Introduction to Symmetric Functions

Chapter 2

Mike Zabrocki
Abstract. A development of the symmetric functions using the plethystic notation.
 CHAPTER 3

The algebra structure of the ring of symmetric functions

Consider the polynomial ring in the variables pi for i ≥ 1, Λ = Q[p1 , p2 , p3 , . . .]. We will
define the degree of a variable pk to be k and so the degree of a monomial pk1 pk2 · · · pk` is
simply k1 + k2 + · · · + k` . Λ is the ring of symmetric functions.

This is a very abstract way to begin, but at the end of this chapter we will draw a connection
between this algebra and the space of class functions of the symmetric group. From this
perspective Λ can be seen as an infinite dimensional graded vector space where the symmetric
functions of degree m are a finite dimensional subspace.

The elements pi are refered to as the power generators and since we are considering them as
the variables in a commutative polynomial ring the space is spanned by the monomials in
these variables. To specify a basis of this space we may assume that the variables are listed
in weakly decreasing order. That is, if we denote Λm by the symmetric functions of degree
m, then the set {pλ : λ ` m} forms a basis for Λm where pλ := pλ1 pλ2 · · · pλ`(λ) . The set
L
m≥0 {pλ }λ`m is called the power basis. Note that the degree of a monomial pλ is given by
|λ|.

Λ has a natural ‘un-multiplication’ operation called a coproduct. In the sense if the product
represents a way of putting elements in the algebra together, the coproduct represents ways
of pulling elements apart. This can be a very interesting operation, especially when the
multiplication and comultiplication interact.

Formally, we define the multiplication function on this algebra µ : Λ ⊗ Λ −→ Λ as

(3.1) µ(pλ ⊗ pµ ) = pλ pµ = pλ1 · · · p`(λ) pµ1 · · · pµ`(µ) = pλ]µ .

The comultiplication will be denoted as ∆ : Λ −→ Λ ⊗ Λ is given by ∆(pk ) = 1 ⊗ pk + pk ⊗ 1

(this is the property that the power generators are primitive in this algebra). We impose
that it is a ring homomorphism, that is for f, g ∈ Λ, ∆(f g) = ∆(f )∆(g) and and c, d ∈ Q,
∆(cf + dg) = c∆(f ) + d∆(g). Hence for an arbitrary basis element we have

Proposition 3.1.

λ1  
X Y mi (λ) X zλ
(3.2) ∆(pλ ) = pµ ⊗ pν = pµ ⊗ pν
µ]ν=λ i=1
mi (µ) µ]ν=λ
zµ zν

23
24 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

By rearranging the coefficients of (3.2) we can see a natural basis to consider is pλ /zλ because
it arises naturally in this formula in the sense that
 
pλ X pµ pν
(3.3) ∆ = ⊗ .
zλ µ]ν=λ
zµ zν

Proof. First we note that the coefficients in the two formulations of equation (3.2) are
equal since
λ1 λ1 λ1 
imi (λ) mi (λ)!

zλ Y Y mi (λ)! Y mi (λ)
= mi (µ) m (µ)!imi (ν) m (ν)!
= = .
zµ zν i=1
i i i i=1
m i (µ)!m i (ν)! i=1
m i (µ)
We show this proposition by induction on the number of parts of λ. We have a base case
since the formula clearly works if λ has only one part by definition.

Let λ be a partition of n and denote λ = (λ2 , λ3 , . . . , λ`(λ) ). It follows that

 
λ2  
X Y mi (λ)
(3.4) ∆(pλ ) = ∆(pλ1 )∆(pλ ) = (pλ1 ⊗ 1 + 1 ⊗ pλ1 )  pµ ⊗ pν  .
i=1
m i (µ)
µ]ν=λ

mi (λ) mi (λ)



If λ1 6= λ2 there is nothing to prove since for all 1 ≤ i ≤ λ2 and the
mi (µ)
= mi (µ)
mλ1 (λ)

expansion of the right hand side is exactly as stated in the proposition since m (µ)
= 1.
λ1

If λ1 = λ2 then mλ1 (λ) = mλ1 (λ) − 1 and mi (λ) = mi (λ) for 1 ≤ i < λ1 , hence we see
 
mλ1 (λ)
 X X mλ (λ) − 1 Y mi (λ) 
1
∆(pλ1 )∆(pλ ) = (pλ1 ⊗ 1 + 1 ⊗ pλ1 ) 
 pµ ⊗ pν 
j=0
j 1≤i<λ
mi (µ) 
µ]ν=λ 1
mλ1 (µ)=j

mλ1 (λ)     Y  
X X mλ1 (λ) − 1 mλ1 (λ) − 1 mi (λ)
= + pµ ⊗ pν
j=0
j j−1 1≤i<λ
m i (µ)
µ]ν=λ 1
mλ1 (µ)=j
X mλ (λ) Y mi (λ)
1
= pµ ⊗ pν .
mλ1 (µ) 1≤i<λ mi (µ)
µ]ν=λ 1

And so it follows by induction that this formula holds for all partitions λ. 

The pk have the property that ∆(pk ) = pk ⊗ 1 + 1 ⊗ pk and hence are called the primitive
elements of this algebra. We remark that there is a map S called the antipode with the
property
(3.5) µ ◦ (id ⊗ S) ◦ ∆(f ) = 0
for all f ∈ Λ such that f has 0 constant term. We set S(pλ ) = (−1)`(λ) pλ and extend this
map linearly and it is easy to check that µ ◦ (id ⊗ S) ⊗ ∆(pk ) = µ(pk ⊗ 1 − 1 ⊗ pk ) = 0 and
 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS 25

similarly for pλ . This implies that µ ◦ (id ⊗ S) ◦ ∆(f ) is equal to the constant term of f for
all f ∈ Λ.

Therefore, so far our algebra of symmetric functions is very simple, but we should develop
some intuitive ideas on how to picture what this algebra is. Now if f ∈ Λ, then f is some
polynomial in variables pi . Since we are using partitions to index our basis we often just
write pλ when we talk about the basis elements when we consider Λ as a vector space over
Q. The indexing set of partitions can be represented by their Young diagrams so when we
take the product of pλ and pµ , µ(pλ ⊗ pµ ) represents shuffling the Young digrams together.
Example 14. Consider for instance λ = (6, 3, 1) and µ = (5, 2, 2). This can be represented
by the picture

This diagram is simply representing the equation p(6,3,1) p(5,2,2) = p(6,5,3,2,2,1) .

We should also develop a combinatorial picture of what happens when ∆ acts on a term pλ .
Because the pk are primitive elements, there will be 2`(λ) terms in the expansion of ∆(pλ ).
Example 15. We compute the action of ∆ on p(5,2,2) and do this by computing the number
of ways of coloring the rows of the partition (5, 2, 2) using two colors so that the whole row
has the same color. This is represented by the following picture.

The blue rows will be in the left tensor and red will be in the right tensor (although the
colors are symmetric) so we have determined
 
∆ p =p ⊗ 1 + 2p ⊗p

+p ⊗p + 2p ⊗p
+p ⊗p +1⊗p .

Here we are splitting the partition up into pieces such that their union is the original partition
in all possible ways. Notice that the sum of the coefficients in this expression is 8 = 2`(5,2,2) .
This picture will help us gain some intuition as we develop this algebra more completely.
Remark 1. The reader is encouraged to try to develop some sort of a picture each time
a formula appears in this presentation since the formulas are difficult to appreciate unless
some meaning is assigned to the symbols we are working with.
26 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

There are generally considered to be 6 ‘standard’ bases of the symmetric functions since
these bases are fundamental in the development of tools to describe the calculus of symmetric
functions. After the power symmetric basis we will introduce the homogeneous basis and the
elementary basis as these are defined as products of generators. We will save the definition
of the Schur basis and the monomial basis for a later section. 1

For n > 0, we set

X
(3.6) hn = pλ /zλ ,
λ`n

these are the homogeneous generators. We also define

X
(3.7) en = (−1)|λ|−`(λ) pλ /zλ
λ`n

which are the elementary generators. For a partition λ we set hλ := hλ1 hλ2 · · · h`(λ) and
eλ := eλ1 eλ2 · · · e`(λ) . Clearly we have the triangularity relations that hλ = pλ / `(λ)
Q
i=1 λi +
terms containing pµ with µ smaller than λ in lexicographic order (and a similar relation
with eλ ). This implies that {hλ }λ and {eλ }λ are bases for the symmetric functions and that
Λ = Q[h1 , h2 , h3 , . . .] = Q[e1 , e2 , e3 , . . .]. Also set as a convention p0 = h0 = e0 = 1 and
p−n = h−n = e−n = 0 for n > 0, so that formulas which require us to refer to these elements
make sense.

There are several ways of picturing what the elements hn and en represent. In some sense,
n!hn is the generating function of all permutations of the symmetric group Symn with weight
1 for each element which we can see in the following formula.
X X n!
(3.8) n!hn = pλ(σ) = . pλ .
σ∈Symn λ`n
zλ

At the same time n!en is a signed generating function with weight equal to (−1)n−`(λ) if the
permutation has cycle type λ.
X
(3.9) n!en = (σ)pλ(σ) .
σ∈Symn

We will see when we introduce the Schur functions that these formulas are a special case of
one where the elements of Λ are generating functions for the irreducible characters of the
symmetric group and hn is representing the trivial character and en is representing the sign
character.

1There is another basis which is typically called the forgotten basis which completes the analogy, ‘the
homogeneous basis is to the monomial basis as the elementary basis is to the (um, I forget) basis.’ There are
few direct formulas for the forgotten basis except for those which are analogous to those for the monomial
basis and hence remains somewhat underdeveloped in our account.
 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS 27

Set P (t) = r≥1 pr tr /r as a generating function for the power generators and set H(t) =
P
exp(P (t)). Notice by the following calculation we have
!
X Y
H(t) = exp pr tr /r = exp(pr tr /r)
r≥1 r≥1
Y X pn
r
(3.10) = n n!
tnr
r≥1 n≥0
r
X X pλ X
= tk = hk tk .
k≥0 λ`k
zλ k≥0
n n
P
Similarly we may easily show that E(t) = exp(−P (t)) = n≥0 (−1) en t . Simply by defini-
tion of these generating functions we have the relation
(3.11) H(t)E(t) = exp(P (t))exp(−P (t)) = 1
We can also consider the product of these generating functions explicitly and take the coef-
ficient of tn . On the right hand side the coefficient is 0 as long as n > 0 and the coefficient
on the left hand side shows that
X n
(3.12) (−1)k hn−k ek = 0.
k=0

Define the ring homomorphism on Λ that sends ω(pk ) = (−1)k−1 pk . Clearly, ω is an invo-
lution and is related to the antipode map on Λ by ω(pλ ) = (−1)|λ| S(pλ ). By going back to
formulas (3.6) and (3.7) for hk and ek in terms of pλ we see that ω relates the {hλ }λ and
{eλ }λ bases by ω(hλ ) = eλ .

By exploiting the generating functions P (t), H(t) and E(t) further we can extract other
algebraic relations between the elements of this ring. For instance, notice that P (t) =
log(H(t)) and hence P 0 (t) = H 0 (t)/H(t). Therefore by taking the coefficient of tn−1 in
P 0 (t)H(t) = H 0 (t) we see that
Xn
(3.13) nhn = hn−k pk .
k=1

By an application of ω on each side of this equation we also see that

Xn
(3.14) nen = (−1)k−1 en−k pk
k=1

Equations (3.12), (3.13) and (3.14) give us a simple recursive method to express any of the
algebraic generators of this space in terms of any other because the term containing hn , en or
pn can be isolated to provide algebraic relations. These recursive definitions will be exactly
the method that we use when we develop computer functions in Maple to change between
bases.
Example 16. If we wish to expand h3 in the elementary basis we note that h3 = h2 e1 −
h1 e2 + e3 , h2 = h1 e1 − e2 and h1 = e1 . Combining these we find that h3 = e31 − 2e2 e1 + e3 .
28 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

We may also use these equations derive simple results using calculations by induction. For
instance, (3.13) may be used to derive by induction by induction the following action of the
coproduct ∆, acting on the symmetric function hk .
Proposition 3.2.
n
X
(3.15) ∆(hn ) = hk ⊗ hn−k
k=0

Proof. Assume by induction that we know that equation (3.15) is true for all k < n.
Then we know that
n
X
∆(nhn ) = ∆(pr hn−r )
r=1
n n−r n−r
!
X X X
= (pr ⊗ 1) hn−r−k ⊗ hk + (1 ⊗ pr ) hk ⊗ hn−r−k
r=1 k=0 k=0
n n−r n−r
!
X X X
= pr hn−r−k ⊗ hk + hk ⊗ pr hn−r−k
r=1 k=0 k=0

n−1 X
X n−k n−1 X
X n−k
= pr hn−r−k ⊗ hk + hk ⊗ pr hn−r−k
k=0 r=1 k=0 r=1
n−1
X n−1
X
= (n − k)hn−k ⊗ hk + hk ⊗ ((n − k)hn−k )
k=0 k=0
n−1
X
= nhn ⊗ 1 + n hk ⊗ hn−k + 1 ⊗ nhn
k=1
n
X
=n hk ⊗ hn−k
k=0


This last result gives us an interesting combinatorial way of looking at the action of ∆ on
the functions hλ . On a single hn , ∆ acts by summing over all possible ways of breaking up
a block of size n into two pieces whose sum is n. Therefore when ∆ acts on an hλ , we can
use this idea to come up with a combinatorial interpretation for the coefficient of hµ ⊗ hν in
∆(hλ ).

Think of the rows of µ as red blocks with labels 1, 2, . . . , `(µ) whose horizontal lengths are
µ1 , µ2 , . . . , µ`(µ) and the rows of ν are represented by blue blocks with horizontal lengths
ν1 , ν2 , . . . , ν`(ν) . When ∆ acts on hλ it splits each of the rows of λ into two parts (with
some of the parts possibly empty) and so we can interpret the coefficient ∆(hλ ) as the
hµ ⊗hν
number of ways of taking at most one red block and at most one blue block placing it next
to each other to get rows of size λ1 , λ2 , . . . , λ`(λ) .
 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS 29

Example 17. We wish to compute ∆(h(4,3,3) ) , then we break up the rows of

h(2,2,1) ⊗h(2,2,1)
the partition (4, 3, 3), each one into a red part and a blue part (possibly empty) such that
red pieces sorted are a partition (2, 2, 1) and the blue pieces sorted are a partition (2, 2, 1).
The can be done in exactly two ways,

Equation (3.14) can also be used to derive the following determinantal formula for pn in
terms of ek .

nen en−1 en−2 · · · e1

(n − 1)en−1 en−2 en−3 ··· 1
(3.16) pn = (n − 2)en−2 en−3 en−4 ··· 0
.. .. .. ..
. . ··· . .
e1 1 0 ··· 0
This follows directly from equation (3.14) by expanding the determinant about the first
row of the equation. It follows that the determinant satisifes the same recurrence as the pk
elements do in equation (3.14). Similarly, we also have
nhn hn−1 hn−2 · · · h1
(n − 1)hn−1 hn−2 hn−3 ··· 1
(3.17) (−1)n−1 pn = (n − 2)hn−2 hn−3 hn−4 ··· 0
.. .. .. .
. . . · · · ..
h1 1 0 ··· 0
which follows most easily by an application of the involution ω or by observing the same
recurrence with equation (3.13).

There is another product defined on symmetric functions known as the ‘Kronecker’ or ‘inner
tensor’ product. We will denote this product by ∗. It is defined on the power sum basis by
pλ pµ pλ
(3.18) ∗ = δλµ .
zλ zµ zλ
This product is associative and preserves the degree of the symmetric function, that is it
maps Λn ⊗Λn → Λn . We also know that the product is commutative f ∗g = g ∗f since clearly
this holds on the power basis. We will go into more detail of this product as we introduce
more of the bases of the symmetric functions. The Kronecker product very naturally arises in
the algebra of class functions and we will show that it is also nicely encoded in our notation.

There is an assocated coproduct with ∗ that defines a bialgebra structure on the symmetric
functions. Define the corresponding coproduct as ∆0 : Λ → Λ ⊗ Λ will be defined on the
power basis
(3.19) ∆0 (pλ ) = pλ ⊗ pλ .
30 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

Proposition 3.3. The vector space Λ endowed with the product µ(f ⊗g) = f g and coproduct
∆0 forms a bialgebra.

Proof. ∆0 (pµ pλ ) = (pµ pλ ) ⊗ (pµ pλ ). Similarly we have that ∆0 (pµ )∆0 (pλ ) = (pµ ⊗
pµ )(pλ ⊗ pλ ) = (pµ pλ ) ⊗ (pµ pλ ) = ∆0 (pµ pλ ). Therefore we have shown that ∆0 is an algebra
homomorphism with respect to the product. 

Note: With the map ε0 (pk ) = 1 (and more generally ε0 (pλ ) = 1) is a counit and satisfies
µ ◦ (id ⊗ ε0 ) ◦ ∆0 = id, however this product/coproduct pair fails to have an antipode and
hence is not a Hopf algebra (see exercise 12).

∆0 is not an algebra homomorphism with respect to the Kronecker product since ∆0 (pµ ∗pλ ) =
δλµ zλ pλ ⊗ pλ while ∆0 (pµ ) ∗ ∆0 (pλ ) = δλµ zλ2 pλ ⊗ pλ and hence it is not a bialgebra.

In addition to the Hopf algebra and bialgebra structures on Λ, one should also think of Λ as
a vector space over Q and so it is convenient to define a scalar product on this to serve as a
tool for computation. Define
 
pµ
(3.20) pλ , = δλµ
zµ
where we use the notation δxy = 0 if x 6= y and δxx = 1. The remarkable property of this
scalar product is that it interacts nicely with the products and coproducts on this space.
Proposition 3.4. The scalar product is positive definite. In addition, it satisfies the fol-
lowing useful properties.
(3.21) hf, gi = hg, f i
If we set hf1 ⊗ f2 , g1 ⊗ g2 i⊗ = hf1 , g1 i hf2 , g2 i, then the coproduct ∆ is dual to multiplication,
(3.22) hf ⊗ g, ∆(h)i⊗ = hf g, hi .
The coproduct ∆0 is dual to the product ∗,
(3.23) hf ⊗ g, ∆0 (h)i⊗ = hf ∗ g, hi .
The involution ω and antipode S are self-dual,
(3.24) hω(f ), ω(g)i = hS(f ), S(g)i = hf, gi .
Moreover,
(3.25) hf, gi = ε0 (f ∗ g)
and
(3.26) hf ∗ g, hi = hg, f ∗ hi

Proof. It suffices to verify these identities for a basis and then the result must extend
by linearity, and for this we choose the basis {pλ }λ . Note that the fact that the scalar
product is symmetric follows since hpλ , pµ i = hpµ , pλ i = δλµ zλ . We show equation (3.22) by
 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS 31

expanding the left hand side using equation (3.2) and comparing it to the right hand side of
the equation.
X Y mi (ν)
(3.27) hpλ ⊗ pµ , ∆(pν )i⊗ = hpλ , pγ i pµ , pν\γ
γ i≥1
mi (γ)
Y mi (ν)
(3.28) = δν,λ]µ zλ zµ
i≥1
m i (λ)
On the right hand side of this equation we have
(3.29) hpλ pµ , pν i = δν,λ]µ zν
and by referring back to the definition of zν it is easy to see that (3.28) and (3.29) are equal.

Similarly we may verify (3.23),

(3.30) hpλ ⊗ pµ , ∆(pν )i⊗ = hpλ , pν i hpµ , pν i = δλµ δµν zν2
= hzµ δλµ pµ , pν i = hpλ ∗ pµ , pν i

(3.31) hω(pλ ), ω(pµ )i = (−1)|λ|+|µ|−`(λ)−`(µ) hpλ , pµ i =

(3.32) hS(pλ ), S(pµ )i = (−1)`(λ)+`(µ) hpλ , pµ i = δλµ zλ = hpλ , pµ i
Equation (3.25) follows because we have set pλ ∗ pµ = δλµ zλ pλ and hence ε0 (pλ ∗ pµ ) =
δλµ zλ ε0 (pλ ) = δλµ zλ = hpλ , pµ i. This implies our last identity as well since the product ∗ is
associative and symmetric and
(3.33) hf ∗ g, hi = ε0 (f ∗ (g ∗ h)) = ε0 (g ∗ (f ∗ h)) = hg, f ∗ hi


Now for any symmetric function homomorphism we can ask what the operation which is
dual with respect to the scalar product. That is, for φ ∈ Hom(Λ, Λ) we ask what is the
operator φ∗ with the property
(3.34) hφ(f ), gi = hf, φ∗ (g)i
Notice that we have already shown that many of the operators which we have considered so
far (e.g. S, ω, the action Kronecker product by a symmetric function f ∗ ·) are self dual. In
the last proposition we also showed that the coproduct ∆ is dual to the product operation
m and the coproduct ∆0 is dual to the Kronecker product.

This leads to a useful computational tool, the operation which is dual to multiplication of a
symmetric function f , which we will denote by f ⊥ . That is, f ⊥ is defined as the operator
with the property
hf · g, hi = g, f ⊥ h .
Since multiplication by f is an operation which raises the degree of a symmetric function by
the degree of f , f ⊥ is an operator which lowers the degree of a symmetric function by the
degree of f .
32 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

We show the following useful properties of this operation.

Proposition 3.5.
X
(3.35) f ⊥ (g) = hf pλ , gi pλ /zλ
λ

p⊥ ∂
k = k ∂pk and, in particular,

(3.36) p⊥
k (pλ ) = kmk (λ)pλ (k)
P
where pλ (k) is zero if λ does not have a part of size k. If ∆(f ) = i ai ⊗ bi , then
X
(3.37) f ⊥ (gh) = a⊥ ⊥
i (g)bi (h)
i

Proof. The first equation follows because the coefficient of pλ inPf ∈ Λ is given by
hpλ /zλ , f i, therefore the expansion of f in the pλ basis is simply f = λ hpλ /zλ , f i pλ . In
the case that we expand f ⊥ (g) in the power basis, we have that
X X
(3.38) f ⊥g = pλ /zλ , f ⊥ (g) pλ = hf pλ , gi pλ /zλ .
λ λ

The coefficient of pµ in p⊥ ⊥
k pλ is given by pk pλ , pµ /zµ = hpλ , pk pµ /zµ i which is equal to 0
unless µ ] (k) = λ. If µ = λ (k), then the scalar product evaluates to zλ /zµ = kmk (λ) and
otherwise the result is 0. It follows that p⊥ ∂
k = k ∂pk since the action of these operators is the
same on the monomial pλ .

The fact that p⊥ ⊥ ⊥

k (f g) = pk (f )g + f pk (g) follows from the product rule for derivatives since
⊥
we can interpret pk as a differential operator. Since ∆(pk ) = pk ⊗ 1 + 1 ⊗ pk we have shown
that equation (3.37) holds for any pk . We know that ∆(pλ ) = ∆(pλ1 )∆(pλ2 ) · · · ∆(pλk ) while
p⊥ ⊥ ⊥ ⊥
λ = pλ1 pλ2 · · · pλk , therefore (3.37) must hold for any pλ . It follows by extending this result
linearly that it also holds for any symmetric function f . 
Proposition 3.6. For k ≥ 0, the action of the operators p⊥ ⊥ ⊥
k , ek and hk for k ≥ 0 on the
symmetric functions pn , en and hn is given by the following table.
hn en pn
h⊥
k hn−k (δk0 + δk1 )en−k δkn + δk0 pn
(3.39)
e⊥
k (δk0 + δk1 )hn−k en−k (−1)k−1 δkn + δk0 pn
p⊥
k hn−k (−1)k−1 en−k nδnk + δk0 pn

Proof. e⊥ ⊥ ⊥ ⊥
k (en ), ek (hn ), ek (pn ) and pk (en ) can all be calculated from the action of the
operator h⊥ ⊥ ⊥
k or pk since we have that ω(f g) = (ω(f )) (ω(g)).
⊥

pλ pλ
p⊥
k ( zλ ) is zλ
(k)
if λ contains a part of size k and 0 if λ does not contain a part of size k.
(k)

p⊥
P P
Therefore if acts on hn =
k λ`n pλ /zλ the result will be hn−k = λ`n−k pλ /zλ . This
justifies the last line of the table.
 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS 33

If k = 0 then h⊥ ⊥
k = 1, but otherwise hk (pn ) = 0 unless k = n since only then will there be a
partition which contains a part of size n so that λ`k p⊥
P
λ (pn )/zλ is non-zero.

We will prove h⊥
k hn using induction and formula (3.13).
k k
1X ⊥ ⊥ 1X ⊥
(3.40) h⊥
k (hn ) = hk−i pi (hn ) = h (hn−i )
k i=1 k i=1 k−i

since we have already calculated that p⊥ i (hn ) = hn−i . If we assume by induction that
h⊥ (h
k−i n−i ) = hn−k for 1 ≤ i ≤ k, then it follows that h⊥
k (hn ) = hn−k .

We can prove in a similar manner the formula for h⊥ k (en ). We handle the k = 0 and k = 1
cases separately for there we already know h0 (en ) = en and h⊥
⊥ ⊥
1 (en ) = p1 (en ) = en−1 .

For k > 1, if we assume that h⊥

k−i (en ) is known for all i > 0 and all n then we calculate that
k k
1X ⊥ ⊥ 1X ⊥
(3.41) h⊥
k (en ) = hk−i pi (en ) = hk−i (−1)i−1 en−i
k i=1 k i=1
Only two terms of this equation will survive, i = k − 1 and i = k. Therefore,
1
(3.42) h⊥
k (en ) = ((−1)k en−k + (−1)k−1 en−k ) = 0
k


We can take this one step further to calculate explicitly the action of h⊥ ⊥ ⊥
k , ek and pk on hλ , eλ ,
and pλ . It is sufficient to give an expression for the expressions p⊥ ⊥ ⊥ ⊥
k (hλ ), pk (pλ ), hk (hλ ), hk (eλ ),
⊥
and hk (pλ ) since the others can be found by applying ω to both sides of the equation. We will
Pk
need to use the relation that h⊥ k (f g) =
⊥ ⊥ ⊥ ⊥ ⊥
i=0 hi (f )hk−i (g) and pk (f g) = pk (f )g + f pk (g).
The real difficulty in this problem is in finding a nice way of elegantly expressing these
quantities.
Proposition 3.7.
`(λ)
X
(3.43) p⊥
k (hλ ) = hλ (λi )](λi −k)
i=1

(3.44) p⊥
k (pλ ) = mk (λ)kpλ (k)

`(λ)
XY
(3.45) h⊥
k (hλ ) = hλi −αi
|α|=k i=1

where the sum over all sequences α such that 0 ≤ αi ≤ λi .

`(λ)
XY
(3.46) h⊥
k (eλ ) = eλi −αi
|α|=k i=1
34 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

where the sum over all sequences α such that 0 ≤ αi ≤ 1.

X Y
(3.47) h⊥
k (pλ ) = pλi
S⊂{1,2,...,`(λ)} i∈S
/

P
where the sum is over all subsets S of {1, 2, . . . , `(λ)} such that i∈S λi = k.

Proof. Most of these expressions follow from the previous proposition and the action
of h⊥ ⊥
k and pk on a product. Notice that

`(λ) `(λ)
X X
(3.48) p⊥
k (hλ ) = p⊥
k (hλi )hλ (λi ) = h(λi −k) hλ (λi )
i=1 i=1

which is the same as equation (3.43).

Also we have

`(λ)
X
p⊥
k (pλ ) = p⊥
k (pλi )pλ (λi )
i=1
`(λ)
X
(3.49) = δλi ,k kp(λi −k) pλ (λi )
i=1

and because there is exactly one non-zero term for each part of size k in λ so this is equal
to the expression in equation (3.44).

Similarly to compute h⊥
k (hλ ), we compute

k
X
(3.50) h⊥
k (hλ ) = h⊥ ⊥
i (hλ1 )hk−i (hλ (λ1 ) )
i=0

We can assume by induction on the length of λ that we know the formula for h⊥ k−i (hλ (λ1 ) )
is given by equation (3.45) (the base case is known because `(λ) = 1 is given in the previous
proposition). We have then that equation (3.50) is equal to
 
k `(λ)
X X Y
(3.51) = hλ1 −i  hλi −α̃i 
i=0 |α̃|=k−i i=2

and this is equivalent to equation (3.45) since if i > λ1 then hλ1 −i = 0 so the size of the first
part of α is restricted by both k and the size of λ1 .
 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS 35

Similarly, we can show again by assuming (3.46) is true for all partitions of length less than
`(λ) by induction, then
k `(λ)
X X Y
h⊥
k (eλ ) = h⊥
i (eλ1 ) eλi −α̃i
i=0 |α̃|=k−i i=2
`(λ) `(λ)
XY X Y
(3.52) = eλ1 eλi −α̃i + eλ1 −1 eλi −α̃i .
|α̃|=k i=2 |α̃|=k−1 i=2

This equation is then equivalent to (3.46) for an equation indexed by a partition equal to
the length of the partition λ.

Now to calculate h⊥ k (pλ ) we again assume by induction that (3.47) holds for partitions of
length less than `(λ).
(3.53) h⊥ ⊥
k (pλ ) = pλ1 hk (pλ (λ1 ) ) + h⊥
k−λ1 (pλ (λ1 ) )

which follows from the action of h⊥

k on pn given in the previous proposition. In this equation,
if k − λ1 < 0 then the second term in this sum is equal to 0. Using the inductive assumption,
(3.53) is equal to
X Y X Y
(3.54) = pλ1 pλi + pλi
S⊂{2,...,`(λ)} i∈S
/ T ⊂{2,...,`(λ)} i∈T
/
i6=1 i6=1
P
where the the first sum is over all subsets S such that i∈S / λi = k and the second sum
P i6=1
is over subsets T such that i∈T/ λi = k − λ1 . This is equivalent to equation (3.47) for a
i6=1
partition of length equal to `(λ). 

There is a third operation of multiplication which we have not yet mentioned which is a
type of composition of symmetric functions. We define pn [pm ] = pnm and then extend this
definition in a natural manner. That is we set,
`(λ)
Y
(3.55) pn [pλ ] = pnλi .
i=1

For c, d ∈ Q and f, g ∈ Λ this operation is linear by

(3.56) pn [cf + dg] = c pn [f ] + d pn [g].
P P Q`(λ)
In particular we have, pn [ λ cλ pλ ] = λ cλ i=1 pnλi . Then for f ∈ Λ and for a partition λ
we define
`(λ)
Y
(3.57) pλ [f ] = pλi [f ].
i=1
P
Finally for f = λ cλ pλ and g ∈ Λ, we define
X
(3.58) f [g] = cλ pλ [g].
λ
36 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

This definition implies that for c, d ∈ Q and f, g, h ∈ Λ, (c f + d g)[h] = c f [h] + d g[h] but
in general f [c g + d h] 6= c f [g] + d f [h] (note this will hold if f = pn ).

3.1. The class functions of the symmetric group

We have defined the algebra of ‘symmetric functions’ without much a hint as why we have
chosen this as the name of the algebra since the elements of Λ are neither symmetric nor
functions. One motivation for studying this algebra is that it is isomorphic to the space of
class functions of the symmetric group.

Let us consider Φn the linear vector space over Q of the class functions of the symmetric
group Symn . We know that Φn is a vector space spanned by the elements Cλ where λ is a
partition of n and

1 if π has cycle type λ
(3.59) Cλ (π) = .
0 otherwise
We also know that Φn is spanned by the set of irreducible characters of Symn .

We will define the Frobenius map between the space of class functions Φn and the space of
symmetric functions of degree n. That is we define F : Φn → Λn by the action on the basis
Cλ as
pλ
(3.60) F(Cλ ) = .
zλ

This map is clearly an isomorphism since the sets {Cλ }λ`n and {pλ /zλ }λ`n are both bases of
Φn and Λn respectively.

Let χtrivP
n
represent the trivial character on the symmetric group Symn . This means that
trivn
χ = λ`n Cλ and therefore we have
X X pλ
F(χtrivn ) = F(Cλ ) = = hn .
λ`n λ`n
zλ
sgnn
As well we may denote the sign character on Symn by χ P . Since the sign of a permutation
|λ|−`(λ) sgnn |λ|−`(λ)
with cycle type λ is (−1) we have that χ = λ`n (−1) Cλ and therefore
X X pλ
F(χsgnn ) = (−1)|λ|−`(λ) F(Cλ ) = (−1)|λ|−`(λ) = en .
λ`n λ`n
zλ

The homogeneous and elementary generators are natural elements to consider in this context
since the trivial and sign characters are the two one dimensional characters of Symn .

The definition of the scalar product on the symmetric functions may have seemed somewhat
arbitary when we introduced it for symmetric fucntions but it is actually motivated by the
scalar product of class functions and the connection by the following proposition.
 3.1. THE CLASS FUNCTIONS OF THE SYMMETRIC GROUP 37

Proposition 3.8. If χ, ψ ∈ Φn , then

(3.61) hF(χ), F(ψ)i = hχ, ψi
where on the left the scalar product is over symmetric functions with hpλ , pµP i = δλµ zλ and on
the right it is the scalar product on the class functions defined as hχ, ψi = n!1 σ∈Symn χ(σ)ψ(σ −1 ).

Proof. Because the map F is linear and the both the scalar products are bilinear it
suffices to show that this result holds for a basis. That is, we need only show that
(3.62) hF(Cλ ), F(Cµ )i = hCλ , Cµ i
P P
since then we know that for χ = λ`n cλ Cλ and ψ = µ`n dµ Cµ and
X X
hF(χ), F(ψ)i = cλ dµ hF(Cλ ), F(Cµ )i = cλ dµ hCλ , Cµ i = hχ, ψi .
λ,µ`n λ,µ`n

Now since F(Cλ ) = pλ /zλ it is easy to establish (3.62) for a fixed λ and µ.
hF(Cλ ), F(Cµ )i = hpλ /zλ , pµ /zµ i = δλµ /zλ ,
while at the same time
1 X
hCλ , Cµ i = Cλ (σ)Cµ (σ −1 ) = δλµ /zλ .
n! σ∈Sym
n

Remember that the irreducible characters of the symmetric group are an orthonormal basis
of class functions. The images of the irreducible characters will be the fundamental basis for
the symmetric functions and we will introduce this basis in a later chapter.

The space of symmetric functions Λ is an algebra since it is a vector space endowed with a
multiplication operation which takes elements in Λn × Λm and sends them to Λn+mL . The set
of class functions is endowed with a similar multiplication operation. We set Φ = n≥0 Φn .

Recall that for χ ∈ Φn and ψ ∈ Φm , we have χ ⊗ ψ is a class function of Symm × Symn

defined by χ ⊗ ψ(π, σ) = χ(π)ψ(σ). To make this a character of Symn+m we consider the
Sym
induced character χ ⊗ ψ ↑Symn+m n ×Symm
∈ Φn+m . This is our analogous operation in the space
of class functions to the operation of multiplication in the symmetric functions. In fact, this
operation is more than just analogous, it satisfies the following property:
Proposition 3.9. For χ ∈ Φn and ψ ∈ Φm ,
Sym
(3.63) F(χ ⊗ ψ ↑Symn+m
n ×Symm
) = F(χ)F(ψ)

Proof. The operation of inducing P two class functions isPlinear in both the first and in
the second position as we have for χ = λ`n cλ Cλ and ψ = µ`m dµ Cµ then
Sym Sym
XX
(3.64) χ ⊗ ψ ↑Symn+m
n ×Symm
= cλ dµ Cλ ⊗ Cµ ↑Symnn+m
×Symm .
λ`n µ`m
38 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

Sym
Therefore we need only show that F(Cλ ⊗ Cµ ↑Symnn+m
×Symm ) = F(Cλ )F(Cµ ). To do this we
Sym
will expand F(Cλ ⊗ Cµ ↑Symn+m n ×Symm
) in the basis {pν }ν`n+m .
D E D E
Symn+m Symn+m
Cλ ⊗ Cµ ↑Symn ×Symm , Cν = Cλ ⊗ Cµ , Cν ↓Symn ×Symm
1 X Sym
= Cλ (σ)Cµ (τ )Cν ↓Symnn+m
×Symm (σ, τ ).
n!m! σ∈Sym ,τ ∈Sym
n m

Every term in this last sum is equal to 0 unless λ ] µ = ν and only then when σ is of cycle
δ
type λ and τ is of cycle type µ. Therefore the right hand side is equal to zλ]µ,ν
λ zµ
and hence
D E δ
Symn+m λ]µ,ν
F(Cλ ⊗ Cµ ↑Symn ×Symm ), F(Cν ) =
zλ zµ
and hence
Sym pλ]µ pλ pµ
F(Cλ ⊗ Cµ ↑Symn+m
n ×Symm
)= = = F(Cλ )F(Cµ ).
zλ zµ zλ zµ


This last proposition gives us an interpretation for hλ and eλ since we have already noted that
the image of the trivial and sign characters in the Frobenius map are hn and en respectively.
Define Symλ to be the subgroup of Symn isomorphic to Symλ1 × Symλ2 × · · · × Symλ`(λ) in
the natural manner. Denote the trivial and sign characters on this subgroup as χtrivλ and
χsgnλ so that for all π ∈ Symλ , χtrivλ (π) = 1 and χsgnλ (π) = sgn(π) and more precisely
triv sgn
χtrivλ = χtrivλ1 ⊗ χtrivλ2 ⊗ · · · ⊗ χ λ`(λ) and similarly χsgnλ = χsgnλ1 ⊗ χsgnλ2 ⊗ · · · ⊗ χ λ`(λ) .
From the previous proposition we have
Sym
hλ = F(χtrivλ ↑Symnλ )
and
Sym
eλ = F(χsgnλ ↑Symnλ ).

We defined a second type of multiplication on symmetric functions which we called the inner
or Kronecker product of symmetric functions. The definition this operation ∗ is given by
pλ
zλ
∗ pzµµ = δλµ pzλλ . It arises naturally in the following sense.

Proposition 3.10. For χ, ψ ∈ P hin and we define χ · ψ as the class function χ · ψ(g) :=
χ(g)ψ(g). This is the inner tensor product of characters. We have
(3.65) F(χ · ψ) = F(χ) ∗ F(ψ).

Proof. Again it suffices to verify this identity on a basis for the class functions because
it will hold for any linear combination of the class functions as well. This is easy to verify
for the class functions Cλ , since

0 if λ 6= µ
Cλ · Cµ (π) = = δλµ Cλ (π)
1 if λ = µ and Cλ (π) = 1
 3.1. THE CLASS FUNCTIONS OF THE SYMMETRIC GROUP 39

This means that Cλ · Cµ = δλµ Cλ and so we have

F(Cλ · Cµ ) = δλµ F(Cλ )
pλ
= δλµ
zλ
pλ pµ
= ∗
zλ zµ
= δλµ F(Cλ ) ∗ F(Cµ )


***** interpretation of ω here?

The coproduct operation also has an interpretation in the algebra of class functions, however
we first need to extend our definition of the Frobenius map to the algebra of class functions
on Symk × Symn−k . Recall that we have for class functions χ ∈ Φk and ψ ∈ Φn−k that the
function χ ⊗ ψ defined to be χ ⊗ ψ(π, σ) := χ(π)ψ(σ) for π ∈ Symk and σ ∈ Symn−k and
χ ⊗ ψ is a class function of Symk × Symn−k . This is called the the outer tensor product
of class functions. Note that for any basis of class functions of G, {C (i) }, and of the class
functions of H, {D(i) }, then {C (i) ⊗ D(j) } is basis for the class functions of G × H.

To extend the definition of the Frobenius map to include class functions of Symk × Symn−k
which has as a basis {Cλ ⊗ Cµ } λ`k we set
µ`n−k
pλ pµ
(3.66) F(Cλ ⊗ Cµ ) := ⊗ = F(Cλ ) ⊗ F(Cµ )
zλ zµ
and this definition is extended linearly. This implies that we have more generally, for χ ∈ Φn
and ψ ∈ Φm ,
(3.67) F(χ ⊗ ψ) = F(χ) ⊗ F(ψ).
Using this extension of notation we have the following interpretation of the coproduct oper-
ation on symmetric functions.
Proposition 3.11.
n
X Sym
(3.68) ∆(F(χ)) = F(χ ↓Symnk ×Symn−k )
k=0

Proof. It suffices to show that this result again holds on a basis and the natural basis
to consider is again Cλ for λ ` n. We have that

 1 if Cµ (π) = 1 and Cν (σ) = 0 for
Symn
Cλ ↓Symk ×Symn−k (π, σ) = µ ` k, ν ` n − k with µ ] ν = λ
0 otherwise


In other words we see that

Sym
X
Cλ ↓Symnk ×Symn−k = δµ]ν,λ Cµ ⊗ Cν
µ`k
ν`n−k
40 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

This implies that

Sym
X 1
F(Cλ ↓Symnk ×Symn−k ) = δµ]ν,λ pµ ⊗ pν .
µ`k
zµ zν
ν`n−k

As well we have
n
X X zλ
∆(zλ F(Cλ )) = ∆(pλ ) = δµ]ν,λ pµ ⊗ pν
k=0 µ`k
zµ zν
ν`n−k
n
X Sym
= zλ F(Cλ ↓Symnk ×Symn−k ),
k=0

and therefore the proposition holds for all class functions. 

3.2. Exercises

(1) (a) Expand p(2,2) in the elementary and homogeneous bases.

(b) Expand e(2,2) in the power basis.
(c) Expand h(2,2) in the homogeneous basis.
(2) Calculate the following scalar products
(a) h(2,2,1) , p(3,2)
(b) h(3,2) , p(3,2)
(c) h(3,2) , p(2,2,1)
(d) h(3,2) , h(4,1)
(e) h(3,2) , h(3,1,1)
(f) h(3,2) , h(2,2,1)
(3) Calculate the following inner products using the formulas given in this section.
Assume that |λ| = n.
(a) hhn , pλ i
(b) hen , pλ i
(c) hpn , hλ i
(d) hp1n , hλ i
(e) hpλ , hλ i
(f) hhn , hn i
(g) hen , hn i
(h) hhn , hλ i
(i) hen , hλ i
(4) Show that ∆ ◦ ω = (ω ⊗ ω) ◦ ∆Pby showing that it holds true on the power basis.
Use this to show that ∆(en ) = nk=0 ek ⊗ en−k .
(5) Show that (1 ⊗ ω) ◦ ∆0 = (ω ⊗ 1) ◦ ∆0 = ∆0 ◦ ω.
(6) Show the following determinental formulas for hn .
 3.2. EXERCISES 41

(a)

p1 −1 0 ··· 0
p2 p1
2 2
−1 ··· 0
hm = .. .. .. ..
. . . ··· .
pm−1 pm−2 p1
m−1 m−1
··· m−1
−1
pm pm−1 pm−2 p1
m m m
··· m

(b)

e1 e2 e3 · · · en
1 e1 e2 · · · en−1
hn = 0 1 e1 · · · en−2
.. .. .. ..
. . . ··· .
0 0 ··· 1 e1

State and prove the corresponding determinental formulas for en in terms of pn and
hn .
(7) Show that ∆(H(t)) = H(t) ⊗ H(t) and ∆(E(t)) = E(t) ⊗ E(t) and ∆(P (t)) =
P (t) ⊗ 1 + 1 ⊗ P (t).
(8) Show that p⊥ k = k ∂pk
∂

(9) Use the fact that E(t)H(t) = 1 to develop a formula for hn in terms of the elementary
basis.
Use the relationship P (t) = log(1 + n≥1 hn tn ) to derive a formula for pn in terms
P
(10)
of the homogeneous basis.
(11) Use the previous result to expand pλ in the homogeneous basis.
(12) Prove that Λ endowed with the bialgebra with product µ and coproduct ∆0 does
not have a corresponding antipode and hence is not a Hopf algebra.
(13) Prove that for any f ∈ Λn , hn ∗ f = f and en ∗ f = ωf .
(14) Prove that if g ∈ Λn has the property that for all f ∈ Λn , g ∗ (g ∗ f )) = f then
hg, pλ i = ±1 for all λ partitions of n.
(15) Show that Z[p1 , p2 , p3 , . . .] ⊆ Z[e1 , e2 , e3 , . . .] and that the converse of this statment
(i.e. that Z[e1 , e2 , e3 , . . .] ⊆ Z[p1 , p2 , p3 , . . .]) is not true.
(16) Show that Z[h1 , h2 , . . . , hk ] = Z[e1 , e2 , . . . , ek ].
(17) (a) Show that the linear span of the symmetric functions {f + ω(f ) : f ∈ Λ} forms
a subalgebra of the symmetric functions under the standard product.
(b) Show that this algebra is not a bialgbra with the coproduct ∆.
(c) Show that a linear basis for this space is given by the set {pλ : |λ|−`(λ) mod 2 =
0}
(d) Show that the space is closed under the Kronecker product of equation (3.18)
and the coproduct ∆0 of equation (3.19).
(18) Show that Q[p1 , p3 , p5 , . . .] is a Hopf subalgebra of the symmetric functions and that
it is also closed under the Kronecker product and coproduct ∆0 . This subalgebra is
sometimes known as the Q-function algebra.
42 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

3.3. Solutions to exercises

(1) (a)

2
p22 = −2 e2 + e1 2 = e1 4 − 4 e2 e1 2 + 4 e2 2

2
= −2 h2 + h2 2 = h1 4 − 4 h2 h1 2 + 4 h2 2
(b)

2
e22 = 1/4 −p2 + p1 2 = 1/4 p1 4 − 1/2 p2 p1 2 + 1/4 p2 2
(c)

2
h22 = −e2 + e1 2 = e1 4 − 2 e2 e1 2 + e2 2
(2) By direct calculation
h(3,2) = 1/12 p1 5 + 1/3 p2 p1 3 + 1/6 p3 p1 2 + 1/4 p2 2 p1 + 1/6 p3 p2
h(2,2,1) = 1/4 p1 5 + 1/2 p2 p1 3 + 1/4 p2 2 p1
h(4,1) = 1/24 p1 5 + 1/4 p2 p1 3 + 1/3 p3 p1 2 + 1/8 p2 2 p1 + 1/4 p4 p1
h(3,1,1) = 1/6 p1 5 + 1/2 p2 p1 3 + 1/3 p3 p1 2
(a) h(2,2,1) , p(3,2) = 0
(b) h(3,2) , p(3,2) = 1
(c) h(3,2) , p(2,2,1) = 2
(d) h(3,2) , h(4,1) = 3
(e) h(3,2) , h(3,1,1) = 4
(f) DhP (3,2) , h(2,2,1) =E5

(3) (a) pµ /zµ , pλ = 1

DPµ`n E
|µ|−`(µ)
(b) µ`n (−1) pµ /zµ , pλ = (−1)|λ|−`(λ)
(c) hpn , hλ i = δλ,(n)
(d) hp1n , hλ i = z1n / `(λ) n
Q

i=1 z1λi = λ1 ,λ2 ,...,λ`(λ)
Q
(e) hpλ , hλ i = Di mi (λ)! E
P P P
(f) hhn , hn i = p
µ`n µ µ /z , p /z
λ`n λ λ = λ`n 1/zλ which is equal to 1 since
P
n!/zλ = thePnumber of permuations with cycle type λ and λ`n n!/zλ = n!.
(g) hen , hn i = Pλ`n (−1)|λ|−`(λ) /zλ = 0 if n > 1 and is equal to 1 if n = 1. This fol-
lows since λ`n (−1)|λ|−`(λ) n!/zλ = the number of permutations of even length
− the number of permutations of odd length. The numbers of these subsets of
these permutations must be equal because composition with a permutation of
length 1 is an involution which interchanges these sets.
(h) hhn , hλ i
(4) For any partition k > 0, ∆◦ω(pk ) = (−1)k−1 ∆(pk ) = (ω ⊗ω)◦∆(pk ). Since ω and ∆
are both ring homomorphisms, this formula holdsPon any f ∈ Λ. Using P this identity,
∆(en ) = ∆ ◦ ω(hn ) = (ω ⊗ ω) ◦ ∆(hn ) = (ω ⊗ ω)( nk=0 hk ⊗ hn−k ) = nk=0 ek ⊗ en−k .
(5) Again we need only show that this property holds on a basis to conclude that it holds
for all symmetric functions ∆0 ◦ ω(pλ ) = ∆0 ((−1)|λ|−`(λ) pλ ) = (−1)|λ|−`(λ) (pλ ⊗ pλ ) =
(ω ⊗ 1)(pλ ⊗ pλ ) = (ω ⊗ 1) ◦ ∆0 (pλ ) Similarly we have ∆0 ◦ ω(pλ ) = (1 ⊗ ω) ◦ ∆0 (pλ ).
 3.3. SOLUTIONS TO EXERCISES 43

(6) Let Mn = [ai,j ] be the n × n matrix with ai,j = pn−i−j+2 if n − i − j + 2 > 0

n
and ai,j = i − 1 if n − i − j + 2 = 0 and ai,j = 0 otherwise. Show (−1) n!
detMn
satisfies the relation of equation (3.13) by expanding the determinant about the
(1,k)
first row. Notice that the (1, k) minor Mn (the minor formed by deleting the 1st
row and k column of Mn ) has determinant equal to (−1)n−1 (n − 1)! if k = 1 and
th
n
(−1)n−k (n − 1)k detMk−1 for 2 ≤ k ≤ n. Therefore, n (−1) n!
Mn = p n p n
(7)
! !
X X
H(t) ⊗ H(t) = hr tr ⊗ hm tm
r≥0 m≥0
X n
X
= tn hn−k ⊗ hk
n≥0 k=0
X
= tn ∆(hn ) = ∆(H(t))
n≥0

Apply the result of problem 4 to show as well that ∆(E(t)) = E(t) ⊗ E(t).
X pr X  pr pr  r
r
∆(P (t)) = ∆( )t = ⊗1+1⊗ t
r≥1
r r≥1
r r
= P (t) ⊗ 1 + 1 ⊗ P (t)
(8) p⊥
k (pλ ) = mk (λ)kpλ (k) from equation (3.44). Notice that
∂
∂pk
(pλ ) = mk (λ)pλ (k) , so
p⊥ ∂ ⊥ ∂
k (pλ ) = k ∂pk (pλ ) and pk (f ) = k ∂pk (f ).
(9)
!`
X X
` k
H(t) = 1 + (−1) ek (−t)
`≥1 k≥1
n
Now the coefficient of t on both sides of this equation will be hn on the left, and
on the right a term appears for every partition with length ` and a coefficient equal
to −1 raised to the size of the partition times a multinomial coefficient.
 
X
n−`(λ) `(λ)
hn = (−1) eλ
m 1 (λ) m2 (λ) · · ·
λ`n
X `(λ)!
= (−1)n−`(λ) Q eλ
λ`n i≥1 mi (λ)!

(10)
!`
X (−1)`−1 X
P (t) = hm tm
`≥1
` m≥1
The coefficient of t on the left hand side of this equation is pnn and on the right
n

hand side for each partition there is a term with a multinomial coefficient which
depends on the length and a factor of (−1)`−1 over the length of the partition.
X (`(λ) − 1)!
pn = n (−1)`(λ)−1 Q hλ
λ`n i≥1 m i (λ)!
44 3. THE ALGEBRA STRUCTURE OF THE RING OF SYMMETRIC FUNCTIONS

(11) With the formula from the previous problem it is a matter of finding a good way of
expressing the product pλ . The coefficient of hµ in pλ will be positive if `(µ) − `(λ)
even and negative otherwise. The coefficient of hλ in pµ is
`(µ)
`(λ)−`(µ)
X Y µi (`(ν (i) ) − 1)!
(−1) Q (i)
ν (1) ]ν (2) ]···]ν (`(µ)) =λ i=1 j≥1 mj (ν )!

where the sum is over all sequences of partitions with ν (i) ` λi .

(12) In order for the bialgebra structure to have a Hopf algebra structure it must hold
that m ◦ (id ⊗ S 0 ) ◦ ∆0 = u ◦ ε0 where m(f ⊗ g) = f ∗ g. Act by this expression on
p1 and we must have that
p1 S 0 (p1 ) = 1
This cannot happen unless S 0 (p1 ) = 1/p1 which Pis not in
 our algebra.
pµ
(13) For λ ` n we clearly have that hn ∗ pλ = µ`n zµ ∗ pλ = pλ and en ∗ pλ =
P 
n−`(µ) pµ
µ`n (−1) zµ
∗ pλ = (−1)n−`(λ) pλ = ω(pλ ). Therefore by linearity it holds
that hn ∗ f = fP and en ∗ f = ω(f ).
(14) Note that g = µ cµ pµ . Therefore, g ∗ pλ = zλ cλ pλ and g ∗ (g ∗ pλ ) = zλ2 c2λ pλ = pλ ,
and so we know that zλ2 c2λ = 1 or zλ cλ = ±1. We also have hg, pλ i = cλ zλ = ±1.
(15) Since each pk ∈ Z[e1 , e2 , e3 , . . .] from equation (3.16) orPproblem number 10, we
know that pλ ∈ Z[e1 , e2 , e3 , . . .]. It follows that each f = λ cλ pλ with each cλ ∈ Z,
then f ∈ Z[e1 , e2 , e3 , . . .]. Many counterexamples to the converse exists (e.g. e2 =
p2 /2 + p21 /2).
(16) From equation (3.12) or problem number 6b or 9 we know that hk ∈ Z[e1 , e2 , . . . , ek ]
and with an application of ω on these equations we know equally that ek ∈ Z[h1 , h2 , . . . , hk ].
(17) (a) {f + ω(f ) : f ∈ Λ} are the set of functions which are invariant under the
involution ω. That property is clearly invariant under products since ω is a
ring homomorphism.
(b) This is not invariant under the coproduct ∆, since for instance ∆(e(2,2) + h(2,2) )
in the degree (2, 2) tensor is not invariant under ω.
(c) Note that pλ + ω(pλ ) = 2pλ if |λ| + `(λ) is even and it is equal to 0 if |λ| + `(λ)
is odd. Since pλ is a linear basis of Λ, {pλ : |λ| − `(λ) mod 2} is a linear basis
for {f + ω(f ) : f ∈ Λ}.
(d) If λ has |λ| − `(λ) even, then pλ ∗ pµ = 0 or is a multiple of pλ and hence the
basis elements are closed. ∆ also sends the basis elements pλ to pλ ⊗ pλ .
(18) The fact that this ring is a Hopf subalgebra and closed under the Kronecker product
and coproduct is true for any subalgebra generated by a subcollection of the the
pk and this need only be checked by verifying these operations are closed on the pλ
basis where the parts of λ are taken from the pk which generate the subalgebra.
 Bibliography

[1] G. Andrews, Euler’s Pentagonal Number Theorem, Mathematics Magazine, 56 no. 5 (1983), 279–284.
[2] G. Chrystal, Algebra, Part II, 2nd ed., Black, London, 1931.
[3] O. Egecioglu and J. B. Remmel: ”Brick tabloids and the connection matrices between bases of symmetric
functions,” Discrete Applied Math, 34 (1991), 107-120.
[4] L. Euler, Evolutio producti infiniti (1 − x)(1 − xx)(1 − x3 )(1 − x4 )(1 − x5 )(1 − x6 ) etc., Opera Omnia,
(1) 3, 472–479.
[5] F. Franklin, Sur le développement du produit infini (1 − x)(1 − x2 )(1 − x3 ) · · · , Comptes Rendus, 82
(1881).
[6] E. Grosswald, Topics in the Theory of Numbers, Macmillan, New York, 1966; Birkhäuser, Boston, 1983.
[7] H. Gupta, Selected Topics in Number Theory, Abacus Press, Tunbridge Wells, 1980.
[8] G. H. Hardy and E. M. Wright, An Introduction to the Theory of Numbers, 4th ed., Oxford University
Press, London 1960.
[9] I. Niven and H. Zuckerman, An Introduction to the Theory of Numbers, 3rd ed., Wiley, New York,
1973.
[10] H. Rademacher, Topics in Analytic Number Theory, Grundlehren series, vol 169, Springer, New York,
1973.

107