Lecture Notes in

Mathematics
Edited by A. Dold and B. Eckmann

682

G. D. James

The Representation Theory

of the Symmetric Groups

Springer-Verlag
Berlin Heidelberg New York 1978
Author
G. D. James
Sidney Sussex College
Cambridge CB2 3HU
Great Britain

AMS Subject Classifications (1970): 20C15, 20C20, 20C30

ISBN 3-540-08948-9 Springer-Verlag Berlin Heidelberg New York

ISBN 0-387-08948-9 Springer-Verlag New York Heidelberg Berlin

This work is subject to copyright. All rights are reserved, whether the whole
or part of the material is concerned, specifically those of translation, re-
printing, re-use of illustrations, broadcasting, reproduction by photocopying
machine or similar means, and storage in data banks. Under § 54 of the
German Copyright Law where copies are made for other than private use,
a fee is payable to the publisher, the amount of the fee to be determined by
agreement with the publisher.
© by Springer-Verlag Berlin Heidelberg 1978
Printed in Germany
Printing and binding: Beltz Offsetdruck, Hemsbach/Bergstr.
2141/3140-543210
 Pre::ace

The representation theory of the symmetric groups was first studied

by Frobenius and Schur, and then developed in a long series of papers
by Young. Althouqh a detailed study of Young's work would undoubtedly
Day dividends, anyone who has attemnted this will realize just how
difficult it is to read his papers. The author, for one, has never
undertaken this task, and so no reference will be found here to any of
Young's proofs, although it is probable that some of the techniques
presented are identical to his.
These notes are based on those given for a Part III course at
Cambridge in 1977, and include all the basic theorems in the subject,
as well as some material previous Iv unpUblished. Many of the results
are easier to explain a blackboard and chalk than with the type-
written word, since combinatorial arguments can often be best presented
to a student bv indicating the correct line, and leaving him to write
out a comnlete proof if he wishes. In many places of this book we have
nreceded a proof bv a worked example, on the nrinciple that the reader
will learn more easily bv translating for himself from the particular
to the general than by reading the sometimes unpleasant notation reqUired
for a full proof. However, the complete argument is always inclUded,
perhaps at the expense of supnlying details which the reader might find
quicker to check for himself. This is especially important when dealing
with one of the central theorems, as the Littlewood­Richardson
Rule, since many who read early proofs of this Rule find it difficult to
fill in the details (see [16] for a description of the problems encount-
ered) •
The aDproach adopted is characteristic­free, except in those
places, such as the construction of the character tables of symmetric
qrouus, where the results themselves deDend upon the ground field. The
reader who is not familiar with representation theory over arbitrary
fields must not be deterred by this; we believe, in fact, that the
ordinary reDresentation theory is easier to understand by looking initi-
ally at the more general situation. Nor should he be put off by the
thought that technical knowledge is reouired for characteristic­free
representation theory, since the symmetric groups enjoy special propert-
ies which make it possible for this book to be largely self­contained.
The most economical wav to learn the imnortant results without using any
general theorems from representation theory is to read sections 1­5,
10­11 (noting the remarks follOWing Example 17.17), then 15­21.
Many of the theorems rely on a certain bilinear form, and towards
 IV

the end we show that this bilinear form must have been known to Young,
by using it in a new construction of Young's Orthogonal Form. It is
remarkable that its significance in the representation theory of the
symmetric grouns was only recently recognized.
I wish to express mv thanks to Mrs. Robyn Brinqans for her careful
and 1"atient typing of my manuscrirt.

G. D. James
 Contents

1. Background from representation theory 1

2. The symmetric group • 5
3. Diagrams, tableaux and tabloids 8
4. Specht modules 13
5. Examples 18
6. The character table of Gn 22
7. The Garnir relations 27
8. The standard basis of the Specht module • 29
9. The Branching Theorem 34
10. p-regular partitions 36
11. The irreducible representations of G 39
n
12. Composition factors 42
13. Semi standard homomorphisms 44
14. Young's Rule 51
15. Sequences • 54
16. The Littlewood-Richardson Rule 60
17. A Specht series for 65
18. Hooks and skew-hooks 73
19. The Determinantal Form 74
20. The Hook Formula for dimensions • 77
21. The Murnaghan-Nakayama Rule 79
22. Binomial coefficients • 87
23. Some irreducible Specht modules 89
24. On the decomposition matrices of 98
25. Young's Orthogonal Form. • 114
26. Representations of the general linear group • • 125

Appendix. The decomposition matrices of the symmetric groups Gr

n
for the primes 2 and 3 with n < 13 136
References 153
Index • 155
1. BACKGROUND FROM REPRESENTATION THEORY

We shall assume that the reader is familiar with the concept of

tne group algebra, FG, of a finite group G over a field F, and with
the most elementary properties of (unital right-)FG-modules. It is
possible to prove all the important theorems in the representation
theory of the symmetric group using only the following:

1.1 THEOREM If M is an irreducible FG-module, then M is a composition

factor of the group algebra, FG.

Let m be a non-zero element of M. Then mFG is a non-zero sub-

module of M, and since M is irreducible, M = mFG. The map
0: r ... mr (r FG)
is easily seen to be an FG-homomorphism from FG onto M. By the first
isomorphism theorem,
FG/ker 0 ;: M
so FG has a top composition factor isomorphic to M.

The first isomorphism theorem will appear on many occasions,

because we shall work over an arbitrary field, when an FG-module can
be reducible but not decomposable.
We often use certain G-invariant bilinear forms, as in the proof
of a special case of Maschke's Theorem:

1.2 MASCHKE"S THEOREM If G is a finite group and F is a subfield of

the field of real numbers, then every FG-module is completely reducible.

Let el, ••• ,e

be an F-basis for our FG-module M. Then there
m
is a unique bilinear form on M such that
= 1 if i = j, and 0 if i j.
Now,a new bilinear form can be defined by
<u,v> = I (ug, v g) for all u,v in M.
gEG
This form is G-invariant, in the sense that
<u 9 ,v g > = <u,v> for all 9 in G.
Given a submodule U of M, v u 1 means, by definition, that <u,v>
o for every u in U. But if u U, then u g-l U. Thus
<u ,v g> <u 9 -1 ,v> = 0,
1
using the fact that our form is G-invariant. This shows that v g u ,
1
which is the condition required for u to be a submodule of M.
If u 0, then <u,u> 0, since F is a subfield of the field of
1
real numbers, so U nUl = O. We shall prove below that dim U + dim u
1
= dim M, and therefore u is an FG-module complementing U in M as
required.
We now remind the reader of some elementary algebra involving
 2

bilinear forms.
Let M be a finite-dimensional vector space over F. The dual of
M is the vector space of linear maps from Minto F, and will be denoted
by M*. Let el, ••• ,e be a basis of a subspace V, and extend to a basis
k
el, ••• ,e ofM. ForlSjsm, definec. 11 * bye. c. = l i f i = j , and
m J J
o if i j. By considering the action on el, ••• ,e m, we see that any
element ¢ of M* can be written uniquely as a linear combination of
cl' •••• ,c m' thus: ¢ (el¢)c l + ••• +(em¢)c m• Therefore, cl, ••• ,c m
is a basis of M* and
dim M = dim 11* •
Further, ¢ belongs to VO , the annihilator of V, if and only if
e = ••• = e = O. Therefore, ck+l, ••• ,c m spans VO and
l¢ k¢
dim V + dim VO = dim 11.
Suppose now that we have a symmetric bilinear form, < , >, on 11
which is non-singular (That is, for every non-zero m in 11 there is an
m' in 11 with <m,m'> 0). Define
8: M + 11 * by m + where
x + <m,x> (x c M).
We see that
M*, since < , > is linear in the second place, and
m
8 is a linear transformation, since < , > is linear in the first place.
Now, ker 8 = {m M I for all x M, <m,x> O}= 0, since the bilinear form
is non-singular. But dim M = d i.r.i 11*, so 8 is an isomorphism between
M and M*. Under this v L corresponds to Vo. Thus, for
every subspace V,

1.3 dim V + dim v L = dim M

LL,
Since V V this equation between dimensions gives
LL
V = V.
More generally, given subspaces 0 cUe V c M, we have vL c UL,
and we may define
dual of UL/V L by v +
8: V + where
L L
x + V + <v,x> (x U ) .
L L
If x + V = x' + v , then x - x' V L, and <v s x> - <v,x' > = <v,x-x' > O.
This shows that is well-defined. In the same way as before,
and 8 are linear, but now
ker8 = Iv Vi for all x UL, <v,x> = O} = V nULL •
LL
Since U = U V, ker 0 = u. We therefore have a monomorphism from
V/ker 8= V/U into the dual of UL/V L• Again, dimensions give:

1.4 vvhen 0 cUi;. V c 1";, V/U.!i! dual of u L/V L • In particular, V'"

dual of Ivl/VL•

If M is an FG-module for the group G, we can turn the dual space

 3

M* into an FG-module by letting

m(1jJg) = (rn (m EN, 1jJ E *, g E G) •

Notice that the inverse of 9 appears to ensure that 1jJ (g h) = (1jJg) h.

This means that the module M* (which we shall call the dual of M ) is
not in general FG-isomorphic to :1. Indeed, if T (g) is the matrix
representing 9 with respect to the basis el, ••• ,e of then T' (g-l)
m
is the matrix representing 9 with respect to the dual basis
of H*. This means that the character of H* is the complex conjugate
of the character of M when we are working over the complex numbers.
assume that the bilinear form < , > is G invariant. If U
and V are FG-submodules of N, then the isomorphisms in 1.4 are FG-
isomorphisms. To verify this, we must show tha t 0: v ... 1jJ is a G-
homomorphism. But (x + V1 ) 1jJ = <x,vg> -1
<xg ,v> = (xg
1
+ V )1jJv
-y
1 -1 1 vg
(x + V ) 9 1jJv = (x + V ) (1jJv g), and 1jJvg 1jJv g, as required.
For every pair of subspaces U and V of N, (U + V)l = U1
can easily be deduced from the definitions. Replacing U and
and v1 , we also find that tr' + v1 = (U n V) 1.
Throughout this book, the next picture will be useful:
M

/ '\'
V + v1

v"vov/ I
o
The second isomorphism theorem gives VI (V n V
1
) :: (V+ V )
1
IV 1 • But
1 1 1 1
(V + V ) IV :: dual of VI (V + V ) 1, by 1. 4 = dual of VI (V n V ) , so
1
1. 5 For every FG-submodule V of M, VI (V n V ) is a self-dual FG-
module.
Every irreducible representation of the symmetric group will turn
up in this fashion.
1
It is very ±mportant to notice that V n V can be non-zero for a
submodule V of M. How can we compute the dimension of V/(V n V 1 ) , given
a basis of V? The answer is simple in theory, but will require a lot
of calculation if V has large dimension. The Gram matrix, A, is def-
ined with respect to a basis el, ••• ,e of V by letting the (i,j)th
k
entry of A be <ei,e > .
j
1
1.6 THEOREM The dimension of VI (V n V ) equals to the rank of the
 4

Gram matrix with respect to a given basis of V.

Proof: As usual, map V .... dual of V by

8: v .... iJJ where uiJJ <v,u> (UEV)
v v
Let e , ... ,e be the given basis of V, and £1' ••• ' £k be the dual
l k
basis of V*. Since e iJJ
j ei
= <ei,e j> , we have

iJJ e. = <ei,e l> £l+···+<ei,e k> £k

Thus the Gram matrix for the basis el, ••• ,e k coincides with the
matrix of 8 taken with respect to the bases el, ••• ,e k of V and £1' •••
£k of V*. But, visibly, ker 8 = V n v", so dim vi (V n VJ.) = dim Im 8 =
the rank of the Gram matrix.
The only results from general representation theory which we shall
use without proof are those telling us how many inequivalent ordinary
and p-modular irreducible representations a finite group possesses,
and the following well-known result about representations of a finite
group over the field of complex numbers (cf. Curtis and Reiner ]
43.18 and Exercise 43.6).

1.7 Let S be an irreducible and M be any Then

the number of composition factors of M isomorphic to S equals
dim Hom{;G (S 1M) •

In fact, it turns out that these results are redundant in our

approach, and Theorem 1.1 gives everything we want, but it would be
foolish to postpone proofs until Theorem 1.1 can be applied.
Readers interested in character values will be familiar with the
Frobenius Reciprocity Theorem and the orthogonality relations for
characters, so we assume these results when discussing characters.
 5

2. THE SYMMETRIC GROUP

The proofs of the results stated in this section can be found in

any elementary book on group theory.
A function from {1,2, ••• ,n} onto itself is called a permutation
of n numbers, and the set of all permutations of n numbers, together
with the usual composition of functions, is the symmetric group of
degree n, which will be denoted by (; n • Note that G n is defined for
n :2= 0, and has n! elements (where 0: = 1). If X is a subset of
{1,2, ••• ,n}, we shall write Gx for the sUbgroup of Gn which fixes every
number outside X.
It is common practice to write a permutation 'JT as follows:
'JT = 2 3
27f 37f

By considering the orbits of the group generated by 7f , it is

simple to see that 7f can be written as a product of disjoint cycles,
as in the example :
2 3 4 5 6 7 8
9 -- (2 5 6 8) (1 3) (4 9) (7)
5 1 9 6 8 7 2 4)

We usually suppress the I-cycles when writing a permutation. For

example, if 7f interchanges the different numbers a,b and leaves the other
numbers fixed, then 'JT is called a transposition and is written as 7f
(a b).

All our maps will be written on the right1 in this way, we have
(1 2) (2 3) = (1 3 2). This point must be noted carefully, as some
mathematicians would interpret the product as (1 2 3).
Since (i l i 2 ••• i k) = (i i (i i ••• (i i any cycle, and hence
l 2) l 3) l k),
any permutation, can be written as a product of transpositions. Better
still,

2.1 The transpositions (x-l,x) with 1 < x <n generate

This is because, when a < b, we can conjugate (b-l,b) by (b-2,b-l)

(b-3,b-2) ••• (a,a+l) to obtain (a b).
If 'JT = 01 02 ••• 0j = '1 '2""k are two ways of writing 7f as a
product of transpositions, then it can be proved that j - k is even.
Hence there is a well-defined function
sgn: G n, -+ {±l}
such that sgn 7f = (-l)J if 7f is a product of j transpositions.

2.2 DEFINITION A = (A ••• ) is a partition of if A •••

l,A 2,A 3, l,A 2,A 3,
are non-negative integers, with A A :2=A '" ••• and. LIA = n ,
l:2= 2 3 i
 6

The permutation TI is said to have cycle-type A if the orbits of

the group generated by TI have lengths AI;' A ;, . . .
2 Thus, (2 5 6 8) (1 3)
(4 9) (7) has cycle-type (4,2,2,1,0,0, ••• ). Abbreviations such as the
following will usually be adopted:
(4,2,2,1,0,0,.,.) = (4,2,2,1) = (4,2 2 , 1 ) .
That is, we often suppress the zeros at the end of A, and indicate
repeated parts by an index.
Since two permutations are conjugate in n
if and only if the
permutations have the same cycle type,

2.3 The number of conjugacy classes of Gn equals the number of par-

titions of n.

Now, for any finite group G, the number of inequivalent irreducible

is equal to the number of conjugacy classes of G, so

2.4 The number of ordinary irreducible representations

of equals the number of partitions of n.

We should therefore aim to construct a representation of G n for

each partition of n. Let us look first at an easy example:

2.5 EXAMPLE There is a natural representation which arises directly

from the fact that G n permutes the numbers 1,2, ••• ,n ; take a vector
space over F of dimension n, with basis elements called I,2, ... ,n ,
and let <;; act on the space by
n
TI = (TI to G l ;
n
r
We shall denote rrr
this representation by M(n-l,l)
We can easily spot a submodule of M(n-l,l); the space U spanned
by I + 2 + ••• + n is a submodule on which Gn acts trivially. It is
not hard to find another submodule, but suppose we wish to eliminate
guesswork. If F = Q, the field of rational numbers, the promf of
Maschke's Theorem suggests we construct an G -invariant inner product
on M(n - l , l ) an d t'nen tr' 'II b e an ' , ncomp 1 ement to U•
< I, j > = 1 if i j and 0 if i ;t j (* )
de f ' an G" t' pro d uc t on M(n-l,l) • Th en

u
i
= 0: ai I I ai to Q a l + ••• + an = O}
Let s(n-l,l) = (2 -
• Then certainly S(n-l,l) is a sub-
I)Fg'
i
n
module of u , and it is easy to see that we have equality. Thus
M(n-l,l) = s(n-l,l) U when F =
Notice though, that (*) gives an G -invariant bilinear form on
(n-l,l) . , ( n - l 1) ."
M wnatever the S ' alwayp a submodule, too (It
is a complement to U if and only if char F tn.) s(n-l,l) is a Specht
module.

Are there any other easy ways of constructing representation

 7

r' (n-2 2)
modules for Consider the vector space H ' , o v e r F spanned
by unordered pairs ij (i j). M(n-2,2) has dimension and becomes
an F G -module if we de fine i j
n
7T i 7T , j 7T • This space should not be
difficult to handle, but it is not irreducible, since I { ij 11 5 i
< j 5 n } is a trivial submodule. We do not go into details for the
moment, but simply observe that M(n-2,2) supplies more scope for inves-
tigation.
More generally, we can work with the vector space M(n-m,m) spanned
by unordered m-tuples il ••• i (where i i unless j = k). Since
m j k
this space is isomorphic to that spanned by unordered (n-m)-tuples,
there is no loss in assuming that n-m m. This means that for every
partition of n with two non-zero parts we have a corresponding (redu-
cible) F G -module at our disposal.
n
Flushed with this success, we should go on and see what else we
(n -_ 2 , 1 2 )
can do. Let M be the space spanned by ordered pairs, which we
shall denote by (i j). The G action is 7T = 12. Let M(n-3,2,l)
1. n 1. i::.
be the space spanned by vectors consisting of an unordered 2-tuple
ij followed by a I-tuple no two of i,j and k are equal. These
vectors may be denoted by but it seems that we should change our
notation and have as a basis vector of M(n-3,2,l) in
i l·······in _ 3
i _ i _
n 2 n l
in
place of

By now, it should be clear how to construct an F G -module M

A for
n
each partition A of n. The notation we need to do this formally is
introduced in the next section. MA is reducible (unless A = (n», but
contains a Specht module SA, which it turns out, is irreducible if
char F = O.
 8

3. DIAGRAMS, TABLEAUX AND TADLOIDS

3.1 DEFINITIONS. If A is a partition of n, then the diagram [AJ is

{(i,j) I i,j E :E 1:; i 1:; j s Ai}
(Here, Z is the set of integers).
If (i,j) E OJ, then (i,j) is called a node of OJ. The k t h row (res-
pectively, column) of a diagram consists of those nodes whose first
(respectively, second) coordinate is k.

We shall draw diagrams as in the following example:

x x x x
OJ x x
x x
x

There is no universal convention about which way round diagrams

should be shown. Some mathematicians work with their first coordinate
axis to the right and the second one upwards: It is customary to drop
the inner brackets when giving examples of diagrams, so we write
[4 ,2 2 ,1 J, not [( 4 ,2 2 ,1) J •
The set of partitions of n is partially ordered by

3.2 DEFINITION. If A and are partitions of n, we say that A dom­

inates and write A provided that

for all j, A. ;,
i=l J. i=l J.

If A and A we write A

3.3 EXAMPLE. The dominance relation on the set of partitions of 6

is shown by the tree:
(6)

(5 1)

"
"(3 "
(3,3)./'
(4 2)

2 1)/
(4,1 2 )

/'
(3,1 3 )
" (2 3)

".(2 2,1 2 ) /

(2! 1
I
(1 6 )

The dominance order is certainly the "correct" order to use for

partitions, but it is sometimes useful to have a total order, >, on
the set of partitions. The one we use is given by

3.4 DEFINITION If A and are partitions of n, write A > if and

only if the least j for which Aj satisfies Aj > (Note that
 9

some authors write this relation as A < This is called the dictio-
nary order on partitions.

It is simple to verify that the total order > contains the partial
order in the sense that A implies A But the reverse
lication is false since

3.5 DEFINITION If [A] is a diagram, the conjugate diagram [A'] is

obtained by interchanging the rOHS and columns in [A]. A' is the par-
tition of n conjugate to A.

The only use of the total order> is to specify, say, the order in
which to take the rows of the character table of G n' Since there may
be more than one self­conjugate partition of n (e.g. (4,2,1 2) and (3 2,2)
are both self­conjugate partitions of 8), there is no "symmetrical"
way of totally ordering partitions, so that the order is reversed by
conjugates. It is interesting to see, though, that
A if and only if t> A'.
The next thing to define is a A­tableau. This can be defined as
a bijection from [A] to {1,2, ••• ,n}, but we prefer the less formal

3.6 DEFINITION A A­tableau is one of the nl arrays of integers

obtained by replacing each node in [A] by one of the integers 1,2, ••• ,n,
allowing no repeats.
For example,
1 2 4 5 and 457 3 are (4,3,1)­tableaux.
3 6 7 218
8 6
bn acts on the set of A­tableaux in the natural way; thus the
permutation (1 4 7 8 6) (2 5 3) sends the first of the tableaux above
to the second. (Of course, the definition of a tableau as a function
wins here. Given a tableau t and a permutation the compositions of
the functions t and gives the new tableau •
Every approach to the representation theory of depends upon a
form of the next result, which relates the dominance order on partitions
to a property of tableaux.

3.7 THE BASIC LEMMA Let A and be partitions of n,

and suppo$e that tl is a A­tableau and t 2 is a Suppose that
for every i the numbers from the ith row of t 2 belong to different
columns of tl' Then A

Imagine that we can place the numbers from the first row of
t in [A] such that no two numbers are in the same column. Then [A]
2
must have at least columns; that is Al Next insert the
 10

numbers from the second row of t 2 in different columns. To have space

to so this, we require Al + A2 + Continuing in this way, we
have A
3.8 DEFINITIONS If t is a tableau, its row-stabilizer, Rt, is the
subgroup of keeping the rows of t fixed setwise.
Le. R = {n Gnl for all i, i and Ln belong to the same row of t}
t
The column stapilizer Ct, of t is defined similarly.

For example, when t = 1 2 4 5 , R

t = (; { l 2 4 5} x
GO 6 7} x <:;;{8}
3 6 7
8
and jRtl = 4: 3: 1:
Note that Rt 7f 7f- 1R 7f
t
and Ct7f = 7f- 1C 7f
t
.
3.9 DEFINITION Define an equivalence relation on the set of A-
tableaux by t - t if and only if t = t for some 7f R The
l 2 l7f 2 tl
tabloid {t} containing t is the equivalence class of t under this
equivalence relation.

It is best to regard a tabloid as a"tableau with unordered row

entries". In examples, we shall denote {t} by drawing lines between
the rows of t. Thus
3'4"5" 24'5 I'45 '2'35 135 125 234 134 12"4 123
13 rs- '2'"""5 3"5""
are the different (3,2) -tabloids, and I32
s-r-
123
rs-
.
(;n acts on the set of A-tabloids by {t.}n = [ t.n } • This action

l} =
is well-defined, since {t {t } implies t tlo for some o in Rt l•
2 2
Then 11 - I 07f 11 - I Rt 111 = Rt 17f , so {t 1l } = {t {t 211}.
1 lO7f}
We totally order the A-tabloids by

3.10 DEFINITION {t l} < {tif and only if for some i

2}
(i) When j > i, j is in the same row of {tl} and {t
2}
(ii) i is in a higher row of {t than {t
l} 2}.
We have written the (3,2)-tabloids in this order, above. There
are many other sensible orderings of A-tabloids, but the chosen method
is sufficient for most of our purposes. As with the dominance order
on partitions, the best tabloid ordering is a partial one:

3.11 DEFINITION Given any tableau t, let mir(t) denote the number
of entries less than or to i in the first r rows of t. Then
write
{t l} s {t 2} if and only if for all i and r mir(t mi r(t 2).
l)
 11

This orders the tabloids of all shapes and sizes, but we shall
compare only tabloids associated with the same partition.
By considering the largest i, then the largest r, such that
mir(t l) < mi r(t2), it follows that

3.12 For A-tabloids {tl} and

3.13 EXAMPLES (i) If t
l
= 1 3 6 and t
2
= 1 2 4
257 356
4 7
then the first 7 rows and 3 columns of the matrices (mi r (t l) ) and
(m » are
i r(t 2
1 1 1 1 1 1
1 2 2 2 2 2
2 3 3 2 3 3
(mirt l) ) Z 3 4 (mir(t Z » 3 4 4
2 4 5 3 5 5
3 5 6 3 6 6
3 6 7 3 6 7

(ii) The tree below shows the relation on the (3,2)-tabloids:

/
I45 '2'"""'35
2 3 rr-
"'-I35/"'-"234
125/
34
-, T3"4;-
"-124/
"3""s-
I
123

Suppose that w < x and w is in the ath row and x is in the bth
row of t. Then the definition of mir(t) gives

3.14 mi r (t (wx) - mi r (t) =( 1 if b s r < a and ws i < x

-1 if a <; r < band w!: i < x
 12

l ° otherwise.
Therefore

3.15 {t} <I {t(wx)} if w < x and w is lower than x in t.

When we prove Young's Othogonal Form, we shall need to know that

the tabloids {t} and {t(x-l,x)} are immediately adjacent in the 4 order
(or are the same tabloid):

3.16 LEMMA If x-l is lower than x in t, and t is a A-tableau, then

there is no A-tableau t l with {t} {tl} 4 {t(x-l,x)}

Proof: First note that for any tableau t* with i* in the r*th row,
mi*r(t*) - mi*-l,r (t*) = the number of numbers equal to i* in the
first r rows of t* {Ol if r < r:
if r 2; r

Now suppose that x-l is lower than x in t, and {t} {t

l}
{t(x-l,x)}. By 3.14,

mir(t) = mir(t(x-l,x)) if i • x-l.

Therefore
mir(t l) = mir(t) if i • x-l
and
mir(t) - mi_1,r(t)mir(t l) - mi-l,r(t l) if i • x-lor x.
By the first paragraph of the proof, all the numbers except x-I
and x appear in the same place in t and t
l• But t and t l are both
A-tableaux. Therefore, {tl} = {t} or {t(x-l,x)} as required.
 13

4. SPECHT MODULES

With each partition of n, we associate a Young subgroup of

C; n by taking
Gu G{l,2, ••• }x G x

The study of representations of Gn starts with the permutation

module of G on C;. The Specht module is a submodule of
n u
and when the base field is Q (the field of rational numbers), the
different Specht modules, as varies over partitions of n, give all
the ordinary irreducible representations of G n '
4.1 DEFINITION Let F be an arbitrary field, and let be the vec-
tor space over F whose basis elements are the various

The action of G on tabloids has already been defined, by {t}Tr

n
{t1T} (1T EG ). £xtending this action to be linear on turns
n
into an FG ­module, and because G is transitive on tabloids, with
n n
G u stabilizing one tabloid,
4.2 is the permutation module of C;; on the subgroup (,;. is
n
a cyclic F C;;n­module, generated by anyone tabloid, and dim = n: /
J..l!.l ... ) .
4.3 DEFINITIONS Suppose that t is a tableau. Then the signed column
K , is the element of the group algebra F G obtained by summing
t n
the elements in the column stabilizer of t, attaching the signature
to each permutation. In short,
L (sgn 1T)1T
K
t =
1TEC
t
The poly tabloid, e t, associated with the tableau t is given by
e = {th
t t
The Specht module for the partition is the submodule of
spanned by poly tabloids.

A poly tabloid, it must be noted, depends on the tableau t, not

just the tabloid it}. All the tabloids involved in e have coefficient
t
± 1 (If v then v is a linear combination of tabloids; we say that
the tabloid it} is involved in v if its coefficient is non­zero.)

4.4 EXAMPLE If t = 251 then K

t
= (1­(23»(1­(45».
3 4
(We always denote the identity permutation by ll. Also
e t '= 25'l T5l 2'4l + "3"4T
n­ "2"'5""
The practical way of writing down e t, given t, is to permute the
 14

numbers in the columns of t in all possible ways, attaching the sig-

nature of the relevant permutation to each tableau obtained that way,
and then draw lines between the rows of each tableau.

Since KtTI =
t TI,
we have etTI = e t TI, so
TIK
4.5 is a cyclic module, generated by anyone poly tabloid.

It we wish to draw attention to the ground field F, we shall.write

and Many results for Specht modules work over an integral
domain, and it is only in Theorem 4.8 and Lemma 11.3 that we have
a field. When F is unspecified, then the ground field is arbitrary.
Since is a permutation module, it is hardly surprising that most of
its properties (for instance, its dimension) are independent of the
base field. What is more remarkable is that many results for the
Specht module are also independent of the field. Two special cases
are immediate. When = (n), = = the trivial FGn­module.
u = (In), is isomorphic to the regular representation of G and
n,
is the alternating representation (i.e. TI sgnTI).
We now use the basic combinatorial Lemma 3.7 to prove

4.6 LEMMA Let A and y be partitions of n. Suppose that t is a A-

tableau and t* is a y­tableau, and that {t*}K O. Then A y, and
t
if A = Y then {t * }K = ±{t}K t (= ±
t
Proof: Let a and b be two numbers in the same row of t*. Then
{ t *} (1­ (a b)) = {t; *} ­ {t, * (a bl} = o.
a and b cannot be in the same colunm of t, otherwise we could
select signed coset representatives 0l, .•• ,oK for the subgroup of the
column stabilizer of t consisting on 1 and (a,b) and obtain
Kt = ( 1- (a b)) (0 1 + ••• +0 K) •

It would then follow that {t*}K = 0, contradicting our hypothesis.

t
We have now proved that for every i, the numbers in the ith row
of t* belong to different columns of t, and Lemma 3.7 gives A
Also, if A = , then {t*} is one of the tabloids involved in {t}K
t,
by construction. Thus, in this case, {t*} = {t}TI for some permutation
TI in Ct, and {t*}K = {t}TI K = ±{t}K
t t t•
4.7 COROLLARY It u is an element of and t is a y­tableau, then
is a multiple of e
t•
Proof: u is a linear combination of {t*} and {t*}K is a
t
multiple of e t ' by the Lemma.
Now let < > be the unique bilinear form on for which
 15

Clearly, this is a symmetric, bn-invariant, non-singular bilinear

form on whatever the field. If the field is Q, then the form is an
inner product (cf. Example 2.5).
We shall often use the following trick;
For u, v E MfA, <UKt,V> I < (sgn TI) UTI, v>
TIEC t
-1
I
TIEC
<u, (s gn TI) VTI >
t
(since the form is G' -invariant.)
n
<u, (sgn TI) VTI>

The crucial result using our bilinear form is

4.8 THE SUBHODULE 'i'liEORLI,l (James [7J). If U is a sullL10dulc of

tDen either U J or U c

Proof: Suppose that u E U and t is a Then by Corollary 4.7,

UK = a multiple of e
t t•
If we can choose u and t so that this multiple is non-zero, then
e E U. Since is generated by e we have U .::.
t t,
If, for every u and t, UK
t
= 0, for all u and t
o = <UK , {t}> = <u,
t
That is, U c S

4.9 THEORLM n is zero or absolutely irreducible. Further

if tnis is non-zero, then n is the unique rClaximal submodule of
and n is self-dual.

Proof; By the Submodule Theorem, any submodule of is either

itself, or is contained in n Using 1.5, all parts of the
Theorem follow at once, except b1,-,t we have still to prove that
n remains irreducible when we extend the field.
Choose a basis el, ••• ,e for where each e is a polytabloid.
k i
(We shall see later how to do this in a special way.) By Theorem 1.6,
n is the rank of the Gram matrix with respect to this
basis. But the Gram matrix has entries from the prime subfield of F,
since the coefficients of tabloids involved in a polytabloid are all
± 1. Therefore, the rank of the Graw matrix is the same over F as over
the prime subfield, and so n does not increase in dimension if
we extend F. Since n ) is always irreducible, it follows
that it is absolutely irreducible.

Remark He shall s h ow that all the irreducible representations of G'n

turn up as I n the Theorem means that vre can work over lQ. or
 16

the field of p e1ements. We now concentrate on completing the case

where char F 0, although the remainder of this section also follows
from the more subtle approach in section 11. The reader impatient for
the more general result can go immediately to sections 10 and 11.
4.10 LEH!'IA If G is an F S , " "A.
f rom ,., t 0 ,.'1
Jl
ana's A .!.
,j.

n A
Ker G, then A V. If A - V, the restriction of G to S is multipli-
cation by a constant.

Remark Ker G c SAL by the Submodule Theorem, since Ker G sA. The
Lemma will later be improved in several ways (cf. 11.3 and 13.17) .

Proof: Suppose that t is a A-tableau. Since e t 4 Ker G,

o e
t
G = {t}K
t
G = {t}G K
t
(a linear combination of Jl-tabloids)K t.
By Lemma 4.6, A '" u , and if A = u , then e t 8 is a multiple of e t•

4.11 COROLLARY If char F = 0, and G is a non-zero element of

A Jl then A Jl. If A V, then 8 is multiplication by a
HomF 1.5
n
constant.

Proof: When F = Q, < , > is an inner product. The rank of the Gram
matrix with respect to a basis of SA therefore equals dim SA for any
field of characteristic O. Thus
when char F = 0, SAn SAL = 0 and !'II.. = SA @ SAL.
Any homomorphism defined on sX can therefore be extended to be
defined on MA by letting i t be zero on SAL. Now apply the Lemma.

4.12 THEOREM (THE ORDINARY IRREDUCIBLE REPRESENTATIONS OF b n ) . The

Specht modules over R are self-dual and absolutely irreducible, and
give all the ordinary irreducible representations of G n •
A u
Proof: If SQ ; SQ' then A Jl by Corollary 4.11. Similarly, Jl '" A
so A = Jl. Since n = 0, the Theorem follows from Theorem 4.9
and 2.4.

Since MJl is completely reducible when char F 0, Corollary 4.11

also gives

4.13 THEOREM If char F 0, the composition factors of MJl are SJl

(once) and some of {sA II.. Jl} (possibly with repeats).

Some authors prefer to work inside the group algebra of 6 , and

n
so we explain how to find a right ideal of the group algebra of G
n
corresponding to the Specht module.
Given a Jl-tableau t, let P t = I a, so that PtE FG'n' and let
aER
t
8: P
t 'IT ... {th ('IT E G'n)'
 17

This is clearly a well-defined F isomorphism from the right

ideal P F onto (It is well-defined, since Pt TI = Pt <=> TIE Rt
t
<=> {t}TI = {t}.) Restricting e to the right ideal P K t F gives an
t
isomorphism from P K F onto Using this isomorphism, every
t t
result can be interpreted in terms of the group algebra. We prefer
the Specht module approach for two reasons. First, the Specht module
depends only on the partition whereas the right ideal P K
t FG n
t
depends on the particular t. Perhaps more important is that
in place of P is a long sum of group elements, we have a single
t,
object {t}; this greatly simplifies manipulations with particular
examples, as will be seen in the next section, where we pause in the
development to work through some examples illustrating many salient
points.
 18

5. EXAMPLES
5.1 Reverting to the notation of Example 2.5, where the first
row of the tabloids in M(n-l,l) is omitted, we have
S (n-l ' 1) - - , \ ' " I ai"F,al+···+an-O
- }
s(n-l,l)L= Sp(I + 2 + ••• + n).
Clearly, s(n-l,l)l c s(n-l,l)if and only if char F divides n. By
the Submodule Theorem
(1 2) (1 2)
o c S (12)L = S cM if char F = 2 and n = 2
o c s(n-l,l)lc s(n-l,l) c M(n-l,l) if char F divides n > 2
are the unique composition series for M(n-l,l) if char F divides n.
The same Theorem shows that when char F does not divide n, S(n-l,l)
is irreducible and M(n-l,l) = s(n-l,l) s(n-l,l)L.
Note that in all cases s(n-l,l)l = s(n) and dim S(n-l,l) = n-l.

5.2 We examine M(3,2) in detail. A (3,2)-tabloid is

determined by the unordered pair of numbers ij which make up its second
row. To get a geometric picture of , consider the set of graphs
(:Nithout loops) on 5 points, where we allow an edge to be "weighted"
by a field coefficient. By identifying ij with the edge joining point
i to point j, we have constructed an isomorphic copy of M(3,2). For
example, 1
+ corresponds to

Any "quadrilateral with alternate edges weighted ± 1" is a gener-

ator for the Specht module s(3,2).
Let tl,t2,t3,t4,t5 1 3 5 1 2 5 1 3 4 1 2 4 1 2 3
2 4 3 4 2 5 3 5 4 5
respectively. Then etl, ••• ,et5 correspond to

D
t 1
1 1 1

S. W1 i1 5""
-t
" 'l. S
1
2 S" 2. S 0.- 2

-1 -1 1 -1 1 -1 , -1
I
4 3 4 1 3 4" :3 4" 3 40 ·3

respectively.
The 10 edges are ordered by 3.10:

The last edges involved in e , ••• ,et are

t1 5
 19

(which correspond to {tl}, ••• ,{t Since these last edges are
S}.)
different, etl, ••• ,ets are linearly independent. Note that it is far
from clear that they also span the Specht module, but we shall prove
this later. Assuming that they do give a basis, the Gram matrix with
respect to this basis is

4 2 2 1 -1
2 4 1 2 1
A 2 1 4 2 1
1 2 2 4 2
-1 1 1 2 4

One checks that i f char F 0 or char F 2: S, rank A S

if char F 3, rank A 1
if char F 2, rank A 4.
Therefore, dim(s(3,2) /s(3,2)n s(3,2)L) = S unless char F 2 or 3,
when the dimension is 4 or 1, respectively.
Let us find S(3,2)L. Certainly,

f = and S graphs like I' (-1)

are orthogonal to "quadrilaterals with alternate edges weighted ± I".

(An unlabelled edge is assumed to have weight 1). That is, they belong
to s(3,2)L. (f(-i) is defined by f(-i) = f(-l)(l i) for 1,:;, i,:;, 5.)
Now, f(-l) + f(-2) +••• + f(-S) = 3f. It is easy to verify that
f(-l) , ••• ,f(-S) are linearly independent if char F 3, and that they
span a space of dimension 4 when char F = 3. Hence
s(3,2)L is spanned by f, f(-1),f(-2), ••• , f(-S)
since s(3,2)L has dimension S (by 1.3).
When char F = 2, et2 + et + et + et = f. Therefore, I' E S (3,2)
3 4
n S(3,2)L in this case, and by dimensions ft spans s(3,2) n s(3,2)L.
When char F 3, e t + et = f(-S), and now f(-l) , ••• span
S(3,2) n s(3,2)L • 1 2

We do not yet have a convenient way of checking whether or not a

graph belongs to s(3,2). However, every such graph certainly satisfies
the two conditions:

5.3 (i) The sum of the coefficients of the edges is zero.

(ii) The valency of each point is zero. (Formally: the sum of the
coefficients of the edges at each point is zero.)
 20

These conditions hold because a generator for s(3,2) satisfies

the conditions. In fact, the properties characterize S(3,2) and enable
us rapidly to check that rES (3,2) when char F = 2 (I' has an even number
. (3 2)
of edges, and each p01nt has even valency), and that r(-S)E S '
when char F = 3 (r(-S) has 6 edges and each point has valency 0 or 3).
So far, we have highlighted two problems to be discussed later:

(a) Find a basis for the general Specht module like that given
above. (N.B. It is not obvious even that dim is independent of the
field. )
(b) Find conditions similar to 5.3 characterizing the Specht
module as a submodule of the second expression for s(n-l,l) in
Example 5.1).

We have proved that etl, ••• ,etS are linearly independent; here, as
in the general case, it is a lot harder to prove that they span s(3,2)
This example is concluded by a simultaneous proof that et , ••• ,et
form a basis of s(3,2) and that conditions 5.3 s(3,2)S.
Define ljIo E Hom (H(3,2) M(S)) and ljI E HOIlL (M(3,2) M(4,1))
FG'S ' 1 '
by
ljI o •• abc d e

abc e + abc d (i.e. de d + e)

d e

Now, conditions S.3(i) or (ii) hold for an element v of H(3,2)if

and only if v E Ker ljI or v E Ker ljIl' repectively. Therefore
(3,2) 0
S Ker ljIo n Ker ljIl (cf. Lemma 4.10), and we want to prove equality.
Write s(3,1), (3,2) for the space spanned by graphs of the form

i j - i k

j k

Now, S (3,1) , (3,2) Ker ljIo and ljIl sends s(3,1),(3,2) onto s(4,1)
(since ljIl: ij - ik i+3"-I-:k = j - k). Therefore, we have the
following series for M(3,2) :
 21

Dimensions
H(3,2)

I
Ker '4J o
} -
S (5) 1

I '" 0

}
S (3,1) , (3,2)
S (4,1)
I
S (3,1) , (3,2)
- 4 (see Example 5.1)

I n
Ker '4J l
'" 0

}
S(3,2)

I
S (3,2)
- '" 5

But dim H(3,2) = 10, so we have equality in all possible places.

In particular, dim s(3,2) = 5 and S(3,2) = Ker '4J
o
n Ker '4J
l
, as we wished
to prove.

5.4 S(2,2) is spanned by the graphs

'1 12-

2-
-1 -1
Zf -1

4- 3 4 j :3

Clearly, the first two form a basis.

When char F = 2, s(2,2) The reason underlying this is
that any poly tabloid contains none or both edges of the following pairs
of edges:
z.

4 4 ...- - - - _ . 3
 22

6. THE CHARACTER TABLE OF Gn

There are many ways of evaluating the ordinary irreducible chara-
cters of 0n. If the character table of Gn­ 1 is known, the Branching
Theorem (section 9) is very useful, but to calculate the character
table of this way ue have to work out all the earlier tables. On
the other hand, if just a few entries are required, the Hurnaghan-
Nakayama Rule (section 21) is the most efficient method, but i t is
hard to use a computer on this formula. The method given here finds
all the entries in the character table of simultaneously. It is
due to R.F.Fox, with some simnlifications by G.Mullineux.
A
Let X denote the ordinary irreducible character of G.n corres-
ponding to the partition A ­ that is, the character of the QG module
n
Let IG denote the trivial character of a group G. Recall that
G'A is a Young subgroup, and that 1 GAt is the character of
by 4.2 (The notation tG means"induced up to G" and +G neans "restricted
to G".)
All the matrices in this section will have rows and columns indexed
by partitions of n, in dictionary order (3.4). Since has as
a composition factor once, and t.ue other factors correspond to parti-
tions with > A (Theoren 4.13) ,

6.1 Tile matrix In = given by = the character inner product

..Q.6 A .!. Sn' ) is lower trian;;ular vlith l's dOvln the diagonaL

(see the example for G below) • It follows at once that the matrix
S'
B = given by

is upper triangular.
Let denote the conjugacy class of Gn corresponding to the
partition and let A = be tile matrix given by
= IGA n
The matrix A is not hard to calculate, and we claim that once i t
is known, the character table C = (c , ) of G: can be calculated by
n
straightforward matrix manipulations. First note that

L
u
= I GvI (X A+ Gv,l G) = b Av'

Therefore, B = CA', where A' is the transpose of A.

I GAil C;v1
But, \'
L (1 GAt On' 1 coo v t G n )
u
I G A I I G vi (1 GAt G n + G v' 1 G: )
v
I G AI L (1 5 A t Gn evaluated on an element of type
u
 23

u) • I sv n &fll

L (n: / I Gfll) I «, n Gv n
u
L (n ; / I G'fll) a
Afl
a
vfl
u
If A is known, v,e can solve these equations by starting at the
top left hand corner of D, working down each column in turn, and pro-
ceeding to the next column on the right. Since B is upper triangular,
there is only one unknown to be calculated at each stage, and this can
be found, since B has non­negative entries. Therefore

6.2 THEOREt1 If the matrix A = (a,

I\fl
), where a,
I\ll
= I G',I \n -&- lIl -
is
-
known, then we can find the unique non­negative upper triangular matrix
B = (b
Afl
) satisfying the equations
L buA b uv = L(n ; / I G u I) a AII a v fl
\l \l
and the character table C of Gn is given by C BA' -1.

6.3 Suppose n = 5. Then

(5) (4,1) (3,2) (2 2,1)

(5) 24 30 20 15
(4,1) 6 o 3
.../ (3,2) 2 3
A (3,1 2 ) o
(2 2,1) 1
3
(2,1 )
5
(1 )

(4,1) (3,2) 2,1)

(5) (2
(5) 120 24 12 4
(4,1) 24 12 8
(3,2) 12 8
B (3,1 2 ) 4
2,1)
(2 4
(2,1 3 )
(1 5 )
 24

(5) (4,1) (3,2) (3,1 2 ) (2 2,1) (2,1 3

)

(5) 1 1 1 1 1 1
(4,1) -1 o -1 1 o 2
(3,2) o -1 1 -1 1 1
C (3,1 2 ) 1 o o o -2 o
(2 2,1) o 1 -1 -1 1 -1
(2,1 3
) -1 o 1 1 o -2
(1 5 ) 1 -1 -1 1 1 -1

The columns of the character table are in the reverse order to the
usual one - in particular, the degrees of the irreducible characters
appear down the last column - because we have chosen to take the dic-
tionary order on both the rows and the columns.

6.4 NOTATION Equations like [3J[2J = [5J + [4,lJ + [3,2J are to be

interpreted as saying that has composition factors isomorphic
to and In general if A is a partition of n,

[A 1J[A 2J[A 3J··· I

u
means that has as a factor with multiplicity (m =
is the matrix defined in 6.1).
By dividing each column of the matrix B by the number at the top
of that column (which equals I I) , and transposing, the matrix m is
obtained. In the above example,

[5J [4,lJ [3,2J [3,12J [2 2,lJ [2,1 3J [15 J

[5J 1
[4][lJ 1 1
[3J[2J 1 1 1
m [3J[1J2 1 2 1 1
[2J2[1J 1 2 2 1 1
[2J[1J3 1 3 3 3 2 1
[1]5 1 4 5 6 5 4 1
Notice that the results [ 4][lJ = [5J + [4,1] and [3J[2J = [5J +
[4,lJ + [3,2J are in agreement with Examples 5.1 and 5.2. Young's
Rule in section 14 shows how to evaluate the matrix m directly.
Theorem 6.2 has the interesting

6.5 COROLLARY The determinant of the character table of c>'n is the

product of all the parts of all the partitions of n.

!:E22!: a AA = lj: (Ai - 1): and bAA = I G'A I = lj: Ai:

Since A and B are upper triangular and B = CA', we have
 25

det C = n II A, , as claimed.
A
i J.
Recall that the partition A' conjugate to A is by "turning
A on its side" (see definition 3.5). The character table of G 5 in
Example 6.3 exhibits the property:
A' A (In)
6.6 X = X 0 X

We prove this in general by showing

6.7 THEOREM A "'" S(ln)"

Q A' •
J.S J.somorp h'J.C to the dual of Sa
A'
Remark Since Sa is self-dual, we may omit the words "the dual of"
from the statement of the Theorem, but we shall later prove the ana-
logous Theorem over an arbitrary field, where the distinction between
SA' and its dual must be made.

Proof: Let t be a given A-tableau, and let t' be the corresponding A'
tableau.
e.g. if t = 1 2 3 then t '= 1 4
4 5 2 5
3
Let P
t,
=
Rt,} and Kt'L{1I!1I
L{ (sgn 11)11111 E Ct,}, as usual.
E

Let u be a generator for so that U1l = (sgn 11) u when 11 E G'n.

It is routine to verify that there is a well-defined iQ. G -epi-
AI A (In) n
morphism 0 from MQ onto SiQ. 0 S @.. sending {t'} to ({t} 0 u)P
t,;
o is given by.
6.8 0: {t ' 1l } .... ({t} 0 u)P , 1I
t

<{t}K ,{t}p
t t>
<{t}K t ,IRtl{t}> = IRtl.

Since,lRtl is a non-zero element of {t'}K 0 O. Thus

t•
Ker 0 1 SaA' ' and, by the Submodule Theorem, Ker 0 A' L Therefore,
A A' A' A'L A'
dim Sa = dim 1m 0 = 0) dim(MQ /SQ ) = dim Sa (*).
AI A" A A A'
Similarly, dim Sa dim SQ dim SQ. Therefore, dim SQ = dim SQ
and we have equality in (*). Thus, Ker0= The theorem is now
proved, since we have constructed an isomorphism between MA'/SA'L (:
A' A (In) Q. IQ
dual of Stl,l , by 1. 4) and SQ 0 S Q . .
A'
Remark Corollary 8.5 will give dim SA dim S , trivially, but this
shortens the proof by only one line.
There is one non-trivial character of G which can always be
n
evaluated qUickly, namely x(n-l,l):
 26

(ri-vL 1)
6.9 LEM1A The value of X ' on a permutation is one less than
the number of fixed points of
(n-l 1)
Proof: The trace of acting on the permutation module M "
is clearly the number of fixed points Since
M(n-l,l) tx S (n) E& S (n-l,l)
Q. - Q !Q.
(cf. Example 5.1), the result follows at once.
(n) (n-l 1) (In)
We can thus write down four characters, X ,X ',X and
(2 , l n - 2 ) ( 1 1) (In)
X (= X n-, @ X ) of G at once. The best way of finding
n
the character table of for small n is to deduce the remaining
characters from these, using the column orthogonality relations.
 27

7. THE GARNIR RELATIONS

For this section, let t be a given We want to find

elements of the group algebra of {;n which annihilate the given poly-
tabloid e t•
Let X be a subset of the ith column of t, and Y be a subset of the
(i + l)th column of t.

Let 0 1 , ••• , ok be coset representatives for GX x by in b x uY '

K
and let Gx Y =.L (sgn 0j)Oj. Gx,Y is called a Garnir element. (Garnir
, J=l
[5]) •
In all applications, X will be taken at the end of the ith column
of t and Y will be at the beginning of the (i+l)th column. The permu-
tations 0l, ••• ,ok are, of course, not unique, but for practical pur-
poses note that we may take 0l, ••• ,okso that to
l,t0 2
, ••• ,to k are all
the tableaux which agree with t except in the positions occupied by
X u Y, and whose entries increase vertically downwards in the posi tions
occupied by X u Y.

7.1 EXAMPLE if t 1 2 , X {4,5} and Y {2 , 3} then to 1 ' ••• , tO k

4 3
5

may be taken as
t = t 1 2 t 1 2 t 1 2 t 1 3 t 1 3 t 1 4
l 2 3 4 5 6
4 3 3 4 3 5 2 4 2 5 2 5
5 5 4 5 4 3

when sgn o. = 1 for i 1,3,4,6, sgn o. = ­1 for i = 2,5 and Gx,Y

1 ­ (3 4) + (3 5 4) + (2 3 4) ­
(2 3 5 4) + (2 4) (3 5).

7.2 THEOREM If IX u YI > u • then e t..9.x , Y (for any base field).

Proof: (See Peel [19J) v,rite G'; for L{Ls qn 0)010 G'x x G'y}

and G for L{(sgn 0)0!0 G'xuY}

,
Since I Xu Y I > for every T in the column stabiliz e r of t, some
pair of numbers in X u Yare in the same row of t r , Hence, in the usual
way, {t.r I = O. Therefore, {t}Kt G' = O.
Now, is a factor of K
t , and Gx,Y
 28

Therefore
o = {th t = [x] ; Iyl :{t}K t GX,y
Thus, {t}K
t
GX,y = 0 when the base field is Q, and since all the
tabloid coefficients here are integers, the same holds over any field.

7.3 EXAMPLE Referring to Example 7.1, we have

so
 29

8. THE STM,DARD BASIS OF THE SPECHT MODULE

8.1 DEFINITIONS t is a standard tableau if the numbers increase along

the rows and down the columns of t. {t} is a standard tabloid if there
is a standard tableau in the equivalence class {t}. e t is a standard
poly tabloid if t is standard.
In Example 5.2, the 5 standard (3,2)-tableaux and the corresponding
standard poly tabloids are listed.
A standard tabloid contains a unique standard tableau, since the
numbers have to increase along the rows of a standard tableau. It is
annoying that a poly tabloid may involve more than one standard tabloid
(In Example 5. 2, et involves and
5
We prove that the standard polytabloids form a basis for the Specht
module, defined over any field.
The have been totally ordered by definition 3.10. The
linear independence of the standard poly tabloids follows from the tri-
vial
8.2 LEMMA
Iti} is the
different,

l} < {t; 2} < ••• < {tmL

Proof: We may assume that {t I f a 1v 1 + ••• + amvm
= 0 (a i E F) and a j + l = ••• = am = 0, then a.J = 0, since {t.} J
is invol-
ved in v and in no v with k < j. Therefore, a l = ••• = am = O.
j k
It is clear that {t} is the last tabloid involved in e
when t
t
is standard, and this is all we need to deduce that the standard poly-
tabloids are linearly independent, but we go for a stronger result,
using the partial order (3.11) on tabloids:

8.3 LEMMA If t has numbers increasing down columns, then all the
tabloids {t'} involved in e satisfy {t'} 4 {t}.
t
Proof: If t' = tn with n a non-identity element of the column stabil-
izer of t, then in some column of t' there are numbers w < x with w
lower than x. Thus, by 3.15, {t'} {t' (wx)}. Since {t' is
involved in e induction shows that {t' (w x)} 9 {t.}, Therefore, {t'}
t,
4 {t}.
8.4 THEOREM ietlt is a standard v-tableau} is a basis for

Proof: (See Peel [19]) We have already proved that the standard
poly tabloids are linearly independent, and we now use the Garnir rela-
tions to prove that any poly tabloid can be written as a linear combi-
nation of standard polytabloids - a glance at Example 7.3 should show
the reader how to do this.
 30

First we write [t] for the column equivalence class of t; that is

[t] = {tl!t = t1T for smile 1T E C The column equivalence classes are
l t}.
totally ordered in a way similar to the order 3.10 on the row equiva-
lence classes.
Suppose that t is not standard. By induction, vie may assume that
e can be written as a linear of standard poly tabloids
t,
wheri [t I] < [t] and prove the s arie result for e Since e t 1T = (sgn 1T) e t
t•
when 1T E C we may that the entries in t are in increasing order
t,
down columns. Unless t is standard, some adjacent pair of columns, say
the jth and (j+l) th columns, have entries a <a < ••• < a b <b <••• <
l 2 r, l 2
b", w i, th a > b for some q
q q

b
1\ S
a
r

Let X = {aq, •••

,a } and Y = {bl, ••• ,b } and consider tne corresponding
r q
Garnir element Gx,Y = L(sgn 0)0, say. By Theoren 7.2

o = e t L(sgn a)a = L(sgn a)c t a •

Because b <••• < b <a < ••• < a , [to] < [t] for a L Since
l q q r
et = -L (sgn a)e
ta
' the result follows from our induction hypothesis.

8.5 COROLLARY The dimension of the Specht module is independent

of'the ground field, and equals the number of standard p­tableaux.

Remark An independent proof of Theorem 8.4 is given in section 17.

8.6 COROLLARY any poly tabloid can be written as an integral

linear combination of standard poly tabloids.

Proof: This result comes from the proof of Theorem 8.4; alternatively,
see 8.9 below.

8.7 COROLLARY The matrices representing S'n over Q with respect to

the standard basis of all have integer coefficients.

Proof: e t1T = e t 1T• Now apply Corollary 8.6.

8.8 COROLLARY If v is a non­zero element of then every last

tabloid (in the partial order on tabloids) involved in v is standard.

Proof: Since v is a linear combination of standard poly tabloids, the

result follows from Lemma 8.3.
 31

8.9 COROLLARY !!.....LE and the coefficients of the tabloids

involved in v are all integers, then v is an integral linear combina-
tion of standard poly tabloids.

Proof: We may assume that v is non-zero. Let {t} be the last (in
the < order) tabloid involved in v-, with coefficient a s :1':, say. By
the last corollary, {t} is standard. Now Lemma 8.3 shows that the
last tabloid in v - a e is before {t}, so by induction v - a e t is
t
an integral linear combination of standard poly tabloids. Therefore,
the same is true of v.

8.10 COROLLARY If v E and the coefficients of the tabloids invol-

ved in v are all integers, then we may reduce all these integers modulo
p and obtain an element where F is the field of p elements.
Proof: By the last Corollary, v is an integral linear combination of
standard poly tabloids, v = Ia e say (a i E Reducing modulo pall
i i,
the tabloid coefficients in v, we obtain v, say. -
Let a i a i modulo
p, The equation v= Ia i e i shows that v E .
Remark If we knew only that the standard poly tabloids span the
proof of Corollary 8.10 shows that any poly tabloid can be written as
a linear combination of standard poly tabloids over any field. There-
fore, we can deduce that the standard poly tabloids span over any
field, knowing only the same information over Q.

8.11 COROLLARY If F is the field of p elements, then is the

p-modular representation of t;; n obtained from
Proof: Apply the last Corollary.

8.12 COROLLARY There is a basis of all of whose elements inv-

olve a unique standard tabloid.

Proof: Let {t l} < {t 2} < ••• be the standard {t is

l}
the only standard tabloid involved in etl by Lemma 8.3. et may
2
involve {t l}, with coefficient a, say. Replace et2 by ft = et - ae
2 2 t l•
Then {t 2} is the only standard tabloid involved in f t . Continuing
2
in this fashion, we construct the desired basis.

Corollary 8.12 is useful in numerical calculations.

8.13 EXAMPLE Taking etl, ••• ,e
as in Example 5.2, each involves
ts
_just one standard tabloid, except et 5 which involves as well as
4 s . Replace etS by ft s = etl + et Then etl,et2,et3,et4,fts involve
S•
respectively with coefficient 1, and no other
standard tabloids.
 32

Consider the following vector

v belongs to s(3,2), since the sum of the edge coefficients is

zero, and each point has valency zero (cf. 5.3). But v involves
Therefore

v -etl - et2 - et3 - et4 + 3fts

2etl - et2 - et3 - et4 + 3etS'

Next we want the rather technical

8.14 LEMMA Suppose that 0EHomQ. G' and that all the tabloids
involved in {t}0 have integer ({t} E Then, reducing
all these integers modulo p« we obtain an element e of HomF G
where F is the field of p elements. If ker 0 = n
Ker e .2
Proof: It is trivial that 0 E G •
n
Take a basis f ... ,fir of and ex-tend
by the standard basis
l, " 'bL A
of SQ to obtain a basis fl, ••• ,fm of MQ• Let {tl}, •••• ,{tm} be the
different A-tabloids. Define the matrix N = (n by
i j)
n i j = < fi,{t j} >

We may assume that N has integer entries, and by row reducing the
first k rows, we may assume that the first k rows of N (which corres-
pond to the basis of are linearly independent modulo p. Reducing
all the entries in N modulo p, we obtain a set of vectors in the
last m - k of which are the standard basis of and the first k of
which are linearly independent and orthogonal to the standard basis of
s;. Since
dim = dim M 1._ dim S;
F
= k ,

we have constructed a basis of whose elements give a basis of SAL

F
when the tabloid coefficients are reduced modulo p.
AL
Now, anyone of our basis elements of SQ is an integral linear
combination of A-tabloids, and is sent to zero by 0. Therefore, when
all integers are reduced modulo p, e certainly sends the basis of
to zero, as required.
 33

We can now complement Theorem 6.7 by proving

8.15 THEOREM Over any field, SA @ S(ln) is isomorphic to the dual

of S A' •

Proof: It is sufficient to consider the case where the ground field

is F, the field of p elements, since we have proved the result when
F =
In the proof of Theorem 6.7, we gave a Q G -homomorphism e from
A' n
MQA' a.n
. t
0
MA _ s(ln) and proved that Ker e = Sa
Q 'Ct Q J. Using the Lemma above,
e, defined by
e: {t'n} + (sgn n) {tn}Ktn u

is an F -homomorphism onto SF@SF(ln)WhosekernelcontainssFA'J..

n _ A'
By dimensions, Ker e = SF J., and the result follows.
 34

9. THE BRANCHING THEOREM

The Branching Theorem tells us how to restrict an ordinary irre-

ducible representation from Gn to Gn ­ l • We have introduced the
symbols .j. b n ­ for restriction to G n­sL and t G'n+l for inducing to G'n+l'
l
Using notation like that in 6.4, we have

9.1 EXAMPLE [4,2 2,lJt G 2,lJ 2]

a [3,2 2,1 + [4,2,12J + [4,2 3 2
2
[4,2 ,1 J t <5 10= [5,2 J + [4,3,2,1 J + [4,2 J + [4,2 2,1 J

These are special cases of

9.2 THE BRANCHING THEOREM

(i) {SA IDJ is a diagram obtained by adding a

.!­.Gn+l '" Ql Q.
node to [\.IJ }.
u
(ii) ­ Ql {sA
lQ.
I DJ is a diagram obtained by taking a
node away from [lJJ }.

Proof: The two parts of the Theorem are equivalent, by the Frobenius
Reciprocity Theorem. Part (ii) follows from the more general:

9.3 THEOREM ""hen s\.I is defined over an arbitrary field, S\.I.j. G _

n l
has a series with each factor isomorphic to a Specht module for n­l'
The factors occurring are those given by part (ii) of the Branching
i
Theorem, and sA occurs above sAj in the series if Ai Aj •

Proof: (See Peel [19]) Let r <r

be the integers such < ••• < r
m l 2
that a node can be removed from the r.th row of [\.IJ to leave a diagram
.
(e.g. when [\.IJ = [4,2 2,1], r = 1, 3, 4). Suppose that is
l,r Z,r 3
the diagram obtained by removing a node from the end of the rith row of

i)
Define 0 £ (M\.I, MA by
i
n­l
0.: {t} ­+ { 0 if n row of {t.}
C£} if n £ r. th row of {t}

where {t} is {t}, with n removed.

When t is standard,

e . . {e­
t t
if n £ r i th row of t

o if n s r lth,r 2th, ••• ,or ri_lth row of t ,

Let Vi be the space spanned by those poly tabloids e where t is

t
a standard \.I­tableau and n is in the ••• , or rith row of t.
Then V. 1 c Ker 0. and V.0. =
­ a,

since the standard Ai_poly tabloids span sAi.

In the series
 35

••• - Vm- l Vm n Ker 8 m c Vm

i
we have dim(Vi/(V
i
n Ker 8
i)l = dim Vi 8 i = dim sA •
But m
I dim sAi = dim
i=l
since the dimension of a Specht module is the nurrber of standard tab-
leaux. Therefore, there is equality in places in the
i
above, and Vi/V is - isomorphic to sA • This is our desired
i_ l
result.
(4 2 2 1)
9.5 EXAMPLE As an F G -module, S ' , has a series with factors,
2), 2,1).
reading from the top, to s(4,2 s(4,2,12), s(3,2
(cf. Example 17.16.)
 36

10. p-REGULAR PARTITIONS

We have seen that n is zero or irreducible,and that

it can be zero only if the ground field has prime characteristic p.
In order to distinguish between those partitions for which is or
is not contained in we make the following

10.1 DEFINITION A partition is p-singular if for some i

= ... > O.
Otherwise, is p-regular.
2
For example, (6 ,1) is p-regular i f and only i f p 2: 5.
A conjugacy class of a group is called a p-regular class i f the
order of an element in that class is coprime to p.

10.2 LEMMA The number of p-regular classes of equals the number

of p-regular partitions of n.

Proof: Writing a permutation as a product of disjoint cycles, we

see that has order coprime to p if and only if no cycle has length
divisible by p. Therefore, the number of p-regular classes of G'n
equals the number of partitions of n where no part of is divi-
sible by p.
Now simplify the following ratio in two ways:
p
x ) (1 - x 2P ) •••
(1 -
(1-x)(1-x 2 ) • • •

(i) Cancel equal factors (1 - xmp)in the numerator and denomin-

ator. This leaves
IT (1 - xi) -1 = IT (1 + xi + (xi) 2 + (xi) 3 + ••• )
pfi pti
and the coefficient of x n is the number of partitions of n where no
summand is divisible by p. (The partition ( ••• 3 c , 2 b , l a ) corresponds
a
to taking x from the first bracket (x 2)b from the second bracket,
and so on.)
mp)
(ii) For each m divide (1 - x m) in the denominator into (1 - x
in the numerator, to give
; (1 + x m + (x m) 2 + ••• + (xm)p-l).
m=l
n
Here the coefficient of x is the number of partitions of n where no
part of the partition occurs p or more times.
Comparing coefficients of x n, we obtain the desired equality (The
reader who is worried about problems of convergence is referred to
section 19.3 of Hardy and Wright [3]).
Remark Like most combinatorial results involving p-regularity, Lemma
 37

10.2 does not require p to be prime, and it is only when we come to

representation theory that we must not allow p to be composite.

We next want to investigate the integer defined by

10.3 = g.c.d.{<et,et*>le t and e t* are poly tabloids in

The importance of this number is that it is the greatest common
divisor of the entries in the Gram matrix with respect to the standard
basis of the Specht module. (Corollary 8.6 shows that any poly tabloid
can be written as an integral linear combination of standard poly tab-
loids).

10.4 LEMMA (James [7J) Suppose that the partition H has parts
equal to j. Then j[l Zj: divides and divides IT
j=l
Remarks Since 0: 1, there is no problem about taking infinite
products. Some of the integers involved in the definition of may
be zero or negative, but we adopt the convention that, for example,
g.c.d. {­3,0,6} = 3.

Proof: Define an equivalence relation on the set of by

{tl} {t 2} if and only if for all i and j, i and j belong to the
same row of {t when i and j belong to the same row of {tl}.
2}
Informally, this is saying that we can go from {tl} to {t by
00 2}
shuffling rows. The equivalence classes have size IT z.:
j=l ]
Now, if {tl} is involved in e and {tl} {t 2}, then the defini-
t
tion of a poly tabloid shows that {t
2} is involved in e t, and whether
the coefficients (which are ±ll are the same or have opposite signs
depends only {tl} and {t
2}. Therefore, any two polytabloids have a
multiple of Zj: tabloids in common, and jli Zj: divides (cf.
l
Example 5.4).
Next, let t be any and obtain t* from t by reversing
the order of the numbers in each row of t. For example,
if t = 1 2 3 4 then t* = 4 3 2 1
5 6 7 7 6 5
8 9 10 10 9 8
11 11

Let 71 be an element of the column stabili zer of t haVing the pro-

perty that for every i, the numbers i and i71 belong to rows of t which
have the same length. (In the example, 71 can be any element of the
group x G{6,9} x S{7,10}l. Then {t7l} is involved in e t and
e t* with the same coefficient in each. It is easy to see that all
tabloids common to e t and e t* have this form. (In the example, every
 38

tabloid involved in e t* has 1 in the first row. Looking at e t, no

common tabloid has 5 or 8 in the first row. Going back to e t*, 2 must
be in the first row of a common tabloid, and so on.) Therefore, < e t,
00 j
e t* > (Zj:> , and the lemma is proved.

10.5 COROLLARY The prime p divides if and only if V is p-singular.

Proof: is p-singular if and only if p divides Zj: for some j, and

this happens if and only if p divides

10.6 COROLLARY If t* is obtained the v-tableau t by reversing the

order of the numbers in each row of t, then et*K is a multiple of e t,
t
and this multiple is coprime to p if and only if V is p-regular.

Proof: Corollary 4.7 shows that et*K t is a multiple of e t, et*K t h e

t
say. Now,
h h < e , {t.} > < h et' {t} >
t
< et*,{t}K >
t
The last line of the proof of Lemma 10.4 shows that h
which is coprime to p if and only if is p-regular.
 39

11. THE IRREDUCIBLE REPRESENTATIONS OF Sn

The ordinary irreducible representations of Gn were constructed

at the end of section 4. We now assume that our ground field has
characteristic p, and the characteristic 0 case can be subsumed in this
one, by allowing p =
11.1 THEOREM Suppose that is defined over a field of characteris-
tic p. Then n is non­zero if and only if H is p­regular.

Proof: c if and only if < et,e > = 0 for every pair of

t*
polytabloids e and e in But this is equivalent to p dividing
t t*
the integer defined in 10.3, and Corollary 10.5 gives the desired
result.

Shortly, we shall prove that all the irreducible Fsn­modules

are given by the modules where

11.2 DEFINITION Suppose that the characteristic of F is p (prime or

and that is p­regular. Let = n

As usual, we shall drop the suffix F when our results are indep-
endent of the field.
To prove that no two are isomorphic, we need a generaliza-
tion of Lemma 4.10, which said that SA is sent to zero by every element
A,
of Hoffip G (M unless A !?: u ,
n
11.3 LEMMA Suppose that A and Hare partitions of n, and A is p­reg-
ular. Let U be a submodule of and suppose that 6 is a non­zero
FG n­homomorphism from SA into Then A H and if A = then
1m e + U)/U.
Remark The submodule U is insignificant in the proof of this result.
The essential part of the Lemma says that, for A p­regular, SA is sent
to zero by every element of Hoffip unless A • (cf. Coro-
llary 13.17). n

Proof: (See Peel [20J) • Let t be a A­tableau and reverse the order of
the row entries in t to obtain the tableau t*. By Corollary 10.6,

et*K t =h e
t
where h f O.

But h e
t6 = e t* Kt6 = e t*6K t
Since h .. 0 and 6 is non­zero, e
t*6K t
.. U. By Lemma 4.6, A
and if A = then
-1 u
e t6 = h e
t*6K t
= a multiple of e
t
+ U (S + U) /U.
The result follows, because SA is generated by e
t•
 40

11.4 COROLLARY Suppose that A and are partitions of n, and A is p-

regular. Let U be a submodule of and­!?JJPJ2.9_se_.th.gt 8 is ..i'LI!on­zero
F homomorphism from DA into Then A and A if U
--n
Proof: He can lift e to a non­zero element of HOr:"FG as fol-
n
lows:
SA + SA/(SA n = DA +
canon. e
Therefore, A by the Lemma. If A = then Iw e is a non­zero
submodule of + U)/U, so U does not contain

11. 5 THEOREM (James [7 J) Suppose that our ground field F has charac-
teristic p (prime or = 00). As varies over p­regular partitions of n,
varies over a complete set of inec;ui valent irreducible F n -fnoduLes ,
Each is self­dual and absolutely irreducible. Every field is split-
ing field for (,5 n'

Proof: Theorews 4.9 and 11.1 show that is self­dual and absolutely
irreducible.
A
Suppose that D ; Then we have a non­zero F bn­homomorphism
A
from D into n and by Corollary 11.4, A Similarly,
II A, so A =
Having shown that no two are isomorphic, we are left with the
question: Why have we got all the irreducible representations over F?
ill section 17 we shall prove that every composition factor of the regular
representation over F is isomorphic to some and then Theorem 1.1
gives our result. Rather than follow this artificial approach, the
reader will probably prefer to accept two results from representation
theory which we quote from Curtis and Reiner [2J:
Curtis and Reiner 83.7: If Q is a splitting field for a group G, then
every field is a splitting field for G.
Curtis and Reiner 83.5: If F is a splitting field for G, then the num-
ber of inequivalent irreducible FG­modules equals the number of p-
regular classes of G.
Since Theorem 4.12 shows Q is a splitting field, Lemma 10.2 now
sees us home. More subtle, (to make use of our knowledge that is
absolutely irreducible), is to combine Curtis and Reiner 83.5 with
Curtis and Reiner 82.6: The number of inequivalent absolutely irred-
ucible FG­modules is less than or equal to the number of p­regular
classes of G.
Theorem 1.6 gives
11.6 THEOREM The dimension of the irreducible representation of
n over a field of characteristic p can be calculated by evaluating
the p­rank of the Gram matrix with respect to the standard basis of
 41

11.7 EXAMPLE We have already illustrated an application of Theorem

11.6 in Example 5.2. Consider now the partition (2,2). The Gram matrix
we obtain is (cf. Example 5.4):

A = [:

The p-rank of this is 0,1 or 2 if P = 2, 3 or >3, respectively.

Therefore, s(2,2) /(s(2,2) n = 0 if char F = 2, and dim D(2,2)

1 or 2 if char F = 3 or >3, respectively.

11.8 THEOREM The dimension of every non-trivial 2-modular

representation of G is even.
n

Proof: If (n) and t is a then < et,e >, being the order
t
of the column stabilizer of t, is even. Hence < , > is an alternating
bilinear form when char F = 2, and it is well-known that an alternating
bilinear form has even rank, so Theorem 11.6 gives the result.

Remark Theorem 11.8 is a special case of a general result which states

that every non-trivial, self-dual, absolutely irreducible 2-modular
representation of a group has even dimension.
The homomorphism e in
the proof of Theorem 8.15 sends {tl}K t , to
- AI
{t}KtP t 3 u, and Ker e = S
L
• Thus, if A' is p-regular, the submod-
ule of sA generated by {t}KtP is isomorphic to DA: In terms of the
t
group algebra FG n , this means that the right ideals generated by
PtKtP t (choosing one t for each partition whose conjugate is p-regular)
give all the irreducible representations of G over F when char F - P
n
(p prime or =
 42

12 COMPOSITION FACTORS

We next examine what can be said about the composition factors of

and in general terms. When the ground field has characteristic
zero, all the composition factors of are known (see section 14).
The problem of finding the composition factors of when the field is
of prime characteristic is still open. (All published algorithms for
calculating the complete decomposition matrices for arbitrary symmetric
groups give incorrect answers.)
First, a generalisation of Theorem 4.13:

12.1 THEOREM All the composition factors of have the form D

A with
A p, except if is p-regular, when occurs precisely once.

Proof: Consider the following picture:

I
+

/
ur
S n S
I
o

By Corollary 11.4, all the composition factors of have the

form DA with A But is isomorphic to the dual of and
so has the same composition factors, in the opposite order. (See 1.4,
and recall that every irreducible F is self-dual.) Now,
n is non-zero if and only if is p-regular, when it equals
Since 0 - n c _ is a series for the Theorem is
proved.

12.2 COROLLARY If P is p-regular, has a unique top composition

factor = n If D is a composition factor of n

then D D
A for some A If P is p-singular, all the composition
factors of A
have the form D with A p.

Proof: This is an immediate corollary of Theorems 4.9 and 12.1.

The decomposition matrix of a group records the multiplicities

of the p-modular irreducible representations in the reductions modulo
p of the ordinary irreducible representations. Corollaries 8.11 and
12.2 give
 43

12.3 COROLLARY The decomposition matrix of for the prime p has

the form:
DIl (11 p-regular)
_ _ _.A....._ _....

5 11 (11 p-regular)
1
o

*
when the p-regular partitions are placed in dictionary order before all
the p-singular partitions.

12.4 EXAMPLE Consider n

3
= 3, S(3) = D(3) is the trivial p-modular
representation. 5(1 ) is the alternating representation, and
3
5 (1 ) e;: S (3) if and only if p = 2. Using Example 5.1, the decomposi-
tion matrices of are:

when p 2,

1 when p > 3

(By convention, omitted matrix entries are always zero.)

 44

13 SEMI STANDARD HOMOMORPHISMS

Carter and Lusztig [1] observed that the ideas in the construction
of the standard basis of the Specht module can be modified to give a
basis for when char F 2. A slightly simplified form
of their is given here, and some cases where the ground field
has characteristic 2 are included.
A A
We keep our previous notation for the modules Sand M , but it is
convenient to introduce a new copy of This requires the introduc-
tion of tableaux T having repeated entries, and we shall use capital
letters to denote such tableaux. A tableau T has if for every
i, the number i occurs times in T. For example

2 2 1 1
1

is a (4,1)­tableau of type (3,2).

13.1 DEFINITION = {TIT is a A­tableau of type ul .

Remark: We allow to be any sequence of non­negative integers, whose

sum is n. For example, if n = 10, can be (4,5,0,1). The definition
of as the permutation module of G on a Young subgroup does not
n
requ ; r e ""'1 >]l
­
>
2 ­ .•.•
and M(4,5,0,1) ;; M(5,4,1).
­

For the remainder of section 13, let t be a given A­tableau (of

type (In)).
If T let (i)T be the entry in T which occurs in the same
position as i occurs in t., Let G'n act on by

The action of n is therefore that of a place permutation, and we are

1
forced to take n­ in the definition to make the well­
defined.

13.2 EXAMPLE Ift 1 345 and T 2 211 then

2 1

T(l 2) 1 211 and T (1 2 3) 2 1 1 1 .

2 2

Since Gn is transitive on and the stabilizer of an ele-

ment is a Young subgroup we may take to be the vector space
 45

over F spanned by the tableaux in It will soon emerge why we

have defined in a way which depends on both A and
If T and T
belong to we say that T l and T 2 are
l 2
(respectively, column) equivalent if T = TITI for some permutation
2
TI in the row (respectively,column) stabilizer of the given A-tableau t.

13.3 DEFINITION If T , define the map 6 T by

6 : {t}S .... l: {'I'lIT is r.ow equivalent to T}S (8 F G'n) •
T l

It is easy to verify that 6 T belongs to HomFG (M A,M i ,

n
13.4 EXAMPLE If t = 1 3 4 5 and T = 2 2 1 1 then
2 1

2 2 1 1 + 2 1 2 1 + 2 1 1 2 + 1 2 2 1 + 1 2 1 2 + 1 1 2 2 and
1 1 I I I 1

{t}(123)El 2 1 1 1 + 1 1 2 1 + 1 1 1 2 + 2 1 2 1 + 2 112 + 1 122

T
2 2 2 1 1 1

Notice that the way to write down {t}8

is simply to sum all the
T
different tableaux whose rows contain the sa@e numbers as the corres-
ponding row of T.
It is clear that

13.5 !...Et o if and only if some column of T contains two identical

numbers.

If we define 8 T by
A
8 T = the restrictionAof 8 T to S , A
then 13.5 suggests that sometimes 8 is zero, since e {t}8 K ,
T t8T T t
To eliminate such trivial elements of we make the
n
following

D.G DEFINITION A tableau T is semistandard if the numbers are non-

decreasing along the rows of T and strictly increasing down the col-
umns of T. Let be the set of semistandard tableaux in ';1

13.7 EXAMPLE If A (4,1) and u (2,2,1), then (A consists

of the two tableaux 1 1 2 2 and 1123.
3 2
A

We aim to prove that the homomorphisms 8 T with T in

 46

usually give a basis for HomF!C (5 A,M]J). These homomorphisms will be

n
called semistandard homomo rph i srns , and, as w i t.h the standard basis of
the Specht module, the difficult part is to decide \'lhether the semi-
standard rrr.cmorpo Ls rcs span HomF G (SA ,H]J). 'I'h e proof that they are
linearly independent uses a order on the column equivalence
classes [T] of tableaux in (cf. 3.11 and 3.15):

13.8 DEFINITION Let [T <!) [T ] if [T ] can be obtained from [T

l] 2 2 l]
by interchanging wand x, \'lhere w belongs to a later column of T than
l
x and \'l < x. Then generates a partial order <l •

13.9 EXAMPLE Hhen A = (3,2) and ]J = (2,2,1), the follo\'ling tree

indicates the partial order on the column equivalence classes:

The crucial, but trivial, property of this partial order is:

13.10 It T is semistandard, and T' is row equivalent to T, then

[T'] q [T] unless T' = T.

13.11 {8 T IT E O(A,y)} is a linearly independent subset of

Horn F (SA ,M]J) •
-- n

Proof: (cf. Lemmas 8.2 and 8.3). If 0 is a linear combination of

T
homomorphisms with T in BrO(A,]J) and not all the field coefficients
equal zero, choose T such that aT 0, but aT = a if [T 4 [T].
l l l]
Then from the definition of 0 and 13.10,
T
{t} 8
T = aT
l
T + a linear combination of tableaux
l
T satisfying [T [T
2 l] 2].
 47

Since the colunm of t preserves colunm equivalence

classes, and T K 0, this shows that
l t
{t}K EaT aT = {t} EaT 8 T Kt 0
t

is a non-zero element of HomF G (S ,M ), as required.

A )J
Therefore, EaT
n
We now have to be careful about the case where our ground field
has characteristic 2:

Suppose that e is a non-zero element of HomF

A )J
(S ,M ),
13.12 LEMMA
n
and write

where t is the given A-tableau. Unless char F =2 and A is 2-singular,

then
(i) £T* o for every tableau T* having a repeated entry in some
colunm.
and (ii) c T 0 for some semistandard tableau Tl.:..
-- - 1

Proof: Part (i) Suppose that i *j are in the same colunm of t, and
(i)T* = (j)T*. We wish to prove that cT* = O.
Since Kt(i,j) = -K
t,
L c T T(i,j) = {t}K t e(i,j) = -L c
T
T

Because T*(i,j) = T*, it follows that c T* = 0 when char F 2.

If char F = 2 and A is 2-regular, let n be the permutation rever-
sing the order of the numbers in each row of t. By Corollary 10.6 ,
K
t
=
Therefore
E C TT = {t.} Kt e = {t.} Kten Kt = E C T Tn Kt •

By 13.5, no tableau which has a colunm containing a repeated

entry appears in E c T T n K so c = O.
t, T*
Part (ii) If n is in the colunm stabilizer of t, then 1 -(sgn n)n
annihilates {t}K • Therefore
t
L c
T T = L cT(sgn n)Tn ,
and so
l
= ± cT when T l and T 2 are colunm equivalent.
CT
2
Since e 0, we may choose a tableau T l such that C r 0, but
1
c T = 0 if [T l] 4 [T]. The previous paragraph and part (i) of the
Lemma show that we may assume that the numbers strictly increase down
the columns of T l •
We shall be home if we can derive a contradiction from assuming
 48

that for some j, al < a 2 < ••• < a r are the entries in the jth column of
Tl' h l < b 2 < ••• < b s are the entries in (j+l)th column of T l and
a > b q for some q ,
q
al bl

"
aq > b"
q
!I.

b
s
!I.
ar

Let X be the entry in the (i,j)th place of the tableau t, and

ij
let a)a be aGarnir element for the sets {x ., ••• ,X .} and
qJ rJ
{Xl,j+l, ••• ,xq,j+l}. Then

cT T a)a = {t}K
t
a)ae = O.
For every tableau T, T (sgn a)a is a linear combination of
tableaux agreeing with T on all except (l,j+l)th, (2,j+l)th, ••• ,
(q,j+l)th, (q,j)th, ••• ,(r,j)th places. All the tableaux involved in
Tl a)a have coefficient ± , and since
c T T Z(sgn a)a is
.Ll
zero, there must be a tableau T T with c T 0 such U1at T agrees
l
with T l on all except the places described above. Since b < •• b q
l
< a < ••• < a , we must have [TIJ q [TJ, and this contradicts our
q r
initial choice of T
l•

13.13 THEOREM Unless char F =

2 and A is 2-singular,
is a basis for HomF 125 (SA,M Il).
n

Suppose e is a non-zero element of (S A,MIl). By Lemma

n
13.12,
{th t e ZC
T T, where c Tl 0 for some T
l
We may assume that c T 0 if T and [TIJ <l [TJ. Then, by
13.10, {t}K t (8 - c T e
Tl) is a linear combination of tableaux T "lith
2
['1'1 J 1 [T 2 J · Ly e - e 8Tl is a linear combination of semi-
T1
standard homomorphisms, and so the same is true of 6. The TheoreQ now
follows from Lernrna 13.1l.
13.14 COROLLARY Unless char F = 2 and A is 2-singular,
dim HomFG (SA,M Il) equals the number of semistandard A-tableaux
n
of type Il •

Remark If v is obtained from Il by reordering the parts (e.g. 11

 49

(4,5,0,1) and v = (5,4,1)), then visibly

A v
dim G (SA ,H\.I) = dim Homp G (S ,H )
n n

Equivalently, we may choose an unusual order of integers in definition

13.6. Therefore, the number of semistandard tableaux of a given
shape and size is independent of the order we choose on the entries.
Por example, we list below the elements in (2,2,1)) for
different orderings of {1,2,3}:

1 122 1 1 2 3
when 1 < 2 < 3
3 2

3 2 1 1 3 221
when 3 < 2 < 1
2 1

1 132 1 1 2 2
when 1 < 3 < 2
2 3

13.15 COROLLARY Unless char P = 2 and A is 2-sinqular, every element

A \.I A \.I
of Homp G: (S ,1'1 ) can be extended to an element of Homp G (1'1 ,H ).
n n
Proof: 0" T can be extended to 0 T•

Of course, Corollary 13.15 is trivial if char P = 0, but we know

of no direct proof for the general case.
That Theorem 13.13 and Corollary 13.15 can be false if char P 2
and A is 2-singular is illustrated by the easy:

char P = 2, + f
-+ 12 defines an element of
which cannot-be extended to an element of

Proof: There is just one semistandard A-tableau of type \.I if A = \.I,

and none at all unless A \.I. (cf. the proof of Lemma 3.7). Corollary
13.14 gives the result.

Corollary 13.17 has already been proved under the hypothesis that
A is p-regular (Lemma 11.3), and we now provide an alternative proof
for the case where char P 2.
Let e Homp G' (S A,M\.I ), and suppose that t is a A-tableau and t l is
n
a \.I-tableau. If A \.I , or if A = \.I and {t l} is not involved in e t,
 50

then some pair of numbers a,b belong to the same row of t l and the same
column of t. Therefore

< etG,{t > -< e

l} t(a,b)8,{t l} >
-< e (a,b) >
t8,{t l}
-< e >
t8,{t l}
Since char F 2, < e t8,{tl} > o. This proves that G = 0 if
A , and that e involves only tabloids involved in e t when A =
t8
If A = and n belongs to the column stabilizer of t, then
< etG,{t}n > = < etG n-l,{t} > = sgn n < etG,{t} > and this shows that
etG = < etG,{t} > e Thus 8 is multiplication by a constant.
t•

13. 18 COROLLARY Unless char F = 2 and A is 2-singular, SA is inde-

composable.

Proof: If SA were decomposable, we could take the projection onto

A)
one component, and produce a non-trivial element of Hoffip G (SA ,H ,
n
contradicting the last Corollary.

Remark: There are decomposable Specht modules - see Example 23.10(iii).

When we investigate the representation theory of the general linear

group, we shall need the simple

13.19 THEOREH {GT IT and the numbers are non-decreasing

along each row of T } is a basis for HomF a ( r·1 A ,11 •
n
Proof: Gur set of homomorphisms has been constructed by taking one
representative T l,T ••• ,T k from each row equivalence class of
2,
The linear independence of the set follows from the definition of G
T•
Suppose that G is an element of HomF<s: If T and T' are
n
row equivalent, then T' = Tn for some n in R and so
t,
< {t}8,T' > < {t}G,Tn > < {t}Gn-l,T >

< {dG,T >

Hence
{t}G = < {t}G,T i > {t}GT.
i=l
A
and since M is a cyclic module, G is a linear combination of GT. 's
as required
k
G L < {t}G,T > GT
i=l i i
 51

14 YOUNG'S RULE

It is now possible to describe the composition factors of

exp1icity.

14.1 YOUNG'S RULE The multiplicity of as a composition factor of

equals the number of semistandard A-tableaux of type y.

Proof: Since is a splitting field for the number we seek is

dim by 1.7. But this is equal to the number of semi-
n
standard A-tableaux of type by Corollary 13.14.

Remark: An independent proof of Young's Rule appears in section 17.

Young's Rule shows that the composition factors of are obtained

by writing down all the semistandard tableaux of type which have the
shape of a partition diagram.

14.2 We calculate the factors of The semistandard

tableaux of type are:

1 112 2 3 3 1 1 122 3 1 1 122

3 3 3

1 1 1 2 3 3 1 1 1 2 3 1 112 3
2 2 3 2
3

1 1 1 2 111 2 1 1 1 3 3
2 3 3 2 3 2 2
3

111 3 111 3 111

223 2 2 223
3 3

111
2 2
3 3

Therefore in the notation of 6.4,

[3][2][2] = [7] + 2[6,1] + 3[5,2] + 2[4,3] + [5,1 2 ] + 2[4,2,1] + [3 2,1]

+ [3,2 2 ]
Remark: Young's Rule gives the same answer whichever way we choose to
 52

order the integers in the definition of "semistandard", and does not

require to be a proper partition:

14.3 EXAMPLE The factors of M(3,2) are given by

Q

by 1 1 122 1 1 1 2 III
2 2 2

or by 2 2 III 2 2 1 1 2 2 1
1 1 1

Therefore, [3][2] = [5] + [4,1] + [3,2] (c f , Example 5.2).

14.4 EXAMPLE If m S n/2 then

[n-m][m] = En] + [n-l,l] + [n-2,2]+ ••• +[n-m,m].

Since dim M(n-m,m) = (n), we deduce that
m

dim S(n-m,m) (n) (n)

= m - m-l·

Notice that Young's Rule gives as a composition factor of

with multiplicity one, and the other Specht modules we get satisfy
A in agreement with Theorem 4.13. Remembering that this shows
that the matrix m = recording factors of as A varies (see 6.1)
is lower triangular with l's down the diagpnal, we can use Young's
Rule to write a given as a linear combination of terms of the form
[A ••• [A (The method of doing this explicitlyis given by the
l][A 2] j]
Determinantal Form - see section 19). Hence we can calculate terms
like •• [\!k] (= @ ••• @ for integers
\!l' ••• '\!k. More generally, Young's Rule enables us to evaluate
Lu Jf.v l ( = @ t n) for any pair of partitions u and v • The pro-
duct is the subject of the Littlewood-Richardson Rule (section
16), and the argument we have just given shows that the Littlewood-
Richardson Rule is a purely combinatorial generalisation of Young's
Rule.

14.5 EXANPLE We calculate [3,2] [2] = S (3,2) @ S (2) t >7 using only
Q. iB.
Young's Rule. By Example 14.4,
[3,2] = [3][2] - [4][1] •

To find [4][1][2], we use Young's Rule:

1 1 112 3 3 11112 3 1 1 1 1 2
3 3 3
 53

1 11133 1 1 1 1 3 1 111 3
2 2 3 2
3

1 1 1 1 1 1 1 1
2 3 3 2 3
3

[3,2][2] = [3][2][2] - [4][1][2] , and using 14.2, we have

[3,2][2] = [7] + 2[6,1] + 3[5,2] + 2[4,3] + [5,1 2 ] + 2[4,2,1]
2,1]
+ [3 + [3,2 2 ] - [7] - 2[6,1] - 2[5,2] - [4,3] - [5,1 2 ] - [4,2,1]
[5,2] + [4,3] + [4,2,1] + [3 2,1] + [3,2 2 ] . (cf. Example 16.6).
 54

15 SEQUENCES

In order to state the Littlewood-Richardson Rule in the next

section, we must discuss properties of finite sequences of integers.
A sequence is said to have 1.1 if, for each i, i occurs 1.1 i times in
the sequence.

15.1 EXAMPLE The sequences of type (3,2) are

2 2 1 1 1 2 1 2 1 1 2 112 1 2 1 1 1 2 1 2 2 1 1
x x / / / x / / / / x / / / / x / / / / / / x / /

1 2 1 2 1 1 2 112 1 1 2 2 1 11212 1 112 2

/ / .' .' .' / .' .' / / .' .' .' / .' .' .' .' .' / / .' .' / .'

15.2 DEFINITION Given a sequence, the quality of each term is deter-

mined as follows (each term in a sequence is either good or bad).
(i) All the l's are good.
(ii) An i + 1 is good if and only if the number of previous
good i's is strictly greater than the number of previous good (i+l)'s.

15.3 EXAMPLES We have indicated the quality of the terms in the

sequences of type (3,2) above. Here is another example:
3 1 1 2 332 3 2 1 2
x / .' .' .' x .' .' x .' .'

It follows immediately from the definition that an i+l is bad if

and only if the number of previous good i's equals the number of prev-
ious good (i+l) 'so Hence we have a result which will be needed later:

15.4 If a sequence contains m good (c-l) 's in succession, then the

next me's in the sequence are all good.

15.5 DEFINITION Let 1.1 = (1.1 ••• ) be a sequence of non-negative

1,1.12'
integers whose sum is n, and let 1.1* = (I.It, 1.1;, ••• ) be a sequence of
non-negative integers such that for all i,

Then I.I¢, 1.1 is called a pair of partitions for n.

Remark: As here, we shall frequently drop the condition 1.11 1.1

2 •••
on a partition 1.1, but will still refer to 1.1 as a partition of n.
If the condition 1.1 1.1 ••• holds we shall call 1.1 a proper partition
1 2
of n. So, for example, 1.1* is a proper partition of some n' n in
definition 15.5. Note that a Specht module SI.I is defined only for 1.1
 55

a proper partition, but the module Nil spanned by ll-tabloids may have
11 improper.

15.6 DEFINITION Given a pair of partitions 11 for n, let s(ll.,ll)

be the set of sequences of type 11 in which for each i, the number of
good i's is at least llr
We write 0 for the partition of 0, so that s(O,Il) consists of all
sequences of type 11. Since the number of good (i+l) 's in any sequence
is at most the number of good i's there has been no loss in assuming
flo ,.
that lli+l ,;; lli'

15.7 If = III and A1 = 11: for i > 1, then S(ll*,ll), since

every 1 in a sequence is good.

Thus we can absorb the first part of 11 into 11#.

15.8 EXAMPLE 5(0, (3,2» = s( (3), (3,2». The sequences in the second
and third columns below give s«3,1) ,(3,2» and the sequences in the
last column give s( (3,2), (3,2».

s«3),(3,2» =.> s «3,1) , (3,2» ::> s «3,2), (3,2»

2 2 1 1 1 2 1 2 1 1 1 2 1 2 1
2 1 1 2 1 1 2 1 1 2
2 1 1 1 2 1 1 2 2 1
1 2 2 1 1 1 1 2 1 2
1 1 1 2 2

Example 5.2, where M(3,2) has a series of submodules with

the factors of dimensions 1,4 and 5. This is no coincidence

Given a pair 11#, 11 of partitions, we record them in a picture

similar to a diagram. I,j"e shall draw a line between each rml and enc-
lose 11# by vertical lines. The picture for 11* ,11 will always be ident-
ified with the picture obtained by enclosing all the nodes in the first
row (cf. 15.7).

15.9 EXAMPLE Referring to Example 15.8, we have

s x X X s IRJ[]! s 5

X X X X U
This nesting suggests that we should have some notation which adds
a node from 11 to We need only consider absorbing a node which is
not in the first row.
 56

lS.lO DEFINITION Suppose ll* ll. Let c be an integer greater than

#
1 such that < and llc-l

(i) If > ll;' then II # A , II is the pair of parti tions

c
obtained from ,ll by changing to + 1. If llc-l = llc' then
ll"" A II is the pair 0,0.
c'
(ii) II • , llR is the pair of partitions obtained from
c
by changing llc to llc# and llc-l to llc-l + llc - llc • *

The operator R merely moves some nodes lying outside ll* to the end
c
of the row above (R stands for "raise" and A stands for "add"). Both
and II are involved in the definitions of Ac and Rc' and note that
we stipulate that ll:_l equals llc-l'

lS.l1 EXAMPLE IXxxxxl

lKTI Ix X X Ix X
X X

Other examples are given in lS.13, 17.1S and 17.16.

Since R raises some nodes, and we always enclose all the nodes in
c
the first row, any sequence of operations Ac,R on a pair of partitions
c
leads eventually to a pair of partitions of the form A,A (when, per-
force, A is a proper partition.) It is also clear that

lS.12 Gi ven any pair of partitions, II '# , IJ, there is a partition v

and a sequence of operations Ac,R leading from O,v to ll# ,ll.
c

lS.13 EXAMPLE To obtain «4,3,1), (4,S,2 2 » , apply

3
A A R R R R R R to (0, (4,3,1,2,1,2»:
2 3 4 3 S 6 4 S
Ix X X xl R4 R3
X X X X X X
.... X
X X
X X X X
X X X
X X X X X X
 57

R4 R5
......,...-__"'--_X ... x

x x

The critical theorem for sequences is

15.14
s(}J#,}J) \

(c-l) 's.

Proof: Recall that our definition of the operators A and Rc required

c
= Therefore, a sequence "i in \ contains

= ... (c-l) 's, all good.

good c's and bad c's.

The bad c's are changed to (c-l)'s to give a sequences s2. We claim
that
15.15 For all j, the number of good (c-l)'s before the jth term of
s2 the number of good (c-l) 's before the jth term in sl.

This is certainly true for j = 1, so assume true for j i. Then

15.15 is obviously true for j = i + 1, except when the ith term is a
(c-l) which is good in sl but bad in s2. But in this case, the inequ-
ality in 15.15 (with j replaced by i) is strict, because the number of
(c-2) 's before the ith term is the same in both sl and s2. Therefore,
15.15 is true for j =i + 1 in this case also.
15.15 shows that s2 has at least good (c-l) 's, and that all
the c's in s2 are good. Hence, for i c-l or c, i is good in s2 if
and only if i is good in sl' and so s2 belongs to
It is more difficult to prove the given map 1-1 and onto.
Given any sequence replace all the (c-l)'s by left-hand brackets,
( , and all the c's by right-hand brackets,). For example, if c = 3

1 2 1 2 312 332 2 113 112 2 3

goes to
1 1 1 1 1 1 1

Now, in any sequence be longing to s , uR

c)' all the c' s are good.
Therefore, every right-hand bracket is preceded by more left-hand brac-
kets than right-hand brackets, and the sequence looks like
 58

where each Pj is a closed parenthesis system, containing sone terms i

with i c-l or c.

It is now clear that there is only one hope for an inverse map;
namely, reverse the first Il - 116 "extra" brackets (precisely the
c
brackets which are reversed must become unpaired right-hand brackets,
to give us an inverse image.)
Let s belong to s(ll# ,IlRc)' We say that a c-l is black in s if it
corresponds to an extra bracket; otherwise it is white.
Let s* be the sequence obtained from s by changing the first Ilc -
11* black (c-l) 's to c's. \"e must prove
c
Every c-l in s* is good.

The Theorem will then follow, since every c appearing in both sand
s* will be good, and s* will be the unique element of s (]..I # ,]..1)
s(]..I# A ,11) mapping to s.
c
We tackle the proof of 15.16 in two steps. First

15.17 For every term x in s, the of white (c-l) 's before x

the of good (c-l) 's before x.

Initially, let x be a black c-l. The of white (c-l) 's before

x = the nurriber of c's before x (by the definition of "black") the
of good (c-l) 's before x, since every c in s is good. This
proves 15.17 in the case where x is a black c-l.
The same proof shows that the number of white (c-l) 's in s the
of good (c-l) 's in s. Thus, we may start at the end of sand
work back, noting that 15.17 is trivially true for the (j-l)th term of
s if it is true for the jth term, except when the (j-l)th term is a
black c-l, which is the case we have already done.

Next we have

15.18 Either c = 2, or for every c-l in s*, the number of previous

good (c-2) 's > the number of previous (c-l) 's in s*.

For the proof of 15.18, assume c > 2. Now, s contains at most

]..Ic - bad (c-l) 's since s belongs to s(]..I# ,]..IR
so for any c-l in
c)'
s, the number of previous good (c-2) 's > the number of previous (c-l) 's
in s - (Il c - Therefore, 15.18 holds for a c-l after the (Il
c -
]..I # ) th black c-l in s ,
c If the term x in s
* is a c-l appearing before the (]..I
c - 11#
c )th black
c-l in s, then x was white in s. Also, the number of (c-l) 's before x
in s* = the number of white (c-l) 's before x in s the number of good
 59

(c-1) 's before x in s by 15.17 (the inequality being strict if x is a

bad c-1 in s, by applying 15.17 to the next term) the number of good
(c-2) 's before x (the inequality being strict if x is a good c-1 in s),
and 15.18 is proved in this case too.
From 15.18, 15.16 follows at once, and this completes the proof of
Theorem 15.14.

15.19 EXAMPLE Referring to Example 15.8, the 1-1 correspondence

between s «3), (3,2» \ s «3,1), (3,2» and s «5), (5» is obtained by:

22111 ... 1 1 1 1 1
x x I I I
The 1-1 correspondence be-tween s«3,1), (3,2» \ s«3,2), (3,2» and
s«4,1),(4,1» is given by

2 1 2 1 1 1 1 2 1 1
x I I I I
2 1 1 2 1 1 1 1 2 1
x I I I I
2 1 1 1 2 1 1 1 1 2
x I I I I
1 2 2 1 1 1 2 1 1 1
I I x I I
 60

16 THE LITTLEh'00D-RICHARD50N RULE

The Littlewood-Richardson Rule is an algorithm for calculating

[A][p] where A is a proper partition of n-r and p is a proper partition
of r. Remember that [A][p] is a linear combination of diagrams with n
nodes, and the interpretation is that when a is the coefficient of
v
tvi , ® t G has as a composition factor with multiplicity
n
a It is a well-known result that every ordinary irreducible repre-
v•
sentation of G x H, for groups G and H is equivalent to 51 x 52' for
some irreducible G-module 51 and some irreducible H-module 52' so the
Littlewood-Richardson Rule enables us to calculate the composition
factors of any ordinarj representation of a Young subgroup, induced up
to G>n'
For the moment, forget any intended interpretation, and consider
the additive group generated by {[A]!A is a proper partition of some
integer}. Given any pair of partitions p# ,p as in definition 15.5,
we define a group endomorphism l u 41: ,p]' of this additive group as foll ...
ows:
16.1 DEFINITION [A][p# ,pJ' = Z av[v] where a = 0 unless Ai vi
v
for every i, and if Ai vi for every i, then a is the number of ways
v
of replacing the nodes of [v]\[AJ by integers such that
(i) The numbers are non-decreasing along rows
and (ii) The numbers are strictly increasing down columns
and (iii)When reading from right to left in successive rows, we have a
sequence in s (u -# ,\1).
If \1# \1, when \1 is a fortiori a proper partition, we write [\1]
for [\1/ P]·.

The operators are illustrated by the next Lemma and by Examples

16.6 arid 16.7.

1 6.2 LEMMA If II -- ( \1 1 ,l!.2 '..:....:...:-.L\1k'

) ..::t::,;h..::e.:;;n,• [ J[O,\1]
If II is a proper partition, then [0][\1] = ell]'

Proof: When \1#= 0, condition (iii) of definition 16.1 merely says

that we have a sequence of type \1. But [\11][P2J ••• [\1k]' by definition,
describes the composition factors of , and the first result follows
from Young's Rule.
Let [vJ be a diagram appearing in [0][\1]'. Then the nodes in (v]
can be replaced by \1 l's, \1 2's, and so on, in such a way that
1 2
conditions (i) to (iii) of 16.1 hold. Suppose that some i appears in
the jth row with j < i, and let i be the least number for which this
happens. There are no (i-l) 's higher than this i, by the minimality of
 61

of i; nor can there be any (i-l) 's to the right of it in the same row,
by condition (i). Thus, this i is preceded by no (i-l) 's when reading
from right to left in successive rows, and the i is bad, contradicting
condition (iii). But no i can appear in the ith row with i > i, Lv condi-
tion (ii). This proves that every i is in the ith row, and [v] =

16.3 LEI1MA
Proof: Assume that is a partition of r, and that A and v are
partitions of n-r and n, respectively, with Ai vi for each i.
Replace each node in [v]\[A] by l's, 2's and so on, such that we
have a sequence in \ A when reading from right to left
c
in successive rows. We must prove blat changing all the bad c's to
(c-l) 's gives a configuration of integers satisfying 16.1 (i) and (ii)
if and only if we start with a configuration of integers satisfying
16.1 (i) and (ii), since the Len®a will then follow from Theorem 15.14.
Suppose we have not yet changed the bad c" s to (c-l) , s and condi-
tions 16.1 (i) and (ii) hold for our configuration of integers. There
are two problems which might occur. A bad c might be to the right of
a good c in the same row. This cannot happen, because a c immediately
after a bad c must itself be bad, llore complicated is the possibility
that there is a bad c in the (i,j)th place and a c-l in the (i-l,jlth
place. To deal with this, let m be maximal such that there are c's in
the (i,j)th,(i,j+l)th, ••• , (i,j+m-l)th places. Then by conditions 16.1
(L) and (ii), there are (c-l) 's in the (i-l,j) th, (i-l,j+l) th, ••• , (i-l,
j+m-l)th places. Since all the (c-l) 's are good in a sequence belonging
to our c in the (i,j)th place must be good, after
all, by 15.4. This shows that all the bad c's can be changed to (c-l) 's
without affecting conditions 16.1(i) and (ii).
Conversely, suppose that after changing the Dad c's to (c-l) 's we
have a configuration satisfying conditions 16.1 (i) and (ii). We dis-
cuss the configuration of integers we started with. This must satisfy
conditions 16.1 (i) and (ii) unless a bad c occurs immediately to the
left of a (good) c-l in the same row, or a bad c lies immediately above
a good c in the same column. The first problem cannot occur by 15.4.
Therefore, we have only to worry about the possibility that a bad c is
in the (i-l,j)th place and a good c is in the (i,j)th place. Reading
from right to left in successive rows, we see that the number of (good)
(c-l) 's in the (i-l)th row to the left of our bad c in the (i-l,j)th
place is at least the number of good c's in the ith row. But every
c-l in the (i-l)th row to the left of the (i-l,j)th place must have a
good c immediately below it in the ith rm'1 (s Lrce there is a good c in
 62

fue (i,j)th place, and we end up with a configuration satisfying condi-

tions 16.1 (i) and (ii». This contradicts the fact that there is a
good c in the (i,j)th place, and completes the proof of the Lemma.

16.4 THE LITTLEWOOD­RICHARDSON RULE

[AJ[lJJ· = [A][lJ]

Proof: (James [10J') If v is a proper partition of n , then applying

operators A and R repeatedly to 0, v we reach a collection of pairs
c c
of partitions w,w. By Lemma 16.3, we may write

[O,vJ a [w1 );
w w
where each a w in an integer, a v = 1 and a w = 0 unless [wJ rvr.
Hence there are integers b a and C such that
s
[AJ

By Lemma 16.2
[AJ[lJJ· [OJ[AJ·[lJJ

[OJ); ba[O,aJ·); cS[O,SJ

); b a [alJ···[a.J ); c [SlJ,,·[Sk J
a J 13 s
[OJ); ba[O,aJ·[OJ); cS[O,SJ·

[OJDJ·[OJ[lJJ·

DJ[lJJ •

16.5 COROLLARY [vJ·[]JJ· = []JJ·[vJ· = ([]JJ[vJ)

Proof: For all [AJ, [AJ[VJ·[lJJ [AJ[VJ[lJJ = [AJ[lJJ[VJ

[AJ[lJ]"[VJ· = DJ ([lJ][vJ)·

The Corollary is extremely hard to prove directly. More generally,

it follows from the Littlewood­Richardson Rule that for every equation
like [3J[2J = [5J + [4,lJ + [3,2J there is a corresponding operator
equation [3J·[2J z [5J + [4,lJ + [3,2J

Of course, the Branching Theorem (part (1» is a special case of

the Littlewood­Richardson Rule.
When applying the Littlewood­Richardson Rule, it is best to draw
the diagram [AJ, then add lJ l's, then lJ 2's and so on, making sure
l 2
that at each stage [AJ, together with the numbers which have been added,
form a proper diagram shape and no two identical numbers appear in the
same column. Then reject the result unless reading from right to left
 63

in successive rows each i is preceded by more (i-l) 's than i's. (This
condition is necessary and sufficient for every term to be good.)

16.6 [3,2J[2J

= [5,2J + [4,3J + [4,2,lJ + [3 2,lJ + [3,2 2J, by looking at the

follo#ing configurations: (cf. Example 14.5).

x X X 1 1 X X X 1 X X X 1 X X X x X X
X X X X 1 X X X X 1 X X
1 1 1 1

16. 7 [3,2J[2J[2J [3,2J[2J·[2J

X X X 1 1 X X X 1 1 X X X 1 1 X X X 1 X X X 1
X X 2 2 X X 2 X X X X 1 2 X X 1
2 2 2 2 2 2

X X X 1 X X X 1 x X X X X X
X X 2 X X X X 1 X X
1 1 2 1 2 1 1
2 2 2 2 2

X X X 1 1 2 X X X 1 1 2 X X X 1 2 X X X 1 2
X X 2 X X X X 1 2 X X 1
2 2

x X X 1 2 X X X 1 2 X X X 1 2 X X X 1 X X X 1
X X 2 X X X X X X 2 2 X X 2
1 1 2 1 1 1 2
2

x X X 2 X X X 2 X X X X X X 2 X X X
X X 1 X X 1 X X 1 X X X X 2
1 2 1 122 1 1 1 1
2 2 2

X X X 1 122 X X X 1 2 2 X X X 1 2 2 X X X 2 2
X X X X 1 X X X X 1
1 1

X X X 2 2 X X X 2
X X X X 2
1 1 1 1
 64

We have arranged the diagrams so that, reading from right to left

in successive rows, the diagrams in the first batch (before the first
line) give sequences in s ((2,2) , (2,2», so

[2,2]·
[3,2][2,2] = [3,2J = 2 2
[5,4J + [5,3,lJ + [5,2 J + [4 ,1J
+ [4,3,2J + [4,3,1 2J + [4,2 2,lJ + [3 2,2,lJ + [3,2 3J

The diagrams before the second line give [3,2J[(2,1) ,(2,2)J

The reader may care to check that changing a bad 2 to a 1 in the sec-
ond batch gives the sarae answer as [3,2J[3,lJ·, in agreement with Lemma
16.3.

[3,2J[3,lj = [3,2J[3,lJ· = [6,3J + [6,2,lJ + [5,4J + 2[5,3,lJ

+ [5,2 2J + [5,2,1 2J + [4 2,lJ + 2[4,3,2J + [4,3,1 2J + [3 3J + [4,2 2,lJ
[3 2 , 2 , 1 ] .

The last batch contains all the configurations where Doth 2's are
bad, and by changing the 2'5 to 1'5, Lemma 16.3 gives

[3,2J[4J [3,2J[4J· = [7,2J + [6,3J + [6,2,lJ + [5,3,lJ + [5,2 2J

+ [4,3,2J ,
which is simple to verify directly.
 65

17. A SPECHT SERIES FOR

A better form of Youngs Rule can be dedved over an arbitrary field.

What happens in this case is that has a series with each factor iso-
morphic to a Specht module; such a series will be called a Specht
series. Since is not completely reducible over some fields, we must
take into account the order of the factors in a Specht series. The
next example shows that the order of the factors does matter:

17.1 EXAMPLE
Let char F divide n > 2, and consider M(n­l,l).
(n­l 1 ) . . . (n) (n­l,l)
Example 5.1 shows that M ' 1S un1ser1al, with factors 0 ,D
D(n) and that s(n­l,l) is uniserial with factors D(n­l,l) ,D(n}, reading
(n­l 1) .. (n­l,l)
from the top. Thus M ' has no Specht ser1es W1th factors S ,
s(n) reading from the top. The Specht series in Example 5.1 has factors
in the order s(n) ,s(n­l,l).

In this important section, we shall use only Theorem 15.14 on seq-

uences, and deduce both Young's Rule and the standard basis of the
Specht module. At the same time, we characterize the Specht module SA
as the intersection of certain F Gn­homomorphisms defined on MA, in the
case where A is a proper partition. Throughout this section F is an
arbitrary field.
Let u 'If , u be a pair of partitions for n, as defined in 15.5. Recall
that 'N" must be a proper partition, while we do not require u to be
proper. We want to define a submodule of and to do this we
construct an object which is "between" a tabloid and a poly-
tabloid.

17.2 DEFINITION Suppose that t is a Let

et = e­
L.
{ sgn 'IT { t } 'IT I 'IT Ct and 'IT fixes the numbers outside Lu # ] }

17.3 EXAMPLE If t = and = (3,2,0), u = (3,4,2)

U­:t 9
8 6
(part of t is boxed­in only to show which numbers will be moved) , then

e 135 2 3 5 175 2 7 5
t
2 7 4 9 1 7 4 9 2 3 4 9 + 1 3 4 9
8 6 8 6 8 6 8 6

""
'#
17.4 DEFINITION is the subspace of spanned by 's
t
as t varies.

Of course, S is an F GI ­submodule of
n since e t , 'IT
t1f ""
 66

If u 41= = 0, then = and if

.ff'

17.5 If Al = and = for i > 1, then SA , so

we can absorb the first part of into (cf. 15.7).

Sequences now come into play by way of

17.6 CONSTRUCTION Given a sequence of type construct a corres-

ponding t as follows. I'lork along the sequence. If the jth
term is a good i, put j as far left in the ith row of t as possible.
If the jth term is a bad i, put j as far right in the ith row as poss-
ible.

17.7 EXAMPLE 311233232121 E s«4,3,2),(4,4,4))

x I I I I x I I x I I I
and corresponds to

Different sequences in correspond to tableaux which belong

to different tabloids, so

The construction gives a 1­1 correspondence between and

the set of tabloids.

Remark We have already encountered the concept of viewing a basis of

II
M as a set of sequences, for in section 13, the tableau T of type
corresponds to the sequence (l)T, (2)T, •••• ,(n)T.

The construction ensures that a sequence in s ,11) corresponds to

a tableau which is standard inside [11"] (Le. the numbers which lie in-
side [11#] increase along rows and down colmnns­ cf. Example 17.7). But,
if t is standard inside {t}is the last tabloid involved in
#
,II (cf. Example 17.3), and so Lemma 8.2 gives
'It
17.9 It corresponds to a sequence in s(Il"" ,11) by 17.6} is
#
a linearly independent subset of Sll ,11.
<#
We shall see soon that we actually have a basis of Sll ,11 here.
#­ ." A 'It R
Our main objective, though, is to prove that c,1l SIl c,
where the operators Ac and Rc are defined in 15.10. First, note that
#A 'It
SII £ SII ,11 This is trivially true if 11# A
c,1l
= 0,0 (i.e. if
= since we adopt the convention that SO,O is the zero module.
"c­l c'
Otherwise, given t, we may take coset representatives 0l, ••• ,ok for
the subgroup of Ct fixing the numbers outside [11:/1'] in the subgroup of
 67

Ct fixing the nwnbers outside Lu # A ], whereupon

• k c
ell ,ll E (sgn 0.)0 . •
t i=l
Now we want an F G'n -homomorphism mapping Sll'" , II onto SII * , llRc •
17.10 DEFINITION Let II = (1l1,1l2' ••• ) and

v =
(1l1,1l2, ••• ,lli_l,lli + lli+l - V,V,lli+2' ••• ) • Then belonging to
V
G'n(Mll,M ) is defined by {t}ljIi,v = E {{tl}l{t
l} agrees with {t} on
all except the ith and (i+l)th rows, and (i+l)th row of {t l} is a
subset of size v in the (i+l)th row of {t}}.

Remark It is slightly simpler to visualize the action of ljIi,v on the

basis of Mil viewed as sequences. ljIi ,v sends a sequence to the sum of
all sequences obtained by changing all but v (i+l) 's to i's. Whichever
way you look at it, ljIi ,v is obviously an F G n -homomorphism. Every
tabloid involved in {t}ljIi,v has coefficient 1, so ljIi,v is "independent
of the ground field."

17.11 EXAMPLES
(i) When II = (3,2), ljIl,o and ljIl,l are the homomorphisms ljIo and
ljIl appearing in Example 5.2.
(ii) If II = (4,3 2 , 2 ) , then

1 2 5 10 1 2 5 10 1 2 5 10
ljI2,1
3 4 9 + 3 4 9 7 8 + 3 4 9 6 8
6 7 8 6 7
11 12 11 12 11 12

+ 1 2 5 10
3 4 9 6 7
8
11 12

(iii) If n 6 and II = (n-3,3) and

v = r-r-3 + + + (replacing each tabloid by its
second row only), we have
v ljJl,o 4 F
v ljIl,l 1 + i + 3 + I + i + 4 + I + 3 + 4 + i + 3 + 4
3(1 + 2 + 3)
2(12 + T3" + T'4 + n + 24 + 34).
Therefore, v Ker ljJ 1,0 n Ker ljI 1,2 if and only if char F 2
and v Ker ljIl,l if and only if char F 3.
 68

(iv) Taking n = 6 in exmaple (iii),

(4 5 6 - "f56)\jJl,l = 4 + '5 + 6 - I - '5 - 6 4 - I

(4""'5'6 "f56 - 426 + 1 2 6)\jJl,1 0.

T:'lat is, for t = , 11'* = (3,1) and 11 .= (3,3) , we have

456

where tR 2 1 2 356
4
and ll
et -
* ,I,
'1'1,1 0.

Compare the last Example with

17.12 LEMr-1A Sll "" ,11 S

11# ,IlR
c

0l!-
and Sll Ac,ll 0.

Proof: Let t be a Il-tableau, and let

K
t
*
= L {sgn n)nln fixes the numbers in t outside [ll-#]}.
Choose a set B of numbers from the cth row of t, and move the rest
of the nuniliers in the cth row of t into the (c-l)th row.
If B consists of the first numbers in the cth row of t, then
we get a tableau, tR c say, and
{tR }K
c t'"
ell
tR c
,IlRc ""
:#
For any other set of llc numbers from the cth row of t, we still get
a llRc -tabloid, {t say, but now one of the numbers, say x, which has
l}
been moved up lies inside [11 #. ] Let y be the number above x in t.,
Tnen(l-(x y» is a factor of Kt ' and so

0.

Now, {t} \jJ # K Of and

c-l'llc t

is the sum of all the tabloids obtained by moving all

{t} 1jJ
c- 1 ,Il c#
except numbers from the cth row of {t} into the (c-l)th row.
Therefore,

Since Ac.1l has one more node enclosed in the cth row (or SIl#Ac,1l
So,o = ° if 11""c-l =
c'
the proof we used to deduce that {tl}K t # = °
shows that ell-'l'Ac,ll\jJ
t
4S=
c-l,llc
°.
 69

17.13 THEOREM (James [10])

(i ) ,I, and

n ker 1/J 1 =
c-
4t #A '#
(ii)
#
(iii) dim = I; indeed,
" t corresponds to a sequence in by 17.6} is a basis of
u
.,.
(iv) has a Specht series. The factors in this series are
given by

Proof: Let O,V be a pair of partitions from which we can reach the pair
by a sequence of Ac and Rc operators (cf. 15.12)
dim SO,v = dim MV = Is(O,v) I by 17.8. We may therefore
assume that dim • = I and prove that dim #Ac' =
I and dim = I.

Now, dim "

dim dim by Lemma 17.12
I+ I by 17.9
Is I by Theorem 15.14 .
Everything falls out! We must have equality everywhere, so results
(L) ,(Ld ) and (iii) follow.
When = and so has a Specht series whose factors
are given by (see Lemma 16.2). Therefore, we may assume induct-
ively that and have Specht series whose factors are
[#A ]• [ R]'
given by [0] and [0] c . Since we have proved conclusion
#
(i), and = + (see Lemma 16.3), has
a Specht series whose factors are given by

All we have used in the above proof are the purely combinatorial
results 15.14 and 16.3 (In fact, it is much easier to show that
= + than to prove Lemma 16.3 in its
full form.) We have therefore given alternative proofs that the standard
poly tabloids form a basis for the Specht module (take = in part
(iii», and of Younq ' a Rule (take u '# == a in part (Lv l ) ,

17.14 has a Specht series. More generally,

COROLLARY
SA 0 ... 0
has a Specht series. The factors and their
order of appearance are independent of the ground field, and can be
 70

calculated by applying the operators Ac and Rc repeatedly to [O,V] and

[A, J, respectively. The factors of SA @ .•• ®
are given by

(By .•• we mean the partition (Al,···,Aj' •.• ,Pk)'

where A. is the last non-zero part of A).
J

Proof: It is simple to see that

SA, (A,Pl,···,Pk) SA e s(Pl)® ..• ® s(Pk)t G'n
and we just apply Theorem 17.13(ii) to obtain a Specht series.' The last
sentence is true because [A][Pl]····[Pk]· as can
be easily verified.

Remark James and Peel have recently constructed a Specht series for
the module sP @ SAt G'n Here again, the factors and their order of
appearance are independent of the ground field. The Specht factors are
given by the Littlewood-Richardson Rule.

17.15 EXAMPLE We construct a Specht series for M(3,2,1). In the tree

below, we always absorb the first part of u into P"* (e.g. M(3,2,1)
sO,(3,2,1) = s(3),(3,2,1); cf. 17.5). Above each picture we give the
dimension of the corresponding module.
60 6 6 1
Ix X Ix X X X Xl Ix X X X Xl Ir"'x-x-x-x-x-xl

-t 54
X
A2
X

24
I
o.f
A
'
;

14
tA 2
X X X xl
5

If/ xl
R3 R,2

t
)It )0 X X
X X

A
2
10
A3

9 tA2

fir
X X I
X X
 71

16 t 5 r
x x
x

Therefore, M(3,2,l) has a Specht series with factors S (6), S (5,1) ,

s(5,l), s(4,2), s(4,12), S(4,7) ,.,(32) ",(3,2,1) reading from the toP.
, 0 I 'j,.)

This holds regardless the field.

2
(4 2 1)
17.16 EXAMPLE Consider S ' , t

R5 R4R3 R2
>- :r-

t A
5 t
A4
X X X X
tA2

X X
X X
X X

2
(4 2 1)
Hence, S ' , t 2, has a series with factors, reading from the
3
top, isomorphic to 5(5,2 ), s(4,3,2,1), S ( 4 , 2 ) , s(4,22,12) ( c f .
Examples 9.1 and 9.5) .

17.17 EXAMPLE Following our algorithm, we find that when n-m,

(n-m m) (n) (n-l 1) (n -rn m)
M ' has a Specht series with factors S , S ' , •.• ,S "
reading from the top (cf. Example 14.4).
There is a point to beware of here. It seems plausible that
M(n - m- l , m+l ) / S(n-m-l,m+l) " " hi t 0 M(n1l1,m) ; a ft er a 11 , both
modules have Specht series with factors as listed above. However, this
is sometimes false. For instance, when char F = 2, s(6,2) has composi-
tion factors 0(6,2) and 0(7,1) (see the decomposition matrices in the
Appendix.) Since 0(6,2) is at the top of S(6,2)
0(7,1) ;;; s(6,2) n S(6,2).L E! M(6,2)/ (s(6,2) + s(6,2).L).

Therefore, M(6,2) / s(6,2) has a top factor isomorphic to 0(7,1), while

M(7,1) does not (see Example 5.1).

Theorem 17.13 provides an alternative method of showing that all

the irreducible representations of Sn appear as a DV , thereby avoiding
 72

the quotes from Curtis and Reiner in the proof of Theorem 11.5. Since
has the same factors as all the composition factors of
come from (if is p-regular), and from But Theorem 17.13
shows that has a series with factors isomorphic to SA,s with
,
A By ' d " SA MA , every 't' f ac t or 0 f
is isomorphic to some oV. Applying this fact to the case where = (In),
when is the regular representation of F n' Theorem 1.1 shows that
every irreducible F
n
-module is isomorphic to some oV.
Theorem 17.13(i) has the useful

17.18 COROLLARY If }! is a proper partition of n, with k non-zero

parts, then
k
= n
i=2 v=o
ker \j!i-l,v

The Corollary is perhaps the most important result of this section,

since it characterizes as a subset of consisting of vectors having
certain properties (cf. Example 5.2). It will be discussed at greater
length in the section dealing with decomposition matrices of
 73

18 HOOKS AND SKEW-HOOKS

Hooks play an important part in the representation th:ory of n'

but it is not clear in terms of modules why they have a role at all!
For example, it would be nice to have a direct proof of the Hook for-
mula for dimensions (section 20), without doing all the work required
for the standard basis of the Specht module.
The (i,j)-hook may be regarded as the intersection of an infinite
r shape (having the (i,j)-node at its corner) with the diagram.

18.1 EXAMPLE X X X X The (2,2)-hook is X X X X

X X X X X ll§ ll§ ll§

X X X X ll§ X

and the hook graph is 6 5 4 2

5 4 3 1
3 2 1

18.2 DEFINITIONS
(i) The (i,j)-hook of consists of the (i,j)-node along with
the nodes to the right of it (called the of the hook) and the
I
.-i nodes below it (called the leg of the hook).
J I
(ii) The length of the (i,j)-hook is h .. = + + 1 - i - j
J
(iii) If we replace the (i,j)-node of by the number h for
i j
each node, we obtain the hook graph.
(iv) A skew-hook is a connected part of the rim of which can
be removed to leave a proper diagram.

18.3 EXAMPLE X X X X and show the only two

skew 4-hooks in [4 2,3]. The diagram also has one skew 6-hook, two
skew 5-hooks, two skew 3-hooks, two skew 2-hooks, and two skew l-hooks.
Comparing this with the hook graph, we have illustrated:

18.4 LEMMA There is a natural 1-1 correspondence between the hooks

of [V] and the skew-hooks of [V],

Proof: The skew hook

+- ith row

jth column
corresponds to the (i,j)-hook.
 74

19 THE DETERMINANTAL

We have seen that when Al •••• ,

and the matrix m = is lower triangular with l's down the diagonal
(see 6.4 and 4.13). It follows that
1)
[A] = (m-
and m- 1 is lower triangular with l's down the diagonal.

19.1 EXAMPLE Inverting the matrix m for (;5 given in section 6, we

find

[5] [4][1] [3][2] [3][lJ2 [2]2[1] [2] [lJ 3 [115

[5] 1
[4,1] -1 1
[3,2] 0 -1 1
-1 [3,1 2]
m 1 -1 -1 1
[2 2,1] 0 1 -1 -1 1
[2,1 3 ] -1 1 2 -1 -2 1
[15 ] 1 -2 -2 3 3 -4 1

The coefficients in the matrix m are given by Young's Rule, and the
-1
entries in m can be found directly by

19.2 THE DETERMINANTAL FORM If A is a proper partition of n, then

[A] = I[A i - i+j]1
where we define Em] = 0 if m < O.

The way to write down the determinant for [A] is to put [A ••

l],[A 2]
in order down the diagonal, and then let the numbers increase by 1 as
we go from one term to the next in each row. Beware of the distinction
between [0] (which behaves as a multiplicative identity) and 0 (0 x any-
thing = 0).

19.3 EXAMPLES

[ 3] [4]/ [3][1] - [4] [3,1] + [4] - [4] [3,1]

/ [0] [ 1]

[ 3] [4]1 [3][2] - [4][1] = [3,2] + [4,1] + [5] - [4,1] - [5]

/ [ 1] [2] E3,2]

19.4 EXAMPLE Suppose we have proved the determinantal form for 2-

 75

part partitions. Then expanding the following determinant up the last

column, we have

[ 3] [4] [5] [2] [ 3]

\[3] [4]\ 1[3] [4]1
[ 1] [ 2] [ 3] [1] [2] [0] [ 1]
[0] [ 1] [2]
+ \[1] [2]1 [5]
[0] [ 1]

which, by induction, is [3,2][2] - [3,1][3] + [1 2][5]

[3,2 2] + [3 2,1] + [4,2,1] + [4,3] + [5,2]

-([32,1] + [4,2,1] + [4,3] + [5,2]) - ([6,1] + [5,1 2])
+ [6,1] + [5,1 2] =[3,2 2]

•
Diagrams
+
Diagrams
t
Diagrams
containing containing containing
X X X X X X
X X

Proof of the Determinantal Form: It is sufficient to prove the result

in the case where A = (A ••• ,A with A > 0, since zero parts at
l,A 2, k) k
the end of A do not change the determinant. The result is true when A
has no non-zero part, so assume that we have proved the result for A
having fewer than k non-zero parts.
The numbers in the last column of ([Ai - i+j]) are the "first
column hook lengths of [A]", hll,hZl, ••• ,hkl' since
,
h i l = Ai + Al + 1 i - 1 = Ai - i + k.
Let s. be the skew hook of [A] corresponding to the (i,l)-hook (In
l.
Example 19.4, s3,s2 and sl are X X X X X X X
X X X X ) .

Omitting the last column and ith row of ([Ai - i+j]) gives a matrix
with diagonal terms

and these are precisely the parts of [A\si]. Therefore, the result
of expanding the determinant I[A - i+j]1 up the last colllirun and using
i
induction is
 76

Now consider [A\ si][h This is evaluated by adding nil nodes

i l].
to [A \si] in all ways such that no two added nodes are in the
column (by the Littlewood-Richardson Rule, or Corollary 17.14).
[A \ si] certainly contains the last node of tile 1st, 2nd, ••• , (i-l) th
rows of [A], so we deduce that all the diagrams in [A \ si ][h
i l]
(i) contain the last nodes of the lst,2nd, ••• , (i-l)th rows of [A],
and (ii) do not contain the last nodes of the (i+l)th, (i+2)th, ••• ,kth
rows of [A].
Split the diagrams in D \ si ][h into 2 set, according to whe che r
i l]
or not the last node of the ith row of [A] is in the diagram. It is
clear that [A] is the only diagram we get containing the last nodes of
all the rows of [A], and a little thought shows that in (*) we get sets
cancelling in pairs to leave [A]. This proves the Determinantal Form.

19.5 COROLLARY 'I

dim S A -_ n ; (Ai -1 i+j): I where = 0 if r < 0

Proof:
).Jl:···).Jk:

(see 4.2), and the Corollary is now immediate.

 77

20 'l'HE HOOK FORMULA FOR DIMENSIONS

20.1 TIlEORE14 (Frame, Robinson and Tiuall [4])

The dimension of the Specht module sA is given by

IT (h , 1 - h k l )
i<k
IT h I
IT(hook lengths in [A])
i ilo

20.2 EXAMPLE The hook graph for [4,3,1] is

6 4 3 1
421
1

Therefore, dim 5(4,3,1) 8! = 70.

6.4.3.4.2

The hook formula is an amazing result. It is hard to prove directly

even that n! is divisible by the product of the hook lengths, let alone
show that the quotient is the number of standard A-tableaux.

Proof of Theorem 20.1 We show that the result is true when A has 3
non-zero parts. It is transparent that the proof works in general, but
a full proof obscures the simplicity of the ideas required.
By Corollary 19.5,
dim SA 1 1 1
n! (h
ll
- 2) ! (h
ll
- I)! h ll =

1 1 1
(h
n - 2) ! (h 2 1 - 1) ! h 21 !

1 1 1
(h 3 1 - 2) ! (h 31 - 1) ! h 31 !

h ll (h ll - 1) h 1
ll
1 1 1
= h ll! h
2 1!
h
31
! h 21(h2
1
- 1) h
21 1

h 31 (h 3 l - 1) h 1
3l

giving the
first result.
 78

(h - 1) (h ll - 2) h ll - 1 1
ll
1 1 1
h ll• h ! h 31 ! (h - 1) (h
21
- 2) h Z1 - 1 1
21 21

(h
31
- 1) (h
31
- 2) h 31 - 1 1

1 1 1 1
ll - 3) :
!
ll -
h h h (h (h - Z) : (h 1)
ll 21 31 ll

1 1 1
(h
Z1
- 3) ! (h
Z1
- Z) ! (h
Z1
- 1) !

1 1 1
(h
31
- 3) ! (h
31
- Z) : (h
31
- 1) !

1 1
hll h Z1 h 3 1 IT(hook lengths in [A1-1,AZ-1,A3 -1])
by induction
1
IT (hook lengths in [A] ) , as required.
 79

21 THE LiURNAGlJAN-NAKAYJ\I"L'\ RULE

'I'he r-lurnaghan-Nakayama Ru Le is a very Leautiful and efficient way

of calculating a single entry in the cbaracter table of (5n.
In the statement below, the leg-length of a skew-hook is defined to
be the s arue as that of the corresponding hook.

21.1 TEE dURN,\GHAN-NAKAYAMi, RULE

SupE,0se that 1Tp E G where p is an r-cyc1e and 1T is a permutation
n
of the remaining n-r nUIllbers. Then
A
X {1T p) = L {(-l)i XV {1T ) I [A] \ [v] is a skew r-hook of leg length i}.
v
As usual, an empty sum is interpreted as zero. 'I'he case where p is
a l-cyc1e is the Branching Theorem.

21.2 EXAMPLES
(i) Suppose we want to find the value of X{5,4,4) on the class
(5,4,3,1) •

There are two ways of removing a skew 5-hook from [5,4,4) and the
Murnaghan-Nakayama Rule gives:
X{5,4,4) on (5,4,3,1) (3,3,2) (5,3)
X - X on (4, 3 ,1)
X( 2 , 1 2 ) _ X( 3 , 1) + X ( 2 2) on ( 3 , 1) ,
applying the rule again
X (2 2) on (3, 1) , because we cannot
remove a skew 3-hook from either [2,1 2 ] or [3,1].
-X (1) on (1)

-1.

(ii) X{5,4,4) is zero on any class containing an 8,9,10,11,12 or

13-cycle, since we cannot remove hooks of these lengths from [5,4,4].

(iii)

X(5,4,4) on ( 7,3,3) X (3 2) on (3, 3 )

_X{2,1) + X(3) on (3)

 80

2.

The only character table required in the construction of the charac-

ter table of Gn using the Murnaghan­Nakayama Rule is that of Go'
Remember that is a group of order 1, and a computer is unnecessroy in
evaluating the character table of
Our proof of the Murnaghan­Nakayama Rule needs several preliminary
lemmas. We first prove the special case where p is an n­cycle, then
examine what the Littlewood­Richardson Rule gives for [v][x,lr­x], and
finally we combine these pieces of information to prove the Rule in
general. See the remarks following 21.12 for an alternative approach.
A hook diagram is one of the form [x,lY].

21.3 LEMMA Unless both [a] and [B] are hook u1agrams, [a][B] contains
n­r­a r­b
no hook diagrams. If [a] = [a,l ] and [B] [b,l ] then
[a][B] = [a + b,ln­a­b] + [a + b ­ l,ln­a­b+l] + some non­hook diagrams.

Proof: If one of [a] and [B] contains the (2,2)­node, the so does
[a][BJ = [a][B]' = [B][a]·. This proves the first result.

Suppose, therefore, that [a] = [a,ln­r­a] and [B] = [b,lr­b]. In

order to obtain a hook diagram in [a][B]', we have to put b l's in
the places shown, then 2,3, ••• in order down the first column:

b
X X.•••• X * * *
X

[a] ...
X

*
The second result follows.

21.4 THEOREM (A special case of the Murnaghan­Nakayama Rule).

Suppose that p is an n­cycle, and v is a partition of n. Then

XV(p) = {o(_l)n­x if [v] = [x,ln­x]

otherwise

Proof: Let [u] and [B] be diagrams for G with 0 < r < n.
r
Then the character inner product
(x[uJ[BJ, x(n)­(n­l,1)+(n­2,12)­ ••• ± (1 n »

is zero, since [a][BJ contains two adjacent hook diagrams, each with
coefficient 1, or no hooks at all by Lemma 21.3.
 81

(n)-(n-l,l)+ ••• ± (In)

By the Frobenius Reciprocity Theorem, X
restricts to be zero on all Young subgroups of the form G(r,n-r) with
o < r < Gn ,
in particular, i t has value zero on all classes of
except perhaps, that containing our n-cycle p. Therefore, the column
n-x (x In-x)
vector which has (-1) opposite X ' and 0 opposite all other
characters is orthogonal to all columns of the character table of Sn'
except that associated with p. Since the character table is non-smgu-
lar, this column vector must be a multiple of the p-column. But the
entry opposite x(n) is 1. Therefore, it is the p-column, as required.

Remark: Theorem 21.4 can also be proved using the Determinantal Form,
but the above proof is more elegant.

21.5 LEMMA Suppose that A is a partition of n and v is a partition

of n-r. Then
r-x ] is zero unless [A]\ [v]
(i) The multiplicity of [A] in [v][x,l
is a union of skew-hooks.
(ii) The multiplicity of [A] in [V][x,lr-x] is the binomial coeffi-
cient (m-l) if [A] \ [v] is a union of m disjoint skew hooks having (in
c-x - -
total) c columns (and r

Proof: The Littlewood-Richardson Rule assures us that the diagram [A]

appears in [V][x,lr-x] if and only if [v] is a subdiagram of [A] and we
can replace the nodes in [A] \[v] by x lIs, one 2, one 3, ••• , one (r-x)
in such a way that

(i) Any column containing a 1 has just one 1, which is at the top
of the column.
(ii) For i > 1, i+l is in a later row than i; in particular, no
two numbers greater than 1 are in the same row.
(iii) The first non-empty row contains no number greater than 1.
(iv) Any row containing a number greater than 1 has it at the end
of the row.

Suppose that the multiplicity of [A] in [v][x,lr-x] is non-zero.

Then [A] \ [v] does not contain four nodes in the shape

x X
X X

since neither left hand node can be replaced by a number greater than 1
(by (iv»; nor can they both be replaced by 1 (by (i». Therefore,
[A] \ [v] is a union of skew hooks.
Suppose that [A] \ [v] is a union of m disjoint skew-hooks, having
 82

c columns. When we try to replace the nodes in [A] \ [v] by numbers, we

notice that certain nodes must be replaced by l's and others by some
numbers b > 1, as in the following example

1 1
b
1 1 X
1 1 b
b
1 1 X c = 11, m 4

b
X
b
b

Each column contains at most one 1 (by (i». Also, each column
contains at least one 1, except the last column of the 2nd, 3rd, ••• ,
mth components (by (ii) ,(iii) and (iv». Therefore, (c-m+l) l's are
forced. There remain (x-c + m-l) l's which can be put in any of the
m-l spaces left at the top of the last columns in the 2nd, 3rd, ••• ,mth
components. The position of each number greater than 1 is determined
by (ii) once the l's have been put in. The multiplicity of [A] in
[\I][x,lr-x] is therefore (m-+l
x-c m-
1) = (m-l) , as we claimed.
c-ix

Proof of the Murnaghan-Nakayama Rule:

Let a
vu
= (X A + G
(n-r,r),
where is a partition of r
and v is a partition of n-r.
If p is an r-cycle and is a permutation of the remaining n-r
numbers, then

v r
L: X r a , (x,lr-x) (-1) , by 21.4.
\I x=l v
x [\I] [x lr-x]
But (X , X ' ) by the Frobenius Reciprocity
Theorem
= (m-l)
o-x b y Lemma 21 •5•

The definitions of m and c give r c m, so

(m-l) (-1) r-x + •••

x=l c-x
 83

However, when m = 1, [A] \ [v] is a single skew r-hook of leg length

r-c. Therefore,
xA(np) = L {(-l)ixV(n) I [A] \ [v] is a skew r-hook of leg length i},
v
which is the Murnaghan-Nakayama Rule.

21.6 COROLLARY
Suppose p is a prime. If no entry in the hook graph
for [A] is divisible by p, then x A is zero on all permutations whose
order is divisible by p.

Proof: The hypothesis shows that no skew kp-hook can be removed from
[A], so the Murnaghan-Nakayama Rule shows that XA is zero on all permu-
tations containing a kp-cycle (k > 0).

Remark The hypothesis of Corollary 21.6 is equivalent to the statement

A
that I / deg x is coprime to p, by the Hook Formula. The Coroll-
ary therefore illustrates the general theorem that if X is an ordinary
irreducible character of a group G and IGI / deg X is coprime to p,
then X is zero on all p-singular elements of G. (In the language of
modular theory, X is in a block of defect 0.)

The Murnaghan-Nakayama Rule can be rephrased in a way which is use-

ful in numerical calculations, especially in the modular theory for

21.7 THEOREM If v is a partition of n-r, then the generalised charac-

ter of corresponding to

L {(-l) i[A] I [A] \ [v] is a skew r-hook of leg-length i }

is zero on all classes except those contaiQing an r-cycle.

Proof: Suppose that [A] is a diagram appearing in

[v]([r] - [r-l,l] + [r-2,12] - ••• ± [lr]).

Then, by Lemma 21.5, [A] \ [v] is a union of m disjoint skew hooks and
its coefficient is
f (m-l) (-1) r-x
x=l c-x

As before, this is (_l)r-c if m = 1, and zero if m 1. Therefore

[v]([r] - [r-l,l] + [r-2,12] - ••• ± [lr])

z {(-l)i[A] I [A] \ [v]
is a skew r-hook of leg length i}.
But, by definition, XV X(r)-(r-l,l)+ ••• ± (l r) t is zero on
"'n
all of G'n except the subgroup G(n-r,r). However, it is zero even
here, except on (p an r-cycle), by Theorem 21.4.
 84

Remark: The proof shows that"the operator [r]· - [r-l,l]·+ ••• ± [lrJ·
wraps skew r-hooks on to the rim of a diagram".

21.8 EXAMPLES (i) Hhen \I = (3,2) and r = 3

x X X ••• X X X • X X.X, X X X
+ ............ . +
X X X X X X X X

shows the ways of skew 3-hooks on to [3,2]. The generalised

2,1)
c h arac t er X
(6,2)
- X
(4)
-
x(3,2 + X(3,2,13) is zero on all classes
of G except those containing a 3-cycle.
8

(ii) For n 4 , X (n) + X (n-2,2) - X (n-2,12)4s

• zero on all classes
of Gn except those containing a 2-cycle.
2) 2,1)
These examples show that X(6,2) + X(3,2,13) = x(4 + x(3,2 as
a 3-modular character, since this equation holds on 3-regular classes,
and x(n-2,12) = x(n-2,2) + X(n) as a 2-modular character. At once, i t
follows that x(n-2,12) x(n-2,2) and x(n) are in the same 2-block of
G'n. Also, X(6,2), X b,2,13), X(4
2)
and X(3,2
2,1)
are in the same 3-
block of G'8' since

A
21.9 THEOREM
A X = 0 be a non-trivial relation between
Let a
characters on p-regular classes. Then a is non-zero for some p-
A
singular A, and if a is non-zero for just one p-singular A, then all
A
the characters with non-zero coefficients are in the same p-block.

Proof: If the only non-zero coefficients belong to p-regular partitions,

consider the last partition whose coefficient is non-zero. The
character contains a modular irreducible character corresponding
to the factor of By Corollary 12.2, is not a constituent of any
other ordinary character in our relation, and this contradicts the fact
that the modular irreducible characters of a groupare linearly Indepanderrt ,
If the partitions with non-zero coefficients lie in more than one
p-block, then there are two non-trivial subrelations of the given one,
and each subrelation must involve a p-singular partition, by what we
have just proved. The Theorem now follows.
Although it is fairly easy to prove that all relations between the
ordinary characters of Sn' regarded as p-modular characters, come from
applying Theorem 21.7, there seems to be no way of completely determin-
ing the p-block structure of G along these lines.
n
 85

21.10 EXAMPLE It is an easy exercise to prove from the Murnaghan-

Nakayama Rule that when n = 2m is even
(n) (n­l,l) + (n­2,2) ± X{m,m)
X ­ X X ­
is zero on all classes of containing an odd cycle. Hence
en) (n­l, 1) (m,m)
X,X ,··.,X are all in the same 2­block of G2m, by
Theorem 21. 9.
This is a convenient point at which to state

21.11 THEOREM ("The Nakayama Conjecture"). and SA are in the

same p­block of G if and only if there is a (finite) permutation 0
n
of {1,2, ••• } such that for all i

1i - i modulo p.
- u Lo _­­=i""o_ _.::;;,;;;.;;;.:o:=;....o;.

We do not prove the Nakayama Conjecture here ­ the interested reader

is referred to Meier and Tappe [17] where the latest proof and refer-
ences to all earlier ones appear. It seems to the author that the
value of this Theorem has been overrated1 it is certainly useful (but
not essential) when trying to find the decomposition matrix of for
a particular small n, but there are few general theorems in which it is
helpful. In fact, there is just one case of the Nakayama Conjecture
needed for a Theorem in this book, and we prove this now:

21.12 LEMMA If n is odd, sen) and S(n­l,l) are in different 2­blocks

of (;
--n
•

Proof: Let n = (l 2) (3 4) ••• (n­2,n­l). Then I 6n l is odd, where Gn

is the conjugacy class of containing n. But x{n) en) =1 and
X(n­l,l) ( n ) ­­ 0 , b y Lemma 6 • 9 • Th ere f ore,

I x{n) en) I &n l x{n­l,l) (n)

mod 2.
x{n) (1) x(n­l,l) (1)

General theory (see Curtis and Reiner [2], 85.12) now tells us that
sen) and s(n­l,l) are in different 2­blocks.

The proof we have given for the Murnaghan­Nakayama Rule has been
designed to demonstrate the way in which skew­hooks come into rlay.
The Rule can also be deduced from the Determinantal Form, and we conclude
this section with an outline of the method.
21.12 LEMMA Suppose that np E: G where p is an r­cvcle and n is a
n
permutation of the remaining n­r numbers. Let be a
partition of n. Then

1=1------------------
 86

Proof: = the number of fixed by np

n
= (the number of fixed by n in which all the numbers
moved by p lie in the ith row), since a is fixed by p if and
only if each orbit of p is contained in a single row of the tabloid.
n
(the number of fixed

As usual, [kJ is taken to be zero if k < 0, and X (n) = O.

o
21.14 EXAMPLE (cf. Example 21.2(i». Suppose that np E S13 where
p is a 5-cycle and n is a permutation of the remaining 8 numbers. Then
[5] [6] [7J
X{5,4,4) (np) = the character of evaluated at np, by the
[3J [4J [5J
[?] [3] [4J
Determinantal Form

[OJ [lJ [2J [5J [6J [7J [5J [6J {7J

[3J [4J [5 J + [-2J [-1] [OJ + [3J [4J [5J at n, by
[2 ] [3J [4J [2J [3J [4] [ -3 J [ - 2 J [ -1 J Lemma 21.13

[3J [4J
[SJ
[2 J [3J [4 J
I [5J
[2J
[6J [7]
[3J [4J at n
[OJ [lJ [2J [-2J [-lJ [OJ
(X(3,3,2)
X (5,3,0»( 'T1" ) , by the Determinantal Form.

By inspecting the above example, the reader will see what is required
to prove the Murnaghan-Nakayama Rule from the Determinantal Form, and
should have no difficulty wdth the details.
 87

22 BINOMIAL COEFFICIENTS

In the next couple of sections, we shall put our mind to the repre-
sentations of G over a field of finite characteristic p. Many of the
n
problems which arise depend upon deciding whether or not the prime p
divides certain binomial coefficients, and the relevant Lemmas are
collected in this section.

22.1 DEFINITION Suppose n = no + nlP + ••• + nrpr wnere, for each i,

o n
i
< p and n
r
O. Then let

(i) vp(n) max {iln j = 0 for j < i}

(ii) 0p(n) = n0 + n l + ... + ll
r
(iii) (n) r + 1.

For a positive rational number n/m, let vp(n/m) v (n) ­ v (m) ,

p p
v,e do
not define vp(O), but we let a (0) = (0) = O.
P P
22.2 LEI·1MA

Proof: The result is true for n = 0, so we may apply induction. If

n = pr, then vp{(pr­l):} = (pr­l­rn+r)/(p­l), by incuchn)'l. Bnt v,..,(pr:)
= r+v r:' {(Dr_l):}
.
=
(pr_1)/(n­l), and the res1.lJt i5 true in this case.
_. r r+' r
AssuMe, therefore, that 0 < n­p < p. - pro Since v (p + x) =
r+l r p
vp(x) for 0 < x < P ­ P
r
v {n (n­l) ••• (p + l)} = v (n_pr):}
p p{
'llErefore \ip(n:) Vp(pr:) + \i {(n _ pr):}
p

0p (n) + 1) / (p­l) ,
by induction, and this is the required result.

22.3 a)
Assume a b > O. Then v (b < (a) ­ v (0).
p -p--p-

Proof: We may apply induction on a, since the result is true for

a = 1.
If pI b, let n' = b/p and a' = (a­ao)/p, where 0 a < p and
o
a ­ a o modulo p. Using the last Lemma, we have

{Op(b) + 0p(a­b) ­ 0p(a) }/(p ­ 1)

{Op(b') + 0p(a'­b') ­ 0p(a')}/(p ­ 1)

a'
vp(b')
a'
But v
p
(b') <
p
(a') ­ Vp(b'), by induction, and = p
(a') + 1 and

Vp(o) = Vp(b') + 1, so < ­ Vp(b), in this case.

 88

a+b-l a
suppose that vp(b) = O. Since b (b-l) ,

= vp(a-b+l) + Vp(b:l).

Because the result is true for b = 1, we may assume that b > 1, and
a Hence, unless v (a-b+l) > 0,
v (0-1) < R, (a) - v (b-l).
p P P p
a
vp(b) < R,p(a).

But if v (a-b+l) > 0, then

p
a
Vp(b:l) = v p ( a- b + 1) < R, p (a) - v P (a-b+l),

by the first paragraph of the proof. R, (a)

p
- V (0) in this case also.
p

22.4 Assume that

a o + a 1P_+---,,-,-,,_+_arPr (0
"
b = b ° + b 1P---,-+--=-...;..:.--,-+-,b r Pr (0
" b, J.

The n ( .2.- ( i)_._._._ In particular, p divides (a)

.....:.:m",o;.::d:.,::u:.,:l:.,::o,--,,:..:....--::.:..:....-::.::::.::...:=.;::..=.;::..=.;::..t.......:;;..--::;::;...:..,=.:;.o;._
P.
b
if and only if a i for some i.

Proof: As a polynomial over the field of p elements, we have

(x+l)a = (x+l)aO(xP+l)al ••• (xpr+l)ar •
Comparing coefficients of x b, we obtain the result.

22.5 COROLLARY Assume a b 1. Then all the binomial coefficients

(ba ) ' (a-l)
b-l , ••• , (a-b+l)
1 are J.VJ.SJ. e b y p J.'f an d on 1 y J.' f
d'" bl

a-b = -1 mod pR,p(b).

Proof: By considering Pascal's Triangle, p divides all the given

binomial coefficients if and only if p divides each of
( a - b1 +l ) ' (a-b+l)
2 , ••• ,
(a-b+l )
b •

Then the last sentence of the Lemma gives our result.

 89

23 SOME IRREDUCIBLE SPECHT MODULES

The Specht module is irreducible over fields of characteristic

zero, and since every field is a splitting field for is irre-
ducible over field of prime characteristic p if and only if it is
irreducible when the ground field has p elements. This then, is the
case we shall investigate and, except where otherwise stated, F is the
field of order p in this section. The complete classification of irre-
ducible Specht modules is still an open problem, but we tackle special
cases below.

23.1 LEMMA Suppose that ­ F. Then is irreducible

if and only if is self dual. n

Proof: If is irreducible, then it is certainly self­dual (since its

modular character is real.)
Let U be an irreducible submodule of If is self­dual, then
there is a submodule V of with = U. Since
U
canon iso
gives a non­zero element of we must have U = so
n
is irreducible.
The hypothesis HomF F cannot be omitted from this Lemma
n
(see Example 23.1diii) below), but Corollary 13.17 shows that the
hypothesis holds for most Specht modules.

.
Before applying the Lemma, we want a result about the integer
defined in 10.3 as the greatest common divisor of the integers
,
et,e t * > where e t and e t * are poly tabloids in being the par-
<
,
tition conjugate to and < , >' being the bilinear form on
Remember that Kt = E (sgn n)n. Let P E n.
n ct t n Rt

23.2 Let the ground field be Q, and t be a Then

(i)
The greatest common divisor of the coefficients of the tabloids
involved in {t}KtP is 9
t
(ii) {t}KtPtK t = n (hook lengths in [pJ)

Proof:
,
(i) By definition, 9 = g.c.d. < e ,e >' as the permuta­
t, t,
tion nvaries. But

E {sqn n sgna s qn r [o , r C 0' n r R

t" t,
E {sgn w It Ct', w t- l n­ l C
t',
w R
t,}
 90

{sgn wi, E R
t
' w,- 1 n -1 E Rt, wE Ct }

-1
< {t},{t}KtP >
t n
< {th, {t}KtP
t >

and result (i) follows.

(ii) Corollary 4.7 shows that {t}KtPtK = C{t}K for some CEQ.
t t
To evaluate c, it is best to consider the group algebra Q (See the
remarks at the end of section 4). We have PtKtPtK t = CPtK t •
The right ideal PtK of Q (which is isomorphic to has
t
a complementary right ideal U, by Maschke's Theorem.
Multiplication on the left by PtK
gives a linear transformation of
t
Taking a basis for PtK t Q Sn' followed by a basis of U, this
linear transformation is represented by the matrix

{
c

0
c
dim
(!

* 0
On the other hand, taking the natural basis {n j n E G'n} for Q.G'n'
the linear transformation is represented by a matrix with l's down the
diagonal, since the identity permutation occurs with coefficient 1 in
the product PtK t•
A comparison of traces gives c dim = By the Hook Formula for
the dimension of c = ll(hook lengths in

Since {t}KtP t n = {tn}K t n P t n, the first part of the Lemma and Coro-
llary 8.10 show that we may give:

23.3 DEFINITION Suppose that F is the field of p elements. Let e be

the non-zero element of in given by
1
e: {t} -+ (-, {t}KtP
t) p

where this means that the image of {t} is obtained from the vector
{t}KtP t in by reducing all the tabloid coefficients modulo p.
g

23.4 THEOREM
(i) If 1m e c equivalently if Ker e J then is reducible.
(ii) If 1m e = equivalently if Ker e and if
HOID-
- t" n F, then
=--..I.''':::''....I-_-=-L._
is irreducible.
 91

Proof: If F the the homomorphism defined by

=
sends {t}K to a non-zero multiple of itself, by Lemma 23.2(ii). There-
t
fore dim Ker = dim , and by the Submodule Theorem, Ker =
By Lemma 8.14, Ker 8 when we work over the field of p elements.
Therefore, Ker 8 if and only if 1m 8 c

The first part of the Theorem is now trivial, since 1m 8 is a proper

submodule of in this case.
If Ker 8 = then 8 gives an isomorphism between and
and result (ii) follows from Lemma 23.1.

23.5 THEOREM Suppose that is p-regular. Then is reducible if

and only if p divides the integer
{II (hook lengths in

Proof: The last Theorem and Corollary 13.17 show that is reducible
if and only if Ker 8 But, since is p-regular, has a
unique minimal submodule + (by Theorem 4.9). Therefore,
is if and only if Ker 8 •
1
But {t}K t 8 (gW {t}KtPtKt)p

(II(hook lengths in {t} K )

g t P
by Lemma 23.2 (ii). Since is a cyclic module, is reducible if
and only if p divides the integer II(hook lengths in l u I) •
g

23.6 EXAMPLES (i) If p does not divide II(hook lengths in then

(u is p-regular and) is irreducible. This is just the case where
is in a block of defect a (cf. The Hook Formula).

(ii) If both and are p-regular, then from Corollary 10.5, p

does not divide g Thus is reducible if and only if p divides
II(hook lengths in For instance, is reducible of = «p_l)x)
where 1 < x < p.

(iii) If = (3,2) and t 1 2 3 , then direct computation shows

4 5
1
 92

The g.c.d. of the edge coefficients is 4, so = 4. But the pro-

duct of the hook lengths in is 24, so is reducible if and only
if char F = 2 or 3. When char F 2, {t}8 is the vector called r in
hxample 5.2, and when char F 3, {t}e = ­r(­4) ­ r(­5).

23.7 THEOREM Suppose that is a hook partition, and let be defined

over the field of p elements. Then is irreducible if and only if
one of the following holds:
(i) \l (n) or (In)

(ii) p t nand \l (n­l,l) or (2,ln­2)

(iii) p { nand p 2.

Proof: Since s(n) and s(ln) have dimension 1, they are certainly irre-
ducible. Thus, we may assume that = (x,lY) with x > 1, Y > 0 and
x + y = n.
1 (y+2) ••• (y+x)
2
Let t

(y+l)

and let K l: {sgn 0)0 a E

{2, 3, •.• , y+ l } } . Then
t

K = (1 ­
t
(12) ­ (13) - ••• - (l,y+l» Kt
For the moment, work Then

Therefore,
y:{t}Ktpt(l ­ (12) ­ ••. ­ (l,y+l» = {t}KtPtK
t
II(hook lengths in by 23.2
(x ­ 1): y:(x+y){t}K
t
I
But g (x ­ 1): by Lemma 10.4, and so
1
­ (12) ­ ... ­ (1, (y+l» = (X+y){t}K
t
g
Let e be the of definition 23.3. Then

{t}(l ­ (12) ­ •.. ­(l,y+l»e = (x+y){t}Kt'

where we are now working over the field of p elements. This shows that
if pIn, 1m e = Therefore,

23.8 If P { n, s(x,lY) is self­dual.

 93

But Hom
F
F if P 2 or if = (n-l,l), by Corollary
13.17. Using Bemma 23.1, is irreducible in the cases where p f n
and p 2 or = (n-l,l) (also when = (2,1 n-2 ), by Theorem 8.15).
Next suppose that pin. Then
{t}(l - (12) - ••• -(l,y+l» Ker e
(y+x) (Y+x - 1) ••• (y+2) 1
2
Let t*

(y+l)
Since x > 1, all the tabloids in e t* have 1 in the first row. Hence
{t} {t*} is the unique tabloid involved in both e t* and
{t.} (1 - (12) - .•• - (l,y+l», and so

< {tll - (12)- •.• -(l,y+l», e t* > = 1-

Therefore, {t}(l - (12) - ••. -(l,y+l» Ker e , and Theorem 23.4
proves is reducible in this case, where pin.
Finally, we prove that is reducible when =
(x.1 Y) with x > 1,
Y > 1 and p = 2. By Theorem 8.15, we may assume that x y. Observe
that
[x+yJ + [x+y-l,lJ + •• +[x,yJ

and [x+l,ly-1J + [x,lYJ

by the Littlewood-Richardson Rule. But when P -- 2, x(y) and X(lY) are

the same 2-modular character, and thus
X(X+l,ly-l) + X(X,lY) = X(x+ Y) + x(X+y-l,l) + ••• + X(x,y)
as a 2-modular character. Whence, by induction,
X (x,lY) = X(x,y) + X(x+2,y-2) + X(x+4,y-4) + •••

and so X(x,lY) is certainly a reducible 2- modular character.

Remark: The last part of the proof shows that

(n) , (n-2,2), (n-4,4),... are in the same 2-block,
and (n-l,l),(n-3,3),(n-5,5), ••• are in the same 2-block of
Gn (see Theorem 21.9). When n is even, all the 2-part partitions of
n are in the same 2-block of since Example 5.1 proves that (n)
and (n-l,l) are in the same 2-block (see also, Example 21.10). When
n is odd, the 2-part partitions of n lie in two different 2-blocks,
since Lemma 21.12 shows that (n) and (n-l,l) are in different 2-blocks.

Theorem 23.7 will help us in our first result in the next chapter
on the decomposition matrices of G For hook partitions, is easy
n•
to calculate; unfortunately, this is not the case for other types of
 94

partition, for example:

23.9 LEMMA If H; (x,y), then

9
I
; y: g.c.d. {x!, (x-l):l:, (x-2):2:, .. .,(x-y):y:l

Proof: Let t and t be Let

l 2
X ; {klk belongs to the ith column of t and to the jth column of
i j l
t }
2
t

The polytabloids e and et in have the tabloid {t in common

t l 2 3}
if an only if no two numbers from anyone of the sets Xl l U X1 2 '
X U X X U X X U X are in the same row of {t Any row
2l 2 2, ll 2 l' 12 22 3}.
of {t must contain a number from X and a number from X or no
3} 12 2l
numbers from X1 2 U X • Therefore, < etl,et2 > ; 0 unless Ix121 ;
2l
Ix 21 l ·
Suppose now that Ix121 ; Ix2ll. The tabloid {t 3 } is common to e
tl
and e if and only if each of the first y rows of {t } is occupied
t2 3
by just one number from X U X and each row containing a number from
2l 22
X2 1 contains a number from X Thus, etl and et2 have
1 2•
y! I x 1 2 1: (x - IX121): common tabloids.
Assume that the tabloid representative t for the common tabloid
3
{t has been chosen such that t ; tln for some n in the column sta-
3} 3 l l
bilizer of t
l•
Let °
be the permutation in the row stamlizerof t
3
interchanging each number in X with a number in X leaving the other
12 2 l'
numbers fixed. Then t 0 = t for some n in the column stabilizer of
3 2n 2 2
t
2,
and sgn = °
(-1) IX121. Therefore, t
l
nlo n;l = t
2,
and (sgn n
l)
(sgn n 2 ) depends only on t and t and not on {t But {t {tl}n
l 2 3}· 3} l
= {t 2}n 2, and hence
< etl' et2 > ; ±y: I x121: (x - I X1 2 ' ) :

By definition, is the greatest common divisor of such integers,

and, since 0 IX1 21 y, the Lemma is proved.

23.10 EXAMPLES
,
(i) If u = (5,2), then g ; 2: g.c.d.(5:,4:1:,3:2:)= 2 '.3. But
IT(hook lengths in = 2 3,3 2,5. Therefore, s(5,2) is reducible if and
only if the grand field has characteristic 3 or 5.
 95

(ii) Similarly, s(5,3) is reducible if and only if the ground field

has characteristic 2 or 5.

(iii) If P 7, s(5,12) is self-dual, by 23.8. Now let the ground

field have characteristic p = 2. Then the first proves s(5,2)
is irreducible, and Example 21.8(ii) shows that s(5,1 ) has composition
factors isomorphic to s(5,2) and s(7). Since s(5,12) is self-dual, these
2)
factors can occur in either order, and so 5(5,1 is decomposable over a
field of characteristic 2.

The last Example that the hypotheses cannot be omitted in

13.17, 13.18, 23.1 or 23.4.

23.11 DEFINITION The p-power diagram for is obtalned by rep-

lacing each integer h;J' in the hook graph for by v (h .. ).
. p

23.12 EXAMPLE If = (8,5,2), then the hook graph is

10 9 7 6 5 3 2 1
6 5 3 2 1
2 1

0 2 0 1 0 100
and 1 0 1 0 0
0 0

1 0 0 1 0 010
1 0 0 1 0
1 0

We now classify the irreducible Specht modules corresponding to

2-part partitions.

23.13 THEOREM Suppose = (x,y) is p-regular (i.e. if p = 2, we assume

x y). Then defined over the field of p elements, is reducible if
and only if some column of contains two different numbers,

Proof: The hook lengths h are given by

i j for
h x - j + 2 for 1 s j s Y
1j
1j = x - j + 1 for y < j s; x
h
h2j Y j + 1 - for 1 s; j s y.

If there is a j with ) , consider the largest j with this

V
p (h l j) V (h
p 2j
r
property and let V (h ) = r. Then j + p s; y + 1 and
p 2j
r
v (h ) = V (h < r for j + 1 s; i < j + p
p li p 2i)
 96

But {hli!j S i < j + pr} is a set of pr consecutive integers, so

\) (h r = v (h Since \) (h \) (h we have \) (x - j+2) >
p l,)
J p 2·,).
J p l,)
J P 2,),
J P
\) (y - j+l). Writing b = x - j+2 and noting that \)p(b) > \)p(b - x+y - 1)
P
if and only if \) (b) > \) (x - y+l), this proves
p p

23.14 Some column of [x,yJ P contains two different numbers if and

if there is an integer b with x - y+2 S b S x+l and \) (b) > \) (x - y+l) •
P P
Now, ll(hook lengths in [x,yJ) = (y: (x+l) :)!(x - y+l) and
g
I
= y: g.c.d.{x:,(x - l):l!, ••. ,(x - y):y!} by Lemma 23. 9, so Theorem
23.5 proves that is reducible if and only if p divides

x+l x x x }
x - y+l Lv c s m, {(x), (x-l)'···' (x-y)

Since (x+l) = b(X:l) ,

23.15 s(x,y) is reducible if and only if there is an integer b with

b x+l
x - y+l S b S x+l and \)p { x _ y+l ( b )} > 0 .

Comparing 23.14 and 23.15, we see that S(x,y)is reducible if some

column of [x,yJP contains two different numbers.
On the other hand, suppose that no column of [x,yJ P contains diffe-
rent numbers. Then, for every b with x - y+2 S b s x+l,
v (b) s v (x - y+l) .
P P
Let r r+l s
x - y+l arP + ar+1P + ••. + asp
(0 s a < p, a 0 as).
i r

Then x - y+l < (a r+l r+2

r+1 + l)p + a r+ 2P +

and \)p«a + l)pr+l + ... + a pS) > \) (x - y+l). Thus our supposi-
r+1 r+l s p s
tion gives x+l < (a
r+1 + l)p + .•. + asp Therefore

x+l r r+l s
Co + c1P + •.• + crP + ar+1P + •.• + asp
(0 s c < p)
i
and if x - Y + 1 s b s x+l, then
b = b q pq + b
q+lP
q+l + ••• + brP r + ar+1P r+l + .•• + asp s
(0 s b < P, b 0).
i q
Therefore, x+l .,;, b q-vL q r
Co + c1P + ••• + c q_ 1 P + dqP + + drP
(0 S d < p)
i
 97

By Lenuna 22.2,
(x+l)
\l p b °
{Op(b) + p (x+l - b) - o (x+l)} /
P
(p - 1)

(b + ••• + b + d + + d - c
q
- cr)/(p - 1)
q r q r
c pq + r
\I
p + crP )
(bqpq + + b Pr
q r
r - q, by Lenuna 22.3 (since b 0)
q
\lp(x - y+l) - \l (b).
p
Therefore, for x - y+l S b S x+l, \lp{x b (x+l)} $ 0
- y+l b
and s(x,y) is irreducible,as required.

23.16 EXAMPLE s(2p-l,p) is irreducible over the field of p elements

if and only if p 2 (cf. Example This is interesting because
an earlier author believed, apparently on the evidence of the case p = 2,
that s(2p-l,p) always has two composition factors, one being the trivial
module D(3p-l). Since dim s(2p-l,p) = 1 mod p3 for p odd - this follows
from the Hook Formula - the mistake would have provided counterexamples
to a conjecture of Brauer which states that \I (!GI/dim D) 0 for each
p
p-modular irreducible representation D of a group G.

R.W. Carter has put forward

23.17 CONJECTURE No column of [\lJ P contains two different numbers if

and only if \l is p-regular and is irreducible over the
elements.

It is trivial that has a column containing two different numbers

if is p-singular. The author [llJ has proved that the given condition
is necessary for a p-regular Specht module to be irreducible, and has
proved it is sufficient in the case where p = 2.
Over the field of 2 elements, it turns out that S(x,x) is irreducible
if and only if x = 1 or 2 (This is the only 2 part partition not consi-
dered in Theorem 23.13). We conjecture that (2,2) is the unique parti-
tion such that is irreducible over the field of 2 elements but
neither nor is 2-regular.
24 ON THE MATRICES OF

There is no known way of determining the composition factors of the

general Specht module when the ground field F has characteristic a prime
p. Thus we cannot decide the entries in the decomposition matrix of
which records the multiplicity of each p modular irreducible repre-
A
sentation D (A p­regular) as a composition factor of except in
some special cases. The theorems we expound give only partial results.

24.1 THEOREM (Peel [18]) Suppose p is odd.

(i) If P t n, all the hook representations of G n remain irreducible

modulo p, and no two are isomorphic.

(ii) If p n, part of the decomposition matrix of is

o
(n) 1
(n­l,l) 1 1
(n­l,12) 1 1

o 1 1
1

Proof: The result is true for n = 0, so we may assume that it is true

for n ­ 1. Note that
y y­l
X(x,lY) oj. <5 X (x­l , 1 ) + X(x, 1 )
if x > 1, Y > 0,
n­l
x+y = n.
Case (i) P does not divide n.
In view of Theorem 23.7, we need prove only that no two hook repre-
sentations are isomorphic. But this follows at once, since they have
non­isomorphic restrictions to G n­ 1.
Case (ii) p divides n.
(x lY)
Suppose x > 1, Y > O. Then by restricting to
l' X ' has at
n­
most two modular constituents, and therefore precisely two, by Theorem
23.7. Let be the modular constituent of X(x,lY) satisfying

oj. Gn ­ l = x(x­l,lY) and be that satisfying oj. X(x,ly­l)

(and let ¢n =0 and = 0). We must show that for every x, =

no other equalities can hold because there are different restrictions
to G'n­l.
The following relation between characters holds on all classes
except (n), in particular on all p­regular classes:
 99

(n) _ X(n-l,l) + (n-2,12) _ + (In) _ 0

X X ···-x -.
(This comes from Theorem 21.7 or direct from Theorem 21.4, by using the
ordinary character orthogonality relations).
In terms of modular characters; we have
'"+ - ('" - + 4>n-l)
+ -
+ (4)n-2 +)
+ 4>n-2 ••• ± 4>1 =0.
- +-
If some 4>X-l were not equal to 4>x' then 4>x-l would appear just once in
this relation, contradicting the fact that the modular irreducible
characters of a group are linearly independent.

From now on, we shall label the rows of our decomposition matrices
by partitions, and the columns by p-regular partitions. Thus the entry
in the row and Ath column is the multiplicity of n A as a composition
factor of over a field of characteristic p. Omitted entries in
de cpmpos i tion matrices are zero. We write X
for tte p-modular character
of and 4>A for the p-modular character of n A

24.2 EXAMPLE When p = 3, the decomposition matrix of GIS is

(5) (4,1) (3,2) (3,1 2 ) (2 2,1)

(5) 1
(4,1) 1
(3,2) 1 1
(3,1 2) 1
(2 2,1) 1 1
(2,1 3 ) 1
(15 ) 1

Proof: The rows corresponding to (5), (4,1) and (3,1 2) corne from
Theorem 24.1.
Taking [v] and r = 3 in Theorem 21.7, we find that
2,1)
- X(2 + X(2,13) = 0 on 3-regular classes.
are irreducible and inequivalent, by Theorem 24.1.
has precisely two factors. Since one of these must be
follows that
4>(5) + 4>(2 2,1)
2,1)
X (2
3 2,1)
and X (2,1 ) 4> (2 •

The rest of the matrix is similarly deduced from the equation:

X(15) _ X(3,2) + X(4,1) =a on 3-regular classes.

24.3 EXAMPLE When p = 3, the decomposition matrix of is that

given in the Appendix.
 100

(4 2) (2 2
12 )
Proof: First note that X ' and X ' are irreducible by Example
23.6 (L) •
By Theorem 24.1, part of the matrix is

(6) (5,1) (4,1 2 )

(6) 1
(5,1) 1 1
(4,1 2 ) 1 1
(3,1 3 ) 1 1
(2, 1 ) 1 1
(1 6
) 1

Applying Theorem 21.7, with r = 3 and [v] = [3] [2,1] and [1 3] in

turn we get,
(6) + X(3 2) _ X(3,2,1) + X(3,13) a
X
X (5,1) (3 2) (2 3)
- X - X + X a
(4,1 2 3) 6
) _ X(3,2,1) + X(2 + X(1 )
=a
X
on 3-regular classes. These equations, together with
(6) X(5,1) + X(4,12) (3,1 3) (1 6 ) a
X - -X -X -X

enable us to deduce that the remaining two columns above should be

labelled (3,2,1) and (3 2 ) , respectively, and the equations let us write
2), 3)
X(3 X(3,2,1) and X(2 in terms of q,(6), q,(5,1), ••• , in the way
shown in the complete decomposition matrix in the Appendix.

Note that Examples 24.2 and 24.3 have been computed without using
the Nakayama Conjecture, and without resorting to induction (except
where it is implicit in Theorem 24.1). We agree that it is quicker to
deduce the decomposition matrix of from that of using the Bran-
ching Theorem and block theory, but this traditional method of finding
decomposition matrices fails to determine the factors of S(2p-l,p),
even for p = 2 (cf. Example 23.16), and very rapidly leads to further
ambiguities.
It seems to us that if a method is eventually devised for finding
the decomposition matrices of it will include information concerning
n
the order of the factors of each Specht module, as well as the multipli-
cities of the composition factors'. For this line of attack, the most
useful Theorems we know are Theorem 13.13, giving a basis of HomF S
and Corollary 17.18, describing as a kernel
It is unfortunate that these two results look rather ugly, and that the
notation which has to be used obscures the simplicity of their applica-
tion, but we embark upon the task of employing them.
We return to the notation of section 13, where is described as the
 101

space spanned by A-tableaux of The remarks following 17.8 and

17.10 show that the homomorphism acts on by sending a tableau
T to the sum of all the tableaux obtained by changing all but v (i+1) 's
to i's.

1 1 122 11111 1 1 112 1 1 121

+ +
2 3 3 233 133 133

The first result we prove could be subsumed in Theorem 24.6, but we

present the special case to help the reader become familiar with the
relevant ideas.

24.4 THEOREM Over a field of prime characteristic p, has a sub-

module isomorphic to the trivial
-
S n -module s(n) if and only if for all
.....- - - - - -
:: -1 mod -pZ:i. where z. = R. (u , +1)- .

Proof: By Theorem 13.13 (or trivially) there is, to within a scalar

multiple, a unique element 0 in Homp G'n (s(n) T is the semi-
T
standard (n)-tableau of type and 0 sends {t} to the sum of the (n)-
T
tableaux of type .
e.g. if (3,2), then
{t}0T = 1 1 1 2 2 + 1 1 2 1 2 + 1 1 2 2 1 + 1 2 1 1 2 + 1 2 1 2 1 +
122 1 1 + 2 1 1 1 2 + 2 1 1 2 1 + 2 1 2 11+ 2 2 111

Now, the crucial step is that when T is an (n)-tableau of type

l
+ - there are

+ - V)
- v
tableaux row equivalent to T in which all but v (i+l) 's can be changed
to i's to give T
l
e.g. 1 1 1 1 1 comes from tableaux above, by changing all the 2's
to l's, and each of 1 1 1 1 2, 1 1 1 2 1, 1 1 2 1 1, 1 2 1 1 1 ,
2 1 1 1 1 comes from (1) tableaux by changing all except one 2 to 1.

Therefore, {t}8 T belongs to n ker v if and only if each of

v=o '
, ... ,

is divisible by p. This is equivalent to :: -1 pZi where

zi = by Corollary 22.5. Thus, Corollary 17.18 shows that
{t}0T belongs to if and only if this congruence holds for all i 2: 1-

24.5 EXAMPLES (i) 5(8,2,2,1) contains a trivial submodu1e if and only

 102

if the ground field F has characteristic 3.

(ii) 5(5,2) does not contain a trivial submodule if char F = 2.
(iii) s(p-l,p-l, ••• ,p-l,r) contains a trivial submodule if char F = p,
p- l- r)
and r < p. Write n = x(p - l)+r. Then «x+l)r,x is the partition
. x. (n) II
ll' conJugate to II = «p - 1) ,r). HomF (5 ,5) 0, and
Sll s(ln) is isomorphic to the dual of sV' it that
, (In) ,
HomF (Sll,S ) O. By construction, Sll is p-regular, so V is the
n , (In) ,
unique partition of n such that Oll 5 (Remember that Oll is the
unique top composition factor of Slit). Compare Example 24.2, where
5(1 5 ) ;;; 0(3,2).

(iv) Consulting the decomposition matrices in the Appendix, we see

that S(4,2) has a trivial composition factor for p = 2, but s(4,2) does
not have a trivial bottom composition factor, by Theorem 24.4.

It is interesting to see that for any given A and V, we can use

Theorem 13.13 and Corollary 17.18 to determine whether or not
HomFGn(S A,S II) is zero (except in the rather uninteresting case where
char F = 2 and A is 2-singular), for we may list the semistandard homo-
A
morphisms from M into Mll and then test whether some linear combination
of them sends {t}K t into the kernel intersection of Corollary 17.18.
This is a tedious task, but not altogether impossible, even for fairly
large partitions. For example, after a little practice on small parti-
tions, the reader should have no difficulty using the technique of Theo-
rem 24.6 below to prove that HomFG' (SA,s(10,5,3,»= 0 when char F = 3
n
and A = (16,2), (13,5) or (10,8). Using the Nakayama Conjecture, this
proves that S(10,5,3) is irreducible over fields of characteristic 3
(cf. Carter's Conjecture 23.17).
When applying Theorem 13.13 and Corollary 17.18, we are usually
interested in the case where SA is p-regular, since then HomF
A
0 implies that D is a composition factor of Sll. Unfortunately, a
completeclassification of the cases where HomF is non-zero is
not sufficient to determine the decomposition matrix of b ; in
Example 24.5(iv) D(6) is a factor of s(4,2) over the fieldnF of 2 elem-
ents, but HomF ,s(4,2» = O. Even so, sometimes a modification
of the method is good enough to classify all the composition factors of
Sll; see Theorem 24.15 below, for example.
In section 13 we saw that there is much choice in the way we define
a semistandard A-tableau of type V. It turns out that it is often most
useful to consider tableaux where the numbers are non-increasing along
the rows and strictly decreasing down the columns; we shall call such
a tableau reverse semistandard. The second part of the next Theorem
 103

probably classifies all cases where there is a reverse semistandard

homomorphism in A,S Il ). When considering linear combinations

of more than one semistandard homomorphism, the situation becomes

horribly complicated:

24.6 THEOREM Assume that A and Il are (proper) partitions of nand

that char F = p. Suppose that T is a reverse semistandard
of type Il, and let N be the number of i's in the jth row of T.
i j
a' .
i 2 and j . = -1 mod p 1J where
1, N. 1
,J A Il 1- A.I.
aT belongs to HomF c;n (M ,S ) and Ker aT =-
bi'
1, N. 1 . = -1 mOd_p J where
00 1- ,J ..
b .. = then aT is a non-zero
-1J
element

Proof: Since T is reverse semistandard, Ker aT 1 SA by Lemma 13.11

and the Remark following Corollary 13.14. Therefore, Ker aT =- SA.I.
by the Submodule Theorem.
Let t be the A-tableau used to define the GIn action on Mil. Then
{tla T is, by definition, the sum of the A-tableaux of type Il which are
row equivalent to T.
Let i 2, 0 s v S Il
i - 1. Since j=l N 1J
.. = 1l
1.,
r
we may choose
v l,v 2 ' •.. such that 0 S v S N for each j and E v = v. Choose a
j i j j
tableau T l row equivalent T, and for each j change all except v i's
j
in the jth row of T into (i-l) 's. Let T be the resulting tableau.
l 2
By definition, each tableau T involved in {tlaTwi_l,v is constructed
2
in this way, and T 2 appears in {tla Wi-l,v from
T

ooIl
j=l (N.1-,J
1 . + N..
1J - v.)J
N .. - V.
1J J
different tableaux row equivalent to T.
00

Since EN .. Il
i
> v = E v., there is an integer k with
j=l 1J j=l J

a· .
I f for all j Ni-l,j -1 mod p 1J then

is divisible by p, by Corollary 22.5. Thus if the hypothesis of part (i)

of the Theorem holds, Corollary 17.18 proves that MAa
T
=- SIl as required.
Under the hypothesis of part (ii), it again follows that
 104

{t}K t does not involve T 2, except if

i-l
Ni k - v k > L (A + - - L N
m=l k m l s=k ms)
00

But for m < i - 1, T 2 has L N numbers equal to m in rows k,k+l, •••

s-k ms
since T 2 come from a tabIeau row equivalent to T. Similarly, T has
2
at least E N s + N'k - v k numbers equal to i - 1 in rows k,k+l, ••• ,
s=k i- l '
since Ni k - v k i's have been changed to (i-l) 's in row k. Altogether,
therefore, T has at least
2
i-I
l: l: N
ms
m=l s=k

numbers less than or equal to i-I in rows k,k+l,... If we assume

that this excedes iZl A it follows that some column of T 2 cont-
k+m- l,
ains two identi.cal Wlifhbers. Therefore, T is annihilated by Kt . This
2
shows that in part (ii) of the Theorem, Kt = 0 when i 2
and 0 v - 1; thus, {t}K belongs to as we wished to prove.
t8 T
Since is isomorphic to the dual of SA, and SA n is the
unique maximal submodule of SA when A is p-regular we have

24.7 COROLLARY Under the hypothesis of part (i) of Theorem 24.6, every
composition factor of SA is a composition factor of Under the
second hypothesis, OA is a composition factor of if A is p-regular.

There are very many applications of Corollary 24.7. We give just

one, but we shall use the Corollary again later to find all the compo-
sition factors of Specht modules corresponding to 2-part partitions.

24.8 EXAMPLE (o f , Example 24.3) . Let = (3,2,1) and char F 3.

Then all the factors of S (5,1) are factors of take T = 3 2 2 1 1
2
(3 ) 1
0 is a factor of take T = 3 2 2
1 1 1
2
0(4,1 ) is a factor of take T = 3 2 1 1
2
1

Theorem 24.6 also gives

z·
24.9 COROLLARY If for all i 2, ]Ji-l := -1 mod p l where
Z4 = - then is i.rreducible over a field of characteristic
£.:.
Proof: The unique reverse semistandard T of type has
Ni j = - Our hypothesis and the first part of Theorem
24.6 show that 8 T belongs to and Ker 8
T
 105

By dimensions, g].l/S].l.L S].l. The parts of 11 must be strictly decreasing,

so 11 is certainly p-regular. The result now follows from Lemma 23.1.

When p = 2, it is straightforward to verify that the hypothesis

of the above Corollary is equivalent to the statement that no column
of the 2-power diagram [IlJ 2 contains two different numbers; cf. the
comments following the Carter Conjecture 23.17.
To describe another special case of Theorem 24.6, we write 11 i A
if we can obtain [AJ from [].lJ by moving some number d 0 of nodes from
the end of the ith row of [IlJ to the end of the (i-l)th row of [IlJ and
each node is moved through a multiple of p£p(d) spaces. (See Example
24.11) .

24.10 CO ROLLARY
partitions of n with
(l ) k (2) k-l (3) k-2 k-r+2 (r)
].l + 11 +].l + + 11

If 1 a b r and A = ll(b), 11 = ].l(a) then Romp O.

Proof: We may suppose that a 1 and b = r, since otherwise we may

(a) (b)
restrict our attention to the sequence 11 + ••• -s- 11
(k-j+l) j (k-j+2)
Let d. be the number of nodes moved in 11 -r- 11
J
d. = 0 i f j > k or j < k
J
- r + 2) . By construction, for all
i,
(r) (1)
].li lli + di+l - d.
l

and p£p(di) divides lll=i - ll{l) - d + d + 1

i+ l i
(1) (r) (r) (r)
Let Nil ].li - ].li+l and Ni j = ].li+j-l ].li+j for j 2,
and let T be the corresponding ].l(r)-tableau of type 11(1) in Theorem
24.6 (It is simple to verify that T is reverse semistandard).

Now, i-l (r)

l: (].l'+m-l l: N ) = d if j 1, and 0 if j 2.
m=l J s=j ms i

_ (1)
Also, Ni-l,l - ].li-l - ].llr) = ].ll=i - ].lll) - d + d -
i+ l i
so Theorem 24.6(ii) gives the result.

24.11 EXAMPLE Suppose char P 3

4 3 2
+ ->- ->-
x x

x x
x

Therefore, RompS' (SA .s") 0 for A",].l and A,].l any pair from (7,3,1),
11
 106

(5 2,1), (5,3 2 ) and (5,3,2,1). Compare the following 4 by 4 submatrix

of the decomposition matrix of G'll for the prime 3.
D(7,3,l) D(5 2,l)
D(5,3 2 ) D(5,3,2,l)
S (7,3,1) 1
S (52 ,1)
1 1
s(5,3 2 ) 1 1 1
s(5,3,2,l) 1 1 1 1

Note that the number of nodes we raise to the rml above need not be
tne same for each \l(k-j+l) \l(k-j+2) i
in Corollary 24.10; in I?arti-
cular, the Corollary includes the case
(1) il (2) i2 (3) ir-l (r)
\l \l \l ••• \l with i l > i 2>···>i r- l
since we are allowed to raise zero nodes at any stage. The hypothesis
i > i 2 > ... > \.-1 cannot be omitted, since when char F = 2,
l

x X 2 X X X 2 X X X X
X X X
2
and while Hom G (S(4) ,s(3,l)) and HomF G4(s(3,l) ,s(2 ) ) are non-zero
F 2
(by the HomF ,s(2 ) ) is zero (by Theorem 24.4).

For our next Theorem we require

24.12 DEFINITION Given two non-negative integers a and b, let

a = a + alP + ... + a P r a, p, a 0)
(0
" < ;t
0 r 1. r
b b 0 + blP + .•. + b sP s (0 s b.1. < p, b S ;t 0) .

we say that a contains b to base P i f s < r and for each i b, 0 or

1.
b. = a.
1. 1.

24.13 EXAMPLE 65 2 + 0.3 + 1.3 2 + 2.3 3 , so 65 contains precisely

0,2,9 1. 3 2
and 11 2 + 1.3 2 to base 3.

24.14 DEFINITION The function fp(n,m) is defined by fp(n,m) 1 if

n + 1 contains m to base p, and = 0, otherwise.

Since the only composition factors of s(n-m,m) have the form

D(n-j,j) with j " m, by Corollary 12.2, a sensible first step towards
evaluating the decomposition matrix for is to prove

THEOREM (James [6] and [8]). The multiplicity of D(n-j,j) as a

factor of s(n-m,m)-e:=.-=
.::..::;::,;==-.::::..=c....:::...- is f p (n -2J', m- J') •

Proof Since the result is true when n o or 1, we may assmae it for

 107

n' < n. Let t be the (n-j,j)-tableau used to define the Sn action on

M(n-m,m). Let T be the (n-j,j)-tableau of type (n-m,m) having 2's in
the (l,l)th, (1,2)th, .•. ,(1,m)th places. As in the proof of Tneorem
24.6 , the maps defined on M(n-m,m) have the property that
m-l
{t}8-T E n ker If,
Yl,i i.f rr-r.•..\ -J' -= -1 mod p\J (m-r) .
i=r
Also
ker 8 c s(n-j,j)L
T m-l
Therefore, all the composition factors of s(n-j,j) occur in n
i=r
ker
But, by the second isomorphism theorem,
m-l m-l m-l r-l
it;r ker I [!o ker E ( ker + ker

r-l
n
i=o
ker

r-l
_c M(n-m,m) I /"\ k ,I.
I' er 'l'l,i
i=o
m-l
ThUS, every composition factor of " ker is either a factor of
i=r ,1.
m-l r-l
S (n-m,m)
= n
i=o
ker lj!l ,1.' or of M(n-m,m) I n
i=o
ker lj!l '
,1.
By Theorem

17.13 we have:

24.16 If n-m-j = -1 mod then every factor of s(n-j,j) is a

factor of s(n-m,m) or of one of {s(n-i,i) 10 i r-l}

Now suppose that f = 1. Then m j 0 and n-2j + 1

p(n-2j,m- j)
contains m-j to base p. If m > j, then there is a unique integer jl
such that
n-2j+ 1 =
(m-j) + (jl-j) mod p P
l'. (m-rj )

and 0 jl-j < m-j .

But then n-2j + 1 contains jl-j to base p. Hence we may find integers
such that
m j

and n - jk - jk+l = -1 mod

Then, by 24.16 every factor of s(n-j,j) is a factor of s(n-js,js)

or one of {s(n-i,i)jo s i s j-l}. But D(n-j,j) is not a factor of
s(n-i,i) for 0 i j-l, by Corollary 12.2, so D(n-j,j) is a factor of
S (n-js,js) •

Applying 24.16 again, every factor s(n-js,js) is a factor of

 108

s(n-js-l,js-l) or of one of {s(n-i,i) [0 i $ j-l} . Therefore,

D(n-j,j) is a factor of s(n-js-l,js-l). Continuing this argument to
jo = m, we have proved

24.17 When f = 1, D(n-j,j) is a factor of s(n-m,m)

p(n-2j,m-j)

Next, consider the case Wlere

I n =_ m-l mod Then let
r-l
m-l a alP + ... + a (0 ai < p, a r- l 0)
r- 1 P
$
o+
r-l r
so n = a o+ alP + ... + ar-1P + brP + ...
where b
r
= 0 if m = pr Thus, n contains m-l to base p, so
fp(n-l,m-l) = 1. Similarly, fp(n-l,m) 0 and fp(n,m) = 1.
Returning to the case of general nand m, we prove

24.18 If m 1 and f (n-l,m) + f (n-l,m-l) > f (n,m), then there is

some integer j with $ l
j $ m that is a factor of
S (n-m,m) and 1 contains the trivial factor D(n-l) with
n-
multiplicity f (n-l,m) + f (n-l,m-l).
p p
To prove 24.18, consider first the case where m is a power of p,
say m = pro The inequality fp(n-l,m) + fp(n-l,m-l) > fp(n,m) easily
implies that pr divides n + 1, and the argument above proves that
r+l (n-m,m)
p does not divide n-m+l. Therefore, vp(n-m+l) = r. Hence S
-, (n-m m) (n-m m)
is irreducib le in this case, by Theorem 23.13, and D ,. = S '.
Since s(n-m,m) S has the same factors as s(n-m-l,m) s(n-m,m-l)
n-l
by the Branching Theorem, 1 contains D(n-l) with multi-
n-
plicity fp(n-l,m) + fp(n-l,m-l), by the induction hypothesis. This shows
that we may take j = m in 24.18 when m is a power of p.

Suppose, therefore, that m is not a power of p. Since fp(n-l,m) +

fp(n-l,m-l) 1, n contains m or m-l to base p. The fact that m is not
a power of p now shows there is a unique j with
o $ j < m n _ m+j-l mod p£P (m)

Further, j 1, since vre have shown that n == m-l mod p£P (m) implies
that f p (n-l,m) + f p (n-l,m-l) = f p (n,In). Now the above congruence shows
that n + 1 contains m to base p if and only if n+l contains j to base p,
and n contains m to base p if and only if n contains j-l to base p, and
n contains m-l to base p if and only if n contains j to base p. Therefore,
f (n-l,j) + f (n-l,j-l) f (n-l,m) + f (n-l,m-l)
p p p p
> f (n,m) f (n ,j).
p p
By induction, there is an i with 1 $ i $ j < m such that D(n-i,i) is a
factor of s(n-j,j) and Sn-l has D(n-l) as a factor with
 109

multiplicity f (n-l,m) + f (n-l,m-l). But, since n = m+j-l mod

p p (n-' ')
24.16 shows that every factor of S J,J is a factor of s(n-m,m). In
(n-i i) (n-m m) and so 24.18 is proved.
particular, D ' is a factor of S '
The multiplicity of o(n) as a factor of s(n-m,m) is at most
fp(n-l,m) + fp(n-l,m-l), since s(n-m,m)+ G - has o(n-l) as a factor
n l
with this multiplicity, by our induction hypothesis. Further, 24.18
shows that o(n) is not a factor of s(n-m,m) when fp(n-l,m) + fp(n-l,m-l)
> fp(u,m). This proves our next main result, namely

24.19 The multiplicity of o(u) as a factor of s(n-m,m) is at most

fp(n,m) .

Finally we prove
24.20 If j 1, o(n-j,j) is a factor of s(n-m,m) with multiplicity at
most f p(n-2j,m- j).

The way we show this is to consider a subgroup H of S , and find

a modular representation O. of H such that o(n-j,j)+ II as a
factor, but s(n-m,m)+ H 0, as a factor with mUltiPliCityJf (n-2j,
m-j). 24.20 then follows at We should like to choose Sn-2 or
as our subgroup H, so that we can apply induction. Since the
prime 2 is we consider first
Case 1 p is odd.
The ordinary representations of S(n-2,2) are given
by
@,.
s(2)
B
and
(2)@l.
sl )
"" (12)
as varies over partitions of n-2.
Since p is odd, 0 and 0 are inequivalent representations. Hence
the p-modular irreducible representations of S( -2 2) are given by
u (2) (1 2 ) n ,
o 0 0 as varies over p-regular partitions of n-2,
2
and the multiplicity of o(n-j-l,j-l) 0(1 ) as a factor of
2
S (n-m-l,m-l) ® s (1 ) is f (n-2J,m-J)
,.
when J
,
1, by
..
p
Now, by the Littlewood-Richardson Rule, s(n-m,m)+ G(n-2 2) has the
(n-m-l m-l) (1 2 ) ,
same composition factors as S ' S ,together with some
modules of the form s(2). In particular, the multiplicity of
(n-j-l, j-l) (1 2 ) (n-m m) G ' "
o 0 as a factor of S '+ (n-2,2) a s f p (n - 2 J , m- J ) .
2
On the other hand s(n-j,j)+ b has o(n-j-l,j-l) 0(1 )
, (0-2,2)
as a factor with multiplicity one (since f p (n - 2 j , O) = 1), and for i < j
2
s(n-i,i)+ does not have o(n-j-l,j-l) 0(1 ) as a factor
(n-2,2) ( ..)
(sinc;:e . f p(n-2j, i-j) = 0). every factor of S n-J, J , besides
o(n-J,J), has the form with i < so it follows that
D(n-j,j)+ G( has o(n-j-l,j-l) 0(1 ) as a factor with multip-
(n-2,2)
licity. one.
 110

The results of the last two paragraphs prove 24.20 in this case.

Case 2a p = 2 and n is even.

(n-m,m) I. r.- f t S (n-m-l,m) as S (n-m,m-l)
S Y \:>n-l has the same ac ors as "', .
By induction, this contains the factor o(n-j-l,j) with multiplicity
f + f It is simple to verify that this
2(n-1-2j,m-j) 2(n-1-2j,m-j-l).
equals f since n is even.
2(n-2j,m-j),
In particular, for 2J' < n s(n-j,j) ... G has o(n-j-l,j) as a
, n-t
factor with multiplicity one, and for i < j, ... E1n - l does not
nave o(n-j-l,j) as a factor. As before, o(n-J,J) ... Gn-l therefore has
o(n-j-l,j) as a factor with multiplicity one, and 24.20 is proved in
this case too.

Case 2b P = 2 and n is odd.

S (n-m,m) ... G has the same factors as S (n-m-2 ,m) e 2 S (n-m-l,m-l)
n-2
s(n-m,m-2) . This contains o(n-j-l,j-l) with multiplicity
f + 2f + f which equals 2f 2(n-2j,
2(n-2j,m-j+l) 2(n-2j,m-j) 2(n-2j,m-j-l),
m-j) when m-j is even,
Thus s(n-j,j) ... has D(n-j-l,j-l) as a factor with multipli-
, n-2, ,
city 2, and for i j-2, s(n-l,l) ... 6 does not have o(n-j-l,j-l) as
a factor. But every factor of besides has the form
D(n-i,i) with i j-2, by the Remark following Theorem 23.7, so
o(n-j,j) ... has o(n-j-l,j-l) as a factor with multiplicity 2.
n-2
The results of the last two paragraphs prove 24.20 in this final
case.
Now 24.17, 24.19 and 24.20 together give Theorem 24.15.

24.21 COROLLARY If j 1, the multiplicity of D(n-j,j) as a factor

of s(n-m,m) is the same as the multiplicity of o(n-j l,j-l) as a factor
ofS(nml,ml)

By the way, we conjecture that Corollary 24.21 is a special case

of a general theorem involving the removal of the first column.

24.22 EXAMPLE Suppose p = 3. The rows of the following table record,

respectively, n, n+l written to base 3, and the numbers contained in
n+l to base 3, for 0 s n 13.

0 1 2 3 4 5 6 7 8 9 10 11 12 13
1 2 10 11 12 20 21 22 100 101 102 110 111 112

0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 2 1 2 1 2 10 1 2
10 10
11 12
 111

Under n = 13, for example, we have 0,2,10,12 which are integers

to base 3. There are l's in the (O+l)th, (2+1)th, (3+1)th and (5+1)th
places (counting from the diagonal) in the column labelled 13 in the
following pair of matrices. Another example: 10+1 contains 0 and 2 to
base 3. There are l's in the (O+l)th and (2+1)th places of the column
labelled 10.

1 1
1 1 1
1 1 1
1 1 1 1 1 1
1 1 1 1 1
1 1 1 1
1 1 1 1
12 10 8 6 4 2 0 13 11 9 7 5 3 1

The part of the decomposition matrix of Sn corresponding to

2-part partitions for p = 3 and n $ 13 can be read off these matrices
at once. Simply truncate the matrix at the column labelled n, and label
the rows and columns by 2-part partitions in dictionary order.
(9) (8,1) (7,2) (6,3) (5,4)
(9) 1
(8,1) 1 1
e.g. n 9 (7,2) 1
(6,3) 1 1
(5,4) 1 1

For p an odd prime and n small, most of the decomposition matrix

of b n is given by Theorems 24.1 and 24.15.

24.33 EXAMPLE Suppose p = 3 and n = 9. Applying Peel's Theorem 24.1,

the column labels can be found as in Example 24.2 . Alternatively, they
are given explicitly in [9J page 52. Combined with the information
above, this gives

(9 ) (8,1) (7,2) (6,3) (5,4) (7,1 2 ) (6,2,1) (5,2 2) (4,3,2) (42,1)

(9 ) 1
(8,1) 1 1
f] , 2) 1
(6,3) 1 1
(5,4) 1 1
f] , 1 2) 1 1
(6,1 3 ) 1 1
(5, 1 1 1
 112

(4,1 5 ) 1 1
(3,1 6 ) 1 1
(2,1 7
) 1 1
(1 9 ) 1

Applying Theorem 8.15 to the first five rows, another part of the
decomposition matrix is
2,1)
(5,4) (4
9
(1 ) 1
(2,1 7 ) 1 1
2,1 5
(2 ) 1
(2 3,1 3 ) 1 1
,1) 1 1
9 7
(The rows corresponding to (1 ) and (2,1 ) already occur above).
Using Theorem 21.7 we find that the last three columns should be labelled
(4,3,1 2 ) , (3 2,2,1) and (9). Incidentally, we do not know how to sort out
efficiently the column labels once we have taken conjugate partitions
as above (although Theorem A in [9] gives some partial answers).
We have now accounted for 12 of the 16 3-regular partitions
2,1)
labelling columns. 8(5,3,1) and 8(3,2 are irreducible, by Example
so we have two more 3-modular irreducibles to find, namely
those corresponding to (4,2 2,1) and (5,2,1 2 ) . But
2,1)
X(7,2) _ X(4,2 + X(4,2,13)

on 3-regular classes (using Theorem 21.7 with [v] = [4,2]). Appealing

to the theory of blocks of defect 1 (or to the Nakayama Conjecture)
part of our decomposition matrix is

(7,2) (4,2 2 , 1 )
(7,2) 1
(4,2 2,1) 1 1
(4,2,1 3) 1

By taking conjugate partitions, we get

(5,2,1 2 )
(4,3,1 2 ) 1
(2 2,1 5 )
1

Now Theorem 21.7 enables us to complete the decomposition matrix,

since we can write every ordinary character which corresponds to a 3-
singular partition in terms of orilinary characters corresponding to 3-
regular partitions, on 3-regular classes.
 113

When p = 2, Theorem 24.1 cannot be applied. Howe ve z , all the

rows of the decomposition matrix for partitions of the form (n-m-l,m,l)
are known for p = 2 (see James [6J).
Our sources for the decomposition matrices in the Appendix are
Kerber [13J (p = 2,n 9), James [6J(p = 2, n = 10), Mac Aogain [15J
(p = 2,n = 11), Stockhofe [21J (p = 2,n = 12,13), Kerber and Peel [14J
(p = 3, 8 n 10) and l1ac Aogain [15J (p = 3,11 n s 13 ,completed by
James [12J). Mac Aogain[15J gives the decomposition matrices for p=

The most difficult cases are p = 2,n = 12 and 13, and for these
Stockho£e used a computer to find dim D(5,4,2,1) and dim D(7,4,2),
employing Theorem 11.6.
 114

25 YOUNG'S ORTHOGONAL FORM

We turn now to the problem of finding the matrices which represent

the action of permutations on the Specht module This has been post-
poned to a late stage in order to emphasize the fact that the represen-
tation theory of Gh can (and we believe should) be presented without
reference to the representing matrices.
Since G'n is generated by the transpositions (x­l,x) for 1 < x ::; n,
is is sufficient to determine the action of these transposition on a
basis of Consider first the basis of standard poly tabloids e
t.
Here we have

25.1 (i) If x­l and x are in the same column of t, then et(x­l,x) = ­e
t•
(ii) If x­l and x are in the same row of t, then et(x­l,x) = e
t
+ a linear combination of standard poly tabloids e t* with {t*} {t}
(by 8.3 and the technique used to prove 8.9).
(iii) If t(x­l,x) is standard, then et(x­l,x) = et(x­l,x)
In case (ii), the relevant standard tableaux t* may be calculated
by applying the Garnir relations.

25.2 EXAMPLE If u (3,2) and we take the standard in the

order tl,t2,t3,t4,t5 1 3 5 1 2 5 1 3 4 1 2 4 1 2 3 then
2 4 3 4 2 5 3 5 4 5

­1 0 0 0 0 0 1 0 0 0
­1 1 0 0 0 1 0 0 0 0
(1 2) <-> 0 0 ­1 0 0 (2 3) <-> 0 0 0 1 0
0 0 -1 1 0 0 0 1 0 0
1 0 ­1 0 1 1 -1 0 0 1

­1 0 0 0 0 0 0 1 0 0
­1 1 0 0 0 0 0 0 1 0
(3 4) +­» ­1 0 1 0 0 (4 5) 1 0 0 0 0
0 0 0 0 1 0 1 0 0 0
0 0 0 1 0 1 0 -1 0 1

In many ways, Young's natural representation, as this is called,

is the best way of describing the matrices which represent permutations.;
for example, it is independent of the field. However, vie must take three
cases into account, and the second one, where x­I and x are in the same
row, involves an unpleasant calculation. It turns out that these prob-
lems can be avoided when we work over the field real numbers, and
the rest of this section will be devoted to the case where the ground
 115

field is lR.

Let t < t < ••• < t be the standard in the order

l 2 d
given by definition 3.10. Wherever possible, we shall use the abbre-
viation e. for the standard poly tabloid et . •
Since we are working over the reals, we may construct from e l,
e ..• ,e an orthonormal basis f •.. ,f of using the Gram-
2, d l,f 2, d
Schmidt orthogonalization process. It is with respect to the new ortho-
normal basis that we get "nice" matrices representing permutations.
To fix notation, we remind the reader of the Gram­Schmidt orthogona-
lization process.
Suppose we have constructed a basis fl, ..• ,f j of the space spanned
by el, ..• ,e over and that fl, ••• ,f are orthonormal relative to the
j j
bilinear form < >. Then there is a non­zero linear combination
f of el, ... ,e + with < ei,f > = 0 for 1 i j (see 1.3). Now, the
j l
tabloid {t + is involved in f (othenlise f would be a linear combina-
j l}
tion of el, •.. ,e by the proof of 8.9, contradicting the fact that
j
< ei,f > = 0 for 1 i j.) Therefore, vie may take

fj+l= (±f)/«
the sign being chosen so that {t + has a positive coefficient in f j +
j l} l.
This determines f + uniquely.
j l
Of course, the new basis f •.. ,f of depends on the order
l,f2, d
of the original basis e .•• ,e However, we prove
l,e 2, d•

25.3 THEOREM The orthonormal basis of constructed

from the standard basis is independent of the total order we choose on
the standard tableaux, provided that the total order contains the partial
order q , given by definition 3.11

At the same time, we prove

25.4 YOUNG'S ORTHOGONAL FORM.

If (x­l, x) is a transposi tion in G n' then for all r

fr(x­l,x) = Plf r + P2 fs
where t s = tr(x­l,x) and Pl(= Pl(x,r)) equals (i-k+ £_j)­l if x­l is in
the (i,j)th position and x is in the (k,£)th position of t and
r,
= 1 with P 2 o.
Remark: It does not matter that there is no t equal to tr(x­l,x) when
s
x­l and x are in the same row or column of t since P = 0 in these
r, 2
cases. Young's Orthogonal Form says that fr(x­l,x) = ±f if x­l and
r
x are in the same row or column of t r, respectively.
 116

Before embarking on the proofs of 25.3 and 25.4, we require a

preliminary Lemma.
25.5 LEMMA Suppose that t and t* are any two v-tableaux, and that
x-I is lower than x in t*. If {t} 4 {t*} then {t}(x-l,x) 4 {t*} (x-l,x) •

Proof: Recall from definition 3.11 that miu(t) is the number of entries
less than or equal to i in the first u rows of t. Since {t} 4 {t*} ,
m. (t) s m. (t*) for all i and u.
au au
Let x-I be in the alth row and x be in the blth row of t. Let
x-I be in the a row and x be in the b row of t*. We are given
2th 2th
that b < a 2 •
2
Using 3.14, we deduce from m . (t) s m. (t*) that m;u(t(x-l,x» s
au au •
miU(t*(x-l,x», except perhaps for i = x-I and either b l s u < min(a·l,b 2 )
or max (b , a s u < a
l 2) l•
For b s u < min(a ),
l l,b 2
m 1 (t(x-l,x» m (t), since x-I is in the alth row and x is in the
x- ,u x,u
blth row of t and b l s u < a l
s m (t*), since {t} 4 {t*}
x,u
m 1 (t*(x-l,x», since u < b 2 < a 2 .
x- ,u
For max (b l,a 2 ) s u < aI'

mx_l,u(t(x-l,x» mx_ 2,u(t) + 1, since b l s u < al

s m
x- 2 ,u (t*) + 1, since {t} 4 {t*}

mx-l,u (t* (x-l,x», since b 2 < a 2 s u •

Therefore, m. (t(x-l,x» s m. (t*(x-l,x» in all cases. Thus
{t(x-l,x)} {t*(x-l,x)}. We do not have equality, since {t} ;< {t*}.

Proofs of Theorem 25.3 and Young's Orthogonal Form:

Assume that both results are true for all lRc='n_l Specht modules
(Both are vacuously true when n = 0). The proof now proceeds in 3 steps.

Step 1 The matrices which we claim represent (x-l,x) are correct for
x < n.
We take our notation for the proof of Theorem 9.3, so that Vi is
the spanned by those et's where t is a standard
and n is in the rith, r
2th, •.• ,or rith row of t. Since VI C V2 c ••• ,

the proof we gave for Maschke's Theorem shows that

Vi Ul $ U2 $ .•. $ Ui,
where U is the spanned by those ft's where n is in the
i
 117

r.th row of t. (Recall that Oelr total order on tabloids puts all those
1.
wi th n in the I' 1 th row before all those with n in the r 2 th row etc.)

In tile proof of Theorem 9.3 we constructed an l-homomorphism

Ai . . n-
ei mapping Vi onto Sm Whose kernel 1.S Vi- l· s i.nce Vi- l '" Ul 6l •••
\D U
i- l
and Vi = Ul E9 ••• E9 U
t,
we therefc e know that 0i is an :R G'n_l- i s o-
morpnism from Ui onto

Define a bilinear form < >*

< U,V >* '" < uO , vOfor u,v in U >

i i i,
Ai
where tile second bilinear form is that on SR • Since Ui is an absolutely
irreducible JRGn_l-module, our new bilinear form on Ui must be a multiple
of the original one, by Schur's Lemma. That is, there is a real constant
c such that
< U,V > * = c< u,v > for all u,v in Ui'

Because both forms are inner products, c is positive.

For each standard t having n in the rith row, let t
denote t with n removed, and write et for e
t
and ft for f t ' Suppose
that tp,tp+l, ... ,t are the standard have n in the rith
q
row. If p $ r $ q then

f I' '" u + a p e P + a P+le p+l + ... + a r e I'

for some u in V and a > O. Therefore, by 9.4,

i- l r
f O.
r 1.
= a e
p P
+ a
p+
le +1 + ... + a e
P I' I'

Since the last tabloid here is {t I' } with a positive coefficient, and
since < f z O.1.,f r O.1. > = c< f Z ,f I' > for p -
< z-< r, we deduce that

f
r
O.1. = Ie f r
Ive are as sumi.n q that Young I s Orthogonal Form is correct for the
R Gn_l-module sA1., so for x < n,

Ie fr(X-l,x)
IC (Plf r+ P 2fsJ = (Plf r+ P2fsJOi .

Here, t
s
= tr(x-l,lJ, and the real numbers P l and P2 are those in the
statement of Young's Orthogonal Form (the positions of x-l and x in t
r
mE the same as their positions in trJ. Since 0i is an isomorphism, we
have proved the desired result of Step 1, namely that

Step 2 The proof of Theorem 25.3

We know that there are real numbers a ••• ,a with

l,a2, r
 118

Theorem 25.3 will follow if we can show that

{t {t By induction, we may assume that when
j} r}.
is a linear combination of standard poly tabloids e i
and prove the corresponding result for fro

Case 1 For some x < n, x is lower than x-l in t and not in the same
r
row or column as x-l.

Let tr(x-l,x) = t
k•
Then {t
k}
{t
r}.
Therefore,

f
k
= cle
l
+ ••• + cke k where c
i
=0 unless {til {t k}·

Using 25.1, and applying Lemma 25.5, fk(x-l,x) is a linear combi-

nation of poly tabloids e for which {til {t r}.
i
Since x < n, Step 1 shows that

f
r
a multiple of f
k
+ a multiple of fk(x-l,x).
Therefore in this case,
fr ale
l
+ ••. + are r where a j =0 unless {t }
j
{t r}.

Case 2 For every x < n, x is higher than x-l in t or is in the same

r
row or column as x-l.

Since t r is standard,
_ it is easy to see that the hypothesis of
Case 2 implies that t (= t with n removed) has l,2, .•• ,n-l in order
r r,
down successive columns.
We may certainly write

fr =
blf l + .•. + br-lf r- l + bre r where b r o.
Let x be the smallest integer such that b. 0 for some j and
J
mxu(t r) < mxu(t j) for some u, if such an integer x exists. We aim to
produce a contradiction.
First, 1 < x < n, since for all u, mlu(t r) = mlu(t j) = 1 (t and
r
t J· being standard),and m (t r) = m (t.) = + ••• + for all
nu nu J
tableaux t r and t .
j
By the minimality of x, mx-l,u(t r) mx-l,u(t j) for all u.
Let x be in the (y,z) place of t Then y > 1 (otherwise, for all
r.
u, mxu(t r) = mX-l,u(t r) + 1 + 1 mxu(t j), contradicting
the definition of x). Since t r has 1,2, .•. ,n-l in order down successive
columns, x-l is in the (y-l,z) place of t Therefore, using Step 1,
r•
er(x-l,x) -e r and fr(x-l,x) = -f •
r
y, mxu(t r) = m 1
r + 1
For u
x- ,u (t) mx-,u
1 (t.)
J
 119

The definition of x therefore shows that

m (t) < m (t.) for some u < y.

xu r xu J

r = uz for u < y (since t r has 1,2, ••. ,n-l in order down

But mx- 1 ,u (t)
successive columns), and the first row of t. contains at most z numbers
J
less than or equal to x-l (since mx-,1 l(t.)
J
$ m
x- 1 , l(t r) = z). Because
t is standard, this means that x must be in the (l,z+l) place of t j,
j
and x-l is in a column of t no later than the zth column.
j

z z z+l

x-l
y
x x-l

If t = tj(x-l,x), then Step 1 gives

k
fj(x-l,x) 0lf j + 02fk where 0 < 01 < 1.

Therefore,
blf l + .•. + 0jf j + ... + Dr-lf r- l + bre r

f -fr(x-l,x)
r
-b f (x-l,x) - - Dj (01 f j + 02 fk) - •.• + bre r •
l l
Since b. 0 and 01 -1, f. must appear elsewhere in the last
J J
line. This means that b is non-zero. But mx-l,l(t k) z + 1 > z =
k
1 l(t r ), and this contradicts our minimal choice of x.
mx-,
vJe have thus proved that in the expression

f r = blf l + .•. + br-lf r- l + bre r

b . = 0 unless {t.} {t}. Our induction hypothesis at the beginning
J J r
of Step 2 shows now that f is a linear combination of polytabloids e i
r
with {til g {t This concludes the proof of Step 2.
r}.

Step 3 Calculation of the matrices representing (n-l,n).

Take a new total order on tabloids, containing 4 , in which {t}

and {t(n-l,n)} are adjacent if both are standard. (This is possible in
view of Lemma 3.16.) < {t 2} <
We fix our notation by saying that
•.• <{t d} are the different standard tabloids ordered by definition 3.10,
and {t } « {t « •• « {t is the new order. Thus, n is a
ln 2 n} d n}
permutation of {1,2, ..• ,d} and if both t. and t. (n-l,n) are standard
an an
then t.a n (n-l,n)= t(. 1) rr ,
 120

We plan to evaluate frn(n-l,n). Asswne, for the moment, that if

trn(n-l,n) is standard, then trn(n-l,n) = t(r+l)n.
Let G denote the group {l, (n-l,n)} .
Let X denote the space spanned by e l n,e 2n, ... ,e(r-l)n

Let Y X + ern JR G (so that dim Y = dim X + 2 or 1, depending

on whether or not both t and trn(n-l,n) are standard.)
rn
Since our new total order contains <J , for every standard t,
neither or both e t and et(n-l,n) belong to X (using 25.1) . Hence both
X and Yare G-invariant.
By Step 2, fln, ... ,f(r-l)n is an orthonormal basis for X and
f l n, ... ,frn,f(r+l)n is an orthonormal basis for Y (Omit f(r+l)n if
dim Y = dim X + 1). The space spanned by f r n and f(r+l)n is the ortho-
gonal complement to X in Y,and because our inner product is G-invariant,
the space spanned by f r n and f(r+l)n is G-invariant (Omit f(r+l)n if
dim Y = X + 1) .
Now,
rn
= an element of X + b e r ' where b > O(since the coeffi-
f
cient of {t } in f is chosen to be positive). Therefore, when n-l
rn rn
and n belong to the same row or column of t
r n,
frn(n-l,n) = an element of X + Eb ern

+ l if n-l and n are in the same row of t

where E = rn
{
-1 if n-l and n are in the same colliOn of t
rn
But we have just proved that frn(n-l,n) is a multiple of f in
rn
these cases, and comparing coefficients of ern' we see that
frn(n-l,n) = Ef r n

and this completes the case where trn(n-l,n) is not standard.

On the other hand, when both t r n and trn(n-l,n) (= t(r+l)n) are
standard,
frn(n-l,n) = an element of X + b e(r+l)n (b > 0)
Since the space spanned by fr'lf and f (r+l)'If is G-invariant,

frn(n-l,n) = Plf r n + P 2f(r+l)n

where P l and P 2 are real nwnbers, and the coefficient of {t(r+l)n} shows
that P 2 is strictly positive. Now
< frn(n-l,n) ,frn(n-l,n) > < frn,f > 1
rn
so pi + 1 w'ith P
2
> O. Also

f P1 f r n (n -1 ,n) + P 2 f (r+1) n (n -1 ,n) ,

rn
whence
f(r+l)n(n-l,n) = p?,fr'lf - P1f(r+l)'If
It therefore, to show that Pi may be calculated as in the
 121

statement of Young's Orthogonal Form in the case under discussion, where

trrr(n-l,n) = t(r+l)rr This will be done using some properties of the
group G'3 .
Since n-l and n are not in the same row or colmnD of t r rr, n 3.
Also, t r rr 4 trrr(n-l,n), so n-l is lower than n in t r rr There are 4
cases to consider
(i) n-2, n-l and n appear in t thus:
r rr

(ii) Some two numbers from {n-2,n-l,n} are in the same row, but
no two are in the same column of t r rr •

(iii) Some two numbers from {n-2,n-l,n} are in the same column,
but no two are in the same row of t
r rr
(iv) No two numbers from {n-2,n-l,n} are in the same row or colman
of t
rrr
We tackle case (ii) first; case (iii) is similar and case (i) is
comparatively trivial. Finally we deal with the hard case (iv).

Case (ii) Let H be the group generated by gl = (n-2,n-l) and g2 (n-l,

n). Since n-l is lower than n in t ' t has the form:
r rr r rr

t or
rrr

In the first case, let t

r rr,
= t
and in the second let t = trrr(n-l,
n). The space spanned by ft,f and f is H-invariant. In fact, our
t 9l t 92

l n
results so far show that, with respect to the basis ft,ft9l,ft92 ' the

l:'
action of H on this space is given by

0
°2
gl (n-2 ,n-l) "1 g2 (n-l,n) _ 1
°2 -°1
0 0 T2 0 Tl

where 01 is known, from Step 1. The axial distance from n-l to n in t

= -(the axial distance from n-2 to n-l in t) + 1. Ne shall therefore
' h e d if we can prove that 01
h ave f1 n1S -1 1 + T-1 •
l
 122

I trace glg2 1
,;;
lal'll + Iall + I'll ,;; + + 1 2.

The character table of is

3
(1 ) (2,1) (3)
(3) 1 1
X 1
(2,1) -1
X 2 0
3
(1 )
-1 1
X 1

The only representation of dimension 3 having trace 1 on the

(3) (2,1)
transpositions and Itracel ,;; 2 on elements of order 3 is X + X •
Therefore, trace glg2 = 0, giving '1 = a 1'1 + 0'1 , as required.

Case (iv) Let H, gl and g2 be as in Case (ii). He may assume that n-2
is higher than and n-l is higher than n in t, and that t r th for
some h in H. Taking f f ft
t ' tgt' g:/ f tQ2g1' f t92g192' f t 9'291"2"'1 as a

basis for ftmH, gl and g2 are represented by

(n-2 ,n-l) -e

0: 2
13 2
0: 1
(n-l,n)
-Y l Y2
Y2 Yl
13 1

(Omitted !=Iltries are zero).

Here we know that each of is non-zero· The
-1 -1 -1
values of vl,w l and are known and vl + = wl '
from Step 1. vJe want 0: = 'lf , 13 = w and Y == V • There seems to
1 l 1 l l l
be no more efficient way of proving this than equating (gl g2)2 with
g2 g1' using the fact that glg2 has order 3 (cf. Thrall [23]). The
(4,1), (5,2) and (3,1) entries in the relevant matrices give

W
2 0: 2 C,l "i - wl w2 0: 1 C,2 - wl w2 Yl 0: 2 = 0
- "-2 13 2 v l 13 1 + -1 132 Yl - 13 1 13 2 = 0
 123

and - wl v
l
0.
1 0. w2
2 + 1
0.
1 0.
2
- w22 Yl 0.
2 = -0.
2
v
l
-1
Substi tu.ting w
2
2
= 1 - W
2
1 and w
l
_1
= v
-1
l + TTl , these rapidly

give the required result: 0.

1 = TT l' 6 1 = wl and Y
l = v l·
This finishes step 3 and completes the proof of Young's Orthogonal
Form.

2S.6 EXAMPLE Here is the orthonormal basis of S(3,2) in terms of

:JR
the graphs used in Example 5.·2:

2 f
l 5... 2. el t
l
1 3 5
2 4

2 13 f 2 !il. 1 -e
l
+ 2e 2 t
2
1 2 S
3 4

S'?-=--+-4---. 2. 134
2 S

{; f S'" 1 e - 2e - 2e + 4e 4
4 l 2 3

t 1 2 4
4
3 S

3 12 f s 4 2e - e - e - e + 3e
S l 2 3 4 S

t 1 2 3
s
4 S
 124

For clarity, we have chosen the graphs (= G •.• , G say)

l,G2, S'
so that the edges have integer coefficients. It is easy to check that
the graphs are orthogonal, and that {t is the last tabloid involved
i}
in G.. The numbers multiplying each f. ensure that < f. ,f. > = 1
1. 1._ 1. 1.
1
(For example, < G > = 1 2, so (213) G has norm 1) .
3,G3 3
Corollary 8.12 has been used to write the graphs in terms of poly'
tabloids. Since {t
2}
p {t 3}, e
2
is not involved in f 3, illustrating
Theorem 25.3.
Writing out in full the matrices representing (1 2}, (2 3), (3 4)
and (4 5) with respect to the orthonormal basis, f •.• ,f we have:
l,f2, S'
-1 1/2 13/2
1 13/2 -1/2
(1 2) <-> -1 (2 3) <-> 1/2 13/2
1 13/2 -1/2
1 1

-1 1/2 13/2
1 1/2 13/2
(3 4) 1 (4 5) <-> 13/2 -1/2
1/3 212/3 13/2 -1/2
212/3 -1/3
1

It is interesting to see that the last element of the orthonormal

basis is always a multiple of the vector {t}KtP used in definition
t
23.3 (cf. Example 23.6(iii) and f above). This is because both are
S
fixed by the Young subgroup G and to within a scalar multiple b
11 11
fixes a unique element of , by Theorem 4.13 (Theorem 4.13 shows that
dim HomlR (!'1iR = 1).
n
 125

26 REPRESENTATIONS OF THE GENERAL LINEAR GROUP

The representation theory of is useful in the study of more

general permutation groups. For example, Frobenius used part of the
character table of G to find that of the group H2 4 There is
24
another, less obvious application of the theory, following from a study
of the group G L d(F) of non-singular d x d matrices over a field F.
Remember that any group which has a representation of dimension dover
F has (by definition) a homomorphic image inside G Ld(F). Although the
results of this section will be stated in terms of the general linear
group, they apply equally well to any subgroup thereof. \Ie plan to
construct, for each n and each partition of n, a representation of
G Ld(F) over F. Hence from any representation of any group, we can
produce infinitely many new representations over the same field.
G Ld(F) acts naturally on a d-dimensional vector space, ,./(1) say,
over F. -
Let 1,
-
- -
2, .•• ,d
- -
. . . (1)
be a bas1s for W . If g = (g .. ) .
1J
.
1S a matr1x
in G Ld(F), then
.1.
g = L: g .. i
j 1J
The general element of W(l) @ w(l)may be written as

L: a .. i
i,j";d 1J I
(The reason for this perverse notation will emerge later.) Let G Ld(F)
act on w(l)@ w(l) by

as usual.
For the moment, assume char F = o. There are two natural G Ld(F)-
invariant subspaces of ,.,(l)@ W(1), namely those spanned by

i + i l l ,,; i ,,; J' ,,; d }

i i

{1-t i l , ;
and by
i < j ,,; d}
These are called the symmetric part of w(l)@ .,(1) and the second
exterior power of w(l) (or the skew-symrnetric part of w(l)@ w(l»
respectively. Since char F = 0

w(l)@ .., (1) (symmetric part) E& (2nd exterior power).

Wri te this as
w(l)@ w(l) _ w(2)E& w(12).
Less wellknown is that
w(l) @ w(l) _ w(3) E& 2W(2,1) E& .., ,(1 3 )
 126

(3) . (2 1)
for some subspaces W (called the 3rd symmetr1c power), W ' (of
which there are two copies) and w(13) (called the 3rd exterior power)
Also
1',(1) \iT(1) E!f w(4)e 3w(3,1)e 2\iT(2,2)e 3W(2,12)
4
e vI (1 )

"and so on". Further

Most of the work needed to prove these results has already been
done, since they are similar to those for the symmetric group (compare
the last example with 8(2)0 8(2)t G
4
=
8(4) e 8(3,1)e 8(2,2), when
char F = 0).
Consider again w(l)s 1',(1). How do we deal with the symmetric and
skew-symmetric parts when F is arbitrary (allowing char F = 2)7 We
adjust our notation, by letting \iT (2) be the space of homogeneous poly-
nomials of degree 2 in commuting variables I, ... We write

i j for the monomial I1

so that
i j i-l and w(2) is spanned by {i j d}. 11 i j
2),
We keep our previous notation for w(l) 0 \iT(1) and for ,,(1 and
now w(1))/w(12) w(2) as vector spaces, since
2)
i i
.i - i
modulo W(1 .

Another way of looking at this is to define the linear trans for-

mation w(1)0 w(l)+ w(2) by
1,0
i
.i + Li
Then ker Wl,o = W
(1 2 ) (2)
• If we let G Ld(F) act on W in the natural way,
then Wl,O turns out to be a G Ld(F)-homomorphism:

I g .
i g
- l:
i j
k,9,

It is the generalization of w(2), described in the way above, to

the kth symmetric power of w(l) which we take as our building block for
the representation theory of G Ld(F).

26.1 DEFINITION The kth symmetric power of w(l) is the vector space
w(k) of homogeneous polynomials of degree k in commuting variables

i, ... with coefficients from F. We write

 127

and we let the G Ld(F) action on \il(k) be defined by

where the sum is over all suffices jl,j2, ••• ,jk between 1 and d, and
g = (gij)·
The reader who is more familiar with the kth symmetric power as
the subspace of \il(l)@ •.• @ \il(l) (k times) spanned by certain
symmetrized vectors, may find it useful to know that the connection
between this and \il(k) is:
\il(k)* =
where * denotes the process of taking duals.
Corresponding to = in the representation theory of G ,
n
we consider the space ..• @ \il(lJ n). There is still a little more
preliminary work, though, before we come to this. It should,
be clear that it is useful to discuss vector spaces spanned by tabloids
with repeated entries (For the time being, it is best to forget any
intended interpretation in terms of the action of G Ld(F)).
Let X = Xl x ••. x be a sequence of non-decreasing positive
2 n
integers. If lJ is a partition of nand t is a (of type (In))
let t X denote the array of integers obtained by making the substitu-
tions i Xi in t (1 $ i
X - t$ n). Let t
-
if and only if for all
-
l 2X
m and r the mumber of m's in the rth row of tlX equals the number of m's
in the rth row of t and let {tX} denote the --class containing tX.
2X,
Then
it} {t}x = {tX}

is clearly a well-defined map from the set of lJ-tabloids of type (In)

onto the set of lJ-tabloids of type where the partition is defined
by
= the number of terms of X equal to i.

(As in some of our earlier work, we do not require lJ and to be proper

partitions of n.) Extend X to be a linear map on SO,lJ, the space spanned
by the lJ-tabloids.

26.2 EXAMPLES (i) If X = 1 1 2, then

SO, (2,1)x is spanned by 1 and r2
-1-

S(2,1), (2,1)x is spanned by Il 2T

-2- - 1
(ii) If X 1 1 1, then
SO, (2,1)x is spanned by 1

S(2,1), (2,1)x
0.
 128

Certain linear were defined on the vector

spaces in section 17. Define the corresponding linear transfor-
mations on by

= X

(It is clear that this is welldefined.)

26.3 THEOREM Suppose that X is a sequence of type A is a proper

partition, and are a pair of partitions as in 15.5.
Then
(i) dim SAX = the number of semistandard A-tableaux of type

(ii) = X

(iii)

Proof: In 17.12, we proved that

e
t
t
p
", ..
-- e tR c
11 ",
C and
°.
Applying X to these equations, we deduce that
s]l",]l X ,I, X

and

By considering last tabloids, as in the construction of the

standard basis of the Specht module, obviously dim SA X
o
I,
where is the set of semistandard A-tableaux of If
this inequality is strict for
#
SODE A, or if X n ker 1 *
c- ,]lc
strictly contains S]l Ac,]lX for some pair of then choose
a pair of partitions O,v and a sequence of operations Ac,R c leading
from O,v to A,A or respectively (using 15.12). For each proper
partition a of n, let aa be the multiplicity of as a factor of
Then there is a series of subspaces of SO,v X with at least aa
factors isomorphic to sax (cf. Corollary 17.14).
Therefore,
the number of v-tabloids of type = dim sO'v X
;0, L:
aa dim Sa X
a
;0, l: a I I
a a
L: dim Homl[ by Corollary 13.14.
a aa
,
n
At least one of the inequalities is strict (the first is strict if
our kernel is too big, and the second is strict if dim SA > I I}.
a v a,v
Recall that aa is the multiplicity of Se as a factor of Me = Se
 129

Therefore,

= the number of v-tabloids of type i;, by Theorem 13 .19.

This contradiction completes the proof.

26.4 DEFINITIONS Let W I be the vector space direct sum of X-

where X runs over all non-decreasing sequences whose terms are 1,2, ••. ,d.
#
Let the l/J maps act on by acting on each component separately. When
is a proper partition of n, let =
We now have

26.5 THEOREM Let A be a proper partition of n. Then

(i) dim wA equals the number of semistandard A-tableaux with
entries from {l,2, ... ,d}

(ii) wA is an intersection of kernels of p-maps defined on wo,A.

Proof: This follows immediately from 'l'heorem 26.3, since Il is the

A-
direct sum of the spaces S X
Next, identify with ... 0 We have
defined the action of G Ld(F) on a symmetric power, and hence G Ld(F)
acts 0n An unpleasant use of suffix notation shows that the
l/J-maps with the action of G Ld(F), and then Theorem 26.5 shows
A
that w is a G Ld(F) module, which we call a Weyl module.

From Theorem 26.3, we have

26.6 THEOREM • has a series, all of whose factors are Weyl modules.
The number of times IVA occurs in this series equals the number of times
the Specht module SA occurs in a Specht series for .

In particular, the number of times IJ A occurs in a Weyl series for

= ... 09 is given by Young's Rule. (Notice that
no "inducing up" takes place here, as it did in the corresponding symme-
tric group case). This justifies all the examples we gave at the beginn-
ing of the section; indeed, we have proved their characteristic-free
analogues. For example, w(1)0 w(1)0 w(l) has a G Ld(F) series with factors
'. . (3) (2, 1) (2 1) (1 3 ) .
r somorpn i c to W ,W ,W',1i order from the top, and this
holds every fielct F.
We now investigate character values. Let

[:"2
 130

If F is algebraically closed, every elements of G Ld(F) is conjugate to

one of the above form, and so it is sufficient to specify the character
of g on a Weyl module.

26.7 DEFINITION For an integer k, let {k} denote the kth homogeneous
symmetric function of al, ... ,a That is,
d.
{k } = 1: ail ai .ai
.•. d 2" k

(By convention {O} 1 and {k} = 0 if k < 0)

26.8 EXAMPLES {l} = a + a + ... +a

l 2 d
2 2 2
{2} a + a + ... + a + a +a
1 2 d la 2 la 3 + .. + ad_lad
3 2 2
{3} a + ••. + a 3 + a 2 + a 2 + .•. + ad_lad + ad_lad + a la 2a 3
1 d la 2 la 2
+ ..• + ad-2ad-lad

26.9 THEOREM {k} is the character of g on w(k).

Proof
-I g = a. i + a combination of l's with j < i. Therefore,
..,---;--
if 1 i :0: ••• :0: i :0: d, then the coefficient of
il .•. i k in il •.. i k g
l k
is ail" .aik Since w(k) has a basis consisting of elements of the
form il ... i k , the result follows.

Now, recall from 6.1 that m = is the matrix whose entries

are indexed by proper partitions, given by
[A l][A 2]···[An] =
From Theorem 26.6, we have
26.11

Since the Determinantal Form gives the inverse of the matrix m,

we have

26.12 THEOREM If A is a proper partition of n, then the character of

A
g on the Weyl module W is I{Ai-i+j}l.

We write {A} = I{A.-i+j}!

a,
= the character of g on wA• Then
immediately

26.13 THEOREM is the character of g on wA 0

The Littlewood-Richardson Rule tells us how to evaluate

as a linear combination of {v}'s (where A is a partition of r, is a
 131

is a partition of n-r and v is a partition of n), since we know that

the Littlewood-Richardson Rule follows from Young's Rule.
It is worth noting that were we to define

{k} = Z . ail ai 2'" ai k

l:5i l
where {a •.. } is countable set of indeterminates, then
l,a 2,

and {A} = I{Ai-i+j}!

are equivalent definitions of {A}, for A a partition of n (since our
results work for al, ... ,a
in an infinite field, the above must be
d
identities in the indeterminates al, ... ,a d).
{A} is called a Schur function, and the algebra of Schur functions
is thus isomorphic to the algebra generated by the [AJ'S, where A varies
over partitions of various n. The Littlewood-Richardson Rule enables us
to multiply Schur functions.
Schur functions can be evaluated explicitly by

26.14 THEOREM If II is a proper partition of n, then

Z m
v vu

Note: In all that follows, Z' denotes the Silla over all unordered sets
of n indices i l,i .•• ,i (no two equal) chosen from {1,2, ... ,d} or
2, n
from {1,2, •.• } depending on whether we wish to define {ll}in terms of
{a l,a 2,· .. ,a d} or of {a ... }.
l,a 2,

Proof of Theorem 26.14 (m m')AV =

O,
(Z mAO X Z mVT XT) this being an inner product of
o T
characters of
(XD1JD2J···DnJ,X[V1J[V2J",[VnJ), by the definition
of m.
V)
dim Home (MA,M
n
the nmnber of A-tabloids of type v, by Theorem 13.19.
t h e coe ff icient of alv 1 a v2 2 ••. a vn·1n { Al } •.. { An } ' by
n
considering how this coefficient is evaluated,
, vi v 2
Therefore, { A } .•. { A } = Z (m m'), Z a. a.
l n v "v 11 12
-1
But {ll}= (m tA {Al} •.. {An} by 26.11,
-1 V2
Z (m) A mAO mv o Z a.
A, v , 0 II 12
 132

vl v2 vn
L: m L:' a. a i 2··· a. , as required.
v Vfl
k
and s'
26.15 DEFINITION Let sk L: a.
i
if k
"1 0
l.

We can now prove the useful

26.16 THEOREM Let P be a permutation of C; n with cycle lengths Pl'

and let C(p) denote the centraliser of P in Gn . Let xfl(p)
be the value of the character of S corresponding to the partition
n
fl, evaluated on p. Then

(i) s
Pl

(Li.) {fl}

Proof x[V l][V 2 J. · .Lv n] (p) the number of tabloids in r-flfixed by p •

= the nuniller of v-tabloids of type (In) where each cycle of p is cont-
ained in a single row of the tabloid.
. . vl 'J2 'I n
= the of a a ... an in s s ... s , by considering
l 2 p1 P2 Pn
how this coefficient is evaluated.

L: Xfl(p) {fl}, from the definition of m.

u
This proves part (i) of the Theorem.
By the orthogonality relations for the columns of the character
table of G ,
n
1 A
P I
'C
IC(p) X (p) sPl sP2·· .sp n {A} •

and this is the second part of the Theorem.

26.17 COROLLARY I f G is any group, and e is an ordinary character of
G, then for all n 0 and all proper partitions u of n, efl
character of G, where
" is a

efl (g) 1
= L:
P I C(p) I x" (p) e (gP 1 ) e (gP 2) .•. e (gP U ) (g G)

The centraliser order IC(p) I and the character Xfl refer to the symmetric
group and the sum is over all proper partitions P of n; P is written
where P l
If e has degree d, then efl has degree equal to the number of semi-
standard fl-tableaux with entries from {1,2, ••• ,d}
 133

Proof: There is a homomorphism ¢ from G into G L (C). If g E G, let

k d k k
¢(g) have eigenvalues a a ••. ,ad. Then a l, a 2, .•. , ad are the
l, 2,
eigenvalues of gk, and so 0(gk) = a k + ••. + The result now
l
follows from Theorem 26.16(ii) and Theorem 26.5(i).

26.18 EXM1PLES Referring to the character tables of (;0' G'l' 62 and

the last of which is
(1 3 ) (2,1) (3)
Cerrtr a Ld.s e r order: 6 2 3
X (3) 1 1 1
(2,1)
X 2 0 -1
(1 3 )
X 1 -1 1

we have, for any ordinary character 0 of any group G, and any g in G,

trivial character of G

1 2
-h0(g) )2 + T0(g )
1 2 1 2
T(0(g) ) - 'T0(g )
l 2 1 3
-t(0(g»3 + )0(g) )
3 2 1
-h0(g) )3 + 0.0(g )0(g) - )
J. 2 l 3
.h0(g) )3 - 20(g )0(g) + ""!0(g ).

Note that 0(1)0 0(1) 0(2) + 0(1 2)

0(2)0 0(1) 0(2,1) + 0(3), etc. (cf. Young's Rule)

If 0 has degree d, then

deg 0(2) (d) + d _ d(d+l)

- 2
2
(1 2 ) (d) d t d L)
deg 0
r

2 2
(1 3 ) (d)
deg 0
3
deg 0(2,1) (d+l) d (d-l)
3
deg 0(3) (d+2)
3
(The last two degrees are most easily calculated by using the next
Theorem. )
Similar to the Hook Formula for dim S A, we have

26.19 THEOREM dim wI. IT (d+j-l)

(i,j)E[A]

Proof: We prove first that dim w(k) = if k is a non-negative

integer.
The natural basis of w(k) consists of (k)-tabloids with entries
 134

from {1,2, ••• ,d}. There is a 1-1 correspondence between this basis and
sequences of "bars" (I) and "stars" (*) with d-l bars and k stars

e.g. * I * * * I * * * I *
1 3 3 4 777 8

There are (k+d-l) such sequences, so this is the dimension of w(k) .

d-l
Since {A} = I {Ai + j - i} I, we have

dim wI,. = I (Ai + dd = 11 + j - ') I = I (Ai + d - 1 + j - i)

Ai + j - i
I
Id(d+l) •.• (d + Ai - 1 + j - ill f(d), say.
(Ai+j-i)!

Let A have h non-zero parts (so we are taking the determinant of

an h x h matrix). It is clear that the polynomial f(d) has degree
1,.1 + 1,.2 + ••. +Ah and leading coefficient

I
I(A.
1 1 , by 19.5 and 20.l.
a,
+ j - 1): IT (hook lengths in [A])

Therefore, the result will follow if we can prove:

When k -h+l, and i* is the largest integer i such that

A. k+i, then
.* divides f(d) for k 0 and . * +k divides
f(d) for k < o.
(k measures "how far right of the diagonal we are", and the above will
ensure that the numerator in the statement of the Theorem is correct.)

Case 1 k o.
For i $ .* , d $ d+k $ d+ Ai-i. Examining the third determinantal
expression for f(d) above, we see that, for i $ i*, (d+k) divides all
the entries in the ith row of our matrix. Therefore, .* divides
f (d).

Case 2 k < O.
Here we claim that f(d) = det(Mk(d» where is a matrix whose
(i,j)th entry for all i, and for all j -k, is

This is certainly true for k -1 (by our first expression for

f(d», so assume, inductively, that it is true for k. For all j -k,
subtract the jth column of Mk(d) from the (j+l)th column of Mk(d). In
the new matrix, for j -k+l, the (i,j)th entry is
 135

A. +d + j-i + k) A. +d + j-i + k-l)

( ci + k ( d + k-l

Thus, our new matrix may be taken as 1'\-1 (d), and tile result claimed
is correct.

-Since if Ai + j-i < 0

=
if Ai + j-i 0

and Ai + j-i

Therefore, the rank of is at most (-k-l) + (h-i* + 1),

whence the nullity of 1'1 ( - k ) is at least i* + k. Thus (d+k)i*+k divides
k
det(Mk(d» = f(d), as required.
26.20 EXMlPLES
d (d+l) ... (d+k-l)
(i) If A = (k ) then dim HA In particular,
(2) = d(d+l)
dim
2:

(ii) I f [A] X X X , then the hook graph is 4 3 1

X X 2 1

Replacing the (i,j) node in [A] by j-i, we have 0 1 2

-1 0
A d (d+l) (d+2) (d-l) d
Then the Theorem gives dim W
4.3.2.1.1.

As with the Hook Formula for the dimension of the Specht module
SA, the formula of Theorem 26.19 is much more practical than the count
of semistandard tableaux when calculating dimensions of Weyl modules
wA•
 APPENDIX

THE DECOMPOSITION MATRICES OF THE SYMMETRIC GROUPS Gn FOR THE PRIMES

2 AND 3 WITH n < 13

We have deliberately presented these decomposition matrices without

sorting the characters into blocks. This makes it easier to spot
patterns which might hold in general: for example, compare the part of
the decomposition matrix of Gl 3 corresponding to partitions having 3
parts with the decomposition matrix of and see the remark follow-
ing Corollary 24.21.
 137

The decomposition matrices of b for the prime 2

n

rl r-! rl

C rl N
n = 0 n = 1 n = 2

f: 1 (0) 1 *1 0) 1 1 (2) 1
1 (l 1

rlN rlN rl.:T .:T

,..... ,.....
rlN
rl
,..... rl
,.....
L!).:T (Y)
n = 3 (Y) N n = 4 .:T (Y)
n = 5 '-' '-' '-'

1 (3) 1 1 (4) 1 1 (5) 1

2 (2,1) 1 3 (3,1) 1 1 4 (4,1) 1
1 0 3) 1 2 (22) 1 5 (3,2) 1 1
3 (21 2 ) 1 i f: 6 (31 2 ) -2 1
1 ( 4)
1 5 (22) ) 1 1
4 (21 3) 1
1 05 ) 1

rl .:T.:T to rl to.:TooC
rl rl N

,.....,..... ,..... ,.......,....... ,......,......

,..... ....
rl N rl
,.....
r--
rl N

to L!) .:T.:T
(Y) rl

n = 6 ill Lf).:T(Y)
n = 7 ........... ......., ............................

1 ( 6) 1 1 (7) 1
5 (5,1) 1 1 6 (6,1) 1
9 (4,2) 1 1 1 14 (5,2) 1
f: 16 (321) 1 14 ( 4,3) 1 1
10 (41 2) 2 1 1 35 (421) 1 1 1
5 (32) 1 1 15 (51 2) 1 1
10 ( 31 3) 2 1 1 21 (321 ) 1 1
5 ( 2 3) 1 1 21 (32 2) 1 1
9 (221 2) 1 1 1 f:20 (41 3) 2 1
5 (21 4 ) 1 1 35 (321 2) 1 1 1
1 (1 6 ) 1 14 (231 ) 1 1
15 (31 4 ) 1 1
14 (2 21 3) 1
6 (21 5) 1
1 0 7
) 1
 138

The decomposition matrix of 6 8 for the prime 2

r-I<D.:t co.:tO
r-I <D .:t
,...,,.... ,....
r--t N cY)r-Ir-I
,..., N ro
cot-<D l1?l1?.:t
'-' '-' '-'

1 (8 ) (18 ) 1
7 (7,1) (21 6) 1 1
20 (6,2) 1 1
28 (5 ,3) (2 31 2) 1 1 1
64 ( 521) (321 3) 1
70 (431) (32 21) 2 1 1 1 1
14 (4 2) (2 1 1
21 (61 2) (31 5 ) 1 1 1
56 (42 2) (32 12) 2 1 1
42 "'(322) 2 1
35 (51 3) 1 2 1 1
90 1'(421 2) 2 2 2 1 1

Block number: 1 1 1 1 2 1

The decomposition matrix of b for the Drime 2

9

r-ICO<D co <D cooo

N .:tr-l c- .:t <D
r-I
,.... ,.... ,....,.... ,....
r-I N cY).:t r-I r-I N
,.... "'''N(Y)(Y}
0) co r-- <D co <D U') .:t
'-' '-' '-' '-' ......... "'-"" ................

1 (9) (19 ) 1
8 (8,1) (21 7 ) 1
27 (7,2) (2 21 5 ) 1 1
48 (6,3) (231 3) 1
42 (5,4) 1) 1 1
105 (621) (321 1 1 1
162 (531) (32 21 2) 2 1 1 1 1
168 (432) (3 221) 1 1
28 (71 2 ) (31 6 ) 2 1
84 (4 21) ( 32 3 ) 2 1 1 1
120 (52 2 ) (3 21 3) 2 1 1
42 ,',(33) 2 1
56 (61 3) (41 5 ) 1 1
189 (521 2) (421 3) 3 2 1 1 1
216 (431 2) (42 21) 1 1 1
70 1'(51 2 2 1

Block number: 1 2 1 2 1 112

 139

The decomposition of for the prime 2

rieo<.O 00 <.0 oeo eo 000

C': .::t ri <.0 O'l N 0<.0
riri ri N L'"--

'""
. . .
,-...,-... ,.....,-...
'"" '""
""" rl N (Y).::triri '""
ri
'"" ri
C'J N
0 .. N (Y) .::t (Y) (Y)
rimoo r- tOr-to Lf) LIO.::t
.................. '-"'......"

1 (10) (110) 1
9 (9,1) (21 e) 1 1
35 ( 8 ,2) (221 6 ) 1 1 1
75 (7,3) (231 4 ) 1 1 1
90 (6,4) (2 41 2) 1 1 1
160 (721) (321 5 ) 1
315 (631) (32 21 3) 1 2 1 1 1
288 ( 541) (32 31) 1 1
450 (532) (3221 2) 2 1 1 1 1 1
768 ;, (4321) 1
42 (52) (25 ) 1 1
36 (81 2) (31 7 ) 2 1 1
225 (62 2) (321 4 ) 1 1 1
252 (4 22) (3 22 2) 2 1 1 1 1
210 (43 2) (331 ) 2 1 1
84 (71 3 ) (41 6 ) 2 1 1 1
350 (621 2 ) (421 4 ) 2 1 3 1 1 1
567 (531 2 ) (42 21 2 ) 3 1 3 1 2 1 1
300 (4 21 2 ) ([12 3) '? 1 1 1 1 1
525 (52 21) (431 3) 3 1 2 1 1 1 1
126 (61 4 ) (51 5 ) 2 1 2 1 1
448 2 1
Block number: 1 1 1 1 1 2 1 2 1 3
 140

The decomposition matrix of G'11-L0r the prime 2

.-10 .::T 0 .::TN<OCO .::t co woo

.-1 .::T 0 <..D (Y) OJ 0) .=t .=t .-1<0
.-1 .-1 .-1rl .-1 co .::T.-1
.-1
,... ,...
.-1 ,... ,-... ,...... ,......,...... ,....... ,....., ,...... ,.....,oi
,... .. N (Y) .::T L.f) rl ,....-i ,.....; N N N
.-10 .. (Y) .=t C") .::TC")
.-1 .-1 (j) co t-- <0 co r-- <0 <0 co '"
'-' '-' ................. ................ ......... ................

1 ( 11) 011) 1
10 00,1) (21 8) 1
44 ( 9 ,2) (221') 1
110 ( 8 ,3) (231 5) 1 1
165 <7,4) (2 41 3 ) 1 1
132 (6,5) (251) 1 1
231 ( 821) (321 6 ) 1 1 1
550 (731) (32 21 4 ) 2 1 1 1
693 ( 641) (32 31 2) 1 1 1 1 1
990 (632) (32 213) 1 1 1 1
990 (542) (3 22 21) 2 1 1 1 1 1
2310 (5321) (4321 2) 3 2 2 1 1
45 (91 2) (31 8 ) 1 1
330 (5 21) (32 4 ) 1 1
385 <72 2 ) (321 5) 1 1 1
660 (53 2) (331 2) 2 1 1 1
462 (4 23) ( 3 32) 2 1 1
120 (81 3) (41') 2 1
594 (721 2) (421 5) 2 1 1 1 1
1232 (631 2) (42 21 3) 2 3 2 1
1155 (541 2) (42 31) 3 1 1 1 1 1 1
1100 (62 21) (431 4 ) 2 2 1 1
1320 (4 221) (432 2) 2 1 1 1
1188 1'(43 21) 2 1
825 (52 3) (4 21 3) 3 1 1 1 1
210 (71 4 ) (51 6 ) 2 1 1
924 (621 3) (521 4 ) 2 1 1 2 1 1
1540 (531 3 ) (52 21 2) 4 1 1 2 2 1 1
252 1'(61 5) 2 2 1

Block number: 1 2 1 2 1 2 1 1 1 2 1 2
 141

The decomposition matrix of G1 2 for the prime 2

riO ;:T 0 ;:TN 0 0 co co <D <DNNCO

ri ;:T 0 <D (Y) N r-- OCO;:T ri m (Y) <D
ri ri (Y) If) ;:T NO ;:T r-- <D (Y)
ri ri If) N
,..... ri
,..... ,.....
,..... rl N,-..."......,,-.. ,-....,-... ,.....,..... ,.....
,-...,-.. r-1 r-i
ft ft(Y) ;:T If) rl ri ririN NMNN
Nrl 0 ft ft "N(Y) ;:T1f)(Y) ;:T ;:T (Y) ;:T
riri rimcor--mco r--<.er--<DIf)<D1f)
.......,.......,........,.......,......., ......... '-' ......... ........ ......... '-' '-' ......... '-' ........

1 (2) (112 ) 1
11 01,1) (21 1 0 ) 1 1
54 00,2) (221 8 ) 1 1
154 (9,3) (231 6 ) 1 1 1
275 (8,4) 1 1 1 1
297 (7,5 ) (25 12) 1 1 1 1
320 ( 921) (321 7 ) 1
891 (831) (32 21 5) 3 1 1 1 1 1
1 1108 (741 ) (32 21 3) 1
1155 (651) 1 1 1 1 1 1
1925 (732 ) (251 2) 3 1 1 1 1 1 1
2673 (642) (3 22 21 2) 3 1 1 1 1 1 1 1 1 1
2112 (543) (3321) 1 1
5632 ( 63',)1) (432]3) 1
5775 (5421) (432 21) 5 5 2 3 1 2 1 1 1 2 1
132 (6 2) ( 26 ) 1 1
55 001 2 ) (31 9 ) 1 1 1
616 (8,:>2) (321 6 ) 2 1 1
1320 (5 22) (3 22 3) 2 1 1 1 1
1650 ( 63 2 ) (33 13) 2 1 1 1 1 1 1
462 (4 3) 2 1 1
165 (91 3) (41 8) 1 2 1 1
945 (821 2) (421 6 ) 3 2 2 1 1 1
2376 (731 2) 4 2 1 3 2 2 1 1
3080 (641 2) (42 31 2) 4 2 1 3 2 2 1 1 1 1
1485 (5 2}2) 3 1 1 1 1 1
2079 (72 21) (1+31 5) 3 2 1 2 1 1 1 1
4158 (53 21) (43 21 2) 2 5 1 2 1 1 1 1
2970 (4 231) (43 22) 2 4 1 1 1 1
1925 (62 3) 3 2 1 2 1 1 1 1
4455 (532 2) (4 221 2) 3 5 1 3 1 2 1 1 1
2640 *(4 22 2) 4 2 1 1
330 (51 7 ) 2 2 1 1 1
3696 (631 3) (52 21 3) 6 2 2 3 2 2 2 1 1 1
3520 (541 3) (52 3 1) 1 1 1
3564 (62 21 2) 6 2 2 2 2 1 2 1 1 1
7700 >"(5321 2) 8 6 2 4 2 3 ? 1 :> 2 1
462 (71 5) (61 6 ) 2 2 1 2 1 1
2100 4 2 2 2 2 1 2 1
1728 (721 3) (521 5) 1 1

Block number: 1 1 1 1 1 1 2 1 2 1 1 1 2 3 1
 142

The decomposition matrix of 5 1 3 for the prime 2

.... N sr co::o
.... .0 0.0.0
sr sr
.0 .0
ONCOC:OO
r--- r-- co
.:t ,...j
co coco
",co
Oco
0::

"" ....
NCO U')
U') U') e-NV N ::
N""'" .., N .-;NCO
"" co

--
_ _ _ _ _ _ _ ...... r-fr"'"i
HNM--
...... ....-irlC'lNNMNC'-lM

----- -- ----- --_ ..............

.. .=t U') 1.0 ('\,I
.. .:t U') sr ('I') sr sr
.... r-i C .. ("').=t tJ) M
""
N "0
r-i rl r-i C) <:::::) t" ......-l c:nOOt'-OOC"'--c.oWr---tDU')

1 (13) (11' ) 1
12 Cl2,l) (21 1 1 ) 1
65 (ll,2) (221' ) 1 1
208 ClO, 3) (2'1 7 ) 1
429 (9,4) (2'1 5 ) 1 1 1
572 (8,5) (2 5 1 ' ) 1 1
429 (7,6 ) (2 61) 1 1 1
429 (1021) (321 8 ) 1 1 1
1365 (931) (32 21 6 ) 3 1 1 1 1
2574 (841) (32'1') 4 1 1 1 1
2860 (751) (32'1 2) 2 1 1 1 1 1
3432 (832 ) (3 221 5 ) 2 1 1
60()6 (742 ) (322 21') 4 1 1 1 2 1 1 1
5148 (652) (322'1) 1 1 1 1
6435 (643) (3'21 2) 3 1 1 1 1 .1 1 1 1
12012 (7321) (4321') 3 2 1 1
17160 ( 6421) (432 21 2) 4 1 2 1 1 1 1
15015 (51+31) (43'21) 7 3 1 1 1 1 1 1 1 1
66 Clll' ) (31 10 ) 2 1
1287 (6 21) (32 5) 1 1 1 1 1
936 (92' ) (3'1 7 ) 2 1 1
3575 (73' ) (3'1') 3 1 1 1 1 1
3432 (5' 3) (3' 2 2 ) 2 1 1 1 1
2574 (51+' ) (3'1) 2 1 1
220 (lOl' ) (41') 1 1
1430 (921' ) (421 7 ) 1+ 2 1 1 1
4212 (831' ) (42 21 5 ) 3 1 2 1
6864 (741' ) (1+231' ) 6 2 3 2 2 1 1 1
5720 (,651') (42'1) 2 2 1 1
3640 (82'1) (431 6 ) '2 1 1 1
8580 (5 221) (1+32') 3 1 1 1 1 1
11440 (63 21) (1+3'1') 2 1 1 1
3432 ( 4'1) (43' ) 2 1 1
4004 (72' ) (4'1 5 ) 4 2 2 1 1 1
12012 (632 2 ) (4'21') 3 1 1 1 1
12870 (542 2) (1+' 2'1) 6 3 2 2 1 1 1 1
11583 (53' 2 ) (4 231 2 ) 5 3 1 1 1 1
8580 *(4 232) 4 2 1
495 (91' ) (51 8 ) 3 2 1
3003 (8n') (521 6 ) 5 2 1 1 1 1
7800 (731') (52'1') 8 2 3 '2 1 3 1 1 1
10296 (61+1 3 ) (52'1' ) 8 2 3 2 1 3 1 2 1 1
5005 (5'1') (52' ) 3 1 1 1 1 1
7371 (72 21') (531 5 ) 7 2 2 1 1 3 1 1 1
20592 (6321') (5321' ) 6 1 3 2 1 1 1
211+50 (51+21' ) (532'1) 12 1+ 3 3 1 3 1 2 2 1 1
16016 *(53'1' ) 8 4 2 2 2 1 2 1
9009 (62'1) (51+1') 7 2 2 1 1 2 1 1 1 1
729 (81 5 ) (61 7 ) '2 1 1
4290 (721' ) (621 5 ) 6 '2 2 1 1 2 1 1
9360 (631') (62'1') 1+ 1 3 2 1
921+ ,.,(71 6 ) 1+ 2 2 1
Block number: 1 2 1 2 1 2 1 1 III 2 1 2 1 2 2 1
 143

rl rl rlrl

c .....
,.....
"'rl
..,.....
n = 0 n = 1 n = 2
*1 un 1 *1 0) 1 1 (2)
2
1
1 0 ) 1

rl rl rl (Y) rl (Y) rl =t rl to =t
,..... ,.....
.
,.......,........ ,......,,......
rl rl ,..... ..
,..... rl ,..... r l " ' ..
rl ..
=t (Y) eo r-,
"" .:t N
(Y) (Y) C'; Lf)
-.... .....................
'"' '"' '"' '"' '"'
n = 3 n = 4 n = 5

1 0) 1 1 (4 ) 1 1 (5) 1
*2 (2,1) 1 1 3 0,1) 1 4 (4,1) 1
1 0 3 ) 1 *2 (2 2 ) 1 1 5 0,2) 1 1
3 (21 2 ) 1 {,6 OF) 1
1 (14 ) 1 5 (2 2 1) 1 1
4 (21 3 ) 1
1 (15) 1

r-1WMr-1 u;C to Lf) co

riri

.,.....
ri

.
rl=t

,..... ,.....,..... ,.... ,.....

.
,..... ,.....
...-l('.:C'0 ri.-iN r-:
,..... ri N -N
rl
,....-!rl
N
,..... N
...' r-I ('.;N C""-l('J
to Lf) .=t M.=t (T)('J r-wu;.:t Lf) =t (Y) (Y) (Y)

-.. .......................................... ....... -.... ..............

'"' '"' '"' '"' '"'
n = 6 n = 7

1 (6 ) 1 1 (7) 1
5 (5,1) 1 1 h (6,1) 1
9 (4,2) 1 14 (5,2) 1 1
5 ( 2)
1 1 14 (4,3) 1 1
10 (41 2) 1 1 15 (51 2 ) 1
*16 ( 321) 1 1 1 1 1 35 (421) 1 1 1 1
9 (221 2 ) 1 21 ( 21) 1 1
5 (2 3) 1 1 ?1 ( 32 2 ) 1 1
1D (31 3 ) 1 1 35 (321 2) 1 1 1 1
5 (21 4 ) 1 1 *?O (41 3 ) 1
1 (16) 1 14 (2 3 1 ) 1 1
15 (31 4 ) 1
14 (/2 13) 1 1
6 (21 5) 1
1 0 7
) 1
 144

.-it- (Y) 00""'; ,....; Ll) t- Ll) ,....; C(Y)OO

rl N N (Y) (Y) C'J O"l ,....; N

,....,.... ,....
,.... ,.... ,.... ,....,.... ,.... ,....,.... OJ OJ ,....;
N (Y) ,.... ,....; ,....i
,.... rl
''OJ
OJ
,......j N (Y)
N N
NOj
,....;
e-, ,'" N
,....; OJ

00 e-- to Ll).::t to Ll) .::;- :.T (Y) .::t r-) (Y)

................ ........................ '-' '-' '-' ........................................

1 (8 ) 1
7 (7,1) 1
?n (6,2) 1 1
28 (5,3)
14 ( tl 2 ) 1 1
21 (61 2) 1
64 r szi ) 1 1 1
7n (431 ) 1 1
56 (1+2 2) 1 1 1 1
*42 (322) 1 1
90 (421 2) ]
56 (32]2) 1 1 1 1
7n (2 21 ) 1 1 1
35 (51 3 ) 1
11+ (24) 1 1
35 (41 4 ) 1
64 0?1 3 ) ,
-'. 1 1
31 2)
28 (2 1
21 (31 5) 1
?'" (221 4 ) 1 1
7 (21 6 ) 1
1 ri 8) 1

Block number: 1 :' ')

1 :' 3 1 1 2 3 1 2
 145

...-;r--r-- ...-;...-; ...-; lJ) c-, r-- L!) ...-; en r--co...-;"

c-: ;j" N (Y) u: (Y) N 0::>
...-;
H H H
,....
,.... ,....
,-.. ......... ,.....",..... N
N HHH
,-..,......... ,,-......,-.. ,...... ,-.. """"'-"N
HN (Y) H H rlN C':H HN "N
,.... (Y)N C'J (Y) N (Y) NN N
co co r-- u: L!) r-- <.D U") zr L!) ::T ;j"(Y)(Y)
......, '-'
'-' '-' "'-"" "'-' '-'" '-" ....... '-" '-" '-' ..................................

1 ( 9) 1
8 (8,1) 1 1
27 o,:n 1
48 ( 6 ,3) 1 1
42 ( 5 ,4) 1 1
28 (71 2 ) 1 1
lr15 (621) 1 1 1 1 1
162 (531) 1
84 ( 4 2 1) 1 1 1 1
120 (52 2) 1 1 1 1 1 1
168 (432) 1 1 1 1 1 1 1 1
189 (521 2) 1
216 (431 2) 1 1
216 (42 21) 1 1
168 (3 221) 1 1 1 1 1 1 1 1
162 (32 21 2) 1
"'42 (33) 1 1
56 (61 3 ) 1 1
84 (32 3) ] 1 1 1
*70 (51 4) 1 1
189 (421 3 ) 1
121') (321 3) 1 1 1 1 1 1
42 C2 41) ] 1
56 (41 5 ) 1 1
1n5 (321 4) 1 1 1 1 1
48 (2 31 3) 1 1
28 (31 6) 1 1
27 (2 21 5 ) 1
8 (21 7 ) ] 1
1 (19 ) 1

Block number: 1 1 2 1 1 1 1 3 1 1 1 4 4 2 1 5
 146

The decomDosition matrix of (;10 for the prime 3

H01 .:tH OH <.0 .:t 01 01 <.0 <.0 <.0 .:t .:t r-- .:t .:t H 0 e- 01
(Y) .:t 01 (Y) co r-- N N (Y) co N <.0 (Y) N.:t01 <.0 e-
N rl H N U) N U) N

,...,...
,... ,...,...
,... ,...,... ,... ,... ,.....,,......,-.. .-...-.. ,..... N N

. . .
,....... ,....,,.....,
0.1 rlHN
.-...
0.1
N HH
,... H N(Y) .::t.-..N HHrl N N N 0.1 H H H 0.1 N N 0.1 N
0 .. 0.1 H N(Y).:tN (Y)N M N M N N(Y)N No.1
H01 rot-<.OU)<X) ["'--. lD LJ) to </).:t.:t<.OU).:tU).:tC"'l.:t(Y)
.............. ............... ..................... ...................... """" .................................................. ""-'f ...............

1 (10) 1
9 (9,1) 1
35 (8,2) 1 1
75 (7,3) 1 1
90 (6,4) 1
42 (52) 1 1
36 (81 2 ) 1
160 (721) 1 1 1 1
315 ( 631) 1 1
288 ( 541) 1 1
225 ( 62 2 ) 1 1 1
450 (532) 1 1 1 1
252 (4 22) 1 1 1
210 (43 2) 1 1 1 1
350 (621 2) 1 1 1 1
567 (531 2) 1
300 (4 21 2) 1 1 1 1
525 (52 21) 1 1 1 1 1 1
(4321) 1 1 1 1 1 1 1 1 1 1
252 (322 2) 1 1 1
567 (42 21 2) 1
450 (32 2 12) 1 1 1 1
84 (71 3) 1
210 (331) 1 1 1 1
300 (42 3) 1 1 1 1
126 (61") 1
*448 (521 3) 1 1
525 (431 3) 1 1 1 1 1 1
288 (32 31) 1 1
42 (25) 1 1
126 (51 5) 1
350 (421") 1 1 1 1
225 (321") 1 1 1
315 (32 21 3) 1 1
90 0"1 2) 1
84 (41 6 ) 1
160 (321 5) 1 1 1 1
75 (231") 1 1
36 (31 7 ) 1
35 (221 6 ) 1 1
9 (21 8 ) 1
1 (1 1 0 ) 1

Block numbers: 1 2 1 1 2 1 3 1 3 3 2 3 2 1 1 4 1 1 1 3 5 2
 147

matr.;x o-F &'11 for the prime 3

"'e:
... "" "'............
0"" :
ec
co COM
"'rl
('.: N 01
NIJ)OC
IJ):1"HC'./
:
'"
rl.::t.:t:rO).,:TH
Cl (T)..-I r-I C Q") M
e- t""-- r--- ,....lV'> M
rl

'" co
r--
eM
N '"
'"
ri""w N N Nri w
n ______________
--------C\j
Ci N r-I ,......j M r--i N
ri '"...
ri
'"
- "C"-lM;:TrJ)('\j Hr-IrlrlN NNN Mr-Ir-It"'"""fN NNC\I NC\I NN
r-f C .. .... H """ (T) .::t C\I C'J ('I") .::t (T) N C'-: M .::t N M N M M N (Y) N

1 (11) 1
10 (10,1) 1
(9,2) 1 1
110 (8,3) 1 1
165 (7 1 1
132 (6,5) 1 1
(912 ) 1
231 (821 ) 2 1 1
550 (731) 1 1 1 1
693 ( 641) 1
330 ( 5 21) 1 1
385 (72 2 ) 1 1 1 1
990 (632) 1 1 1
990 (542) 1 1 1
660 (53 2) 1 1 1 1
1 1 1 1
(7212) 1
1232 (631 2) 1 1 1 1
1155 1 1 1 1
1100 (62 21) 1 1 1 1 1 1
2310 (5321) 2 1 1 1 1 1 1 1 1
1320 (4 221) 1 1 2 1 1 1 1
*1188 (43 21) 1 1
1320 (432 2) 2 1 1 1 1 1 1
1540 (52 21 2) 1 1 1 1
2310 (4321 2 ) 1 1 2 1 1 1 1 1 1
990 (322 21) 1 1 1
120 (81 3 ) 1
825 (52)) 2 1 1
(332) 1 1 1 1
210 (71" ) 1
(6213) 1 1
(531 3) 1 1 1 1
825 2 1 1
660 (331 2 ) 1 1 1 1
1155 31) 1 1 1 1
330 (32' ) 1 1
*252 (615 ) 1
(521') 1 1
1100 (431') 1 1 1 1 1 1
1232 (42 21 3) 1 1 1 1
990 (32213) 1 1 1
693 (32 3}2 ) 1
132 (251) 1 1
210 (51') 1
594 1
385 (3215) 1 1 1 1
550 (3221") 1 1 1 1
165 (2'13) 1 1
120 (41' ) 1
231 (321') 2 1 1
110 C2 31 5) 1 1
(31" ) 1
(221 7 ) 1 1
10 (219) 1
1 (1'1) 1

Block numoers: 1 2 2 1 2 2 3 1 1 3 1 2 3 312 112 124 1 2 2 3

 The 3-regular part of the decomposition matrix of for the prime 3

ri
0

1 (12) 1 Ul '"
11 (11 ,1) 1 1 riri
on,2 ) 1 '" <-
rien
N
(9,3) 1 1 1
275 1 1 1 ri
UlO
297 (7,5) 1
132 (6 2) 1 1 riri
en '"
55 ( 1"'1 2) 1 1 0) ri
0
320 (921 ) 2 1 1 1 1 riO
891 (831) 1
ri 0
ri
(741 ) 1 1 1 1 1 N N'"
Ul <-
1155 (651) 1 1 1 1 NtO
616 (82 2) 2 1 1 1 1 1 N
l/)
1925 (732 ) 1 1 1 1 1 1 1 1 1 .::TO
riC'
2673 1 N N
1320 (5 22) 1 1 1 1
ri l/) ri
1650 (63 2) 1 1 1 1 1 1
.::T '"
en.::TtO
ri'"
.t>
00
2112 (543) 1 1 1 1 1 1 1 1 1 1 en
ri
(82J2) 1 .::T 0)
2376 (7312) 1 1 l/) NO)
3080 e-- N
(6412 ) 1 1 1 1 1 ri.::T
(5 2 J2 ) 1 1 ri '"
.::TO)
2079 (72 21) 1 1 1 ri N
r-, Ul
5632 (6321) 3 1 1 1 112 1 1 1 1 1 1 1 ri.::T.::T
5775 (5421) 2 1 1 2 1 1 2 1 1 III 1 1 1
en to
(53 21)
l/)

1 1 1 1 "'ri
2970 (4 231) 1 1 1 '"
ri <-
en .::T
(532 2) 1 1 NtOtO
2 1 1 2 1 1 1 1 1 1 1
\l) ' "
"'en
2970 1 1 1
riri
enri
(62 2 J2 ) 1 0)
.::Tri
'" '"
*7700 (5321 2 ) 3 1 1 3 1 1 1 1 1 1 2 1 1 ri 0 '"
ri <-
4158 1 1 to
1 1 1 1
N
5775 2 2 1 211 1 1 1 1 1 1 1 1 1
2673 (322 21 2) 1
 The 3-singular part of the decomposition matrix of G 1 2 for the prime 3

MMN NNW NrlW3m

rl N rirl rirl rl M Mn rlrlN

462 (4 3 ) 1 1 1 1
165 (91 3 1 1 1
1925 (6: 3) 2 1 1 1 1 1
462 (3' ) 1 1 1 1
330 (81') 1 1
1728 (nIl ) 1
3696 (631 3) 7. 1 1 1 1
3520 (541 3) 1 2 1 1 1 1
3520 (52 31) 2 1 1 1 1 1
2112 (33?1) 1 1 1 1 1 1 1 1 1 1
1'185 (42' ) 1 1
1320 (322 3) 1 1 1 1
462 (71 5 ) 1 1
*2100 (621') 1 1 1 1
3564 (531')
1925 (4 21' ) 1 2 1 1 1 1 1
3696 (52 21 3) 2 1 1 1 1
5632 (4121 3) 1 1 3 1 1 1 1 1 2 1 1 1 1 1
Ol
v 1650 (331 3) 1 1 1 1 1 1
308n (42 31 2) 1 1 1 1 1
1155 (32'1) 1 1 1 1
132 ( 26 ) 1 1
462 (61 6 ) 1 1
1728 (521 5) 1
2 n79 (1131 5) 1 1 1
2376 (42 21') 1 1
1925
1408
(3221')
(32 31 3 )
1
1
1 1 111
1
1
1
1
1
,1
297 (?51 2) 1
330 (51 7 ) 1 1
945 (4 ?J.6)
616 (321 6 ) 2 1 1 1 1
891 (2 21 5) 1
275 0'1" ) 1 1 1
165 (41 8 ) 1 1
320 (21 7 ) 2 1 1 1 1
154 (2 31 6 ) 1 1 1
S5 (31 9 ) 1 1
54 ( 2 1 8) 1
11 (2110) 1 1
1 (112) 1

Rlock number: 112 112 I I I 3 1 1 114 1 1 1 5 5 1 5 2 115 2 315 6 1 6 2 1 7

 .-I
N
.-I
:T ...,
1 (13 ) 1 W :T
.-I .....
12 (I2,]) 1 .-leo
The 3-reE:ul_al'_ part of the de cornr-osI t ion natrix of
:TN
E5 (11,7) 1 1 :T
2"8 (1f'1,3) 1 1 1
§:13 for the nrime 3
.-I
479 (9,4) 1 1 W
woO'>
572 ( 8 ,5 ) 1 1 1 NO'>
429 (7,6) 1 1 N N U)
.-I .....
66 (11,1 2) 1 N e--
.-leo
429 (10,7,1) 2 1 1 1 N
1365 (931) 1 1 N N
.-IU)
7574 ( 841) 1 1 0'>
:T <'I N
2860 (751) 1 1 1 1 1 0'> eo
1" 8 7 ( 6 21) 1 1 ..... 0
U)
936 ( 97 2 ) 2 1 1 W
W:T
74 (832) 1 1 1 1 <'IN
6f'106 (742 ) 1 1 1 1 0'>0'>
..... 0
8 US?) 1 1 N
N U) U)
3575 (73 2 ) 1 1 1 1 1 CD
0'>

6435 (643) l 1 1 1 "'ON

.-1.-1
3'132 ( 523 ) 1 1 1 :l. Nt-

?574 ( 54 2 ) 1 1 1 1
:TCD
...,
(971 2)
...,,.. eo
143f'1 2 1 1 1 CD ...,
4712 (831 2) 1 0'> eo
.-INCD
01
0
6864 (7111 2 ) 2 1 1 1 1 ,.. 0'>
.-I N
57?n (651 2 ) 1 1 1 1 1 C ...,
3f40 (8?21) 2 1 1 1 1 1 .-I ,.. eo
.-I ..., U)
1?f'l12 (7321) 4 1 2 1 1 1 1 1 1 1 1 1 0'> CD

1716n (5 11? 1) 1 1 1 1 1 1 1 .-10

.-I
eo
CD
85SO (5 221) 2 1 2 1 1 1 1 1 1 1 Ll)
eo r--
11440 (63 21) 3 1 1 1 1 1 1 1 1 1 1 .-Irl
.:tt-
15n15 (5431) 2 1 1 3 1 1 1 1 1 111 111 ..., 00

(632 2 ) 1 1 1 1 1 1 r-- ('Jrl

12012 .:tt-co
1?870 (542 2) 1 1 2 1 1 1 1 ..., CD
r-- Ll) C"1
1]583 (53 27) 1 1 co C"1

"'SS8n (4 232) 2 1 1 2 1 1 1 1 1 1 1 c-r r--

rlCDN
7371 02 2]2) 1 M rl
("') e-,
CD
0)

2"sn (63V 2
) 2 1 1 1 1 1 .:t N
HI150 (S'i?1 2) 1 1 1 1 1 1 1 C'Ll)
rl r-. r-,
*Hf'l16 (53 21 2 ) 3 1 1 3 1 111 2 111 1 N
rl
ro
N N
11S 83 (11 2 31 2 ) 1 1 NCO
?l45n (532 2 1) 1 1 1 1 1 1 1 0
U)
]?87f'1 (4 22 2 ] ) 2 1 1 1 1 1 1
15f'115 (43 221) :3 1 1 2 1 1 1 1 1 111 1 1 1
17161') ('432 21 2 ) 1 1 1 1 1 1 1
 The decomnosition matrix of for the prime 3 (continued)

.::TOO COlD M (X)L/)CO r- MOOM a:>O'lC'N lDlf)r---N

,.;tD:!"r-tN to C'l0'l r--- OOM mcnco (,0 NO) N O1lO r-tlO c.DM N 01.::t MW<'o rl t-NE' lDcnc.Dr-t CT>[""-o CO 00
N N N N m f"'o- N:::- 0
.....
N(Y')
...-l ..... O ...-l...-la> .... r-- co r-i .:t 0 r-i ('.ILl')

'"
(Y)
...-l...-lN ...-l
...-l
0 01 ' ) 1
3 11 ( 11 '
1) 1 1 1 1
4004 (71' ) 2 1 1 1 1 1 1
2 S 7'1 ( 113' ) 1 1 1
495 (91") 1
3O"3 (821') 1 1
78n n on') 2 1 1 1 1
10296 (6 111' ) 1
5 n0S (5 21') 1 2 1 1 1 1
9009 (62'1) ? 1 1
2 5 7 11 0"1) 1 1 1 1
S""5 (S2") 2 1 1 1 1 1
LO 1 1 1 1
858 0 (432') 7. 1 1 2 1 1
(3'2 2) 1 1 1
3432 1
792 ( 81 5 ) 1
4?9 0 on" ) 1 1 1
9360 (631") 1 1
9 009 (SIll') 2 1 1
1 1
936O 1
2 1 1 1 1
?"59? 1 1 1
12"12 ( II2 ? 1 ' ) 1 1 1
1 1 1 1 1 1 1 1
11440 ( '13 21 ' ) 1 1 3
1
(52'1 2) 1
6 1135 (3'n 2) 1 1 1
1
1
5720 ('12 "1) 1 1 1 1
51'18 (322'1) 1 1
1287 ( 32 5 ) 1
,.,924 (71 6 ) 1
 The decomposition matrix of for the prime 3 (concluded)

.:teo co to (T) CX)LI)CO f"oo. r-ico,...-l cocnr---C'J (OLO t"--N

r-t<D3rlN NCl COM cnClCQ <D NOl N cnc.o r-IID CDM N en =t Mto<D r-i f'Nt' WallO rl OlE"'-- cow
..,
U) E""-
,-L:t .:t N N N
.-l.-lN
N .:t .... 0 Olt'-oN30NM
.-l.:t(")
0'l3Nr-ImOLO:J'
.-l.-l0
.-l
.-l.-lco ....M;:1"('I')
t--
U')NMN
co,-l("l')::r
NNNO
OM ('\,ILO
.-l
4290 (621 5 ) 1 1 1
7371 (531 5) 1
4004 1 2 1 1 1 1 -I
7800 (5?21') 2 1 1 1 1
POI? (4311' ) 1 4 1 1 1 1 211 1 1 1
3575 (131') 1 1 1 1 1
6864 (112 31 3) 2 1 1 1 1
6 (1() 6 (322 21 3) 1 1 1 1
2860 (3;'1 2 ) 1 1 1 1 1
429 (261) 1 1
792 (61') 1
31)03 (521 6 ) 1 1
3640 (411 6) 2 1 111 1 en
ro
11212 (q?21 5) 1
3432 (3221 5) 1 1 1 1
257 11 (3?31') 1 1
512 (251 3) 1 1 1
495 (51 8 ) 1
1430 (lin') 2 1 1 1
936 (321') 2 1 1
1365 (32 21 6 ) 1 1
1129 0'1 5) 1 1
?I 0 (41 9 ) 1
429 (321 8 ) 2 1 1 1
208 (2 31') 1 1 1
66 (3JI0) 1
65 0 2] 1 1
12 (2111) 1
1 (113) 1

n]ock nUMhers: 1 211 2 1 1 3 1 3 3 132 3 2 2 1 2 1 314 1 1 1 1 2 I I I 334 1 5 2 2 1 5 3 7. 1 3

 References

1. R.W. CARTER and G. LUSZTIG, On the modular representations of

the general linear and symmetric groups, Math Z. 136 (1974), 193-242.

2. C.W. CURTIS and I. REINER, "Representation theory of finite

groups and associative algebras," Interscience PUblishers, New York,
1962.

3. G. H. HARDY and E. M. T/ffiIGHT, "An introduction to the theory of

numbers," Oxford Univ. Press, Oxford, 1960.

4. J.S. FRAME, G. de B. ROBINSON and R.M. THRALL, The hook graphs

of the symmetric group, Canad. J. Math. 6 (1954), 316-324.

5. H. GARNIR, de la representation lineaire des groupes

symetriques, Memoires de la Soc. Rovale des Sc. de Liege, (4), 10
(1950) •

6. G.D. JAMES, Representations of the symmetric groups over the

field of order 2, J. Algebra 38 (1976), 280-308.

7. G.D. JAMES, The irreducible representations of the symmetric

groups, Bull. London Math. Soc. 8 (l976), 229-23:L

8. G.D. eJA-"1ES, On the decomposition matrices of the symme t r Lc groups

I, .J. Algebra 43 (1976), 42-44.

9. G.D •.JAMES, On the decomposition matrices of the syrnme t r Lc groups

II, ,J. Algebra 43 (1976), 45-54.

10. G.D. ,JAMES, A characteristic-:free approach to the representation

theory of (;n,J. Algebra 46 (l977) 430-450.

11. G.D. ,JAMES, On a conjecture of Carter concerningirreducihle

Specht modules, Math. Proc. Camb. Phil. Soc. 83 (1978), 11-17.

12. G.D. JAMES, A note on the decomposition matrices of b 1 2 and

for the prime 3, J. Algebra, to appear.

13. A. KERBER, "Representations of permutation groups I," Lecture

Notes in Mathematics, no. 240, Springer-Verlag.

14. A. KERBER and M.H. PEEL, On the decomposition numbers of symmetric

and alternating groups, Mitt. Math. Sem. Univ. Giessen 91 (197l), 45-81.

15. E. MAC AOGAIN, Decomposition matrices of symmetric and alternating

groups, Trinity College Dublin Research Notes, TCD 1976-10.

16. J. McCONNELL, Note on multiplication theorems for Schur functions

"Combinatoire et representation du groupe symetrique, Strasbourg 1976,"
 154

Proceedings 1976, Ed. by D. Foata, Lecture Notes in Mathematics, no.

579, Springer-Verlag, 252-257.
17. N. MEIER and J. TAPPE, Ein neuer Beweis der Nakayama-Vermutung
tiber die B10ckstruktur Symmetrischer Gruppen, Bull. London Math. Soc.
8 (1976), 34-37.

18. M.H. PEEL, Hook of symmetric groups, Glasgow

Math. ,7. 12 (1971), 136-149.

19. M.H. PEEL, Specht modules and the symmetric groups, J. Algebra
36 (1975), 88-97.

20. M.H. PEEL, Modular representations of the symmetric groups,

Univ. of Ca1qarv Research Paper no. 292, 1975.

21. D. STOCKHOFE, Die Zer1egungsmatrizen der Symmetrischen Gruppen

S12 und 5 zur primzah1 2, Communications in Algebra, to appear.
13
22. W. SPECHT, Die irreduzib1en Darste11ungen der Symmetrischen
Gruppe, Math Z. 39 (1935), 696-711.

23. R.M. THRALL, Young's semi-normal representation of the symmetric

group, Duke J. Math. 8 (1941), 611-624.
 Index

Basic combinatorial lemma 9 Hook 73, 77, 89

basis, orthonormal 115 diagram 80, 92, 98
, standard 29, 69 formula 77, 135
bilinear form, invariant 1 graph 73
, non-singular 2 , skew- 73
binomial coefficients 87
block 84, 85, 93 Involve 13
Branching Theorem 34, 62, 79 irreducible representation 16
39,40,71
Carter Conjecture 97, 102, 105 Specht module 89, 104
character 23, 79
stabilizer 10 Littlewood-Richardson Rule 52
composition factor 16, 42, 51 62, 130
60, 104, 110
trivial 101 Haschke s Theorem
I 1
conjugate diagram 9 Murnaghan-Nakayama Rule 79
partition 9, 25 80, 85
cycle type 6
Nakayama Conjecture 85, 102
Determinantal Form 74
decomposition matrix 42, 43 Order, dictionary 9
98, 111, 113, 136 , dominance 8
diagram 8 on tabloids 10
, conjugate 9 ordinary irreducible
hook 80,92,98 representation 16
, p-power 95, 97 Orthogonal Form 114
dictionary order 9 orthonormal basis 115
dominance order 8
dual module 3 1jJ-maps 67
vector space p-power diagram 95, 97
p-regular partition 36
Exterior power 126 class 36
pair of partitions 54
Garnir relations 27 partition 5
general linear group 125 , 2-part 94,95, 97, 106
graph 18 s, pair of 54
Gram matrix 3 , proper 54
group algebra 16, 41 permutation 5
 156

permutation module 13 standard 29

polytabloid 13 Submodule Theorem 15
, standard 29 symmetric group 5
power, exterior 126 power 126
, symmetric 126
Tableau 9

Row stabilizer 10 , standard 29

tabloid 10
Schur function 131 , standard 29
semi standard homomorphism 46 transposition 5
tableau 45 type of tableau 44
, reverse 102 of sequence 54
signature 5
signed column sum 13 Neyl module 1:'.9
skew-hook 73 dimension 129, 133
Specht module 13
dimension 30, 76 Young's natural representation
52, 77 114
, irreducible 89, 104 Orthogonal Form 114
Specht series 65, 69 Rule 51, 69
stabilizer 10 Young subgroup 13