Symmetric Functions

Symmetric Functions and Rectangular
Catalan Combinatorics
F. Bergeron
September 7, 2019
2
Contents
Introduction 5
1 Combinatorial Background 7
1.1 Ferrers diagrams and partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 Young tableaux, hook length formula, and Kostka numbers . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 Robinson-Schensted-Knuth . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 Charge and cocharge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.5 Exercises and problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2 “Classical” Symmetric functions 19

2.1 Basic notions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 From symmetric polynomials to symmetric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.3 Schur functions, a combinatorial approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.4 Dual basis and Cauchy kernel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.5 Product, Kronecker product, and Schur positivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.6 Transition matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.7 Jacobi-Trudi determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8 Plethysm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.9 Some ties with geometric complexity theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.10 Counting with symmetric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3 Macdonald symmetric functions and operators 39

3.1 Macdonald symmetric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2 Macdonald eigenoperators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4 Rectangular combinatorics, and symmetric functions 47

4.1 Rectangular Catalan combinatorics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.2 Counting (m, n)-Dyck paths and (m, n)-parking functions . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.3 Elliptic Hall algebra operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.4 Construction of the operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.5 Interesting seeds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.6 What comes next . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
Bibliography 62
3
4
Introduction
These notes are intended as a complement for my course in the AEC 4th Algorithmic and Enumerative
Combinatorics Summer School, July 30 – August 3, 2018; RISC, Hagenberg, Austria. For sure their
content is certainly more extensive than what I will realistically be able to cover in the five hours of
the course, for which I will assume very little background knowledge.
I propose many exercises and problems in these notes; but, once again, there are many more of these
than what can be solved in the two half-hour periods allocated. Furthermore, some problems are
hard, assuming that one only knows the background in these notes. My aim is for the participants
to have something to take home and think about. Beside the exercises that one typically considers
in order to become familiar with the notions introduced, I have hence included some problems that
may lead to small research projects for the participants that may wish to undertake them. I have
also included a wide enough bibliography so that the interested reader is oriented in finding the
necessary background, and what is currently happening in the subject.
Much of the notions considered in these notes are interesting to explore using computer algebra
tools1 , and I give indications of where one may start to learn about these. In particular, there are two
special tutorials that were prepared2 by graduate students of UQAM: Pauline Hubert and Mélodie
Lapointe, on the occasion of a School/Workshop held at CRM in 2017 on Equivariant Combinatorics.
Other computer algebra tools are in need of being formatted for general consumption (chapter 4),
and this might be one of the outcomes of the school.
Please be lenient with me, since I wrote all of this a bit at the last minute (recycling pieces from
other texts of mine). In other words, feel free to underline any typos or mistakes.
1
In fact, this is a great way to familiarize oneself with the subject.
2
See opening paragraphs of chapters 1, 2, and 3.
5
6
Chapter 1
Combinatorial Background
All of this chapter is pretty much classical. For more, see one of the monographs: FB. [3], Fulton [20],
or Stanley [43]. Bruce Sagan also has a nice online text:
http://users.math.msu.edu/users/sagan/Papers/Old/uyt.pdf
As a fast startup, the web site https://en.wikipedia.org/wiki/Young_tableau is not too bad. Also,
see in the Notices of the AMS of February 2007: What is . . . a Young Tableau? by Alexander Young
(not related). For a fast introduction to partitions in Sage, a nice Sage Partition Tutorial is available1
(It is in part better than the original for our purpose). It was prepared by Pauline Hubert and
Mélodie Lapointe (UQAM).
1.1 Ferrers diagrams and partitions
Considering the component-wise partial order on N × N, a Ferrers diagram µ (with n cells), is a

cardinal n finite subset of N × N (with pointwise partial order) such that
(i, j) ≤ (k, l) and (k, l) ∈ µ implies (i, j) ∈ µ.
Clearly a Ferrers diagram is characterized by the decreasing integer sequence (µ1 , µ2 , . . . , µk ) with
µi denoting the number of cells in the ith row of µ (reading these row lengths from the bottom to
the top). Elements of µ are called cells. The conjugate of a Ferrers diagram µ, denoted by µ0 , is
the set
µ0 := {(j, i) | (i, j) ∈ µ}.
Thus, the rows of µ0 are the columns of µ, and vice-versa.
The hook of a cell c = (i, j) of a Ferrers diagram µ, is the set of cells of µ that either lie in the
same row, to the right of c, or lie in the same column, and above c. Moreover, c itself belongs to its
1
https://more-sagemath-tutorials.readthedocs.io/en/latest/tutorial-integer-partitions.html
7
hook. The corresponding hook length is h(c) = hij , is the number of the cells in question. It may
be calculated as follows
hij := µj+1 + µ0i+1 − i − j − 1. (1.1.1)
The number of cells to the right of c (resp. above) on its row (resp. column) is also called the arm
of c (resp. the leg) in µ, denoted by a(c) = aµ (c) (resp. `(c) = `µ (c)). Thus,
h(c) = a(c) + `(c) + 1.
These notions are illustrated in figure 1.1, with the hook of the red cell highlighted in yellow and
green. The green cells is the arm of c, and the yellow ones correspond to its lef. Here, c = (3, 1),
Figure 1.1: A cell and its hook (with arm and leg).
and its hook length is h21 = 8. The row lengths of µ are (9, 8, 8, 8, 6, 3, 2, 1), whereas those of µ0 are
(8, 7, 6, 5, 5, 5, 4, 4, 1). The south-west most cell is (0, 0).
We are here following the (Cartesian style) right side up “French” convention, rather than the
(matrix-style) upside down “English” convention, for drawing diagrams. The readers who would like
to follow “Macdonald’s advice” about this should read these notes upside down in a mirror.
To a n-cell Ferrers diagram there corresponds a partition2 of the integer n. This is simply the
decreasing ordered sequence (µ1 , µ2 , . . . , µk ) of row lengths of the diagram. Typically, there is no
confusion about also denoting by µ the partition associated to a Ferrers diagram µ. Each µi is
said to be a part of µ, and |µ| := µ1 + µ2 + . . . + µk = n. We write µ ` n to indicate that µ is a
partition of n. The length, `(µ), of µ is simply the number of (non-zero) parts of µ; hence we have
`(µ0 ) = µ1 .
Partitions are often presented as words µ = µ1 µ2 · · · µk , when all parts are less or equal to 9. One
also considers the empty partition, denoting it 0. The partition sets Part(n) := {µ | µ ` n}, for
small n, are respectively
Part(0) = {0}
Part(1) = {1}
Part(2) = {2, 11}
Part(3) = {3, 21, 111}
Part(4) = {4, 31, 22, 211, 1111}
Part(5) = {5, 41, 32, 311, 221, 2111, 11111}
Part(6) = {6, 51, 42, 411, 33, 321, 3111, 222, 2211, 21111, 111111}
2
See https://en.wikipedia.org/wiki/Partition_(number_theory) for more on partitions.
8
Another description of partitions consists in writing µ = 1d1 2d2 · · · j dj , where di is the number of
parts of size i in µ, and a useful number in the sequel is
zµ := 1d1 d1 !2d2 d2 ! · · · j dj dj !
A corner of µ, is any cell of the form c = (µi − 1, i − 1) for which if µi > µi+1 . Corners are exactly
the cells that can be removed from the associated Ferrers diagram so that the resulting diagram is
also a Ferrers diagram. For example, the corners of the partition µ = 4 4 2 1 are the three dark blue
cells in figure 1.2.
Figure 1.2: Corners of 4421.
If ν is obtained from µ by removing one of its corners, we write ν → µ. The transitive closure
of this (covering) relation is a partial order called the Young poset (see figure 1.3). This simply
corresponds to set-inclusion of the corresponding Ferrers diagrams. Since the characteristic property
of Ferrers diagram is compatible with union and intersection, the Young poset has the structure of a
lattice (as a sub-lattice of the lattice of finite subsets of N × N). Hence, for two Ferrers diagrams (or
partitions) we have ν → µ if and only if ν ⊂ µ, with ν ` n and µ ` n + 1.
For α ⊆ β, the interval [α, β] is the set {µ | α ⊆ µ ⊆ β}. An interesting special case is the interval
[0, nk ] of partitions contained in the rectangular partition nk . The number of such partitions is
n+k

easily seen to be equal to the binomial coefficient k . This can be refined to give the classical
q-analog of the binomial coefficient:

X
|µ| n+k
q = . (1.1.2)
k
k q
µ⊆n
In general, it is not easy to describe the polynomial

X
Pµ (q) := q |ν| (1.1.3)
ν⊆µ
for a given partition µ. In Sage (using x as a variable so that nothing has to be declared) the
polynomial in question may be calculated as:
Pµ (x)=“add(xk ∗Partitions(k, outer=µ).cardinality() for k in range(n + 1))”.
When ν is contained in µ, we consider the skew partition, denoted by µ/ν, having as skew Ferrers
diagram the set difference µ \ ν.
Dominance order on partitions
For two partitions µ and λ of n, we say that µ dominates λ, and write λ µ, if and only if for all k
λ1 + λ2 + · · · + λk ≤ µ1 + µ2 + · · · + µk .
9
.. .
.. .. . . . ..
.. . . . .
. ..
I
@
@ @
I
@
@ @
@
I
@
@
↑
0
Figure 1.3: Young’s lattice.
If needed one adds parts µi = 0 or λi = 0 so that inequalities make sense. The following figure gives
the dominance order on partitions of n = 6, with an arrow µ → λ indicating that µ is covered by λ.
% & % &
→ → → →
& % & %
This example underlines that the dominance order is not a total order (although it is for all n ≤ 5).
1.2 Young tableaux, hook length formula, and Kostka numbers
Let d be a finite subset of N × N. A tableau of shape d, with value in a set A (typically

A = {1, 2, . . . , n}), is simply a function τ : d → A. This function is “displayed” by filling each cell c
of d by its value τ (c), as illustrated in figure 1.4. We mostly consider the case when d is a Ferrers
diagram µ (or a skew diagram).
6
4
2 3
3 8
1 1 5-
Figure 1.4: A semi-standard tableau.
A tableau τ is said to be semi-standard if its entries are non-decreasing along rows (from left to
10
right), and strictly increasing along columns (from bottom to top) of d:
i<k implies τ (i, j) ≤ τ (k, j), and j < ` implies τ (i, j) < τ (i, `);
for (i, j), (k, j), and (i, `) lying in d.

A n-cell tableau τ is standard if it is a bijective semi-standard tableau with values in {1, 2, . . . , n}.
Thus, for τ to be standard we need τ (i, j) < τ (k, `) whenever (i, j) < (k, `) coordinate-wise. figure 1.5
gives an example of a standard tableau of shape 431.
6
7
2 5 8
1 3 4 6-
Figure 1.5: A standard tableau.
The reading word, ρ(τ ), of a tableau τ is obtained as the row by row reading of the entries of τ .
Each row is read from left to right, and rows are read from top to bottom. Thus, the reading word
of the tableau in figure 1.5 is: 72581346. One may check that a partition shaped semi-standard
tableau is entirely characterized by its reading word.
A standard tableau of shape µ corresponds to a maximal chain
0 = µ(0) → µ(1) → . . . → µ(n) = µ,
where the partition µ(i) is made out of the cells taking values ≤ i. For instance, the standard tableau
in figure 1.5 correspond (bijectively) to the maximal chain of figure 1.6.
0→ → → → → → → →
Figure 1.6: A maximal chain in Young’s lattice
1.2.1 Hook length formula
The number f µ of standard tableaux of shape µ (µ ` n) is given by the Frame-Robinson-Thrall (see

[19]) hook length formula.
n!
fµ = Q , (1.2.1)
c∈µ h(c)
11
By a direct application of this formula, there are exactly 16 standard tableaux of shape 321. These
are:
3 3 4 4 4 4 5 5
2 5 2 6 2 5 2 6 3 5 3 6 2 4 2 6
1 4 6 1 4 5 1 3 6 1 3 5 1 2 6 1 2 5 1 3 6 1 3 4
5 5 5 6 6 6 6 6
3 4 3 6 4 6 2 4 2 5 3 4 3 5 4 5
1 2 6 1 2 4 1 2 3 1 3 5 1 3 4 1 2 5 1 2 4 1 2 3
It is easy to check that the sum of the hook lengths of a partition µ is given by the formula
X X
h(c) = n(µ) + n(µ0 ) + |µ|, where n(d) := j.
c∈µ (i,j)∈d
The classical Robinson-Schensted-Knuth correspondence (see next section) shows that

X
(f µ )2 = n!,
µ`n
µ
P
and its properties imply that µ`n f is the number of involutive permutations.
1.2.2 Kostka numbers
The content γ(τ ) of a tableau τ is the sequence γ(τ ) = (m1 , m2 , m3 , . . .) of multiplicities of each
entry i in the tableau τ . For example, the content of the semi-standard tableau
4 4
2 2 4 4
1 1 1 1 2
is γ(τ ) = (4, 3, 0, 4, 0, 0, . . .). If λ and µ are two partitions of n, we define the Kostka number Kλ,µ
to be the number of semi-standard tableaux of shape λ and content µ. For instance, the 4 possible
semi-standard tableaux having content 2211 and shape 321 are
4 3 4 3
2 3 2 4 2 2 2 2
1 1 2 1 1 2 1 1 3 1 1 4
The content of a semi-standard tableau is 1n if and only if the tableau is standard. Hence Kλ,1n = f λ .
We may show that Kλ,λ = 1, and that Kλ,µ 6= 0 forces µ λ, in dominance order (you are asked to
show this in one of the exercises).
The above observations give a somewhat more natural characterization of the dominance order.
Indeed, we have λ µ if and only if Kλ,µ 6= 0. Moreover we see that the matrix (Kλ,µ )λ,µ`n is upper
12
triangular when partitions are sorted in any decreasing order which is a linear extension (why?) of
the dominance order. This immediately implies that the matrix (Kλ,µ )λ,µ`n is easily invertible. For
n = 4, the Kostka matrix is
4 31 22 211 1111
 
1 1 1 1 1
 0 1 1 2 3 
  (1.2.2)
 0 0 1 1 2 
 
 0 0 0 1 3 
0 0 0 0 1
1.3 Robinson-Schensted-Knuth
Let A and B be two ordered sets, and consider lexicographic two row matrices:

b1 b2 . . . bn
w= such that either (bi < bi+1 ) or (bi = bi+1 and ai ≤ ai+1 ), (1.3.1)
a1 a2 . . . an
with the ai in A, and the bi in B. We denote by ε the empty case (i.e.: n = 0).
The RSK correspondence (see Knuth [32] for more on this) w ←→ (P, Q), associates bijectively a
pair (P (w), Q(w)) = (P, Q) of same shape semi-standard tableaux to each such lexicographic matrix.
The entries of P (w) are the ai ’s, and those of Q(w) are the bi ’s. The correspondence is recursively
defined as follows.
1. First, we start with setting ε ←→ (0, 0).
2. For a nonempty

b1 b2 . . . bk−1 b b1 b2 . . . bk−1
w= set v :=
a1 a2 . . . ak−1 a a1 a2 . . . ak−1
and recursively sets v ←→ (P 0 , Q0 ).
3. Then construct
P := (P 0 ← a)
by tableau insertion (see subsection below) of a into P 0 . The tableau Q is then obtained by
adding b to Q0 in the position corresponding to the unique cell by which the shape of P differs
from that of P 0 .
This may be displayed as follows

 
 
5 6
3 5 5 3 4 4 
6
3 3 5
, 3 4 4
 
, 2 2 2 ←
1 = 2 2 3
2 3 3  
2 2 2
  
1 1 1 2 3 3 1 1 1 1 2 2  
1 1 1 1 3 3 1 1 1 1 2 2
13
In the “classical” RSK correspondence, w takes the form

1 2 ··· n
a1 a2 · · · an
and it is then customary to identify w with “word” a1 a2 · · · an . Since the bi are all distinct, it follows
that the tableau Q is actually a standard tableau. When a1 a2 . . . an is a permutation of {1, 2, . . . , n},
the tableau P is also standard. This establishes a bijection between permutations in Sn and pairs of
standard tableaux of same shape. As a by-product, we see that
X
n! = (f µ )2 , (1.3.2)
µ`n
since (f µ )2 counts the number of pairs of standard tableaux of shape µ.

We consider the inverse w−1 of w:

−1 a1 a2 . . . ak
w = lex ,
b1 b2 . . . bk
with “lex” standing for the increasing lexicographic reordering. Thus with

1 1 1 2 2 3 3 4 4 4
w=
3 4 5 1 4 1 4 1 2 2
we have
−1 1 1 1 2 2 3 4 4 4 5
w =
2 3 4 4 4 1 1 2 3 1
One can show that
w ←→ (P, Q) iff w−1 ←→ (Q, P ) (1.3.3)
Tableau insertion
One constructs as follows the tableau τ 0 := (τ ← a) obtained by insertion of a ∈ A in a semi-

standard τ : µ → A.
• First, if τ = 0 is the empty tableau, then τ 0 := (0 ← a) is simply the tableau of shape 1 with
value a in its single cell.
• Otherwise, one inserts a in the first row of τ , and two cases are to be considered depending on
how a compares to the entries occurring in this first row of τ .
(a) If a is larger or equal to the largest entry of the first row, then a is simply appended to
the end of the first row of τ ;
(b) Otherwise, a replaces the leftmost entry x of the first row which is larger than a. One
says that a bumps x. In this case, x is recursively inserted by the same process in the
tableau consisting of the rows of τ from the second to the top. This results in a bump
path, consisting in the successive cells altered by the insertion process, which ends in a
new (corner) cell for the underlying shape of τ .
In the example above, the bump path corresponding to the insertion of 1 is highlighted in green in
the P -tableau, as well as the new corner in the Q-tableau.
14
1.4 Charge and cocharge
We say that a standard tableau τ has a reading descent at a cell (i, j) if τ (i, j) + 1 lies in a cell
that is to the left of (i, j). For example, the reading descents of the tableau on the left-hand side of
figure 1.7 correspond to the cells that contain the values 2, 4, 7 and 8.
5 9 2 4
3 4 8 7−→ 1 1 3
1 2 6 7 0 0 2 2
Figure 1.7: Minimization of a tableau.
The minimization m(τ ), of a µ-shaped semi-standard tableau τ , is the “minimal” standard tableau
of shape µ which mimics the reading descent pattern of τ . It is obtained by successively reading off
the entries of τ from 1 to n, and filling the corresponding cells of µ with entries that stay constant
as long as one goes eastward, but rise up by 1 otherwise. This process may readily be extended to
semi-standard tableaux, reading equal entries from left to right.
The co-charge coch(τ ) of a standard tableau τ is simply the sum of the entries of m(τ ). The charge
ch(τ ) of a semi-standard tableau τ is simply n(λ) − coch(τ ), so that charge is not truly worth an
independent description. The maximal value for the co-charged statistic, for a shape λ tableau,Pis n(λ).
This maximal value is realized by the standard tableau τ , whose entries are τ (i, j) = j − 1 + k<i λ0i .
Many formulas involving the enumeration of tableaux admit a charge/co-charge refinement. For
instance, we have (see [37])
X [n]q !
f µ (q) := q coch(τ ) = q n(µ) Q , (1.4.1)
c∈µ (1 − q h(c) )
λ(τ )=µ
where the sum is over the set of standard tableaux of shape µ. One gets back the usual hook
length formula by taking the limit as q tends to 1. We will encounter later the Kostka-Foulkes
polynomials:
X
Kλµ (q) = q ch(τ ) ,
τ
with sum over the set of semi-standard tableaux of shape λ and content µ. The q-Kostka matrix for
n = 4 is:
1 q q2 q3 q6
 
0 1 q q 2 + q q 5 + q 4 + q 3 
 
0 0 1
 q q4 + q2  
0 0 0 1 q3 + q2 + q 
0 0 0 0 1
15
1.5 Exercises and problems
Exercice 1.1. Writing [n]q for the q-integer 1 + q + . . . + q n−1 , one considers the q-factorial
[n]q ! := [1]q [2]q · · · [n]q ,
with [0]q ! set to be equal to 1. One then defines the q-binomial coefficient:
n [n]q !
k q := [k]q ! [n−k]q ! .
(a) Show that we have the q-Pascal-triangle recursion

n+1 k n
n
k q = q k q + k−1 q ,
n
with appropriate initial conditions. Conclude that k q is a polynomial in q with positive
integer coefficients.
(b) Prove the q-analog of Newton’s theorem:
Qn−1 k
Pn k(k−1)/2 n z k .

k=0 (1 + q z) = k=0 q k q (1.5.1)
Exercice 1.2. Show that formula (1.1.2) holds, by proving that the left-hand side satisfies the same
recursion as the right-hand side.
Exercice 1.3. Show that Pµ (q) = Pµ0 (q), see definition (1.1.3), for all partition µ.
Exercice 1.4. Let
δn := (n − 1, n − 2, · · · , 2, 1) (1.5.2)
be the n-staircase partition (see figure below), and consider the q-enumeration of partition sitting
inside δn : X
Cn (q) := q |µ| .
µ⊆δn
Using the “first return” to diagonal decomposition illustrated in the following figure:
p p
first return
( (
n−j n−j
| {z } p | {z } p
j j
prove the recursion:

n
X
Cn (q) = q j(n−j) Cj−1 (q)Cn−j (q), (1.5.3)
j=1
hence that Cn (q) lies in N[q]. Can you see that this implies that
Cn (1) = n+1
1 2n

n (Catalan numbers)?
16
Exercice 1.5. (explore) For k < ` and n < m, find an expression for the q-enumeration of
partitions µ such that nk ⊆ µ ⊆ n` .
Exercice 1.6. (a bit harder) Prove that
1
2n
Cn (q) := [n+1]q n q lies in N[q]. (1.5.4)
At least check that Cn (q) 6= Cn (q) in general.

Exercice 1.7. (open) For any a ≤ b, c ≤ d in N (no restriction between b and c), such that ad = bc,
show that b+c a+d
b q − a q ∈ N[q]. (1.5.5)
You may check this in Sage. Can you prove it for some simple families of cases?
Exercice 1.8. (open exploration) Find all the pairs of partitions of n such that either the
difference Pµ (q) − Pν (q) or Pν (q) − Pµ (q) lies in N[q]. Experiments suggest that this is the case
very often, so that it may be more interesting to find when it fails to hold. Observe that exercise
(d) reduces the number of cases to consider. It may be interesting the evident fact that when both
Pµ (q) − Pν (q) and Pν (q) − Pρ (q) lie in N[q], then so does Pµ (q) − Pρ (q).
Exercice 1.9. In this exercise, we assume that we are in the context of the classical RSK corre-
spondence (so that bi = i). Show that for all semi-standard tableau τ we have (ε ← ρ(τ )) = τ with
ρ(τ ) standing for the reading word of τ . For example
3 3
 
 2 ← 21132412  = 2 2 2 4
1 3 1 1 1 1 2
Exercice 1.10. Consider the skew partition obtained as the difference µ/λ
with µ equal to the shape of (τ ← v), for some semi-standard tableau τ of shape λ:
5
 
 2 5 ← 1 1 2 3 5 = 2 3 4 6
1 3 4 6 1 1 1 2 3 5
Show that there is a bijection between pairs (τ, v), with τ semi-standard tableau of shape λ, and
v non-decreasing sequence of k intergers; and pairs (τ 0 , λ), where τ 0 is a semi-standard tableau of
shape µ such that µ/λ is an horizontal k strip (a skew tableau with k cells, none of which lies on
top of the other).
Exercice 1.11. The pointwise sum, µ + λ, of partitions µ and λ, is the partition whose parts are
µi + λi (adding zero parts if needed). For µ having at most n parts, find a multinomial formula for
the number of standard tableaux of skew-shape (µ + 1n )/µ (which is a vertical strip with n-cells, as
illustrated in figure 1.8).
17
5
1
3
7
4
2
6
Figure 1.8: A standard tableau of shape (µ + 1n )/µ, with µ = 53332 and n = 7.
Exercice 1.12. Show that the Kostka matrix (Kλµ )λ,µ`n is upper unitriangular, if partitions are
ordered with any linear extension of the dominance order. If you use the lexicographic order, e.g.:
4, 31, 22, 211, 1111 for n = 4, you get:
 
1 1 1 1 1
0 1 1 2 3
 
0 0 1 1 2
 
0 0 0 1 3
0 0 0 0 1
Exercice 1.13. Show that, for µ ` n,
f µ = n! det (aij ) ,
where aij = 1/(µi + j − i).
Exercice 1.14. (open) Explore the value of µ`n (f µ )k , for k ≥ 2.

P
18
Chapter 2
“Classical” Symmetric functions
All of the following (and much more) may be found in Macdonald’s book [37], Sagan’s [41], or
Stanley’s [43]. See also [38] for links to algebraic geometry. We mostly use Macdonald’s notation.
The usual Sage tutorial for symmetric functions is available here. An improved Sage Symmetric
Function Tutorial1 was prepared by Pauline Hubert and Mélodie Lapointe (UQAM). It is more
adapted to our presentation. In Maple, one typically uses John Stembridge’s package SF.
2.1 Basic notions
We write x for the list (vector) n variables (x1 , x2 , . . . , xn ), and use “vector” notation for monomials:
xa = xa11 xa22 · · · xann , for a = (a1 , a2 , . . . , an ) ∈ Nn .
The degree of xa is deg(xa ) = |a| = a1 +a2 +. . .+an . Thus for a scalar q, we have (q ·x)a = q d (xa)
with d = deg(xa ). For certain, q · x := (qx1 , qx2 , . . . , qxn ). It is well known that there are n+d−1
d
monomials of degree d in n variables. The ring of polynomials in n-variables, R := Q[x], is graded
by degree M
R= Rd , with Rd = Q{xa | deg(xa ) = d}.
d∈N
n+d−1
-dimensional Q-vector space Rd is the d-degree homogeneous component of R.

The d
Equivalently, we have
p(x) ∈ Rd , iff p(q x) = q d p(x).
We then say that p(x) is homogeneous of degree d. Any polynomial p(x) is uniquely decomposed
as a sum of homogeneous polynomials, i.e.:
p(x) = p0 (x) + p1 (x) + p2 (x) + . . . + pn (x),
where n is the degree of p(x), and pd (x) lies in Rd . We say that pd (x) is the degree d homogeneous
component of p(x).
1
https://more-sagemath-tutorials.readthedocs.io/en/latest/tutorial-symmetric-functions.html
19
2.2 From symmetric polynomials to symmetric functions
Recall that symmetric polynomials are polynomials f (x) in R such that, for any permutation σ
of the set {1, 2, . . . , n} (or σ ∈ Sn ), we have
f (xσ(1) , xσ(2) , . . . , xσ(n) ) = f (x1 , x2 , . . . , xn ).
The set of symmetric polynomials forms a graded subring of R, denoted by Λ, and Λ = d∈N Λd .
L
Indeed, It is clear that the sums and products of symmetric polynomials are also symmetric
polynomials. Moreover, if f is a symmetric polynomial, so are all of its homogeneous components.
In other words, Λd = Rd ∩ Λ. A linear basis of Λd is formed by the set of (non-vanishing) monomial
symmetric polynomials X
mλ (x) := xa , λ`d
a∈r(λ)
with a varying in the set r(λ) of rearrangements of the length n vector (λ1 , . . . , λk , 0, . . . , 0). For
example
m211 (x1 , x2 , x3 ) = x21 x2 x3 + x1 x22 x3 + x1 x2 x23 ,
Observe that our definition forces mλ = 0 when `(λ) > n. As we will see, this makes it so that
identities within Λd are simpler to describe when n > d. Moreover, they specialize with no surprises
when one restricts the number of variables. For this reason, and after some initial care, it is customary
to send n to infinity, and consider a denumerable set of variables x = (x1 , x1 , x3 , . . .). The resulting
“polynomials” in infinitely many variables are then called symmetric functions. In fact, most of
the time we will forego actual mention of the variables, and say for example that {mλ }λ`d forms a
linear basis of Λd . Thus the dimension of Λd is the number of partitions of d.
Among the interesting bases of Λd , one first considers the following three “multiplicative” bases, all
indexed by partitions µ = (µ1 , µ2 , . . . , µk ) of d:
• The complete homogeneous symmetric functions are defined to be the product hµ :=

hµ1 hµ2 · · · hµk , with simply indexed hd defined as:
X
hd := mλ .
λ`d
For example, we have h1 = m1 ;
h2 = m11 + m2 , h11 = h21 = 2 m11 + m2 ;

h3 = m111 + m21 + m3 , h21 = 3 m111 + 2 m21 + m3 , h111 = 6 m111 + 3 m21 + m3 ;
and
h4 = m4 + m31 + m22 + m211 + m1111 ,

h31 = m4 + 2 m31 + 2 m22 + 3 m211 + 4 m1111
h22 = m4 + 2 m31 + 3 m22 + 4 m211 + 6 m1111 ,
h211 = m4 + 3 m31 + 4 m22 + 7 m211 + 12 m1111 ,
h1111 = m4 + 4 m31 + 6 m22 + 12 m211 + 24 m1111 .
20
• The elementary symmetric functions defined as the product eµ := eµ1 eµ2 · · · eµk , with simply
indexed ed defined as ed := m1d . Thus e1 = m1 = h1 , implying that e1d = h1d . Other values
(not already explicitly given) are e21 = 3 m111 + m21 , and
e211 = m31 + 2 m22 + 5 m211 + 12 m1111 ,

e22 = m22 + 2 m211 + 6 m1111 ,
e31 = m211 + 4 m1111 .
• The power sum symmetric functions defined as the product pµ := pµ1 pµ2 · · · pµk , with simply
indexed pd defined as pd := md . Again, p1 = m1 = h1 = e1 , so that p1d = h1d = e1d . Other
values are p21 = m21 + m3 , and
p31 = m4 + m31 ,
p22 = m4 + 2 m22 ,
p211 = m4 + 2 m31 + 2 m22 + 2 m211 .
There are nice combinatorial descriptions for all the transition matrices between the four bases above
(i.e., mµ , hµ , eµ , and pµ ). In some instances these come out of the following identities, best coined in
terms of generating series in an extra variable z. First, the above definition is equivalent to (why?)
X Y 1 X Y
H(z) := hd z d = and E(z) := ed z d = 1 + xi z (2.2.1)
1 − xi z
d≥0 i≥1 d≥0 i≥1
It is thus immediate that H(z) E(−z) = 1, and a simple calculation gives

a) H(z) = exp P (z) , and b) E(z) = exp − P (−z) , (2.2.2)
if one sets P (z) := d≥1 pd z d /d. Comparing coefficients of z d in the above identities gives:
P
d X pµ
X X pµ
a) (−1)k hd−k ek = 0, b) hd = , c) ed = (−1)d−`(µ) , (2.2.3)
zµ zµ
k=0 µ`d µ`d
We can recursively solve (2.2.3.a) either to write the hk in terms of ej , or vice-versa. Also, from (2.2.2.a)
we get H 0 (z) = P 0 (z) H(z), from which it follows that
d
X
d hd = pr hd−r . (2.2.4)
r=1
Using Cramer’s rule to solve this system of equations for the indeterminate pk , one finds the following
expansion in terms of the hd ’s:
h1 1 0 ... 0
2 h2 h1 1 ... 0
pk = (−1)k−1 det 3 h3 h2 h1 ... 0 . (2.2.5)
.. .. .. .. ..
. . . . .
k hk hk−1 hk−2 . . . h1
21
Pd r−1 p e
Similarly, we get d ed = r=1 (−1) r d−r from (2.2.2.b), and derive
e1 1 0 ... 0
2 e2 e1 1 ... 0
pk = det 3 e3 e2 e1 ... 0 (2.2.6)
.. .. .. .. ..
. . . . .
k ek ek−1 ek−2 . . . e1
The omega involution is the linear and multiplicative operator that sends pd to (−1)d−1 pd . Clearly,
for µ ` d,
ω(pµ ) = (−1)d−`(µ) pµ .
From the above, one sees easily that ω(hd ) = ed , or equivalently that ω(ed ) = hd (since ω 2 = Id).
2.3 Schur functions, a combinatorial approach
The Schur function sλ (x) is combinatorially defined as:

X
sλ (x) := xτ , (2.3.1)
τ
Q
with the sum being over all semi-standard tableaux τ : λ → N, and xτ := c∈λ xτ (c) . This
P definition
is naturally extended to skew shapes λ/µ to get skew Schur polynomials sλ/µ (x) := τ xτ , with
the sum over semi-standard tableaux of shape λ/µ. If one assumes (as is shown in the exercises)
that sλ (x) is a symmetric polynomial, then Equation (2.3.1) implies that
X
sλ (x) = Kλµ mµ (x), (2.3.2)
µ`n
In particular, we always have s(n) = hn , and s1n = en . One may show (see exercise) that ωsµ = sµ0 .
Often, it is natural to consider a tableau being filled with the variables xi , rather than their indices i.
With this point of view, the evaluation monomial is simply the product of all entries of the tableau.
Then one may consider sλ (x + y) as the evaluation of the Schur function sλ on the variables
x + y = (x1 + x2 + x3 + . . .) + (y1 + y2 + y3 + . . .), with the convention that all yk are larger than
all xi . It follows that (also see section 2.8 on plethysm)
X
sλ (x + y) = sµ (x) sλ/µ (y), (2.3.3)
µ⊆λ
since any semi-standard tableau τ : λ → x + y, can naturally be separated into the semi-standard
tableau corresponding to the sub-partition µ ⊆ λ, where only the xi appear, and the skew semi-
standard tableau of shape λ/µ, where only the yk appear.
For the special cases λ = (n) and λ = 1n , we get
n
X n
X
hn (x + y) = hk (x) hn−k (y), and en (x + y) = ek (x) en−k (y). (2.3.4)
k=0 k=0
22
Exploiting the RSK correspondence, we find interesting formulas. Indeed, associate to

b1 b2 . . . bd
w= ,
a1 a2 . . . ad
the product of degree d monomials xa y b , with a = (a1 , . . . , ad ) and b = (b1 , . . P

. , bd ). Thus we
identify the set of length d lexicographic two row matrices, with the terms of the sum |a|=|b|=d xa y b .
This sum is readily seen to be the same as degree d homogenous component of
X X 1
hd (x y) = ,
1 − x i yj
d≥0 i,j≥1
with hd expanded in the “variables” xi yj .

We now apply the RSK correspondence to each monomial xa y b (translated back into a lexicographic
two row matrices) to get a pair of same shape
P semi-standard
P tableaux (P, Q), for which we have
xa y b = xP y Q . We may rewrite this sum a,b xa y b as P,Q xP y Q , with the sum now being over
the pairs of semi-standard tableaux of same shape. It follows that we have the identities
X X
hk (xy) = sλ (x)sλ (y). and ek (xy) = sλ0 (x)sλ (y). (2.3.5)
λ`k λ`k
applying ω to the x component, to get the second from the first.

Now, summing up both sides of (2.3.5) for n, we get Cauchy-Littlewood’s formula
Y 1 X
= sλ (x)sλ (y). (2.3.6)
1 − x i yj
i,j≥1 λ

1 2 . . . k
If we restrict this argument to matrices of the form w = a a . . . a , and consider the
1 2 k
evaluation monomial of w to be xa , and the evaluation monomial of a pair (P, Q) to be xP . We get
the formula (see exercises) X
hn1 = f λ sλ (x) (2.3.7)
λ`n
with fλ standing for the number of standard tableaux of shape λ.

Another interesting formula (see Stanley [43, Proposition 7.19.11]) says that
Kµ,1n (q)
sλ/µ (1, q, q 2 , q 3 , . . .) = . (2.3.8)
(1 − q)(1 − q 2 ) · · · (1 − q n )
2.4 Dual basis and Cauchy kernel
On Λ, we consider the Hall scalar product characterized by

(
zλ , if λ = µ,
hpλ , pµ i := (2.4.1)
0, if λ 6= µ.
23
Which is readily seen to be preserved by ω, i.e.: hω(f ), ω(g)i = hf, gi. We may calculate directly
that
X 1
hhn , hn i = hen , en i = .
zµ
µ`n
In fact, this sum is equal to 1 (see exercise).

The Schur functions may also be obtained via the Gram-Schmidt orthogonalization process applied
to the basis of monomial symmetric {mλ }λ`n , written in increasing lexicographic order of partitions.
Indeed, the Schur polynomials are uniquely determined by the following two properties:
(1) hsλ , sµ i = 0, whenever λ 6= µ,

(2.4.2)
X
(2) sλ = mλ + cλµ mµ ,
µ≺λ
for some coefficients cλµ to be calculated so that the first property holds. Although it is not
immediate, this makes it so that
hsµ , sµ i = 1, (2.4.3)
for all partitions µ. One way to “see” this is through the following discussion. We start with the
simple computation
Y 1 X
= exp − log(1 − xi yj )
1 − xi yj
i,j≥1 i,j
XX
= exp (xi yj )k /k
i,j k≥1
X
= exp pk (x) pk (y)/k
k≥1
X pλ (y)
= pλ (x) . (2.4.4)
zλ
λ
The left-hand side of (2.3.6) and (2.4.4) is the Cauchy kernel, henceforth denoted by Ω(xy). This
is a symmetric series in the variables “xi yj ” (as will become clearer in section 2.8). Its degree n
homogeneous component (with xi yj being considered as having degree 1) is
X pλ (y)
hn (xy) = pλ (x) . (2.4.5)
zλ
λ`n
Various similar formulas characterize dual basis pairs {uλ }λ and {vµ }µ . These are pairs of bases
such that (
1, if λ = µ,
huλ , vµ i = (2.4.6)
0, otherwise.
The relevant statement here is that the set conditions (2.4.6) is equivalent to the single identity
X X
Ω(xy) = hn (xy) = uλ (x) vλ (y). (2.4.7)
n≥0 λ
24
You are asked to show this in an exercise. In particular, identities (2.3.6) and (2.11.6) are equivalent
to hsλ , sµ i = δλ,µ , and hmλ , hµ i = δλ,µ , where δλ,µ is the usual2 Kronecker delta function. In
other words, {sλ }λ is a self dual orthonormal basis. Exploiting the duality between {mλ }λ and
{hµ }µ , and equality (2.3.2), we get X
hµ = Kλ,µ sλ (2.4.8)
λ
Comparing this with Equation (2.3.2) we notice that the index of summation is now λ rather than µ.
2.5 Product, Kronecker product, and Schur positivity
Since the product of two Schur functions is also a symmetric function, it may be expanded in terms
of the Schur basis. In formula, for µ ` m and ν ` n, we have
X
sµ sν = cλµν sλ . (2.5.1)
λ`(m+n)
It is a fact that the coefficients cλµν , known as the Littlewood-Richardson coefficients, are
all positive integers3 . They also occur in the Schur expansion of skew Schur functions, i.e.: for
λ ` (m + n) and µ ` m, we have X
sλ/µ = cλµν sλ . (2.5.2)
ν`n
Taken together, (2.5.1) and (2.5.2) imply that, for all µ, ν and λ, one has
hsµ sν , sλ i = hsν , sλ/µ i. (2.5.3)
This makes it natural to consider, for any symmetric function f , the skewing operator, denoted
by f ⊥ , which is simply the adjoint of the multiplication operator f · (−), with respect to the scalar
product. In formula, since the set of Schur functions forms a linear basis, it is characterized by the
property
hf · sν , sλ i = hsν , f ⊥ sλ i. (2.5.4)
In particular, s⊥
µ sλ = sλ/µ .
There is a nice combinatorial rule (not presented here) due to Littlewood and Richardson to express
the cλµν as enumerators for certain semi-standard tableaux. Two special cases are the following Pieri
formulas, where all coefficients are either 0 or 1:
X X
hk sµ = sθ , (Pieri) ek sµ = sθ (dual Pieri), (2.5.5)
θ θ
with the sum being over partitions θ for which θ/µ is a k-cell horizontal strip (see exercise 1.10) in
the first formula, and over those for which θ/µ is a k-cell vertical strip (no two cells in the same
row) in the second. One may use these Pieri rules to “calculate” any sµ (see exercise).
2
Which is so usual that one always seems to be required to mention the fact.
3
They are hard to calculate, and in fact closely tied to deep questions in Geometric Complexity Theory: a.k.a. the
study of the (VP vs VNP) problem.
25
The Kronecker product of symmetric functions, denoted by f ∗ g, is the unique bilinear product
(commutative and associative) such that
(
zλ pλ , if λ = µ,
pλ ∗ pµ := (2.5.6)
0, otherwise.
Observe that
( (
f, if f ∈ Λn , ω(f ), if f ∈ Λn ,
hn ∗ f := and en ∗ f := (2.5.7)
0, otherwise; 0, otherwise.
P P
Hence, H = n hn acts as identity for this product; and multiplication by E = n en corresponds
to ω. We easily check that ω(f ) ∗ g = f ∗ ω(g). Calculating sµ ∗ sλ is hard.
Littlewood showed the nice identity
X
(sα sβ ) ∗ sθ = cθλµ (sα ∗ sλ ) (sβ ∗ sµ ), (2.5.8)
λ,µ
with the sum being over the pairs λ, µ such that: cθλµ is not zero, λ ` |α|, and µ ` |β|.
A common feature of all of these is that they have expansions with positive integer coefficients in
terms of the Schur function basis. When this is the case, we say that we have Schur positivity, or
that the function in question is Schur positive. Illustrating with instances of operations discussed
above, we have
s⊥
21 s4321 = s43 + 2 s421 + s4111 + 2 s331 + 2 s322 + 2 s3211 + s2221 ,
s3111 ∗ s321 = s51 + 2 s42 + 2 s411 + s33 + 4 s321 + 2 s3111 + s222 + 2 s2211 + s21111 ,
s21 s32 = s53 + s521 + s44 + 2 s431 + s422 + s4211 + s332 + s3311 + s3221 .
The analogous stronger notions of h-positive and e-positive are also considered. In view of (2.3.2),
it is clear that we have monomial positivity for any Schur positive functions. The global result (see
exercises) is that
Theorem 2.5.1. For any Schur positive symmetric functions f and g, the symmetric functions
f + g, f g, f ∗ g, and f ⊥ g are also Schur positive. Any h-positive (or e-positive) symmetric function
is also Schur positive.
In fact, there is a nice theorem that makes apparent that Schur positivity is a rare phenomenon.
Foregoing a technical description4 , it may be stated as follows.
Theorem 2.5.2 (B-Patrias-Reiner 2017). The probability of choosing at random (with uniform
distribution) a Schur positive symmetric function in the set of homogeneous symmetric functions of
degree n which are monomial positive, is equal to
1
Q P , (2.5.9)
µ`n λ`n Kλµ
with Kλµ the Kostka numbers.

4
One must make precise the sampling space, see exercises.
26
For n = 1, 2, 3, . . ., the first terms of the sequence of denominators are
1, 2, 9, 560, 480480, 1027458432000, 2465474364698304960000, . . .
making the point that Schur positivity is indeed rare. There is a similar statement making the point
that h-positivity (or e-positivity) is rare among Schur positivity functions.
For symmetric functions with coefficients in Q(q, t) (such as we will consider in the sequel), it is
natural to say that we have Schur positivity when functions have Schur expansions with coefficients
in N[q, t].
2.6 Transition matrices
The matrices that express the possible changes of basis between the 6 fundamental bases that have
been discussed may be described as follows, borrowing figure 2.1 from Macdonald [37, page 104].
Each oriented edge encodes a transition matrix between bases, with bases labeling the vertices. Thus
the arrow labeled L = (Lµλ ) stands for the matrix describing how write the pµ ’s in terms of the
hλ , K stands for the Kostka matrix, and J for the matrix that corresponds to the involution ω. As
usual we write M Tr for the transposed of the matrix M . By composition of the red arrows (or their
inverse), all unlabeled edges may be filled in to complete the graph.
s
K0J K0
e h
JK K
f m
L
p
Figure 2.1: Transition matrices.
For the n = 3 case, a systematic presentation (excluding the Schur basis, and the forgotten one) is
27
given by the following table where all expressions along a given line are equal.
m3 e31 − 3 e1 e2 + 3 e3 h31 − 3 h1 h2 + 3 h3 p3
m21 e1 e2 − 3 e3 −2 h31 + 5 h1 h2 − 3 h3 p1 p2 − p3
1 3
m111 e3 h31 − 2 h1 h2 6 (p1 − 3 p1 p2 + 2 p3 )
1 3
m21 + 3 m111 e1 e2 h31 − h1 h2 2 (p1 − p1 p2 )
m3 + 3 m21 + 6 m111 e31 h31 p31
1 3
m3 + m21 + m111 e31 − 2 e1 e2 + e3 h3 6 (p1 + 3 p1 p2 + 2 p3 )
1 3
m3 + 2 m21 + 3 m111 e31 − e1 e2 h1 h2 2 (p1 + p1 p2 )
m3 + m21 e31 − 2 e1 e2 −h31 + 2 h1 h2 p1 p2
2.7 Jacobi-Trudi determinants
Jacobi-Trudi’s formula (resp. its dual) gives an explicit expansion of the Schur functions in terms
of complete homogeneous functions (resp. elementary). We have
sµ = det(hµi +j−i )1≤i,j≤n , and resp. sµ0 = det(eµi +j−i )1≤i,j≤n , (2.7.1)
with hk = ek = 0 whenever k < 0. This may be nicely deduced from the tableaux definition of Schur
functions using the technology of Lindström [36] and Gessel-Viennot [23]. It is articulated around
an interpretation of Schur functions as configurations of non-intersecting north-east lattice paths in
Z × Z, which correspond to tableaux as illustrated in figure 2.2.
µ3 − 3 µ2 − 2 µ1 − 1
5
4
4 ↔ 3
2 5 2 2 2
1 2 2 3 s s s1
−3 −2 −1
Figure 2.2: Non-crossing configuration associated to a semi-standard tableau.
2.8 Plethysm
Plethysm was first defined by Littlewood (see [37]), as an operation between symmetric functions.
More generally, we consider here a symmetric function f as an operator that transforms a rational
expression A into an expression f [A]. Let A and B be rational fraction expressions in variables
28
x = (x1 , x2 , . . .), and consider symmetric functions f and g in Λ. Plethysm is characterized by the
following “calculation” rules, with α, and β denoting scalars:
Rules for Plethysm
1. (αf + βg)[A] = α f [A] + β g[A],
2. (f · g)[A] = f [A] · g[A],
3. pk [αA + βB] = α pk [A] + β pk [B], for any power sum pk ,
4. pk [A · B] = pk [A] · pk [B],
5. pk [A/B] = pk [A]/pk [B],
6. pk [x] = xk , for x a variable,
7. pk [1] = 1.
Since any symmetric function f may be expanded as a polynomial in the pk , the first two rules
clearly reduce the calculation of f [A] to instances of pk [A]. Then, A being a rational expression
in the xi ’s, the last five rules may be used to finish any needed calculation. Words of caution
are necessary here. First, one needs to be clear about the distinction between variables and scalars.
Indeed, on variables z one applies the rule x 7→ xk , whereas scalars are left invariant under plethysm.
Second, as f [−] is an operator it needs not commute with other operators, such as specialization of
the variables.
We may consider Formula (2.3.3) as an instance of a plethystic calculation. Also, it follows from the
rules that
X X zk
hn [A] z n = exp pk [A] , and (2.8.1)
k
n≥0 k≥1
X X zk
en [A] z n = exp (−1)k−1 pk [A] , (2.8.2)
k
n≥0 k≥1
so that we may calculate many instances of plethysm in one stroke. It is then helpful to remember
that, considering z as a variable, we have f [z A] = z d f [A], whenever f is homogeneous of degree d.
Observe that a special case of plethysm may be considered an operation between symmetric functions
(considering the case when A lies in Λ). In this situation one also writes f ◦ g for f [g]. This is an
associative operation, and the above rules imply that
pk ◦ pj = pkj . (2.8.3)
The “unique” homogeneous symmetric function of degree 1, say p1 (= m1 = h1 = e1 = s1 ), acts as

identity for plethysm. In the context of plethysm, it is also very handy to consider variables sets as
formal sums, writing
x = x1 + x2 + x3 + . . . + xn + . . .
29
Indeed, we then have f [x] = f (x1 , x2 , . . .), since pk [x1 + x2 + . . .] = xk1 + xk2 + . . . = pk (x). In the
sequel, it will be convenient to use the star notation
h x i
f ∗ (x) := f . (2.8.4)
1−q
Some examples
• Calculating f [x1 + x2 + . . . + xk ] corresponds to the specialization of symmetric functions to

a finite number of variables. We already know that mµ [x1 + x2 + . . . + xk ] vanishes when
`(µ) > k, and this is also the case for sµ .
• Calculating f [−1] comes as our first “surprise”. Indeed, using (2.8.1) we get:
X X zk X zk
hn [−1] z n = exp pk [−1] = exp − =1−z
k k
n≥0 k≥1 k≥1
X
n
X
k−1 zk X (−z)k 1
en [−1] z = exp (−1) pk [−1] = exp = ,
k k 1+z
n≥0 k≥1 k≥1
so that

1, if n = 0,
 (
1, if n is even,
hn [−1] = −1, if n = 1, and en [−1] =
 −1, otherwise.
0, otherwise,

Using the Jacobi-Trudi identities, we may deduce that sµ [−1] = 0 for all partitions of n ≥ 3,
except when µ = 1n . Using the fact that pk [−x] = −pk (x), it may likewise be seen that
sµ [−x] = (−1)n sµ0 (x). (2.8.5)
• For the plethysm f [x − y], mixing Formula (2.3.4) with (2.8.5) we find that
n
X
hn [x − y] = (−1)n−k hk (x) en−k (y). (2.8.6)
k=0
More generally, for all partition µ, we have

X
sµ [x − y] = (−1)n−k sµ/ν (x) sν (y). (2.8.7)
ν⊆µ
• For any partition µ of n, one may show that for a positive scalar k:
Y k+j−i
sµ [k] = . (2.8.8)
hij
(i,j)∈µ
30
In fact, this is the number of semi-standard tableaux of shape µ with values in {1, 2, . . . , k}.
One may also show that:
1−qk Y 1 − q k+j−i
(a) sµ 1−q = q n(µ) , and (2.8.9)
1 − q hij
(i,j)∈µ
q n(µ)
(b) s∗µ (1) = sµ 1

1−q =Q hij
. (2.8.10)
(i,j)∈µ 1 − q
• We have (see exercises)

(
sµ [1 − u] (−u)k , if µ = (n − k, 1k ),
= (2.8.11)
1−u 0, otherwise,
• We calculate that
X z (1 − q) (1 − t)
en [(1 − q)(1 − t)]z n = ,
(1 + t z) (1 + q z)
n≥1
which implies that

(−1)n−1
en [(1 − q)(1 − t)] = [n]q,t , (2.8.12)
(1 − q)(1 − t)
where we use the (q, t)-integer notation
q n − tn
[n]q,t := = q n−1 + q n−2 t + . . . + q tn−2 + tn−1 .
q−t
Similarly, we may see that

h 1 i Yn
1
h∗n (1) = hn = (2.8.13)
1−q 1 − qk
k=1
Observe that
1
= (1 − q)n [n]q ! (2.8.14)
h∗n (1)
• Starting with
X 1 `(µ)
Y pµ (x)
h∗n (x) = i
, (2.8.15)
zµ 1 − q µi
µ`n i=1
together with (2.3.8), and the Cauchy kernel formula, we may calculate that
h∗n (x) X s∗µ (1)

= sµ (x), (2.8.16)
h∗n (1) h∗n (1)
µ`n
X
= Kµ,1n (q) sµ (x). (2.8.17)
µ`n
31
This last N[q]-coefficient symmetric polynomial is an important instance of the “Combinatorial
Macdonald polynomials”, denoted by Hµ (q; x), that will arise in the next chapter. Thus, we have
h∗n (x)
Hn (q; x) := . (2.8.18)
h∗n (1)
Some examples are:
H1 = s1 ,
H2 = s2 + q s11 ,
H3 = s3 + q 2 + q s21 + q 3 s111 ,

H4 = s4 + q 3 + q 2 + q s31 + q 4 + q 2 s22 + q 5 + q 4 + q 3 s211 + q 6 s1111 ,

H5 = s5 + q 4 + q 3 + q 2 + q s41 + q 6 + q 5 + q 4 + q 3 + q 2 s32 ,

+ q 6 + q 5 + q 4 + q 3 + q 2 s32 + q 7 + q 6 + 2q 5 + q 4 + q 3 s311 ,

+ q 8 + q 7 + q 6 + q 5 + q 4 s221 + q 9 + q 8 + q 7 + q 6 s2111 + q 10 s11111 .

2.9 Some ties with geometric complexity theory
As explained in [39], fundamental P 6= N P questions of complexity theory may be “reformulated”

(this has to be clarified) in the context of calculations of the Schur function expansions of either sµ sν ,
sµ ∗ sν , or of sµ ◦ sν . The aim is to prove typical complexity statements about explicit calculations
of the Littlewood-Richardson coefficients cλµν in the expansion (2.5.1):
X
sµ sν = cλµν sλ ,
λ`(m+n)
or similar expansions for Kronecker product, or plethysm. Questions regarding these include showing
that cλµν =
6 0, for µ, ν, and λ satisfying some constraints. The Littlewood-Richardson rule does give
a procedure, but it is hard to apply when partitions involved become large. The following is chief
among results in this context.
Theorem 2.9.1 (Knutson-Tao, De Loera-McAllister). The problem of deciding the nonvanishing of
cλµν is in P , this is to say that it can be solved in time that is polynomial in the bit-lengths of µ, ν
and λ.
There are several open conjecture in this area, which is currently very active. An overview may be
found in [14], and more complete presentations in [33, 34]. You may find there an explanation of
how the Foulke’s conjecture plays a role in this story.
2.10 Counting with symmetric functions
There are several classical formulas and identities from enumerative combinatorics that may be
obtained using symmetric functions. Indeed, we can get interesting integers (or q-integers) out of a
32
symmetric function f (typically homogeneous), in the following way. First, one may calculate some
scalar product hf, gi choosing a “special” g, such as
hf, en i, hf, hn i, hf, hk en−k i, or hf, sk1 i.
For example,
n
hhn1 , hn1 i = n!, hhn1 , hk en−k i = ,
k
One may also evaluate f plethysticaly at an integer or q-integer:
f [n], or f [1 + q + . . . + q n−1 ].
For instance, formulas (2.8.8) and (2.8.9) give:

n n+k−1
ek [n] = , hk [n] = , (2.10.1)
k k

k n n+k−1
ek [1 + q + . . . + q n−1 ] = q (2) , hk [1 + q + . . . + q n−1 ] = (2.10.2)
k q k q
It follows that we get combinatorial identities out of any symmetric function identity, and this applies
to series. To illustrate, consider what happens to the series in (2.2.1) when one assumes that there
are n variables xi , which are all set to equal 1. We get, using (2.10.1),
X n X n + k − 1 n
k k k 1
z = (1 + z) , and z = . (2.10.3)
k k 1−z
k≥0 k≥0
similarly, with xi = q i−1 , we find that

n X n + k − 1 n
X
( k
) n k Y Y 1
q 2 z = (1 + q i−1 z), and k
z = . (2.10.4)
k q k q 1 − q i−1 z
k≥0 i=1 k≥0 i=1
Exploiting remarks of our first chapter, it may be seen that the last formula counts partitions with
some restriction (which is it?), taking into account the number of parts that they have.
The theory of species (for which the typical reference is [5]) furnishes a very general systematic
approach to the part of enumerative combinatorics that considers constructions on finite sets, and
the corresponding Pólya theory. Recall that Pólya theory gives general tools for the enumeration
of isomorphism type of structures (a.k.a. structures on unlabelled objects). See chapter
6 of Bergeron/Combinatoire_Algebrique (in french), or wiki/Combinatorial_species for simple
introductions.
To each species of combinatorial objects there corresponds a symmetric function series, known as its
Pólya-Joyal cycle index series. Via a combinatorial calculus on species, which allows the construction
of more complex species out of simpler ones, one gets symmetric function series identities (involving,
sums, product, plethysm, and Kronecker product). This makes it possible to obtain both the
enumeration of labelled and unlabelled structures, simply from specializations of formulas entirely
coined in terms of symmetric functions series, in which the “basic” variables are power-sums.
33
Examples of typical species include: graphs, connected graphs, oriented graphs, permutations, cyclic
permutations, derangements, orders, lists, k-subsets, set partitions, trees of various kinds, etc. In
the algebra of (combinatorial constructions on) species, one may solve equations, consider Lagrange
inversion, and interpret combinatorially many other classical calculations. For sure, all of this
corresponds to interesting calculations on symmetric functions series.
Exercice 2.1. (a) Adapt RSK to show that

X
(x1 + x2 + . . . + xk )n = f µ sµ (x), (2.11.1)
µ`n
with f µ the number of standard tableaux of shape µ.
(b) Find the monomial functions expansion of the above formula (as a linear combination of the
mµ ), and check that it is essentially Newton’s binomial formula.
Exercice 2.2. For any partition µ, let us set
πµ (x) = πµ1 (x)πµ2 (x) · · · πµk (x), (2.11.2)
where πn (x) is the following alternating sum of Schur functions indexed by hooks:
k−1 n−1
πn (q, t; x) := s1n (x) + . . . + −1/q t s(k,1n−k ) (x) + . . . + −1/q t sn (x). (2.11.3)
(a) Exploit (2.2.2) to check that P (z) = −E 0 (−z)H(z), and use the Pieri rule to deduce from this
the Schur expansion of pn .
(b) Check that πn specializes to (−1)n−1 pn whenever qt = 1.
(c) Conclude that the set {πµ }µ`d constitutes a basis of Λd .
Exercice 2.3. The aim here is to prove that

bn/2c
X
pn−2
1
k k
p2 is Schur-positive, (2.11.4)
k=0
for all n ≥ 1. Recall that p2 = s2 − s11 . Consider the series

1 2
2 3
3 4 2 2
4
= 1 + p1 z + p 1 + p2 z + p1 + p 1 p 2 z + p1 + p 2 p 1 + p 2 z + ...
(1 − p1 z) (1 − p2 z 2 )
(a) Show that

1 1 + s1 z 1 1
= − .
(1 − p1 z) (1 − p2 z 2 ) 2 s11 z 2 1 − (s2 + s11 ) z 2 1 − (s2 − s11 ) z 2
34
(b) Show that the above expression may be written without division by 2 s11 z 2 .
(c) Conclude that (2.11.4) is Schur positive.
Exercice 2.4. (a) Assuming the Pieri rules, prove that h-positivity (or e-positivity) implies Schur
positivity, and that the product of a h-positive symmetric function with a Schur positive one
is also Schur positive.
(b) Show that one may recursively calculate the Schur functions using the Pieri rule (or dual Pieri
rule). Compare your approach to the Jacobi-Trudi formula.
(c) Exploit the above to prove that ωsµ = sµ0 .
Exercice 2.5. (a) With scalars in R, consider the set of symmetric functions of the form
X
cµ mµ , with 0 ≤ cµ ≤ 1,
µ`n
such that µ`n cµ = 1. One may thus think that this is a simplex in Rp(n) , where p(n) is the
P
number of partitions of n. Show that the Schur positive symmetric functions form a convex
subset of this simplex.
(b) Defining the probability of Schur positivity as the quotient of the volume of this convex subset,
by the volume of the simplex, prove theorem 2.5.2.
(c) (explore) Find a formula for the probability of being e-positive among Schur positive symmetric
functions of degree n. Likewise for h-positivity.
Exercice 2.6. (a) Show that the determinant of the Hankel matrix5
det(hi+j+k )0≤i,j≤n−1
is equal (up to a sign) to a single Schur function (What is the sign? What is the shape?). For
example, for n = 3 and k = 1, we have the Hankel matrix
 
h1 h2 h3
hi+j+1 = h2 h3 h4  .
0≤i,j≤2
h3 h4 h5
(b) Same question for det(ei+j+k )0≤i,j≤n−1 .
Exercice 2.7. (a) Extend formula (2.3.2) to the context of skew Schur functions. Prove that
sλ/µ (x) is invariant under the exchange of the variables xi and xi+1 , by constructing for each
λ/µ-shape semi-standard tableau τ a new semi-standard tableau of same shape but in which
the respective number of occurrences of i and i + 1 are exchanged. Conclude that sλ/µ (x) is
symmetric.
(b) There is an analog of formula (2.3.3) for skew Schur functions. Can you figure out what it is?
5
See [15, 30, 31] for many interesting related identities.
35
(c) There is an analog of the Jacobi-Trudi formula for skew Schur functions. Can you figure out
what it is?
Exercice 2.8. (a) Use the (dual) Pieri formula, and the fact that the Schur functions are or-
thonormal, to show that for all partition µ of n, we have heµ , en i = 1.
(b) Recall that we have set hpλ , pµ i = 0, whenever λ 6= µ. Hence hf, gi = 0 whenever f and g are
both homogeneous, but of different degree. Conclude that,
X X
hf, Ei = cµ if f (x) = cµ eµ (x). (2.11.5)
µ µ
P
where, as before, E := n en .
Exercice 2.9. (a) By a direct calculation, show that
Y 1 X
= hλ (x)mλ (y). (2.11.6)
1 − x i yj
i,j≥1 λ
(b) Show that the bases {uλ }λ and {vµ }µ are dual, if and only if the bases {ω uλ }λ and {ω vµ }µ
are also dual.
(c) Conclude that the set of fµ := ω mµ , for µ ` n, (aka the forgotten basis) constitutes a basis
which is dual to the eλ .
(d) The dual Cauchy Kernel Ω0 (xy) is
Y
Ω0 (xy) := (1 + xi yj ) (2.11.7)
i,j≥1
Check that its degree n component is

X pµ (y)
en (xy) = (−1)n−`(µ) pµ (x) (2.11.8)
zµ
µ`n
X
= s0µ (x) sµ (y) (2.11.9)
µ`n
X
= eµ (x) mµ (y) (2.11.10)
µ`n
X
= hµ (x) fµ (y). (2.11.11)
µ`n
Exercice 2.10. Show the equivalence P of (2.4.6) and (2.4.7) P by expanding pλ and pµ /zµ in the
given basis uλ and vµ : to get pλ = ρ aλρ vρ , and pµ /zµ = ν bµν vν . and that we must have
Tr = Id, where
P
δλ,µ = hpλ , pµ /zµ i = ρ,ν aρλ huλ , vν i bνµ . In matrix form, this becomes A Z B
A := (aλρ ), B := (bµν ), and Z := (huλ , vµ i). Check that statement (2.4.6) is equivalent to Z = Id,
and thus it is also equivalent to A B Tr = Id. Using the fact that
X X pλ (y)
uλ (x) vλ (y) = pλ (x) .
zλ
λ λ
P
and that the pλ (x) pλ (y)/zλ ’s are linearly independent, conclude that λ aλρ bλν = δρ,ν , is equivalent
to (2.4.7).
36
Exercice 2.11. (a) Using (2.8.14) and (2.8.10), show that for all µ, the expression s∗µ (1)/h∗n (1) is
a polynomial in q.
(b) Give an expression for s∗µ (1)/h∗n (1) in terms of [n]q !, and the various [hij ]q for (i, j) ∈ µ.
(c) Find a nice formula for e∗n (1).
Exercice 2.12. (a) Our aim here is to prove (2.8.11). To this end, find the value of each hn [1 − u]
by calculating the series
X
hn [1 − u] z n (2.11.12)
n≥0
using (2.8.1), and pk [1 − u] = 1 − uk . Conclude using Jacobi-Trudi.
(b) Show that

n−1
hn (x) X h x i
= (−q)k s(n−k,1k ) . (2.11.13)
1−q 1−q
k=0
To this end, calculate

hn (x) 1 h x i
= hn (1 − q) ,
1−q 1−q 1−q
1
using the Cauchy kernel formula with x = 1 − q and y = 1−q , and conclude using (2.8.11).
Exercice 2.13. (a) Show that plethysm of symmetric functions is associative.
(b) Show that, for any a and b,

a+b
(ha ◦ hb )[1 + q] = . (2.11.14)
b q
(c) Prove Hermite reciprocity, i.e.:
(ha ◦ hb )(x, y) = (hb ◦ ha )(x, y). (2.11.15)
(d) Show that, for all n

bn/2c
X
(h2 ◦ hn ) = s(2n−2k,2k) . (2.11.16)
k=0
(e) Show that, for all n

bn/2c
X
(hn ◦ h2 ) = sµ . (2.11.17)
µ`2n
µ has even parts
(f) (open) Foulkes conjecture, for any a < b,
(hb ◦ ha ) − (ha ◦ hb ) is Schur postive. (2.11.18)
37
(g) (open) Generalized Foulkes conjecture (due to Vessenes), for any a ≤ b, c ≤ d,
(hb ◦ hc ) − (ha ◦ hd ) is Schur postive. (2.11.19)
(g) Calculate sµ [1 + q], and conclude that (2.11.19) implies (1.5.5).
(h) Let n = ab. Show that hha ◦ hb , pab

1 i is the number of set partitions of a n-set into a blocks of
size b.
Exercice 2.14. In this exercise we consider plethysm for series like H(z; x) = H(z) (see (2.2.1)),
here underlining the role of x. We also write H for H(1), and observe that H(x; z) may be obtained
as the plethysm H[z x], thinking that z x = z · (x1 + x2 + . . .).
(a) Considering x as a (single) variable, so that pk [x] = xk , find closed form formulas (not series)
for H[x], P [x], and E[x].
(b) Let x and y be two sets of variables. Show that H[x + y] = H(x) H(y).
(c) Considering t as a constant, show that H[tx] = H(x)t .
(d) Considering x as a (single) variable, show that

Y 1
(H ◦ (H − 1))[x] = .
1 − xj
j≥1
(d) Consider any symmetric function series of the form

X
F (x) = p1 (x) + fd (x), with fd ∈ Λd .
d≥2
Can you prove that F affords an inverse for plethysm?
38
Chapter 3
Macdonald symmetric functions and

operators
For more background and details, see FB. [3] or Haglund [24]. The Sage Symmetric Function Tutorial
includes examples about Macdonald symmetric functions, and related operators.
3.1 Macdonald symmetric functions
The notions that we plan to explore are closely linked to the theory of Macdonald symmetric
functions, and operators for which they are joint eigenfunctions.
3.1.1 Macdonald’s original definition
Our context here is the ring Λqt of symmetric functions with coefficients in the fraction field C(q, t)
in the parameters, q and t, equipped with the scalar product

zλ (q, t) if λ = µ, h i
hpλ , pµ iq,t := where zλ (q, t) := zλ · pλ 1−q
1−t . (3.1.1)
0 otherwise,

In his original paper of 1988, Macdonald establishes the existence and uniqueness of symmetric
functions (polynomials) Pµ = Pµ (q, t; x) such that
X
(1) Pµ = mµ + γµλ (q, t) mλ , with coefficients γµλ (q, t) in C(q, t); and
λ≺µ
(2) hPλ , Pµ iq,t = 0, whenever λ 6= µ.
Since at q = t, the scalar product (3.1.1) specializes to the usual scalar product for symmetric
functions, we get Pµ (q, q; x) = sµ (x). The Hall-Littlewood symmetric functions are obtained by
39
setting q = 0, and the Jack symmetric functions by setting q = tα (α ∈ R and α > 0) and letting
t → 1. Zonal symmetric functions are just the special case when α = 2; and we have
Pµ (q, 1; x) = mµ , and Pµ (1, t; x) = eµ0 .
3.1.2 The “combinatorial” renormalization
We are interested here in the combinatorial renormalization of the Macdonald symmetric functions,
denoted by Hµ , as introduced by Garsia. One sets
Y
x
Hµ (q, t; x) := Pµ q, t−1 ; (q a(c) − t`(c)+1 ), (3.1.2)
1 − t c∈µ
where a(c) = aµ (c) and `(c) = `µ (c) denote respectively the arm length and the leg length of the
cell c of µ, see section 1.1. After a long process, it has been shown that the Schur expansion of the
functions Hµ (q, t; x) has coefficients in N[q, t]. We have already encountered Hn in (2.8.18). There is
an interesting combinatorial description of the Hµ given in [25]. Small explicit values are as follows:
have
H2 = s2 + q s11 ,
H11 = s2 + t s1 ;
H3 = s3 + q 2 + q s21 + q 3 s111 ,

H21 = s3 + (q + t) s21 + q t s111 ,

H111 = s3 + t2 + t s21 + t3 s111 ;

H4 = s4 + (q 3 + q 2 + q) s31 + (q 4 + q 2 ) s22 + (q 5 + q 4 + q 3 ) s211 + q 6 s1111 ,

H31 = s4 + (q 2 + q + t) s31 + (q 2 + q t) s22 + (q 3 + q 2 t + q t) s211 + q 3 t s1111 ,
H22 = s4 + (q t + q + t) s31 + (q 2 + t2 ) s22 + (q 2 t + q t2 + q t) s211 + q 2 t2 s1111 ,
H211 = s4 + (q + t + t2 ) s31 + (q t + t2 ) s22 + (q t + q t2 + t3 ) s211 + q t3 s1111 ,
H1111 = s4 + (t + t2 + t3 ) s31 + (t2 + t4 ) s22 + (t3 + t4 + t5 ) s211 + t6 s1111 .
P
The coefficients Kλ,µ (q, t) of the expansion Hµ (q, t; x) = λ`n Kλ,µ (q, t) sλ (x), are said to be
q, t-Kostka polynomials. We have the specializations
Kλ,µ (1, 1) = f µ , Kλ,µ (0, 1) = Kλ,µ (3.1.3)

Hµ (0, 0; x) = sn , Hµ (0, 1; x) = hµ , Hµ (1, 1; x) = sn1 , (3.1.4)
The next properties that stand out are the symmetries

0
Hµ0 (q, t; x) = Hµ (t, q; x), Hµ (q, t; x) = q n(µ ) tn(µ) ω(Hµ (1/q, 1/t; x)) , (3.1.5)
n(µ0 )
Kλ,µ (q, t) = Kλ,µ0 (t, q), Kλ,µ (q, t) = q tn(µ) Kλ0 ,µ (1/q, 1/t). (3.1.6)
40
These symmetries are made apparent in the q, t-Kostka matrix given here for n = 4:
1 q3 + q2 + q q4 + q2 q5 + q4 + q3 q6
 
1 q 2 + q + t q 2 + qt q 3 + q 2 t + qt q 3 t 
 
1 qt + q + t q 2 + t2 q 2 t + qt2 + qt t2 q 2  .
 
1 q + t2 + t qt + t2 qt2 + qt + t3 qt3 
1 t3 + t2 + t t4 + t2 t5 + t4 + t3 t6
Their specialization t = 1, are multiplicative (see exercises), i.e.:
Hµ (q, 1; x) = Hµ1 (q, 1; x) · · · Hµk (q, 1; x), (3.1.7)
and thus characterized by the simple functions Hn (q; x) = Hn (q, 1; x) = h∗n (x)/h∗n (1) (see for-
mula (2.8.18)). A second interesting specialization is at t = 1/q. We then get the following Schur
expansion Y
Hµ (q, 1/q; x) = q −n(µ) (1 − q a(c)+`(c)+1 ) s∗µ (x). (3.1.8)
c∈µ
Other formulas are obtained by specializing x (using plethystic substitutions). For instance, we have
Y
Hµ [q, t; 1 − u] = (1 − q i tj u), (3.1.9)
(i,j)∈µ
which implies that

X
hHµ , s(n−k,1k ) i = ek [Bµ − 1] where Bµ = Bµ (q, t) := q i tj . (3.1.10)
(i,j)∈µ
Since the cell (0, 0) always belongs to µ, the polynomial Bµ (q, t) − 1 lies in N[q, t]. In particular,

( k
) n−1
hHn , s(n−k,1k ) i = q 2 . (3.1.11)
k q
An analog of the Cauchy-formula for the Hµ is:

X
xy Hµ (x) Hµ (y)
en = (3.1.12)
(1 − q)(1 − t) wµ (q, t)
µ`n
where Y
wµ (q, t) := (q a(c) − t`(c)+1 )(t`(c) − q a(c)+1) .
c∈µ
In particular, X
x Hµ (x)
en = (3.1.13)
(1 − q)(1 − t) wµ (q, t)
µ`n
We may also characterize the Hµ as being the unique solution of the system of equations:
(i) hsλ (x), Hµ [q, t; (1 − q) x]i = 0, if λ 6 µ,

(ii) hsλ (x), Hµ [q, t; (1 − t) x]i = 0, if λ 6 µ0 , and (3.1.14)
(iii) hsn (x), Hµ (q, t; x)i = 1,
41
It is also useful to have the expansion
X Πµ B µ Y
en (x) = (1 − q)(1 − t) Hµ (x), with Πµ := (1 − q i tj ); (3.1.15)
µ
hµ h0µ (i,j)∈µ,
(i,j)6=(0,0)
and
X Πµ
(−1)n−1 pn (x) = (1 − tn )(1 − q n ) Hµ (x). (3.1.16)
µ
hµ h0µ
3.1.3 Macdonald scalar product
The ?-scalar product is defined on the power-sum basis by

(
Zµ (q, t) if µ = λ,
hpµ , pλ i? :=
0 otherwise,
where Zµ (q, t) is defined by
Zµ (q, t) := (−1)|µ|−`(µ) pµ [(1 − q) (1 − t)] zµ .
Observe that our two scalar products are linked by the relation (to be checked on power-sums)
hf, gi = hf, ω g ? i? ,
where
? x
g (x) := g .
(1 − t)(1 − q)
The Macdonald polynomials form an orthogonal family for this scalar product. More precisely, we
have
hHµ , Hλ i? = 0, whenever λ= 6 µ.
3.2 Macdonald eigenoperators
Many operators that we will consider share the Macdonald functions Hµ has eigenfunctions. We
have taken the habit of calling them Macdonald eigenoperators to underline this fact. Thus, these
operators are entirely characterized by their eigenvalues for the Hµ . The first such, here denoted
by ∆, was essentially introduced by Macdonald himself (up to a reformulation in terms of the
renormalized Hµ ), and it has eigenvalues as follows
∆(Hµ ) := Bµ Hµ . (3.2.1)
i tj
P
Observe that all these eigenvalues are different (since Bµ := (i,j)∈µ q characterizes µ). It may be
shown that
n
X q k − tk
∆(en ) = en−k ek . (3.2.2)
q−t
k=1
42
3.2.1 The Nabla operator
Next, we considered (the author and A. Garsia, see [6, 7]) the operator ∇, such that
0
∇(Hµ ) := q n(µ ) tn(µ) Hµ . (3.2.3)
0
Its eigenvalues may also be described as q n(µ ) tn(µ) = (i,j)∈µ q i tj . It has been shown by Haiman
Q
(see [28]) that ∇(en ) is Schur positive, and that it corresponds to the bigraded Frobenius of the
Sn -module of diagonal harmonic polynomials. Powers ∇r , for r in N, also have nice representation
theory interpretations. It is interesting that we have

r 1 (r + 1)n
h∇ (en ), en iq=t=1 = , h∇r (en ), en1 iq=t=1 = (r n + 1)n−1 , (3.2.4)
rn+1 n
n
q −r( 2 ) (r + 1)n

n
r
h∇ (en ), en it=1/q = , h∇r (en ), en1 it=1/q = q −r( 2 ) [r n + 1]n−1
q . (3.2.5)
[r n + 1]q n q
These formulas come from the two following more global ones
1
∇r (en )q=t=1 = en [(r n + 1)x], and (3.2.6)
rn+1
n 1
∇r (en )t=1/q = q −r( 2 )

en [r n + 1]q x . (3.2.7)
[r n + 1]q
We write ∇ e for the specialization of ∇ at t = 1. The operator ∇ e is multiplicative (see exercise),

∗ 0) ∗
and it is characterized by the fact that ∇hµ (x) = q
e n(µ hµ (x), since h∗µ (x) is proportional to
Hµ (q; 1, x). The specialization of ∇ at t = 1/q is denoted by ∇, b and it is characterized by the fact
∗ n(µ 0 )−n(µ) ∗ ∗
that ∇s
b (x) = q
µ sµ (x), since sµ (x) is proportional to Hµ (q; 1/q, x).
It may be shown that (see definition 1.5.2 of δn )
X
∇(e
e n) = q area(µ) s(µ+1n )/µ (x), (3.2.8)
µ⊆δn
n

where area(µ) is defined as 2 − |µ|. One may understand this in terms of q-Lagrange inversion.
More explicitly, the series
X
F(q, z; x) = e n ) z n+1 ,
∇(e
n≥0
satisfies the functional equation
F = z E ◦q F, with E(z) = 1 + e1 z + e2 z 2 + . . . (3.2.9)
involving the q-composition1

X X
f ◦q g := fn g(z) g(qz) · · · g(q n−1 z), when f (z) = fn z n .
n≥0 n≥0
1
Careful, this operation is not associative See [21, 22].
43
From this we may deduce (see also Flajolet [18, page 330, formula (74)]) that generating series
X
F (q; z) = e n ), en i z n ,
h∇(e
n≥0
is given by the formula

eq (z) X (−z)n q n2
F (q; z) = , where eq (z) := . (3.2.10)
eq (z/q) (1 − q)n [n]q !
n≥0
Beside these interesting facts, our original experimental observation about ∇ was the following. Let
us write −1 ι(µ)
sbµ := sµ ,
qt
where ι(µ) is equal to the sum of the values µi − i which are positive. Then, we have
Conjecture 3.2.1 (FB. and A. Garsia 1999). For all µ, the symmetric function ∇(b
sµ ) is Schur
positive.
This is still widely open. Observe that we can readily calculate ∇(b
e sµ ) using Jacobi-Trudi.
3.2.2 The ∆f -operators
In general, for any symmetric function f , we set

∆f (Hµ ) := f [Bµ ]Hµ ,
It is easy to see that, on homogeneous symmetric function of degree n, we have2 ∇ := ∆en . It is
also clear that
∆f +g = ∆f + ∆g , and ∆f g = ∆f ◦ ∆g .
It is known that
∇(πn ) = ∆en−1 en , and πn = ∆e1 b
hn , (3.2.11)
with bhn = sb(n) , and thus stands for (−1/q t)n−1 hn . Many relevant operator identities may be found
in [7], see in particular those listed in (I.12) loc. cit. These imply that
n
X
e⊥
1 ∇(en ) =∇ sk (q, t) ek en−k , and e⊥
1 ∇(hn ) = ∇(πn ),
b (3.2.12)
k=0
as well as
(−1)n−1 M ∆en−1 (pn [x/M ]) = ∇(b
hn ),
where M := (1 − q)(1 − t). We also consider D0 the Macdonald eigenoperator whose eigenvalue for
Hµ is X
1 − (1 − q)(1 − t)Bµ = 1 − (1 − t) (1 − q) q i tj . (3.2.13)
(i,j)∈µ
The Delta Conjecture (see [27]) proposes an explicit combinatorial formula for ∆ek (en ). See
exercises.
2
Careful though: this is not so for other symmetric functions.
44
Exercice 3.1. (a) Show that

h∗µ (x)
Hµ (q, 1; x) = . (3.3.1)
h∗µ (1)
(b) Conclude that

0
X
∇(e
e n) = fµ [1 − q] q n(µ ) h∗µ (x). (3.3.2)
µ`n
(c) Show that

s∗µ (x)
Hµ (q, 1/q; x) = ∗ . (3.3.3)
sµ (1)
(d) Conclude that

n
b n) = q − 1
X
∇(e (−1)n−k q (n+1)(2k−n)/2 s∗(k,1n−k ) (x). (3.3.4)
q
k=1
Exercice 3.2. The riser sequence ρ(µ) = (r1 , . . . , rk ), of a partition µ siting inside δn , is defined
by setting
ri := µ0i−1 − µ0i , (with µ00 := n).
These are the column height differences in µ. See that eρ(µ) (x) = s(µ+1n )/µ (x).
Exercice 3.3. Recall that the Macdonald polynomial Hn does not depend on t (see (2.8.18)).
(a) Show that Hb ◦ Ha − Ha ◦ Hb is divisible by 1 − q. To this end, calculate (Hb ◦ Ha )q=1 .
(b) Calculate the limit, as q → 1, of
Hb ◦ Ha − Ha ◦ Hb
,
1−q
and prove that it is Schur positive when b > a.
(c) Show that

Hb ◦ Ha − Ha ◦ Hb
lim = hb ◦ ha − ha ◦ hb .
q→0 1−q
(d) (Conjecture FB., 2017) Show that (Hb ◦ Ha − Ha ◦ Hb )/(1 − q) is Schur positive.
Exercice 3.4. (a) Check that

X µi
0
n(µ ) = , (3.3.5)
2
i
and conclude that ∇

e is multiplicative.
(b) Give a determinantal formula for ∇(be sµ ) (of a matrix with entries of the form ∇(e
e n )), and for
e sµ ), en i (of a matrix with entries of the form Cn (q)).
h∇(b
45
Exercice 3.5. (a) Show that the function F (q; z) in (3.2.10) satisfies the difference equation
1
F (q; z/q) = (3.3.6)
1 − z F (q; z)
e n ), en i = Cn (q) (see equation (1.5.3)).

(b) Deduce from this equality that h∇(e
(c) Compare the above with the specialization of (3.2.9) at q = 1 when one takes the scalar product
with E = E(1).
(d) Make sense of the statement:
“The series n≥0 ∇(en )q=t=1 z n+1 is the generic Lagrange inverse.”
P
by exploiting the fact that the en are algebraically independent (so that they may be specialized
as one wishes). See [35] for more on this.
Exercice 3.6. (a) Calculate ∆e1 (en ), and conclude that

P
1  k ek+j+1 ek−j , if n = 2k + 1,
j=0
∆e1 (en ) = (3.3.7)
n q=t=1 e /2 + Pk e e , if n = 2k.
kk j=1 k+j k−j
(b) (explore) Show that

1 X 1
∆e (en ) = eµ (3.3.8)
(n)k k q=t=1
µ`n
d0 !d1 ! · · · dn !
`(mu)≤k
where µ is packed with 0’s to make it of length k, and di = di (µ) is the multiplicity of i in µ
(including the 0 parts). We use here the notation (n)k = n(n − 1) · · · (n − k + 1).
46
Chapter 4
Rectangular combinatorics, and

symmetric functions
In recent years, a lot of attention has been given to “Rational” Catalan combinatorics (see [1])
associated to pairs (a, b) of co-prime integers, with a and b the proportions of a (m, n)-stairshape
partition (µ ⊆ δmn ), inside which one consider sub-partitions (aka (a, b)-Dyck paths). More recently,
this has been expanded (see [2]) to “Rectangular” Catalan combinatorics (removing the co-primality
condition), considering any pair (m, n) in N × N. The greatest common divisor d = gcd(m, n) playing
a special role here, it is natural to think that (m, n) = (ad, bd), so that the two approaches are linked
together.
This is closely tied to the study of (m, n)-indexed symmetric functions that arise from “creation”
operators1 on symmetric functions. These symmetric functions play a fundamental role in many
areas of recent research, such as: Algebraic Combinatorics (Delta Conjecture [27]), Representation
Theory (graded modules of bivariate harmonic polynomials [2]), Knot Theory (Khovanov-Rozansky
homology of (m, n)-torus links [29]), Algebraic Geometry (Hilbert scheme of points in the plane [28]),
Theoretical Physics (quantum field theory [16, 42]), and more.
4.1 Rectangular Catalan combinatorics
In preparation for our upcoming extensions of “Macdonald operator” theory, we consider the (m × n)-
rectangular extension of the notions of “Dyck paths” and “parking functions”. In our context, these
correspond to sub-partitions µ of the (m, n)-staircase shape δmn , and standard tableaux of skew
shape (µ + 1n )/µ. The (m, n)-staircase shape, denoted by δmn , (see figure 4.1) is set to be
δmn := r1 r2 · · · rn , with rk := bm (n − k)/nc. (4.1.1)
More precisely, we define the (m, n)-Dyck path associated to sub-partition µ of δmn , as the path
that goes from (0, n) to (m, 0) with either south or east steps following the grid N × N, following the
1
In Theoretical Physics, this is a classical notion which typically builds an object out of the empty one (for us this
will be the symmetric function s0 = 1), thus the name.
47
Figure 4.1: The (9, 7)-staircase: 765321, and the (4, 9)-staircase: 332211.
boundary of µ. In other terms, the cells of µ are precisely those that lie bellow the Dyck path. We
denote by µ both notions. The area of a (m, n)-Dyck path µ is set to be
area(µ) := |δmn | − |µ|, (4.1.2)
hence, it is the number of cells that lie between the path µ and the path δmn . Further along, we will
need to encode δmn as a monomial in the variables z = z0 , z1 , z2 , . . . , zm−1 , setting
zmn := zr1 zr2 · · · zrn , for δmn = r1 r2 · · · rn ,
where we are careful to include 0 parts (observe that the largest possible value of the ri ’s is m − 1).
Given a (m, n)-Dyck path µ, a (m, n)-parking function π on µ is a bijective labeling of the n
vertical steps in the path µ, by the elements of the set {1, 2, . . . , n}, such that consecutive vertical
steps have increasing labels reading bottom up. This is clearly the same as a standard tableau
π : (µ + 1n )/µ → {1, 2, . . . , n},
of shape (µ + 1n )/µ, as is illustrated in figure 4.2.
5
4
1
3
7
6
2
Figure 4.2: The (9, 7)-parking function 0420043 on 4432.
Equivalently, a (m, n)-parking function π on µ corresponds to a permutation π1 π2 · · · πn of the rows

of µ (packed with 0’s so that it be of length n), with πi equal to the number of cells (of µ) to the
left of cells in which i sits. We will also say that the parking function π is a µ-parking function.
We will respectively denote by Dyckmn , Parkµ , and by Parkmn , the sets of (m, n)-Dyck paths,
µ-parking functions, and the (m, n)-parking functions. The cardinalities of these sets are respectively
denoted by Catmn , Pµ , and by Pmn ; and the q-enumeration by area of Dyck paths is set to be
X
Catmn (q) := q area(µ) . (4.1.3)
µ⊆δmn
48
For instance, we have
Cat10 (q) = 1,
Cat21 (q) = 1,
Cat32 (q) = 1 + q,
Cat43 (q) = 1 + 2 q + q 2 + q 3 ,
Cat54 (q) = 1 + 3 q + 3 q 2 + 3 q 3 + 2 q 4 + q 5 + q 6 ,
Cat65 (q) = 1 + 4 q + 6 q 2 + 7 q 3 + 7 q 4 + 5 q 5 + 5 q 6 + 3 q 7 + 2 q 8 + q 9 + q 10 .
4.2 Counting (m, n)-Dyck paths and (m, n)-parking functions
Our exposition here will be made clearer if we consider that (m, n) = (ad, bd) with a and b co-prime
(relatively prime), so that d is equal to gcd(m, n).
4.2.1 The co-prime case
To start with, the enumeration of (a, b)-Dyck path (and parking functions) is much easier than the
more general case. In a nutshell the number Catab of (a, b)-Dyck path is simply given by the formula

1 a+b
Catab = . (4.2.1)
a+b a
This may be obtained by a “classical” cyclic argument (see exercises), which appears to be due to
Dvoretzky-Motzkin (see [17]), or even earlier to Lukasiewicz. It follows from the same argument
that the number Pab of (a, b)-parking functions is given by the formula
Pab = ab−1 . (4.2.2)
Both of the above formulas may be deduced from (any) of the following (equivalent) more general
formulas (see exercises).
X 1
(a) s(µ+1n )/µ (x) = eb [a x],
a
µ⊆δa,b
1 X `(λ) (−1)b−`(λ)
(b) = a pλ (x),
a zλ
λ`b

1 XY a
(c) = mλ (x), (4.2.3)
a k
λ`b k∈µ
1 X 0
(d) = sλ (a) sλ (x),
a
λ`b
X (a + 1) · · · (a + `(λ) − 1)
(e) = (−1)`(λ)−1 hλ (x),
d1 (λ)! · · · dn (λ)!
λ`n
where di (λ) is the number of parts of size i in λ in this last expression.
49
4.2.2 Bizley’s formula
It is a bit harder is to figure out the formula without the co-prime condition. Although this was
entirely done in the 1950’s by Grossmann and Bizley (see [11]), it remained largely unknown2 by the
community until the early 2000’s. One finds the following values
n\m 1 2 3 4 5 6 7 8 9
1 1 1 1 1 1 1 1 1 1
2 1 2 2 3 3 4 4 5 5
3 1 2 5 5 7 12 12 15 22
4 1 3 5 14 14 23 30 55 55
5 1 3 7 14 42 42 66 99 143
6 1 4 12 23 42 132 132 227 377
7 1 4 12 30 66 132 429 429 715
Table 4.1: Number of (m, n)-Dyck paths.
The general formulas is

X 1 Y 1 ak + bk
Catmn = , (4.2.4)
zµ a+b ak
µ`d k∈µ
with the product over the parts of µ. In other words, for each fixed a and b co-prime, one has the
generating function
 
∞
X X 1 aj + bj
Cat(ad,bd) xd = exp xj /j  . (4.2.5)
a+b aj
d=1 j≥1
For (m, n)-parking function enumeration, the analogous formula is
X 1 n Y 1
Pmn = (ka)kb−1 . (4.2.6)
zµ µ a
µ`d k∈µ
All the above formulas follow from

X X 1 Y1
s(µ+1n )/µ (x) = ekb [ka x]. (4.2.7)
zµ a
µ⊆δmn µ`d k∈µ
2
Maybe this is in part because the paper was published in a journal for actuaries.
50
n\m 1 2 3 4 5 6 7 8
1 1 1 1 1 1 1 1 1
2 1 3 3 5 5 7 7 9
3 1 4 16 16 25 49 49 64
4 1 11 27 125 125 243 343 729
5 1 16 81 256 1296 1296 2401 4096
6 1 42 378 1184 3125 16807 16807 35328
7 1 64 729 4096 15625 46656 262144 262144
Table 4.2: Number of (m, n)-parking functions.
The left hand side may be modified in several ways, to account for interesting families of partitions.
For instance, we have a similar formula
X X (−1)d−`(µ) Y 1
s(µ+1n )/µ (x) = (−1)n−1 ekb [ka x], (4.2.8)
0
zµ a
µ⊆δmn µ`d k∈µ
0
where µ ⊆ δmn corresponds to restricting the summation to “diagonal avoiding” partitions µ.
4.2.3 Constant term formula
Let us now set

m
Y 1
Ω(z) := .
1 − zk
k=0
We get the q-enumeration of (m, n)-Dyck path as follows (see exercises).
Proposition 4.2.1. For any m and n, we have the constant term formula
m
!
Ω(z) Y 1
Catmn (q) = , (4.2.9)
zmn 1 − q zi+1 /zi z0
i=0
where (−)|z0 means that we take the constant term with respect to the variables zk .
4.2.4 Cell rank
For our upcoming constructions, we consider the notion of (m, n)-rank on cells (x, y) in N × N:
ρmn (x, y) := mn − nx − my,
51
and outline some of its properties. Clearly, ρm,n (x, y) = 0 if and only if (x, y) sits on the line of
equation nx + my = mn, which crosses the x-axis at m, and positive rank cells sit below this line.
For example, Figure 4.3 presents the (7, 5)-ranks for cells in the first quadrant. When a and b are
.. .. .. .. .. .. .. ..
. . . . . . . .
0 −5 −10 −15 −20 −25 −30 −35 · · ·
7 2 −3 −8 −13 −18 −23 −28 · · ·
14 9 4 −1 −6 −11 −16 −21 · · ·
21 16 11 6 1 −4 −9 −14 · · ·
28 23 18 13 8 3 −2 −7 · · ·
35 30 25 20 15 10 5 0 ···
Figure 4.3: Example of (m, n)-rank (m = 7 and n = 5).
co-prime, all the cells of the (a × b)-rectangle have different (a, b)-ranks. We can thus order them
(without ambiguity) in decreasing values of their (a, b)-rank.
4.3 Elliptic Hall algebra operators
For any given degree d symmetric function g(x), that we call a seed, we construct in this section
a family of symmetric function gmn (q, t; x) with (m, n) = (ad, bd), for any a and b co-prime as in
the previous section. For the respective seeds ed and b hd := (−1/(qt))d−1 hd , we respectively get
symmetric functions emn and hb mn that are such that
X X
emn (1, 1; x) = s(µ+1n )/µ (x), and h(1,
b 1; x) = s(µ+1n )/µ (x).
µ⊆δmn 0
µ⊆δmn
The sum in the second formula is restricted to “diagonal avoiding” partitions µ, like at the end of
the previous section. As we will see, the occurence of the parameters q and t turn these symmetric
functions in a much refined enumeration of (m, n)-parking functions. In this, the parameter q
accounts for the area, and the parameter t for the number of “diagonal inversions”.
1. Our first step is to calculate the coefficients cµ = cµ (q, t) of the expansion of g(x):
X
g(x) = cµ πµ (x) (4.3.1)
µ`d
in the multiplicative basis πµ (x) = πµ (q, t; x) (see definition (2.11.2)).

2. Next, for any co-prime pair (a, b) in N × N, having constructed (in the next section) raising
degree operators Q(a,b)µ on symmetric functions, we set
X
gmn (q, t; x) := cµ Q(a,b)µ (1), where (m, n) = (ad, bd). (4.3.2)
µ`d
52
The operator Q(a,b)µ raises degrees by n. As it is here applied to the symmetric function “1”,
we get deg(gmn ) = n.
We thus have a process

gd → {gmn }(m,n)=(ad,bd) ,
that maps a symmetric function g = gd in Λd , to a family {gmn } indexed by pairs of the form
(ad, bd), with a and b co-prime.
Among of the interesting identities in this setup (see [2]) we have the operator equality
∇Q(a,b)µ ∇−1 = Q(a+b,b)µ . (4.3.3)
Moreover, Q(0,1)µ (1) = πµ (x). It follows that
g0d (x) = gd (x), and ∇(gmn ) = gm+n,n , (4.3.4)
hence grd,d = ∇r (gd ). We have thus generalized to the (m, n)-rectangular context and to other seeds,
our previous “squarre” situation corresponding to the study of ∇(ed ), hence the seed ed .
4.4 Construction of the operators
Our calculations are going to be done inside an operator algebra realization of the elliptic Hall
algebra (see [2]), which is graded3 over N × N. We will not actually need to describe the whole
algebra, but use (without further details) the fact that collinearly graded operators commute, as
well as property (4.3.3). In fact, we only work with the portion of the algebra that is algebraically
generated by the two operators:
1. multiplication by p1 (x), and
2. the D0 Macdonald eigenoperator, see (3.2.13).
To each pair (m, n) in N × N, we recursively associate a bracketings of these operators, via the
following splitting procedure of (m, n). One may show (see exercise) that there are precisely d
positive integer entries matrices such that

r s
det = gcd(m, n), and (m, n) = (r, s) + (u, v). (4.4.1)
u v
One then recursively sets

1
Qmn := [Qrs , Quv ], with Q01 := p1 , and Q10 := D0 ; (4.4.2)
(1 − q)(1 − t)
and then finally
Q(a,b)µ := Q(aµ1 ,bµ1 ) · · · Q(aµk ,bµk ) , for µ = µ1 · · · µk ` d. (4.4.3)

3
In fact it is graded over Z × Z, but we will only use the “positive” part.
53
This is a composition of operators having collinear gradings, hence they commute (making the
definition independent of a choice of order). From general properties of the elliptic Hall algebra, it
may also be shown that this construction does not depend on the choice of a solution for (4.4.1).
For example, writing M := (1 − q)(1 − t) as before, we have
1
Q43 = [[p1 , D0 ], [[p1 , D0 ], [[p1 , D0 ], D0 ]]],
M6
and
1
Q63 = [[p1 , D0 ], [[[p1 , D0 ], D0 ], [[[p1 , D0 ], D0 ], D0 ]]].
M8
4.5 Interesting seeds
There are many seeds gd , for which the corresponding gmn ’s all appear to be Schur positive. Many
are also tied to interesting combinatorics. We consider here the three cases ed , b
hd , and πd , with the
respective families denoted by emn , h
b mn , and πmn . In view of (4.3.4), we have
ern,n = ∇r (en ), b rn,n = ∇r (b

h hn ), and πrn,n = ∇r (πn ).
It may be seen that, for all r and n,
ern+1,n = ern,n , and ern−1,n = h

b rn,n . (4.5.1)
Observe that in the co-prime situation, one has eab = h

b ab = πab , illustrating that the more interesting
situation is when co-primality does not hold. Part of the conjectures in [8] is to propose the following
explicit combinatorial formula
X X
emn (q, t; x) = q area(µ) tdinv(τ ) xτ , (4.5.2)
µ⊆δmn τ ∈Pµ
where Pµ stands for the set of semi-standard tableaux of skew-shape (µ + 1n )/µ (recall that µ + λ
denotes the partition obtained by pointwise addition of parts), having values in N. Broadly speaking,
dinv is a statistics that depends (in some manner not described here) on how entries of cells of τ
compare, taking into account some “rank function” for cells that depends on m and n. For more on
“dinv”, see [26]. This relates to the “sweep” map considered in [44].
4.5.1 Specializations at t = 1
It appears (although this is still to be proven, see [2]) that, for any symmetric function g(x)
independent of t, one has
Q(a,b)µ (g(x)) t=1

= Q(a,b)µ (1) t=1
· g(x). (4.5.3)
In other words, at t = 1, the effect of the operator Q(a,b)µ t=1 on any g(x) is to multiply it by the
fixed symmetric function Y
π(a,b)µ (q, 1; x) = π(ak,bk) (q, 1; x),
k∈µ
54
reducing all calculations to that of polynomials in the πmn (q, 1; x)’s. In fact, this “multiplicativity”
property holds for any other multiplicative basis. For instance, we have that
X X
g(x) = cµ eµ (x) implies gmn (q, 1; x) = cµ e(a,b)µ (q, 1; x),
µ`d µ`d
Q
for any co-prime (a, b), and e(a,b)µ (q, 1; x) = k∈µ e(ak,bk) (q, 1; x). The individual emn (q, 1; x) afford
a combinatorial description (see [8]) of the form
X
emn (q, 1; x) = q area(µ) s(µ+1n )/µ (x). (4.5.4)
µ⊆δmn
4.5.2 Specializations at t = 1/q
It is also interesting to recall (see [2]) that

[d]q h i
q α(m,n) πmn (q, 1/q; x) = en [m]q x , (4.5.5)
[m]q
with d = gcd(m, n), and where α(m, n) = ((n − 1)(m − 1) + d − 1)/2. Further setting q = 1, we find
that
d
πm,n (1, 1; x) = en [m x],
m
where one considers m, r and n as a constant for the calculation of this plethysm. It follows
from (4.5.1) that
1 1
ern,n (1, 1; x) = en [(r n + 1) x], and h
b rn,n (1, 1; x) = en [(r n − 1) x].
(r n + 1) (r n − 1)
These formulas imply that

1 (r + 1) n
hern,n (1, 1; x), en1 i = (r n + 1) n−1
, hern,n (1, 1; x)), en i = ,
rn + 1 n

b rn,n (1, 1; x)), en i = (r n − 1)n−1 , 1 (r + 1) n − 2
hh 1 hhrn,n (1, 1; x)), en i =
b ,
rn − 1 n

d m+n
hπmn (1, 1; x)), en1 i = d mn−1 , hπm,n (1, 1; x)), en i = .
m+n n
Moreover, for m and n co-prime we have
(−1)n−1 q (m+1)(n−1)/2
πmn (q, 1/q; x) = (∆em pn (x))t=1/q . (4.5.6)
[n]q em [[n]q ]
4.6 What comes next
The next step is to expand the previous considerations to 3 parameters. This leads to the study
of intervals in the Tamari lattice, or its extension to the (m, n)-context. There are several ways
55
to define the Tamari order4 on (classical m = n) Dyck paths. The simplest may be to consider
these as well parenthesized expressions, with ()() · · · () as smallest element, and associativity
used for covering relation. Thus the largest element is (((· · · ))). See next subsection for an explicit
description in terms of partitions.
There are also many (well-known) ways of constructing a convex polytope, called the associahedron5
whose vertices an edges correspond to the Hasse diagram of the Tamari lattice. The Tamari order
corresponds to an orientation of this polytope, that may be specified by choosing a minimal vertex
(why?). An example with m = n = 4 corresponds to figure 4.4.
0000
1100
1000
1110
2110 2000 2210 2200
2100
3000
3100
3110 3200
3210
Figure 4.4: The (4, 4)-Tamari poset for n = 4, bottom element is 3210.
Similarly, the (m, n)-Tamari lattice, Tmn , corresponds to an order on the set of (m, n)-Dyck paths,
with the m = n case corresponding to the classical Tamari lattice. Examples include the (8, 4)-Tamari
lattice of figure 4.5 (a bit “exploded” here to better display its structure). Its smallest element is
the one at the forefront (the one with only blue faces adjacent). The blue portion of this figure
is a copy of 4.4. Increasing order corresponds to following shortest paths away from the smallest
node. Likewise, we have the (4, 6)-Tamari lattice in figure 4.6. Nice geometrical realizations and
combinatorial properties of these posets are studied in [10].
For m equal to rn + 1, the following formulas (see [9, 13, 12]) account for the number of intervals,
pairs (ν, µ) with ν µ in the (m, n)-Tamari lattice, and the number of pairs (ν, π) with π a parking
function on µ:
(r + 1)2 n + r

X (r + 1)
1= , (4.6.1)
n (rn + 1) n−1
νµ ∈Trn+1,n

X n
= (r + 1)n (r n + 1)n−2 , (4.6.2)
ρ(µ)
νµ ∈Trn+1,n
where both sums are over pairs ν µ for ν and µ in the (rn + 1, n)-Tamari lattice, and where
ρ(µ) := (c1 , c2 , . . . , ck ) is the sequence of column heights in the skew-shape (µ + 1n )/µ. Both
4
See for instance https://en.wikipedia.org/wiki/Tamari_lattice.
5
See https://en.wikipedia.org/wiki/Associahedron.
56
Figure 4.5: The (8, 4)-Tamari lattice, seen from the bottom.
211000 210000
200000
111000 110000 100000
000000
311000
211100
221000 220000 111100
321000
221100 311100
222000
321100
222100
322000
322100
Figure 4.6: The (4, 6)-Tamari lattice, bottom element is 322100, and top is 000000.
identities come out of the symmetric function formula

X X (r+1) k pλ (x)
(−1)n−`(λ) (r n + 1)`(λ)−2
Q
s(µ+1n )/µ (x) = k∈λ k . (4.6.3)
zλ
νµ ∈Trn+1,n λ`n
There is a similar formula for m = rn − 1, namely

X X k (r+1)−1 pλ (x)
(−1)n−`(λ) ((r + 1) n − 1)`(λ)−2 k∈λ
Q
s(µ+1n )/µ (x) = k . (4.6.4)
zλ
νµ ∈Trn−1,n λ`n
These formulas extend in a natural manner (explained further below) formulas (4.2.7) and (4.2.7)
(see also (4.2.3)). They also afford extensions with three parameter q, t, and r, such that the case
r = 0 corresponds to the emn (q, t; x) of this chapter. In particular, formulas (4.6.3) and (4.6.4) are
related in this way to (4.5.1).
The number of intervals in the (m, n)-Tamari lattices are given in table 4.3, for m ≤ n ≤ 7.
57
n\m 1 2 3 4 5 6 7
1 1
2 1 3
3 1 3 13
4 1 6 13 68
5 1 6 23 68 399
6 1 10 58 161 399 2530
7 1 10 58 248 866 2530 16965
Table 4.3: Number of intervals in the (m, n)-Tamari lattice.
Similarly table 4.4 gives the number of pairs (ν, π) (decorated intervals), with π a parking function
on µ ν.
n\m 1 2 3 4 5 6 7
1 1
2 1 4
3 1 5 32
4 1 17 49 400
5 1 23 167 729 6912
6 1 72 1048 4407 14641 153664
7 1 102 1818 15626 90079 371293 4194304
Table 4.4: Number of decorated intervals in the (m, n)-Tamari lattice.
4.7 Conclusion
There is a global algebraic point of view (see [4]) that unifies most of what we have seen at play
in these notes, and ties it to the study of interesting representation theory problems, as well as
many other subjects. The formulas in this broader approach take the form of linear combinations
Emn (q; x), with positive integer coefficients, of products sλ (q)sµ (x), with q = (q1 , q2 , · · · , qk ); and
the various specific cases occurring in these notes are obtained by specialization of the qi . We
illustrate with an example, with m = n = 4.
58
Let us set
E44 (q; x) := s4 (x)

+ (s1 (q) + s2 (q) + s3 (q)) s31 (x)
+ (s2 (q) + s4 (q) + s11 (q)) s22 (x) (4.7.1)
+ (s3 (q) + s4 (q) + s5 (q) + s11 (q) + s21 (q) + s31 (q)) s211 (x)
+ (s6 (q) + s41 (q) + s31 (q) + s111 (q)) s1111 (x)
Then, we have
E44 (q; x) = Hn (q, x), E44 (q, t; x) = ∇(e4 ) = e4,4 (q, t; x),
and X
E44 (1, 1, 1; x) = s(µ+1n )/µ (x).
νµ ∈T4,4
In each of these specializations, some of the Schur functions sλ (q) evaluate to 0, simply because
there are not enough variables (the number is less than `(λ)). Otherwise they all are essentially the
“same” formula. It is interesting to observe that setting all the qi = 1 gives the formula

E44 (k; x) = s4 + k3 + 3 k2 + 3 k1 s31

+ k4 + 5 k3 + 6 k2 + 2 k1 s22

+ k5 + 8 k4 + 18 k3 + 15 k2 + 3 k1 s211

+ k6 + 9 k5 + 25 k4 + 29 k3 + 12 k2 + k1 s1111 ,

with the convention that (or thinking of this as a plethysm with k constant):
E44 (k; x) := E44 (1, 1, · · · , 1; x).

| {z }
k copies
In particular, we find that the formulas
hErn,n (1; x), en i = 1, hErn,n (1; x), pn1 i = n!,

(r+1)n
hErn,n (2; x), en i = 1
hErn,n (2; x), pn1 i = (rn + 1)n−1 ,

rn+1 n ,
(r+1) (r+1)2 n+r
hErn,n (3; x), en i = hErn,n (3; x), pn1 i = (r + 1)n (rn + 1)n−2 ,

n (rn+1) n−1 ,
come out of the same “molds”. For instance, these are respectively
hE4,4 (k; x), e4 i = k1 + 12 k2 + 29 k3 + 25 k4 + 9 k k

5 + 6 ,
and
k k k k k k
hE4,4 (k; x), p41 i = 1 + 23

1 + 78 2 + 96 3 + 51 4 + 12 5 + 6 .
The expressions Emn (q; x) have many interesting structural properties and hidden symmetries that
are currently being studied.
59
Exercice 4.1. (a) Prove that Catrn+1,n = Catrn,n and Prn+1,n = Prn,n by a simple combinatorial
argument.
(b) Prove formula 4.2.1 via the “cyclic approach”. For this, you encode a Dyck path as a sequence
of a + b letters either equal to “s” (coding a “south” step in the path) or “e” (coding an “east”
step in the path). This sequence is periodically repeated “infinitely” in both directions. See
that there are precisely (a + b) different length (a + b) subsequences of consecutive letters in
this periodic infinite sequence. Each of these correspond to a path in the a × b-rectangle. Show
that there is a unique one that stays below the diagonal.
(c) Suitably adapt your previous argument to prove formula 4.2.2. You may be inspired by
figure 4.7.
2
3
5
7
1
4
6
8
2
3
5
7
1
4
6
8
Figure 4.7: A parking function “periodically” extended.
(d) For a (m, n)-Dyck path µ, let ρ(µ) := (c1 , c2 , . . . , ck ) be the sequence of column heights in
the skew-shape (µ + 1n )/µ. Show that the number of (m, n)-parking function is given by the
combinatorial formula
X n
Pmn = , (4.8.1)
ρ(µ)
µ⊆δmn
with the usual multinomial notation

n n n!
= = .
ρ(µ) c1 , c2 , . . . , ck c1 ! c2 ! · · · ck !
(e) Extend this argument to semi-standard fillings of the shape (µ + 1n )/µ, to deduce that
X 1
s(µ+1n )/µ (x) = eb [a x] (4.8.2)
a
µ⊆δa,b
prove that all expressions in (4.2.3) are equivalent.
60
Exercice 4.2. (a) Prove proposition 4.2.1. You may start your reflexion as follows. Associate
to cells (i, j) the weight q zi /zi−1 , and globally weight a partition µ with the product of the
weights of it cells. Here, one sets z−1 := 1. Find how one may get a constant term out of the
expression, and link this to partitions that lie inside δmn .
(b) Formulate a similar statement for partitions lying inside any fixed partition.
(c) Assuming (4.5.4), prove the constant term formula (see [40])
m−1
!
E[(z0 + . . . + zm ) x ] Y 1
emn (q, 1; x) := (4.8.3)
zmn 1 − q zi /zi+1 z0
i=1
Exercice 4.3. (a) Show that

qt x(qt − 1)
πn = en . (4.8.4)
qt − 1 qt
(b) Use this to find that the dual basis of πµ is given be the formula 6
((qt − 1)/qt)`(µ) fµ [x qt/(qt − 1)]. (4.8.5)
(c) Show that

X h qt i h x(qt − 1) i
en (x) = mµ eµ ,
qt − 1 qt
µ`n
from which you may deduce the π-basis expansion of en .
(d) Find a plethystic formula for the π-expansion of hn .

Exercice 4.4. Show that, at t = 1, formula (4.5.2) specializes to formula (4.5.4).
Exercice 4.5. (a) Show that the solutions of (4.4.1) may be indexed by the integer points that are
closest (below) to the “diagonal” (from (0, 0) to (m, n)) in the m × n rectangle. See figure 4.8.
(3,5)
(2,3)
(0,0)
Figure 4.8: A splitting of (3, 5)
(b) Show that, when (m, n) splits as (r, s) + (u, v), then both the pairs (r, s) and (u, v) are co-prime.
Exercice 4.6. See in Sage the function “GeneralizedTamariLattice” for the (m, n)-Tamari poset
when m > n in the coprime case.
6
This is useful for the calculation of π-basis expansions of symmetric functions.
61
62
Bibliography
[1] D. Armstrong, N. Loehr, and G. Warrington, Rational parking functions and Catalan
numbers, Annals of Combinatorics, 20 (2016), no. 1, 21–58. See arXiv:1403.1845. 47
[2] F. Bergeron, Open Questions for Operators Related to Rectangular Catalan Combinatorics,
Journal of Combinatorics Vol. 8, No. 4 (2017), 673–703. See arXiv:1603.04476. 47, 53, 54, 55
[3] F. Bergeron, Algebraic Combinatorics and Coinvariant Spaces, CMS Treatise in Mathematics,
CRC Press, 2009. 221 pages. 7, 39
[4] F. Bergeron, Multivariate Diagonal Coinvariant Spaces for Complex Reflection Groups,
Advances in Mathematics, Volume 239 (2013) 97–108. (see arXiv:1105.4358) 58
[5] F. Bergeron, P. Leroux, and G. Labelle, Combinatorial Species and Tree-like Structures,
Encyclopedia of Mathematics, Cambridge University Press, 1998. 497 pages. 33
[6] F. Bergeron and A. M. Garsia, Science Fiction and Macdonald’s Polynomials, in Algebraic
Methods and q-Special Functions, J.P. Van Dejen, L.Vinet (eds.), CRM Proceedings & Lecture
Notes, AMS (1999), 1–52. 43
[7] F. Bergeron, A.M. Garsia, M. Haiman and G. Tesler, Identities and Positivity Con-
jectures for some Remarkable Operators in the Theory of Symmetric Functions, Methods and
Applications of Analysis, 6 (1999), 363–420. 43, 44
[8] F. Bergeron, A. M. Garsia, E. Leven, and G. Xin, Compositional (km, kn)–
Shuffle Conjectures, International Mathematics Research Notices, (2015), 33 pages. doi:
10.1093/imrn/rnv272. arXiv:1404.4616. 54, 55
[9] F. Bergeron, L.-F. Préville-Ratelle, Higher Trivariate Diagonal Harmonics via gener-
alized Tamari Posets, Journal of Combinatorics, Volume 3 (2012), Number 3, 317–341. (see
arXiv:1105.3738) 56
[10] C. Ceballos, A. Padrol, and C. Sarmiento, Geometry of ν-Tamari lattices in types A
and B, (see arXiv:1611.09794) 56
[11] T. L. Bizley, Derivation of a new formula for the number of minimal lattice paths from (0, 0) to
(km, kn) having just t contacts with the line and a proof of Grossman’s formula for the number
of paths which may touch but do not rise above this line, J. Inst. Actuar. 80, (1954), 55–62. 50
[12] M. Bousquet-Mélou, G. Chapuy, L.-F. Préville-Ratelle, The representation of the
symmetric group on m-Tamari intervals, Proceedings FPSAC 2012–Nagoya, DMTCS 2012,
349–362. (see arXiv:1202.5925v2) 56
63
[13] M. Bousquet-Mélou, É. Fusy, and L.-F. Préville-Ratelle, The number of intervals in
the m-Tamari lattices. Electron. J. Combin., 18(2):Research Paper 31, 26 pp. (electronic), 2011.
(see arXiv:1109.2398) 56
[14] Peter Bürgisser, J.M. Landsberg, L. Manive, and J. Weyman, An Overview of
Mathematical Issues Arising in The Geometric Complexity Theory Approach to V P 6= V N P ,
Siam J. Comput. Vol. 40 (2011), No. 4, 1179–1209. arXiv:0907.2850 32
[15] J. Cigler, and C. Krattenthaler, Some determinants of path generating functions. Adv.
in Appl. Math. 46 (2011), no. 1-4, 144–174. 35
[16] V. Drinfeld, A new realization of Yangians and of quantum affine algebras, Soviet Math.
Dokl. 36 (1988), no. 2, 212–216. 47
[17] A. Dvoretzky and Th. Motzkin, A problem of arrangements, Duke Math. J. 14 (1947),
305–313. 49
[18] Ph. Flajolet, and R. Sedgewick, Analytic Combinatorics, Cambridge University Press,
2009. (Freely accessible http://ac.cs.princeton.edu/home/) 44
[19] J.S. Frame, G. de B. Robinson and R.M. Thrall, The Hook Graphs of the Symmetric
Groups, Canadian J. Math. 6 (1954), 316–324. 11
[20] W. Fulton, Young Tableaux, Cambridge University Press, 1997. 7
[21] A.M. Garsia, A q-analogue of the Lagrange Inversion Formula, Houston J. Math. 7 No. 2
(1981), 205–237. 43
[22] A.M. Garsia and I. Gessel, Permutation statistics and partitions, Adv. Math., 31 (1979),
288–305. 43
[23] I. Gessel and G.X. Viennot, Binomial Determinants, Paths, and Hook Length Formulae,
Adv. in Math. 58 (1985), 300–321. 28
[24] J. Haglund, The q, t-Catalan Numbers and the Space of Diagonal Harmonics, AMS University
Lecture Series, 2008. (Freely accessible http://www.math.upenn.edu/j̃haglund/) 39
[25] J. Haglund, M. Haiman and N. Loehr, A Combinatorial formula for Macdonald polynomials,
Jour. Amer. Math. Soc. 18 (2005), 735–761. arXiv/math/0409538 40
[26] J. Haglund, M. Haiman, N. Loehr, J. Remmel, and A. Ulyanov, A Combinatorial For-
mula for the Character of the Diagonal Coinvariants, Duke Mathematical Journal, 126 (2005),
195–232. MR2115257. 54
[27] J. Haglund, J. Remmel, and A. Wilson, The Delta Conjecture, Transactions of the
American Mathematical Society 370 (2018), 4029–4057. doi.org/10.1090/tran/7096. (See also
arXiv:1509.07058.) 44, 47
[28] M. Haiman, Combinatorics, symmetric functions and Hilbert schemes, In CDM 2002: Current
Developments in Mathematics in Honor of Wilfried Schmid & George Lusztig, International
Press Books (2003) 39–112. MR2051783. 43, 47
[29] M. Hogancamp, Khovanov-Rozansky homology and higher Catalan sequences, 39 pages.
arXiv:1704.01562 47
[30] C. Krattenthaler, and D. Yaqubi, Some determinants of path generating functions, II,
2018. arXiv:1802.05990 35
64
[31] C. Krattenthaler, Determinants of (generalised) Catalan numbers, J. Statist. Plann. Infer-
ence 140 (2010), no. 8, 2260–2270. arXiv:0709.3044v2 35
[32] D.E. Knuth, Permutations, Matrices and Generalized Young Tableaux, Pacific J. Math. 34
(1970), 709–727. 13
[33] J. M. Landsberg, Geometry and complexity theory, Cambridge Studies in Advanced Mathe-
matics, 169. Cambridge University Press, Cambridge, 2017. 32
[34] J. M. Landsberg, Geometric complexity theory: an introduction for geometers, Ann. Univ.
Ferrara Sez. VII Sci. Mat. 61 (2015), no. 1, 65–117. (also arXiv:1305.7387) 32
[35] C. Lenart, Lagrange Inversion and Schur Functions, J. Algebraic Combin. 11 (2000), 69–78.
46
[36] B. Lindström, On vector representations of induced matroids, Bull. London Math. Soc. 5
(1973), 85–90. 28
[37] I.G. Macdonald, Symmetric functions and Hall polynomials, second ed., Oxford Mathematical
Monographs, The Clarendon Press Oxford University Press, New York, 1995, With contributions
by A. Zelevinsky, Oxford Science Publications. 15, 19, 27, 28
[38] L. Manivel, Symmetric Functions, Schubert Polynomials and Degeneracy Loci, SMS/AMS
Texts and Monographs, vol. 6, 2001. 19
[39] K. D. Mulmuley, and M. Sohoni, Geometric Complexity Theory: Introduction
arXiv:0709.0746 32
[40] A. Negut, The Shuffle Algebra Revisited, International Mathematics Research Notices, Vol.
2014, No. 22, pp. 6242–6275. (see arXiv:1209.3349), (2012). 61
[41] B. Sagan, The Symmetric Group: Representations, Combinatorial Algorithms, and Symmetric
Functions, Graduate Texts in Mathematics, Springer, 2001. 19
[42] O. Schiffmann, Drinfeld realization of the elliptic Hall algebra, Journal of Algebraic Combina-
torics, 2012, Volume 35, Issue 2, pp 237–262. See arXiv:1004.2575) 47
[43] R.P. Stanley, Enumerative Combinatorics, Volume 1 and 2, Cambridge University Press, 1999.
7, 19, 23
[44] H. Thomas, and N. Williams, Sweeping up zeta, Séminaire Lotharingien de Combinatoire,
78B (2017), Art. 10, 12 pp. http://www.mat.univie.ac.at/ slc/ See also the latest much augmented
version: arXiv:1512.01483 54
65

Symmetric Functions

Uploaded by

Copyright:

Available Formats

Symmetric Functions

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Symmetric Functions

Uploaded by

Copyright:

Available Formats

Symmetric Functions and Rectangular

2 “Classical” Symmetric functions 19

3 Macdonald symmetric functions and operators 39

4 Rectangular combinatorics, and symmetric functions 47

1.1 Ferrers diagrams and partitions

Considering the component-wise partial order on N × N, a Ferrers diagram µ (with n cells), is a

(i, j) ≤ (k, l) and (k, l) ∈ µ implies (i, j) ∈ µ.

Figure 1.2: Corners of 4421.

In general, it is not easy to describe the polynomial

Dominance order on partitions

Figure 1.3: Young’s lattice.

1.2 Young tableaux, hook length formula, and Kostka numbers

Let d be a finite subset of N × N. A tableau of shape d, with value in a set A (typically

Figure 1.4: A semi-standard tableau.

for (i, j), (k, j), and (i, `) lying in d.

Figure 1.5: A standard tableau.

0 = µ(0) → µ(1) → . . . → µ(n) = µ,

Figure 1.6: A maximal chain in Young’s lattice

1.2.1 Hook length formula

The number f µ of standard tableaux of shape µ (µ ` n) is given by the Frame-Robinson-Thrall (see

The classical Robinson-Schensted-Knuth correspondence (see next section) shows that

1.2.2 Kostka numbers

1. First, we start with setting ε ←→ (0, 0).

and recursively sets v ←→ (P 0 , Q0 ).

This may be displayed as follows

since (f µ )2 counts the number of pairs of standard tableaux of shape µ.

One constructs as follows the tableau τ 0 := (τ ← a) obtained by insertion of a ∈ A in a semi-

Figure 1.7: Minimization of a tableau.

[n]q ! := [1]q [2]q · · · [n]q ,

(a) Show that we have the q-Pascal-triangle recursion

prove the recursion:

At least check that Cn (q) 6= Cn (q) in general.

Figure 1.8: A standard tableau of shape (µ + 1n )/µ, with µ = 53332 and n = 7.

Exercice 1.13. Show that, for µ ` n,

Exercice 1.14. (open) Explore the value of µ`n (f µ )k , for k ≥ 2.

“Classical” Symmetric functions

2.1 Basic notions

xa = xa11 xa22 · · · xann , for a = (a1 , a2 , . . . , an ) ∈ Nn .

p(x) = p0 (x) + p1 (x) + p2 (x) + . . . + pn (x),

f (xσ(1) , xσ(2) , . . . , xσ(n) ) = f (x1 , x2 , . . . , xn ).

• The complete homogeneous symmetric functions are defined to be the product hµ :=

For example, we have h1 = m1 ;

h2 = m11 + m2 , h11 = h21 = 2 m11 + m2 ;

h4 = m4 + m31 + m22 + m211 + m1111 ,

e211 = m31 + 2 m22 + 5 m211 + 12 m1111 ,

It is thus immediate that H(z) E(−z) = 1, and a simple calculation gives

2.3 Schur functions, a combinatorial approach

The Schur function sλ (x) is combinatorially defined as:

the product of degree d monomials xa y b , with a = (a1 , . . . , ad ) and b = (b1 , . . P

with hd expanded in the “variables” xi yj .

applying ω to the x component, to get the second from the first.

with fλ standing for the number of standard tableaux of shape λ.

2.4 Dual basis and Cauchy kernel

On Λ, we consider the Hall scalar product characterized by

In fact, this sum is equal to 1 (see exercise).

(1) hsλ , sµ i = 0, whenever λ 6= µ,

2.5 Product, Kronecker product, and Schur positivity

hsµ sν , sλ i = hsν , sλ/µ i. (2.5.3)

with Kλµ the Kostka numbers.

(b) Show that, for any a and b,