Chapter 4

Permutation
Representations

In this chapter we study permutation representations and their invariants. Because of their combinatorial nature there is a lot one
can prove about them without any machinery from abstract algebra.
This is what we exploit in this chapter. In the rst section we dene permutation representations. Then in Section 4.2 we prove that
every symmetric polynomial can be written as a polynomial in the
elementary symmetric polynomials. In Section 4.3 we generalize this
result to arbitrary permutation representations.

4.1. Permutation Representations

The symmetric group on n letters n acts on the set {1, . . . , n} by
permutations. Let V = Fn be the n-dimensional vector space over
a eld F. Let e1 , . . . , en be a basis. Then n also acts on the set
S = {e1 , . . . , en } by permutation.
n S S, (, ei )  e(i)
for all i = 1, . . . , n. Thus n acts on V by permuting the basis vectors.
Let v = (v1 , . . . , vn ); then
n V V, (, v)  (v(1) , . . . , v(n) ).
63

64

4. Permutation Representations

In the terminology introduced in Chapter 1, this means that we have

rediscovered the dening representation of n :
: n  GL(n, F)
aorded by the matrices

0
1

..

(1i) =
1

1
0
1
..

i = 2, . . . , n,

where the 1s in the rst row and column appear in the ith place.
Since n can be generated by the transpositions (1i) for i = 2, . . . , n,
it is enough to give these images. Note that by construction every
matrix in (n ) has exactly one 1 in each row and column. Matrices
with this property are called permutation matrices. Let G be an
arbitrary group. We call a representation
: G GL(n, F)
a permutation representation if for each element g G its image
(g) GL(n, F) is a permutation matrix. Identifying the symmetric
group n with its image in GL(n, F) in its dening representation
allows us to rephrase this as
(g) n

g G.

Example 4.1. Take the cyclic group of order 4, Z/4, with generator
3 = 1. Then

0 1 0 0
0 0 1 0

: Z/4 GL(4, F), 3 

0 0 0 1
1 0 0 0
denes a faithful permutation representation of Z/4.

4.1. Permutation Representations

65

Indeed, every nite group G has a permutation representation

because G can be viewed as a subgroup of some symmetric group.
This can be seen as follows. Let |G| = n. Then enumerate the
elements of G, so that we have G = {g1 , . . . , gn }. Take an arbitrary
element g G. Then for every i = 1, . . . , n there is a g(i) {1, . . . , n}
such that
ggi = gg(i) .
Moreover,
ggi = ggj

i = j.

Thus G acts by permutation on its underlying set {g1 , . . . , gn }. We

identify the gi s with the basis vectors of an n-dimensional vector
space V , then group multiplication leads to a permutation of the basis
vectors. The representation of G that is obtained in this way is called
the regular representation of G. It is probably the most important
representation of G since it encodes all other representations in a way
we will not explain here, since this would lead us too far o the path.
However, this is an interesting topic and you should denitely learn
about this!
The representation in Example 4.1 is the regular representation
of Z/4: The group Z/4 consists of 0, 1, 2, 3. We identify the group
elements with basis elements of a vector space
{0, . . . , 3} {e1 , . . . , e4 }, i  ei+1 .
The group is generated by 1 = 3. Since (i)1 = i1, this translates
into
+3

ei+1  ei
for the basis vectors. So, the group permutes cyclicly the basis vectors. The matrix above describes precisely this: a cyclic permutation
of the basis vectors
e4  e3  e2  e1  e4 .
Thus the above Example 4.1 is the regular representation of Z/4.

66

4. Permutation Representations

Example 4.2 (Regular Representation of Z/n). Take the cyclic group

of order n. Then

0 1

0 1

.
.
.. ..
: Z/n GL(n, F), n 1 

0 1
1
0
denes a faithful permutation representation of Z/n. This matrix
describes a cyclic permutation of the basis vectors
en  en1   e2  e1  en .
Thus this is the regular representation of Z/n.
Before we proceed, let us give another example:
Example 4.3 (Regular Representation of 3 ). The symmetric group
in three letters has six elements. We enumerate them in the following
way:
g1 = (1), g2 = (12), g3 = (13), g4 = (23), g5 = (123), and g6 = (132).
Then the regular representation of

0
1

0
(g1 ) = I, (g2 ) =
0

0
0

0
0

1
(g3 ) =
0

0
0

0
0
0
0
1
0

1
0
0
0
0
0

0
0
0
0
0
1

0
1
0
0
0
0

0
0

0
,
1

0
0

3 is given by the matrices

1 0 0 0 0
0 0 0 0 0

0 0 0 0 1
,
0 0 0 1 0

0 0 1 0 0
0 1 0 0 0

0
0

0
(g4 ) =
1

0
0

0
0
0
0
0
1

0
0
0
0
1
0

1
0
0
0
0
0

0
0
1
0
0
0

0
1

0
,
0

0
0

4.2. Newton, Waring, and Gauss

0 0 0 0 0 1
0
0 0 0 1 0 0
0

0 1 0 0 0 0
0
(g5 ) =
, and (g6 ) =
0 0 1 0 0 0
0

1 0 0 0 0 0
0
0 0 0 0 1 0
1

0
0
0
1
0
0

0
1
0
0
0
0

0
0
1
0
0
0

1
0
0
0
0
0

67

0
0

0
.
0

1
0

Remark 4.4. It does not matter in which order we enumerate our

group elements to obtain the regular representation, because a change
of the order translates into a permutation of the basis. Thus we
receive an equivalent representation. This justies speaking about
the regular representation.
Note that the degree of the regular representation always coincides with the order of the group. Note also that not every permutation representation is the regular representation.
Example 4.5. The dening representation of n is, of course, a
permutation representation of degree n. However, it is not the regular
representation: this has degree n!.

4.2. Newton, Waring, and Gauss

In this section we look at the polynomial invariants of the dening
representation of the symmetric group on n letters n .
Consider the ring F[V ] = F[x1 , . . . , xn ] of polynomials in n variables. We call the sum
s1 = x1 + + xn F[V ]
the rst elementary symmetric polynomial in x1 , . . . , xn . More
generally we dene the ith elementary symmetric polynomial as

si =
xI ,
I

where the sum runs over all exponent sequences I = (i1 , . . . , in ) consisting of i 1s and zeros otherwise. These are also sometimes called
Vi`
ete polynomials. For example
s2 = x1 x2 + x1 x3 + + x1 xn + x2 x3 + + xn1 xn

68

4. Permutation Representations

and
s n = x 1 xn .
As we saw in the previous section the dening representation of n
induces an action of n on the n-dimensional vector space V = Fn
by permuting the basis elements e1 , . . . , en . For all n we have
(ei ) = e(i)

i = 1, . . . , n,

and by linear extension

(v) = (v1 , . . . , vn ) = (v(1) , . . . , v(n) ) v V.
Thus the induced action on the dual space V also permutes the basis
elements x1 , . . . , xn . For n we have
(xi (v)) = xi ( 1 (v)) = xi (v1 (1) , . . . , v1 (n) )
= v1 (i) = x1 (i) (v)
for all i = 1, . . . , n. Hence for a monomial xi11 xinn F[V ] we obtain
(xi11 xinn ) = xi11 (1) xin1 (n) ,
another monomial of the same degree with permuted indices. Ex
tending this linearly to all polynomials gives for f =
aI xI



(f ) =
aI (xI ) =
aI (xi11 xinn ) =
aI xi11 (1) xin1 (n) .
We apply an element n to an elementary symmetric polynomial
(si ) = (x1 x2 xi + x1 x2 xi1 xi+1 + + xni+1 xn )
= x1 (1) x1 (2) x1 (i) + + x1 (ni+1) x1 (n) .
Thus the only thing that has changed is the order of summation,
i.e., (si ) = si is a polynomial invariant under n . An arbitrary
polynomial invariant under n is called symmetric.
Example 4.6 (Symmetric Power Sums). The symmetric power sums
are dened by
pi = xi1 + + xin .
They are symmetric because any element in n permutes the summands.
The following formulae go back to Sir Isaac Newton [1646-1723].

4.2. Newton, Waring, and Gauss

69

Proposition 4.7 (Newtons Formulae). The symmetric power sums

and elementary symmetric functions satisfy the following relations:
s1 = p 1 ,
2s2 = p1 s1 p2 ,
3s3 = p1 s2 p2 s1 + p3 ,

isi =

(1)k1 pk sik ,

k=1

where we set s0 = 1.
Proof. We consider the following polynomial in an indeterminant t
and with coecients from F[V ]:
()

(t) =

n


(1 xi t) F[V ][t].

i=1

Its logarithmic derivative with respect to t is

n
n


(t)
1
t
xi (
(1 xj t))
=
(t)
(t) i=1
j=1,j=i

n

i=1

xi
= p 1 + p 2 t + p 3 t2 + ,
1 xi t

where the last equality holds because

xi
= xi + x2i t + x3i t2 + x4i t3 + .
1 xi t
Thus we obtain

(1)
(t) = (t)(p1 + p2 t + p3 t2 + ).
t
On the other hand, we can expand the product in equation () and
get
n

(1)i si ti .
(t) =
i=0

Therefore

(t) =
(1)i isi ti1 ,
t
i=0
n

(2)

70

4. Permutation Representations

and we obtain the result by equating coecients of (1) and (2).

These formulae express implicitly the symmetric power sums in

terms of the elementary symmetric functions. A little bit later Edward Waring [1734-1798] gave a formula expressing the pi s explicitly
in terms of the elementary symmetric functions. More generally, he
gave formulae expressing any symmetric polynomial in terms of the
elementary symmetric ones. This gives the elementary symmetric
functions their signicance: any symmetric function is a polynomial
in the si s. We want to prove this result, but instead of giving Warings (very long) formulae, we present the algorithmic proof due to
Carl Friedrich Gauss [1777-1855]. For this we need some preliminary
steps.
Example 4.8. Consider : Z/2 GL(4, F) with

0 1

1 0
.
(1) =

0 1
1 0

The dual space V is the disjoint union of {0} and the orbits of the
following nine types:
o[x1 ] = {x1 , x2 },
o[x3 ] = {x3 , x4 },
o[x1 + x2 ] = {x1 + x2 , x2 + x1 },
o[x1 + x3 ] = {x1 + x3 , x2 + x4 },
o[x1 + x4 ] = {x1 + x4 , x2 + x3 },
o[x3 + x4 ] = {x3 + x4 , x4 + x3 },
o[x1 + x2 + x3 ] = {x1 + x2 + x3 , x2 + x1 + x4 },
o[x1 + x3 + x4 ] = {x1 + x3 + x4 , x2 + x4 + x3 },
and
o[x1 + x2 + x3 + x4 ] = {x1 + x2 + x3 + x4 ,
x2 + x1 + x4 + x3 },

4.2. Newton, Waring, and Gauss

71

for nonzero eld elements , , , and . In particular, the set of basis

elements consists of two orbits:
{x1 , . . . , x4 } = {x1 , x2 } {x3 , x4 }.
Thus the polynomials x1 + x2 and x3 + x4 are also invariant.
The two polynomials we constructed in the preceding example are
orbit sums. They are what you think they are: Take any monomial
xI F[V ]. Then gxI F[V ] for g G is also a monomial, since we
are dealing with permutation representations. Thus the orbit of xI is
the set of monomials
o[xI ] = {gxI |g G}.
The orbit sum is the sum of all orbit elements

gxI .
o(xI ) =
It is, by construction, an invariant polynomial. This is our rst
method to construct invariant polynomials. As we see below, in the
case of permutation invariants this suces to nd all invariant polynomials.
Before we proceed note that in general


o(xI ) =
gxI =
gxI
gG/StabG (xI )

gG

as we illustrate with the next example.

Example 4.9. Reconsider Example 4.8. The orbit of x1 x2 consists of
one element, because it is invariant. Thus o(x1 x2 ) = x1 x2 . However,

g(x1 x2 ) = 2x1 x2 .
gZ/2

The elementary symmetric polynomials are nothing but orbit

sums (under the action of the full symmetric group)
si = o(x1 xi )

i = 1, . . . , n.

Indeed in general we have the following result.

Proposition 4.10. Permutation invariants are sums of orbit sums.

72

4. Permutation Representations

Proof. Let : G  GL(n, F) be a permutation representation. Let

f F[V ]G be a homogeneous invariant of degree d. Without loss of
generality we assume that f is not the sum of two other invariant
polynomials. We need to show that f is the orbit sum of one of its
terms. Set

aI xI , with aI = 0.
f=
I

Observe that
i

g(xI ) = g(xi11 xinn ) = xig11 (1) xign1 (n) = x1g

1 (1)

xng

1 (n)

Thus the action of G can be interpreted as a permutation of the

exponent sequence:
1
g(xI ) = xg (I) .
Thus for any g G we have that



1
gf =
aI g(xI ) =
aI xg (I) =
ag(I) xI .
I

Therefore we obtain
0 = f gf =

aI xI

ag(I) xI .

Since the set of monomials of degree d forms a basis of the vector

space of all polynomials of degree d, they are linearly independent.
Moreover, the coecients aI = 0 by assumption. Thus every term
aI xI in the rst sum is cancelled by the term ag(I) xI , i.e.,
aI = ag(I)
as claimed.

Example 4.11. We come back to Example 4.8 and consider monomials of degree two. The second elementary symmetric polynomial
is
s 2 = x 1 x 2 + x1 x 3 + x 1 x 4 + x 2 x 3 + x 2 x 4 + x 3 x 4
= (x1 x2 ) + (x1 x3 + x2 x4 ) + (x1 x4 + x2 x3 ) + (x3 x4 ).
We set the brackets around expressions that are Z/2-orbit sums. We
see that s2 is the sum of four Z/2-invariants.
Recall from Exercise 3 of Chapter 3 that the set of homogeneous
invariants of degree d, F[V ]G
(d) , forms an F-vector space, d N0 .

4.2. Newton, Waring, and Gauss

73

Proposition 4.12. The orbit sums of any given degree d form a basis
of the vector space F[V ]G
(d) .
Proof. In the preceding Proposition 4.10 we have seen that every
invariant polynomial is a sum of orbit sums. Thus the orbit sums
of degree d form a generating set of F[V ]G
(d) . To show that they are
linearly independent, assume that there is a linear relation
()

a1 o1 + + an on = 0,

where o1 , . . . , on are orbit sums of degree d and a1 , . . . , an F. Expand () as a sum of distinct monomials of degree d. Since those
are linearly independent and the orbits are disjoint, the coecients


a1 , . . . , an vanish as desired.
In other words, we have proven that rings of permutation invariants are generated by orbit sums of monomials. With this at hand
we can proceed to Gauss proof.
We say a monomial xI is smaller in lexicographic order than
xJ , written as
xI <lex xJ ,
if the rst nonzero entry in the dierence of the exponent sequences
J I = (j1 i1 , . . . , jn in )
is positive.
Theorem 4.13 (Ring of Invariants of the Dening Representation of
n ). Any polynomial invariant under the dening representation of
the symmetric group on n letters can be written as a polynomial in
the elementary symmetric polynomials.
Proof. Let f be a homogeneous symmetric polynomial. If f is divisible by sn , then
f = sn f

for some symmetric polynomial f
. Thus without loss of generality
we assume that sn does not divide f .
We want to proceed by induction on the lexicographic order. If
f F, then there is nothing to show. Thus our induction starts. Let

74

4. Permutation Representations

axI , a F \ 0, be the leading (i.e., highest) term of f with respect to

the lexicographic order. We claim that
in in1 i2 i1 .
Assume to the contrary that there is an index k {1, . . . , n} such
that ik < ik+1 . Then for = (k k + 1) n , the term
i

k
(m) = (axi11 xinn ) = axi11 xkk+1 xik+1
xinn

has higher lexicographic order. Since this also appears in f by Proposition 4.12, we obtain a contradiction.
Next observe that the symmetric polynomial
asi11 i2 si22 i3 sinn
has the same leading term as f . Thus
f asi11 i2 si22 i3 sinn
is a symmetric polynomial with lower highest term. So we are done
by induction.


Thus we can write our ring of polynomial invariants as
F[V ]n = F[s1 , . . . , sn ].
A set of polynomials is called algebraically independent if
there are no algebraic relations between them.
Example 4.14. There are no algebraic relations between the variables x1 , . . . , xn F[V ]. For otherwise there would be a polynomial
expression
f (x1 , . . . , xn ) 0.
But this means that f is the zero polynomial. There is also no algebraic relation between x2 and y 2 in F[x, y].
Indeed there are no algebraic relations among the elementary
symmetric functions.
Proposition 4.15. The elementary symmetric polynomials are algebraically independent.

4.3. G
obels Bound

75

Proof. Suppose there is an algebraic relation between the elementary symmetric functions. Let f (X1 , . . . , Xn ) F[X1 , . . . , Xn ] be a
polynomial such that
f (s1 , . . . , sn ) 0.
We have to show that f is the zero polynomial. Rewrite f (s1 , . . . , sn )
as a polynomial in the xi s. Choose the term aXI of f of highest
lexicographic order. Thus f (s1 , . . . , sn ), regarded as a polynomial in
the xi s, has leading term
axi11 ++in xi22 ++in xinn ,
which, by assumption, is zero. Thus a = 0 as desired.

Remark 4.16. The issue here is that this means that every symmetric polynomial can be uniquely written as a polynomial in the
s1 , . . . , sn : Let p F[V ]n be an arbitrary invariant. Assume that p
can be written in two ways as a polynomial in the elementary symmetric polynomials. Then
p = f1 (s1 , . . . , sn ) = f2 (s1 , . . . , sn )
for some f1 , f2 F[X1 , . . . , Xn ]. Thus
0 = f1 (s1 , . . . , sn ) f2 (s1 , . . . , sn ),
which in turn means that 0 = f1 f2 by the preceding Proposition 4.15. Thus f1 and f2 are the same polynomials.
Remark 4.17. The generators of a ring of invariants are in general
not algebraically independent; cf. Exercise 13 in this chapter.

4.3. G
obels Bound
In this section we want to nd a complete set of generators for an
arbitrary permutation representation
: G  GL(n, F).
This result was proven by Manfred G
obel in his PhD thesis.1 The
proof we present here is based on a reworking of the original proof
1Manfred G
obel, Computing Bases for Permutation-Invariant Polynomials,
PhD thesis, University of T
ubingen, Germany, 1996.

76

4. Permutation Representations

by Nelson Killius, who was at that time a rst year graduate student
visiting Northwestern University.
We note that (G) n GL(n, F) (where we identied the
symmetric group with its image under its dening representation). In
particular, this means that every polynomial that is invariant under
the full symmetric group is invariant under the smaller group G as
well.
F[V ]n = F[s1 , . . . , sn ]  F[V ]G  F[V ].
Thus the elementary symmetric polynomials are present in every ring
of permutation invariants. In other words, they form a universal set
of invariants for the class of permutation invariants.
We know already that F[V ]G is generated by orbit sums by Proposition 4.12. Our next goal is to describe more precisely the set of orbit
sums we actually need. For that we introduce more terminology.
Take an exponent sequence I = (i1 , . . . , in ) with i1 , . . . , in N0 ,
and rearrange it in weakly decreasing order. We obtain an n-tuple
(1 (I), . . . , n (I)) that is a permutation of the original (i1 , . . . , in ).
The underlying ordered set
(I) = (1 (I) 2 (I) n (I))
is called the associated partition of I.
We call a monomial xI special if the associated partition (I)
satises
(i) i (I) i+1 (I) 1 for all i = 1, . . . , n 1, and
(ii) n (I) = 0.
Example 4.18. The monomial x21 x32 x13 F[x1 , x2 , x3 ] is not special,
because the associated partition (3 2 1) does not satisfy property
(ii). However, considered as an element in F[x1 , . . . , x4 ] it is special.
The associated partition is (3 2 1 0). The monomial x11 x42 x13 is
not special in F[x1 , x2 , x3 ] nor in F[x1 , . . . , x4 ]. The respective associated partitions are (4 1 1) and (4 1 1 0). In both cases
the rst condition for special monomials is not satised: 4 1 > 1,
i.e., there is a gap.
Since the n-tuples of the associated partitions are permutations
of the original exponent sequence, two exponent sequences J and K

4.3. G
obels Bound

77

are permutations of each other if and only if the associated partitions

are equal: (J) = (K).
We want to prove that the ring of invariants of a permutation
representation is generated as an algebra by the orbit sums of special
monomials and the elementary symmetric polynomial2 sn = x1 xn .
Thus we need to show that any permutation invariant can be rewritten
as a polynomial in the orbit sums of special monomials and sn .
The proof relies on the same idea as Gauss proof for the n invariants: we give an algorithm that rewrites an arbitrary invariant
in the desired form. For this algorithm to work (i.e., to terminate)
we need to choose a suitable order on the set of monomials.
Let I = (i1 , . . . , in ) and J = (j1 , . . . , jn ) be n-tuples of nonnegative integers. Then we say that the monomial xI is smaller in the
dominance order than xJ , denoted by xI dom xJ , if the associated
partitions satisfy
1 (I) 1 (J),
1 (I) + 2 (I) 1 (J) + 2 (J),
..
.
1 (I) + + n (I) 1 (J) + + n (J).
If xI dom xJ and xJ dom xI we say xI =dom xJ . (Note that
xI =dom xJ if and only if I and J are permutations of each other; see
Exercise 8 in this chapter.)
We explain next how we rewrite the orbit sum of an arbitrary
monomial in terms of orbit sums of special monomials and sn .
Let xI be a monomial. Let (I) be the associated partition.
Assume that xI does not satisfy the rst condition for being special.
Then there is a gap in the associated partition. We denote by
tI = min{i|i (I) i+1 (I) > 1}
the rst occurrence of such a gap. We dene the reduced monomial
of xI to be xIred , and denote it by
(Ired ) = (1 (I) 1, . . . , tI (I) 1, tI +1 (I), . . . , n (I)).
2Note that s is not special. However, all other elementary symmetric polynomin
als are (sums of) orbit sums of special monomials.

78

4. Permutation Representations

Thus we obtained the exponent sequence Ired from I by lowering the

tI largest exponents by 1. Note that the gap in the new partition is
smaller,
tI (Ired ) tI +1 (Ired ) = tI (I) tI +1 (I) 1,
thus the reduced monomial xIred is closer to being special than xI
was.
Example 4.19. Let us reconsider the example x11 x42 x13 F[x1 , . . . , x4 ].
We have tI = 1. The reduced monomial is x11 x32 x13 with associated
partition (3 1 1 0). This is still not special. However, if we
reduce this again (again tIred = 1) we obtain the special monomial
x11 x22 x13 with associated partition (2 1 1 0).
We want to use this reduction method in our algorithm. So that
this can work we need to show that this reduction method leads to
monomials that are lower with respect to the dominance order than
the original one. This is the content of the next proposition.
Proposition 4.20. Let I = (i1 , . . . , in ) be an n-tuple of nonnegative
integers. Any monomial xJ occurring in o(xIred ) stI is lower in the
dominance order than xI . We have equality if and only if xJ is a
term in o(xI ).
Proof. Denote by t = tI the rst occurrence of a gap in the partition
of I. Let xJ be a monomial occurring in o(xIred ) st . Let Ired =
red
(ired
1 , . . . , in ). We have
x J = x K xl 1 xl t ,
where
()

ired
1 (1)

xk11 xknn = xK = g(xIred ) = x1g

ired
1 (n)

xng

for some element g G. Thus K is a permutation of Ired , and

(K) = (Ired ). Moreover, note that

kl
if l  {l1 , . . . , lt },
()
jl =
kl + 1 if l {l1 , . . . , lt }.
We express the associated partition (J) = (1 (J) n (J))
in terms of the original partition (I) = (1 (I) n (I)) as
follows.

4.3. G
obels Bound

79

Set L = {l1 , . . . , lt } {1, . . . , n}. Denote by L the characteristic function of the subset L, i.e.,

1 if i L,
L : {1, . . . , n} {0, 1}, L (i) =
0 if i {1, . . . , n} \ L.
By () we have
ki = ired
g 1 (i) , i = 1, . . . , n,
for some g G. Thus () says that

if i  {l1 , . . . , lt },
ired
1
ji = gred (i)
ig1 (i) + 1 if i {l1 , . . . , lt }
= ired
g 1 (i) + L (i).
This means the following for the associated partition (J):

g1 (i) (I) 1 + L (i) if 1 i t,
()
i (J) =
g1 (i) (I) + L (i)
if t + 1 i n.
Next, in order to show that J dom I, we need to show that
s

i (J)

i=1

i (I)

i=1

for all s = 1, . . . , n. Assume that s t. We have

s

i=1

(1)

i (J) =

(g1 (i) (I) 1 + L (i))

i=1

=(

g1 (i) (I)) s +

i=1
(2)

s

i=1

L (i)

i=1
s
(3)

g1 (i) (I)

i (I),

i=1

s
where (1) is true because of (), (2) is true because i=1 L (i) s,
and (3) is true because reordering of (I) = (1 (I) n (I))
can only lead to a smaller sum for the rst s terms.

80

4. Permutation Representations
If s > t, then with () we obtain
s

i (J) =

i=1

i (J) +

i=1

i (J)

i=t+1



t 
s 


=
g1 (i) (I) 1 + L (i) +
g1 (i) (I) + L (i)
i=1

i=t+1


g1 (i) (I) t +

i=1
s

g1 (i) (I) t +

i=1

L (i) +

i=1
s


g1 (i) (I)

i=t+1

L (i).

i=1

Therefore
s

i=1

i (J) =

s

i=1

g1 (i) (I)t+

s

i=1

(1)

L (i)

s

i=1

g1 (i) (I) =

i (I),

i=1


where (1) follows from si=1 L (i) t. Finally, we have to show that
xJ is a term in o(xI ) if and only if xI =dom xJ .
If xJ is a term in o(xI ), then I and J are permutations of each
other and their associated partitions are equal: (I) = (J). Thus of
course xI =dom xJ .
To prove the converse, observe that if xI =dom xJ , then (I) =
(J). Therefore, J is a permutation of I. We need to prove that there
is an element g G that permutes I into J. To this end, note that by
denition the monomial xJ occurs in the orbit sum o(xIred )st . Thus,
we subtract 1 from the t largest exponents in the exponent sequence
I to obtain xIred . Then we permute the result by an element g of
G to get some monomial in the orbit of xIred . Finally we add 1 to
the same tI exponents in the result as we subtracted in the rst step,
because we are multiplying with st . Since the resulting monomial is
xJ we have that the element g G permutes I into J. Thus xJ is a


term of o(xI ).
We illustrate this result with an example.

4.3. G
obels Bound

81

Example 4.21. Let : Z/2 GL(4, R) be the faithful representation aorded by the matrix

0 1

1 0
.

0 1

1 0

Let xI = x1 x32 x3 ; thus I = (1, 3, 1, 0). This is a nonspecial monomial

with orbit sum
o(xI ) = x1 x32 x3 + x31 x2 x3 .
The gap occurs at t = 1, and its reduced exponent sequence is
Ired = (1, 2, 1, 0).
We compute
o(x1 x22 x3 )s1 = (x1 x22 x3 + x21 x2 x3 )(x1 + + x4 )
= x21 x22 x3 + x31 x2 x3 + x1 x32 x3 + x21 x22 x3
+ x1 x22 x23 + x21 x2 x23 + x1 x22 x3 x4 + x21 x2 x3 x4 .
The exponent sequences occurring in this polynomial have partitions
(2, 2, 1, 0),

(2, 1, 1, 1),

or (3, 1, 1, 0).

The sequences are lower in dominance order than the exponent sequence of our original xI . Also note that the two monomials
x31 x2 x3

and

x1 x32 x3

with the same dominance order occur in o(xI ). Hence,

o(xI ) = o(x1 x22 x3 )s1 o(x21 x22 x3 ) o(x1 x22 x23 ) o(x21 x2 x3 x4 ).
Furthermore, the rst three orbit sums on the right hand side are
orbit sums of special monomials. Thus we are left with the problem
of breaking up the last orbit sum:
o(x21 x2 x3 x4 ) = o(x1 )s4 .
Theorem 4.22 (M. G
obel). Let : G  GL(n, F) be a permutation
representation of a nite group. Then the ring of invariants F[V ]G
is generated as an algebra by the top elementary symmetric function
sn = x1 xn and the orbit sums of special monomials.

82

4. Permutation Representations

Proof. By Proposition 4.12 we need to show that the orbit sum of

any monomial can be written as a polynomial in sn and orbit sums
of special monomials. We denote by A F[V ]G the subalgebra generated by sn and all orbit sums of special monomials.
Let xI F[V ] be a nonspecial monomial. We use induction on
the dominance order to show that o(xI ) A.
The monomials with smallest dominance order are those of degree
zero. Thus our induction starts.
Assume that the result is proven for all monomials smaller in
dominance order than xI .
If n (I) = 0, then xI is divisible by sn . We have

xI = xI snn (I) ,

where xI is no longer divisible by sn . Since sn is invariant under G,

we have that

o(xI ) = o(xI ) snn (I) .

Thus it is enough to show the result for xI . In other words we assume,

without loss of generality, that n (I) = 0. Since xI was assumed to
be nonspecial, there is a gap in the associated partition (I) at tI .
We set
o(xI ) = o(xIred ) stI r(xI ).
The tI th elementary symmetric polynomial stI is a sum of orbit sums
of special monomials, because tI < n. Thus stI A. Moreover, the
orbit sum o(xIred ) belongs to A by induction, because the reduced
monomial is strictly lower in dominance order. Therefore, o(xIred )stI
belongs to A. Finally, we have
r(xI ) = o(xIred ) stI o(xI ).
Thus by Proposition 4.20 all the monomials that occur in r(xI ) are
lower in the dominance order than xI . Hence r(xI ) is in A by induc

tion, and therefore so is o(xI ).
What have we gained by this result?
First, special monomials have degree at most
the following corollary.

n
2 . Thus we have

4.3. G
obels Bound

83

Corollary 4.23 (G
obels Bound). Let : G  GL(n, F) be a permutation representation of a nite group. Then the ring of invariants
F[V ]G is generated as an
 algebra by homogeneous polynomials of degree at most max{n, n2 }.

Proof. The maximal degree of a special monomial is n2 . The degree
of sn is n.


Note that Gobels bound is valid for all permutation representations independent of the group or the ground eld. However, it
depends on n, the size of the representation (or the size of the set
being permuted). We will see in Chapter 6 that in characteristic zero
there is always an a priori degree bound on the generators. This
bound will be called Noethers bound.
The next example illustrates that G
obels bound is sharp, i.e.,
there exists a permutation representation of a nite group so that
any minimal generating set contains an element of highest possible
order:
Example 4.24 (Dening Representation of the Alternating Group).
Consider the alternating group An , n 4, in its dening representation. Certainly we have s1 , . . . , sn F[x1 , . . . , xn ]An . Furthermore,
the Vandermonde determinant

n =
(xi xj )
i<j

is invariant under An , because An consists of all even permutations.

The polynomial n is not symmetric, because for any transposition
n we have that
n = n .
Therefore
F[s1 , . . . , sn ]
F[s1 , . . . , sn , n ] F[x1 , . . . , xn ]An .
By what we have proven in this section, we know that in order to nd
the full ring of invariants we just have to calculate the orbit sums of
all special monomials and put them together with sn into an algebra.
If you do that (or write a small computer routine that does that for
you) you nd that F[V ]An is equal to F[s1 , . . . , sn , n ]. This is of
course not very satisfying. It is one of the goals of the next chapters

84

4. Permutation Representations

to nd methods to determine the ring

 of invariants more eectively.
obels bound is sharp.
For now, note that n has degree n2 . Thus G
See Exercise 13 in this chapter for the case of n = 3.
The preceding results also tell us that a ring of permutation invariants is always nitely generated, because they are only nitely
many special monomials (even though their number grows rapidly
with the number of variables).
Corollary 4.25. Let : G  GL(n, F) be a permutation representation of a nite group. Then the ring of invariants F[V ]G is nitely
generated as an algebra.
This statement remains true for all invariant rings of nite groups.
We will present two proofs of this fact in this text: an algorithmic
proof in Chapter 6 and an abstract proof in Chapter 10.
Finally, note that the results of this chapter yield an algorithm
to calculate any ring of permutation invariants:
Let : G  GL(n, F) be a permutation representation of a nite
group G.
rst: Calculate the orbit sums of all special monomials.
second: The desired ring is the F-algebra generated by all orbit sums
calculated in the rst step and sn .
Alternatively, you could
calculate the orbit sums of all monomials
of degree at most max{n, n2 }.
Note carefully that this algorithm terminates, because there are
only nitely many special monomials,
resp. nitely many monomials

n
of degree at most max{n, 2 }.
Example 4.26. Recall the regular representation of Z/4 from Example 4.1:

0
0
: Z/4 GL(4, F), 3 
0
1

1
0
0
0

0
1
0
0

0
0
.
1
0

4.3. G
obels Bound

85

By G
obels Theorem the ring of invariants is generated by orbit sums
of special monomials and s4 = x1 x2 x3 x4 . Since n = 4, special monomials have degree at most 6. The possible partitions are
(1 0 0 0), (1 1 0 0), (2 1 0 0), (1 1 1 0),
(2 1 1 0), (2 2 1 0), and (3 2 1 0).
So the following is a complete list of algebra generators (where we
noted when two orbit sums are obviously equal): In degree one there
is only the rst elementary symmetric function
o(x1 ) = o(x2 ) = o(x3 ) = o(x4 ) = s1 .
In degree two there are two orbit sums of special monomials (their
sum is s2 - why is that supposed to be so?)
o(x1 x2 ) = o(x2 x3 ) = o(x3 x4 ) = o(x4 x1 )
= x1 x2 + x2 x3 + x3 x4 + x4 x1 , and
o(x1 x3 ) = o(x2 x4 ) = x1 x3 + x2 x4 .
In degree three we nd
o(x21 x2 ) = o(x22 x3 ) = o(x23 x4 ) = o(x24 x1 )
= x21 x2 + x22 x3 + x23 x4 + x24 x1 ,
o(x21 x3 ) = o(x22 x4 ) = x21 x3 + x22 x4 ,
o(x21 x4 ) = o(x22 x1 ) = o(x23 x2 ) = o(x24 x3 )
= x21 x4 + x22 x1 + x23 x2 + x24 x3 , and
o(x1 x2 x3 ) = o(x2 x3 x4 ) = o(x3 x4 x1 ) = o(x4 x1 x2 )
= x1 x 2 x 3 + x 2 x 3 x 4 + x 3 x 4 x 1 + x 4 x 1 x 2
= s3 .
In degree four we have
o(x21 x2 x3 ) = o(x22 x3 x4 ) = o(x23 x4 x1 ) = o(x24 x1 x2 )
= x21 x2 x3 + x22 x3 x4 + x23 x4 x1 + x24 x1 x2 ,
o(x1 x22 x3 ) = o(x2 x23 x4 ) = o(x3 x24 x1 ) = o(x4 x21 x2 )
= x1 x22 x3 + x2 x23 x4 + x3 x24 x1 + x4 x21 x2 , and
o(x1 x2 x23 ) = o(x2 x3 x24 ) = o(x3 x4 x21 ) = o(x4 x1 x22 )
= x1 x2 x23 + x2 x3 x24 + x3 x4 x21 + x4 x1 x22 .

86

4. Permutation Representations

In degree ve we have
o(x21 x22 x3 ) = o(x22 x23 x4 ) = o(x23 x24 x1 ) = o(x24 x21 x2 )
= x21 x22 x3 + x22 x23 x4 + x23 x24 x1 + x24 x21 x2 ,
o(x21 x22 x4 ) = o(x22 x23 x1 ) = o(x23 x24 x2 ) = o(x24 x21 x3 )
= x21 x22 x4 + x22 x23 x1 + x23 x24 x2 + x24 x21 x3 , and
o(x21 x2 x23 ) = o(x22 x3 x24 ) = o(x23 x4 x21 ) = o(x24 x1 x22 )
= x21 x2 x23 + x22 x3 x24 + x23 x4 x21 + x24 x1 x22 .
Finally, in degree six we have
o(x31 x22 x3 ) = o(x32 x23 x4 ) = o(x33 x24 x1 ) = o(x34 x21 x2 )
= x31 x22 x3 + x32 x23 x4 + x33 x24 x1 + x34 x21 x2 ,
o(x31 x22 x4 ) = o(x32 x23 x1 ) = o(x33 x24 x2 ) = o(x34 x21 x3 )
= x31 x22 x4 + x32 x23 x1 + x33 x24 x2 + x34 x21 x3 ,
o(x31 x23 x2 ) = o(x32 x24 x3 ) = o(x33 x21 x4 ) = o(x34 x22 x1 )
= x31 x23 x2 + x32 x24 x3 + x33 x21 x4 + x34 x22 x1 ,
o(x31 x23 x4 ) = o(x32 x24 x1 ) = o(x33 x21 x4 ) = o(x34 x22 x1 )
= x31 x23 x4 + x32 x24 x1 + x33 x21 x2 + x34 x22 x3 ,
o(x31 x24 x2 ) = o(x32 x21 x3 ) = o(x33 x22 x4 ) = o(x34 x23 x1 )
= x31 x24 x2 + x32 x21 x3 + x33 x22 x4 + x34 x23 x1 , and
o(x31 x24 x3 ) = o(x32 x21 x4 ) = o(x33 x22 x1 ) = o(x34 x23 x2 )
= x31 x24 x3 + x32 x21 x4 + x33 x22 x1 + x34 x23 x2 .
Note that there are several orbit sums with the same associated partition. Note also that we do not know whether we really need all of
those: there might be an orbit sum that can be expressed in terms of
the others, like, e.g.,
o(x21 x2 x3 ) = s3 s1 4s4 o(x21 x3 x4 ) o(x21 x2 x4 ).
So G
obels Theorem gives us a complete set of generators, but
some of them might be redundant.
We close this chapter with another example.

4.4. Exercises

87

Example 4.27. The dihedral group of order 10 has a ve-dimensional

permutation representation
: D10  GL(5, F)
aorded by the

0
0

D=
0
0
1

matrices
1
0
0
0
0

0
1
0
0
0

0
0
1
0
0

0
0

1
0

and

0
1

S=
0
0
0

1
0
0
0
0

0
0
0
0
1

0
0
0
1
0

0
0

1
.
0
0

Following G
obels bound we calculate
the orbit sums of all special
5
monomials up to degree 10 = 2 . We obtain in degree one only s1 ;
in degree two we have s2 . The third elementary symmetric function
splits into two invariants, namely
x1 x2 x3 + x 2 x3 x4 + x 3 x4 x5 + x 4 x5 x1 + x 5 x1 x2
and
x1 x2 x4 + x2 x3 x5 + x3 x4 x1 + x4 x5 x2 + x5 x1 x3 .
Furthermore, we have the orbit sums of the monomials with partition
(2 1 0 0 0). In degree four there is the fourth elementary
symmetric function s4 and the orbit sums of the monomials with partition (2 1 1 0 0). In degree ve there is the fth elementary
symmetric function s5 and the orbit sums of the monomials with partitions (2 1 1 1 0) and (2 2 1 0 0). In degree six we
need the orbit sums of monomials with partitions (2 2 1 1 0)
and (3 2 1 0 0); in degree seven the orbit sums of monomials
with partitions (2 2 2 1 0) and (3 2 1 1 0); in
degree eight (3 2 2 1 0); in degree nine (3 3 2 1 0)
and in degree ten (4 3 2 1 0); see Exercise 14 in this chapter.

4.4. Exercises
(1)
(2)
(3)
(4)

Find the regular representation of V4 .

Find the regular representation of Q8 .
Find the regular representation of D8 .
Find a permutation representation of Z/6 of degree 6 that is not
the regular representation.

88

4. Permutation Representations

(5) Find a degree 4 representation of Z/4 that is not the regular

representation.
(6) Show that lexicographic order is a partial order.
(7) Show that dominance order is a partial order.
(8) Show that xI =dom xJ if and only if I is a permutation of J.
(9) Consider the three monomials x31 x2 x53 , x61 x2 x3 , and x71 x2 x3 in
C[x1 , . . . , x4 ].
(i) Which one is the largest in lexicographic order?
(ii) Which one is the largest in dominance order?
(iii) Are they special? If not, nd the corresponding reduced
monomials.
(10) Rewrite the following polynomials in C[x1 , x2 , x3 ]3 in terms of
the elementary symmetric polynomials:
f1 = x21 x2 + x22 x3 + x23 x1 + x21 x3 + x22 x1 + x23 x2 ,
f2 = x31 x2 x3 + x1 x32 x3 + x1 x2 x33 ,
f3 = x41 + x42 + x43 .
(Why do you know that these polynomials are invariant under
the 3 -action?)
(11) Consider the representation of Z/2 of Example 4.21. Rewrite
the orbit sum of x1 x42 x3 as a polynomial in orbit sums of special
monomials. Do the same for the monomial x1 x43 x74 .
(12) Find the ring of invariants of the Z/2-representation given in
Example 4.8.
(13) Find the ring of invariants of the regular representation of Z/3.
(Hint: This representation coincides with the dening representation of the alternating group in three letters.) Show that there
are invariants that can be written in more than one way in terms
of the algebra generators.
(14) Complete Example 4.27:
(i) Show that the orbit length of a monomial is 1, 2, 5, or 10.
(ii) Convince yourself that the partitions given in that example
are all that we have to consider.
(iii) Show that there are 30 monomials with the partition (2
1 1 0 0).
(iv) Count the number of special monomials of degree at least
5.

4.4. Exercises

89

(v) Make a complete list of generators of the ring of invariants.

(15) Find the ring of invariants of the regular representation of V4 .
(16) Let T = {t1 < t2 < < tk } be an ordered subset of the set
{1, . . . , n 1}. Set tk+1 = n. We dene the Young subgroup
YT of n associated to the descent set T by
YT = { n |([ti , ti+1 ]) = [ti , ti+1 ] i = 1, . . . , k},
where [ti , ti+1 ] = {ti , ti + 1, . . . , ti+1 } denotes the interval of integers from ti to ti+1 . (In other words, we put n things into k
boxes and then permute the things in the boxes independently.)
(i) Show that YT is indeed a group.
(ii) What is the order of YT ?
(iii) Take the dening representation of n and restrict it to YT .
What is the corresponding ring of invariants?
(17) Prove the following generalization of Newtons formulae (Proposition 4.7): Let l1 , . . . , lr V be linear forms. Then
isi (l1 , . . . , lr ) =

(1)k1 pk (l1 , . . . , lr )sik (l1 , . . . , lr ),

k=1

where s0 (l1 , . . . , lr ) = 1, si (l1 , . . . , lr ) is the ith elementary symmetric polynomial in the li s, and pi (l1 , . . . , lr ) is the ith symmetric power sum in the li s.
(18) Read about Sir Isaac Newton [1646-1723].
(19) Read about Edward Waring [1734-1798].
(20) Read about Carl Friedrich Gauss [1777-1855].
(21) Who gave the Vi`ete polynomials their name?
(22) The Vandermonde determinant is named after Alexandre-Theophile Vandermonde [1735-1796]. Why is the Vandermonde determinant a determinant?
(23) Alfred Young [1873-1940] gave Young groups their name. Read
about him.

Application: Decay of a
Spinless Particle

We want to illustrate the results we achieved so far with an example

from particle physics taken from the diploma thesis of Dirk Engelmann.1
Consider the decay of a spinless boson into four identical spinless
bosons. Each of these four nal particles has a 4-momentum pi (i =
1, . . . , 4). Mathematically speaking the pi s are just 4-dimensional
vectors pi = (Ei , pi1 , . . . , pi3 ), where the rst component describes the
total energy and the last three components describe the 3-momentum.
We obtain the six basic Minkowski scalar products
pi pj = Ei Ej (pi1 pj1 + pi2 pj2 + pi3 pj3 )
for 1 i < j 4, which encode the basic physical properties.
Since our four bosons are identical, any permutation of them does
not change the physical situation. However, say we interchange boson 1 and boson 2, then the scalar product p1 p3 is interchanged with
p2 p3 . Thus we obtain a permutation of the set of basic scalar products. In other words, setting vij = pi pj , we are looking at a 4 -action
on the 6-dimensional vector space spanned by vij for 1 i < j 4
1Dirk Engelmann, Polynominvarianten einer Darstellung der symmetrischen
Gruppe, University of G
ottingen, 1995.

91

92

Application: Decay of a Spinless Particle

induced by the dening representation of 4 :


v(i)(j) if (i) < (j),
(vij ) =
v(j)(i) otherwise.
Now you want to know which expressions in the vij s are invariant
under this 4 -action.
Let us step back and assume we have three nal identical particles. Then we have to study a 3 -action on a 3-dimensional vector
space spanned by v23 , v13 , v12 . Which representation is that? Let us
check that: We keep the basis vectors in the above order. Then we
have
: 3 GL(3, C),
where

0 1 0
(12) = 1 0 0
0 0 1

and

0 0 1
(13) = 0 1 0 .
1 0 0

This is just the dening representation of the symmetric group on

three letters. By Theorem 4.13 its ring of invariants C[x23 , x13 , x12 ]3
is generated by the three elementary symmetric functions in the basis
elements x23 , x13 , x12 .
However, for four nal particles the situation is a bit more complicated: Our vector space has dimension 6 with basis elements
v12 , v13 , v14 , v23 , v24 , v34 .
Our representation is
: 4 GL(6, C)
with

1
0

0
(12) =
0

0
0

0
0
0
1
0
0

0
0
0
0
1
0

0
1
0
0
0
0

0
0
1
0
0
0

0
0
0
0

0
0
, (13) =
1
0

0
0
1
0

0
1
0
0
0
0

0
0
0
0
0
1

1
0
0
0
0
0

0
0
0
0
1
0

0
0

1
, and
0

0
0

Application: Decay of a Spinless Particle

93

0 0 0 0 1 0
0 0 0 0 0 1

0 0 1 0 0 0
(14) =
.
0 0 0 1 0 0

1 0 0 0 0 0
0 1 0 0 0 0
So, what is the ring of invariants? According to G
obels Theorem,
Theorem 4.22, we have to check the orbit sums of all special monomials. They have degree at most 15, and there are 332 of them!
In his diploma thesis Dirk Engelmann developed methods to treat
representations
like this for arbitrary symmetric groups: Let V be

a n2 -dimensional complex vector space. Denote its basis by vij for
1 i < j n. Then consider the representation of the symmetric
group on n letters
 
n
: n GL(
, C)
2
given by

v(i)(j) if (i) < (j),
(vij ) =
v(j)(i) otherwise
for all n . Now nd its ring of invariants! In our case, n = 4,
he obtained the following set of C-algebra generators of the ring of
invariants C[x12 , . . . , x34 ]4 :
s1 = x12 + + x34 ,
p2 = x212 + + x234 ,
p3 = x312 + + x334 ,
p4 = x412 + + x434 ,
o(x12 x34 ) = x12 x34 + x13 x24 + x14 x23 ,
o(x12 x13 x14 ) = x12 x13 x14 + x12 x23 x24 + x13 x23 x34 + x14 x24 x34
and the orbit sums
o(x212 x13 ), o(x212 x13 x23 ), o(x412 x13 ), o(x412 x213 ), and o(x312 x313 x223 x14 ).
We note that the rst elementary symmetric function s1 and the
power sums p2 , p3 , and p4 are invariant under the full symmetric
group on six letters. Furthermore, we note that some of his invariants
are not orbit sums of special monomials, because he had developed

94

Application: Decay of a Spinless Particle

other methods avoiding the calculation of all orbit sums of special

monomials and better suited for this type of n -action.

Application: Counting
Weighted Graphs

The representation of the symmetric group n in the preceding application appears also in graph theory as we show in this section.1
A graph consists of a set of vertices
V () = {v1 , v2 , . . .}
and a set of edges
E() = {eij if vi , vj are connected}
between them. Two graphs and
are isomorphic if there exists
a bijective map
: V () V (
)
between the sets of vertices such that it induces a bijection
: E() E(
), eij  e
(i)(j)
between the sets of edges.
A weighted graph is a graph such that every edge eij has a
weight mij C. Two weighted graphs are weighted isomorphic if
the bijection maps edges onto edges of the same weight.
1The material of this section is taken out of the PhD thesis of Nicolas M. Thi
ery,
Invariants alg
ebriques de graphes et reconstruction. Une
etude exp
erimentale., PhD
thesis, University of Lyon I, 1999.

95

96

Application: Counting Weighted Graphs

We want to count the number of isomorphism classes of weighted

graphs with, say, n vertices. This translates into counting invariants
of the same type of representations we encountered in the previous
application:
Consider the ring of polynomials
n

C[xij |1 i < j n]

in 2 variables. Then n acts on C[xij |i = j, i, j = 1, . . . , n] by

acting on the indices

x(i)(j) if (i) < (j),
(xij ) =
x(j)(i) otherwise.
We arrange the basis elements xij in a symmetric matrix

xij for i < j,

Xij = xji for i > j,

0
otherwise
with zeros on the diagonal. The symmetric group acts by conjugation,
Xij 1 , where we have identied n with its image under the
dening representation. This is exactly the representation of n we
obtain when considering the decay of a particle. In graph theory this
becomes useful by the following observation:
Assume that you have two weighted isomorphic graphs and

with n vertices. The isomorphism : V () V (
) is nothing but
a permutation of the vertices inducing a permutation on the edges as
well. Thus setting m = (mij ) we obtain
f (m) = f (m
)
for all invariant polynomials f C[xij |1 i < j n]n .
The converse is also true: Assume that you have two weighted
graphs and
such that f (m) = f (m
) for all polynomials f
C[xij |1 i < j n]n . Then
m = (m
)
for some n . Thus provides us with the desired isomorphism
between and
.