . . . ak )(b1 b2 . . . bl ) . This means that the conjugacy classes of Sn are characterized by the cycle types, and thus they correspond to partitions of n, which are equivalent to Young diagrams of size n. Recall from representation theory that the number of irreducible representations of a nite group is equal to the number of its conjugacy classes. So our goal for the next two sections is to construct an irreducible representation of Sn corresponding to each Young diagram. and 1 3 . 2

4.3

Tabloids and the Permutation Module M

We would like to consider certain permutation representations of Sn . There is the obvious one: the permutation action of Sn on the elements {1, 2, . . . , n}, which extends to the dening representation. In this section, we construct other representations of Sn using equivalence classes of tableaux, known as tabloids. Denition 5. Two -tableaux t1 and t2 are row-equivalent, denoted t1 t2 , if the corresponding rows of the two tableaux contain the same elements. A tabloid of shape , or -tabloid is such an equivalence class, denoted by {t} = {t1 | t1 t} where t is a -tabloid. The tabloid {t} is drawn as the tableaux t without vertical bars separating the entries within each row.

36 For instance, if

T HE H ARVARD C OLLEGE M ATHEMATICS R EVIEW 2.2 t= 1 2 3

then {t} is the tabloid drawn as 1 3 2

which represents the equivalence class containing the following two tableaux: 1 2 3 2 1 3

The notation is suggestive as it emphasizes that the order of the entries within each row is irrelevant, so that each row may be shufed arbitrarily. For instance: 1 3 2 4 6 5 7 4 = 6 2 7 3 5 1 4 = 6 2 7 5 3 1 4 = 2 6 7 3 5 1

We want to dene a representation of Sn on a vector space whose basis is exactly the set of tabloids of a given shape. We need to nd a way for elements of Sn to act on the tabloids. We can do this in the most obvious manner, that is, by letting the permutations permutate the entries of the tabloid. For instance, the cycle (1 2 3) S3 acts on a tabloid by changing replacing its 1 by a 2, its 2 by a 3, and its 3 by a 1, as shown below: (1 2 3) 1 3 2 = 2 1 3

We should check that this action is well dened, that is, if t1 and t2 are row-equivalent, so that {t1 } = {t2 }, then the result of permutation should be the same, that is, {t1 } = {t2 }. This is clear, as simply gives the instruction of moving some number from one row to another. Now that we have dened a way for Sn to act on tabloids, we are ready to dene a representation of Sn . Recall that a representation of a group G on a complex vector space V is equivalent to extending V to a C[G]-module, so we often use the term module to describe representations. Denition 6. Suppose n. Let M denote the vector space whose basis is the set of -tabloids. Then M is a representation of Sn known as the permutation module corresponding to . Let us show a few example of permutation modules. We see that the M corresponding to the following Young diagrams are in fact familiar representations.

Example 7. Consider = (n). We see that M is the vector space generated by the single tabloid 1 2 n .

Since this tabloid is xed by Sn , we see that M (n) is the one-dimensional trivial representation. Example 8. Consider = (1n ) = (1, 1, . . . , 1). Then a -tabloid is simply a permutation of {1, 2, . . . , n} into n rows and Sn acts on the tabloids by acting on the corresponding permutation. n It follows that M (1 ) is isomorphic to the regular representation C[Sn ].

Y. Z HAO YOUNG TABLEAUX AND THE R EPRESENTATIONS OF THE S YMMETRIC G ROUP 37 Example 9. Consider = (n 1, 1). Let {ti } be the -tabloid with i on the second row. Then M has basis {t1 }, {t2 }, . . . , {tn }. Also, note that the action of Sn sends ti to t(i) . And so M (n1,1) is isomorphic to the dening representation C{1, 2, . . . , n}. For example, in the n = 4 case, the representation M (3,1) has the following basis: t1 = 2 1 3 4 , t2 = 1 2 3 4 , t3 = 1 3 2 4 , t4 = 1 4 2 3 .

Now we consider the dimension and characters of the representation M . First, we shall give a formula for the number of tabloids of each shape. Proposition 10. If = (1 , 2 , , l ), dim M = n! . 1 !2 ! l !

We leave the proof of this proposition to the readers. It is a simple combinatorial exercise of counting the number of -tabloids. Now we give a formula for the characters of M . Proposition 11. Suppose = (1 , . . . , l ), = (1 , . . . , m ) are partitions of n. The character l 1 2 of M evaluated at an element of Sn with cycle type is equal to the coefcient of x 1 x2 xl in m Y i i i (x 1 + x2 + + xl ).
i=1

To prove this formula, note that since M can be realized as a permutation representation on the -tabloids, its character at an element Sn is equal to the number of tabloids xed by . The rest of the proof consists of a simple generating function argument, which we leave to the readers. Note that Proposition 10 also follows as a corollary to the above result. Indeed, the dimension of a representation is simply the value of the character at the identity element, which has cycle type l 1 2 = (1n ). So Proposition 11 tells us that the dimension of M is the coefcient of x 1 x2 xl n n! in (x1 + + xn ) , which is equal to dim M = 1 !2 !l ! by the multinomial expansion formula. Example 12. Let us compute the full list of the characters of the permutation modules for S4 . The character at the identity element is equal to the dimension, and it can found through Proposition 10. For instance, the character of M (2,1,1) at e s4 is 4!/2! = 12. Say we want to compute the character of M (2,2) at the permutation (12), which has cycle type 2 (2, 1, 1). Using Proposition 11, we see that the character is equal to the coefcient of x2 1 x2 in 2 2 (x 2 + x )( x + x ) , which is 2. Other characters can be similarly computed, and the result is 1 2 1 2 shown in the following table. permutation cycle type M (4) M (3,1) M (2,2) M (2,1,1) M (1,1,1,1) e (1, 1, 1, 1) 1 4 6 12 24 (12) (2, 1, 1) 1 2 2 2 0 (12)(34) (2, 2) 1 0 2 0 0 (123) (3, 1) 1 1 0 0 0 (1234) (4) 1 0 0 0 0

Note that in the above example, we did not construct the character table for S4 , as all the M are in fact reducible with the exception of M (4) . In the next section, we take a step further and construct the irreducible representations of Sn .

38

T HE H ARVARD C OLLEGE M ATHEMATICS R EVIEW 2.2

4.4

Specht Modules

In the previous section, we constructed representations M of Sn known as permutation modules. In this section, we consider an irreducible subrepresentation of M that corresponds uniquely to . The group Sn acts on the set of Young tableaux in the obvious manner: for a tableaux t of size n and a permutation Sn , the tableaux t is the tableaux that puts the number (i) to the box where t puts i. For instance, 1 2 4 5 2 3 5 4 (1 2 3)(4 5) 3 6 = 1 6 . 7 7 Observe that a tabloid is xed by the permutations which only permute the entries of the rows among themselves. These permutations form a subgroup of Sn , which we call the row group. We can similarly dene the column group. Denition 13. For a tableau t of size n, the row group of t, denoted Rt , is the subgroup of Sn consisting of permutations which only permutes the elements within each row of t. Similarly, the column group Ct is the the subgroup of Sn consisting of permutations which only permutes the elements within each column of t. For instance, if t= 4 1 2 3 5 then Rt = S{1,2,4} S{3,5} , and Ct = S{3,4} S{1,5} S{2} .

Let us select certain elements from the space M that we use to to span a subspace. Denition 14. If t is a tableau, then the associated polytabloid is X et = sgn( ) {t}.
Ct

So we can nd et by summing all the tabloids that come from column-permutations of t, taking into account the sign of the column-permutation used. For instance, if t= 4 1 2, 3 5 then et = 4 3 1 5 2 3 4 1 5 2 4 3 5 1 2 + 3 4 5 1 2

Now, through the following technical lemma, we see that Sn acts on the set of polytabloids. Lemma 15. Let t be a tableau and be a permutation. Then et = et . Proof. First observe that Ct = Ct 1 , which can be viewed as a relabeling similar to the discussion at the end of Section 4.2. Then, we have X X et = sgn( ) {t} = sgn( ) {t}
Ct Ct 1

X
Ct

sgn(

) 1 {t} =

X
Ct

sgn( ) {t} = et .

Now we are ready to extract an irreducible subrepresentation from M .

Y. Z HAO YOUNG TABLEAUX AND THE R EPRESENTATIONS OF THE S YMMETRIC G ROUP 39 Denition 16. For any partition , the corresponding Specht module, denoted S , is the submodule of M spanned by the polytabloids et , where t is taken over all tableaux of shape . Again, let us look at a few examples. We see that the Specht modules corresponding to the following Young diagrams are familiar irreducible representations.

Example 17. Consider = (n). Then there is only one polytabloid, namely 1 2 n

Since this polytabloid is xed by Sn , we see that S (n) is the one-dimensional trivial representation. Example 18. Consider = (1n ) = (1, 1, . . . , 1). Let 1 2 t= . . . n Observe that et is a sum of all the -tabloids multiplied by the sign of permutation it took to get there. For any other -tableau t , we have either et = et if t is obtained from t through an even permutation, or et = et if t is obtained from t through an odd permutation. So S is a onedimensional representation. From Lemma 15 we have et = et = sgn( )et . From this we see n that S (1 ) is the sign representation. Example 19. Consider = (n 1, 1). Continuing the notation from Example 9 where we use {ti } to denote the -tabloid with i on the second row, we see that the polytabloids have the form {ti } {tj }. Indeed, the polytabloid constructed from the tableau i j a b

is equal to {ti } {tj }. Let us temporarily use ei to denote the tabloid {ti }. Then S is spanned by elements of the form ei ej , and it follows that S (n1,1) = {c1 e1 + c2 e2 + cn en | c1 + c2 + + cn = 0}. This is an irreducible representation known as the standard representation. The direct sum of the standard representation and the trivial representation gives the dening representation, that is, S (n1,1) S (n) = M (n1,1) . We know that the S3 has three irreducible representations: trivial, sign, and standard. These are exactly the ones described above. Furthermore, there are exactly three partitions of 3: (3), (1, 1, 1), (2, 1). So in this case, the irreducible representations are exactly the Specht modules. Amazingly, this is true in general. Theorem 20. The Specht modules S for of Sn over C. n form a complete list of irreducible representations

The proof may be found in [Sa]. Recall that at the end of Section 4.2 we noted that the number of irreducible representations of Sn equals the number of Young diagrams with n boxes. This Theorem gives a natural bijection between the two sets.

40

T HE H ARVARD C OLLEGE M ATHEMATICS R EVIEW 2.2

Note that the polytabloids are generally not independent. For instance, as we saw in Example n 18, any pair of polytabloids in S (1 ) are in fact linearly dependent. Since we know that S is spanned by the polytabloids, we may ask how to select a basis for vector space from the set of polytabloids. There is an elegant answer to this question: the set of polytabloids constructed from standard tableaux form a basis for S . Recall that a standard tableau is a tableau with increasing rows and increasing columns. Theorem 21. Let be any partition. The set {et : t is a standard -tableau} forms a basis for S as a vector space. The proof may be found in Sagan [Sa]. We only sketch an outline here. First, an ordering is imposed on tabloids. If some linear combination of et is zero, summed over some standard tableaux t, then by looking at a maximal tabloid in the sum, one can deduce that its coefcient must be zero and conclude that {et : t is a standard -tableau} is independent. Next, to prove that the set spans S , a procedure known as the straightening algorithm is used to write an arbitrary polytabloid as a linear combination of standard polytabloids. Now we look at some consequences of the result. Let f denote the number of standard tableaux. Then the following result follows immediately from Theorem 21. Corollary 22. Suppose n, then dim S = f . Let us end this section with a few results concerning f . Theorem 23. If n is a positive integer, then X
n

(f )2 = n!

where the sum is taken over all partitions of n. Proof. Recall from representation theory that the sum of the squares of the irreducible representation is equal to the order of the group. This theorem follows from that fact and Corollary 22. Theorem 23 also has an elegant combinatorial proof using the celebrated RSK correspondence. See [Fu] or [Sa] for details. Given the partition , the number dim S = f can be computed easily using the hook-length formula of Frame, Robinson, and Thrall, which we state now. Denition 24. Let be a Young diagram. For a square u in the diagram (denoted by u ), we dene the hook of u (or at u) to be the set of all squares directly to the right of u or directly below u, including u itself. The number of squares in the hook is called the hook-length of u (or at u), and is denoted by h (u). For example, consider the partition = (5, 5, 4, 2, 1). The gure on the left shows a typical hook, and the gure on the right shows all the hook-lengths. u Theorem 25 (Hook-length formula). Let 9 8 6 3 1 7 5 4 1 6 4 3 1 4 2 1 1

n be a Young diagram. Then n! . u h (u)

dim S = f = Q

Y. Z HAO YOUNG TABLEAUX AND THE R EPRESENTATIONS OF THE S YMMETRIC G ROUP 41 For instance, from the above example, we get dim S (5,5,4,2,1) = f (5,5,4,2,1) = 17! = 3403400. 9 8 7 62 5 43 32 2 15

For proof of the hook-length formula, see [Sa]. Finally, we state a formula for the characters of the representation S . Theorem 26 (Frobenius formula). Suppose = (1 , . . . , l ), = (1 , . . . , m ) are partitions of n. The character of S evaluated at an element of Sn with cycle type is equal to the coefcient 1 +l1 2 +l2 xl l in x2 of x 1 Y
1i<j l

(xi xj )

m Y

i i i (x 1 + x2 + + xl ).

i=1

See [FH] for proof. Observe the similarity between the statements of Proposition 10 and the hook-length formula, and also between Proposition 11 and the Frobenius formula. The hook-length formula can also be derived from the Frobenius formula by evaluating the character at the identity element. Again, see [FH] for details.

4.5

Young Lattice and Branching Rule

Now let us consider the relationships between the irreducible representations of Sn and those of Sn+1 . Consider the set of all Young diagrams. These diagrams can be partially ordered by inclusion. The resulting partially ordered set is known as Youngs lattice. We can represent Youngs lattice graphically as follows. Let denote that can be obtained by adding a single square to . At the nth level, all the Young diagrams with n boxes are drawn. In addition, to connected to if . Here is a gure showing the bottom portion of Youngs lattice (of course, it extends innitely upwards).

;
Now we consider the following question: given S a representation of Sn , how can we determine its restricted representation in Sn1 and its induced representation in Sn+1 ? There is a beautiful answer to this question, given by Youngs branching rule. Theorem 27 (Branching Rule). Suppose n, then M M Sn+1 ResSn1 S S and IndSn S = S . =
: :

42

T HE H ARVARD C OLLEGE M ATHEMATICS R EVIEW 2.2 For instance, if = (5, 4, 4, 2), so that = ,

then the diagrams that can be obtained by removing a square are

So

ResS14 S (5,4,4,2) = S (4,4,4,2) S (5,4,3,2) S (5,4,4,1) .

Similarly, the diagrams that can be obtained by adding a square are

So

(5,4,4,2) 16 IndS = S (6,4,4,2) S (5,5,4,2) S (5,4,4,3) S (5,4,4,2,1) . S15 S

The proof of Theorem 27 may be found in [Sa]. We shall only mention that the two parts of the branching rules are equivalent through the Frobenius reciprocity theorem. There is an interesting way to view this result. If we consider S only as a vector space, then the branching rule implies that M M M S. S S S = = = =
: : =(0) (1) (n) =

The nal sum is indexed over all upward paths from to in Youngs lattice. Since S is simply an one-dimensional vector space, it follows that we can construct a basis for S where each basis vector corresponds to a upward path in the Young lattice from to . However, observe that upward paths in the Young lattice from to correspond to standard -tableaux! Indeed, for each standard -tableaux, we can associate to it a path in the Young lattice constructed by adding the boxes in order as labeled in the standard tableaux. The reverse construction is similar. As an example, the following path in the Young lattice corresponds to the following standard tableau 1 2 4 . 3 5 So we have recovered a basis for S which turned out to be the same as the one found in Theorem 21. Now, one may object that this argument contains some circular reasoning, namely because the proof of the branching rule (as given in [Sa]) uses Theorem 21, that a basis of S can be found through standard tableaux. This is indeed the case. However, there is an alternative view on the subject, given recently by [VO], in which we start in an abstract algebraic setting with some generalized form of the Young lattice. Then, we can form a basis known as the Gelfand-Tsetlin basis by taking upward paths as we did above. We then specialize to the symmetric group and discover the standard tableaux. This means that the standard tableaux in some sense form a natural basis for S .

Y. Z HAO YOUNG TABLEAUX AND THE R EPRESENTATIONS OF THE S YMMETRIC G ROUP 43

4.6

Decomposition of M and Youngs Rule

First, we constructed the permutation modules M , and from it we extracted irreducible subrepresentations S , such that S forms a complete list of irreducible representations of Sn as varies over all partitions of n. Let us revisit M and ask, how does M decompose into irreducible representations. It turns out that M only contains the irreducible S if is, in some sense, greater than . To make this notation more precise, let us dene a partial order on partitions of n. (Note that this is not the same as the one used to dene Youngs lattice!) Denition 28. Suppose that = (1 , 2 , . . . , l ) and = (1 , 2 , . . . , n ) are partitions of n. Then dominates , written , if 1 + 2 + i 1 + 2 + + i for all i 1. If i > l (respectively, i > m), then we take i (respectively, i ) to be zero. In other words, if, for every k, the rst k rows of the Young diagram of contains more squares than that of . Intuitively, this means that diagram for is short and fat and the diagram for is long and skinny. For example, when n = 6, we have (3, 3) (2, 2, 1, 1). However, (3, 3) and (4, 1, 1) are incomparable, as neither dominates the other. The dominance relations for partitions of 6 is depicted using the following gure. Such diagrams are known as Hasse diagrams and are used to represent partially ordered sets. (6)

(5)

(3, 3)

?    

(4, 2)

_? ? ? ? ? ?    
(4, 1, 1)

_? ? ? ? ? ?    

(3, 2, 1)

_? ? ? ? ? ?    
(2, 2, 2)

(3, 1, 1, 1)

_? ? ? ? ? O

(2, 2, 1, 1)

(2, 1, 1, 1, 1)

(1, 1, 1, 1, 1, 1) Now we can precisely state what we wanted to say at the beginning of the section. Proposition 29. M contains S as a subrepresentation if and only if exactly one copy of S . . Also, M contains

We may ask how many copies of S is contained in M . It turns out that this answer has a nice combinatorial interpretation. In order to describe it, we need a few more denitions.

44

T HE H ARVARD C OLLEGE M ATHEMATICS R EVIEW 2.2

Denition 30. A semistandard tableau of shape is an array T obtained by lling in the boxes of with positive integers, repetitions allowed, and such that the rows weakly increase and the columns strictly increase. The content of T is the composition = (1 , 2 , . . . , m ), where i equals the number of is in T . For instance, the semistandard tableau shown below may be seen to have shape (4, 2, 1) and content (2, 2, 1, 0, 1, 1): 1 1 2 5 . 2 3 6 The number of semistandard tableau of a given type and content is known as the Kostka number. Denition 31. Suppose , n, the Kostka number K is the number of semistandard tableaux of shape and content . For instance, if = (3, 2) and = (2, 2, 1), then K = 2 since there are exactly two semistandard tableaux of shape and content : 1 1 2 2 3 and 1 1 3 . 2 2

We are almost ready to state the result, but let us rst make the following observation, whose proof we leave as a combinatorial exercise for the readers. Proposition 32. Suppose that , n. Then K = 0 if and only if

. Also, K = 1.

We are now ready to state the result about the decomposition of M into irreducible representations. M K S . Theorem 33 (Youngs Rule). M =

For instance, from the table above, we see that

(2,2,1) S (3,1,1) 2S (3,2) 2S (4,1) S (5) . M (2,2,1) =S

Note that Proposition 32 is a consequence of Youngs rule. We shall end with a couple of examples illustrating Youngs rule. Example 34. Note that K(n) = 1 as there is only one (n)-semistandard tableau of content , formed by lling in all the required entries in order. Then Youngs Rule implies that every M contains exactly one copy of the trivial representation S (n) (see Example 17). Example 35. Since a semistandard tableau with content (1n ) is just a standard tableau, we have L n K(1n ) = f (the number of standard -tableaux). So Youngs rule says that M (1 ) = f S. n But from Example 8 we saw that M (1 ) is simply the regular representation. By P taking the magnitude of the characters of both sides, we get another proof of the identity n! = n (f )2 that we saw in Theorem 23.

References
[Fu] W. Fulton: Young Tableaux: With Applications to Representation Theory and Geometry. Cambridge: Cambridge University Press 1997. [FH] W. Fulton and J. Harris: Representation Theory: A First Course. New York: Springer 1991 (GTM 129). [Sa] B. E. Sagan: The Symmetric Group: Representations, Combinatorial Algorithms, and Symmetric Functions. New York: Springer 2001 (GTM 203).

Y. Z HAO YOUNG TABLEAUX AND THE R EPRESENTATIONS OF THE S YMMETRIC G ROUP 45 [Se] J-P. Serre: Linear Representations of Finite Groups. New York: Springer 1977 (GTM 42). [VO] A. M. Vershik and A. Yu. Okounkov: A new approach to the representation theory of symmetric groups. II, Zapiski Nauchn. Semin. POMI 307 (2004), 5798. English transl.: J. Math. Sci. (New York) 131 #2 (2005), 54715494.