Algebraic Combinatorics: Richard P. Stanley

Undergraduate Texts in Mathematics
Richard P. Stanley
Algebraic
Combinatorics
Walks, Trees, Tableaux, and More
Second Edition
Series Editors:
Sheldon Axler
San Francisco State University, San Francisco, CA, USA
Kenneth Ribet
University of California, Berkeley, CA, USA
Advisory Board:
Colin Adams, Williams College

David A. Cox, Amherst College
L. Craig Evans, University of California, Berkeley
Pamela Gorkin, Bucknell University
Roger E. Howe, Yale University
Michael E. Orrison, Harvey Mudd College
Lisette G. de Pillis, Harvey Mudd College
Jill Pipher, Brown University
Fadil Santosa, University of Minnesota
Undergraduate Texts in Mathematics are generally aimed at third- and fourth-

year undergraduate mathematics students at North American universities. These
texts strive to provide students and teachers with new perspectives and novel
approaches. The books include motivation that guides the reader to an appreciation
of interrelations among different aspects of the subject. They feature examples that
illustrate key concepts as well as exercises that strengthen understanding.
More information about this series at http://www.springer.com/series/666

Richard P. Stanley
Algebraic Combinatorics
Walks, Trees, Tableaux, and More
Second Edition
123
Richard P. Stanley
Department of Mathematics
Massachusetts Institute of Technology
Cambridge, MA, USA
ISSN 0172-6056 ISSN 2197-5604 (electronic)

ISBN 978-3-319-77172-4 ISBN 978-3-319-77173-1 (eBook)
https://doi.org/10.1007/978-3-319-77173-1
Library of Congress Control Number: 2018941202
Mathematics Subject Classification (2010): Primary 05Exx; Secondary 15-01
© Springer Science+Business Media New York 2013

© Springer International Publishing AG, part of Springer Nature 2018
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, express or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.
Printed on acid-free paper
This Springer imprint is published by the registered company Springer International Publishing AG part
of Springer Nature.
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
to
Kenneth and Sharon
Look’d at each other with a wild surmise—
Preface to the Second Edition
The primary change from the first edition is the addition of a chapter entitled
“A Glimpse of Combinatorial Commutative Algebra” (Chapter 12). Writing this
chapter was an interesting challenge. The “standard” applications of commutative
algebra to combinatorics require a background in algebraic topology and homo-
logical algebra. I wanted to give a substantial application without using these two
subjects and without developing a lot of commutative algebra. It was still necessary
to present quite a bit of material on simplicial complexes so that the main result
(Theorem 12.25) can be adequately appreciated. The result has been that Chapter 12
is the longest chapter in the book. I hope that I have succeeded in giving some
of the flavor of the remarkable connections between commutative algebra and the
combinatorics of simplicial complexes.
I have also added a section to Chapter 13 (Section 13.8) involving commutative
algebra, but at a much simpler level than Chapter 12. A few new exercises have been
added throughout the book, and numerous typos and minor inaccuracies have been
corrected. There is now too much material for a one-semester course. The instructor
therefore has the pleasure of choosing among diverse treasures, while the student
has something to look forward to when the course has ended.
Cambridge, MA, USA Richard P. Stanley
ix
Updated Preface to the First Edition
This book is intended primarily as a one-semester undergraduate text for a course in

algebraic combinatorics. The main prerequisites are a basic knowledge of linear
algebra (eigenvalues, eigenvectors, etc.) over a field, existence of finite fields,
and some rudimentary understanding of group theory and (for Chapter 12 and
Section 13.8) ring theory. The one exception is Section 13.6, which involves finite
extensions of the rationals including a little Galois theory. Prior knowledge of
combinatorics is not essential but will be helpful.
Why write an undergraduate textbook on algebraic combinatorics? One obvious
reason is simply to gather some material that I find very interesting and hope
that students will agree. A second reason concerns students who have taken an
introductory algebra course and want to know what can be done with their new-
found knowledge. Undergraduate courses that require a basic knowledge of algebra
are typically either advanced algebra courses or abstract courses on subjects like
algebraic topology and algebraic geometry. Algebraic combinatorics offers a byway
off the traditional algebraic highway, one that is more intuitive and more easily
accessible.
Algebraic combinatorics is a huge subject, so some selection process was
necessary to obtain the present text. The main results, such as the weak Erdős–
Moser theorem and the enumeration of de Bruijn sequences, have the feature that
their statement does not involve any algebra. Such results are good advertisements
for the unifying power of algebra and for the unity of mathematics as a whole.
All but the last two chapters are vaguely connected to walks on graphs and linear
transformations related to them. The final chapter is a hodgepodge of some unrelated
elegant applications of algebra to combinatorics. The sections of this chapter are
independent from each other and the rest of the text. There are also three chapter
appendices on purely enumerational aspects of combinatorics related to the chapter
material: the RSK algorithm, plane partitions, and the enumeration of labelled trees.
Almost all the material covered here can serve as a gateway to much additional
algebraic combinatorics. We hope in fact that this book will serve exactly this
purpose, that is, to inspire its readers to delve more deeply into the fascinating
interplay between algebra and combinatorics.
xi
xii Updated Preface to the First Edition
Many persons have contributed to the writing of this book, but special thanks
should go to Christine Bessenrodt and Sergey Fomin for their careful reading of
portions of earlier manuscripts.
Cambridge, MA, USA Richard P. Stanley

Contents
Preface to the Second Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Updated Preface to the First Edition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi
Basic Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv
1 Walks in Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Cubes and the Radon Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3 Random Walks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4 The Sperner Property . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5 Group Actions on Boolean Algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
6 Young Diagrams and q-Binomial Coefficients. . . . . . . . . . . . . . . . . . . . . . . . . . . 57
7 Enumeration Under Group Action . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
8 A Glimpse of Young Tableaux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
9 The Matrix-Tree Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
10 Eulerian Digraphs and Oriented Trees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
11 Cycles, Bonds, and Electrical Networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
11.1 The Cycle Space and Bond Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
11.2 Bases for the Cycle Space and Bond Space . . . . . . . . . . . . . . . . . . . . . . . . . 168
11.3 Electrical Networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
11.4 Planar Graphs (Sketch). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177
11.5 Squaring the Square . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180
12 A Glimpse of Combinatorial Commutative Algebra . . . . . . . . . . . . . . . . . . . 187
12.1 Simplicial Complexes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
12.2 The Face Ring . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
xiii
xiv Contents
13 Miscellaneous Gems of Algebraic Combinatorics . . . . . . . . . . . . . . . . . . . . . . 219

13.1 The 100 Prisoners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219
13.2 Oddtown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
13.3 Complete Bipartite Partitions of Kn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
13.4 The Nonuniform Fisher Inequality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
13.5 Odd Neighborhood Covers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
13.6 Circulant Hadamard Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226
13.7 P -Recursive Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
13.8 Affine Monoids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Hints and Comments for Some Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
Basic Notation
P Positive integers
N Nonnegative integers
Z Integers
Q Rational numbers
R Real numbers
C Complex numbers
[n] The set {1, 2, . . . , n} for n ∈ N (so [0] = ∅)
Zn The group of integers modulo n
R[x] The ring of polynomials in the variable x with coefficients

in the ring R
YX For sets X and Y , the set of all functions f : X → Y
:= Equal by definition
Fq The finite field with q elements
(j ) 1 + q + q 2 + · · · + q j −1
(j )! (1)(2) · · · (j )
n (n)!
k (k)!(n−k)! , for 0 ≤ k ≤ n
#S or |S| Cardinality (number of elements) of the finite set S
xv
xvi Basic Notation
·
S ∪T The disjoint union of S and T , i.e., S ∪ T , where S ∩ T = ∅
2S The set of all subsets of the set S

S
k The set of k-element subsets of S

S
k The set of k-element multisets on S
KS The vector space with basis S over the field K
Bn The poset of all subsets of [n], ordered by inclusion
ρ(x) The rank of the element x in a graded poset
[x n ]F (x) Coefficient of x n in the polynomial or power series F (x)
x y, y x y covers x in a poset P
δij The Kronecker delta, which equals 1 if i = j and 0 otherwise
|L| The sum of the parts (entries) of L, if L is any array of

nonnegative integers
(λ) Length (number of parts) of the partition λ
p(n) Number of partitions of the integer n ≥ 0
ker ϕ The kernel of a linear transformation or group homomorphism
Sn Symmetric group of all permutations of 1, 2, . . . , n
ι The identity permutation of a set X, i.e., ι(x) = x for all x ∈ X

Chapter 1
Walks in Graphs

Given a finite set S and integer k ≥ 0, let Sk denote the set of k-element subsets
of S. A multiset may be regarded, somewhat informally, as a set with repeated
elements, such as {1, 1, 3, 4, 4, 4, 6, 6}. We are only concerned with how many
times each element occurs and not on any ordering of the elements. Thus for instance
{2, 1, 2, 4, 1, 2} and {1, 1, 2, 2, 2, 4} are the same multiset: they each contain two
1’s, three 2’s, and one 4 (and no other elements). We say that a multiset M is on a
set S if every element of M belongs to S. Thus the multiset in the example
above is
S
on the set S = {1, 3, 4, 6} and also on any set containing S. Let k denote the set
of k-element multisets on S. For instance, if S = {1, 2, 3}, then (using abbreviated
notation),

S S
= {12, 13, 23}, = {11, 22, 33, 12, 13, 23}.
2 2
We now define what is meant by a graph. Intuitively, graphs have vertices and
edges, where each edge “connects” two vertices (which may be the same). It is
possible for two different edges e and e to connect the same two vertices. We want
to be able to distinguish between these two edges, necessitating the following more
precise definition. A (finite) graph G consists of a vertex set V=
{v1 , . . . , vp } and
edge set E = {e1 , . . . , eq }, together with a function ϕ : E → V2 . We think that
if ϕ(e) = uv (short for {u, v}), then e connects u and v or equivalently e is incident
to u and v. If there is at least one edge incident to u and v, then we say that the
vertices u and v are adjacent. If ϕ(e) = vv, then we call e a loop at v. If several
edges e1 , . . . , ej (j > 1) satisfy ϕ(e1 ) = · · · = ϕ(ej ) = uv, then we say that
there is a multiple edge between u and v. A graph without loops or multiple edges
is called simple. In this case we can think of E as just a subset of V2 [why?].
The adjacency matrix of the graph G is the p×p matrix A = A(G), over the field
of complex numbers, whose (i, j )-entry aij is equal to the number of edges incident
© Springer International Publishing AG, part of Springer Nature 2018 1

R. P. Stanley, Algebraic Combinatorics, Undergraduate Texts in Mathematics,
https://doi.org/10.1007/978-3-319-77173-1_1
2 1 Walks in Graphs
to vi and vj . Thus A is a real symmetric matrix (and hence has real eigenvalues)
whose trace is the number of loops in G. For instance, if G is the graph
1 2
4 5
then
⎡ ⎤
2 1 0 2 0
⎢1 1⎥
⎢ 0 0 0 ⎥
⎢ ⎥
A(G) = ⎢ 0 0 0 0 0⎥.
⎢ ⎥
⎣2 0 0 0 1⎦
0 1 0 1 1
A walk in G of length from vertex u to vertex v is a sequence v1 , e1 , v2 , e2 , . . . ,

v , e , v+1 such that:
• Each vi is a vertex of G.
• Each ej is an edge of G.
• The vertices of ei are vi and vi+1 , for 1 ≤ i ≤ .
• v1 = u and v+1 = v.
1.1 Theorem. For any integer ≥ 1, the (i, j )-entry of the matrix A(G) is equal
to the number of walks from vi to vj in G of length .
Proof. This is an immediate consequence of the definition of matrix multiplication.

Let A = (aij ). The (i, j )-entry of A(G) is given by

(A(G) )ij = aii1 ai1 i2 · · · ai−1 j ,
where the sum ranges over all sequences (i1 , . . . , i−1 ) with 1 ≤ ik ≤ p. But
since ars is the number of edges between vr and vs , it follows that the summand
aii1 ai1 i2 · · · ai−1 j in the above sum is just the number (which may be 0) of walks of
length from vi to vj of the form
vi , e1 , vi1 , e2 , . . . , vi−1 , e , vj
1 Walks in Graphs 3
(since there are aii1 choices for e1 , ai1 i2 choices for e2 , etc.) Hence summing over
all (i1 , . . . , i−1 ) just gives the total number of walks of length from vi to vj , as
desired.
We wish to use Theorem 1.1 to obtain an explicit formula for the number
(A(G) )ij of walks of length in G from vi to vj . The formula we give will depend
on the eigenvalues of A(G). The eigenvalues of A(G) are also called simply the
eigenvalues of G. Recall that a real symmetric p × p matrix M has p linearly
independent real eigenvectors, which can in fact be chosen to be orthonormal (i.e.,
orthogonal and of unit length). Let u1 , . . . , up be real orthonormal eigenvectors for
M, with corresponding eigenvalues λ1 , . . . , λp . All vectors u will be regarded as
p × 1 column vectors, unless specified otherwise. We let t denote transpose, so ut
is a 1 × p row vector. Thus the dot (or scalar or inner) product of the vectors u
and v is given by ut v (ordinary matrix multiplication). In particular, uti uj = δij
(the Kronecker delta). Let U = (uij ) be the matrix whose columns are u1 , . . . , up ,
denoted U = [u1 , . . . , up ]. Thus U is an orthogonal matrix, so
⎡ ⎤
ut1
⎢ . ⎥
U t = U −1 = ⎣ .. ⎦ ,
utp
the matrix whose rows are ut1 , . . . , utp . Recall from linear algebra that the matrix U
diagonalizes M, i.e.,
U −1 MU = diag(λ1 , . . . , λp ),
where diag(λ1 , . . . , λp ) denotes the diagonal matrix with diagonal entries

λ1 , . . . , λp (in that order).
1.2 Corollary. Given the graph G as above, fix the two vertices vi and vj . Let
λ1 , . . . , λp be the eigenvalues of the adjacency matrix A(G). Then there exist real
numbers c1 , . . . , cp such that for all ≥ 1, we have
(A(G) )ij = c1 λ1 + · · · + cp λp . (1.1)
In fact, if U = (urs ) is a real orthogonal matrix such that U −1 AU =

diag(λ1 , . . . , λp ), then we have
ck = uik uj k .
Proof. We have [why?]
U −1 A U = diag(λ1 , . . . , λp ).
4 1 Walks in Graphs
Hence
A = U · diag(λ1 , . . . , λp )U −1 .
Taking the (i, j )-entry of both sides (and using U −1 = U t ) gives [why?]

(A )ij = uik λk uj k ,
k
as desired.
In order for Corollary 1.2 to be of any use we must be able to compute the
eigenvalues λ1 , . . . , λp as well as the diagonalizing matrix U (or eigenvectors ui ).
There is one interesting special situation in which it is not necessary to compute U .
A closed walk in G is a walk that ends where it begins. The number of closed walks
in G of length starting at vi is therefore given by (A(G) )ii , so the total number
fG () of closed walks of length is given by

p
fG () = (A(G) )ii
i=1
= tr(A(G) ),
where tr denotes trace (sum of the main diagonal entries). Now recall that the trace
of a square matrix is the sum of its eigenvalues. If the matrix M has eigenvalues
λ1 , . . . , λp then [why?] M has eigenvalues λ1 , . . . , λp . Hence we have proved the
following.
1.3 Corollary. Suppose A(G) has eigenvalues λ1 , . . . , λp . Then the number of
closed walks in G of length is given by
fG () = λ1 + · · · + λp .
We now are in a position to use various tricks and techniques from linear algebra
to count walks in graphs. Conversely, it is sometimes possible to count the walks by
combinatorial reasoning and use the resulting formula to determine the eigenvalues
of G. As a first simple example, we consider the complete graph Kp with vertex set
V = {v1 , . . . , vp } and one edge between any two distinct vertices. Thus Kp has p

vertices and p2 = 12 p(p − 1) edges.
1.4 Lemma. Let J denote the p × p matrix of all 1’s. Then the eigenvalues of J
are p (with multiplicity one) and 0 (with multiplicity p − 1).
Proof. Since all rows are equal and nonzero, we have rank(J ) = 1. Since a p × p
matrix of rank p − m has at least m eigenvalues equal to 0, we conclude that J has
1 Walks in Graphs 5
at least p − 1 eigenvalues equal to 0. Since tr(J ) = p and the trace is the sum of the
eigenvalues, it follows that the remaining eigenvalue of J is equal to p.
1.5 Proposition. The eigenvalues of the complete graph Kp are as follows: an

eigenvalue of −1 with multiplicity p −1 and an eigenvalue of p −1 with multiplicity
one.
Proof. We have A(Kp ) = J − I , where I denotes the p × p identity matrix. If the

eigenvalues of a matrix M are μ1 , . . . , μp , then the eigenvalues of M + cI (where c
is a scalar) are μ1 + c, . . . , μp + c [why?]. The proof follows from Lemma 1.4.
1.6 Corollary. The number of closed walks of length in Kp from some vertex vi
to itself is given by
1
(A(Kp ) )ii = ((p − 1) + (p − 1)(−1) ). (1.2)
p
(Note that this is also the number of sequences (i1 , . . . , i ) of numbers 1, 2, . . . , p

such that i1 = i, no two consecutive terms are equal, and i = i1 [why?].)
Proof. By Corollary 1.3 and Proposition 1.5, the total number of closed walks in
Kp of length is equal to (p − 1) + (p − 1)(−1) . By the symmetry of the graph
Kp , the number of closed walks of length from vi to itself does not depend on
i. (All vertices “look the same.”) Hence we can divide the total number of closed
walks by p (the number of vertices) to get the desired answer.
A combinatorial proof of Corollary 1.6 is quite tricky (Exercise 1.1). Our

algebraic proof gives a first hint of the power of algebra to solve enumerative
problems.
What about non-closed walks in Kp ? It’s not hard to diagonalize explicitly the
matrix A(Kp ) (or equivalently, to compute its eigenvectors), but there is an even
simpler special argument. We have

−k
(J − I ) =
(−1) J k, (1.3)
k
k=0
by the binomial theorem.1 Now for k > 0 we have J k = pk−1 J [why?], while
J 0 = I . (It is not clear a priori what is the “correct” value of J 0 , but in order for
(1.3) to be valid we must take J 0 = I .) Hence
1 We can apply the binomial theorem in this situation because I and J commute. If A and B are
p × p matrices that don’t necessarily commute, then the best we can say is (A + B)2 = A2 +
AB + BA + B 2 and similarly for higher powers.
6 1 Walks in Graphs

−k
(J − I ) =

(−1) pk−1 J + (−1) I.
k
k=1
Again by the binomial theorem we have
1
(J − I ) = ((p − 1) − (−1) )J + (−1) I. (1.4)
p
Taking the (i, j )-entry of each side when i = j yields
1
(A(Kp ) )ij = ((p − 1) − (−1) ). (1.5)
p
If we take the (i, i)-entry of (1.4), then we recover (1.2). Note the curious fact that
if i = j then
(A(Kp ) )ii − (A(Kp ) )ij = (−1) .
We could also have deduced (1.5) from Corollary 1.6 using
p
p
A(Kp ) = p(p − 1) ,
ij
i=1 j =1
the total number of walks of length in Kp . Details are left to the reader.
We now will show how (1.2) itself determines the eigenvalues of A(Kp ). Thus
if (1.2) is proved without first computing the eigenvalues of A(Kp ) (which in fact
is what we did two paragraphs ago), then we have another means to compute the
eigenvalues. The argument we will give can in principle be applied to any graph G,
not just Kp . We begin with a simple lemma.
1.7 Lemma. Suppose α1 , . . . , αr and β1 , . . . , βs are nonzero complex numbers
such that for all positive integers , we have
α1 + · · · + αr = β1 + · · · + βs . (1.6)
Then r = s and the α’s are just a permutation of the β’s.
Proof. We will use the powerful method of generating functions. Let x be a complex
number whose absolute value (or modulus) is close to 0. Multiply (1.6) by x and
sum on all ≥ 1. The geometric series we obtain will converge, and we get
α1 x αr x β1 x βs x
+ ··· + = + ··· + . (1.7)
1 − α1 x 1 − αr x 1 − β1 x 1 − βs x
Exercises for Chapter 1 7
This is an identity valid for sufficiently small (in modulus) complex numbers. By
clearing denominators we obtain a polynomial identity. But if two polynomials in x
agree for infinitely many values, then they are the same polynomial [why?]. Hence
(1.7) is actually valid for all complex numbers x (ignoring values of x which give
rise to a zero denominator).
Fix a complex number γ = 0. Multiply (1.7) by 1 − γ x and let x → 1/γ . The
left-hand side becomes the number of αi ’s which are equal to γ , while the right-
hand side becomes the number of βj ’s which are equal to γ [why?]. Hence these
numbers agree for all γ , so the lemma is proved.
1.8 Example. Suppose that G is a graph with 12 vertices and that the number of
closed walks of length in G is equal to 3 · 5 + 4 + 2(−2) + 4. Then it follows
from Corollary 1.3 and Lemma 1.7 [why?] that the eigenvalues of A(G) are given
by 5, 5, 5, 4, −2, −2, 1, 1, 1, 1, 0, 0.
Notes for Chapter 1
The connection between graph eigenvalues and the enumeration of walks is

considered “folklore.” The subject of spectral graph theory, which is concerned with
the spectrum (multiset of eigenvalues) of various matrices associated with graphs,
began around 1931 in the area of quantum chemistry. The first mathematical paper
was published by L. Collatz and U. Sinogowitz in 1957. A good general reference
is the book2 [28] by Cvetković et al. Two textbooks on this subject are by Cvetković
et al. [29] and by Brouwer and Haemers [14].
Exercises for Chapter 1
NOTE. An exercise marked with (*) is treated in the Hints section beginning on
page 245.
1. (tricky) Find a combinatorial proof of Corollary 1.6, i.e., the number of closed
walks of length in Kp from some vertex to itself is given by p1 ((p − 1) +
(p − 1)(−1) ).
2. Suppose that the graph G has 15 vertices and that the number of closed walks
of length in G is 8 + 2 · 3 + 3 · (−1) + (−6) + 5 for all ≥ 1. Let G be the
graph obtained from G by adding a loop at each vertex (in addition to whatever
loops are already there). How many closed walks of length are there in G ?
(Use linear algebraic techniques. You can also try to solve the problem purely
by combinatorial reasoning.)
2 All citations to the literature refer to the bibliography beginning on page 251.
8 1 Walks in Graphs
3. A bipartite graph G with vertex bipartition (A, B) is a graph whose vertex set
is the disjoint union A∪B · of A and B, such that every edge of G is incident to
one vertex in A and one vertex in B. Show by a walk-counting argument that
the nonzero eigenvalues of G come in pairs ±λ.
An equivalent formulation can be given in terms of the characteristic poly-
nomial f (x) of the matrix A(G). Recall that the characteristic polynomial
of a p × p matrix A is defined to be det(A − xI ). The present exercise is
then equivalent to the statement that when G is bipartite, the characteristic
polynomial f (x) of A(G) has the form g(x 2 ) (if G has an even number of
vertices) or xg(x 2 ) (if G has an odd number of vertices) for some polynomial
g(x).
NOTE. Sometimes the characteristic polynomial of a p × p matrix A is defined
to be det(xI −A) = (−1)p det(A−xI ). We will use the definition det(A−xI ),
so that the value at x = 0 is det A.
4. Let r, s ≥ 1. The complete bipartite graph Krs has vertices u1 , u2 , . . . , ur ,
v1 , v2 , . . . , vs , with one edge between each ui and vj (so rs edges in all).
(a) By purely combinatorial reasoning, compute the number of closed walks
of length in Krs .
(b) Deduce from (a) the eigenvalues of Krs .
5. (*) Let Hn be the complete bipartite graph Knn with n vertex-disjoint edges
removed. Thus Hn has 2n vertices and n(n − 1) edges, each of degree (number
of incident edges) n − 1. Show that the eigenvalues of Hn are ±1 (n − 1 times
each) and ±(n − 1) (once each).
6. Let n ≥ 1. The complete p-partite graph K(n, p) has vertex set V =
V1 ∪· · · · ∪V
· p (disjoint union), where each |Vi | = n, and an edge from every
element of Vi to every element of Vj when i = j . (If u, v ∈ Vi then there is no
edge uv.) Thus K(1, p) is the complete graph Kp , and K(n, 2) is the complete
bipartite graph Knn .
(a) (*) Use Corollary 1.6 to find the number of closed walks of length in
K(n, p).
(b) Deduce from (a) the eigenvalues of K(n, p).
7. Let G be any finite simple graph, with eigenvalues λ1 , . . . , λp . Let G(n) be
the graph obtained from G by replacing each vertex v of G with a set Vv of
n vertices, such that if uv is an edge of G, then there is an edge from every
vertex of Vu to every vertex of Vv (and no other edges). For instance, Kp (n) =
K(n, p). Find the eigenvalues of G(n) in terms of λ1 , . . . , λp .
8. Let G be a (finite) graph on p vertices. Let G be the graph obtained from G
by placing a new edge ev incident to each vertex v, with the other vertex of ev
being a new vertex v . Thus G has p new edges and p new vertices. The new
vertices all have degree one. By combinatorial or algebraicreasoning, show
that if G has eigenvalues λi then G has eigenvalues (λi ± λ2i + 4)/2. (An
algebraic proof is much easier than a combinatorial proof.)
9. Let G be a (finite) graph with vertices v1 , . . . , vp and eigenvalues λ1 , . . . , λp .

We know that for any i, j there are real numbers c1 (i, j ), . . . , cp (i, j ) such that
for all ≥ 1,

p
A(G) = ck (i, j )λk .
ij
k=1
(a) Show that ck (i, i) ≥ 0.

(b) Show that if i = j then we can have ck (i, j ) < 0. (The simplest possible
example will work.)
10. Let G be a finite graph with eigenvalues λ1 , . . . , λp . Let G be the graph with
the same vertex set as G and with η(u, v) edges between vertices u and v
(including u = v), where η(u, v) is the number of walks in G of length two
from u to v. For example,
G G*
Find the eigenvalues of G in terms of those of G.

11. (*) Let Kno denote the complete graph with n vertices, with one loop at each

vertex. (Thus A(Kno ) = Jn , the n × n all 1’s matrix, and Kno has n+1
2 edges.)
Let Kno − Kmo denote K o with the edges of K o removed, i.e., choose m vertices
n m
of Kno and remove all edges between these vertices (including loops). (Thus

o has n+1 − m+1 edges.) Find the number C() of closed walks in
Kno − Km 2 2
= K21 o − K o of length ≥ 1.
18
12. (a) Let G be a finite graph, and let be the maximum degree of any vertex
of G. Let λ1 be the largest eigenvalue of the adjacency matrix A(G). Show
that λ1 ≤ .
(b) (*) Suppose that G is simple
√ (no loops or multiple edges) and has a total of
q edges. Show that λ1 ≤ 2q.
13. Let G be a finite simple graph with at least two vertices. Suppose that for some
≥ 1, the number of walks of length between any two vertices u, v (including
u = v) is odd. Show that there is a nonempty subset S of the vertices such that
S has an even number of elements and such that every vertex v of G is adjacent
to an even number of vertices in S. (In a simple graph, no vertex is adjacent to
itself.)
Chapter 2
Cubes and the Radon Transform
Let us now consider a more interesting example of a graph G, one whose

eigenvalues have come up in a variety of applications. Let Z2 denote the cyclic
group of order 2, with elements 0 and 1 and group operation being addition modulo
2. Thus 0 + 0 = 0, 0 + 1 = 1 + 0 = 1, and 1 + 1 = 0. Let Zn2 denote the direct
product of Z2 with itself n times, so the elements of Zn2 are n-tuples (a1 , . . . , an )
of 0’s and 1’s, under the operation of component-wise addition. Define a graph
Cn , called the n-cube, as follows: the vertex set of Cn is given by V (Cn ) = Zn2 ,
and two vertices u and v are connected by an edge if they differ in exactly one
component. Equivalently, u + v has exactly one nonzero component. If we regard
Zn2 as consisting of real vectors, then these vectors form the set of vertices of
an n-dimensional cube. Moreover, two vertices of the cube lie on an edge (in
the usual geometric sense) if and only if they form an edge of Cn . This explains
why Cn is called the n-cube. We also see that walks in Cn have a nice geometric
interpretation—they are simply walks along the edges of an n-dimensional cube.
We want to determine explicitly the eigenvalues and eigenvectors of Cn . We will
do this by a somewhat indirect but extremely useful and powerful technique, the
finite Radon transform. Let V denote the set of all functions f : Zn2 → R, where R
denotes the field of real numbers.1 Note that V is a vector space over R of dimension
2n [why?]. If u = (u1 , . . . , un ) and v = (v1 , . . . , vn ) are elements of Zn2 , then define
their dot product by
u · v = u1 v1 + · · · + un vn , (2.1)
where the computation is performed modulo 2. Thus we regard u·v as an element of

Z2 . The expression (−1)u·v is defined to be the real number +1 or −1, depending
on whether u · v = 0 or 1, respectively. Since for integers k the value of (−1)k
1 For abelian groups other than Zn

it is necessary to use complex numbers rather than real numbers.
2
We could use complex numbers here, but there is no need to do so.

https://doi.org/10.1007/978-3-319-77173-1_2
12 2 Cubes and the Radon Transform
depends only on k (mod 2), it follows that we can treat u and v as integer vectors
without affecting the value of (−1)u·v . Thus, for instance, formulas such as
(−1)u·(v+w) = (−1)u·v+u·w = (−1)u·v (−1)u·w
are well defined and valid. From a more algebraic viewpoint, the map Z → {−1, 1}
sending n to (−1)n is a group homomorphism, where of course the product on
{−1, 1} is multiplication.
We now define two important bases of the vector space V. There will be one basis
element of each basis for each u ∈ Zn2 . The first basis, denoted B1 , has elements fu
defined as follows:
fu (v) = δuv , (2.2)
the Kronecker delta. It is easy to see that B1 is a basis, since any g ∈ V satisfies

g= g(u)fu (2.3)
u∈Zn2
[why?]. Hence B1 spans V, so since #B1 = dim V = 2n , it follows that B1 is a basis.

The second basis, denoted B2 , has elements χu defined as follows:
χu (v) = (−1)u·v .
In order to show that B2 is a basis, we will use an inner product on V (denoted ·, ·)
defined by

f, g = f (u)g(u).
u∈Zn2
Note that this inner product is just the usual dot product with respect to the basis B1 .
2.1 Lemma. The set B2 = {χu : u ∈ Zn2 } forms a basis for V.
Proof. Since #B2 = dim V (= 2n ), it suffices to show that B2 is linearly
independent. In fact, we will show that the elements of B2 are orthogonal.2 We
have

χu , χv = χu (w)χv (w)
w∈Zn2

= (−1)(u+v)·w .
w∈Zn2
2 Recall
from linear algebra that nonzero orthogonal vectors in a real vector space are linearly
independent.
2 Cubes and the Radon Transform 13
It is left as an easy exercise to the reader to show that for any y ∈ Zn2 , we have

2n , if y = 0,
(−1) y·w
=
0, otherwise,
w∈Zn2
where 0 denotes the identity element of Zn2 (the vector (0, 0, . . . , 0)). Thus
χu , χv = 0 if and only u + v = 0, i.e., u = v, so the elements of B2 are orthogonal
(and nonzero). Hence they are linearly independent as desired.
We now come to the key definition of the Radon transform.

Given a subset of Zn2 and a function f ∈ V, define a new function f ∈ V by

f (v) = f (v + w).
w∈
The function f is called the (discrete or finite) Radon transform of f (on the
group Zn2 , with respect to the subset ).
We have defined a map : V → V. It is easy to see that is a linear
transformation; we want to compute its eigenvalues and eigenvectors.
2.2 Theorem. The eigenvectors of are the functions χu , where u ∈ Zn2 . The
eigenvalue λu corresponding to χu (i.e., χu = λu χu ) is given by

λu = (−1)u·w .
w∈
Proof. Let v ∈ Zn2 . Then

χu (v) = χu (v + w)
w∈

= (−1)u·(v+w)
w∈

= (−1) u·w
(−1)u·v
w∈

= (−1) u·w
χu (v).
w∈
Hence

χu = (−1) u·w
χu ,
w∈
as desired.
Note that because the χu ’s form a basis for V by Lemma 2.1, it follows that
Theorem 2.2 yields a complete set of eigenvalues and eigenvectors for . Note
also that the eigenvectors χu of are independent of ; only the eigenvalues
depend on .
Now we come to the payoff. Let = {δ1 , . . . , δn }, where δi is the ith unit
coordinate vector (i.e., δi has a 1 in position i and 0’s elsewhere). Note that the
j th coordinate of δi is just δij (the Kronecker delta), explaining our notation δi . Let
[ ] denote the matrix of the linear transformation : V → V with respect to
the basis B1 of V given by (2.2).
2.3 Lemma. We have [ ] = A(Cn ), the adjacency matrix of the n-cube.
Proof. Let v ∈ Zn2 . We have

fu (v) = fu (v + w)
w∈

= fu+w (v),
w∈
since u = v + w if and only if u + w = v. There follows

fu = fu+w . (2.4)
w∈
Equation (2.4) says that the (u, v)-entry (short for (fu , fv )-entry) of the matrix [ ]
is given by

1, if u + v ∈
( )uv =
0, otherwise.
Now u + v ∈ if and only if u and v differ in exactly one coordinate. This is just
the condition for uv to be an edge of Cn , so the proof follows.
2.4 Corollary. The eigenvectors Eu (u ∈ Zn2 ) of A(Cn ) (regarded as linear

combinations of the vertices of Cn , i.e., of the elements of Zn2 ) are given by

Eu = (−1)u·v v. (2.5)
v∈Zn2
The eigenvalue λu corresponding to the eigenvector Eu is given by
λu = n − 2ω(u), (2.6)
where ω(u) is the number of 1’s in u. (The integer ω(u)

is called the Hamming
weight or simply the weight of u.) Hence A(Cn ) has ni eigenvalues equal to n − 2i,
for each 0 ≤ i ≤ n.
Proof. For any function g ∈ V we have by (2.3) that

g= g(v)fv .
v
Applying this equation to g = χu gives

χu = χu (v)fv = (−1)u·v fv . (2.7)
v v
Equation (2.7) expresses the eigenvector χu of (or even for any ⊆ Zn2 ) as
a linear combination of the functions fv . But has the same matrix with respect
to the basis of the fv ’s as A(Cn ) has with respect to the vertices v of Cn . Hence the
expansion of the eigenvectors of in terms of the fv ’s has the same coefficients
as the expansion of the eigenvectors of A(Cn ) in terms of the v’s, so (2.5) follows.
According to Theorem 2.2 the eigenvalue λu corresponding to the eigenvector
χu of (or equivalently, the eigenvector Eu of A(Cn )) is given by

λu = (−1)u·w . (2.8)
w∈
Now = {δ1 , . . . , δn } and δi · u is 1 if u has a one in its ith coordinate and is 0

otherwise. Hence the sum in (2.8) has n − ω(u) terms equal to +1 and ω(u) terms
equal to −1, so λu = (n − ω(u)) − ω(u) = n − 2ω(u), as claimed.
We have all the information needed to count walks in Cn .

2.5 Corollary. Let u, v ∈ Zn2 , and suppose that ω(u + v) = k (i.e., u and v disagree
in exactly k coordinates). Then the number of walks of length in Cn between u and
v is given by

1
n k
j k n−k
(A )uv = (−1) (n − 2i) , (2.9)
2n j i−j
i=0 j =0
n−k
where we set i−j = 0 if j > i. In particular,
n
1 n

(A )uu = n (n − 2i) . (2.10)
2 i
i=0
Proof. Let Eu and λu be as in Corollary 2.4. In order to apply Corollary 1.2, we need
the eigenvectors to be of unit length (where we regard the fv ’s as an orthonormal
basis of V). By (2.5), we have

|Eu |2 = ((−1)u·v )2 = 2n .
v∈Zn2
Hence we should replace Eu by Eu = 1

E
2n/2 u
to get an orthonormal basis.
According to Corollary 1.2, we thus have
1
(A )uv = Euw Evw λw .
2n n w∈Z2
Now Euw by definition is the coefficient of fw in the expansion (2.5), i.e., Euw =
(−1)u·w (and similarly for Ev ), while λw = n − 2ω(w). Hence
1
(A )uv = (−1)(u+v)·w (n − 2ω(w)) . (2.11)
2n n
w∈Z2
The number of vectors

w of Hamming weight i which have j 1’s in common with
u + v is jk n−k , since we can choose the j 1’s in u + v which agree with w in
k i−j
j ways, while the remaining i − j 1’s of w can be inserted in the n − k remaining

positions in n−k
i−j ways. Since (u + v) · w ≡ j (mod 2), the sum (2.11) reduces to
(2.9) as desired. Clearly setting u = v in (2.9) yields (2.10), completing the proof.
It is possible to give a direct proof of (2.10) avoiding linear algebra, though we

do not do so here. Thus by Corollary 1.3 and Lemma 1.7 (exactly as was done for
Kn ) we have another determination of the eigenvalues of Cn . With a little more work
one can also obtain a direct proof of (2.9). Later in Example 9.12, however, we will
use the eigenvalues of Cn to obtain a combinatorial result for which a nonalgebraic
proof was found only recently and is by no means easy.
2.6 Example. Setting k = 1 in (2.9) yields
n
1 n−1 n−1

(A )uv = n − (n − 2i)
2 i i−1
i=0
n−1
1 n − 1 (n − 2i)+1
= .
2n i n−i
i=0
NOTE (for those familiar with the representation theory of finite groups). The
functions χu : Zn2 → R are just the irreducible (complex) characters of the group
Zn2 , and the orthogonality of the χu ’s shown in the proof of Lemma 2.1 is the usual
orthogonality relation for the irreducible characters of a finite group. The results of
this chapter extend readily to any finite abelian group. Exercise 2.5 does the case
Zn , the cyclic group of order n. For nonabelian finite groups the situation is much
more complicated because not all irreducible representations have degree one (i.e.,
are homomorphisms G → C∗ , the multiplicative group of C), and there do not exist
formulas as explicit as the ones for abelian groups.
We can give a little taste of the situation for arbitrary groups as follows. Let G be
a finite group, and let M(G) be its multiplication table. Regard the entries of M(G)
as commuting indeterminates, so that M(G) is simply a matrix with indeterminate
entries. For instance, let G = Z3 . Let the elements of G be a, b, c, where say a is
the identity. Then
⎡ ⎤
abc
M(G) = ⎣ b c a ⎦ .
cab
We can compute that det M(G) = (a + b + c)(a + ωb + ω2 c)(a + ω2 b + ωc),

where ω = e2π i/3 . In general, when G is abelian, Dedekind knew that det M(G)
factors into certain explicit linear factors over C. Theorem 2.2 is equivalent to this
statement for the group G = Zn2 [why?]. Equation (13.5) gives the factorization for
G = Zn . (For each w ∈ G one needs to interchange the row indexed by the group
element w with the row indexed by w −1 in order to convert M(Zn ) to the circulant
matrices of (13.5), but these operations only affect the sign of the determinant.)
Dedekind asked Frobenius about the factorization of det M(G), known as the group
determinant, for nonabelian finite G. For instance, let G = S3 (the symmetric group
of all permutations of 1, 2, 3), with elements (in cycle notation) a = (1)(2)(3),
b = (1, 2)(3), c = (1, 3)(2), d = (1)(2, 3), e = (1, 2, 3), and f = (1, 3, 2). Then
det M(G) = f1 f2 f32 , where
f1 = a + b + c + d + e + f
f2 = −a + b + c + d − e − f
f3 = a 2 − b2 − c2 − d 2 + e2 + f 2 − ae − af + bc + bd + cd − ef.
Frobenius showed that in general there is a set P of irreducible homogeneous

polynomials f , of some degree df , where #P is the number of conjugacy classes of
G, for which

det M(G) = f df .
f ∈P

Note that taking the degree of both sides gives #G = f df2 . Frobenius’ result was
a highlight in his development of group representation theory. The numbers df are
just the degrees of the irreducible (complex) representations of G. For the symmetric
group Sn , these degrees are the numbers λ
fλ 2of Theorem 8.1, and Appendix 1 of
Chapter 8 gives a bijective proof that λ (f ) = n!.
Notes for Chapter 2
The Radon transform first arose in a continuous setting in the paper [107] of Radon
and has been applied to such areas as computerized tomography. The finite version
was first defined by Bolker [10]. For some further applications to combinatorics see
Kung [80]. For the Radon transform on the n-cube Zn2 , see Diaconis and Graham
[34]. For the generalization to Znk , see DeDeo and Velasquez [33].
For an exposition of the development of group representation theory by Frobe-
nius and other pioneers, see the survey articles of Hawkins [63–65].
1. (a) Start with n coins heads up. Choose a coin at random (each equally likely)
and turn it over. Do this a total of times. What is the probability that all
coins will have heads up? (Don’t solve this from scratch; rather use some
previous results.)
(b) Same as (a), except now compute the probability that all coins have tails up.
(c) Same as (a), but now we turn over two coins at a time.
2. (a) (difficult) (*) For k < n/2 let Cn,k be the subgraph of the cube Cn spanned
by all vertices of Cn with k − 1 or k 1’s (so the edges of Cn,k consist
of all
edges of Cn that connect two vertices of Cn,k ; there are a total of k nk edges).
Show that the characteristic polynomial of A = A(Cn,k ) is given by
n n
k
n n
det(A − xI ) = ±x (k )−(k−1) (x 2 − i(n − 2k + i + 1))(k−i )−(k−i−1) ,
i=1
n
where we set −1 = 0.
(b) Find the number of closed walks in Cn,k of length beginning and ending
with a fixed vertex v.
3. (very difficult, and unrelated to the text) (*) Let n = 2k + 1. Show that the
graphs Cn,k+1 of Problem 2 above have a Hamiltonian cycle, i.e., a closed path
that contains every vertex exactly once. A closed path in a graph G is a closed
walk that does not repeat any vertices except at the last step.
4. Let G be the graph with vertex set Zn2 (the same as the n-cube) and with edge
set defined as follows: {u, v} is an edge of G if u and v differ in exactly two
coordinates (i.e., if ω(u, v) = 2). What are the eigenvalues of G?
5. This problem is devoted to the graph Zn with vertex set Zn (the cyclic group of
order n, with elements 0, 1, . . . , n − 1 under the operation of addition modulo
n) and edges consisting of all pairs {i, i + 1} (with i + 1 computed in Zn , so
(n − 1) + 1 = 0). The graph Zn is called an n-cycle. We will develop properties
of its adjacency matrix analogously to what was done for the n-cube Cn . It will
be necessary to work over the complex numbers C. Recall that there are exactly n
complex numbers z (called nth roots of unity) satisfying zn = 1. They are given
by ζ 0 = 1, ζ 1 = ζ, ζ 2 , . . . , ζ n−1 , where ζ = e2π i/n .
(a) Draw the graphs Z3 , Z4 , and Z5 .
(b) Let V be the complex vector space of all functions f : Zn → C. What is the
dimension of V?
(c) (*) If k ∈ Z, then note that ζ k depends only on the value of k modulo n.
Hence if u ∈ Zn then we can define ζ u by regarding u as an ordinary integer,
and the usual laws of exponents such as ζ u+v = ζ u ζ v (where u, v ∈ Zn ) still
hold. For u ∈ Zn define χu ∈ V by χu (v) = ζ uv . Let B = {χu : u ∈ Zn }.
Show that B is a basis for V.
(d) Given ⊆ Zn and f ∈ V, define f ∈ V by

f (v) = f (v + w).
w∈
Show that the eigenvectors

of are the functions χu , with corresponding
eigenvalue λu = w∈ ζ uw .
(e) Let = {1, n − 1} ⊆ Zn . Define fu ∈ V by fu (v) = δuv . Let F = {fu : u ∈
Zn }. It is clear that F is a basis for V (just as for Cn ). Show that the matrix
[ ] of with respect to the basis F is just A(Zn ), the adjacency matrix
of Zn .
(f) Show that the eigenvalues of A(Zn ) are the numbers 2 cos( 2πj n ), where 0 ≤
j ≤ n − 1. What are the corresponding eigenvectors?
(g) How many closed walks in Zn are of length and start at 0? Give the answers
in the cases n = 4 and n = 6 without using trigonometric functions, complex
exponentials, etc.
(2)
(h) Let Zn be the graph with vertex set Zn and edges {i, j } for j − i = 1 or
(2)
j − i = 2. How many closed walks in Zn are of length and start at 0? Try
to express your answer in terms of trigonometric functions, and not involving
complex numbers.
n be the graph obtained from the n-cube graph Cn by adding an edge
6. Let C
between every vertex v and its antipode (the vertex which differs from v in all
n coordinates). Find the number of closed walks in C n of length which begin
(and hence end) at the origin 0 = (0, 0, . . . , 0).
Chapter 3
Random Walks
Let G be a finite graph. We assume throughout this chapter that G has at least two
vertices and is connected, i.e., there exists a walk between any two vertices of G.
We consider a random walk on the vertices of G of the following type. Start at a
vertex u. (The vertex u could be chosen randomly according to some probability
distribution or could be specified in advance.) Among all the edges incident to u,
choose one uniformly at random (i.e., if there are k edges incident to u, then each of
these edges is chosen with probability 1/k). Travel to the vertex v at the other end of
the chosen edge and continue as before from v. Readers with some familiarity with
probability theory will recognize this random walk as a special case of a finite-state
Markov chain. Many interesting questions may be asked about such walks; the basic
one is to determine the probability of being at a given vertex after a given number
of steps.
Suppose vertex u has degree du , i.e., there are du edges incident to u (counting
loops at u once only). Let M = M(G) be the matrix whose rows and columns are
indexed by the vertex set {v1 , . . . , vp } of G and whose (u, v)-entry is given by
μuv
M uv = , (3.1)
du
where μuv is the number of edges between u and v (which for simple graphs will
be 0 or 1). Thus M uv is just the probability that if one starts at u, then the next step
will be to v. We call M the probability matrix associated with G. An elementary
probability theory argument (equivalent to Theorem 1.1) shows that if is a positive
integer, then (M )uv is equal to the probability that one ends up at vertex v in steps
given that one has started at u. Suppose now that the starting vertex is not specified,
but rather we are given probabilities ρu summing to 1 and that we start at vertex
u with probability ρu . Let P be the row vector P = [ρv1 , . . . , ρvp ]. Then again an
elementary argument shows that if P M = [σv1 , . . . , σvp ], then σv is the probability
of ending up at v in steps (with the given starting distribution). By reasoning as in

https://doi.org/10.1007/978-3-319-77173-1_3
22 3 Random Walks
Chapter 1, we see that if we know the eigenvalues and eigenvectors of M, then we

can compute the crucial probabilities (M )uv and σu .
Since the matrix M is not the same as the adjacency matrix A, what does all this
have to do with adjacency matrices? The answer is that in one important case M is
just a scalar multiple of A. We say that the graph G is regular of degree d if each
du = d, i.e., each vertex is incident to d edges. In this case it’s easy to see that
M(G) = d1 A(G). Hence the eigenvectors Eu of M(G) and A(G) are the same, and
the eigenvalues are related by λu (M) = d1 λu (A). Thus random walks on a regular
graph are closely related to the adjacency matrix of the graph.
3.1 Example. Consider a random walk on the n-cube Cn which begins at the
“origin” (the vector (0, . . . , 0)). What is the probability p that after steps
one is again at the origin? Before applying any formulas, note that after an
even (respectively, odd) number of steps, one must be at a vertex with an even
(respectively, odd) number of 1’s. Hence p = 0 if is odd. Now note that Cn
is regular of degree n. Thus by (2.6), we have
1
λu (M(Cn )) = (n − 2ω(u)).
n
By (2.10) we conclude that
n
1 n
p = n (n − 2i) .
2 n i
i=0
Note that the above expression for p does indeed reduce to 0 when is odd.
It is worth noting that even though the probability matrix M need not be a
symmetric matrix, nonetheless it has only real eigenvalues.
3.2 Theorem. Let G be a finite graph. Then the probability matrix M = M(G) is
diagonalizable and has only real eigenvalues.
Proof. Since we are assuming that G is connected and has at least two vertices, it
follows that dv > 0 for every vertex v of G. Let D be the diagonal
√ matrix whose
rows and columns are indexed by the vertices of G, with D vv = dv . Then
μuv 1
(DMD −1 )uv = du · ·√
du dv
μuv
= √ .
du dv
Hence DMD −1 is a symmetric matrix and thus has only real eigenvalues. But if
B and C are any p × p matrices with C invertible, then B and CBC −1 have the
same characteristic polynomial and hence the same eigenvalues. Therefore all the
eigenvalues of M are real. Moreover, B is diagonalizable if and only if CBC −1
3 Random Walks 23
is diagonalizable. (In fact, B and CBC −1 have the same Jordan canonical form.)
Since a symmetric matrix is diagonalizable, it follows that M is also diagonalizable.
Let us give one further example of the connection between linear algebra and
random walks on graphs. Let u and v be vertices of a connected graph G. Define
the access time or hitting time H (u, v) to be the expected number of steps that a
random walk (as defined above) starting at u takes to reach v for the first time. Thus
if the probability is pn that we reach v for the first time in n steps, then by definition
of expectation we have

H (u, v) = npn . (3.2)
n≥1
Conceivably this sum could be infinite, though we will see below that this is not the
case. Note that H (v, v) = 0.
As an example, suppose that G has three vertices u, v, w with an edge between
u and w and another edge between w and v. We can compute H (u, v) as follows.
After one step we will be at w. Then with probability 12 we will step to v and with
probability 12 back to u. Hence [why?]
1 1
H (u, v) = · 2 + (2 + H (u, v)). (3.3)
2 2
Solving this linear equation gives H (u, v) = 4.
We want to give a formula for the access time H (u, v) in terms of linear
algebra. The proof requires some basic results on eigenvalues and eigenvectors of
nonnegative matrices, which we will explain and then state without proof. An r × r
real matrix B is called nonnegative if every entry is nonnegative. We say that B is
irreducible if it is not the 1 × 1 matrix [0] and if there does not exist a permutation
matrix P (a matrix with one 1 in every row and column, and all other entries 0) such
that

CD
P BP −1 = ,
0 E
where C and E are square matrices of size greater than zero. For instance, the
adjacency matrix A and probability matrix M of a graph G are irreducible if and
only if G is connected and is not an isolated vertex (i.e., a vertex v incident to no
edges, not even a loop from v to itself). We now state without proof a version of
the Perron–Frobenius theorem. There are some other parts of the Perron–Frobenius
theorem that we don’t need here and are omitted.
3.3 Theorem. Let B be a nonnegative irreducible square matrix. If ρ is the
maximum absolute value of the eigenvalues of B, then ρ > 0, and there is an
24 3 Random Walks
eigenvalue equal to ρ. Moreover, there is an eigenvector for ρ (unique up to

multiplication by a positive real number) all of whose entries are positive.
Now let M be the probability matrix defined by (3.1). Let M[v] denote M with
the row and column indexed by v deleted. Thus if G has p vertices, then M[v] is
a (p − 1) × (p − 1) matrix. Let T [v] be the column vector of length p − 1 whose
rows are indexed by the vertices w = v, with T [v]w = μwv /dw . Write Ip−1 for the
identity matrix of size p − 1.
3.4 Theorem. The matrix Ip−1 − M[v] is invertible, and
H (u, v) = ((Ip−1 − M[v])−2 T [v])u , (3.4)
the u-entry of the column vector (Ip−1 − M[v])−2 T [v].
Proof. We first give a “formal” argument and then justify its validity. The probabil-
ity that when we take n steps from u, we never reach v and end up at some vertex
w is (M[v]n )uw [why?]. The probability that once we reach w the next step is to v
is μ(w, v)/dw . Hence by definition of expectation we have
μwv
H (u, v) = (n + 1) (M[v]n )uw . (3.5)
dw
w=v n≥0
We claim that if x is a complex number satisfying |x| < 1, then

(n + 1)x n = (1 − x)−2 . (3.6)
n≥0
This identity is a simple exercise in calculus.

For instance, we can compute the
coefficient of x n in the product (1 − x)2 n≥0 (n + 1)x n . We can also differentiate
the familiar identity
1
xn = . (3.7)
1−x
n≥0
Another proof is obtained by expanding (1 − x)−2 by the binomial theorem for the
exponent −2. Convergence for |x| < 1 follows for example from the corresponding
result for (3.7).
Let us “blindly” apply (3.6) to (3.5). We obtain
μwv
H (u, v) = ((Ip−1 − M[v])−2 )uw
dw
w=v
= ((Ip−1 − M[v])−2 T [v])u , (3.8)
as claimed.
3 Random Walks 25
It remains to justify our derivation

of (3.8). For an arbitrary real (or
complex)
r × r matrix B, we can define n≥0 (n + 1)B n entry-wise, that is, we set n≥0 (n +
1)B n = C if

(n + 1)(B n )ij = C ij
n≥0
for all i and j indexing the rows and columns of B and C.

It is straightforward to verify by induction on m the identity

(Ir − B)2 Ir + 2B + 3B 2 + · · · + mB m−1 = Ir − (m + 1)B m + mB m+1 .
(3.9)
Suppose that B is diagonalizable and that all eigenvalues λ1 , . . . , λr of B satisfy
|λj | < 1. Note that our proof of (1.1) extends to any diagonalizable matrix. (The
matrix U need not be orthogonal, but this is irrelevant to the proof.) Hence
(B n )ij = c1 λn1 + · · · + cr λnr ,
where c1 , . . . , cr are complex numbers (independent of n). Hence from (3.9) we see
the limit as nm → ∞ of the right-hand
that side approaches Ir . It follows [why?] that
−2 .
n≥0 (n + 1)B converges to (Ir − B)
NOTE. The above argument shows that Ir − B is indeed invertible. This fact
is also an immediate consequence of the hypothesis that all eigenvalues of B have
absolute value less than one, since in particular there is no eigenvalue λ = 1.
From the discussion above, it remains to show that M[v] is diagonalizable, with
all eigenvalues of absolute value less than one. The diagonalizability of M[v] is
shown in exactly the same way as for M in Theorem 3.2. (Thus we see also that
M[v] has real eigenvalues, though we don’t need this fact here.) It remains to show
that the eigenvalues θ1 , . . . , θp−1 of M[v] satisfy |θj | < 1. We would like to apply
Theorem 3.3 to the matrix M[v], but this matrix might not be irreducible since the
graph G−v (defined by deleting from G the vertex v and all incident edges) need not
be connected or may be just an isolated vertex. If G − v has connected components
H1 , . . . , Hm , then we can order the vertices of G − v so that M[v] has the block
structure
⎡ ⎤
N1 0 · · · 0
⎢ 0 N2 · · · 0 ⎥
⎢ ⎥
M[v] = ⎢ .. ⎥,
⎣ . ⎦
0 0 · · · Nm
where each N i is irreducible or is the 1 × 1 matrix [0] (corresponding to Hi being

an isolated vertex). The eigenvalues of M[v] are the eigenvalues of the N i ’s.
26 3 Random Walks
We need to show that each eigenvalue of N i has absolute value less than one. If
Ni = [0] then the only eigenvalue is 0, so we may assume that Hi is not an isolated
vertex. Suppose that Hi has k vertices, so N i is a k × k matrix. Let ρi be the largest
real eigenvalue of N i , so by Theorem 3.3 all eigenvalues λ of N i satisfy |λ| ≤ ρi .
Let U = [u1 , . . . , uk ] be a left eigenvector for ρi with positive entries (which exists
by Theorem 3.3), so U N i = ρi U . Let V be the column vector of length k of all 1’s.
Consider the matrix product U N i V . On the one hand we have
U N i V = (ρi U )V = ρi (u1 + · · · + uk ). (3.10)
On the other hand, if σj denotes the j th row sum of N i , then
U N i V = U [σ1 , . . . , σk ]t = σ1 u1 + · · · + σk uk , (3.11)
where t denotes transpose. Now every σj satisfies 0 ≤ σj ≤ 1, and at least one σh

satisfies σh < 1 [why?]. Since each uj > 0, it follows from (3.11) that U N i V <
u1 + · · · + uk . Comparing with (3.10) gives ρi < 1.
Since the eigenvalues of M[v] are just the eigenvalues of the N i ’s, we see that
all eigenvalues θ of M[v] satisfy |θ | < 1. This completes the proof of Theorem 3.4.
3.5 Example. Let G be the graph of Figure 3.1 with v = v4 . Then

⎡ ⎤
1 1 1
⎢3 3 0
⎢ 3⎥⎥
⎢1 ⎥
⎢ 0 1 1⎥
⎢ ⎥
⎢4 4 2⎥
M=⎢ ⎥
⎢ 1 1⎥
⎢0 ⎥
⎢ 2 0 2⎥
⎢ ⎥
⎣1 1 1 ⎦
0
4 2 4
⎡ 2 1 ⎤
− 0
⎢ 3 3 ⎥
⎢ ⎥
⎢ 1 1⎥
I3 − M[v] = ⎢ −
⎢ 4 1 − ⎥
⎢ 4⎥⎥
⎣ ⎦
1
0− 1
2
⎡ 55 13 17 ⎤
⎢ 16 6 24 ⎥
⎢ ⎥
⎢ 13 7 11 ⎥
(I3 − M[v])−2 =⎢
⎢ 8 3
⎥
⎢ 12 ⎥
⎥
⎣ ⎦
17 11 13
16 6 8
Fig. 3.1 A graph for v2

Example 3.5
v3
v1
v4
⎡1⎤ ⎡ 31 ⎤
⎢ 3 ⎥ ⎢ 12 ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ 13 ⎥
−2 ⎢ 1 ⎥
(I3 − M[v]) ⎢ ⎥ = ⎢ ⎢ ⎥.
⎥
⎢2⎥ ⎢ 6 ⎥
⎣ ⎦ ⎣ ⎦
1 25
2 12
Thus H (v1 , v) = 31/12, H (v2 , v) = 13/6, and H (v3 , v) = 25/12.

NOTE. The method used to prove that n≥0 (n + 1)B n converges when all
eigenvalues of B have absolute value less than one can be extended, with a little
more work (mostly
concerned with non-diagonalizability), to show the following.
Let F (x) = n≥0 an x n be a power series with complex coefficients an . Let α > 0
be such that F (x) converges whenever |x| < α. Let B be a square matrix (over the
numbers) whose eigenvalues λ all satisfy |λ| < α. Then the matrix power
complex
series n≥0 an B n converges in the entry-wise sense described above.
Notes for Chapter 3
Random walks on graphs is a vast subject, of which we have barely scratched the
surface. Two typical questions considerably deeper than what we have considered
are the following: how rapidly does a random walk approach the stationary
distribution of Exercise 3.1? Assuming G is connected, what is the expected number
of steps needed to visit every vertex? For a nice survey of random walks in graphs,
see Lovász [84]. The topic of matrix power series is part of the subject of matrix
analysis. For further information, see for instance Chapter 5 of the text by Horn
and Johnson [68]. Our proof of Theorem 3.4 is somewhat “naive,” avoiding the
development of the theory of matrix norms.
1. Let G be a (finite) graph with vertices v1 , . . . , vp . Assume that some power of

the probability matrix M(G) defined by (3.1) has positive entries. (It’s not hard
to see that this is equivalent to G being connected and containing at least one
28 3 Random Walks
cycle of odd length, but you don’t have to show this.) Let dk denote the degree
(number of incident edges) of vertex vk . Let D = d1 + d2 + · · · + dp = 2q − r,
where G has q edges and r loops. Start at any vertex of G and do a random walk
on the vertices of G as defined in the text. Let pk () denote the probability of
ending up at vertex vk after steps. Assuming the Perron–Frobenius theorem
(Theorem 3.3), show that
lim pk () = dk /D.

→∞
This limiting probability distribution on the set of vertices of G is called the

stationary distribution of the random walk.
2. (a) Let G be a finite graph (allowing loops and multiple edges). Suppose that
there is some integer > 0 such that the number of walks of length from
any fixed vertex u to any fixed vertex v is independent of u and v. Show
that G has the same number k of edges between any two vertices (including
k loops at each vertex).
(b) Let G be a finite graph (allowing loops and multiple edges) with the
following property. There is some integer > 0 such that if we start at
any vertex of G and do a random walk (in the sense of the text) for steps,
then we are equally likely to be at any vertex. In other words, if G has p
vertices then the probability that the walk ends at vertex v is exactly 1/p
for any v. Show that we have the same conclusion as (a), i.e., G has the
same number k of edges between any two vertices.
3. (a) Let P (x) be a nonzero polynomial with real coefficients. Show that the
following two conditions are equivalent.
• There exists a nonzero polynomial Q(x) with real coefficients such that
all coefficients of P (x)Q(x) are nonnegative.
• There does not exist a real number a > 0 such that P (a) = 0.
(b) (difficult) Let G be a connected finite graph, and let M be the probability
matrix defined by (3.1). Show that the following two conditions are
equivalent.
• There exists a probability distribution P on P (so P (k) is the probability
of choosing k ∈ P) such that if we first choose k from the distribution P
and then start at any vertex of G and walk exactly k steps according to
the random walk described in the text, then we are equally likely to be
at any vertex of G.
• The graph G is regular, and no positive real number except 1 is an
eigenvalue of M.
4. (*) Fix 0 ≤ p ≤ 1. Start at the vertex (0, 0, . . . , 0) of the n-cube Cn . Walk
along the edges of the cube according to the following rule: after each unit
of time, either stay where you are with probability p or step to a neighboring
vertex randomly (uniformly). Thus the probability of stepping to a particular
neighboring vertex is (1 − p)/n. Find a formula for the probability P () that
after units of time you are again at (0, 0, . . . , 0). For instance, P (0) = 1 and
P (1) = p. Express your formula as a finite sum.
5. This problem is not directly related to the text but is a classic problem with a
very clever elegant solution. Let G be the graph with vertex set Zn (the integers
modulo n), with an edge between i and i + 1 for all i ∈ Zn . Hence G is just an
n-cycle. Start at vertex 0 and do a random walk as in the text, so from vertex
i walk to i − 1 or i + 1 with probability 1/2 each. For each i ∈ Zn , find the
probability that vertex i is the last vertex to be visited for the first time. In other
words, at the first time we arrive at vertex i, we have visited all the other vertices
at least once each. For instance, p0 = 0 (if n > 1), since vertex 0 is the first
vertex to be visited.
6. Let G be the 3 × 3 “grid graph” of Example 10.7. Using your favorite software
for linear algebra, compute the access times H (u, v) for every pair u, v of
vertices of G.
7. (a) Show that if u and v are two vertices of a connected graph G, then we need
not have H (u, v) = H (v, u), where H denotes access time. What if G is
also assumed to be regular?
(b) (difficult) For each n ≥ 1, what is the maximum possible value of
H (u, v) − H (v, u) for two vertices u, v of a connected simple graph with
n vertices?
8. (*) Let u and v be distinct vertices of the complete graph Kn . Show that
H (u, v) = n − 1.
9. (*) Let Pn be the graph with vertices v1 , . . . , vn and an edge between vi and
vi+1 for all 1 ≤ i ≤ n − 1. Show that H (v1 , vn ) = (n − 1)2 . What about
H (vi , vj ) for any i = j ? What if we also have an edge between v1 and vn ?
10. Let Kmn be a complete bipartite graph with vertex bipartition (A1 , A2 ), where
#A1 = m and #A2 = n. Find the access time H (u, v) between every pair of
distinct vertices. There will be two inequivalent cases: both u and v lie in the
same Ai , or they lie in different Ai ’s.
11. (*) For any three vertices u, v, w of a graph G, show that
H (u, v) + H (v, w) + H (w, u) = H (u, w) + H (w, v) + H (v, u).
12. Let k ≥ 0, and let u and v be vertices of a graph G. Define the kth binomial
moment
n Hk (u, v) of the access time to be the average value (expectation) of
k , where n is the number of steps that a random walk starting at u takes to
reach v for the first time. Thus in the notation of (3.2) we have
n
Hk (u, v) = pn .
k
n≥1
30 3 Random Walks
Let x be an indeterminate. Following the notation of (3.4), show that

Hk (u, v)x k = ((Ip−1 − (x + 1)M[v])−1 T [v])u .
k≥0
13. (*) Generalizing Exercise 3.8 above, show that for any two distinct vertices u, v
of the complete graph Kn , the kth binomial moment of the access time is given
by Hk (u, v) = (n − 1)(n − 2)k−1 , k ≥ 1. (When n = 2 and k = 1, we should
set 00 = 1.)
Chapter 4
The Sperner Property
In this chapter we consider a surprising application of certain adjacency matrices

to some problems in extremal set theory. An important role will also be played
by finite groups in Chapter 5, which is a continuation of the present chapter. In
general, extremal set theory is concerned with finding (or estimating) the most or
least number of sets satisfying given set-theoretic or combinatorial conditions. For
example, a typical easy problem in extremal set theory is the following: what is the
most number of subsets of an n-element set with the property that any two of them
intersect? (Can you solve this problem?) The problems to be considered here are
most conveniently formulated in terms of partially ordered sets or posets for short.
Thus we begin with discussing some basic notions concerning posets.
4.1 Definition. A poset P is a finite set, also denoted P , together with a binary
relation denoted ≤ satisfying the following axioms:
(P1) (reflexivity) x ≤ x for all x ∈ P
(P2) (antisymmetry) If x ≤ y and y ≤ x, then x = y
(P3) (transitivity) If x ≤ y and y ≤ z, then x ≤ z
One easy way to obtain a poset is the following. Let P be any collection of sets.
If x, y ∈ P , then define x ≤ y in P if x ⊆ y as sets. It is easy to see that this
definition of ≤ makes P into a poset. If P consists of all subsets of an n-element
set S, then P is called a (finite) Boolean algebra of rank n and is denoted by BS . If
S = {1, 2, . . . , n}, then we denote BS simply by Bn . Boolean algebras will play an
important role throughout this chapter and the next.
There is a simple way to represent small posets pictorially. The Hasse diagram
of a poset P is a planar drawing, with elements of P drawn as dots. If x < y in
P (i.e., x ≤ y and x = y), then y is drawn “above” x (i.e., with a larger vertical
coordinate). An edge is drawn between x and y if y covers x, i.e., x < y and no
element z satisfies x < z < y. We then write x y or y x. By the transitivity
property (P3), all the relations of a finite poset are determined by the cover relations,

https://doi.org/10.1007/978-3-319-77173-1_4
32 4 The Sperner Property
so the Hasse diagram determines P . (This is not true for infinite posets; for instance,
the real numbers R with their usual order is a poset with no cover relations.) The
Hasse diagram of the Boolean algebra B3 looks like
123
12 13 23
1 2 3
φ
We say that two posets P and Q are isomorphic if there is a bijection (one-to-one
and onto function) ϕ : P → Q such that x ≤ y in P if and only if ϕ(x) ≤ ϕ(y) in
Q. Thus one can think that two posets are isomorphic if they differ only in the names
of their elements. This is exactly analogous to the notion of isomorphism of groups,
rings, etc. It is an instructive exercise (see Exercise 4.1) to draw Hasse diagrams
of the one poset of order (number of elements) one (up to isomorphism), the two
posets of order two, the five posets of order three, and the sixteen posets of order
four. More ambitious readers can try the 63 posets of order five, the 318 of order
six, the 2,045 of order seven, the 16,999 of order eight, the 183,231 of order nine,
the 2,567,284 of order ten, the 46,749,427 of order eleven, the 1,104,891,746 of
order twelve, the 33,823,827,452 of order thirteen, the 1,338,193,159,771 of order
fourteen, the 68,275,077,901,156 of order fifteen, and the 4,483,130,665,195,087 of
order sixteen. Beyond this the number is not currently known.
A chain C in a poset is a totally ordered subset of P , i.e., if x, y ∈ C then either
x ≤ y or y ≤ x in P . A finite chain is said to have length n if it has n + 1 elements.
Such a chain thus has the form x0 < x1 < · · · < xn . We say that a finite poset
is graded of rank n if every maximal chain has length n. (A chain is maximal if
it’s contained in no larger chain.) For instance, the Boolean algebra Bn is graded
of rank n [why?]. A chain y0 < y1 < · · · < yj is said to be saturated if each
yi+1 covers yi . Such a chain need not be maximal since there can be elements of
P less than y0 or greater than yj . If P is graded of rank n and x ∈ P , then we say
that x has rank j , denoted ρ(x) = j , if the largest saturated chain of P with top
element x has length j . Thus [why?] if we let Pj = {x ∈ P : ρ(x) = j }, then P is
a disjoint union P = P0 ∪P · 1 ∪· · · · ∪P
· n , and every maximal chain of P has the form
x0 < x1 < · · · < xn where ρ(xj ) = j . We call Pj the j th level of P . We write
pj = #Pj , the number of elements of P of rank j . For example, if P = Bn then
ρ(x) = |x| (the cardinality of x as a set) and
4 The Sperner Property 33

n
pj = #{x ⊆ {1, 2, . . . , n} : |x| = j } = .
j
(Note that we use both |S| and #S for the cardinality of a finite set S.) If a graded
poset P of rank n has pi elements of rank i, then define the rank-generating function

n
F (P , q) = pi q i = q ρ(x) .
i=0 x∈P
For instance, F (Bn , q) = (1 + q)n [why?].

We say that a graded poset P of rank n (always assumed to be finite) is rank-
symmetric if pi = pn−i for 0 ≤ i ≤ n and rank-unimodal if p0 ≤ p1 ≤ · · · ≤ pj ≥
pj +1 ≥ pj +2 ≥ · · · ≥ pn for some 0 ≤ j ≤ n. If P is both rank-symmetric and
rank-unimodal, then we clearly have
p0 ≤ p1 ≤ · · · ≤ pm ≥ pm+1 ≥ · · · ≥ pn , if n = 2m
p0 ≤ p1 ≤ · · · ≤ pm = pm+1 ≥ pm+2 ≥ · · · ≥ pn , if n = 2m + 1.
We also say that the sequence p0 , p1 , . . . , pn itself or the polynomial F (q) = p0 +

p1 q + · · · + pn q n is symmetric or unimodal, as the case may be. For instance, Bn is
rank-symmetric and rank-unimodal,
since it is well-known (and easy to prove) that
the sequence n0 , n1 , . . . , nn (the nth row of Pascal’s triangle) is symmetric and
unimodal. Thus the polynomial (1 + q)n is symmetric and unimodal.
A few more definitions, and then finally some results! An antichain in a poset P
is a subset A of P for which no two elements are comparable, i.e., we can never have
x, y ∈ A and x < y. For instance, in a graded poset P the “levels” Pj are antichains
[why?]. We will be concerned with the problem of finding the largest antichain in
a poset. Consider for instance the Boolean algebra Bn . The problem of finding the
largest antichain in Bn is clearly equivalent to the following problem in extremal set
theory: find the largest collection of subsets of an n-element set such that no element
of the collection contains another (as a subset). A good guess would be to take all the
subsets of cardinality n/2 (where t denotes the greatest integer ≤ t), giving a
n
total of n/2 sets in all. But how can we actually prove there is no larger collection?
Such a proof was first given by Emanuel Sperner in 1927 and is known as Sperner’s
theorem. We will give three proofs of Sperner’s theorem in this chapter: one proof
uses linear algebra and will be applied to certain other situations; the second proof
is an elegant combinatorial argument due to David Lubell in 1966; while the third
proof is another combinatorial argument closely related to the linear algebra proof.
We present the last two proofs for their “cultural value.” Our extension of Sperner’s
theorem to certain other situations will involve the following crucial definition.
4.2 Definition. Let P be a graded poset of rank n. We say that P has the Sperner
property or is a Sperner poset if
max{#A : A is an antichain of P } = max{#Pi : 0 ≤ i ≤ n}.
In other words, no antichain is larger than the largest level Pi .

Thus Sperner’s theorem is equivalent to saying that Bn has the Sperner property.
Note that if P has the Sperner property then there may still be antichains of
maximum cardinality other than the biggest Pi ; there just can’t be any bigger
antichains.
4.3 Example. A simple example of a graded poset that fails to satisfy the Sperner
property is the following:
We now will discuss a simple combinatorial condition which guarantees that

certain graded posets P are Sperner. We define an order-matching from Pi to Pi+1
to be a one-to-one function μ : Pi → Pi+1 satisfying x < μ(x) for all x ∈ Pi .
Clearly if such an order-matching exists then pi ≤ pi+1 (since μ is one-to-one).
Easy examples (such as the diagram above) show that the converse is false, i.e., if
pi ≤ pi+1 then there need not exist an order-matching from Pi to Pi+1 . We similarly
define an order-matching from Pi to Pi−1 to be a one-to-one function μ : Pi → Pi−1
satisfying μ(x) < x for all x ∈ Pi .
4.4 Proposition. Let P be a graded poset of rank n. Suppose there exists an integer
0 ≤ j ≤ n and order-matchings
P0 → P1 → P2 → · · · → Pj ← Pj +1 ← Pj +2 ← · · · ← Pn . (4.1)
Then P is rank-unimodal and Sperner.
Proof. Since order-matchings are one-to-one it is clear that
p0 ≤ p1 ≤ · · · ≤ pj ≥ pj +1 ≥ pj +2 ≥ · · · ≥ pn .
Hence P is rank-unimodal.
Define a graph G as follows. The vertices of G are the elements of P . Two
vertices x, y are connected by an edge if one of the order-matchings μ in the
statement of the proposition satisfies μ(x) = y. (Thus G is a subgraph of the
Hasse diagram of P .) Drawing a picture will convince you that G consists of a
disjoint union of paths, including single-vertex paths not involved in any of the
order-matchings. The vertices of each of these paths form a chain in P . Thus we
have partitioned the elements of P into disjoint chains. Since P is rank-unimodal
with biggest level Pj , all of these chains must pass through Pj [why?]. Thus the
number of chains is exactly pj . Any antichain A can intersect each of these chains
at most once, so the cardinality |A| of A cannot exceed the number of chains, i.e.,
|A| ≤ pj . Hence by definition P is Sperner.
It is now finally time to bring some linear algebra into the picture. For any
(finite) set S, we let RS denote the real vector space consisting of all formal linear
combinations (with real coefficients) of elements of S. Thus S is a basis for RS, and
in fact we could have simply defined RS to be the real vector space with basis S.
The next lemma relates the combinatorics we have just discussed to linear algebra
and will allow us to prove that certain posets are Sperner by the use of linear algebra
(combined with some finite group theory).
4.5 Lemma. Suppose there exists a linear transformation U : RPi → RPi+1 (U
stands for “up”) satisfying:
• U is one-to-one.
• For all x ∈ Pi , U (x) is a linear combination of elements y ∈ Pi+1 satisfying
x < y. (We then call U an order-raising operator.)
Then there exists an order-matching μ : Pi → Pi+1 .
Similarly, suppose there exists a linear transformation U : RPi → RPi+1
satisfying:
• U is onto.
• U is an order-raising operator.
Then there exists an order-matching μ : Pi+1 → Pi .
Proof. Suppose U : RPi → RPi+1 is a one-to-one order-raising operator. Let [U ]

denote the matrix of U with respect to the bases Pi of RPi and Pi+1 of RPi+1 . Thus
the rows of [U ] are indexed by the elements y1 , . . . , ypi+1 of Pi+1 (in some order)
and the columns by the elements x1 , . . . , xpi of Pi . Since U is one-to-one, the rank
of [U ] is equal to pi (the number of columns). Since the row rank of a matrix equals
its column rank, [U ] must have pi linearly independent rows. Say we have labelled
the elements of Pi+1 so that the first pi rows of [U ] are linearly independent.
Let A = (aij ) be the pi ×pi matrix whose rows are the first pi rows of [U ]. (Thus
A is a square submatrix of [U ].) Since the rows of A are linearly independent, we
have

det(A) = ±a1π(1) · · · api π(pi ) = 0,
where the sum is over all permutations π of 1, . . . , pi . Thus some term

±a1π(1) · · · api π(pi ) of the above sum is nonzero. Since U is order-raising, this
means that [why?] yk > xπ(k) for 1 ≤ k ≤ pi . Hence the map μ : Pi → Pi+1
defined by μ(xk ) = yπ −1 (k) is an order-matching, as desired.
The case when U is onto rather than one-to-one is proved by a completely
analogous argument. It can also be deduced from the one-to-one case by considering
the transpose of the matrix [U ].
NOTE. Although it does not really help in understanding the theory, it is

interesting to regard a one-to-one order-raising operator as a “quantum order-
matching.” Rather than choosing a single element y = μ(x) that is matched with
x ∈ Pi , we choose
all possible elements y ∈ Pi+1 satisfying y > x at the same
time. If U (x) = y>x cy y (where cy ∈ R), then we are choosing y with “weight”
cy . As explained in the proof of Lemma 4.5 above, we “break the symmetry” and
obtain a single matched element μ(x) by choosing some nonvanishing term in the
expansion of a determinant.
We now want to apply Proposition 4.4 and Lemma 4.5 to the Boolean algebra
Bn . For each 0 ≤ i < n, we need to define a linear transformation Ui : R(Bn )i →
R(Bn )i+1 , and then prove it has the desired properties. We simply define Ui to be
the simplest possible order-raising operator, namely, for x ∈ (Bn )i , let

Ui (x) = y. (4.2)
y∈(Bn )i+1
y>x
Note that since (Bn )i is a basis for R(Bn )i , (4.2) does indeed define a unique linear
transformation Ui : R(Bn )i → R(Bn )i+1 . By definition Ui is order-raising; we
want to show that Ui is one-to-one for i < n/2 and onto for i ≥ n/2. There are
several ways to show this using only elementary linear algebra; we will give what is
perhaps the simplest proof, though it is quite tricky. The idea is to introduce “dual”
or “adjoint” operators Di : R(Bn )i → R(Bn )i−1 to the Ui ’s (D stands for “down”),
defined by

Di (y) = x, (4.3)
x∈(Bn )i−1
x<y
for all y ∈ (Bn )i . Let [Ui ] denote the matrix of Ui with respect to the bases (Bn )i
and (Bn )i+1 , and similarly let [Di ] denote the matrix of Di with respect to the bases
(Bn )i and (Bn )i−1 . A key observation which we will use later is that
[Di ] = [Ui−1 ]t , (4.4)
i.e., the matrix [Di ] is the transpose of the matrix [Ui−1 ] [why?]. Now let
Ii : R(Bn )i → R(Bn )i denote the identity transformation on R(Bn )i , i.e., Ii (u) =
u for all u ∈ R(Bn )i . The next lemma states (in linear algebraic terms) the
fundamental combinatorial property of Bn which we need. For this lemma set
Un = 0 and D0 = 0 (the 0 linear transformation between the appropriate vector
spaces).
4.6 Lemma. Let 0 ≤ i ≤ n. Then
Di+1 Ui − Ui−1 Di = (n − 2i)Ii . (4.5)
(Linear transformations are multiplied right-to-left, so AB(u) = A(B(u)).)

Proof. Let x ∈ (Bn )i . We need to show that if we apply the left-hand side of (4.5)
to x, then we obtain (n − 2i)x. We have
⎛⎞
⎜ ⎟
Di+1 Ui (x) = Di+1 ⎝ y⎠
|y|=i+1
x⊂y

= z.
|y|=i+1 |z|=i
x⊂y z⊂y
If x, z ∈ (Bn )i satisfy |x ∩ z| < i − 1, then there is no y ∈ (Bn )i+1 such that x ⊂ y

and z ⊂ y. Hence the coefficient of z in Di+1 Ui (x) when it is expanded in terms of
the basis (Bn )i is 0. If |x ∩ z| = i − 1, then there is one such y, namely, y = x ∪ z.
Finally if x = z then y can be any element of (Bn )i+1 containing x, and there are
n − i such y in all. It follows that

Di+1 Ui (x) = (n − i)x + z. (4.6)
|z|=i
|x∩z|=i−1
By exactly analogous reasoning (which the reader should check), we have for x ∈
(Bn )i that

Ui−1 Di (x) = ix + z. (4.7)
|z|=i
|x∩z|=i−1
Subtracting (4.7) from (4.6) yields (Di+1 Ui − Ui−1 Di )(x) = (n − 2i)x, as desired.
4.7 Theorem. The operator Ui defined above is one-to-one if i < n/2 and is onto
if i ≥ n/2.
Proof. Recall that [Di ] = [Ui−1 ]t . From linear algebra we know that a (rectangular)
matrix times its transpose is positive semidefinite (or just semidefinite for short)
and hence has nonnegative (real) eigenvalues. In particular, by (4.4) the matrix
[Ui−1 ][Di ] is semidefinite. By Lemma 4.6 we have
Di+1 Ui = Ui−1 Di + (n − 2i)Ii .
Thus the eigenvalues of Di+1 Ui are obtained from the eigenvalues of Ui−1 Di by
adding n−2i. Since the eigenvalues of Ui−1 Di are nonnegative (by the semidefinite
property) and we are assuming that n − 2i > 0, it follows that the eigenvalues
of Di+1 Ui are strictly positive. Hence Di+1 Ui is invertible (since it has no 0
eigenvalues). But this implies that Ui is one-to-one [why?], as desired.
The case i ≥ n/2 is done by a “dual” argument (or in fact can be deduced directly
from the i < n/2 case by using the fact that the poset Bn is “self-dual,” though we
will not go into this). Namely, from the fact that
Ui Di+1 = Di+2 Ui+1 + (2i + 2 − n)Ii+1
we get that Ui Di+1 is invertible, so now Ui is onto, completing the proof.
Combining Proposition 4.4, Lemma 4.5, and Theorem 4.7, we obtain the famous
theorem of Sperner.
4.8 Corollary. The Boolean algebra Bn has the Sperner property.
It is natural to ask whether there is a less indirect proof of Corollary 4.8. In fact,
several nice proofs are known; we first give one due to David Lubell, mentioned
before Definition 4.2.
Lubell’s Proof of Sperner’s Theorem. First we count the total number of maxi-
mal chains ∅ = x0 < x1 < · · · < xn = {1, . . . , n} in Bn . There are n choices for x1 ,
then n − 1 choices for x2 , etc., so there are n! maximal chains in all. Next we count
the number of maximal chains x0 < x1 < · · · < xi = x < · · · < xn which contain
a given element x of rank i. There are i choices for x1 , then i − 1 choices for x2 ,
up to one choice for xi . Similarly there are n − i choices for xi+1 , then n − i − 1
choices for xi+2 , etc., up to one choice for xn . Hence the number of maximal chains
containing x is i!(n − i)!.
Now let A be an antichain. If x ∈ A, then let Cx be the set of maximal chains
of Bn which contain x. Since A is an antichain, the sets Cx , x ∈ A are pairwise
disjoint. Hence

Cx = |Cx |

x∈A x∈A

= (ρ(x))!(n − ρ(x))!.
x∈A
Since the total number of maximal chains in the Cx ’s cannot exceed the total number
n! of maximal chains in Bn , we have

(ρ(x))!(n − ρ(x))! ≤ n!.
x∈A
Divide both sides by n! to obtain

1

n ≤ 1.
x∈A ρ(x)
n
Since i is maximized when i = n/2, we have
1 1
n ≤ n ,
n/2 ρ(x)
for all x ∈ A (or all x ∈ Bn ). Thus

1

n ≤ 1,
x∈A n/2
or equivalently,

n
|A| ≤ .
n/2
n
Since n/2 is the size of the largest level of Bn , it follows that Bn is Sperner.
There is another nice way to show directly that Bn is Sperner, namely, by

constructing an explicit order-matching μ : (Bn )i → (Bn )i+1 when i < n/2.
We will define μ by giving an example. Let n = 21, i = 9, and S =
{3, 4, 5, 8, 12, 13, 17, 19, 20}. We want to define μ(S). Let (a1 , a2 , . . . , a21 ) be a
sequence of ±1’s, where ai = 1 if i ∈ S, and ai = −1 if i ∈ S. For the set S above
we get the sequence (writing − for −1)
− − 111 − −1 − − − 11 − − − 1 − 11 − .
Replace any two consecutive terms 1 − with 0 0:
− − 1 1 0 0 − 0 0 − − 1 0 0 − − 0 0 1 0 0.
Ignore the 0’s and replace any two consecutive terms 1 − with 0 0:
− − 1 0 0 0 0 0 0 − − 0 0 0 0 − 0 0 1 0 0.
Continue:
− − 0 0 0 0 0 0 0 0 − 0 0 0 0 − 0 0 1 0 0.
At this stage no further replacement is possible. The nonzero terms consist of a

sequence of −’s followed by a sequence of 1’s. There is at least one − since i < n/2.
Let k be the position (coordinate) of the last −; here k = 16. Define μ(S) = S ∪
{k} = S ∪{16}. The reader can check that this procedure gives an order-matching. In
particular, why is μ injective (one-to-one), i.e., why can we recover S from μ(S)?
It can be checked that if we glue together the order-matchings (Bn )i → (Bn )i+1
for i < n/2 just defined, along with an obvious dual construction (Bn )i → (Bn )i−1
for i > n/2 then we obtain more than just a partition of Bn into saturated chains
passing through the middle level (n even) or middle two levels (n odd), as in
the proof of Proposition 4.4. We in fact have the additional property that these
chains are all symmetric, i.e., they begin at some level i ≤ n/2 and end at level
n − i. Such a decomposition of a rank-symmetric, rank-unimodal graded poset
P into saturated chains is called a symmetric chain decomposition. A symmetric
chain decomposition implies that for any j ≥ 1, the largest size of a union of j
antichains is equal to the largest size of a union of j levels of P (Exercise 4.6).
(The Sperner property corresponds to the case j = 1). It can be a challenging
problem to decide whether certain posets have a symmetric chain decomposition
(e.g., Exercises 5.5(b), 5.6 and 6.6), though we will not discuss this topic further
here.
In view of the above elegant proof of Lubell and the explicit description of an
order-matching μ : (Bn )i → (Bn )i+1 , the reader may be wondering what was
the point of giving a rather complicated and indirect proof using linear algebra.
Admittedly, if all we could obtain from the linear algebra machinery we have
developed was just another proof of Sperner’s theorem, then it would have been
hardly worth the effort. But in the next chapter we will show how Theorem 4.7, when
combined with a little finite group theory, can be used to obtain many interesting
combinatorial results for which simple, direct proofs are not known.
Notes for Chapter 4
For further information on combinatorial aspects of partially ordered sets in general,

see Caspard–Leclerc–Monjardet [21], Fishburn [40], Stanley [130, Ch. 3], and
Trotter [136]. Sperner’s theorem (Corollary 4.8) was first proved by Sperner [119].
The elegant proof of Lubell appears in [85]. A general reference on the Sperner
property is the book by Engel [39]. For more general results on the combinatorics of
finite sets, see Anderson [1]. The linear algebraic approach to the Sperner property
discussed here is due independently to Pouzet [103] (further developed by Pouzet
and Rosenberg [104]) and Stanley [123, 126]. For further information on explicit
order matchings, symmetric chain decompositions, etc., see the text [1] of Anderson
mentioned above.
1. Draw Hasse diagrams of the 16 nonisomorphic four-element posets. For a more

interesting challenge, draw also the 63 five-element posets. For those with lots of
time to kill, draw the 318 six-element posets, the 2,045 seven-element posets, the
16,999 eight-element posets, up to the 4,483,130,665,195,087 sixteen-element
posets.
2. (a) Let P be a finite poset and f : P → P an order-preserving bijection. That is,
f is a bijection (one-to-one and onto), and if x ≤ y in P then f (x) ≤ f (y).
Show that f is an automorphism of P , i.e., f −1 is order-preserving. (Try to
use simple algebraic reasoning, though it’s not necessary to do so.)
(b) Show that the result of (a) need not be true if P is infinite.
3. Let F (q) and G(q) be symmetric unimodal polynomials with nonnegative
real coefficients. Show that F (q)G(q) is also symmetric (easy) and unimodal
(harder).
4. Let q be a prime power, and let Fq denote the finite field with q elements.
Let V = Vn (q) = Fnq , the n-dimensional vector space over Fq of n-tuples of
elements of Fq . Let Bn (q) denote the poset of all subspaces of V , ordered by
inclusion. It’s easy to see that Bn (q) is graded of rank n, the rank of a subspace
of V being its dimension.
(a) Draw the Hasse diagram of B3 (2). (It has 16 elements.)
(b) (*) Show that the number of elements of Bn (q) of rank k is given by the
q-binomial coefficient

n (q n − 1)(q n−1 − 1) · · · (q n−k+1 − 1)
= .
k (q k − 1)(q k−1 − 1) · · · (q − 1)
(c) (*) Show that Bn (q) is rank-symmetric.

(d) Show that every element x ∈ Bn (q)k covers [k] = 1 + q + · · · + q k−1
elements and is covered by [n − k] = 1 + q + · · · + q n−k−1 elements.
(e) Define operators Ui : RBn (q)i → RBn (q)i+1 and Di : RBn (q)i →
RBn (q)i−1 by

Ui (x) = y
y∈Bn (q)i+1
y>x

Di (x) = z.
z∈Bn (q)i−1
z<x
Show that
Di+1 Ui − Ui−1 Di = ([n − i] − [i])Ii .
(f) Deduce that Bn (q) is rank-unimodal and Sperner.

5. (difficult) Let S1 , S2 , . . . , Sk be finite sets with #S1 = #S2 = · · · = #Sk . Let P
be the poset of all sets T contained in some Si , ordered by inclusion. In symbols,
P = 2S1 ∪ 2S2 ∪ · · · ∪ 2Sk ,
where 2S denotes the set of subsets of S. Is P always rank-unimodal?

6. Let P be a rank-symmetric, rank-unimodal poset. Show that if P has a symmetric
chain decomposition, then for any j ≥ 1 the largest size of a union of j antichains
is equal to the largest size of a union of j levels of P .
7. (a) Let P be a finite poset whose largest chain has m elements. Show that P is a
union of m antichains.
(b) (difficult) (*) Let P be a finite poset whose largest antichain has m elements.
Show that P is a union of m chains.
Chapter 5
Group Actions on Boolean Algebras
Let us begin by reviewing some facts from group theory. Suppose that X is an n-
element set and that G is a group. We say that G acts on the set X if for every
element π of G we associate a permutation (also denoted π ) of X, such that for all
x ∈ X and π, σ ∈ G we have
π(σ (x)) = (π σ )(x).
Thus [why?] an action of G on X is the same as a homomorphism ϕ : G → SX ,

where SX denotes the symmetric group of all permutations of X. We sometimes
write π · x instead of π(x).
5.1 Example. (a) Let the real number α act on the xy-plane by rotation coun-
terclockwise around the origin by an angle of α radians. It is easy to check
that this defines an action of the group R of real numbers (under addition) on
the xy-plane. The kernel of this action, i.e., the kernel of the homomorphism
ϕ : R → SR2 , is the cyclic subgroup of R generated by 2π .
(b) Now let α ∈ R act by translation by a distance α to the right, i.e., adding (α, 0).
This yields a completely different action of R on the xy-plane. This time the
action is faithful, i.e., the kernel is the trivial subgroup {0}.
(c) Let X = {a, b, c, d} and G = Z2 × Z2 = {(0, 0), (0, 1), (1, 0), (1, 1)}. Let G
act as follows:
(0, 1) · a = b, (0, 1) · b = a, (0, 1) · c = c, (0, 1) · d = d
(1, 0) · a = a, (1, 0) · b = b, (1, 0) · c = d, (1, 0) · d = c.
The reader should check that this does indeed define an action. In particular,
since (1, 0) and (0, 1) generate G, we don’t need to define the action of (0, 0)
and (1, 1)—they are uniquely determined.
(d) Let X and G be as in (c), but now define the action by

https://doi.org/10.1007/978-3-319-77173-1_5
44 5 Group Actions on Boolean Algebras
(0, 1) · a = b, (0, 1) · b = a, (0, 1) · c = d, (0, 1) · d = c

(1, 0) · a = c, (1, 0) · b = d, (1, 0) · c = a, (1, 0) · d = b.
Again one can check that we have an action of Z2 × Z2 on {a, b, c, d}. The
two actions of G = Z2 × Z2 that we have just defined are quite different;
for instance, in the first action we have some elements of X fixed by some
nonidentity element of G (such as (0, 1) · c = c), while the second action fails
to have this property. See also Example 5.2(c, d) below for another fundamental
way in which the two actions differ.
Recall what is meant by an orbit of the action of a group G on a set X. Namely,
we say that two elements x, y of X are G-equivalent if π(x) = y for some π ∈ G.
The relation of G-equivalence is an equivalence relation, and the equivalence classes
are called orbits. Thus x and y are in the same orbit if π(x) = y for some π ∈ G.
The orbits form a partition of X, i.e., they are pairwise-disjoint, nonempty subsets of
X whose union is X. The orbit containing x is denoted Gx; this is sensible notation
since Gx consists of all elements π(x) where π ∈ G. Thus Gx = Gy if and only
if x and y are G-equivalent (i.e., in the same G-orbit). The set of all G-orbits is
denoted X/G.
5.2 Example. (a) In Example 5.1(a), the orbits are circles with center (0, 0),
including the degenerate circle whose only point is (0, 0).
(b) In Example 5.1(b), the orbits are horizontal lines. Note that although in (a) and
(b) the same group G acts on the same set X, the orbits are different.
(c) In Example 5.1(c), the orbits are {a, b} and {c, d}.
(d) In Example 5.1(d), there is only one orbit {a, b, c, d}. Again we have a situation
in which a group G acts on a set X in two different ways, with different orbits.
We wish to consider the situation where X = Bn , the Boolean algebra of rank
n (so |Bn | = 2n ). We begin by defining an automorphism of a poset P to be an
isomorphism ϕ : P → P . (This definition is exactly analogous to the definition
of an automorphism of a group, ring, etc.) The set of all automorphisms of P
forms a group, denoted Aut(P ) and called the automorphism group of P , under
the operation of composition of functions (just as is the case for groups, rings, etc.)
Now consider the case P = Bn . Any permutation π of {1, . . . , n} acts on Bn as
follows: if x = {i1 , i2 , . . . , ik } ∈ Bn , then
π(x) = {π(i1 ), π(i2 ), . . . , π(ik )}. (5.1)
This action of π on Bn is an automorphism [why?]; in particular, if |x| = i, then

also |π(x)| = i. Equation (5.1) defines an action of the symmetric group Sn of all
permutations of {1, . . . , n} on Bn [why?]. (In fact, it is not hard to show that every
automorphism of Bn is of the form (5.1) for π ∈ Sn .) In particular, any subgroup
G of Sn acts on Bn via (5.1) (where we restrict π to belong to G). In what follows
this action is always meant.
5 Group Actions on Boolean Algebras 45
5.3 Example. Let n = 3, and let G be the subgroup of S3 with elements ι and
(1, 2). Here ι denotes the identity permutation, and (using disjoint cycle notation)
(1, 2) denotes the permutation which interchanges 1 and 2 and fixes 3. There are six
orbits of G (acting on B3 ). Writing, e.g., 13 as short for {1, 3}, the six orbits are {∅},
{1, 2}, {3}, {12}, {13, 23}, and {123}.
We now define the class of posets which will be of interest to us here. Later we
will give some special cases of particular interest.
Let G be a subgroup of Sn . Define the quotient poset Bn /G as follows. The
elements of Bn /G are the orbits of G. If o and o are two orbits, then define o ≤ o
in Bn /G if there exist x ∈ o and y ∈ o such that x ≤ y in Bn . It’s easy to check
that this relation ≤ is indeed a partial order.
5.4 Example. (a) Let n = 3 and G be the group of order two generated by the
cycle (1, 2), as in Example 5.3. Then the Hasse diagram of B3 /G is shown
below, where each element (orbit) is labelled by one of its elements.
123
13 12
3 1
(b) Let n = 5 and G be the group of order five generated by the cycle (1, 2, 3, 4, 5).
Then B5 /G has Hasse diagram
12345
1234
123 124
12 13
One simple property of a quotient poset Bn /G is the following.

5.5 Proposition. The quotient poset Bn /G defined above is graded of rank n and
rank-symmetric.
Proof. We leave as an exercise the easy proof that Bn /G is graded of rank n, and
that the rank of an element o of Bn /G is just the rank in Bn of any of the elements
x ∈ o. Thus the number pi (Bn /G) of elements of rank i is equal to the number
of orbits o ∈ (Bn )i /G. If x ∈ Bn , then let x̄ denote the set-theoretic complement

of x, i.e.,
x̄ = {1, . . . , n} − x = {1 ≤ i ≤ n : i ∈ x}.
Then {x1 , . . . , xj } is an orbit of i-element subsets of {1, . . . , n} if and only if

{x̄1 , . . . , x̄j } is an orbit of (n − i)-element subsets [why?]. Hence |(Bn )i /G| =
|(Bn )n−i /G|, so Bn /G is rank-symmetric.
Let π ∈ Sn . We associate with π a linear transformation (still denoted π )

π : R(Bn )i → R(Bn )i by the rule
⎛ ⎞

π⎝ cx x ⎠ = cx π(x),
x∈(Bn )i x∈(Bn )i
where each cx is a real number. This defines an action of Sn , or of any subgroup G

of Sn , on the vector space R(Bn )i . The matrix of π with respect to the basis (Bn )i
is just a permutation matrix, i.e., a matrix with one 1 in every row and column, and
0’s elsewhere. We will be interested in elements of R(Bn )i which are fixed by every
element of a subgroup G of Sn . The set of all such elements is denoted R(Bn )G i , so
i = {v ∈ R(Bn )i : π(v) = v for all π ∈ G}.

R(Bn )G
5.6 Lemma. A basis for R(Bn )G

i consists of the elements

vo := x,
x∈o
where o ∈ (Bn )i /G, the set of G-orbits for the action of G on (Bn )i .
Proof. First note that if o is an orbit and x ∈ o, then by definition of orbit we have
π(x) ∈ o for all π ∈ G (or all π ∈ Sn ). Since π permutes the elements of (Bn )i ,
it follows that π permutes the elements of o. Thus π(vo ) = vo , so vo ∈ R(Bn )G i .
It is clear that the vo ’s are linearly independent since any x ∈ (Bn )i appears with
nonzero coefficient in exactly one vo .
It remains to show that the vo ’s span R(Bn )G i , i.e., any v = x∈(Bn )i cx x ∈
R(Bn )G i can be written as a linear combination of v o ’s. Given x ∈ (B n )i , let Gx =
{π ∈ G : π(x) = x}, the stabilizer of x. We leave as an easy exercise the standard
fact that π(x) = σ (x) (where π, σ ∈ G) if and only if π and σ belong to the same
left coset of Gx , i.e., π Gx = σ Gx . It follows that in the multiset of elements π(x),
where π ranges over all elements of G and x is fixed, every element y in the orbit
Gx appears #Gx times, and no other elements appear. In other words,

π(x) = |Gx | · vGx .
π ∈G
(Do not confuse the orbit Gx with the subgroup Gx !) Now apply π to v and sum on
all π ∈ G. Since π(v) = v (because v ∈ R(Bn )Gi ), we get

|G| · v = π(v)
π ∈G
⎛ ⎞

= ⎝ cx π(x)⎠
π ∈G x∈(Bn )i

= cx π(x)
x∈(Bn )i π ∈G

= cx · (#Gx ) · vGx .
x∈(Bn )i
Dividing by |G| expresses v as a linear combination of the elements vGx (or vo ), as

desired.
Now let us consider the effect of applying the order-raising operator Ui to an

element v of R(Bn )G
i .
i , then Ui (v) ∈ R(Bn )i+1 .

5.7 Lemma. If v ∈ R(Bn )G G
Proof. Note that since π ∈ G is an automorphism of Bn , we have x < y in Bn if

and only if π(x) < π(y) in Bn . It follows [why?] that if x ∈ (Bn )i then
Ui (π(x)) = π(Ui (x)).
Since Ui and π are linear transformations, it follows by linearity that Ui π(u) =

π Ui (u) for all u ∈ R(Bn )i . In other words, Ui π = π Ui . Then
π(Ui (v)) = Ui (π(v))

= Ui (v),
so Ui (v) ∈ R(Bn )G
i+1 , as desired.
We come to the main result of this chapter and indeed our main result on the
Sperner property.
5.8 Theorem. Let G be a subgroup of Sn . Then the quotient poset Bn /G is graded

of rank n, rank-symmetric, rank-unimodal, and Sperner.
Proof. Let P = Bn /G. We have already seen in Proposition 5.5 that P is graded of
rank n and rank-symmetric. We want to define order-raising operators Ûi : RPi →
RPi+1 and order-lowering operators D̂i : RPi → RPi−1 . Let us first consider just
Ûi . The idea is to identify the basis element vo of RBnG with the basis element o of
RP , and to let Ûi : RPi → RPi+1 correspond to the usual order-raising operator
Ui : R(Bn )i → R(Bn )i+1 . More precisely, suppose that the order-raising operator
Ui for Bn given by (4.2) satisfies

Ui (vo ) = co,o vo , (5.2)
o ∈(Bn )i+1 /G
where o ∈ (Bn )i /G. (Note that by Lemma 5.7, Ui (vo ) does indeed have the form
given by (5.2).) Then define the linear operator Ûi : R((Bn )i /G) → R((Bn )i /G)
by

Ûi (o) = co,o o .
o ∈(Bn )i+1 /G
NOTE. We can depict the “transport of Ui to Ûi ” by a commutative diagram:
Ui
(RBn )G
i −−−−→ (RBn )G
i+1
⏐ ⏐
⏐
∼" ⏐∼
= "=
Ûi
R(Bn /G)i −−−−→ R(Bn /G)i+1
The arrows pointing down are the linear transformations induced by vo → o. The
map obtained by applying the top arrow followed by the rightmost down arrow is
the same as applying the leftmost down arrow followed by the bottom arrow.
order-raising. We need to show that if co,o = 0, then o > o
We claim that Ûi is
in Bn /G. Since vo = x ∈o x , the only way co,o = 0 in (5.2) is for some x ∈ o
to satisfy x > x for some x ∈ o. But this is just what it means for o > o, so Ûi is
order-raising.
Now comes the heart of the argument. We want to show that Ûi is one-to-one for
i < n/2. Now by Theorem 4.7, Ui is one-to-one for i < n/2. Thus the restriction of
Ui to the subspace R(Bn )G i is one-to-one. (The restriction of a one-to-one function
is always one-to-one.) But Ui and Ûi are exactly the same transformation, except
for the names of the basis elements on which they act. Thus Ûi is also one-to-one
for i < n/2.
Fig. 5.1 The poset BX /G of

nonisomorphic graphs with
four vertices
An exactly analogous argument can be applied to Di instead of Ui . We obtain

one-to-one order-lowering operators D̂i : R(Bn )G
i → R(Bn )i−1 for i > n/2. It
G
follows from Proposition 4.4, Lemma 4.5, and (4.4) that Bn /G is rank-unimodal
and Sperner, completing the proof.
We will consider two interesting applications of Theorem 5.8. For our first
application, we let n = m2 for some m ≥ 1 and let M = {1, . . . , m}. Set X = M 2 ,
the set of all two-element subsets of M. Think of the elements of X as (possible)
edges of a simple graph with vertex set M. If BX is the Boolean algebra of all subsets
of X (so BX and Bn are isomorphic), then an element x of BX is a collection of edges
on the vertex set M, in other words, just a simple graph on M. Define a subgroup G
of SX as follows. Informally, G consists of all permutations of the edges M 2 that
are induced from permutations of the vertices M. More precisely, if π ∈ Sm , then
define π̂ ∈ SX by π̂ · {i, j } = {π · i, π · j }. Thus G is isomorphic to Sm .
When are two graphs x, y ∈ BX in the same orbit of the action of G on BX ?
Since the elements of G just permute vertices, we see that x and y are in the same
orbit if we can obtain x from y by permuting vertices. This is just what it means for
two simple graphs x and y to be isomorphic—they are the same graph except for the
names of the vertices (thinking of edges as pairs of vertices). Thus the elements of
BX /G are isomorphism classes of simple graphs on the vertex set M. In particular,
#(BX /G) is the number of nonisomorphic m-vertex simple graphs, and #(BX /G)i
is the number of nonisomorphic such graphs with i edges. We have x ≤ y in BX /G
if there is some way of labelling the vertices of x and y so that every edge of x
is an edge of y. Equivalently, some spanning subgraph of y (i.e., a subgraph of y
with all the vertices of y) is isomorphic to x, as illustrated in Figure 5.1 for the case
m = 4. Hence by Theorem 5.8 there follows the following result, which is by no
means obvious and has no known non-algebraic proof.
5.9 Theorem. (a) Fix m ≥ 1. Let pi be the number of nonisomorphic simple
graphs with m vertices and i edges. Then the sequence p0 , p1 , . . . , p(m) is
2
symmetric and unimodal.
(b) Let T be a collection of simple graphs with m vertices such that no element
of T is isomorphic to a spanning subgraph of another element of T . Then #T
is maximized
by taking T to consist of all nonisomorphic simple graphs with
12 m2 edges.
Our second example of the use of Theorem 5.8 is more subtle and will be the
topic of the next chapter.
Digression. Edge reconstruction. Much work has been done on “reconstruction
problems,” that is, trying to reconstruct a mathematical structure such as a graph
from some of its substructures. The most famous of such problems is vertex
reconstruction: given a simple graph G on p vertices v1 , . . . , vp , let Gi be
the subgraph obtained by deleting vertex vi (and all incident edges). Given the
multiset {G1 , . . . , Gp } of vertex-deleted subgraphs graphs, can G be uniquely
reconstructed? It is important to realize that the vertices are unlabelled, so given
Gi we don’t know for any j which vertex is vj . The famous vertex-reconstruction
conjecture (still open) states that for p ≥ 3 any graph G can be reconstructed from
the multiset {G1 , . . . , Gp }.
Here we will be concerned with edge reconstruction, another famous open
problem. Given a simple graph G with edges e1 , . . . , eq , let Hi = G − ei , the
graph obtained from G by removing the edge ei .
Edge-Reconstruction Conjecture. A simple graph G can be uniquely recon-
structed from its number of vertices and the multiset {H1 , . . . , Hq } of edge-deleted
subgraphs.
NOTE. As in the case of vertex reconstruction, the subgraphs Hi are unlabelled.
The reason for including the number of vertices is that for any graph with no
edges, we have {H1 , . . . , Hq } = ∅, so we need to specify the number of vertices to
obtain G.
NOTE. It can be shown that if G can be vertex-reconstructed, then G can be
edge reconstructed. Hence the vertex-reconstruction conjecture implies the edge-
reconstruction conjecture.
The techniques developed above to analyze group actions on Boolean algebra
can be used to prove a special case of the edge-reconstruction
conjecture. Note that
a simple graph with p vertices has at most p2 edges.

5.10 Theorem. Let G be a simple graph with p vertices and q > 12 p2 edges. Then
G is edge-reconstructible.
Proof. Let Pq be the set of all simple graphs with q edges on the vertex set [p] =
p
{1, 2, . . . , p}, so #Pq = (q2 ) . Let RPq denote the real vector space with basis Pq .
Define a linear transformation ψq : RPq → RPq−1 by
ψq ( ) = 1 + ··· + q,
where 1 , . . . , q are the (labelled) graphs obtained from by deleting a single

edge. By Theorem 4.7, ψq is injective for q > 12 p2 . (Think of ψq as adding edges

to the complement of , i.e., the graph with vertex set [p] and edge set [p]
2 −E( ).)
The symmetric group Sp acts on Pq by permuting the vertices, and hence acts
on RPq , the real vector space with basis Pq . A basis for the fixed space (RPq )Sp

consists of the distinct sums ˜ = π ∈Sp π( ), where ∈ Pq . We may identify ˜
with the unlabelled graph isomorphic to , since ˜ = ˜ if and only if and are
isomorphic. Just as in the proof of Theorem 5.8, when we restrict ψq to (RPq )Sp

for q > 12 p2 we obtain an injection ψq : (RPq )Sp → (RPq−1 )Sp . In particular,
for nonisomorphic unlabelled graphs ˜ , ˜ with p vertices, we have
˜ 1 + · · · + ˜ q = ψq ( ˜ ) = ψq ( ˜ ) = ˜ 1 + · · · + ˜ q .
Hence the unlabelled graphs ˜ 1 , . . . , ˜ q determine ˜ , as desired.

Polynomials with Real Zeros. There are many techniques other than the linear
algebra used to prove Theorem 5.8 for showing that sequences are unimodal. Here
we will discuss a technique based on simple analysis (calculus) for showing that
sequences are unimodal. In fact, we will consider some stronger properties than
unimodality.
A sequence a0 , a1 , . . . , an of real numbers is called logarithmically concave, or
log-concave for short, if ai2 ≥ ai−1 ai+1 for 1 ≤ i ≤ n−1. We say that a0 , a1 , . . . , an

is strongly log-concave if bi2 ≥ bi−1 bi+1 for 1 ≤ i ≤ n − 1, where bi = ai / ni .
Strong log-concavity is equivalent to [why?]

1 1
ai2 ≥ 1 + 1+ ai−1 ai+1 , 1 ≤ i ≤ n − 1,
i n−i
from which it follows that strong log-concavity implies log-concavity.

Assume now that each ai ≥ 0. Does log-concavity then imply unimodality?
The answer is no, a counterexample being 1, 0, 0, 1. However, only this type
of counterexample can occur, as we now explain. We say that the sequence
a0 , a1 , . . . , an has no internal zeros if whenever we have i < j < k, ai = 0,
and ak = 0, then aj = 0.
5.11 Proposition. Let α = (a0 , a1 , . . . , an ) be a sequence of nonnegative real
numbers with no internal zeros. If α is log-concave, then α is unimodal.
Proof. If there are at most two values of j for which aj = 0 then we always have
ai−1 ai+1 = 0 so the conclusion is clear. Now assume that there are at least three
values of j for which aj = 0 and assume that the proposition is false. Then there
exists 1 ≤ i ≤ n − 1 for which ai−1 > ai ≤ ai+1 and ai+1 > 0, so ai2 < ai−1 ai+1 ,
a contradiction.
Now we come to a fundamental method for proving log-concavity.
5.12 Theorem (I. Newton). Let

n n
n
P (x) = bi x =
i
ai x i
i
i=0 i=0
be a real polynomial all of whose zeros are real numbers. Then the sequence
b0 , b1 , . . . , bn is strongly log-concave, or equivalently, the sequence a0 , a1 , . . . , an
is log-concave. Moreover, if each bi ≥ 0 (so the zeros of P (x) are nonpositive
[why?]) then the sequence b0 , b1 , . . . , bn has no internal zeros.
Proof. Let deg P (x) = m ≤ n. By the Fundamental Theorem of Algebra, P (x) has
exactly m real zeros, counting multiplicities. Suppose that α is a zero of multiplicity
r > 1, so P (x) = (x − α)r L(x) for some polynomial L(x) satisfying L(α) = 0.
A simple computation shows that α is a zero of P (x) (the derivative of P (x)) of
multiplicity r − 1. Moreover, if α < β are both zeros of P (x), then Rolle’s theorem
shows that P (x) has a zero γ satisfying α < γ < β. It follows [why?] that P (x)
has at least m − 1 real zeros. Since deg P (x) = m − 1 we see that P (x) has exactly
m − 1 real zeros and no other zeros.
d i−1
Let Q(x) = dx i−1 P (x). Thus Q(x) is a polynomial of degree at most m − i + 1
with only real zeros. Let R(x) = x m−i+1 Q(1/x), a polynomial of degree at most
m − i + 1. The zeros of R(x) are just reciprocals of those zeros of Q(x) not equal
to 0, with possible new zeros at 0. At any rate, all zeros of R(x) are real. Now
d m−i−1
let S(x) = dx m−i−1 R(x), a polynomial of degree at most two. By Rolle’s theorem
(with a suitable handling of multiple zeros as above), every zero of S(x) is real. An
explicit computation yields
m!
S(x) = (ai−1 x 2 + 2ai x + ai+1 ).
2
If ai−1 = 0 then trivially ai2 ≥ ai−1 ai+1 . Otherwise S(x) is a quadratic polynomial.
Since it has real zeros, its discriminant is nonnegative. But
= (2ai )2 − 4ai−1 ai+1 = 4(ai2 − ai−1 ai+1 ) ≥ 0,
so the sequence a0 , a1 , . . . , an is log-concave as claimed.

It remains to show that if each ai ≥ 0 then the sequence a0 , a1 , . . . , an has
no internal zeros. Suppose to the contrary that for some i < j < k we have
ai > 0, aj = 0, ak > 0. By arguing as in the previous paragraph we will obtain
a polynomial of the form c + dx k−i with only real zeros, where c, d > 0. But
since k − i ≥ 2 we have that every such polynomial has a nonreal zero [why?], a
contradiction which completes the proof.
In order to give combinatorial applications of Theorem 5.12 we need to find

polynomials with real zeros whose coefficients are of combinatorial interest. One
such example appears in Exercise 9.10, based on the fact that the characteristic
polynomial of a symmetric matrix has only real zeros.
Notes for Chapter 5
The techniques developed in this chapter had their origins in papers of Harper
[62] and Pouzet and Rosenberg [104]. The closest treatment to ours appears in a
paper of Stanley [126]. This latter paper also contains the proof of Theorem 5.10
(edge reconstruction) given here. This result was first proved
by Lovász [83] by
an inclusion–exclusion argument. The condition q > 12 p2 in Theorem 5.10 was
improved to q > p(log2 p−1) by Müller [95] (generalizing the argument of Lovász)
and by Krasikov and Roditty [78] (generalizing the argument of Stanley).
For further information on Newton’s Theorem 5.12, see, e.g., Hardy et al. [61,
p. 52]. For a general survey on unimodality, log-concavity, etc., see Stanley [128],
with a sequel by Brenti [13].
1. (a) Let G = {ι, π } be a group of order two (with identity element ι). Let G act
on {1, 2, 3, 4} by π · 1 = 2, π · 2 = 1, π · 3 = 3, and π · 4 = 4. Draw the
Hasse diagram of the quotient poset B4 /G.
(b) Do the same for the action π · 1 = 2, π · 2 = 1, π · 3 = 4, and π · 4 = 3.
2. Draw the Hasse diagram of the poset of nonisomorphic simple graphs with five
vertices (with the subgraph ordering). What is the size of the largest antichain?
How many antichains have this size?
3. Give an example of a finite graded poset P with the Sperner property, together
with a group G acting on P , such that the quotient poset P /G is not Sperner.
(By Theorem 5.8, P cannot be a Boolean algebra.)
4. Consider the poset P whose Hasse diagram is given by
Find a subgroup G of the symmetric group S7 for which P ∼

= B7 /G or else
prove that such a group does not exist.
5. A (0, 1)-necklace of length n and weight i is a circular arrangement of i 1’s and

n−i 0’s. For instance, the (0, 1)-necklaces of length 6 and weight 3 are (writing
a circular arrangement linearly) 000111, 001011, 010011, and 010101. (Cyclic
shifts of a linear word represent the same necklace, e.g., 000111 is the same
as 110001.) Let Nn denote the set of all (0, 1)-necklaces of length n. Define a
partial order on Nn by letting u ≤ v if we can obtain v from u by changing
some 0’s to 1’s. It’s easy to see (you may assume it) that Nn is graded of rank
n, with the rank of a necklace being its weight.
(a) (*) Show that Nn is rank-symmetric, rank-unimodal, and Sperner.
(b) (difficult) Show that Nn has a symmetric chain decomposition, as defined
on page 40.
6. (unsolved) Show that every quotient poset Bn /G has a symmetric chain
decomposition.
7. Let M be a finite multiset, say with ai i’s for 1 ≤ i ≤ k. Let BM denote the
poset of all submultisets of M, ordered by multiset inclusion. For instance, the
figure below illustrates the case a1 = 2, a2 = 3.
11222
1222 1122
222 122 112
22 12 11
2 1
Use Theorem 5.8 to show that BM is rank-symmetric, rank-unimodal, and

Sperner. (There are other ways to do this problem, but you are asked to use
Theorem 5.8. Thus you need to find a subgroup G of Sn for suitable n for
which BM ∼ = Bn /G.)
8. (unsolved) Let G be the group related to Theorem 5.9, so G acts on BX where X
consists of all two-element subsets of an
m-set.
Find an explicit order-matching
μ : (BX /G)i → (BX /G)i+1 for i < 12 m2 . Even the weaker problem of finding
an explicit injection (BX /G)i → (BX /G)i+1 is open.
9. (a) (*) Let Gp be the set of all simple graphs on the vertex set [p], so #Gp =
p
2( 2 ) . Given a graph G ∈ Gp , let Gi be the graph obtained by switching at
vertex i, i.e., deleting every edge incident to vertex i and adding every edge
from vertex i that isn’t an edge of G. Define a linear transformation
φ : RGp → RGp
by φ(G) = G1 + · · · + Gp . Show that φ is invertible if and only if p ≡

0 (mod 4).
(b) The graph G is switching-reconstructible if it can be uniquely reconstructed
from the multiset of unlabelled vertex switches Gi . That is, we are given
each Gi as an unlabelled graph. Show that G is switching-reconstructible
if p ≡ 0 (mod 4).
(c) (unsolved) Show that G is switching-reconstructible if p = 4.
(d) Show that the number of edges of G can be determined from the multiset of
unlabelled Gi ’s if p = 4. Find two graphs with four vertices and a different
number of edges, but with the same unlabelled Gi ’s.
(e) Define G to be weakly switching-reconstructible if it can be uniquely
reconstructed from the multiset of labelled vertex switches Gi . That is, we
are given each Gi as a labelled graph, but we aren’t told the vertex i that
was switched. Show that G is weakly switching-reconstructible if p = 4,
but that G need not be weakly switching-reconstructible if p = 4.
10. Suppose X is a finite n-element set and G a group of permutations of X. Thus
G acts on the subsets of X. We say that G acts transitively on the j -element
subsets if for every two j -element subsets S and T , there is a π ∈ G for which
π · S = T . Show that if G acts transitively on j -element subsets for some
j ≤ n/2, then G acts transitively on i-element subsets for all 0 ≤ i ≤ j .
(Although this can be done directly, there is a very easy proof using results
proven in the text.)
11. In Example 5.4(b) is drawn the Hasse diagram of B5 /G, where G is generated
by the cycle (1, 2, 3, 4, 5). Using the vertex labels shown in this figure, compute
explicitly U #2 (12) and U
#2 (13) as linear combinations of 123 and 124, where U#2
is defined as in the proof of Theorem 5.8. What is the matrix of U #2 with respect
to the bases (B5 /G)2 and (B5 /G)3 ?
n
12. A real polynomial F (x) = i
i=0 ai x is called log-concave if the sequence
a0 , a1 , . . . , an of coefficients is log-concave. Let F (x) and G(x) be log-
concave polynomials whose coefficients are positive. Show that the same is
true for F (x)G(x).
13. (*) Let F (x) be a real polynomial with positive leading coefficient whose zeros
all have the polar form reiθ , where 2π 3 ≤ θ ≤ 3 . Show that F (x) has positive,
4π
log-concave coefficients.
Chapter 6
Young Diagrams and q-Binomial
Coefficients
A partition λ of an integer n ≥0 is a sequence λ = (λ1 , λ2 , . . . ) of integers λi ≥ 0

satisfying λ1 ≥ λ2 ≥ · · · and i≥1 λi = n. Thus all but finitely many λi are equal
to 0. Each λi > 0 is called a part of λ. We sometimes suppress 0’s from the notation
for λ, e.g., (5, 2, 2, 1), (5, 2, 2, 1, 0, 0, 0), and (5, 2, 2, 1, 0, 0, . . . ) all represent the
same partition λ (of 10, with four parts). If λ is a partition of n, then we denote this
by λ n or |λ| = n.
6.1 Example. There are seven partitions of 5, namely (writing, e.g., 221 as short
for (2, 2, 1)): 5, 41, 32, 311, 221, 2111, and 11111.
The subject of partitions of integers has been extensively developed, but we
will only be concerned here with a small part related to our previous discussion.
Given positive integers m and n, let L(m, n) denote the set of all partitions
with at most m parts and with largest part at most n. For instance, L(2, 3) =
{∅, 1, 2, 3, 11, 21, 31, 22, 32, 33}. (Note that we are denoting by ∅ the unique
partition (0, 0, . . . ) with no parts.) If λ = (λ1 , λ2 , . . . ) and μ = (μ1 , μ2 , . . . )
are partitions, then define λ ≤ μ if λi ≤ μi for all i. This makes the set
of all partitions into a very interesting poset, denoted Y and called Young’s
lattice (named after the British mathematician Alfred Young, 1873–1940). (It is
called “Young’s lattice” rather than “Young’s poset” because it turns out to have
certain properties which define a lattice. However, these properties are irrelevant
to us here, so we will not bother to define the notion of a lattice.) We will be
looking at some properties of Y in Chapter 8. The partial ordering on Y , when
restricted to L(m, n), makes L(m, n) into a poset which also has some fascinating
properties. Figure 6.1 shows L(1, 4), L(2, 2), and L(2, 3), while Figure 6.2 shows
L(3, 3).

https://doi.org/10.1007/978-3-319-77173-1_6
58 6 Young Diagrams
Fig. 6.1 The posets L(1, 4), 33

L(2, 2), and L(2, 3)
4 22 32
3 21 22 31
2 11 2 21 3
1 1 11 2
φ φ 1
There is a nice geometric way of viewing partitions and the poset L(m, n). The
Young diagram (sometimes just called the diagram) of a partition λ is a left-justified
array of squares, with λi squares in the ith row. For instance, the Young diagram of
(4, 3, 1, 1) looks like:
If dots are used instead of boxes, then the resulting diagram is called a Ferrers
diagram. Thus the Ferrers diagram of (4, 3, 1, 1) looks like
The advantage of Young diagrams over Ferrers diagrams is that we can put
numbers in the boxes of a Young diagram, which we will do in Chapter 8. Observe
that L(m, n) is simply the set of Young diagrams D fitting in an m × n rectangle
(where the upper-left (northwest) corner of D is the same as the northwest corner
of the rectangle), ordered by inclusion. We will always assume that when a Young
diagram D is contained in a rectangle R, the northwest corners agree. It is also clear
from the Young diagram point of view that L(m, n) and L(n, m) are isomorphic
partially ordered sets, the isomorphism being given by transposing the diagram
(i.e., interchanging rows and columns). If λ has Young diagram D, then the partition
whose diagram is D t (the transpose of D) is called the conjugate of λ and is denoted
λ . For instance, (4, 3, 1, 1) = (4, 2, 2, 1), with diagram
6 Young Diagrams 59
Fig. 6.2 The poset L(3, 3) 333
332
322 331
222 321 33
221 311 32
211 22 31
111 21 3
11 2
6.2 Proposition. The poset L(m, n) is graded of rank mn and rank-symmetric. The
rank of a partition λ is just |λ| (the sum of the parts of λ or the number of squares
in its Young diagram).
Proof. As in the proof of Proposition 5.5, we leave to the reader everything except
rank-symmetry. To show rank-symmetry, consider the complement λ̄ of λ in an
m × n rectangle R, i.e., all the squares of R except for λ. (Note that λ̄ depends on m
and n and not just λ.) For instance, in L(5, 4), the complement of (4, 3, 1, 1) looks
like
If we rotate the diagram of λ̄ by 180◦ then we obtain the diagram of a partition

λ̃ ∈ L(m, n) satisfying |λ| + |λ̃| = mn. This correspondence between λ and λ̃
shows that L(m, n) is rank-symmetric.
60 6 Young Diagrams
Our main goal in this chapter is to show that L(m, n) is rank-unimodal and
Sperner. Let us write pi (m, n) as short for pi (L(m, n)), the number of elements
of L(m, n) of rank i. Equivalently, pi (m, n) is the number of partitions of i with
largest part at most n and with at most m parts, or, in other words, the number of
distinct Young diagrams with i squares which fit inside an m × n rectangle (with the
same northwest corner, as explained previously). Though not really necessary for
our goal, it is nonetheless interesting to obtain some information on these numbers
pi (m, n). First let us consider the total number #L(m, n) of elements in L(m, n).

6.3 Proposition. We have #L(m, n) = m+n m .

Proof. We will give an elegant combinatorial proof, based on the fact that m+n m is
equal to the number of sequences a1 , a2 , . . . , am+n , where each aj is either N or
E, and there are m N ’s (and hence n E’s) in all. We will associate a Young diagram
D contained in an m × n rectangle R with such a sequence as follows. Begin at the
lower left-hand corner of R and trace out the southeast boundary of D, ending at
the upper right-hand corner of R. This is done by taking a sequence of unit steps
(where each square of R is one unit in length), each step either north or east. Record
the sequence of steps, using N for a step to the north and E for a step to the east.
Example. Let m = 5, n = 6, and λ = (4, 3, 1, 1). Then R and D are given by:
The corresponding sequence of N ’s and E’s is NENNEENENEE.

It is easy to see (left to the reader) that the above correspondence gives a bijection
between Young diagrams D fitting in an m × n rectangle R and sequences of m
N ’s and nE’s. Hence the number of diagrams is equal to m+n m , the number of
sequences.
We now consider how many elements of L(m, n) have rank i. To this end, let q
be an indeterminate; and given j ≥ 1 define (j ) = 1 + q + q 2 + · · · + q j −1 . Thus
(1) = 1, (2) = 1+q, (3) = 1+q +q 2 , etc. Note that (j ) is a polynomial in q whose
value at q = 1 is just j (denoted (j )q=1 = j ). Next define (j )! = (1)(2) · · · (j ) for
j ≥ 1 and set (0)! = 1. Thus (1)! = 1, (2)! = 1 + q, (3)! = (1 + q)(1 + q + q 2 ) =
1 + 2q + 2q 2 + q 3 , etc., and (j )!q=1 = j !. Finally define for k ≥ j ≥ 0

k (k)!
= .
j (j )!(k − j )!
k
The expression j is called a q-binomial coefficient (or Gaussian coefficient).
6 Young Diagrams 61
When q is regarded as a prime power rather than as an indeterminate, then

Exercise 4 in Chapter 4 gives a definition of nk in terms of the field Fq . In this

chapter we have no need of this algebraic interpretation of nk .
Since (r)!q=1 = r!, it is clear that

k k
= .
j q=1 j

One sometimes says that jk is a “q-analogue” of the binomial coefficient jk . There
is no precise definition of a q-analogue P (q) of some mathematical object P (such
as a formula or definition). It should have the property that there is a reasonable way
to interpret P (1) as being P . Ideally P (q) should have some interpretation involving
Fq when q is regarded as a prime power. The q-analogue of the set {1} is the finite
field Fq , and the q-analogue of the set [n] = {1, 2, . . . , n} is the vector space Fnq .
k
6.4 Example. We have jk = k−j [why?]. Moreover,

k k
= =1
0 k

k k
= = (k) = 1 + q + q 2 + · · · + q k−1
1 k−1

4 (4)(3)(2)(1)
= = 1 + q + 2q 2 + q 3 + q 4
2 (2)(1)(2)(1)

5 5
= = 1 + q + 2q 2 + 2q 3 + 2q 4 + q 5 + q 6 .
2 3

In the above example, jk was always a polynomial in q (and with nonnegative
integer coefficients). It is not obvious that this is always the case, but it will follow
easily from the following lemma.
6.5 Lemma. We have

k k−1 k−j k − 1
= +q , (6.1)
j j j −1
0 k
whenever k ≥ 1, with the initial conditions = 1, = 0 if j < 0 or j > k (the
j
0

same initial conditions satisfied by the binomial coefficients jk ).
Proof. This is a straightforward computation. Specifically, we have

k−1 k−1 (k − 1)! (k − 1)!
+ q k−j = + q k−j
j j −1 (j )!(k − 1 − j )! (j − 1)!(k − j )!

(k − 1)! 1 q k−j
= +
(j − 1)!(k − 1 − j )! (j ) (k − j )
62 6 Young Diagrams
(k − 1)! (k − j ) + q k−j (j )
=
(j − 1)!(k − 1 − j )! (j )(k − j )
(k − 1)! (k)
=
(j − 1)!(k − 1 − j )! (j )(k − j )

k
= .
j
Note that if we put q = 1 in (6.1) we obtain the well-known formula

k k−1 k−1
= + ,
j j j −1
which is just the recurrence defining Pascal’s triangle. Thus (6.1) may be regarded
as a q-analogue of the Pascal triangle recurrence.
We can regard equation (6.1) as a recurrence relation for the q-binomial coeffi-
cients. Given
the initial conditions of Lemma 6.5, we can use (6.1) inductively to
compute jk for any k and j . From this it is obvious by induction that the q-binomial

coefficient jk is a polynomial in q with nonnegative integer coefficients. The
following theorem gives an even stronger result, namely, an explicit combinatorial
interpretation of the coefficients.
6.6 Theorem. Let pi (m, n) denote the number of elements of L(m, n) of rank i.
Then

m+n
pi (m, n)q =
i
. (6.2)
m
i≥0
NOTE. The sum on the left-hand side is really a finite sum, since pi (m, n) = 0 if
i > mn.
Proof. Let P (m, n) denote the left-hand side of (6.2). We will show that
P (0, 0) = 1, and P (m, n) = 0 if m < 0 or n < 0 (6.3)
P (m, n) = P (m, n − 1) + q n P (m − 1, n). (6.4)
Note that (6.3) and (6.4) completely determine P (m,n). On the other hand,
substituting k = m + n and j = m in (6.1) shows that m+n m also satisfies (6.4).
m+n
Moreover, the initial conditions of Lemma 6.5 show that m also satisfies (6.3).

Hence (6.3) and (6.4) imply that P (m, n) = m+n m , so to complete the proof we
need only establish (6.3) and (6.4).
6 Young Diagrams 63
Equation (6.3)
is clear, since L(0, n) consists of a single point (the empty
partition ∅), so i≥0 pi (0, n)q i = 1; while L(m, n) is empty (or undefined, if you
prefer) if m < 0 or n < 0.
The crux of the proof is to show (6.4). Taking the coefficient of q i of both sides
of (6.4), we see [why?] that (6.4) is equivalent to
pi (m, n) = pi (m, n − 1) + pi−n (m − 1, n). (6.5)
Consider a partition λ i whose Young diagram D fits in an m × n rectangle R. If

D does not contain the upper right-hand corner of R, then D fits in an m × (n − 1)
rectangle, so there are pi (m, n − 1) such partitions λ. If on the other hand D does
contain the upper right-hand corner of R, then D contains the whole first row of R.
When we remove the first row of R, we have left a Young diagram of size i − n
which fits in an (m − 1) × n rectangle. Hence there are pi−n (m − 1, n) such λ, and
the proof follows [why?].
q=
Note that if we set 1 in (6.2), then the left-hand side becomes #L(m, n) and
the right-hand side m+n m , agreeing with Proposition 6.3.
As the reader may have guessed by now, the poset L(m, n) is isomorphic to a
quotient poset Bs /G for a suitable integer s > 0 and finite group G acting on Bs .
Actually, it is clear that we must have s = mn since L(m, n) has rank mn and in
general Bs /G has rank s. What is not so clear is the right choice of G. To this end,
let R = Rmn denote an m × n rectangle of squares. For instance, R35 is given by the
15 squares of the diagram
We now define the group G = Gmn as follows. It is a subgroup of the group SR of

all permutations of the squares of R. A permutation π in G is allowed to permute
the elements in each row of R in any way and then to permute the rows among
themselves in any way. The elements of each row can be permuted in n! ways, so
since there are m rows there are a total of n!m permutations preserving the rows.
Then the m rows can be permuted in m! ways, so it follows that the order of Gmn
is given by m!n!m . The group Gmn is called the wreath product of Sn and Sm ,
denoted Sn Sm or Sn wr Sm . However, we will not discuss the general theory of
wreath products here.
6.7 Example. Suppose m = 4 and n = 5, with the boxes of R labelled as follows.
64 6 Young Diagrams
1 2 3 4 5
6 7 8 9 10
11 12 13 14 15
16 17 18 19 20
Then a typical permutation π in G45 looks like
16 20 17 19 18
4 1 5 2 3
12 13 15 14 11
7 9 6 10 8
i.e., π(16) = 1, π(20) = 2, etc.

We have just defined a group Gmn of permutations of the set R = Rmn of squares
of an m × n rectangle. Hence Gmn acts on the boolean algebra BR of all subsets of
the set R. The next lemma describes the orbits of this action.
6.8 Lemma. Every orbit o of the action of Gmn on BR contains exactly one Young
diagram D, i.e., exactly one subset D ⊆ R such that D is left-justified, and if λi is
the number of elements of D in row i of R, then λ1 ≥ λ2 ≥ · · · ≥ λm .
Proof. Let S be a subset of R, and suppose that S has αi elements in row i. If π ∈

Gmn and π · S has βi elements in row i, then β1 , . . . , βm is just some permutation of
α1 , . . . , αm [why?]. There is a unique ordering λ1 , . . . , λm of α1 , . . . , αm satisfying
λ1 ≥ · · · ≥ λm , so the only possible Young diagram D in the orbit π · S is the one
of shape λ = (λ1 , . . . , λm ). It’s easy to see that the Young diagram Dλ of shape λ
is indeed in the orbit π · S. Namely, by permuting the elements in the rows of R we
can left-justify the rows of S, and then by permuting the rows of R themselves we
can arrange the row sizes of S to be in weakly decreasing order. Thus we obtain the
Young diagram Dλ as claimed.
We are now ready for the main result of this chapter.

6.9 Theorem. Set R = Rmn . Then the quotient poset BR /Gmn is isomorphic to
L(m, n).
Proof. Each element of BR /Gmn contains a unique Young diagram Dλ by

Lemma 6.8. Moreover, two different orbits cannot contain the same Young diagram
D since orbits are disjoint. Thus the map ϕ : BR /Gmn → L(m, n) defined by
ϕ(Dλ ) = λ is a bijection (one-to-one and onto). We claim that in fact ϕ is an
isomorphism of partially ordered sets. We need to show the following: let o and o∗
be orbits of Gmn (i.e., elements of BR /Gmn ). Let Dλ and Dλ∗ be the unique Young
diagrams in o and o∗ , respectively. Then there exist D ∈ o and D ∗ ∈ o∗ satisfying
D ⊆ D ∗ if and only if λ ≤ λ∗ in L(m, n).
The “if” part of the previous sentence is clear, for if λ ≤ λ∗ then Dλ ⊆ Dλ∗ . So
assume there exist D ∈ o and D ∗ ∈ o∗ satisfying D ⊆ D ∗ . The lengths of the rows
6 Young Diagrams 65
of D, written in decreasing order, are λ1 , . . . , λm , and similarly for D ∗ . Since each

row of D is contained in a row of D ∗ , it follows that for each 1 ≤ j ≤ m, D ∗ has at
least j rows of size at least λj . Thus the length λ∗j of the j th largest row of D ∗ is at
least as large as λj . In other words, λj ≤ λ∗j , as was to be proved.
Combining the previous theorem with Theorem 5.8 yields the following result.
6.10 Corollary. The posets L(m, n) are rank-symmetric, rank-unimodal, and
Sperner.
Note that the rank-symmetry and rank-unimodality
of L(m, n) can be rephrased
as follows: the q-binomial coefficient m+n m has symmetric and unimodal coeffi-
cients. While rank-symmetry
is easy to prove (see Proposition 6.2), the unimodality
of the coefficients of m+n m is by no means apparent. It was first proved by J.
Sylvester in 1878 by a proof similar to the one above, though stated in the language
of the invariant theory of binary forms. For a long time
it was an open problem to
find a combinatorial proof that the coefficients of m+nm are unimodal. Such a proof
would give an explicit injection (one-to-one function) μ : L(m, n)i → L(m, n)i+1
for i < 12 mn. (One difficulty in finding such maps μ is to make use of the hypothesis
that i < 12 mn.) Finally around 1989 such a proof was found by K. M. O’Hara.
However, O’Hara’s proof has the defect that the maps μ are not order-matchings.
Thus her proof does not prove that L(m, n) is Sperner, but only that it’s rank-
unimodal. It is an outstanding open problem in algebraic combinatorics to find an
explicit order-matching μ : L(m, n)i → L(m, n)i+1 for i < 12 mn.
Note that the Sperner property of L(m, n) (together with the fact that the largest
level is in the middle) can be stated in the following simple terms: the largest
possible collection C of Young diagrams fitting in an m × n rectangle such that
no diagram in C is contained in another diagram in C is obtained by taking all
the diagrams of size 12 mn. Although the statement of this fact requires almost
no mathematics to understand, there is no known proof that doesn’t use algebraic
machinery. The several known algebraic proofs are all closely related, and the one
we have given is the simplest. Corollary 6.10 is a good example of the efficacy of
algebraic combinatorics.
An Application to Number Theory. There is an interesting application of
Corollary 6.10 to a number-theoretic problem. Fix a positive integer k. For a
finite subset S of R+ = {β ∈ R : β > 0}, and for a real number α > 0, define
$ %
S
fk (S, α) = # T ∈ : t =α .
k
t∈T
In other words, fk (S, α) is the number of k-element subsets of S whose elements

sum to α. For instance, f3 ({1, 3, 4, 6, 7}, 11) = 2, since 1 + 3 + 7 = 1 + 4 + 6 = 11.
Given positive integers k < n, our object is to maximize fk (S, α) subject to the
condition that #S = n. We are free to choose both S and α, but k and n are fixed.
Call this maximum value hk (n). Thus
66 6 Young Diagrams
hk (n) = max fk (S, α).

α∈R+
S⊂R+
#S=n
What sort of behavior can we expect of the maximizing set S? If the elements of S
are “spread out,” say S = {1, 2, 4, 8, . . . , 2n−1 }, then all the subset sums of S are
distinct. Hence for any α ∈ R+ we have fk (S, α) = 0 or 1. Similarly, if the elements
of S√are√“unrelated” (e.g., linearly independent over the rationals, such as S =
{1, 2, 3, π, π 2 }), then again all subset sums are distinct and fk (S, α) = 0 or 1.
These considerations make it plausible that we should take S = [n] = {1, 2, . . . , n}
and then choose α appropriately. In other words, we are led to the conjecture that
+
for any S ∈ Rn and α ∈ R+ , we have
fk (S, α) ≤ fk ([n], β), (6.6)
for some β ∈ R+ to be determined.

First let us evaluate fk ([n], α) for any α. This will enable us to determine the
value of β in (6.6). Let S = {i1 , . . . , ik } ⊆ [n] with
1 ≤ i1 < i2 < · · · < ik ≤ n, i1 + · · · + ik = α. (6.7)

k+1
Let jr = ir − r. Then (since 1 + 2 + · · · + k = 2 )

k+1
n − k ≥ jk ≥ jk−1 ≥ · · · ≥ j1 ≥ 0, j1 + · · · + jk = α − . (6.8)
2
Conversely, given j1 , . . . , jk satisfying (6.8) we can recover i1 , . . . , ik satis-

fying (6.7). Hence fk ([n], α) is equal to the number of sequences j1 , . . . , jk
satisfying (6.8). Now let
λ(S) = (jk , jk−1 , . . . , j1 ).

k+1
Note that λ(S) is a partition of the integer α − 2 with at most k parts and with
largest part at most n − k. Thus
fk ([n], α) = pα−(k+1) (k, n − k), (6.9)

2
or equivalently,

k+1 n
fk ([n], α)q α−( 2)= .
k
α≥(k+1
2 )
(and rank-symmetry) of L(n − k, k) (Corollary 6.10), the

By the rank-unimodality
largest coefficient of nk is the middle one, that is, the coefficient of k(n − k)/2.
6 Young Diagrams 67
It follows
k+1 that for fixed k and n, fk ([n], α) is maximized for α = k(n − k)/2 +
2 = k(n + 1)/2. Hence the following result is plausible.
+
6.11 Theorem. Let S ∈ Rn , α ∈ R+ , and k ∈ P. Then
fk (S, α) ≤ fk ([n], k(n + 1)/2).
Proof. Let S = {a1 , . . . , an } with 0 < a1 < · · · < an . Let T and U be distinct
k-element subsets of S with the same element sums, say T = {ai1 , . . . , aik } and
U = {aj1 , . . . , ajk } with i1 < i2 < · · · < ik and j1 < j2 < · · · < jk . Define T ∗ =

{i1 , . . . , ik } and U ∗ = {j1 , . . . , jk }, so T ∗ , U ∗ ∈ [n]
k . The crucial observation is
the following:
Claim. The elements λ(T ∗ ) and λ(U ∗ ) are incomparable in L(k, n − k), i.e.,
neither λ(T ∗ ) ≤ λ(U ∗ ) nor λ(U ∗ ) ≤ λ(T ∗ ).
Proof of claim. Suppose not, say λ(T ∗ ) ≤ λ(U ∗ ) to be definite. Thus by
definition of L(k, n − k) we have ir − r ≤ jr − r for 1 ≤ r ≤ k. Hence
ir ≤ jr for 1 ≤ r ≤ k, so also air ≤ ajr (since a1 < · · · < an ). But
ai1 + · · · + aik = aj1 + · · · + ajk by assumption, so air = ajr for all r. This
contradicts the assumption that T and U are distinct and proves the claim.
It is now easy to complete the proof of Theorem 6.11. Suppose that S1 , . . . , Sr
are distinct k-element subsets of S with the same element sums. By the claim,
{λ(S1∗ ), . . . , λ(Sr∗ )} is an antichain in L(k, n − k). Hence r cannot exceed the size
of the largest antichain in L(k, n − k). By Theorem 6.6 and Corollary 6.10, the size
of the largest antichain in L(k, n − k) is given by pk(n−k)/2 (k, n − k). By (6.9) this
number is equal to fk ([n], k(n + 1)/2). In other words,
r ≤ fk ([n], k(n + 1)/2),
which is what we wanted to prove.

Note that an equivalent statement
of Theorem 6.11 is that hk (n) is equal to the
coefficient of q k(n−k)/2 in nk [why?].
Variation on a theme. Suppose that in Theorem 6.11 we do not want to specify the
cardinality of the subsets of S. In other words, for any α ∈ R and any finite subset
S ⊂ R+ , define

f (S, α) = #{T ⊆ S : t = α}.
t∈T
How large can f (S, α) be if we require #S = n? Call this maximum value h(n).
Thus
h(n) = max f (S, α). (6.10)
α∈R+
S⊂R+
#S=n
68 6 Young Diagrams
4321
432
321 431
32 43 421
21
31 42 321
1 2
21 41
3 32
φ 1
2 4 31
M(1) φ
M(2) 1 3 21
φ 2
M(3)
1
φ
M(4)
Fig. 6.3 The posets M(1), M(2), M(3), and M(4)
For instance, if S = {1, 2, 3} then f (S, 3) = 2 (coming from the subsets {1, 2} and
{3}). This is easily seen to be best possible, i.e., h(3) = 2.
We will find h(n) in a manner analogous to the proof of Theorem 6.11. The big
difference is that the relevant poset M(n) is not of the form Bn /G, so we will have
to prove the injectivity of the order-raising operator Ui from scratch. Our proofs will
be somewhat sketchy; it shouldn’t be difficult for the reader who has come this far
to fill in the details.
Let M(n) be the set of all subsets of [n], with the ordering A ≤ B if the elements
of A are a1 > a2 > · · · > aj and the elements of B are b1 > b2 > · · · > bk , where
j ≤ k and ai ≤ bi for 1 ≤ i ≤ j . (The empty set ∅ is the bottom element of M(n).)
Figure 6.3 shows M(1), M(2), M(3), and M(4).
It is easy to see that M(n) is graded of rank n+1 2 . The rank of the subset T =
{a1 , . . . , ak } is
rank(T ) = a1 + · · · + ak . (6.11)
It follows [why?] that the rank-generating function of M(n) is given by

n+1
(2 )
F (M(n), q) = (#M(n)i )q i = (1 + q)(1 + q 2 ) · · · (1 + q n ).
i=0
6 Young Diagrams 69
Define linear transformations
Ui : RM(n)i → RM(n)i+1 , Di : RM(n)i → RM(n)i−1
by

Ui (x) = y, x ∈ M(n)i
y∈M(n)i+1
x<y

Di (x) = c(v, x)v, x ∈ M(n)i ,
v∈M(n)i−1
v<x
where the coefficient c(v, x) is defined as follows. Let the elements of v be

a1 > · · · > aj > 0 and the elements of x be b1 > · · · > bk > 0. Since x
covers v, there is a unique r for which ar = br − 1 (and ak = bk for all other k).
In the case br = 1 we set ar = 0. (e.g., if x is given by 5 > 4 > 1 and v by 5 > 4,
then r = 3 and a3 = 0.) Set
⎧ n+1
⎨
2 , if ar = 0
c(v, x) =
⎩ (n − a )(n + a + 1), if a > 0.
r r r
It is a straightforward computation (proof omitted) to obtain the commutation

relation

n+1
Di+1 Ui − Ui−1 Di = − 2i Ii , (6.12)
2
where Ii denotes the identity linear transformation on RM(n)i . Clearly by definition

Ui is order-raising. We want to show that Ui is injective (one-to-one) for i < 12 n+1
2 .
We can’t argue as in the proof of Lemma 4.6 that Ui−1 Di is semidefinite since the
matrices of Ui−1 and Di are no longer transposes of one another. Instead we use the
following result from linear algebra.
6.12 Lemma. Let V and W be finite-dimensional vector spaces over a field. Let
A : V → W and B : W → V be linear transformations. Then
x dim V det(AB − xI ) = x dim W det(BA − xI ).
In other words, AB and BA have the same nonzero eigenvalues.

We can now prove the key linear algebraic result.

1 n+1
6.13 Lemma. The linear transformation Ui is injective for i < and
2 2
surjective (onto) for i ≥ 12 n+1
2 .
Proof. We prove by induction on i that Di+1 Ui has positive real eigenvalues for
i < 12 n+1
2 . For i = 0 this is easy to check since dim RM(n)0 = 1. Assume the
70 6 Young Diagrams

induction hypothesis for some i < 12 n+1
2 − 1, i.e., assume that Di Ui−1 has positive
eigenvalues. By Lemma 6.12, Ui−1 Di has nonnegative eigenvalues. By (6.12), we
have

n+1
Di+1 Ui = Ui−1 Di + − 2i Ii .
2

Thus the eigenvalues of Di+1 Ui are n+1 − 2i more than those of Ui−1 Di . Since
n+1 2
2 − 2i > 0, it follows that Di+1 Ui has positive eigenvalues. Hence it is
invertible,
so Ui is injective. Similarly (or by “symmetry”) Ui is surjective for
i ≥ 12 n+1
2 .
The main result on the posets M(n) now follows by a familiar argument.

6.14 Theorem. The poset M(n) is graded of rank n+1 2 , rank-symmetric, rank-
unimodal, and Sperner.

Proof. We have already seen that M(n) is graded of rank n+1
2 and rank-symmetric.

By the previous lemma, Ui is injective for i < 2 2 and surjective for i ≥ 12 n+1
1 n+1
2 .
The proof follows from Proposition 4.4 and Lemma 4.5.
NOTE. As a consequence of Theorem 6.14, the polynomial F (M(n), q) = (1 +

q)(1 + q 2 ) · · · (1 + q n ) has unimodal coefficients. No combinatorial proof of this
fact is known, unlike the situation for L(m, n) (where we mentioned the proof of
O’Hara above).
We can now determine h(n) (as defined by (6.10)) by an argument analogous to
the proof of Theorem 6.11.
+
6.15 Theorem. Let S ∈ Rn and α ∈ R+ . Then
) *
1 n+1
f (S, α) ≤ f [n], = h(n).
2 2
Proof. Let S = {a1 , . . . , an } with 0 < a1 < · · · < an . Let T and U be

distinct subsets of S with the same element sums, say T = {ar1 , . . . , arj } and
U = {as1 , . . . , ask } with r1 < r2 < · · · < rj and s1 < s2 < · · · < sk . Define
T ∗ = {r1 , . . . , rj } and U ∗ = {s1 , . . . , sk }, so T ∗ , U ∗ ∈ M(n). The following fact is
proved exactly in the same way as the analogous fact for L(m, n) (the claim in the
proof of Theorem 6.11) and will be omitted here.
Fact. The elements T ∗ and U ∗ are incomparable in M(n), i.e., neither T ∗ ≤ U ∗
nor U ∗ ≤ T ∗ .
It is now easy to complete the proof of Theorem 6.15. Suppose that S1 , . . . , St are
distinct subsets of S with the same element sums. By the above fact, {S1∗ , . . . , St∗ }
is an antichain in M(n). Hence t cannot exceed the size of the largest antichain
in M(n). By Theorem 6.14, the size of the largest antichain in M(n) is the size
Notes for Chapter 6 71
n+1
p+ 1 n+1
, of the middle rank. By (6.11) this number is equal to f ([n], 12 2 ).
( )
2 2
In other words,
) *
1 n+1
t ≤f [n], ,
2 2
which is what we wanted to prove.
NOTE. Theorem 6.15 is known as the weak Erdős–Moser conjecture. The

original (strong) Erdős–Moser conjecture deals with the case S ⊂ R rather than
S ⊂ R+ . There is a difference between these two cases; for instance, h(3) = 2
(corresponding to S = {1, 2, 3} and α = 3), while the set {−1, 0, 1} has four subsets
whose elements sum to 0 (including the empty set). (Can you see where the proof
of Theorem 6.15 breaks down if we allow S ⊂ R?) The original Erdős–Moser
conjecture asserts that if #S = 2m + 1, then
f (S, α) ≤ f ({−m, −m + 1, . . . , m}, 0). (6.13)
This result can be proved by a somewhat tricky modification of the proof given
above for the weak case; see Exercise 6.5. No proof of the Erdős–Moser conjecture
(weak or strong) is known other than the one indicated here (sometimes given in a
more sophisticated context, as explained in the next Note).
NOTE. The key to the proof of Theorem 6.15 is the definition of Ui and Di which
gives the commutation relation (6.12). The reader may be wondering how anyone
managed to discover these definitions (especially that of Di ). In fact, the original
proof of Theorem 6.15 was based on the representation theory of the orthogonal Lie
algebra o(2n + 1, C). In this context, the definitions of Ui and Di are built into the
theory of the “principal subalgebras” of o(2n + 1, C). R. A. Proctor was the first
to remove the representation theory from the proof and present it solely in terms of
linear algebra.
Notes for Chapter 6
For an undergraduate level introduction to the theory of partitions, see Andrews and
Eriksson [3]. A more extensive treatment is given by Andrews [2], while a brief
introduction appears in [130, Section 1.8].
As already mentioned in the text, the rank-unimodality
of L(m, n), that is, of the
coefficients of the q-binomial coefficient m+nm , is due to J. J. Sylvester [135], with
a combinatorial proof later given by K. M. O’Hara [98]. An explication of O’Hara’s
work was given by D. Zeilberger [147].
The unimodality of the coefficients of the polynomial (1+q)(1+q 2 ) · · · (1+q n )
is implicit in the work of E.B. Dynkin [37], [38, p. 332]. J.W.B. Hughes was the first
72 6 Young Diagrams
to observe explicitly that this polynomial arises as a special case of Dynkin’s work.
The Spernicity of L(m, n) and M(n), and a proof of the Erdős–Moser conjecture,
were first given by Stanley [123]. It was mentioned in the text above that R.A.
Proctor [105] was the first to remove the representation theory from the proof and
present it solely in terms of linear algebra.
For two proofs of Lemma 6.12, see W.V. Parker [99] and J. Schmid [116].
1. (a) Let A(m, n) denote the adjacency matrix (over R) of the Hasse diagram of
L(m, n). Show that if A(m, n) is nonsingular, then m+n
m is even.
(b) (unsolved) For which m and n is A(m, n) nonsingular? The pairs (m, n) with
this property for m ≤ n and m+n ≤ 13 are (1, 1), (1, 3), (1, 5), (3, 3), (1, 7),
(1, 9), (3, 7), (5, 5), (1, 11), (3, 9), and (5, 7).
(c) (very difficult) Show that every irreducible (over Q) factor of the
characteristic polynomial of the matrix A(m, n) has degree at most
2 φ(2(m + n + 1)), where φ is the Euler phi-function (defined on page 85).
1
2. (a) (moderately difficult) Show that the number c(m, n) of cover relations in
L(m, n), i.e., the number of pairs (λ, μ) of partitions in L(m, n) for which
μ covers λ, is given by
(m + n − 1)!
c(m, n) = .
(m − 1)! (n − 1)!
(b) (considerably more difficult) (*) Show that the number d(m, n) of pairs
(λ, μ) of elements in L(m, n) for which λ ≤ μ is given by
(m + n)! (m + n + 1)!
d(m, n) = .
m! (m + 1)! n! (n + 1)!
3. (difficult) (*) Note that L(m, n) is the set of all partitions λ in Young’s lattice Y
satisfying λ ≤ nm , the partition with m parts equal to n. Let Yμ denote the set
of all partitions λ ≤ μ. Is Yμ always rank-unimodal?
4. (a) Find an explicit order matching μ : L(2, n)i → L(2, n)i+1 for i < n.
(b) (more difficult) Do the same for L(3, n)i → L(3, n)i+1 for i < 3n/2.
(c) (even more difficult) Do the same for L(4, n)i → L(4, n)i+1 for i < 2n.
(d) (unsolved) Do the same for L(5, n)i → L(5, n)i+1 for i < 5n/2.
5. Assume that M(j )×M(k)∗ is rank-symmetric, rank-unimodal, and Sperner. Here
M(k)∗ denotes the dual of M(k), i.e., x ≤ y in M(k)∗ if and only if y ≤ x in
M(k). (Actually M(k) ∼
= M(k)∗ , but this is not needed here.) Deduce the original
Erdős–Moser conjecture given by (6.13), namely, if S ⊂ R and #S = 2m + 1,
then
f (S, α) ≤ f ({−m, −m + 1, . . . , m}, 0).
NOTE. If P and Q are posets, then the direct product P × Q is the poset on the
set {(x, y) : x ∈ P , y ∈ Q} satisfying (x, y) ≤ (x , y ) if and only if x ≤ x in
P and y ≤ y in Q.
6. (unsolved) Show that L(m, n) has a symmetric chain decomposition. This is
known to be true for m ≤ 4.
Chapter 7
Enumeration Under Group Action
In Chapters 5 and 6 we considered the quotient poset Bn /G, where G is a subgroup

of the symmetric group Sn . If pi is the number of elements of rank i of this poset,
then the sequence p0 , p1 , . . . , pn is rank-symmetric and rank-unimodal. Thus it is
natural to ask whether there is some nice formula for the numbers pi . For instance,
in Theorem
5.9 pi is the number of nonisomorphic graphs with m vertices (where
n = m2 ) and i edges; is there some nice formula for this number? For the group
Gmn = Sn Sm of Theorem 6.6 we obtained a simple generating function for pi
(i.e., a formula for the rank-generating function F (Bmn /Gmn , q) = i pi q i ), but
this was a very special situation. In this chapter we will present a general theory
for enumerating inequivalent objects subject to a group of symmetries, which will
include a formula for the rank-generating functions F (Bn /G, q). The chief architect
of this theory is G. Pólya (though much of it was anticipated by J. H. Redfield) and
hence is often called Pólya’s theory of enumeration or just Pólya theory. See the
references at the end of this chapter for further historical information.
Pólya theory is most easily understood in terms of “colorings” of some geometric
or combinatorial object. For instance, consider a row of five squares:
In how many ways can we color the squares using n colors? Each square can be
colored any of the n colors, so there are n5 ways in all. These colorings can by
indicated as
A B C D E
where A, B, C, D, and E are the five colors. Now assume that we are allowed to
rotate the row of five squares 180◦ and that two colorings are considered the same
if one can be obtained from the other by such a rotation. (We may think that we
have cut the row of five squares out of paper and colored them on one side.) We
say that two colorings are equivalent if they are the same or can be transformed into

https://doi.org/10.1007/978-3-319-77173-1_7
76 7 Enumeration Under Group Action
one another by a 180◦ rotation. The first naive assumption is that every coloring
is equivalent to exactly one other (besides itself), so the number of inequivalent
colorings is n5 /2. Clearly this reasoning cannot be correct since n5 /2 is not always
an integer! The problem, of course, is that some colorings stay the same when we
rotate 180◦ . In fact, these are exactly the colorings
A B C B A
where A, B, and C are any three colors. There are n3 such colorings, so the total
number of inequivalent colorings is given by
1
(number of colorings which don’t equal their 180◦ rotation)
2
+(number of colorings which equal their 180◦ rotation)
1 5
= (n − n3 ) + n3
2
1
= (n5 + n3 ).
2
Pólya theory gives a systematic method for obtaining formulas of this sort for any
underlying symmetry group.
The general setup is the following. Let X be a finite set, and G a subgroup of the
symmetric group SX . Think of G as a group of symmetries of X. Let C be another
set (which may be infinite), which we think of as a set of “colors.” A coloring of X
is a function f : X → C. For instance, X could be the set of four squares of a 2 × 2
chessboard, labelled as follows:
Let C = {r, b, y} (the colors red, blue, and yellow). A typical coloring of X
would then look like
The above diagram thus indicates the function f : X → C given by f (1) =

r, f (2) = b, f (3) = y, f (4) = r.
7 Enumeration Under Group Action 77
NOTE. We could work in the slightly greater generality of a group G acting

on the set X, i.e., we are given a homomorphism ϕ : G → SX that need not be
injective. However, we then have a well-defined induced injective homomorphism
ψ : H → SX , where H = G/(ker ϕ). The results obtained below for H are
identical to those we get for G, so nothing is lost by assuming that ϕ is injective. In
this case we can identify G with its image ϕ(G).
We define two colorings f and g to be equivalent (or G-equivalent, when it is
G
necessary to specify the group), denoted f ∼ g or f ∼ g, if there exists an element
π ∈ G such that
g(π(x)) = f (x) for all x ∈ X.
We may write this condition more succinctly as gπ = f , where gπ denotes the

composition of functions (from right to left). It is easy to check, using the fact that
G is a group, that ∼ is an equivalence relation. One should think that equivalent
functions are the same “up to symmetry.”
7.1 Example. Let X be the 2 × 2 chessboard and C = {r, b, y} as above. There
are many possible choices of a symmetry group G, and this will affect when two
colorings are equivalent. For instance, consider the following groups:
• G1 consists of only the identity permutation (1)(2)(3)(4).
• G2 is the group generated by a vertical reflection. It consists of the two elements
(1)(2)(3)(4) (the identity element) and (1, 2)(3, 4) (the vertical reflection).
• G3 is the group generated by a reflection in the main diagonal. It consists of the
two elements (1)(2)(3)(4) (the identity element) and (1)(4)(2, 3) (the diagonal
reflection).
• G4 is the group of all rotations of X. It is a cyclic group of order four with
elements (1)(2)(3)(4), (1, 2, 4, 3), (1, 4)(2, 3), and (1, 3, 4, 2).
• G5 is the dihedral group of all rotations and reflections of X. It has eight
elements, namely, the four elements of G4 and the four reflections (1, 2)(3, 4),
(1, 3)(2, 4), (1)(4)(2, 3), and (2)(3)(1, 4).
• G6 is the symmetric group of all 24 permutations of X. Although this is a
perfectly valid group of symmetries, it no longer has any connection with the
geometric representation of X as the squares of a 2 × 2 chessboard.
Consider the inequivalent colorings of X with two red squares, one blue square, and
one yellow square, in each of the six cases above.
(G1 ) There are 12 colorings in all with two red squares, one blue square, and one
yellow square, and all are inequivalent under the trivial group (the group with
one element). In general, whenever G is the trivial group then two colorings
are equivalent if and only if they are the same [why?].
(G2 ) There are now six inequivalent colorings, represented by
r r r b r y b y r b r y
b y r y r b r r y r b r
Each equivalence class contains two elements.

(G3 ) Now there are seven classes, represented by
r r r r b y y b r b b r y r
b y y b r r r r y r r y r b
The first five classes contain two elements each and the last two classes
only one element. Although G2 and G3 are isomorphic as abstract groups,
as permutation groups they have a different structure. Specifically, the
generator (1, 2)(3, 4) of G2 has two cycles of length two, while the generator
(1)(4)(2, 3) has two cycles of length one and one of length two. As we will
see below, it is the lengths of the cycles of the elements of G that determine
the sizes of the equivalence classes. This explains why the number of classes
for G2 and G3 is different.
(G4 ) There are three classes, each with four elements. The size of each class
is equal to the order of the group because none of the colorings have any
symmetry with respect to the group, i.e., for any coloring f using the colors
r, r, y, b, the only group element π that fixes f (so f π = f ) is the identity
(π = (1)(2)(3)(4)).
r r r r r b
y b b y y r
(G5 ) Under the full dihedral group there are now two classes.
r r r b
b y y r
The first class has eight elements and the second four elements. In general, the
size of a class is the index in G of the subgroup fixing some fixed coloring in
that class [why?]. For instance, the subgroup fixing the second coloring above
is {(1)(2)(3)(4), (1, 4)(2)(3)}, which has index four in the dihedral group of
order eight.
(G6 ) Under the group S4 of all permutations of the squares there is clearly only
one class, with all 12 colorings. In general, for any set X if the group is the
symmetric group SX then two colorings are equivalent if and only if each
color appears the same number of times [why?].
Our object in general is to count the number of equivalence classes of colorings

which use each color a specified number of times. We will put the information into
a generating function—a polynomial whose coefficients are the numbers we seek.
Consider for example the set X, the group G = G5 (the dihedral group), and the
set C = {r, b, y} of colors in Example 7.1 above. Let κ(i, j, k) be the number of
inequivalent colorings using red i times, blue j times, and yellow k times. Think of
the colors r, b, y as variables, and form the polynomial

FG (r, b, y) = κ(i, j, k)r i bj y k .
i+j +k=4
Note that we sum only over i, j, k satisfying i + j + k = 4 since a total of four

colors will be used to color the four-element set X. The reader should check that
FG (r, b, y) = (r 4 + b4 + y 4 ) + (r 3 b + rb3 + r 3 y + ry 3 + b3 y + by 3 )
+2(r 2 b2 + r 2 y 2 + b2 y 2 ) + 2(r 2 by + rb2 y + rby 2 ).
For instance, the coefficient of r 2 by is two because, as we have seen above, there
are two inequivalent colorings using the colors r, r, b, y. Note that FG (r, b, y) is a
symmetric function of the variables r, b, y (i.e., it stays the same if we permute the
variables in any way), because insofar as counting inequivalent colorings goes, it
makes no difference what names we give the colors. As a special case we may ask
for the total number of inequivalent colorings with four colors. This is obtained by
setting r = b = y = 1 in FG (r, b, y) [why?], yielding FG (1, 1, 1) = 3 + 6 + 2 · 3 +
2 · 3 = 21.
What happens to the generating function FG in the above example when we use
the n colors r1 , r2 , . . . , rn (which can be thought of as different shades of red)?
Clearly all that matters are the multiplicities of the colors, without regard for their
order. In other words, there are five cases: (a) all four colors the same, (b) one color
used three times and another used once, (c) two colors used twice each, (d) one color
used twice and two others once each, and (e) four colors used once each. These five
cases correspond to the five partitions of 4, i.e., the five ways of writing 4 as a sum
of positive integers without regard to order: 4, 3 + 1, 2 + 2, 2 + 1 + 1, 1 + 1 + 1 + 1.
Our generating function becomes

FG (r1 , r2 , . . . , rn ) = ri4 + ri3 rj
i i =j

+2 ri2 rj2 + 2 ri2 rj rk + 3 ri rj rk rl ,
i<j i =j i<j <k<l
i =k
j <k
where the indices in each sum lie between 1 and n. If we set all variables equal to one
(obtaining the total number of colorings with n colors), then simple combinatorial
reasoning yields

n n−1 n
FG (1, 1, . . . , 1) = n + n(n − 1) + 2 + 2n +3
2 2 4
1 4
= (n + 2n3 + 3n2 + 2n). (7.1)
8
Note that the polynomial (7.1) has the following description: the denominator 8 is
the order of the group G5 , and the coefficient of ni in the numerator is just the
number of permutations in G5 with i cycles! For instance, the coefficient of n2 is
3, and G5 has the three elements (1, 2)(3, 4), (1, 3)(2, 4), and (1, 4)(2, 3) with two
cycles. We want to prove a general result of this nature.
The basic tool which we will use is a simple result from the theory of permutation
groups known as Burnside’s lemma. It was actually first proved by Cauchy when G
is transitive (i.e., |Y /G| = 1 in Lemma 7.2 below) and by Frobenius in the general
case, and is sometimes called the Cauchy–Frobenius lemma.
7.2 Lemma (Burnside’s lemma). Let Y be a finite set and G a subgroup of SY . For
each π ∈ G, let
Fix(π ) = {y ∈ Y : π(y) = y},
so #Fix(π ) is the number of cycles of length one in the permutation π . Let Y /G be

the set of orbits of G. Then
1
|Y /G| = #Fix(π ).
#G
π ∈G
An equivalent form of Burnside’s lemma is the statement that the average number
of elements of Y fixed by an element of G is equal to the number of orbits. Before
proceeding to the proof, let us consider an example.
7.3 Example. Let Y = {a, b, c, d},
G = {(a)(b)(c)(d), (a, b)(c, d), (a, c)(b, d), (a, d)(b, c)},
and
G = {(a)(b)(c)(d), (a, b)(c)(d), (a)(b)(c, d), (a, b)(c, d)}.
Both groups are isomorphic to Z2 × Z2 (compare Example 5.1(c) and (d)). By

Burnside’s lemma the number of orbits of G is 14 (4 + 0 + 0 + 0) = 1. Indeed,
given any two elements i, j ∈ Y , it is clear by inspection that there is a π ∈ G
(which happens to be unique) such that π(i) = j . On the other hand, the number of
orbits of G is 14 (4 + 2 + 2 + 0) = 2. Indeed, the two orbits are {a, b} and {c, d}.
Proof of Burnside’s lemma. For y ∈ Y let Gy = {π ∈ G : π · y = y} (the set of

permutations fixing y). Then
1 1
#Fix(π ) = 1
#G #G
π ∈G y∈Y π ∈G
π ·y=y
1
= 1
#G π ∈G
y∈Y
π ·y=y
1
= #Gy .
#G
y∈Y
Now (as in the proof of Lemma 5.6) the multiset of elements π · y, π ∈ G, contains
every element in the orbit Gy the same number of times, namely #G/#Gy times.
Thus y occurs #G/#Gy times among the π · y, so
#G
= #Gy .
#Gy
Therefore
1 1 #G
#Fix(π ) =
#G #G #Gy
π ∈G y∈Y
1
= .
#Gy
y∈Y
How many times does a term 1/#O appear in the above sum, where O is a fixed
orbit? We are asking for the number of y such that Gy = O. But Gy = O if and
only if y ∈ O, so 1/#O appears #O times. Thus each orbit gets counted exactly
once, so the above sum is equal to the number of orbits.
7.4 Example. How many inequivalent colorings of the vertices of a regular hexagon
H are there using n colors, under cyclic symmetry? Let Cn be the set of all n-
colorings of H . Let G be the group of all permutations of Cn which permute the
colors cyclically, so G ∼
= Z6 . We are asking for the number of orbits of G [why?].
We want to apply Burnside’s lemma, so for each of the six elements σ of G we need
to compute the number of colorings fixed by that element. Let π be a generator
of G.
• σ = 1 (the identity): All n6 colorings are fixed by σ .
• σ = π, π −1 : Only the n colorings with all colors equal are fixed.
• σ = π 2 , π 4 : Any coloring of the form ababab is fixed (writing the colors
linearly in the order they appear around the hexagon, starting at any fixed vertex).
There are n choices for a and n for b, so n2 colorings in all.
• σ = π 3 : The fixed colorings are of the form abcabc, so n3 in all.
Hence by Burnside’s lemma, we have

1 6
number of orbits = (n + n3 + 2n2 + 2n).
6
The reader who has followed the preceding example will have no trouble
understanding the following result.
7.5 Theorem. Let G be a group of permutations of a finite set X. Then the number
NG (n) of inequivalent (with respect to G) n-colorings of X is given by
1 c(π )
NG (n) = n , (7.2)
#G
π ∈G
where c(π ) denotes the number of cycles of π .
Proof. Let πn denote the action of π ∈ G on the set Cn of n-colorings of X. We

want to determine the set Fix(πn ), so that we can apply Burnside’s lemma. Let C
be the set of n colors. If f : X → C is a coloring fixed by π , then for all x ∈ X
we have
f (x) = πn · f (x) = f (π(x)).
Thus f ∈ Fix(πn ) if and only if f (x) = f (π(x)). Hence f (x) = f (π k (x)) for
any k ≥ 1 [why?]. The elements y of X of the form π k (x) for k ≥ 1 are just the
elements of the cycle of π containing x. Thus to obtain f ∈ Fix(πn ), we should
take the cycles σ1 , . . . , σc(π ) of π and color each element of σi the same color.
There are n choices for each σi , so nc(π ) colorings in all fixed by π . In other words,
#Fix(πn ) = nc(π ) , and the proof follows by Burnside’s lemma.
We would now like not just to count the total number of inequivalent colorings
with n colors but more strongly to specify the number of occurrences of each color.
We will need to use not just the number c(π ) of cycles of each π ∈ G, but rather
the lengths of each of the cycles of π . Thus given a permutation π of an n-element
set X, define the type of π to be
type(π ) = (c1 , c2 , . . . , cn ),
where π has ci i-cycles. For instance, if π = 4, 7, 3, 8, 2, 10, 11, 1, 6, 9, 5, then
type(π ) = type (1, 4, 8)(2, 7, 11, 5)(3)(6, 10, 9)

= (1, 0, 2, 1, 0, 0, 0, 0, 0, 0, 0).

Note that we always have i ici = n [why?]. Define the cycle indicator of π to be
the monomial
Zπ = z1c1 z2c2 · · · zncn .

(Many other notations are used for the cycle indicator. The use of Zπ comes from the
German word Zyklus for cycle. The original paper of Pólya was written in German.)
Thus for the example above, we have Zπ = z1 z32 z4 .
Now given a subgroup G of SX , the cycle indicator (or cycle index polynomial
or cycle enumerator) of G is defined by
1
ZG = ZG (z1 , . . . , zn ) = Zπ .
#G
π ∈G
Thus ZG (also denoted PG , Cyc(G), etc.) is a polynomial in the variables z1 , . . . , zn .

7.6 Example. If X consists of the vertices of a square and G is the group of rotations
of X (a cyclic group of order 4), then
1 4
ZG = (z + z22 + 2z4 ).
4 1
If reflections are also allowed (so G is the dihedral group of order 8), then
1 4
ZG = (z + 3z22 + 2z12 z2 + 2z4 ).
8 1
We are now ready to state the main result of this chapter.

7.7 Theorem (Pólya’s Theorem, 1937). Let G be a group of permutations of the
n-element set X. Let C = {r1 , r2 , . . . } be a set of colors. Let κ(i1 , i2 , . . . ) be the
number of inequivalent (under the action of G) colorings f : X → C such that
color rj is used ij times. Define

FG (r1 , r2 , . . . ) = κ(i1 , i2 , . . . )r1i1 r2i2 · · · .
i1 ,i2 ,...
(Thus FG is a polynomial or a power series in the variables r1 , r2 , . . . , depending

on whether or not C is finite or infinite.) Then
FG (r1 , r2 , . . . ) =
j j j
ZG (r1 + r2 + r3 + · · · , r12 + r22 + r32 + · · · , . . . , r1 + r2 + r3 + · · · , . . . ).
j
(In other words, substitute i ri for zj in ZG .)
Before giving the proof let us consider an example.
7.8 Example. Suppose that in Example 7.6 our set of colors is C = {a, b, c, d},
and that we take G to be the group of cyclic symmetries. Then
1
FG (a, b, c, d) = (a + b + c + d)4 + (a 2 + b2 + c2 + d 2 )2 + 2(a 4 + b4 + c4 + d 4 )
4
= (a 4 + · · · ) + (a 3 b + · · · ) + 2(a 2 b2 + · · · ) + 3(a 2 bc + · · · ) + 6abcd.
An expression such as (a 2 b2 + · · · ) stands for the sum of all monomials in the

variables a, b, c, d with exponents 2, 2, 0, 0 (in some order). The coefficient of all
such monomials is 2, indicating two inequivalent colorings using one color twice
and another color twice. If instead G were the full dihedral group, we would get
1
FG (a, b, c, d) = (a + b + c + d)4 + 3(a 2 + b2 + c2 + d 2 )2
8

+ 2(a + b + c + d)2 (a 2 + b2 + c2 + d 2 ) + 2(a 4 + b4 + c4 + d 4 )
= (a 4 + · · · ) + (a 3 b + · · · ) + 2(a 2 b2 + · · · ) + 2(a 2 bc + · · · ) + 3abcd.
Proof of Pólya’s theorem. Let #X = t and i1 + i2 + · · · = t, where each ij ≥ 0.

Let i = (i1 , i2 , . . .), and let Ci denote the set of all colorings of X with color rj
used ij times. The group G acts on Ci , since if f ∈ Ci and π ∈ G, then π · f ∈ Ci .
(“Rotating” a colored object does not change how many times each color appears.)
Let πi denote the action of π on Ci . We want to apply Burnside’s lemma to compute
the number of orbits, so we need to find #Fix(πi ).
In order for f ∈ Fix(πi ), we must color X so that (a) in any cycle of π , all
the elements get the same color, and (b) the color rj appears ij times. Consider the
product
j j
Hπ = (r1 + r2 + · · · )cj (π ) ,
j
where cj (π ) is the number of j -cycles (cycles of length j ) of π . When we expand

j j
this product as a sum of monomials r11 r22 · · · , we get one of these monomials by
j
choosing a term rk from each factor of Hπ and multiplying these terms together.
j
Choosing rk corresponds to coloring all the elements of some j -cycle with rk . Since
j j j
a factor r1 + r2 + · · · occurs precisely cj (π ) times in Hπ , choosing a term rk from
every factor corresponds to coloring X so that every cycle is monochromatic (i.e.,
j
all the elements of that cycle get the same color). The product of these terms rk
j1 j2
will be the monomial r1 r2 · · · , where we have used color rk a total of jk times. It
follows that the coefficient of rii1 r2i2 · · · in Hπ is equal to #Fix(πi ). Thus

Hπ = #Fix(πi )r1i1 r2i2 · · · . (7.3)
i
Now sum both sides of (7.3) over all π ∈ G and divide by #G. The left-hand side
becomes
1 j j
(r1 + r2 + · · · )cj (π ) = ZG (r1 + r2 + · · · , r12 + r22 + · · · , . . . ).
#G
π ∈G j
On the other hand, the right-hand side becomes

- .
1
#Fix(πi ) r1i1 r2i2 · · · .
#G
i π ∈G
By Burnside’s lemma, the expression in brackets is just the number of orbits of πi

acting on Ci , i.e., the number of inequivalent colorings using color rj a total of ij
times, as was to be proved.
7.9 Example (necklaces). A necklace of length is a circular arrangement of

(colored) beads. Two necklaces are considered the same if they are cyclic rotations
of one another. Let X be a set of (uncolored) beads, say X = {1, 2, . . . , }.
Regarding the beads as being placed equidistantly on a circle in the order 1, 2, . . . , ,
let G be the cyclic group of rotations of X. Thus if π is the cycle (1, 2, . . . , ), then
G = {1, π, π 2 , . . . , π −1 }. For example, if = 6 then the elements of G are
π 0 = (1)(2)(3)(4)(5)(6)
π = (1, 2, 3, 4, 5, 6)
π 2 = (1, 3, 5)(2, 4, 6)
π 3 = (1, 4)(2, 5)(3, 6)
π 4 = (1, 5, 3)(2, 6, 4)
π 5 = (1, 6, 5, 4, 3, 2).
In general, if d is the greatest common divisor of m and (denoted d =

gcd(m, )), then π m has d cycles of length /d. An integer m satisfies 1 ≤ m ≤
and gcd(m, ) = d if and only if 1 ≤ m/d ≤ /d and gcd(m/d, /d) = 1. Hence
the number of such integers m is given by the Euler phi-function (or totient function)
φ(/d), which by definition is equal to the number of integers 1 ≤ i ≤ /d such
that gcd(i, /d) = 1. As an aside, recall that φ(k) can be computed by the formula
1

φ(k) = k 1− . (7.4)
p
p|k
p prime
For instance, φ(1000) = 1000(1 − 12 )(1 − 15 ) = 400. Putting all this together gives
the following formula for the cycle enumerator ZG (z1 , . . . , z ):
1
ZG (z1 , . . . , z ) = d
φ(/d)z/d ,

d|
or (substituting /d for d),
1 /d
ZG (z1 , . . . , z ) = φ(d)zd .

d|
There follows from Pólya’s theorem the following result (originally proved by
MacMahon (1854–1929) before Pólya discovered his general result).
7.10 Theorem. (a) The number N (n) of n-colored necklaces of length is
given by
1
N (n) = φ(/d)nd . (7.5)

d|
(b) We have
1
FG (r1 , r2 , . . . ) = φ(d)(r1d + r2d + · · · )/d .

d|
NOTE. (b) reduces to (a) if r1 = r2 = · · · = 1. Moreover, since clearly N (1) =

1, putting n = 1 in (7.5) yields the well-known identity

φ(/d) = .
d|
What if we are allowed to flip necklaces over, not just rotate them? Now the
group becomes the dihedral group of order 2, and the corresponding inequivalent
colorings are called dihedral necklaces. We leave to the reader to work out the cycle
enumerators
⎛ ⎞
1 ⎝
φ(d)zd + mz12 z2m−1 + mz2m ⎠ , if = 2m
/d
2
d|
⎛ ⎞ (7.6)
1 ⎝
φ(d)zd + z1 z2m ⎠ , if = 2m + 1.
/d
2
d|
7.11 Example. Let G = S , the group of all permutations of {1, 2, . . . , } = X.

Thus for instance
1 3
ZS3 (z1 , z2 , z3 ) = (z + 3z1 z2 + 2z3 )
6 1
1 4
ZS4 (z1 , z2 , z3 , z4 ) = (z + 6z12 z2 + 3z22 + 8z1 z3 + 6z4 ).
24 1
It is easy to count the number of inequivalent colorings in Ci . If two colorings of

X use each color the same number of times, then clearly there is some permutation
of X which sends one of the colorings to the other. Hence Ci consists of a single
orbit. Thus

FS (r1 , r2 , . . . ) = r1i1 r2i2 · · · ,
i1 +i2 +···=
the sum of all monomials of degree .

To count the total number of inequivalent n-colorings, note that
1
FS (r1 , r2 , . . . )x = , (7.7)
(1 − r1 x)(1 − r2 x) · · ·
≥0
j
since if we expand each factor on the right-hand side into the series j ≥0 ri x j and
multiply, the coefficient of x will just be the sum of all monomials of degree .
For fixed n, let fn () denote the number of inequivalent n-colorings of X. Since
fn () = FS (1, 1, . . . , 1) (n 1’s in all), there follows from (7.7) that
1
fn ()x = . (7.8)
(1 − x)n
≥0
The right-hand side can be expanded (e.g., by Taylor’s theorem or by the binomial
theorem for the exponent −n) as
1 n + − 1
= x.
(1 − x)n
≥0
Hence

n+−1
fn () = .

It is natural to ask whether there might be a more direct proof of such a simple result.
This is actually a standard result in elementary enumerative combinatorics. For fixed
and n we want the number of solutions to i1 + i2 + · · · + in = in nonnegative
n+−1 Suppose that we arrange n − 1 vertical bars and dots in a line. There are
integers.
such arrangements since there are a total of n + − 1 positions, and we
choose of them in which to place a dot. An example of such an arrangement for
= 8 and n = 7 is
The number of dots in each “compartment,” read from left to right, gives the num-
bers i1 , . . . , in . For the example above, we get (i1 , . . . , i7 ) = (0, 0, 2, 1, 0, 3, 2).
Since this correspondence between solutions to i1 + i2 + · · · + in = and
arrangements of bars and dots is clearly a bijection, we get n+−1 solutions as
claimed.
Recall (Theorem 7.5) that the number of inequivalent n-colorings of X (with
respect to any group G of permutations of X) is given by
1 c(π )
n ,
#G
π ∈G
where c(π ) denotes the number of cycles of π . Hence for G = S we get the
identity

1 c(π ) n+−1
n =
!
π ∈S
1
= n(n + 1)(n + 2) · · · (n + − 1).
!
Multiplying by ! yields

nc(π ) = n(n + 1)(n + 2) · · · (n + − 1). (7.9)
π ∈S
Equivalently [why?], if we define c(, k) to be the number of permutations in S

with k cycles (called a signless Stirling number of the first kind), then

c(, k)x k = x(x + 1)(x + 2) · · · (x + − 1).
k=1
For instance, x(x+1)(x+2)(x+3) = x 4 +6x 3 +11x 2 +6x, so (taking the coefficient

of x 2 ) 11 permutations in S4 have two cycles, namely, (123)(4), (132)(4), (124)(3),
(142)(3), (134)(2), (143)(2), (234)(1), (243)(1), (12)(34), (13)(24), (14)(23).
Although it was easy to compute the generating function FS (r1 , r2 , . . . ) directly
without the necessity of computing the cycle indicator ZS (z1 , . . . , z ), we can still
ask whether there is a formula of some kind for this polynomial. First we determine
explicitly its coefficients.

7.12 Theorem. Let ici = . The number of permutations π ∈ S with ci cycles
of length i (or equivalently, the coefficient of z1c1 z2c2 · · · in ! ZS (z1 , . . . , z )) is
equal to !/1c1 c1 ! 2c2 c2 ! · · · .
Example The number of permutations in S15 with three 1-cycles, two 2-cycles,
and two 4-cycles is 15!/13 · 3! · 22 · 2! · 42 · 2! = 851,350,500.
Proof of Theorem 7.12. Fix c = (c1 , c2 , . . . ) and let Xc be the set of all
permutations π ∈ S with ci cycles of length i. Given a permutation σ =
a1 a2 · · · a in S , construct a permutation f (σ ) ∈ Xc as follows. Let the
1-cycles of f (σ ) be (a1 ), (a2 ), . . . , (ac1 ). Then let the 2-cycles of f (σ ) be
(ac1 +1 , ac1 +2 ), (ac1 +3 , ac1 +4 ), . . . , (ac1 +2c2 −1 , ac1 +2c2 ). Then let the 3-cycles of
f (σ ) be (ac1 +2c2 +1 , ac1 +2c2 +2 , ac1 +2c2 +3 ), (ac1 +2c2 +4 , ac1 +2c2 +5 , ac1 +2c2 +6 ), . . . ,
(ac1 +2c2 +3c3 −2 , ac1 +2c2 +3c3 −1 , ac1 +2c2 +3c3 ), etc., continuing until we reach a and
have produced a permutation in Xc . For instance, if = 11, c1 = 3, c2 = 2, c4 = 1,
and σ = 4, 9, 6, 11, 7, 1, 3, 8, 10, 2, 5, then
f (σ ) = (4)(9)(6)(11, 7)(1, 3)(8, 10, 2, 5).
We have defined a function f : S → Xc . Given π ∈ Xc , what is #f −1 (π ), the

number of permutations sent to π by f ? A cycle of length i can be written in i ways,
namely,
(b1 , b2 , . . . , bi ) = (b2 , b3 , . . . , bi , b1 ) = · · · = (bi , b1 , b2 , . . . , bi−1 ).
Moreover, there are ci ! ways to order the ci cycles of length i. Hence
#f −1 (π ) = c1 ! c2 ! c3 ! · · · 1c1 2c2 3c3 · · · ,
the same number for any π ∈ Xc . It follows that
#S
#Xc =
c1 !c2 ! · · · 1c1 2c2 · · ·
!
= ,
c1 !c2 ! · · · 1c1 2c2 · · ·
as was to be proved.
As for the polynomial ZS itself, we have the following result. Write exp y = ey .
7.13 Theorem. We have

x2 x3
ZS (z1 , z2 , . . . )x = exp z1 x + z2 + z3 + ··· .
2 3
≥0
Proof. There are some sophisticated ways to prove this theorem which “explain”
why the exponential function appears, but we will be content here with a “naive”
computational proof. Write

x2 x3
exp z1 x + z2 + z3 + ···
2 3
x2 x3
= ez1 x · ez2 2 · ez3 3 · · ·
⎛ ⎞⎛ ⎞⎛ ⎞
zn x n zn x 2n zn x 3n
=⎝ 1 ⎠⎝ 2 ⎠⎝ 3 ⎠··· .
n! 2n n! 3n n!
n≥0 n≥0 n≥0
When we multiply this product out, the coefficient of z1c1 z2c2 · · · x , where = c1 +
2c2 + · · · , is given by

1 1 !
= .
1c1 c1 ! 2c2 c2 ! · · · ! 1c1 c1 ! 2c2 c2 ! · · ·
By Theorem 7.12 this is just the coefficient of z1c1 z2c2 · · · in ZS (z1 , z2 , . . . ), as was
to be proved.
As a check of Theorem 7.13, set each zi = n to obtain

x2 x3
ZS (n, n, . . . )x = exp nx + n + n + · · ·
2 3
≥0

x2 x3
= exp n(x + + + ···)
2 3

= exp n log(1 − x)−1
1
= ,
(1 − x)n
agreeing with Theorem (7.5) and (7.8).
Theorem 7.13 has many enumerative applications. We give one such result here
as an example.
7.14 Proposition. Let f (n) be the number of permutations π ∈ Sn of odd order.
Equivalently, π k = ι (the identity permutation) for some odd k. Then
$
12 · 32 · 52 · · · (n − 1)2 , n even
f (n) =
12 · 32 · 52 · · · (n − 2)2 · n, n odd.
Proof. A permutation has odd order if and only if all its cycle lengths are odd. Hence
[why?]
f (n) = n! ZSn (zi = 1, i odd; zi = 0, i even).
Making this substitution in Theorem 7.13 gives

xn x3 x5
f (n) = exp x + + + ··· .
n! 3 5
n≥0
x2 x3
Since − log(1 − x) = x + 2 + 3 + · · · , we get [why?]

xn 1
f (n) = exp (− log(1 − x) + log(1 + x))
n! 2
n≥0

1 1+x
= exp log
2 1−x
/
1+x
= .
1−x
We
√ therefore need to find the coefficients in the power series expansion of
(1 + x)/(1 − x) at x = 0. There is a simple trick for doing so:
/
1+x
= (1 + x)(1 − x 2 )−1/2
1−x
−1/2
= (1 + x) (−x 2 )m
m
m≥0

−1/2
= (−1)m (x 2m + x 2m+1 ),
m
m≥0
where by definition

−1/2 1 1 3 2m − 1
= − − ··· − .
m m! 2 2 2
It is now a routine computation to check that the coefficient of x n /n! in

√
(1 + x)/(1 − x) agrees with the claimed value of f (n).
Quotients of Boolean Algebras. We will show how to apply Pólya theory to the
problem of counting the number of elements of given rank in a quotient poset
BX /G. Here X is a finite set, BX is the Boolean algebra of all subsets of X, and G is
a group of permutations of X (with an induced action on BX ). What do colorings of
X have to do with subsets? The answer is very simple: a 2-coloring f : X → {0, 1}
corresponds to a subset Sf of X by the usual rule
s ∈ Sf ⇐⇒ f (s) = 1.
Note that two 2-colorings f and g are G-equivalent if and only if Sf and Sg are in
the same orbit of G (acting on BX ). Thus the number of inequivalent 2-colorings f
of X with i values equal to 1 is just #(BX /G)i , the number of elements of BX /G
of rank i. As an immediate application of Pólya’s theorem (Theorem 7.7) we obtain
the following result.
7.15 Corollary. We have

#(BX /G)i q i = ZG (1 + q, 1 + q 2 , 1 + q 3 , . . . ).
i
Proof. If κ(i, j ) denotes the number of inequivalent 2-colorings of X with the colors
0 and 1 such that 0 is used j times and 1 is used i times (so i + j = #X), then by
Pólya’s theorem we have

κ(i, j )x i y j = ZG (x + y, x 2 + y 2 , x 3 + y 3 , . . . ).
i,j
Setting x = q and y = 1 yields the desired result [why?].
Combining Corollary 7.15 with the rank-unimodality of BX /G (Theorem 5.8)

yields the following corollary.
7.16 Corollary. For any finite group G of permutations of a finite set X, the
polynomial ZG (1 + q, 1 + q 2 , 1 + q 3 , . . . ) has symmetric, unimodal, integer
coefficients.
7.17 Example. (a) For the poset P of Example 5.4(a) we have G = {(1)(2)(3),
(1, 2)(3)}, so ZG (z1 , z2 , z3 ) = 12 (z13 + z1 z2 ). Hence
1
3
(#Pi )q i = (1 + q)3 + (1 + q)(1 + q 2 )
2
i=0
= 1 + 2q + 2q 2 + q 3 .
(b) For the poset P of Example 5.4(b) we have G = {(1)(2)(3)(4)(5),

(1, 2, 3, 4, 5), (1, 3, 5, 2, 4), (1, 4, 2, 5, 3), (1, 5, 4, 3, 2)}, so ZG (z1 , z2 , z3 , z4 ,
z5 ) = 15 (z15 + 4z5 ). Hence
1
5
(#Pi )q i = (1 + q)5 + 4(1 + q 5 )
5
i=0
= 1 + q + 2q 2 + 2q 3 + q 4 + q 5 .
Note that we are equivalently counting two-colored necklaces (as defined

in Example 7.9), say with colors red and blue, of length five according to the
number of blue beads.
(c) Let X be the squares of a 2 × 2 chessboard, labelled as follows:
Let G be the wreath product S2 S2 , as defined in Chapter 6. Then
G = {(1)(2)(3)(4), (1, 2)(3)(4), (1)(2)(3, 4), (1, 2)(3, 4),
(1, 3)(2, 4), (1, 4)(2, 3), (1, 3, 2, 4), (1, 4, 2, 3)},
so
1 4
ZG (z1 , z2 , z3 , z4 ) = (z + 2z12 z2 + 3z22 + 2z4 ).
8 1
Hence
1
4
(#Pi )q i = (1 + q)4 + 2(1 + q)2 (1 + q 2 ) + 3(1 + q 2 )2 + 2(1 + q 4 )
8
i=0
= 1 + q + 2q 2 + q 3 + q 4

4
= ,
2
agreeing with Theorem 6.6.

Using more sophisticated methods (such as the representation theory of the
symmetric group), the following generalization of Corollary 7.16 can be proved:
let P (q) be any polynomial with symmetric, unimodal, nonnegative, integer
coefficients, such as 1 + q + 3q 2 + 3q 3 + 8q 4 + 3q 5 + 3q 6 + q 7 + q 8 or
q 5 + q 6 (= 0 + 0q + · · · + 0q 4 + q 5 + q 6 + 0q 7 + · · · + 0q 11 ). Then the
polynomial ZG (P (q), P (q 2 ), P (q 3 ), . . . ) has symmetric, unimodal, nonnegative,
integer coefficients.
Graphs. A standard application of Pólya theory is to the enumeration of noniso-
morphic graphs. We saw at the end of Chapter 5 that if M is an m-element vertex
(2)
set, X = M 2 , and Sm is the group of permutations of X induced by permutations
of M, then an orbit of i-element subsets of X may be regarded as an isomorphism
(2)
class of graphs on the vertex set M with i edges. Thus #(BX /Sm )i is the number
of nonisomorphic graphs (without loops or multiple edges) on the vertex set M
with i edges. It follows from Corollary 7.15 that if gi (m) denotes the number of
nonisomorphic graphs with m vertices and i edges, then
(m2 )

gi (m)q i = ZS(2) (1 + q, 1 + q 2 , 1 + q 3 , . . . ).
m
i=0
Thus we would like to compute the cycle enumerator ZS(2) (z1 , z2 , . . . ). If two
m
permutations π and σ of M have the same cycle type (number of cycles of each
length), then their actions on X also have the same cycle type [why?]. Thus for
each possible cycle type of a permutation of M (i.e., for each partition of m) we

need to compute the induced cycle type on X. We also know from Theorem 7.12
the number of permutations of M of each type. For small values of m we can pick
some permutation π of each type and compute directly its action on X in order to
determine the induced cycle type. For m = 4 we have:
Cycle Induced Cycle

lengths permutation lengths
of π Number π π of π
1, 1, 1, 1 1 (1)(2)(3)(4) (12)(13)(14)(23)(24)(34) 1, 1, 1, 1, 1, 1
2, 1, 1 6 (1, 2)(3)(4) (12)(13, 23)(14, 24)(34) 2, 2, 1, 1
3, 1 8 (1, 2, 3)(4) (12, 23, 13)(14, 24, 34) 3, 3
2, 2 3 (1, 2)(3, 4) (12)(13, 24)(14, 23)(34) 2, 2, 1, 1
4 6 (1, 2, 3, 4) (12, 23, 34, 14)(13, 24) 4, 2
It follows that
1 6
ZS(2) (z1 , z2 , z3 , z4 , z5 , z6 ) = (z + 9z12 z22 + 8z32 + 6z2 z4 ).
4 24 1
If we set zi = 1 + q i and simplify, we obtain the polynomial

6
gi (4)q i = 1 + q + 2q 2 + 3q 3 + 2q 4 + q 5 + q 6 .
i=0
Indeed, this polynomial agrees with the rank-generating function of the poset of
Figure 5.1.
Suppose that we instead want to count the number hi (4) of nonisomorphic graphs
with four vertices and i edges, where now we allow at most two edges between any
(2)
two vertices. We can take M, X, and G = S4 as before, but now we have three
colors: red for no edges, blue for one edge, and yellow for two edges. A monomial
r i bj y k corresponds to a coloring with i pairs of vertices having no edges between
them, j pairs having one edge, and k pairs having two edges. The total number e of
edges is j + 2k. Hence if we let r = 1, b = q, y = q 2 , then the monomial r i bj y k
becomes q j +2k = q e . It follows that

i(i−1)
hi (4)q i = ZS(2) (1 + q + q 2 , 1 + q 2 + q 4 , 1 + q 3 + q 6 , . . . )
4
i=0
1
= (1 + q + q 2 )6 + 9(1 + q + q 2 )2 (1 + q 2 + q 4 )2
24

+8(1 + q 3 + q 6 )2 + 6(1 + q 2 + q 4 )(1 + q 4 + q 8 )
= 1 + q + 3q 2 + 5q 3 + 8q 4 + 9q 5 + 12q 6 + 9q 7 + 8q 8 + 5q 9
+3q 10 + q 11 + q 12 .
The total number

of nonisomorphic graphs on four vertices with edge multiplicities
at most two is i hi (4) = 66.
It should now be clear that if we restrict the edge multiplicity to be r, then the
corresponding generating function is ZS(2) (1 + q + q 2 + · · · + q r−1 , 1 + q 2 + q 4 +
4
· · · + q 2r−2 , . . . ). In particular, to obtain the total number N (r, 4) of nonisomorphic
graphs on four vertices with edge multiplicity at most r, we simply set each zi = r,
obtaining
N(r, 4) = ZS(2) (r, r, r, r, r, r)

4
1 6
= (r + 9r 4 + 14r 2 ).
24

This is the same as number of inequivalent r-colorings of the set X = M 2 (where
#M = 4) [why?].
Of course the same sort of reasoning can be applied to any number of vertices.
For five vertices our table becomes the following (using such notation as 15 to denote
a sequence of five 1’s).
Cycle Induced Cycle

lengths permutation lengths
of π Number π π of π
15 1 (1)(2)(3)(4)(5) (12)(13) · · · (45) 110
2, 13 10 (1, 2)(3)(4)(5) (12)(13, 23)(14, 25)(15, 25)(34)(35)(45) 23 , 14
3, 12 20 (1, 2, 3)(4)(5) (12, 23, 13)(14, 24, 34)(15, 25, 35)(45) 33 , 1
22 , 1 15 (1, 2)(3, 4)(5) (12)(13, 24)(14, 23)(15, 25)(34)(35, 45) 24 , 12
4, 1 30 (1, 2, 3, 4)(5) (12, 23, 34, 14)(13, 24)(15, 25, 35, 45) 42 , 2
3, 2 20 (1, 2, 3)(4, 5) (12, 23, 13)(14, 25, 34, 15, 24, 35)(45) 6, 3, 1
5 24 (1, 2, 3, 4, 5) (12, 23, 34, 45, 15)(13, 24, 35, 14, 25) 52
Thus
1 10
ZS(2) (z1 ,. . . ,z10 ) = (z +10z14 z23 +20z1 z33 +15z12 z24 +30z2 z42 +20z1 z3 z6 +24z52 ),
5 120 1
from which we compute

10
gi (5)q i = ZS(2) (1 + q, 1 + q 2 , . . . , 1 + q 10 )
5
i=0
= 1 + q + 2q 2 + 4q 3 + 6q 4 + 6q 5 + 6q 6 + 4q 7 + 2q 8 + q 9 + q 10 .
For an arbitrary number m = #M of vertices there exist explicit formulas for the
cycle indicator of the induced action of π ∈ SM on M 2 , thereby obviating the need
to compute π explicitly as we did in the above tables, but the overall expression for
ZS(2) cannot be simplified significantly or put into a simple generating function as
m
we did in Theorem 7.13. For reference we record
1 15
ZS(2) = (z + 15z17 z24 + 40z13 z34 + 45z13 z26 + 90z1 z2 z43 + 120z1 z2 z32 z6
6 6! 1
+144z53 + 15z13 z26 + 90z1 z2 z43 + 40z35 + 120z3 z62 )
(g0 (6), g1 (6), . . . , g15 (6)) = (1, 1, 2, 5, 9, 15, 21, 24, 24, 21, 15, 9, 5, 2, 1, 1).
Moreover if u(n) denotes the number of nonisomorphic simple graphs with n

vertices, then
(u(0), u(1), . . . , u(11))
= (1, 1, 2, 4, 11, 34, 156, 1044, 12346, 274668, 12005168, 1018997864).
A table of u(n) for n ≤ 75 is given at
http : //oeis.org/A000088/b000088.txt
In particular,
u(75) = 91965776790545918117055311393231179873443957239
0555232344598910500368551136102062542965342147
8723210428876893185920222186100317580740213865
7140377683043095632048495393006440764501648363
4760490012493552274952950606265577383468983364
6883724923654397496226869104105041619919159586
8518775275216748149124234654756641508154401414
8480274454866344981385848105320672784068407907
1134767688676890584660201791139593590722767979
8617445756819562952590259920801220117529208077
0705444809177422214784902579514964768094933848
3173060596932480677345855848701061537676603425
1254842843718829212212327337499413913712750831
0550986833980707875560051306072520155744624852
0263616216031346723897074759199703968653839368
77636080643275926566803872596099072,
a number of 726 digits! Compare

75
2( 2 )
= .9196577679054591809 · · · × 10726 ,
75!
which agrees with u(75) to 17 significant digits [why?].
Notes for Chapter 7
Burnside’s lemma (Lemma 7.2) was actually first stated and proved by Frobenius
[47, end of §4]. Frobenius in turn credits Cauchy [22, p. 286] for proving the lemma
in the transitive case. Burnside, in the first edition of his book [16, §118–119],
attributes the lemma to Frobenius, but in the second edition [17] this citation is
absent. For more on the history of Burnside’s lemma, see [96] and [145]. Many
authors now call this result the Cauchy–Frobenius lemma. The cycle indicator
ZG (z1 , z2 , . . . ) (where G is a subgroup of Sn ) was first considered by Redfield
[108], who called it the group reduction function, denoted Grf(G). Pólya [101]
independently defined the cycle indicator, proved the fundamental Theorem 7.7,
and gave numerous applications. For an English translation of Pólya’s paper,
see [102]. Much of Pólya’s work was anticipated by Redfield. For interesting
historical information about the work of Redfield and its relation to Pólya theory,
see [58, 60, 82, 109] (all in the same issue of Journal of Graph Theory). The
Wikipedia article “John Howard Redfield” also gives information and references
on the interesting story of the rediscovery and significance of Redfield’s work.
The application of Pólya’s theorem to the enumeration of nonisomorphic graphs
appears in Pólya’s original paper [101]. For much additional work on graphical
enumeration, see the text of Harary and Palmer [59].
Subsequent to Pólya’s work there have been a huge number of expositions, appli-
cations, and generalizations of Pólya theory. An example of such a generalization
appears in Exercise 7.14. We mention here only the nice survey [31] by de Bruijn.
Theorem 7.13 (the generating function for the cycle indicator ZS of the
symmetric group S ) goes back to Frobenius (see [48, bottom of p. 152 of GA])
and Hurwitz [70, §4]. It is clear that they were aware of Theorem 7.13, even if they
did not state it explicitly. For a more conceptual approach and further aspects see
Stanley [131, §§5.1–5.2].
1. Verify (7.6), i.e., the formula for the cycle enumerator of the dihedral group
with its defining action.
2. For a simple graph with vertex set V , we can define an automorphism of to
be a bijection ϕ : V → V such that u and v are adjacent if and only if ϕ(u) and
ϕ(v) are adjacent. The automorphisms form a group under composition, called
the automorphism group Aut( ) of . Let be the graph shown below.
Let G be the automorphism group of , so G has order eight.

(a) What is the cycle index polynomial of G, acting on the vertices of ?
(b) In how many ways can one color the vertices of in n colors, up to the
symmetry of ?
3. (*) What is the total number of inequivalent ways to color the vertices of a
regular dodecahedron with the colors dark red and burgundy, up to the 120
rotations and reflections of the dodecahedron? Do this computation in your
head.
4. A regular tetrahedron T has four vertices, six edges, and four triangles. The
rotational symmetries of T (no reflections allowed) form a group G of order
12.
(a) What is the cycle index polynomial of G acting on the vertices of T ?
(b) In how many ways can the vertices of T be colored in n colors, up to
rotational symmetry?
(c) What about coloring the six edges of T , up to rotational symmetry?
5. Let G be the graph on the vertex set {1, 2, . . . , 4r} which is a cycle of length
4r, r ≥ 2, together with an edge from some vertex to its antipode. (Thus G has
4r + 1 edges.) How many ways are there to color the vertices of G from a set
of n colors, up to isomorphism? (Colors may be repeated. Not all n colors need
to be used. Adjacent vertices are allowed to have the same color.)
6. How many necklaces (up to cyclic symmetry) have n red beads and n blue
beads? (Express your answer as a sum over all divisors d of n.)
7. (not directly related to the text) A primitive necklace is a necklace with no
symmetries, i.e., no nonidentity rotation of the necklace preserves the necklace.
Let M (n) denote the number of primitive n-colored necklaces with beads.
Show that
1
M (n) = μ(/d)nd ,

d|
where μ denotes the Möbius function from number theory. (Compare (7.5).)
8. Ten balls are stacked in a triangular array with 1 atop 2 atop 3 atop 4. (Think of
billiards.) The triangular array is free to rotate in two dimensions.
(a) Find the generating function for the number of inequivalent colorings using
the ten colors r1 , r2 , . . . , r10 . (You don’t need to simplify your answer.)
(b) How many inequivalent colorings have four red balls, three green balls, and
three chartreuse balls? How many have four red balls, four turquoise balls,
and two aquamarine balls?
9. The dihedral group D4 of order 8 acts on the set X of 64 squares of an 8 × 8
chessboard B. Find the number of ways to choose two subsets S ⊆ T of X,
up to the action of D4 . For instance, all eight ways to choose S to be a single
corner square s and T to be {s, t}, where t is adjacent to s (i.e., has an edge in
common with s), belong to the same orbit of D4 . Write your answer as a (short)
finite sum.
10. For any finite group G of permutations of an -element set X, let f (n) be the
number of inequivalent (under the action of G) colorings of X with n colors.
Find limn→∞ f (n)/n . Interpret your answer as saying that “most” colorings
of X are asymmetric (have no symmetries).
11. Let X be a finite set, and let G be a subgroup of the symmetric group SX .
Suppose that the number of orbits of G acting on n-colorings of X is given by
the polynomial
1 p
f (n) = (n + bnp−2 + · · · + (p − 1)n),
a
where p is prime.
(a) What is the order (number of elements) of G?
(b) What is the size #X of X?
(c) How many transpositions are in G? A transposition is a permutation that
transposes (interchanges) two elements of X and leaves the remaining
elements fixed.
(d) How many orbits does G have acting on X?
(e) Show that G is either a cyclic group or is not a simple group. A group G
with more than one element is simple if its only normal subgroups are G
and {1}.
12. It is known that there exists a nonabelian group G of order 27 such that x 3 = 1
for all x ∈ G. Use this fact to give an example of two nonisomorphic finite
subgroups G and H of SX for some finite set X such that ZG = ZH .
13. (somewhat difficult) Let NG (n) be the polynomial of Theorem 7.5, and let
#X = d. Show that (−1)d NG (−n) is equal to the number of inequivalent n-
colorings f : X → [n] of X such that the subgroup H of G fixing f (i.e.,

π · f = f for all π ∈ H ) is contained in the alternating group AX . This result
could be called a reciprocity theorem for the polynomial NG (n).
14. (difficult) Suppose that a finite group G acts on a finite set X and another finite
group H acts on a set C of colors. Call two colorings f, g : X → C equivalent
if there are permutations π ∈ G and σ ∈ H such that
f (x) = σ · g(π · x), for all x ∈ X.
Thus we are allowed not only to permute the elements of X by some element
of G but also to permute the colors by some element of H . Show that the total
number of inequivalent colorings is given by

∂ ∂ ∂
ZG , , ,··· ZH (ez1 +z2 +z3 +··· , e2(z2 +z4 +z6 +··· ) , e3(z3 +z6 +z9 +··· ) , . . . ),
∂z1 ∂z2 ∂z3
evaluated at z1 = z2 = z3 = · · · = 0.
Example. Let n be the number of two-colored necklaces of four beads, where
we may also interchange the two colors to get an equivalent coloring. Thus
ZG = 14 (z14 + z22 + 2z4 ) and ZH = 12 (z12 + z2 ). Hence

1 1 ∂4 ∂2 ∂
n= · + 2 +2 (e2(z1 +z2 +z3 +··· ) + e2(z2 +z4 +z6 +··· ) )|zi =0
4 2 ∂z14 ∂z2 ∂z4

1 ∂4 ∂2 ∂ (2z1 )4 (2z2 )2 2z4 (2z2 )4 2z4
= + +2 + + + +
8 ∂z14 ∂z22 ∂z4 4! 2! 1! 4! 1!
z =0 i
1
= (16 + 4 + 4 + 4 + 4)
8
= 4.
The four different necklaces are 0000, 1000, 1100, and 1010.
15. (a) Let e6 (n) denote the number of permutations π ∈ Sn satisfying π 6 = ι (the
identity permutation). Find a simple formula for the generating function
xn
E6 (x) = e6 (n) .
n!
n≥0
(b) Generalize to ek (n) = #{π ∈ Sn : π k = ι} for any k ≥ 1.

16. Let f (n) be the number of permutations in the symmetric group Sn all of
whose cycles have even length. For instance, f (4) = 9 and f (11) = 0.
(a) Let
∞
xn
F (x) = f (n) .
n!
n=0
Find a simple expression for F (x). Your answer should not involve any
summation symbols (or their equivalent), logarithms, or the function ex .
(b) Use (a) to find a simple formula for f (n).
(c) Give a combinatorial proof of (b).
17. (difficult) Give a combinatorial proof of Proposition 7.14. Despite the similarity
between Proposition 7.14 and Exercise 7.16, the latter is much easier to prove
combinatorially than the former.
18. Let c(w) denote the number of cycles of a permutation w ∈ Sn . Let f (n)
denote the average value of c(w)(c(w) − 1) for w ∈ Sn , i.e.,
1
f (n) = c(w)(c(w) − 1).
n!
w∈Sn
(Set f (0) =
1.) Find a simple formula for the generating function
n
n≥0 f (n)t .
19. (a) (*) Let n ≥ 1, and let G be a subgroup of Sn of odd order. Show that the
quotient poset Bn /G has the same number of elements of even rank as of
odd rank.
(b) Generalize (a) as follows: give a necessary and sufficient condition on a
subgroup G of Sn , in terms of the cycle lengths of elements of G, for
Bn /G to have the same number of elements of even rank as of odd rank.
20. Let c(, k) denote the number of permutations in S with k cycles. Show that
the sequence
c(, 1), c(, 2), . . . , c(, )
is strongly log-concave.
Chapter 8
A Glimpse of Young Tableaux
We defined in Chapter 6 Young’s lattice Y , the poset of all partitions of all

nonnegative integers, ordered by containment of their Young diagrams.
11111 2111 221 311 32 41 5
1111 211 22 31 4
111 21 3
11 2
Here we will be concerned with the counting of certain walks in the Hasse diagram
(considered as a graph) of Y . Note that since Y is infinite, we cannot talk about its
eigenvalues and eigenvectors. We need different techniques for counting walks. It
will be convenient to denote the length of a walk by n, rather than by as in previous
chapters.
Note that Y is a graded poset (of infinite rank), with Yi consisting of all partitions
of i. In other words, we have Y = Y0 ∪Y · 1 ∪· · · · (disjoint union), where every
maximal chain intersects each level Yi exactly once. We call Yi the ith level of
Y , just as we did for finite graded posets.
Since the Hasse diagram of Y is a simple graph (no loops or multiple edges), a
walk of length n is specified by a sequence λ0 , λ1 , . . . , λn of vertices of Y . We will
call a walk in the Hasse diagram of a poset a Hasse walk. Each λi is a partition
of some integer, and we have either (a) λi < λi+1 and |λi | = |λi+1 | − 1, or

https://doi.org/10.1007/978-3-319-77173-1_8
104 8 A Glimpse of Young Tableaux
(b) λi > λi+1 and |λi | = |λi+1 | + 1. (Recall that for a partition λ, we write
|λ| for the sum of the parts of λ.) A step of type (a) is denoted by U (for “up,”
since we move up in the Hasse diagram), while a step of type (b) is denoted by
D (for “down”). If the walk W has steps of types A1 , A2 , . . . , An , respectively,
where each Ai is either U or D, then we say that W is of type An An−1 · · · A2 A1 .
Note that the type of a walk is written in the opposite order to that of the walk.
This is because we will soon regard U and D as linear transformations, and
we multiply linear transformations right-to-left (opposite to the usual left-to-right
reading order). For instance (abbreviating a partition (λ1 , . . . , λm ) as λ1 · · · λm ), the
walk ∅, 1, 2, 1, 11, 111, 211, 221, 22, 21, 31, 41 is of type U U DDU U U U DU U =
U 2 D 2 U 4 DU 2 .
There is a nice combinatorial interpretation of walks of type U n which begin at
∅. Such walks are of course just saturated chains ∅ = λ0 λ1 · · · λn . In other
words, they may be regarded as sequences of Young diagrams, beginning with the
empty diagram and adding one new square at each step. An example of a walk of
type U 5 is given by
We can specify this walk by taking the final diagram and inserting an i into square
s if s was added at the ith step. Thus the above walk is encoded by the “tableau”
1 2
3 5
4
Such an object τ is called a standard Young tableaux (or SYT). It consists of the
Young diagram D of some partition λ of an integer n, together with the numbers
1, 2, . . . , n inserted into the squares of D, so that each number appears exactly once,
and every row and column is increasing. We call λ the shape of the SYT τ , denoted
λ = sh(τ ). For instance, there are five SYT of shape (2, 2, 1), given by
1 2 1 2 1 3 1 3 1 4
3 4 3 5 2 4 2 5 2 5
5 4 5 4 3
Let f λ denote the number of SYT of shape λ, so for instance f (2,2,1) = 5. The
numbers f λ have many interesting properties; for instance, there is a famous explicit
formula for them known as the Frame–Robinson–Thrall hook length formula. For
the sake of completeness we state this formula without proof, though it is not needed
in what follows.
8 A Glimpse of Young Tableaux 105
Let u be a square of the Young diagram of the partition λ. Define the hook H (u)
of u (or at u) to be the set of all squares directly to the right of u or directly below
u, including u itself. The size (number of squares) of H (u) is called the hook length
of u (or at u), denoted h(u). In the diagram of the partition (4, 2, 2) below, we have
inserted the hook length h(u) inside each square u.
6 5 2 1
3 2
2 1
8.1 Theorem (Hook Length Formula). Let λ n. Then
n!
fλ = 0 .
u∈λ h(u)
Here the notation u ∈ λ means that u ranges over all squares of the Young diagram
of λ.
For instance, the diagram of the hook lengths of λ = (4, 2, 2) above gives
8!
f (4,2,2) = = 56.
6·5·2·1·3·2·2·1
In this chapter we will be concerned with the connection between SYT and
counting walks in Young’s lattice. If w = An An−1 · · · A1 is some word in U and D
and λ n, then let us write α(w, λ) for the number of Hasse walks in Y of type w
which start at the empty partition ∅ and end at λ. For instance, α(U DU U, 11) = 2,
the corresponding walks being ∅, 1, 2, 1, 11 and ∅, 1, 11, 1, 11. Thus in particular
α(U n , λ) = f λ [why?]. In a similar fashion, since the number of Hasse walks of
type D n U n which begin at ∅, go up to a partition λ n, and then back down to ∅ is
given by (f λ )2 , we have

α(D n U n , ∅) = (f λ )2 . (8.1)
λn
Our object is to find an explicit formula for α(w, λ) of the form f λ cw , where cw
does not depend on λ. (It is by no means a priori obvious that such a formula should
exist.) In particular, since f ∅ = 1, we will obtain by setting λ = ∅ a simple formula
for the number of (closed) Hasse walks of type w from ∅ to ∅ (thus including a
simple formula for (8.1)).
There is an easy condition for the existence of any Hasse walk of type w from ∅
to λ, given by the next lemma.
8.2 Lemma. Suppose w = D sk U rk · · · D s2 U r2 D s1 U r1 , where ri ≥ 0 and si ≥ 0.

Let λ n. Then there exists a Hasse walk of type w from ∅ to λ if and only if:

k
(ri − si ) = n
i=1

j
(ri − si ) ≥ 0 for 1 ≤ j ≤ k.
i=1
Proof.
Since each U moves up one level and each D moves down one level, we see
that ki=1 (ri − si ) is the level at which a walk of type w beginning at ∅ ends. Hence
k
i=1 (ri − si ) = |λ| = n.
j j
After i=1 (ri + si ) steps we will be at level i=1 (ri − si ). Since the lowest
j
level is level 0, we must have i=1 (ri − si ) ≥ 0 for 1 ≤ j ≤ k.
The easy proof that the two conditions of the lemma are sufficient for the
existence of a Hasse walk of type w from ∅ to λ is left to the reader.
If w is a word in U and D satisfying the conditions of Lemma 8.2, then we say

that w is a valid λ-word. Note that the condition of being a valid λ-word depends
only on |λ|.
The proof of our formula for α(w, λ) will be based on linear transformations
analogous to those defined by (4.2) and (4.3). As in Chapter 4 let RYj be the real
vector space with basis Yj . Define two linear transformations Ui : RYi → RYi+1
and Di : RYi → RYi−1 by

Ui (λ) = μ
μi+1
λ<μ

Di (λ) = ν,
νi−1
ν<λ
for all λ i. For instance (using abbreviated notation for partitions)
U21 (54422211) = 64422211 + 55422211 + 54432211 + 54422221 + 544222111
D21 (54422211) = 44422211 + 54322211 + 54422111 + 5442221.
It is clear [why?] that if r is the number of distinct (i.e., unequal) parts of λ, then
Ui (λ) is a sum of r + 1 terms and Di (λ) is a sum of r terms. The next lemma is an
analogue for Y of the corresponding result for Bn (Lemma 4.6).
8.3 Lemma. For any i ≥ 0 we have
Di+1 Ui − Ui−1 Di = Ii , (8.2)
the identity linear transformation on RYi .
Proof. Apply the left-hand side of (8.2) to a partition λ of i, expand in terms of

the basis Yi , and consider the coefficient of a partition μ. If μ = λ and μ can
be obtained from λ by adding one square s to (the Young diagram of) λ and then
removing a (necessarily different) square t, then there is exactly one choice of s and
t. Hence the coefficient of μ in Di+1 Ui (λ) is equal to 1. But then there is exactly
one way to remove a square from λ and then add a square to get μ, namely, remove
t and add s. Hence the coefficient of μ in Ui−1 Di (λ) is also 1, so the coefficient of
μ when the left-hand side of (8.2) is applied to λ is 0.
If now μ = λ and we cannot obtain μ by adding a square and then deleting a
square from λ (i.e., μ and λ differ in more than two rows), then clearly when we
apply the left-hand side of (8.2) to λ, the coefficient of μ will be 0.
Finally consider the case λ = μ. Let r be the number of distinct (unequal) parts
of λ. Then the coefficient of λ in Di+1 Ui (λ) is r + 1, while the coefficient of λ in
Ui−1 Di (λ) is r, since there are r + 1 ways to add a square to λ and then remove it,
while there are r ways to remove a square and then add it back in. Hence when we
apply the left-hand side of (8.2) to λ, the coefficient of λ is equal to 1.
Combining the conclusions of the three cases just considered shows that the left-
hand side of (8.2) is just Ii , as was to be proved.
We come to one of the main results of this chapter.

8.4 Theorem. Let λ be a partition and w = An An−1 · · · A1 a valid λ-word. Let
Sw = {i : Ai = D}. For each i ∈ Sw , let ai be the number of D’s in w to the right
of Ai , and let bi be the number of U ’s in w to the right of Ai . Thus ai − bi is the
level we occupy in Y before taking the step Ai = D. Then

α(w, λ) = f λ (bi − ai ).
i∈Sw
Before proving Theorem 8.4, let us give an example. Suppose w =

U 3 D 2 U 2 DU 3 = U U U DDU U DU U U and λ = (2, 2, 1). Then Sw = {4, 7, 8}
and a4 = 0, b4 = 3, a7 = 1, b7 = 5, a8 = 2, b8 = 5. We have also seen earlier that
f 221 = 5. Thus
α(w, λ) = 5(3 − 0)(5 − 1)(5 − 2) = 180.
Proof. Proof of Theorem 8.4. For notational simplicity we will omit the subscripts
from the linear transformations Ui and Di . This should cause no confusion since the
subscripts will be uniquely determined by the elements on which U and D act. For
instance, the expression U DU U (λ) where λ i must mean Ui+1 Di+2 Ui+1 Ui (λ);
otherwise it would be undefined since Uj and Dj can only act on elements of RYj ,
and moreover Uj raises the level by one while Dj lowers it by one.
By (8.2) we can replace DU in any word y in the letters U and D by U D+I . This
replaces y by a sum of two words, one with one fewer D and the other with one D
moved one space to the right. For instance, replacing the first DU in U U DU DDU
by U D + I yields U U U DDDU + U U DDU . If we begin with the word w and
iterate this procedure, replacing a DU in any word with U D + I , eventually there
will be no U ’s to the right of any D’s and the procedure will come to an end. At this
point we will have expressed w as a linear combination (with integer coefficients)
of words of the form U i D j . Since the operation of replacing DU with U D + I
preserves the difference between the number of U ’s and D’s in each word, all the
words U i D j which appear will have i − j equal to some constant n (namely, the
number of U ’s minus the number of D’s in w). Specifically, say we have

w= rij (w)U i D j , (8.3)
i−j =n
where each rij (w) ∈ Z. (We also define rij (w) = 0 if i < 0 or j < 0.) We claim
that the rij (w)’s are uniquely determined by w. Equivalently [why?], if we have

dij U i D j = 0 (8.4)
i−j =n
(as an identity of linear transformations acting on the space RYk for any k), where
each dij ∈ Z (or dij ∈ R, if you prefer), then each dij = 0. Let j be the least
integer for which dj +n,j = 0. Let μ j , and apply both sides of (8.4) to μ. The
left-hand side has exactly one nonzero term, namely, the term with j = j [why?].
The right-hand side, on the other hand,1 is 0, a contradiction. Thus the rij (w)’s are
unique.
Now apply U on the left to (8.3). We get

Uw = rij (w)U i+1 D j .
i,j
Hence (using uniqueness of the rij ’s) there follows [why?]
rij (U w) = ri−1,j (w). (8.5)
We next want to apply D on the left to (8.3). It is easily proved by induction on

i (left as an exercise) that
DU i = U i D + iU i−1 . (8.6)
1 The phrase “the right-hand side, on the other hand” does not mean the left-hand side!
(We interpret U −1 as being 0 and U 0 = I , so that (8.6) is true for i = 0, 1.) Hence

Dw = rij (w)DU i D j
i,j

= rij (w)(U i D + iU i−1 )D j ,
i,j
from which it follows [why?] that
rij (Dw) = ri,j −1 (w) + (i + 1)ri+1,j (w). (8.7)
Setting j = 0 in (8.5) and (8.7) yields
ri0 (U w) = ri−1,0 (w) (8.8)
ri0 (Dw) = (i + 1)ri+1,0 (w). (8.9)
Now let (8.3) operate on ∅. Since D j (∅) = 0 for all j > 0, we get w(∅) =
rn0 (w)U n (∅). Thus the coefficient of λ in w(∅) is given by
α(w, λ) = rn0 (w)α(U n , λ) = rn0 f λ ,
where as usual λ n. It is clear from (8.8) and (8.9) that

rn0 (w) = (bj − aj ),
j ∈Sw
and the proof follows.
NOTE. It is possible to give a simpler proof of Theorem 8.4, but the proof we
have given is useful for generalizations not appearing here.
An interesting special case of the previous theorem allows us to evaluate (8.1).

α(D n U n , ∅) = (f λ )2 = n!.
λn
Proof. When w = D n U n in Theorem 8.4 we have Sw = {n + 1, n + 2, . . . , 2n},

ai = i − n − 1, and bi = n, from which the proof is immediate.
NOTE (for those familiar with the representation theory of finite groups). It can
be shown that the numbers f λ , for λ n, are the degrees of the irreducible
representations of the symmetric group Sn . Given this, Corollary 8.5 is a special
case of the result that the sum of the squares of the degrees of the irreducible
representations of a finite group G is equal to the order #G of G. There are many

other intimate connections between the representation theory of Sn , on the one
hand, and the combinatorics of Young’s lattice and Young tableaux, on the other.
There is also an elegant combinatorial proof of Corollary 8.5, based on the RSK
algorithm (after Gilbert de Beauregard Robinson, Craige Schensted, and Donald
Knuth) or Robinson–Schensted correspondence, with many fascinating properties
and with deep connections to representation theory. In the first Appendix at the end
of this chapter we give a description of the RSK algorithm and the combinatorial
proof of Corollary 8.5.
We now consider a variation of Theorem 8.4 in which we are not concerned
with the type w of a Hasse walk from ∅ to λ, but only with the number of steps.
For instance, there are three Hasse walks of length three from ∅ to the partition
1, given by ∅, 1, ∅, 1; ∅, 1, 2, 1; and ∅, 1, 11, 1. Let β(, λ) denote the number of
Hasse walks of length from ∅ to λ. Note the two following easy facts:
(F1) β(, λ) = 0 unless ≡ |λ| (mod 2).
(F2) β(, λ) is the coefficient of λ in the expansion of (D + U ) (∅) as a linear
combination of partitions.
Because of (F2) it is important to write (D +U ) as a linear combination of terms
U i Dj ,
just as in the proof of Theorem 8.4 we wrote a word w in U and D in this
form. Thus define integers bij () by

(D + U ) = bij ()U i D j . (8.10)
i,j
Just as in the proof of Theorem 8.4, the numbers bij () exist and are well defined.
8.6 Lemma. We have bij () = 0 if − i − j is odd. If − i − j = 2m then
!
bij () = . (8.11)
2m i! j ! m!
Proof. The assertion for −i −j odd is equivalent to (F1) above, so assume −i −j
is even. The proof is by induction on . It’s easy to check that (8.11) holds for = 1.
Now assume true for some fixed ≥ 1. Using (8.10) we obtain

bij ( + 1)U i D j = (D + U )+1
i,j

= (D + U ) bij ()U i D j
i,j

= bij ()(DU i D j + U i+1 D j ).
i,j
In the proof of Theorem 8.4 we saw that DU i = U i D + iU i−1 (see (8.6)). Hence
we get

bij ( + 1)U i D j = bij ()(U i D j +1 + iU i−1 D j + U i+1 D j ). (8.12)
i,j i,j
As mentioned after (8.10), the expansion of (D +U )+1 in terms of U i D j is unique.

Hence equating coefficients of U i D j on both sides of (8.12) yields the recurrence
bij ( + 1) = bi,j −1 () + (i + 1)bi+1,j () + bi−1,j (). (8.13)
It is a routine matter to check that the function !/2m i!j !m! satisfies the same
recurrence (8.13) as bij (), with the same initial condition b00 (0) = 1. From this
the proof follows by induction.
From Lemma 8.6 it is easy to prove the following result.

8.7 Theorem. Let ≥ n and λ n, with − n even. Then

β(, λ) = (1 · 3 · 5 · · · ( − n − 1))f λ .
n
Proof. Apply both sides of (8.10) to ∅. Since U i D j (∅) = 0 unless j = 0, we get

(D + U ) (∅) = bi0 ()U i (∅)
i

= bi0 () f λ λ.
i λi

Since by Lemma 8.6 we have bi0 () = i (1 · 3 · 5 · · · ( − i − 1)) when − i is
even, the proof follows from (F2).
NOTE. The proof of Theorem 8.7 only required knowing the value of bi0 ().
However, in Lemma 8.6 we computed bij () for all j . We could have carried out
the proof so as only to compute bi0 (), but the general value of bij () is so simple
that we have included it too.
8.8 Corollary. The total number of Hasse walks in Y of length 2m from ∅ to ∅ is
given by
β(2m, ∅) = 1 · 3 · 5 · · · (2m − 1).
Proof. Simply substitute λ = ∅ (so n = 0) and = 2m in Theorem 8.7.

The fact that we can count various kinds of Hasse walks in Y suggests that there
may be some finite graphs related to Y whose eigenvalues we can also compute. This
is indeed the case, and we will discuss the simplest case here. (See Exercise 8.21
for a generalization.) Let Yj −1,j denote the restriction of Young’s lattice Y to ranks
j − 1 and j . Identify Yj −1,j with its Hasse diagram, regarded as a (bipartite) graph.
Let p(i) = #Yi , the number of partitions of i.
8.9 Theorem. The eigenvalues of Yj −1,j are given as follows: 0√is an eigenvalue of
multiplicity p(j ) − p(j − 1); and for 1 ≤ s ≤ j , the numbers ± s are eigenvalues
of multiplicity p(j − s) − p(j − s − 1).
Proof. Let A denote the adjacency matrix of Yj −1,j . Since RYj −1,j = RYj −1 ⊕RYj
(vector space direct sum), any vector v ∈ RYj −1,j can be written uniquely as v =
vj −1 + vj , where vi ∈ RYi . The matrix A acts on the vector space RYj −1,j as
follows [why?]:
A(v) = D(vj ) + U (vj −1 ). (8.14)
Just as Theorem 4.7 followed from Lemma 4.6, we deduce from Lemma 8.3 that for
any i we have that Ui : RYi → RYi+1 is one-to-one and Di : RYi → RYi−1 is onto.
It follows in particular that
dim ker(Di ) = dim RYi − dim RYi−1

= p(i) − p(i − 1),
where ker denotes kernel.

Case 1. Let v ∈ ker(Dj ), so v = vj . Then Av = Dv = 0. Thus ker(Dj ) is an
eigenspace of A for the eigenvalue 0, so 0 is an eigenvalue of multiplicity at least
p(j ) − p(j − 1).
Case 2. Let v ∈ ker(Ds ) for some 0 ≤ s ≤ j − 1. Let

v ∗ = ± j − sU j −1−s (v) + U j −s (v).
√
Note that v ∗ ∈ RYj −1,j , with vj∗−1 = ± j − sU j −1−s (v) and vj∗ = U j −s (v).
Using (8.6), we compute
A(v ∗ ) = U (vj∗−1 ) + D(vj∗ )

= ± j − s U j −s (v) + DU j −s (v)

= ± j − s U j −s (v) + U j −s D(v) + (j − s)U j −s−1 (v)

= ± j − s U j −s (v) + (j − s)U j −s−1 (v)

= ± j − s v∗. (8.15)
Appendix 1: The RSK Algorithm 113
It’s easy to verify (using the fact that U is one-to-one) that if v(1), . . . , v(t)
is a basis for ker(Ds ), then ∗ ∗
√ v(1) , . . . , v(t) are linearly independent. Hence
by (8.15) we have that ± j − s is an eigenvalue of A of multiplicity at least
t = dim ker(Ds ) = p(s) − p(s − 1).
We have found a total of
j −1

p(j ) − p(j − 1) + 2 (p(s) − p(s − 1)) = p(j − 1) + p(j )
s=0
√
eigenvalues
√ of A. (The factor 2 above arises from the fact that both + j − s and
− j − s are eigenvalues.) Since the graph Yj −1,j has p(j − 1) + p(j ) vertices, we
have found all its eigenvalues.
An elegant combinatorial consequence of Theorem 8.9 is the following.
8.10 Corollary. Fix j ≥ 1. The number of ways to choose a partition λ of j , then
delete a square from λ (keeping it a partition), then insert a square, then delete a
square, etc., for a total of m insertions and m deletions, ending back at λ, is given by

j
[p(j − s) − p(j − s − 1)]s m , m > 0. (8.16)
s=1
Proof. Exactly half the closed walks in Yj −1,j of length 2m begin at an element
of Yj [why?]. Hence if Yj −1,j has eigenvalues θ1 , . . . , θr , then by Corollary 1.3
the desired number of walks is given by 12 (θ12m + · · · + θr2m ). Using the values of
θ1 , . . . , θr given by Theorem 8.9 yields (8.16).
For instance, when j = 7, (8.16) becomes 4 + 2 · 2m + 2 · 3m + 4m + 5m + 7m .
When m = 1 we get 30, the number of edges of the graph Y6,7 [why?].
Appendix 1: The RSK Algorithm
We will describe a bijection between permutations π ∈ Sn and pairs (P , Q) of SYT

of the same shape λ n. Define a near Young tableau (NYT) to be the same as an
SYT, except that the entries can be any distinct integers, not necessarily the integers
1, 2, . . . , n. Let Pij denote the entry in row i and column j of P . The basic operation
of the RSK algorithm consists of the row insertion P ← k of a positive integer k
into an NYT P = (Pij ). The operation P ← k is defined as follows: let r be the
least integer such that P1r > k. If no such r exists (i.e., all elements of the first row
of P are less than k), then simply place k at the end of the first row. The insertion
process stops, and the resulting NYT is P ← k. If, on the other hand, r does exist
then replace P1r by k. The element k then “bumps” P1r := k into the second row,
i.e., insert k into the second row of P by the insertion rule just described. Either k
is inserted at the end of the second row, or else it bumps an element k to the third
row. Continue until an element is inserted at the end of a row (possibly as the first
element of a new row). The resulting array is P ← k.
8.11 Example. Let
3 7 9 14
6 11 12
P = 10 16
13
15
Then P ← 8 is shown below, with the elements inserted into each row (either by
bumping or by the final insertion in the fourth row) in boldface. Thus the 8 bumps
the 9, the 9 bumps the 11, the 11 bumps the 16, and the 16 is inserted at the end of
a row. Hence
3 7 8 14
6 9 12
(P ← 8) = 10 11 .
13 16
15
We omit the proof, which is fairly straightforward, that if P is an NYT, then

so is P ← k. We can now describe the RSK algorithm. Let π = a1 a2 · · · an ∈
Sn . We will inductively construct a sequence (P0 , Q0 ), (P1 , Q1 ), . . . , (Pn , Qn ) of
pairs (Pi , Qi ) of NYT of the same shape, where Pi and Qi each have i squares.
First, define (P0 , Q0 ) = (∅, ∅). If (Pi−1 , Qi−1 ) have been defined, then set Pi =
(Pi−1 ← ai ). In other words, Pi is obtained from Pi−1 by row inserting ai . Now
define Qi to be the NYT obtained from Qi−1 by inserting i so that Qi and Pi
have the same shape. (The entries of Qi−1 don’t change; we are simply placing i
into a certain new square and not row-inserting it into Qi−1 .) Finally let (P , Q) =
RSK
(Pn , Qn ). We write π −→ (P , Q).
8.12 Example. Let π = 4273615 ∈ S7 . The pairs (P1 , Q1 ), . . . , (P7 , Q7 ) =
(P , Q) are as follows:
Pi Qi
4 1
2 1
4 2
Appendix 2: Plane Partitions 115
27 13
4 2
23 13
47 24
236 135
47 24
136 135
27 24
4 6
135 135
26 24
47 67
8.13 Theorem. The RSK algorithm defines a bijection between the symmetric group
Sn and the set of all pairs (P , Q) of SYT of the same shape, where the shape λ is a
partition of n.
Sketch. The key step is to define the inverse of RSK. In other words, if π → (P , Q),
then how can we recover π uniquely from (P , Q)? Moreover, we need to find π for
any (P , Q). Observe that the position occupied by n in Q is the last position to
be occupied in the insertion process. Suppose that k occupies this position in P .
It was bumped into this position by some element j in the row above k that is
currently the largest element of its row less than k. Hence we can “inverse bump” k
into the position occupied by j , and now inverse bump j into the row above it by
the same procedure. Eventually an element will be placed in the first row, inverse
bumping another element t out of the tableau altogether. Thus t was the last element
of π to be inserted, i.e., if π = a1 a2 · · · an then an = t. Now locate the position
occupied by n − 1 in Q and repeat the procedure, obtaining an−1 . Continuing in
this way, we uniquely construct π one element at a time from right-to-left, such that
π → (P , Q).
The RSK-algorithm provides a bijective proof of Corollary 8.5, that is,

(f λ )2 = n!.
λn
Appendix 2: Plane Partitions
In this appendix we show how a generalization of the RSK algorithm leads to

an elegant generating function for a two-dimensional generalization of integer
partitions. A plane partition of an integer n ≥ 0 is a two-dimensional array
π = (πij )i,j ≥1 of integers πij ≥ 0 that is weakly decreasing in rows and

columns, i.e.,
πij ≥ πi+1,j , πij ≥ πi,j +1 ,

such that i,j πij = n. It follows that all but finitely many πij are 0, and these 0’s
are omitted in writing a particular plane partition π . Given a plane partition π , we
write |π | = n to denote that π is a plane partition of n. More generally, if L is any
array of nonnegative integers we write |L| for the sum of the parts (entries) of L.
There is one plane partition of 0, namely, all πij = 0, denoted ∅. The plane
partitions of the integers 0 ≤ n ≤ 3 are given by
∅ 1 2 11 1 3 21 111 11 2 1
1 1 1 1.
1
If pp(n) denotes the number of plane partitions of n, then pp(0) = 1, pp(1) = 1,

pp(2) = 3, and pp(3) = 6.
Our object is to give a formula for the generating function

F (x) = pp(n)x n = 1 + x + 3x 2 + 6x 3 + 13x 4 + 24x 5 + · · · .
n≥0
More generally, we will consider plane partitions with at most r rows and at most s
columns, i.e., πij = 0 for i > r or j > s. As a simple warmup, let us first consider
the case of ordinary partitions λ = (λ1 , λ2 , . . . ) of n.
8.14 Proposition. Let ps (n) denote the number of partitions of n with at most s
parts. Equivalently, ps (n) is the number of plane partitions of n with at most one
row and at most s columns [why?].Then

s
ps (n)x n = (1 − x k )−1 .
n≥0 k=1
Proof. First note that the partition λ has at most s parts if and only if the conjugate
partition λ defined inChapter 6 has largest part at most s. Thus it suffices to find the
generating function n≥0 ps (n)x n , where ps (n) denotes the number of partitions
of n whose largest part is at most s. Now expanding each factor (1 − x k )−1 as a
geometric series gives
⎛ ⎞

s
1
s
= ⎝ x mk k ⎠ .
1 − xk
k=1 k=1 mk ≥0
How do we get a coefficient of x n ? We must choose a term x mk k from each factor

of the product, 1 ≤ k ≤ s, so that

s
n= mk k.
k=1
But such a choice is the same as choosing the partition λ of n such that the part k
occurs mk times. For instance, if s = 4 and we choose m1 = 5, m2 = 0, m3 = 1,
m4 = 2, then we have chosen the partition λ = (4, 4, 3, 1, 1, 1, 1, 1) of 16. Hence
the coefficient of x n is the number of partitions λ of n whose largest part is at most
s, as was to be proved.
Note that Proposition 8.14 is “trivial” in the sense that it can be seen by
inspection. There is an obvious correspondence between (a) the choice of terms
contributing to the coefficient of x n and (b) partitions of n with largest part at most
s. Although the generating function we will obtain for plane partitions is equally
simple, it will be far less obvious why it is correct.
Plane partitions have a certain similarity with standard Young tableaux, so
perhaps it is not surprising that a variant of RSK will be applicable. Instead of
NYT we will be dealing with column-strict plane partitions (CSPP). These are
plane partitions for which the nonzero elements strictly decrease in each column.
An example of a CSPP is given by
7743331
4331
32 . (8.17)
21
1
We say that this CSPP has shape λ = (7, 4, 2, 2, 1), the shape of the Young diagram
which the numbers occupy, and that it has five rows, seven columns, and 16 parts
(so λ 16).
If P = (Pij ) is a CSPP and k ≥ 1, then we define the row insertion P ← k
as follows: let r be the least integer such that P1,r < k. If no such r exists (i.e.,
all elements of the first row of P are greater than or equal to k), then simply place
k at the end of the first row. The insertion process stops, and the resulting CSPP
is P ← k. If, on the other hand, r does exist, then replace P1r by k. The element
k then “bumps” P1r := k into the second row, i.e., insert k into the second row
of P by the insertion rule just described, possibly bumping a new element k into
the third row. Continue until an element is inserted at the end of a row (possibly
as the first element of a new row). The resulting array is P ← k. Note that this
rule is completely analogous to row insertion for NYT: for NYT an element bumps
the leftmost element greater than it, while for CSPP an element bumps the leftmost
element smaller than it.
8.15 Example. Let P be the CSPP of (8.17). Let us row insert 6 into P . The set of
elements which get bumped are shown in bold:
7743331
4331
32 .
21
1
The final 1 that was bumped is inserted at the end of the fifth row. Thus we obtain
7763331
4431
(P ← 6) = 3 3 .
22
11
We are now ready to describe the analogue of RSK needed to count plane
partitions. Instead of beginning with a permutation π ∈ Sn , we begin with an r × s
matrix A = (aij ) of nonnegative integers, called for short an r × s N-matrix. We
convert A into a two-line array

u1 u2 · · · uN
wA = ,
v1 v2 · · · vN
where
• u1 ≥ u2 ≥ · · · ≥ uN
• If i < j and ui = uj , then vi ≥ vj .
i
• The number of columns of wA equal to j is aij . (It follows that N = aij .)
It is easy to see that wA is uniquely determined by A, and conversely. As an example,
suppose that
⎡ ⎤
0102
A = ⎣1 1 1 0⎦. (8.18)
2100
Then

333222111
wA = .
211321442
We now insert the numbers v1 , v2 , . . . , vN successively into a CSPP. That is, we

start with P0 = ∅ and define inductively Pi = Pi−1 ← vi . We also start with
Q0 = ∅, and at the ith step insert ui into Qi−1 (without any bumping or other
altering of the elements of Qi−1 ) so that Pi and Qi have the same shape. Finally let
RSK
(P , Q) = (PN , QN ) and write A −→ (P , Q).
8.16 Example. Let A be given by (8.18). The pairs (P1 , Q1 ), . . . , (P9 , Q9 ) =
(P , Q) are as follows:
Pi Qi
2 3
21 33
211 333
311 333
2 2
321 333
21 22
3211 3332
21 22
4211 3332
31 22
2 1
4411 3332
32 22
21 11
4421 3332
321 221
21 11
RSK
It is straightforward to show that if A −→ (P , Q), then P and Q are CSPP of
the same shape. We omit the proof of the following key lemma, which is analogous
to the proof of Theorem 8.13. Let us just note a crucial property (which is easy to
RSK
prove) of the correspondence A −→ (P , Q) which allows us to recover A from
(P , Q), namely, equal entries of Q are inserted from left to right. Thus the last
number placed into Q is the rightmost occurrence of the least entry. Hence we can
inverse bump the number in this position in P to back up one step in the algorithm,
RSK
just as for the usual RSK correspondence π −→ (P , Q).
RSK
8.17 Lemma. The correspondence A −→ (P , Q) is a bijection from the set of
r × s matrices of nonnegative integers to the set of pairs (P , Q) of CSPP of the
same shape, such that the largest part of P is at most s and the largest part of Q is
at most r.
The next step is to convert the pair (P , Q) of CSPP of the same shape into a
single plane partition π . We do this by “merging” the ith column of P with the ith
column of Q, producing the ith column of π . Thus we first describe how to merge
two partitions λ and μ with distinct parts and with the same number of parts into
a single partition ρ = ρ(λ, μ). Draw the Ferrers diagram of λ but with each row
indented one space to the right of the beginning of the previous row. Such a diagram
is called the shifted Ferrers diagram of λ. For instance, if λ = (5, 3, 2) then we get
the shifted diagram
Do the same for μ, and then transpose the diagram. For instance, if μ = (6, 3, 1)
then we get the transposed shifted diagram
Now merge the two diagrams into a single diagram by identifying their main
diagonals. For λ and μ as above, we get the diagram (with the main diagonal drawn
for clarity):
Define ρ(λ, μ) to be the partition for which this merged diagram is the Ferrers
diagram. The above example shows that
ρ(532, 631) = 544211.
The map (λ, μ) → ρ(λ, μ) is clearly a bijection between pairs of partitions (λ, μ)
with k distinct parts and partitions ρ whose main diagonal (of the Ferrers diagram)
has k dots. Equivalently, k is the largest integer j for which ρj ≥ j . Note that
|ρ| = |λ| + |μ| − (λ). (8.19)
We now extend the above bijection to pairs (P , Q) of reverse SSYT of the same
shape. If λi denotes the ith column of P and μi the ith column of Q, then let
π(P , Q) be the array whose ith column is ρ(λi , μi ). For instance, if
4421 5322
P = 3 1 1 and Q = 4 2 1 ,
2 1
then
4421
4221
π(P , Q) = 4 2 .
2
2
It is easy to see that π(P , Q) is a plane partition. Replace each row of π(P , Q) by
its conjugate to obtain another plane partition π (P , Q). With π(P , Q) as above we
obtain
4 322
4 311
π (P , Q) = 2 2 1 1.
1 1
1 1
Write |P | for the sum of the elements of P , and write max(P ) for the largest element
of P , and similarly for Q. When we merge P and Q into π(P , Q), max(P ) becomes
the largest part of π(P , Q). Thus when we conjugate each row, max(P ) becomes the
number col(π (P , Q)) of columns of π (P , Q) [why?]. Similarly, max(Q) becomes
the number row(π (P , Q)) of rows of π(P , Q) and of π (P , Q). In symbols,
max P = col(π (P , Q))

(8.20)
max Q = row(π (P , Q)).
Moreover, it follows from (8.19) that
|π (P , Q)| = |π(P , Q)| = |P | + |Q| − ν(P ), (8.21)
where ν(P ) denotes the number of parts of P (or of Q).

We now have all the ingredients necessary to prove the main result of this
appendix.
8.18 Theorem. Let pprs (n) denote the number of plane partitions of n with at most
r rows and at most s columns. Then

r
s
pprs (n)x n = (1 − x i+j −1 )−1 .
n≥0 i=1 j =1
Proof. Let A = (aij ) be an r ×s N-matrix. We can combine the bijections discussed

above to obtain a plane partition π(A) associated with A. Namely, first apply RSK
RSK
to obtain A −→ (P , Q), and then apply the merging process and row conjugation
to obtain π(A) = π (P , Q). Since a column ji of the two-line array wA occurs aij
times and results in an insertion of j into P and i into Q, it follows that

|P | = j aij
i,j

|Q| = iaij
i,j
max(P ) = max{j : aij = 0}

max(Q) = max{i : aij = 0}
Hence from (8.20) and (8.21), we see that the map A → π(A) is a bijection from
r × s N-matrices A to plane partitions with at most r rows and at most s columns.
Moreover,
|π(A)| = |P | + |Q| − ν(P )

= (i + j − 1)aij .
i,j
Thus the enumeration of plane partitions is reduced to the much easier enumeration
of N-matrices. Specifically, we have

pprs (n)x n = x |π |
n≥0 π
row(π )≤r
col(π )≤s

(i+j −1)aij
= x
r×s N-matrices A
⎛ ⎞

r
s
= ⎝ x (i+j −1)aij ⎠
i=1 j =1 aij ≥0

r
s
= (1 − x i+j −1 )−1 .
i=1 j =1
Write ppr (n) for the number of plane partitions of n with at most r rows. Letting
s → ∞ and then r → ∞ in Theorem 8.18 produces the elegant generating
functions of the next corollary.

ppr (n)x n = (1 − x i )− min(i,r) (8.22)
n≥0 i≥1

pp(n)x n = (1 − x i )−i . (8.23)
n≥0 i≥1
NOTE. Once one has seen the generating function

1
(1 − x)(1 − x 2 )(1 − x 3 ) · · ·
for one-dimensional (ordinary) partitions and the generating function
1
(1 − x)(1 − x 2 )2 (1 − x 3 )3 . . .
for two-dimensional (plane) partitions, it is quite natural to ask about higher-

dimensional partitions. In particular, a solid partition of n is a three-dimensional
array π = (πij k )i,j,k≥1 of nonnegative integers, weakly decreasing in each of the
three coordinate directions, and with elements summing to n. Let sol(n) denote the
number of solid partitions of n. It is easy to see that for any integer sequence a0 = 1,
a1 , a2 , . . . , there are unique integers b1 , b2 , . . . for which

an x n = (1 − x i )−bi .
n≥0 i≥1
For the case an = sol(n), we have
b1 = 1, b2 = 3, b3 = 6, b4 = 10, b5 = 15,
which looks quite promising. Alas, the sequence of exponents continues
20, 26, 34, 46, 68, 97, 120, 112, 23, −186, −496, −735, −531, 779, . . . .
The problem of enumerating solid partitions remains open and is considered most
likely to be hopeless.
Notes for Chapter 8
Standard Young tableaux (SYT) were first enumerated by MacMahon [89, p. 175]
(see also [90, §103]). MacMahon formulated his result in terms of “generalized
ballot sequences” or “lattice permutations” rather than SYT, but they are easily seen
to be equivalent. He stated the result not in terms of the products of hook lengths
as in Theorem 8.1, but as a more complicated product formula. The formulation in
terms of hook lengths is due to Frame and appears first in the paper [45, Thm. 1] of
Frame, Robinson, and Thrall; hence it is sometimes called the “Frame-Robinson-
Thrall hook-length formula.” (The actual definition of standard Young tableaux is
due to Young [146, p. 258].)
Independently of MacMahon, Frobenius [48, eqn. (6)] obtained the same formula
for the degree of the irreducible character χ λ of Sn as MacMahon obtained for the
number of lattice permutations of type λ. Frobenius was apparently unaware of the
combinatorial significance of deg χ λ , but Young showed in [146, pp. 260–261] that
deg χ λ was the number of SYT of shape λ, thereby giving an independent proof of
MacMahon’s result. (Young also provided his own proof of MacMahon’s result in
[146, Thm. II].)
A number of other proofs of the hook-length formula were subsequently found.
Greene et al. [57] gave an elegant probabilistic proof. A proof of Hillman and Grassl
[66] shows very clearly the role of hook lengths, though the proof is not completely
bijective. A bijective version was later given by Krattenthaler [76]. Completely
bijective proofs of the hook-length formula were first given by Franzblau and
Zeilberger [46] and by Remmel [111]. An exceptionally elegant bijective proof was
later found by Novelli et al. [97].
The use of the operators U and D to count walks in the Hasse diagram of Young’s
lattice was developed independently, in a more general context, by Fomin [43, 44]
and Stanley [127, 129]. See also [130, §3.21] for a short exposition.
The RSK algorithm (known by a variety of other names, either “correspondence”
or “algorithm” in connection with some subset of the names Robinson, Schensted,
and Knuth) was first described, in a rather vague form, by Robinson [112, §5], as
a tool in an attempted proof of a result now known as the “Littlewood–Richardson
Rule.” The RSK algorithm was later rediscovered by C.E. Schensted (see below),
but no one actually analyzed Robinson’s work until this was done by van Leeuwen
[143, §7]. It is interesting to note that Robinson says in a footnote on page 754
that “I am indebted for this association I to Mr. D.E. Littlewood.” Van Leeuwen’s
analysis makes it clear that “association I” gives the recording tableau Q of the
RSK
RSK algorithm π −→ (P , Q). Thus it might be correct to say that if π ∈ Sn and
RSK
π −→ (P , Q), then the definition of P is due to Robinson, while the definition of
Q is due to Littlewood.
No further work related to Robinson’s construction was done until Schensted
published his seminal paper [115] in 1961. (For some information about the unusual
life of Schensted, see [5].) Schensted’s purpose was the enumeration of permu-
tations in Sn according to the length of their longest increasing and decreasing
subsequences. According to Knuth [77, p. 726], the connection between the work
of Robinson and that of Schensted was first pointed out by M.-P. Schützenberger,
though as mentioned above the first person to describe this connection precisely was
van Leeuwen.
Plane partitions were discovered by MacMahon in a series of papers which were
not appreciated until much later. (See MacMahon’s book [90, Sections IX and X] for
an exposition of his results.) MacMahon’s first paper dealing with plane partitions
was [88]. In Article 43 of this paper he gives the definition of a plane partition
(though not yet with that name). In Article 51 he conjectures that the generating
function for plane partitions is the product
(1 − x)−1 (1 − x 2 )−2 (1 − x 3 )−3 (1 − x 4 )−4 · · ·
(our (8.23)). In Article 52 he conjectures our (8.22) and Theorem 8.18, finally
culminating in a conjectured generating function for plane partitions of n with at
most r rows, at most s columns, and with largest part at most t. (See Exercise 8.36.)
MacMahon goes on in Articles 56–62 to prove his conjecture in the case of plane
partitions with at most 2 rows and s columns (the case r = 2 of our Theorem 8.18),
mentioning on page 662 that an independent solution was obtained by A.R. Forsyth.
(Though a publication reference is given to Forsyth’s paper, apparently it never
actually appeared.)
We will not attempt to describe MacMahon’s subsequent work on plane parti-
tions, except to say that the culmination of his work appears in [90, Art. 495], in
which he proves his main conjecture from his first paper [88] on plane partitions,
viz., our Exercise 8.36. MacMahon’s proof is quite lengthy and indirect.
In 1972 Bender and Knuth [6] showed the connection between the theory of
symmetric functions and the enumeration of plane partitions. They gave simple
proofs based on the RSK algorithm of many results involving plane partitions,
including the first bijective proof (the same proof that we give) of our Theorem 8.18.
The process of merging two partitions with distinct parts into a single partition,
discussed after Lemma 8.17, was first described by Frobenius [48] for a different
purpose.
For further aspects of Young tableaux and the related topics of symmetric
functions, representation theory of the symmetric group, Grassmann varieties, etc.,
see the expositions of Fulton [49], Sagan [114], and Stanley [131, Ch. 7].
1. Draw all the standard Young tableaux of shape (4, 2).

2. Using the hook-length formula, show
that the number of SYT of shape (n, n)
1 2n
is the Catalan number Cn = n+1 n .
3. How many maximal chains are in the poset L(4, 4), where L(m, n) is defined
in Chapter 6? Express your answer in a form involving products and quotients
of integers (no sums).
4. A corner square of a partition λ is a square in the Young diagram of λ whose
removal results in the Young diagram of another partition (with the same upper-
left corner). Let c(λ) denote the number of corner squares (or distinct parts) of
the partition λ. For instance, c(5, 5, 4, 2, 2, 2, 1, 1) = 4. (The distinct parts are
5, 4, 2, 1.) Show that

c(λ) = p(0) + p(1) + · · · + p(n − 1),
λn
where p(i) denotes the number of partitions of i (with p(0) = 1). Try to give
an elegant combinatorial proof.
5. Show that the number of odd hook lengths minus the number of even hook
lengths of a partition λ is a triangular number (a number of the form k(k+1)/2).
6. (moderately difficult)
n Show
that the total number of SYT with n entries and at
most two rows is n/2 . Equivalently,
n/2

n
f (n−i,i)
= .
n/2
i=0
Try to give an elegant combinatorial proof.

7. (difficult) (*) Let f (n) be the number of partitions λ of 2n whose Young
diagram can be covered with n nonoverlapping dominos (i.e., two squares with
a common edge). For instance, the figure below shows a domino covering of
the partition 43221.
Let

F (x) = f (n)x n = 1 + 2x + 5x 2 + 10x 3 + 20x 4 + 36x 5 + · · · .
n≥0
Show that

F (x) = (1 − x n )−2 .
n≥1
8. (difficult) Let λ be a partition. Let mk (λ) denote the number of parts of λ that
are equal to k, and let ηk (λ) be the number of hooks of length k of λ. Show that

ηk (λ) = k mk (λ).
λn λn
9. (moderately difficult) Let μ be a partition, and let Aμ be the infinite shape

consisting of the quadrant Q = {(i, j ) : i < 0, j > 0} with the shape
μ removed from the lower right-hand corner. Thus every square of Aμ has a
finite hook and hence a hook length. For instance, when μ = (3, 1) we get the
diagram
10 9 8 6 5 3
9 8 7 5 4 2
8 7 6 4 3 1
6 5 4 2 1
3 2 1
Show that the multiset of hook lengths of Aμ is equal to the union of the
multiset of hook lengths of Q (explicitly given by {11 , 22 , 33 , . . . }) and the
multiset of hook lengths of μ.
10. In how many ways can we begin with the empty partition ∅, then add 2n squares
one at a time (always keeping a partition), then remove n squares one at a time,
then add n squares one at a time, and finally remove 2n squares one at a time,
ending up at ∅?
11. (difficult) Fix n. Show that the number of partitions λ n for which f λ is odd
is equal to 2k1 +k2 +... , where k1 < k2 < · · · and n = 2k1 + 2k2 + · · · (the
binary expansion of n). For instance, 75 = 20 + 21 + 23 + 26 , so the number of
partitions λ of 75 for which f λ is odd is 26+3+1+0 = 1024.
12. Let U and D be the linear transformations
associated with Young’s lattice.
Write D 2 U 2 and D 3 U 3 in the form aij U i D j .
13. Let U and D be the linear transformations associated with Young’s lattice.
Suppose that f is some (noncommutative) polynomial in U √ and D satisfying
f (U, D) = 0, e.g., f (U, D) = DU − U D − I . Let i = −1. Show that
f (iD, iU ) = 0.
14. (*) Show that
U n D n = (U D − (n − 1)I )(U D − (n − 2)I ) · · · (U D − I )U D, (8.24)
where U and D are the linear transformations associated with Young’s lattice
(and I is the identity transformation), and where both sides of (8.24) operate on
the vector space RYj (for some fixed j ).
15. (difficult) Give a bijective proof of Corollary 8.8, i.e., β(2m, ∅) = 1 · 3 ·
5 · · · (2m − 1). Your proof should be an analogue of the RSK algorithm. To
start with, note that [why?] 1 · 3 · 5 · · · (2m − 1) is the number of complete
matchings of [2m], i.e., the number of graphs on the vertex set [2m] with m
edges such that every vertex is incident to exactly one edge.
16. Fix a partition λ n − 1. Find a simple formula for the sum t (λ) = μλ f μ
in terms of f λ . The sum ranges over all partitions μ that cover λ (i.e., μ > λ
and nothing is in between, so μ n) in Young’s lattice Y . Give a simple proof
using linear algebra rather than a combinatorial proof.
17. (a) (*) The Bell number B(n) is defined to be the number of partitions of an
1 number of sets {B1 , . . . , Bk } where Bi = ∅, Bi ∩
n-element set S, i.e., the
Bj = ∅ if i = j , and Bi = S. Find a simple formula for the generating
function
xn x2 x3 x4
F (x) = B(n) = 1 + x + 2 + 5 + 15 + · · · .
n! 2! 3! 4!
n≥0
(b) (moderately difficult) Let f (n) be the number of ways to move from the
empty partition ∅ to ∅ in n steps, where each step consists of either (i)
adding a box to the Young diagram, (ii) removing a box, or (iii) adding and
then removing a box, always keeping the diagram of a partition (even in
the middle of a step of type (iii)). For instance, f (3) = 5, corresponding to
the five sequences
∅ (1, ∅) (1, ∅) (1, ∅)

∅ (1, ∅) 1 ∅
∅ 1 (2, 1) ∅ .
∅ 1 (11, 1) ∅
∅ 1 ∅ (1, ∅)
Find (and prove) a formula for f (n) in terms of Bell numbers.

18. (difficult) (*) For n, k ≥ 0 let κ(n → n + k → n) denote the number of closed
walks in Y that start at level n, go up k steps to level n + k, and then go down
k steps to level n. Thus for instance κ(n → n + 1 → n) is the number of cover
relations between levels n and n + 1. Show that

κ(n → n + k → n)q n = k! (1 − q)−k F (Y, q).
n≥0
Here F (Y, q) is the rank-generating function of Y , which by Proposition 8.14

(letting s → ∞) is given by

F (Y, q) = (1 − q i )−1 .
i≥1
19. Let X denote the formal sum of all elements of Young’s lattice Y . The operators
U and D still act in the usual way on X, producing infinite linear combinations
of elements of Y . For instance, the coefficient of the partition (3, 1) in DX is 3,
coming from applying D to (4, 1), (3, 2), and (3, 1, 1).
(a) Show that DX = (U + I )X, where as usual I denotes the identity linear
transformation.
(b) Express the coefficient sn of ∅ (the empty partition) in D n X in terms of the
numbers f λ for λ n. (For instance, s0 = s1 = 1, s2 = 2, s3 = 4.)
(c) Show that
D n+1 X = (U D n + D n + nD n−1 )X, n ≥ 0,
where D −1 = 0, D 0 = I .
(d) Find a simple recurrence relation satisfied by sn .
(e) Find a simple formula for the generating function
xn
F (x) = sn .
n!
n≥0
(f) Show that sn is the number of involutions in Sn , i.e., the number of

elements π ∈ Sn satisfying π 2 = ι.
RSK RSK
(g) (quite difficult) Show that if π ∈ Sn and π −→ (P , Q), then π −1 −→
(Q, P ).
(h) Deduce your answer to (f) from (g).
20. (a) Consider the linear transformation Un−1 Dn : RYn → RYn . Show that its
eigenvalues are the integers i with multiplicity p(n − i) − p(n − i − 1), for
0 ≤ i ≤ n − 2 and i = n.
(b) (*) Use (a) to give another proof of Theorem 8.9.
21. (a) (moderately difficult) Let Y[j −2,j ] denote the Hasse diagram of the restric-
tion of Young’s lattice Y to the levels j − 2, j − 1, j . Let p(n) denote the
number of partitions of n, and write p(n) = p(n) − p(n − 1). Show that
the characteristic polynomial of the adjacency matrix of the graph Y[j −2,j ]
is given by

j
±x p(j ) (x 2 − 1)p(j −1) (x 3 − (2s − 1)x)p(j −1) ,
s=2
where the sign is (−1)#Y[j −2,j ] = (−1)p(j −2)+p(j −1)+p(j ) .

(b) (difficult) Extend to Y[j −i,j ] for any i ≥ 0. Express your answer in terms
of the characteristic polynomial of matrices of the form
⎡ ⎤
0a 0 00
⎢1 0 a + 1 0 0⎥
⎢ ⎥
⎢ . .. 0 0 ⎥
⎢0 1 0 ⎥
⎢ ⎥
⎢ .. ⎥.
⎢ . ⎥
⎢ ⎥
⎢ . ⎥
⎣ .. 0 b⎦
10
22. (moderately difficult)
(a) Let U and D be operators (or just noncommutative variables)
satisfying
DU − U D = I . Show that for any power series f (U ) = an U n whose
coefficients an are real numbers, we have
eDt f (U ) = f (U + t)eDt .
In particular,
eDt eU = et+U eDt . (8.25)
Here t is a variable (indeterminate) commuting with U and D. Regard

both sides as power series in t whose coefficients are (noncommutative)
polynomials in U and D. Thus for instance
⎛ ⎞⎛ ⎞
Dnt n Un
e e =⎝
Dt U ⎠⎝ ⎠
n! n!
m≥0 n≥0
DmU nt m
= .
m! n!
m,n≥0
1 2
(b) Show that e(U +D)t = e 2 t +U t eDt .
(c) Let δn be the total number of walks of length n in Young’s lattice Y (i.e., in
the Hasse diagram of Y ) starting at ∅. For instance, δ2 = 3 corresponding
to the walks (∅, 1, 2), (∅, 1, 11),and (∅, 1, ∅). Find a simple formula for
n
the generating function F (t) = n≥0 δn tn! .
23. Let w be a balanced word in U and D, i.e., the same number of U ’s as
D’s. For instance, U U DU DDDU is balanced. Regard U and D as linear
transformations on RY in the usual way. A balanced word thus takes the
space RYn to itself,
where Yn is the nth level of Young’s lattice Y . Show that
the element En = λn f λ ∈ RYn is an eigenvector for w, and find the
λ
eigenvalue.
24. (*) Prove that any two balanced words (as defined in the previous exercise)
commute.
25. Define a graded poset Z inductively as follows. The bottom level Z0 consists of
a single element. Assume that we have constructed the poset up to level n. First
“reflect” Zn−1 through Zn . More precisely, for each element x ∈ Zn−1 , let x
be a new element of Zn+1 , with x y (where y ∈ Zn ) if and only if y x.
Then for each element y ∈ Zn , let y be a new element of Zn+1 covering y (and
covering no other elements of Zn ). Figure 8.1 shows the poset Z up to level 5.
The cover relations obtained by the reflection construction are shown by solid
lines, while those of the form y y are shown by broken lines.
(a) Show that #Zn = Fn+1 (a Fibonacci number), so the rank-generating
function of Z is given by
1
F (Z, q) = .
1 − q − q2
(b) Define Ui : RZi → RZi+1 and Di : RZi → RZi−1 exactly as we did for
Y , namely, for x ∈ Zi we have

Ui (x) = y
yx

Di (x) = y.
yx
Show that Di+1 Ui −Ui−1 Di = Ii . Thus all the results we have obtained for
Y based on this commutation relation also hold for Z! (For results involving
p(n), we need only replace p(n) by Fn+1 .)
RSK
26. (a) Suppose that π ∈ Sn and π −→ (P , Q). Let f (π ) be the largest integer k
for which 1, 2, . . . , k all appear in the first row of P . Find a simple formula
for the number of permutations π ∈ Sn for which f (π ) = k.
Fig. 8.1 The poset Z up to

level 5
(b) Let E(n) denote the expected value of f (π ) for π ∈ Sn , i.e.,
1
E(n) = f (π ).
n!
π ∈Sn
Find limn→∞ E(n).

RSK
27. Suppose that π ∈ Sn and π −→ (P , Q). Let E12 (n) be the expected
value of the (1, 2)-entry of P (i.e., the second entry in the first row). Find
limn→∞ E12 (n).
28. (a) An increasing subsequence of a permutation a1 a2 · · · an ∈ Sn is a
subsequence ai1 ai2 · · · aij such that ai1 < ai2 < · · · < aij . For instance,
2367 is an increasing subsequence of the permutation 52386417. Suppose
that the permutation w ∈ Sn is sent into an SYT of shape λ = (λ1 , λ2 , . . . )
under the RSK algorithm. Show that λ1 is the length of the longest
increasing subsequence of w.
(b) (much harder) Define decreasing subsequence similarly to increasing
subsequence. Show that λ1 (the number of parts of λ) is equal to the length
of the longest decreasing subsequence of λ.
(c) Assuming (a) and (b), show that for m, n ≥ 1, a permutation w ∈ Smn+1
has an increasing subsequence of length m + 1 or a decreasing subsequence
of length n + 1.
(d) How many permutations w ∈ Smn have longest increasing subsequence of
length m and longest decreasing subsequence of length n? (Use the hook
length formula to obtain a simple explicit answer.)
29. Write down the 13 plane partitions of 4 and the 24 plane partitions of 5.
RSK
30. Prove the statement preceding Lemma 8.17 that in the bijection A −→ (P , Q),
equal elements of Q are inserted from left to right.
31. Let A be the r × s matrix of all 1’s. Describe the plane partition π (A).
32. (a) Find the N-matrix A for which
64433
π (A) = 5 3 3 2 .
321
(b) What message is conveyed by the nonzero entries of A?

33. (*) Let f (n) denote the number of plane partitions π = (πij ) of n for which
π22 = 0. Show that
n+1
n≥0 (−1)n x ( 2 )
f (n)x n = 0 .
i≥1 (1 − x i )2
n≥0
RSK
34. (a) (quite difficult) Let A be an r × s N-matrix, and let A −→ (P , Q). If At
RSK
denotes the transpose of A, then show that At −→ (Q, P ).
NOTE. This result is quite difficult to prove from first principles. If you
can do Exercise 8.19(g), then the present exercise is a straightforward
modification. In fact, it is possible to deduce the present exercise from
Exercise 8.19(g).
(b) A plane partition π = (πij ) is symmetric if πij = πj i for all i and j . Let
sr (n) denote the number of symmetric plane partitions of n with at most r
rows. Assuming (a), show that
r
−1 −1
sr (n)x n = 1 − x 2i−1 · 1 − x 2(i+j −1) .
n≥0 i=1 1≤i<j ≤r
(c) Let s(n) denote the total number of symmetric plane partitions of n. Let
r → ∞ in (b) to deduce that
1
s(n)x n = .
(1 − x 2i−1 )(1 − x 2i )i/2
n≥0 i≥1
(d) (very difficult; cannot be done using RSK) Let srt (n) denote the number of
symmetric plane partitions of n with at most r rows and with largest part at
most t. Show that
t
1 − x (2−δij )(i+j +k−1)
srt (n)x n = .
n≥0 1≤i<j ≤r k=1
1 − x (2−δij )(i+j +k−2)

35. The trace of a plane partition π = (πij ) is defined as tr(π ) = i πii . Let
pp(n, k) denote the number of plane partitions of n with trace k. Show that

pp(n, k)q k x n = (1 − qx i )−i .
n≥0 k≥0 i≥1
36. (very difficult; cannot be done using RSK) Let pprst (n) be the number of plane
partitions of n with at most r rows, at most s columns, and with largest part at
most t. Show that

r
s
t
1 − x i+j +k−1
pprst (n)x n = .
1 − x i+j +k−2
n≥0 i=1 j =1 k=1
37. Let f (n) denote the number of solid partitions

of n with largest part at most 1.
Find the generating function F (x) = n≥0 f (n)x n .
Chapter 9
The Matrix-Tree Theorem
The Matrix-Tree Theorem is a formula for the number of spanning trees of a graph
in terms of the determinant of a certain matrix. We begin with the necessary graph-
theoretical background. Let G be a finite graph, allowing multiple edges but not
loops. (Loops could be allowed, but they turn out to be completely irrelevant.) Recall
that G is connected if there exists a walk between any two vertices of G. A cycle
is a closed walk with no repeated vertices or edges, except for the first and last
vertex. A tree is a connected graph with no cycles. In particular, a tree cannot have
multiple edges, since a double edge is equivalent to a cycle of length two. The three
nonisomorphic trees with five vertices are shown in Figure 9.1.
A basic theorem of graph theory (whose easy proof we leave as an exercise) is
the following.
9.1 Proposition. Let G be a graph with p vertices. The following conditions are
equivalent:
(a) G is a tree.
(b) G is connected and has p − 1 edges.
(c) G has no cycles and has p − 1 edges.
(d) There is a unique path (= walk with no repeated vertices) between any two
vertices.
A spanning subgraph of a graph G is a graph H with the same vertex set as G,
and such that every edge of H is an edge of G. If G has q edges, then the number
of spanning subgraphs of G is equal to 2q , since we can choose any subset of the
edges of G to be the set of edges of H . (Note that multiple edges between the same
two vertices are regarded as distinguishable, in accordance with the definition of a
graph in Chapter 1.) A spanning subgraph which is a tree is called a spanning tree.
Clearly G has a spanning tree if and only if it is connected [why?]. An important
invariant of a graph G is its number of spanning trees, called the complexity of G
and denoted κ(G).

https://doi.org/10.1007/978-3-319-77173-1_9
136 9 The Matrix-Tree Theorem
Fig. 9.1 The three trees with five vertices
9.2 Example. Let G be the graph illustrated below, with edges a, b, c, d, and e.
b
a e c
Then G has eight spanning trees, namely, abc, abd, acd, bcd, abe, ace, bde, and
cde (where, e.g., abc denotes the spanning subgraph with edge set {a, b, c}).
9.3 Example. Let G = K5 , the complete graph on five vertices. A simple counting
argument shows that K5 has 60 spanning trees isomorphic to the first tree in
Figure 9.1, 60 isomorphic to the second tree, and 5 isomorphic to the third tree.
Hence κ(K5 ) = 125. It is even easier to verify that κ(K1 ) = 1, κ(K2 ) = 1,
κ(K3 ) = 3, and κ(K4 ) = 16. Can the reader make a conjecture about the value
of κ(Kp ) for any p ≥ 1?
Our object is to obtain a “determinantal formula” for κ(G). For this we need
an important result from matrix theory, known as the Binet–Cauchy theorem or
Cauchy–Binet theorem and which is often omitted from a beginning linear algebra
course. Later (Theorem 10.4) we will prove a more general determinantal formula
without the use of the Binet–Cauchy theorem. However, the use of the Binet–
Cauchy theorem does afford some additional algebraic insight. The Binet–Cauchy
theorem is a generalization of the familiar fact that if A and B are n × n matrices,
then det AB = (det A)(det B), where det denotes determinant. We want to extend
this formula to the case where A and B are rectangular matrices whose product is a
square matrix (so that det AB is defined). In other words, A will be an m × n matrix
and B an n × m matrix, for some m, n ≥ 1.
We will use the following notation involving submatrices. Suppose A = (aij ) is
an m × n matrix, with 1 ≤ i ≤ m, 1 ≤ j ≤ n, and m ≤ n. Given an m-element
subset S of {1, 2, . . . , n}, let A[S] denote the m × m submatrix of A obtained by
taking the columns indexed by the elements of S. In other words, if the elements of
S are given by j1 < j2 < · · · < jm , then A[S] = (ai,jk ), where 1 ≤ i ≤ m and
1 ≤ k ≤ m. For instance, if
⎡ ⎤
1 2 3 4 5
A = ⎣ 6 7 8 9 10 ⎦
11 12 13 14 15
9 The Matrix-Tree Theorem 137
and S = {2, 3, 5}, then

⎡ ⎤
2 3 5
A[S] = ⎣ 7 8 10 ⎦ .
12 13 15
Similarly, let B = (bij ) be an n × m matrix with 1 ≤ i ≤ n, 1 ≤ j ≤ m,
and m ≤ n. Let S be an m-element subset of {1, 2, . . . , n} as above. Then B[S]
denotes the m × m matrix obtained by taking the rows of B indexed by S. Note that
At [S] = A[S]t , where t denotes transpose.
9.4 Theorem (the Binet–Cauchy Theorem). Let A = (aij ) be an m×n matrix, with
1 ≤ i ≤ m and 1 ≤ j ≤ n. Let B = (bij ) be an n × m matrix with 1 ≤ i ≤ n and
1 ≤ j ≤ m. (Thus AB is an m × m matrix.) If m > n, then det(AB) = 0. If m ≤ n,
then

det(AB) = (det A[S])(det B[S]),
S
where S ranges over all m-element subsets of {1, 2, . . . , n}.

Before proceeding to the proof, let us give an example. We write |aij | for the
determinant of the matrix (aij ). Suppose
⎡ ⎤
c1 d1
a a a
A= 1 2 3 , B = ⎣ c2 d2 ⎦ .
b1 b2 b3
c3 d3
Then

a a c d a a c d a a c d
det AB = 1 2 · 1 1 + 1 3 · 1 1 + 2 3 · 2 2 .
b1 b2 c2 d2 b1 b3 c3 d3 b2 b3 c3 d3
sketch. First suppose m > n. Since from linear algebra we know that rank AB ≤
rank A and that the rank of an m × n matrix cannot exceed n (or m), we have that
rank AB ≤ n < m. But AB is an m × m matrix, so det AB = 0, as claimed.
Now assume m ≤ n. We use notation such as Mrs to denote an r × s matrix M.
It is an immediate consequence of the definition of matrix multiplication (which the
reader should check) that

Rmm Smn Vmn Wmm RV + SX RW + SY
= . (9.1)
Tnm Unn Xnn Ynm T V + UX T W + UY
In other words, we can multiply “block” matrices of suitable dimensions as if their

entries were numbers. Note that the entries of the right-hand side of (9.1) all have
well-defined dimensions (sizes), e.g., RV + SX is an m × n matrix since both RV
and SX are m × n matrices.
Now in (9.1) let R = Im (the m × m identity matrix), S = A, T = Onm (the

n × m matrix of 0’s), U = In , V = A, W = Omm , X = −In , and Y = B. We get

Im A A Omm Omn AB
= . (9.2)
Onm In −In B −In B
Take the determinant of both sides of (9.2). The first matrix on the left-hand side is
upper triangular with 1’s on the main diagonal. Hence its determinant is one. Since
the determinant of a product of square matrices is the product of the determinants
of the factors, we get

A Omm Omn AB
= (9.3)
−In B −In B .
It is easy to see [why?] that the determinant on the right-hand side of (9.3) is
equal to ± det AB. So consider the left-hand side. A nonzero term in the expansion
of the determinant on the left-hand side is obtained by taking the product (with a
certain sign) of m + n nonzero entries, no two in the same row and column (so
one in each row and each column). In particular, we must choose m entries from
the last m columns. These entries belong to m of the bottom n rows [why?], say
rows m + s1 , m + s2 , . . . , m + sm . Let S = {s1 , s2 , . . . , sm } ⊆ {1, 2, . . . , n}. We
must choose n − m further entries from the last n rows, and we have no choice but
to choose the −1’s in those rows m + i for which i ∈ S. Thus every term in the
expansion of the left-hand side of (9.3) uses exactly n − m of the −1’s in the bottom
left block −In .
What is the contribution to the expansion of the left-hand side of (9.3) from
those terms which use exactly the −1’s from rows m + i where i ∈ S? We obtain
this contribution by deleting all rows and columns to which these −1’s belong (in
other words, delete row m + i and column i whenever i ∈ {1, 2, . . . , n} − S), taking
the determinant of the 2m × 2m matrix MS that remains, and multiplying by an
appropriate sign [why?]. But the matrix MS is in block-diagonal form, with the
first block just the matrix A[S] and the second block just B[S]. Hence det MS =
(det A[S])(det B[S]) [why?]. Taking all possible subsets S gives

det AB = ±(det A[S])(det B[S]).
S⊆{1,2,...,n}
|S|=m
It is straightforward but somewhat tedious to verify that all the signs are +; we omit
the details. This completes the proof.
In Chapter 1 we defined the adjacency matrix A(G) of a graph G with vertex

set V = {v1 , . . . , vp } and edge set E = {e1 , . . . , eq }. We now define two related
matrices. Continue to assume that G has no loops. (This assumption is harmless
since loops have no effect on κ(G).)
Fig. 9.2 A graph G with an e

orientation o
1 4
f
a b d
2 c 3
9.5 Definition. Let G be as above. Give G an orientation o, i.e., for every edge
e with vertices u, v, choose one of the ordered pairs (u, v) or (v, u). If we choose
(u, v), say, then we think of putting an arrow on e pointing from u to v; and we say
that e is directed from u to v, that u is the initial vertex and v the final vertex of
e, etc.
(a) The incidence matrix M(G) of G (with respect to the orientation o) is the p × q
matrix whose (i, j )-entry M ij is given by
⎧
⎨ −1, if the edge ej has initial vertex vi
M ij = 1, if the edge ej has final vertex vi
⎩
0, otherwise.
(b) The Laplacian matrix L(G) of G is the p × p matrix whose (i, j )-entry Lij is
given by
$
−mij , if i = j and there are mij edges between vi and vj
Lij =
deg(vi ), if i = j,
where deg(vi ) is the number of edges incident to vi . Note that L(G) is

symmetric and does not depend on the orientation o.
As an example, let (G, o) be the oriented graph of Figure 9.2. Then
⎡ ⎤
1 −1 0 −1 −1 −1
⎢ −1 1 −1 0 0 0 ⎥
M(G) = ⎢
⎣ 0 0 1 1 0 0⎦
⎥
0 0 0 0 1 1
⎡ ⎤
4 −2 −1 −2
⎢ −2 3 −1 0 ⎥
L(G) = ⎢ ⎥
⎣ −1 −1 2 0 ⎦ .
−2 0 0 2
For any graph G, every column of M(G) contains one 1, one −1, and q − 2
0’s; and hence the sum of the entries in each column is 0. Thus all the rows sum to
the 0 vector, a linear dependence relation which shows that rank(M(G)) < p. Two
further properties of M(G) and L(G) are given by the following lemma.
9.6 Lemma. (a) We have MM t = L.
(b) If G is regular of degree d, then L(G) = dI − A(G), where A(G) denotes the
adjacency matrix of G. Hence if G (or A(G)) has eigenvalues λ1 , . . . , λp , then
L(G) has eigenvalues d − λ1 , . . . , d − λp .
Proof. (a) This is immediate from the definition of matrix multiplication. Specifi-
cally, for vi , vj ∈ V (G) we have

(MM t )ij = M ik M j k .
ek ∈E(G)
If i = j , then in order for M ik M j k = 0, we must have that the edge ek connects

the vertices vi and vj . If this is the case, then one of M ik and M j k will be 1 and
the other −1 [why?], so their product is always −1. Hence (MM t )ij = −mij ,
as claimed.
There remains the case i = j . Then M ik M ik will be 1 if ek is an edge
with vi as one of its vertices and will be 0 otherwise [why?]. So now we get
(MM t )ii = deg(vi ), as claimed. This proves (a).
(b) Clear by (a), since the diagonal elements of MM t are all equal to d.
Now assume that G is connected, and let M 0 (G) be M(G) with its last row
removed. Thus M 0 (G) has p − 1 rows and q columns. Note that the number of
rows is equal to the number of edges in a spanning tree of G. We call M 0 (G) the
reduced incidence matrix of G. The next result tells us the determinants (up to sign)
of all (p − 1) × (p − 1) submatrices N of M 0 . Such submatrices are obtained by
choosing a set X = {ei1 , . . . , eip−1 } of p − 1 edges of G, and taking all columns of
M 0 indexed by the set S = {i1 , . . . , ip−1 }. Thus this submatrix is just M 0 [S]. For
convenience we will not bother to distinguish between the set S of indices with the
corresponding set X of edges.
9.7 Lemma. Let S be a set of p − 1 edges of G. If S does not form the set of edges
of a spanning tree, then det M 0 [S] = 0. If, on the other hand, S is the set of edges
of a spanning tree of G, then det M 0 [S] = ±1.
Proof. If S is not the set of edges of a spanning tree, then some subset R of S
forms the edges of a cycle C in G. Suppose that the cycle C defined by R has edges
f1 , . . . , fs in that order. Multiply the column of M 0 [S] indexed by fi by 1 if in
going around C we traverse fi in the direction of its arrow; otherwise multiply the
column by −1. Then add these modified columns. It is easy to see (check a few
small examples to convince yourself) that we get the 0 column. Hence the columns
of M 0 [S] are linearly dependent, so det M 0 [S] = 0, as claimed.
Now suppose that S is the set of edges of a spanning tree T . Let e be an edge
of T which is connected to vp (the vertex which indexed the bottom row of M,
i.e., the row removed to get M 0 ). The column of M 0 [S] indexed by e contains
exactly one nonzero entry [why?], which is ±1. Remove from M 0 [S] the row and
column containing the nonzero entry of column e, obtaining a (p − 2) × (p − 2)
matrix M 0 . Note that det M 0 [S] = ± det M 0 [why?]. Let T be the tree obtained
from T by contracting the edge e to a single vertex (so that vp and the remaining
vertex of e are merged into a single vertex u). Then M 0 is just the matrix obtained
from the incidence matrix M(T ) by removing the row indexed by u [why?]. Hence
by induction on the number p of vertices (the case p = 1 being trivial), we have
det M 0 = ±1. Thus det M 0 [S] = ±1, and the proof follows.
NOTE. An alternative way of seeing that det M 0 [S] = ±1 when S is the set of
edges of a spanning tree T is as follows. Let u1 , u2 , . . . , up−1 be an ordering of
the vertices v1 , . . . , vp−1 such that ui is an endpoint of the tree obtained from T by
removing vertices u1 , . . . , ui−1 . (It is easy to see that such an ordering is possible.)
Permute the rows of M 0 [S] so that the ith row is indexed by ui . Then permute the
columns in the order e1 , . . . , ep−1 so that ei is the unique edge adjacent to ui after
u1 , . . . , ui−1 have been removed. Then we obtain a lower triangular matrix with
±1’s on the main diagonal, so the determinant is ±1.
We have now assembled all the ingredients for the main result of this chapter.
Recall that κ(G) denotes the number of spanning trees of G.
9.8 Theorem (the Matrix-Tree Theorem). Let G be a finite connected graph without
loops, with laplacian matrix L = L(G). Let L0 denote L with the last row and
column removed (or with the ith row and column removed for any i). Then
det L0 = κ(G).
Proof. Since L = MM t (Lemma 9.6(a)), it follows immediately that L0 = M 0 M t0 .

Hence by the Binet–Cauchy theorem (Theorem 9.4), we have

det L0 = (det M 0 [S])(det M t0 [S]), (9.4)
S
where S ranges over all (p − 1)-element subsets of {1, 2, . . . , q} (or equivalently,

over all (p − 1)-element subsets of the set of edges of G). Since in general At [S] =
A[S]t , (9.4) becomes

det L0 = (det M 0 [S])2 . (9.5)
S
According to Lemma 9.7, det M 0 [S] is ±1 if S forms the set of edges of a spanning
tree of G, and is 0 otherwise. Therefore the term indexed by S in the sum on the
right-hand side of (9.5) is 1 if S forms the set of edges of a spanning tree of G, and
is 0 otherwise. Hence the sum is equal to κ(G), as desired.
The operation of removing a row and column from L(G) may seem somewhat
contrived. We would prefer a description of κ(G) directly in terms of L(G). Such a
description will follow from the next lemma.
9.9 Lemma. Let M be a p × p matrix (with entries in a field) such that the sum
of the entries in every row and column is 0. Let M0 be the matrix obtained from
M by removing the last row and last column (or more generally, any row and any
column). Then the coefficient of x in the characteristic polynomial det(M − xI ) of
M is equal to −p · det(M0 ). (Moreover, the constant term of det(M − xI ) is 0.)
Proof. The constant term of det(M − xI ) is det M, which is 0 since the rows of M
sum to 0.
For simplicity we prove the rest of the lemma only for removing the last row
and column, though the proof works just as well for any row and column. Add all
the rows of M − xI except the last row to the last row. This doesn’t affect the
determinant and will change the entries of the last row all to −x (since the rows of
M sum to 0). Factor out −x from the last row, yielding a matrix N (x) satisfying
det(M − xI ) = −x det N(x). Hence the coefficient of x in det(M − xI ) is given
by − det N (0). Now add all the columns of N(0) except the last column to the last
column. This does not affect det N(0). Because the columns of M sum to 0, the
last column of N (0) becomes the column vector [0, 0, . . . , 0, p]t . Expanding the
determinant by the last column shows that det N(0) = p · det M0 , and the proof
follows.
9.10 Corollary. (a) Let G be a connected (loopless) graph with p vertices. Sup-
pose that the eigenvalues of L(G) are μ1 , . . . , μp−1 , μp , with μp = 0. Then
1
κ(G) = μ1 μ2 · · · μp−1 .
p
(b) Suppose that G is also regular of degree d, and that the eigenvalues of A(G)
are λ1 , . . . , λp−1 , λp , with λp = d. Then
1
κ(G) = (d − λ1 )(d − λ2 ) · · · (d − λp−1 ).
p
Proof. (a) We have
det(L − xI ) = (μ1 − x) · · · (μp−1 − x)(μp − x)

= −(μ1 − x)(μ2 − x) · · · (μp−1 − x)x.
Hence the coefficient of x is −μ1 μ2 · · · μp−1 . By Lemma 9.9, we get

−μ1 μ2 · · · μp−1 = −p · det(L0 ). By Theorem 9.8 we have det(L0 ) = κ(G),
(b) Immediate from (a) and Lemma 9.6(b).
Let us look at a couple of examples of the use of the Matrix-Tree Theorem.
Appendix: Three Elegant Combinatorial Proofs 143
9.11 Example. Let G = Kp , the complete graph on p vertices. Now Kp is regular

of degree d = p − 1, and by Proposition 1.5 its eigenvalues are −1 (p − 1 times)
and p − 1 = d. Hence from Corollary 9.10 there follows
1
κ(Kp ) = ((p − 1) − (−1))p−1 = pp−2 .
p
Naturally a combinatorial proof of such an elegant result is desirable. In the

Appendix to this chapter we give three such proofs.
9.12 Example. Let G = Cn , the n-cube discussed in Chapter 2. Now Cn is regular

of degree n, and by Corollary 2.4 its eigenvalues are n − 2i with multiplicity ni for
0 ≤ i ≤ n. Hence from Corollary 9.10 there follows the amazing result
1
n
n
κ(Cn ) = n
(2i)( i )
2
i=1
n −n−1
n
n
= 22 i(i ).
i=1
A direct combinatorial proof (though not an explicit bijection) was found by O.

Bernardi in 2012.
Appendix: Three Elegant Combinatorial Proofs
In this appendix we give three elegant combinatorial proofs that the number of
spanning trees of the complete graph Kp is pp−2 (Example 9.11). The proofs are
given in chronological order of their discovery.
First Proof (Prüfer). Given a spanning tree T of Kp , i.e., a tree on the vertex set
[p], remove the largest endpoint (leaf) v and write down the vertex a1 adjacent
to v. Continue this procedure until only two vertices remain, obtaining a sequence
(a1 , . . . , ap−2 ) ∈ [p]p−2 , called the Prüfer sequence of T . For the tree below, we
first remove 11 and then record 8. Next remove 10 and record 1. Then remove 8 and
record 4, etc., ending with the sequence (8, 1, 4, 4, 1, 4, 9, 1, 9) and leaving the two
vertices 1 and 9.
8 6
11
5 10
4 9 1
7 2 3
We claim that the map just defined from trees T on [p] to sequences
(a1 , . . . , ap−2 ) ∈ [p]p−2 is a bijection, thereby completing the proof since clearly
[p]p−2 has pp−2 elements. The crucial observation is that the first vertex to be
removed from T is the largest vertex of T missing from the sequence [why?—this
takes a little thought]. This vertex is adjacent to a1 . For our example, we get that
11 was the first vertex removed, and that 11 is adjacent to 8. We can now proceed
recursively. If T1 denotes T with the largest missing vertex removed, then the Prüfer
sequence of T1 is (a2 , . . . , ap−2 ). The first vertex to be removed from T1 is the
largest vertex of T1 missing from (a2 , . . . , ap−2 ). This missing vertex is adjacent to
a2 . For our example, this missing vertex is 10 (since 11 is not a vertex of T1 ), which
is adjacent to 1. Continuing in this way, we determine one new edge of T at each
step. At the end we have found p − 2 edges, and the remaining two unremoved
vertices form the (p − 1)st edge.
Second Proof (Joyal). A doubly rooted tree is a tree T with one vertex u labelled S
(for “start”) and one vertex v (which may equal u) labelled E (“end”). Let t (p) be
the number of trees T on the vertex set [p], and let d(p) be the number of doubly
rooted trees on [p]. Thus
d(p) = p2 t (p), (9.6)
since once we have chosen T there are p choices for u and p choices for v.
Let T be a doubly-rooted tree. There is a unique path from S to E, say with
vertices S = b1 , b2 , . . . , bk = E (in that order). The following diagram shows such
a doubly-rooted tree.
11 10 15 7 5 2 3
S E
1 16
6 9 4 12 17
14 13 8
Let a1 < a2 < · · · < ak be the increasing rearrangement of the numbers

b1 , b2 , . . . , bk . Let π be the permutation of the set {a1 , . . . , ak } given by π(ai ) = bi .
Let Dπ be the digraph of π , that is, the vertex set of Dπ is {a1 , . . . , ak }, with a
directed edge ai → bi for 1 ≤ i ≤ k. Since any permutation π of a finite set
is a disjoint product of cycles, it follows that Dπ is a disjoint union of directed
cycles (all edges of each cycle point in the same direction as we traverse the cycle).
For the example above, we have k = 7, (b1 , . . . , b7 ) = (11, 10, 15, 7, 5, 2, 3) and
(a1 , . . . , a7 ) = (2, 3, 5, 7, 10, 11, 15). The digraph Dπ is shown below.
Appendix: Three Elegant Combinatorial Proofs 145
2 3 10
7
11 15 5
Now attach to each vertex v of Dπ the same subgraph Tv that was attached
“below” v in T and direct the edges of Tv toward v, obtaining a digraph DT . For
our example we get
4 12 17
3 10
2
7
11 15 5
1 9 16
6
14 13 8
The graph DT has the crucial property that every vertex has outdegree one, that is,
one arrow pointing out. In other words, DT is the graph of a function f : [p] → [p],
with vertex set [p] and edges i → f (i). Conversely, given a function f : [p] → [p],
all the above steps can be reversed to obtain a unique doubly rooted tree T for which
DT is the graph of f . We have therefore found a bijection from doubly rooted trees
on [p] to functions f : [p] → [p]. Since the number of such functions f is pp , it
follows that d(p) = pp . Then from (9.6) we get t (p) = pp−2 .
Third Proof (Pitman). A forest is a graph without cycles; thus every connected
component is a tree. A planted forest is a forest F for which every component T
has a distinguished vertex rT (called the root of T ). Thus if a component T has k
vertices, then there are k ways to choose the root of T .
Let Pp be the set of all planted forests on [p]. Let uv be an edge of a forest
F ∈ Pp such that u is closer than v to the root r of its component. Define F to cover
the planted forest F if F is obtained by removing the edge uv from F , and rooting
the new tree containing v at v. This definition of cover defines the covering relation
of a partial order on Pp . Under this partial order Pp is graded of rank p − 1. The
rank of a forest F in Pp is its number of edges. The following diagram shows the
poset P3 , with the root of each tree being its top vertex.
1 1 2 2 3 3
2 1 3 1 3 1 3 2
3 2 3 2 3 1 3 1 2 1 2 1
1 2 2 3
1 3
3 3 2 3 1 2 1
2 1 3 1 2
123
It is an easy exercise to see that an element F of Pp of rank i covers i elements

and is covered by (p − i − 1)p elements. We now count in two ways the number
Mp of maximal chains of Pp . On the one hand, we can start at the top. The number
of maximal elements of Pp is p · t (p), where t (p) as above is the number of trees
on the vertex set [p], since there are p ways to choose the root of such a tree. Once
a maximal element F is chosen, then there are p − 1 elements F that it covers, then
p − 2 elements that F covers, etc., giving
Mp = p · t (p)(p − 1)! = p! t (p). (9.7)
On the other hand, we can start at the bottom. There is a unique element F of rank
zero (the planted forest with no edges), then (p − 1)p elements F that cover F ,
then (p − 2)p elements that cover F , etc., giving
Mp = pp−1 (p − 1)!. (9.8)
Comparing (9.7) and (9.8) gives t (p) = pp−2 .

Our third proof isn’t an explicit bijection like the first two proofs. On the other
hand, it has the virtue of not depending on the names of the vertices. Note that in
the first two proofs it is necessary to know when one vertex is larger than another.
Notes for Chapter 9
The concept of tree as a formal mathematical object goes back to G. Kirchhoff

and K.G.C. von Staudt. Trees were first extensively investigated by A. Cayley,
to whom the term “tree” is due. In particular, in [24] Cayley states the formula
κ(Kp ) = pp−2 for the number of spanning trees of Kp , and he gives a vague
idea of a combinatorial proof. Because of this paper, Cayley is often credited with
the enumeration of labelled trees. Cayley pointed out, however, that an equivalent
result had been proved earlier by Borchardt [11]. Moreover, this result appeared
even earlier in a paper of Sylvester [134]. Undoubtedly Cayley and Sylvester could
have furnished a complete, rigorous proof had they had the inclination to do so. The
elegant combinatorial proofs given in the appendix are due to Prüfer [106], Joyal
[71, Exam. 12, pp. 15–16], and Pitman [100].
The Matrix-Tree Theorem (Theorem 9.8) was first proved by Borchardt [11]
in 1860, though a similar result had earlier been published by Sylvester [134] in
1857. Cayley [23, p. 279] in fact in 1856 referred to the not-yet-published work of
Sylvester. For further historical information on the Matrix-Tree theorem, see Moon
[94, p. 42]. See Bernardi [8] for the combinatorial proof mentioned in Example 9.12.
1. (*) Let Gp be the complete graph Kp with one edge removed. How many
spanning trees does Gp have?
2. Let L = L(Krs ) be the laplacian matrix of the complete bipartite graph Krs .
(a) Find a simple upper bound on rank(L − rI ). Deduce a lower bound on the
number of eigenvalues of L equal to r.
(b) Assume r = s, and do the same as (a) for s instead of r.
(c) (*) Find the remaining eigenvalues of L.
(d) Use (a)–(c) to compute κ(Krs ), the number of spanning trees of Krs .
(e) Give a combinatorial proof of the formula for κ(Krs ), by modifying either
the proof of Prüfer or Joyal that κ(Kp ) = pp−2 .
3. (a) (*) Let 1 ≤ m ≤ n. Let Kn − Km denote the graph Kn with all the edges
of some subgraph Km removed. (In particular, Kn − K1 = Kn .) Thus
Kn − Km has n2 − m2 edges. Use directly the Matrix-Tree Theorem to
find the number of spanning trees of Kn − Km .
(b) Give another proof using Exercise 6.
4. Let p ≥ 5, and let Gp be the graph on the vertex set Zp with edges {i, i + 1}
and {i, i + 2}, for i ∈ Zp . Thus Gp has 2p edges. Show that κ(Gp ) = pFp2 ,
where Fp is a Fibonacci number (F1 = F2 = 1, Fp = Fp−1 +Fp−2 for p ≥ 3).
5. Let C n be the edge complement of the cube graph Cn , i.e., C n has vertex set
{0, 1}n , with an edge uv if u and v differ in at least two coordinates. Find a
formula for κ(C n ), the number of spanning trees of C n . Your answer should be
expressed as a simple product.
6. Let G be a finite graph on p vertices with laplacian matrix L(G). Let G be
obtained from G by adding a new vertex v and connecting it to each vertex of
G (so we have p new edges). Express κ(G ) (the number of spanning trees of
G ) in terms of the eigenvalues μ1 , . . . , μp of L(G).
7. (a) Let G be a bipartite graph with vertex bipartition (A, B). Suppose that
deg v = a for all v ∈ A, and deg v = b for all v ∈ B. Let A and L
denote the adjacency matrix and laplacian matrix of G, respectively. Show
that if the eigenvalues of L are λ1 , . . . , λp , then the eigenvalues of A2 are
(λ1 − a)(λ1 − b), . . . , (λp − a)(λp − b).
(b) (*) Find the number of spanning trees of the graph Cn,k of Exercise 2.2.
8. (a) (*) Let G be a finite loopless graph with p vertices. Suppose that the
eigenvalues of the Laplacian matrix L(G) are θ1 , . . . , θp−1 and θp = 0. Let
J be the p × p matrix of all 1’s, and let α ∈ R. Show that the eigenvalues
of L + αJ are θ1 , . . . , θp−1 , αp.
(b) Let G∪Kp be the graph obtained from G by adding one new edge between
every pair of distinct vertices. Express the number of spanning trees of
G ∪ Kp in terms of θ1 , . . . , θp−1 .
(c) Suppose that G is simple, and let G be the complementary graph, i.e., G
and G have the same vertex set, and two distinct vertices are adjacent in G
if and only if they are not adjacent in G. Express the number of spanning
trees of G in terms of θ1 , . . . , θp−1 .
(d) (*) Let G be simple with p vertices, and define the polynomial

P (G, x) = x c(F )−1 ,
F
where F ranges over all spanning planted forests of G, and where c(F ) is
the number of components of F . Show that
P (G, x) = (−1)p−1 P (G, −x − p).
9. (*) Let V be the subset of Z × Z on or inside some simple closed polygonal

curve whose vertices belong to Z × Z, such that every line segment that makes
up the curve is parallel to either the x-axis or y-axis. Draw an edge e between
any two points of V at distance one apart, provided e lies on or inside the
boundary curve. We obtain a planar graph G, an example being
Let G be the dual graph G∗ with the “outside” vertex deleted. (The vertices
of G are the interior regions of G. For each edge e of G, say with regions R
and R on the two sides of e, there is an edge of G between R and R . See
Section 11.4 for more information on duals to planar graphs.) For the above
example, G is given by
Let λ1 , . . . , λp denote the eigenvalues of G (i.e., of the adjacency matrix

A(G )). Show that

p
κ(G) = (4 − λi ).
i=1
10. (a) (*) Let L be the laplacian matrix of a graph G on a p-element vertex set V .
For a k-element subset S of the vertices, let L[S, S] be the k × k submatrix
of L whose rows and columns are indexed by the vertices in S. Show that
det L[S, S] is equal to the number of planted spanning forests (as defined
in the third proof of the Appendix to this chapter) of G whose set of roots
is V − S.
(b) Deduce from (a) that the coefficient of x j in the characteristic polynomial
det(L − xI ) is equal to (−1)j times the number fj (G) of planted spanning
forests of G with j components.
(c) Deduce from Theorem 5.12 that the sequence f1 (G), f2 (G), . . . , fp (G) is
strongly log-concave.
(d) (extremely difficult) Let gk (G) denote the number of spanning forests (but
not planted) of G with j components. Show that the sequence g1 (G),
g2 (G), . . . , gp (G) is log-concave. (The graph K3 shows that this sequence

need not be strongly log-concave.)
11. Let G be a vertex-transitive graph with p vertices. (This means that for any
two vertices u, v there is an automorphism ϕ of G satisfying ϕ(u) = v.) Let A
denote the adjacency matrix of G. Show that for any integer n ≥ 1, tr(An ) is
divisible by p (where tr denotes trace).
12. (a) (moderately difficult) (*) Let G be a (finite) connected graph on a 2m-
element vertex set V . For any graph with vertices u and v, let μ(u, v)
denote the number of edges adjacent to both u and v. Suppose that there is
an automorphism ϕ : V → V of G all of whose cycles have length two. (In
other words, ϕ is a fixed-point free involution.) Define the quotient graph
G/ϕ as follows. The vertices of G/ϕ are the orbits of ϕ. Thus G/ϕ has m
vertices. Write [v] for the orbit containing vertex v, so [v] = {v, ϕ(v)}. Set
μ([u], [v]) = μ(u, v) + μ(ϕ(u), v).
For instance, if G is a 4-cycle and ϕ takes each vertex to its antipode, then
G/ϕ consists of a double edge. If G is a 6-cycle and ϕ takes each vertex to
its antipode, then G/ϕ is a triangle. If G = K4 , then G/ϕ is a double edge
(for any ϕ). If G = K3,3 then G/ϕ is a 3-cycle for any ϕ. Show that 2κ(G)
is divisible by κ(G/ϕ), where κ denotes the number of spanning trees.
(b) (difficult) Show in fact that κ(G) is always divisible by κ(G/ϕ).
13. (a) (difficult) (*) Show that the number s(n, q) of invertible n × n symmetric
matrices over the field Fq is given by
$
q m(m−1) (q − 1)(q 3 − 1) · · · (q 2m−1 − 1), n = 2m − 1
s(n, q) =
q m(m+1) (q − 1)(q 3 − 1) · · · (q 2m−1 − 1), n = 2m.
(b) Find a formula for the number f (p) of simple graphs on the vertex set [p]
with an odd number of spanning trees.
14. (*) Let S be a k-element subset of [p]. Show that the number fS (p) of planted
forests on the vertex set [p] with exactly k components, whose set of roots is S,
is given by fS (p) = knn−k−1 . Deduce that the total number fk (p) of planted
forests on [p] with k components is given by

p p−k−1 p − 1 p−k
fk (p) = k p = p .
k k−1
Chapter 10
Eulerian Digraphs and Oriented Trees
A famous problem which goes back to Euler asks for what graphs G is there a
closed walk which uses every edge exactly once. (There is also a version for non-
closed walks.) Such a walk is called an Eulerian tour (also known as an Eulerian
cycle). A graph which has an Eulerian tour is called an Eulerian graph. Euler’s
famous theorem (the first real theorem of graph theory) states that a graph G without
isolated vertices (which clearly would be irrelevant) is Eulerian if and only if it
is connected and every vertex has even degree. Here we will be concerned with
the analogous theorem for directed graphs. We want to know not just whether
an Eulerian tour exists, but also how many there are. We will prove an elegant
determinantal formula for this number closely related to the Matrix-Tree Theorem.
For the case of undirected graphs no analogous formula is known, explaining why
we consider only the directed case.
A (finite) directed graph or digraph D consists of a vertex set V = {v1 , . . . , vp }
and edge set E = {e1 , . . . , eq }, together with a function ϕ : E → V × V (the set
of ordered pairs (u, v) of elements of V ). If ϕ(e) = (u, v), then we think of e as
an arrow from u to v. We then call u the initial vertex and v the final vertex of e,
denoted init(e) and fin(e), respectively. (These concepts arose in the definition of an
orientation in Definition 9.5.)
A walk in a digraph D is defined analogously to the undirected case.
Namely, a walk in D of length from a vertex u to a vertex v is a sequence
v1 , e1 , e2 , . . . , v , e , v+1 such that:
• each vi is a vertex of D
• each ej is an edge of D with initial vertex vi and final vertex vi+1
• init(ei ) = vi and fin(ei ) = vi+1 , for 1 ≤ i ≤
• v1 = u and v+1 = v.
There are two notions of connected in a digraph D. We say that D is strongly
connected if there is a walk (as defined above) between any two vertices. We also
say that D is weakly connected, or just connected, if D is connected as an undirected

https://doi.org/10.1007/978-3-319-77173-1_10
152 10 Eulerian Digraphs and Oriented Trees
graph, i.e., regard each edge as an undirected edge. Thus a strongly connected
digraph is (weakly) connected.
A tour in D is a sequence e1 , e2 , . . . , er of distinct edges such that the final
vertex of ei is the initial vertex of ei+1 for all 1 ≤ i ≤ r − 1, and the final vertex
of er is the initial vertex of e1 . A tour is Eulerian if every edge of D occurs at
least once (and hence exactly once). A digraph which has no isolated vertices and
contains an Eulerian tour is called an Eulerian digraph. Clearly an Eulerian digraph
is strongly connected. The outdegree of a vertex v, denoted outdeg(v), is the number
of edges of D with initial vertex v. Similarly the indegree of v, denoted indeg(v),
is the number of edges of D with final vertex v. A loop (edge e for which ϕ(e) =
(v, v)) contributes one to both the indegree and outdegree. A digraph is balanced if
indeg(v) = outdeg(v) for all vertices v.
10.1 Theorem. A digraph D without isolated vertices is Eulerian if and only if it is
connected (strongly or weakly) and balanced.
Proof. Assume D is Eulerian, and let e1 , . . . , eq be an Eulerian tour. As we move

along the tour, whenever we enter a vertex v we must exit it, except at the very end
we enter the final vertex v of eq without exiting it. However, at the beginning we
exited v without having entered it. Hence every vertex is entered as often as it is
exited and so must have the same outdegree as indegree. Therefore D is balanced,
and as noted above D is clearly strongly (and therefore weakly) connected.
Now assume that D is balanced and weakly connected. We may assume that
D has at least one edge. We first claim that for any edge e of D, D has a tour
for which e = e1 . If e1 is a loop we are done. Otherwise we have entered the
vertex fin(e1 ) for the first time, so since D is balanced there is some exit edge e2 .
Either fin(e2 ) = init(e1 ) and we are done, or else we have entered the vertex fin(e2 )
once more than we have exited it. Since D is balanced there is new edge e3 with
fin(e2 ) = init(e3 ). Continuing in this way, either we complete a tour or else we
have entered the current vertex once more than we have exited it, in which case we
can exit along a new edge. Since D has finitely many edges, eventually we must
complete a tour. Thus D does have a tour which uses e1 .
Now let e1 , . . . , er be a tour C of maximum length. We must show that r = q, the
number of edges of D. Assume to the contrary that r < q. Since in moving along C
every vertex is entered as often as it is exited (with init(e1 ) exited at the beginning
and entered at the end), when we remove the edges of C from D we obtain a digraph
H which is still balanced, though it need not be connected. However, since D is
connected, at least one connected component H1 of H contains at least one edge
and has a vertex v in common with C [why?]. Since H1 is balanced, there is an edge
e of H1 with initial vertex v. The argument of the previous paragraph shows that H1
has a tour C of positive length beginning with the edge e. But then when moving
along C, when we reach v we can take the “detour” C before continuing with C.
This gives a tour of length longer than r, a contradiction. Hence r = q, and the
theorem is proved.
10 Eulerian Digraphs and Oriented Trees 153
Our primary goal is to count the number of Eulerian tours of a connected

balanced digraph. A key concept in doing so is that of an oriented tree. An oriented
tree with root v is a (finite) digraph T with v as one of its vertices, such that there
is a unique directed path from any vertex u to v. In other words, there is a unique
sequence of edges e1 , . . . , er such that (a) init(e1 ) = u, (b) fin(er ) = v, and (c)
fin(ei ) = init(ei+1 ) for 1 ≤ i ≤ r − 1. (All three of these conditions are considered
to be vacuously true if the sequence e1 , . . . , er is empty.) It’s easy to see that this
means that the underlying undirected graph (i.e., “erase” all the arrows from the
edges of T ) is a tree, and that all arrows in T “point toward” v. From now on, an
oriented subtree of D will always mean a subdigraph of D that is an oriented tree
with the full vertex set V , or in other words, a spanning oriented subtree of D. There
is a surprising connection between Eulerian tours and oriented trees, given by the
next result.
10.2 Theorem. Let D be a connected balanced digraph with vertex set V . Fix
an edge e of D, and let v = init(e). Let τ (D, v) denote the number of oriented
(spanning) subtrees of D with root v, and let (D, e) denote the number of Eulerian
tours of D starting with the edge e. Then

(D, e) = τ (D, v) (outdeg(u) − 1)!. (10.1)
u∈V
Proof. Let e = e1 , e2 , . . . , eq be an Eulerian tour E in D. For each vertex u = v,

let e(u) be the “last exit” from u in the tour, i.e., let e(u) = ej where init(ej ) = u
and init(ek ) = u for any k > j .
Claim #1. The vertices of D, together with the edges e(u) for all vertices u = v,
form an oriented subtree of D with root v.
Proof of Claim #1. This is a straightforward verification. Let T be the spanning
subgraph of D with edges e(u), u = v. Thus if #V = p, then T has p vertices and
p − 1 edges [why?]. There are three items to check to insure that T is an oriented
tree with root v:
(a) T does not have two edges f and f satisfying init(f ) = init(f ). This is clear
since both f and f can’t be last exits from the same vertex.
(b) T does not have an edge f with init(f ) = v. This is clear since by definition
the edges of T consist only of last exits from vertices other than v, so no edge
of T can exit from v.
(c) T does not have a (directed) cycle C. For suppose C were such a cycle. Let f
be that edge of C which occurs after all the other edges of C in the Eulerian
tour E. Let f be the edge of C satisfying fin(f ) = init(f ) (= u, say). We
can’t have u = v by (b). Thus when we enter u via f , we must exit u. We can’t
exit u via f since f occurs after f in E. Hence f is not the last exit from u,
contradicting the definition of T .
It’s easy to see that conditions (a)–(c) imply that T is an oriented tree with root
v, proving the claim.
Claim #2. We claim that the following converse to Claim #1 is true. Given a
connected balanced digraph D and a vertex v, let T be an oriented (spanning)
subtree of D with root v. Then we can construct an Eulerian tour E as follows.
Choose an edge e1 with init(e1 ) = v. Then continue to choose any edge possible to
continue the tour, except we never choose an edge f of T unless we have to, i.e.,
unless it’s the only remaining edge exiting the vertex at which we stand. Then we
never get stuck until all edges are used, so we have constructed an Eulerian tour E.
Moreover, the set of last exits of E from vertices u = v of D coincides with the set
of edges of the oriented tree T .
Proof of Claim #2. Since D is balanced, the only way to get stuck is to end up at
v with no further exits available, but with an edge still unused. Suppose this is the
case. At least one unused edge must be a last exit edge, i.e., an edge of T [why?].
Let u be a vertex of T closest to v in T such that the unique edge f of T with
init(f ) = u is not in the tour. Let y = fin(f ). Suppose y = v. Since we enter y as
often as we leave it, we don’t use the last exit from y. Thus y = v. But then we can
leave v, a contradiction. This proves Claim #2.
We have shown that every Eulerian tour E beginning with the edge e has
associated with it a “last exit” oriented subtree T = T (E) with root v = init(e).
Conversely, given an oriented subtree T with root v, we can obtain all Eulerian
tours E beginning with e and satisfying T = T (E) by choosing for each vertex
u = v the order in which the edges from u, except the edge of T , appear in E; as
well as choosing the order in which all the edges from v except for e appear in E.
0 for each vertex u we have (outdeg(u) − 1)! choices, so for each T we have
Thus
u (outdeg(u) − 1)! choices. Since there are τ (D, v) choices for T , the proof is
complete.
10.3 Corollary. Let D be a connected balanced digraph, and let v be a vertex of

D. Then the number τ (D, v) of oriented subtrees with root v is independent of v.
Proof. Let e be an edge with initial vertex v. By (10.1), we need to show that
the number (D, e) of Eulerian tours beginning with e is independent of e. But
e1 e2 · · · eq is an Eulerian tour if and only if ei ei+1 · · · eq e1 e2 · · · ei−1 is also an
Eulerian tour, and the proof follows [why?].
What we obviously need to do next is find a formula for τ (D, v). This result
turns out to be very similar to the Matrix-Tree Theorem, and indeed we will show
(Example 10.6) that the Matrix-Tree Theorem is a simple corollary to Theorem 10.4.
10.4 Theorem. Let D be a digraph with vertex set V = {v1 , . . . , vp } and with li
loops at vertex vi . Let L(D) be the p × p matrix defined by
⎧
⎪
⎪ −mij , if i = j and there are mij edges with
⎨
Lij = initial vertex vi and final vertex vj
⎪
⎪
⎩
outdeg(vi ) − li , if i = j.
(Thus L is the directed analogue of the laplacian matrix of an undirected graph.)

Let L0 denote L with the last row and column deleted. Then
det L0 = τ (D, vp ). (10.2)
NOTE. If we remove the ith row and column from L instead of the last row and
column, then (10.2) still holds with vp replaced with vi .
Sketch. We first prove the case where D is not connected. In this case L0 has a set
of rows summing to 0, so det L0 = 0. Since a disconnected graph has no spanning
trees, we also have τ (D, vp ) = 0, proving the theorem when D is not connected.
Now assume that D is connected. Induction on q, the number of edges of D.
The fewest number of edges which D can have is p − 1 (since D is connected).
Suppose then that D has p − 1 edges, so that as an undirected graph D is a tree. If
D is not an oriented tree with root vp , then some vertex vi = vp of D has outdegree
0 [why?]. Then L0 has a zero row, so det L0 = 0 = τ (D, vp ). If on the other
hand D is an oriented tree with root vp , then an argument like that used to prove
Lemma 9.7 (in the case when S is the set of edges of a spanning tree) shows that
det L0 = 1 = τ (D, vp ).
Now assume that D has q > p − 1 edges, and assume the theorem for digraphs
with at most q − 1 edges. We may assume that no edge f of D has initial vertex
vp , since such an edge belongs to no oriented tree with root vp and also makes no
contribution to L0 . It then follows, since D has at least p edges, that there exists a
vertex u = vp of D of outdegree at least two. Let e be an edge with init(e) = u.
Let D1 be D with the edge e removed. Let D2 be D with all edges e removed
such that init(e) = init(e ) and e = e. (Note that D2 is strictly smaller than D
since outdeg(u) ≥ 2.) By induction or because D1 and/or D2 is not connected, we
have det L0 (D1 ) = τ (D1 , vp ) and det L0 (D2 ) = τ (D2 , vp ). Clearly τ (D, vp ) =
τ (D1 , vp ) + τ (D2 , vp ), since in an oriented tree T with root vp , there is exactly
one edge whose initial vertex coincides with that of e. On the other hand, it follows
immediately from the multilinearity of the determinant [why?] that
det L0 (D) = det L0 (D1 ) + det L0 (D2 ).
From this the proof follows by induction.

10.5 Corollary. Let D be a connected balanced digraph with vertex set V =
{v1 , . . . , vp }. Let e be an edge of D. Then the number (D, e) of Eulerian tours
of D with first edge e is given by

(D, e) = (det L0 (D)) (outdeg(u) − 1)!.
u∈V
Equivalently (since D is balanced, so Lemma 9.9 applies), if L(D) has eigenvalues

μ1 , . . . , μp with μp = 0, then
1
(D, e) = μ1 · · · μp−1 (outdeg(u) − 1)!.
p
u∈V
Proof. Combine Theorems 10.2 and 10.4.
10.6 Example (the Matrix-Tree Theorem revisited). Let G = (V , E, ϕ) be a

connected loopless undirected graph. Let G# = (V , E,
# #ϕ ) be the digraph obtained
from G by replacing each edge e of G, where ϕ(e) = {u, v}, with a pair e and
e of directed edges satisfying #ϕ (e ) = (u, v) and #ϕ (e ) = (v, u). Clearly G#
is balanced and connected. Choose a vertex v of G. There is an obvious one-to-
one correspondence between spanning trees T of G and oriented spanning trees
T# of G
# with root v, namely, direct each edge of T toward v. Moreover, L(G) =
#
L(G) [why?]. Hence the Matrix-Tree Theorem is an immediate consequence of
Theorem 10.4.
10.7 Example (the efficient mail carrier). A mail carrier has an itinerary of city
blocks to which he (or she) must deliver mail. He wants to accomplish this by
walking along each block twice, once in each direction, thus passing along houses
on each side of the street. He also wants to end up where he started, which is where
his car is parked. The blocks form the edges of a graph G, whose vertices are the
intersections. The mail carrier wants simply to walk along an Eulerian tour in the
digraph G# of the previous example. Making the plausible assumption that the graph
is connected, not only does an Eulerian tour always exist, but also we can tell the
mail carrier how many there are. Thus he will know how many different routes he
can take to avoid boredom. For instance, suppose G is the 3 × 3 grid illustrated
below.
This graph has 192 spanning trees. Hence the number of mail carrier routes
beginning with a fixed edge (in a given direction) is 192 · 1!4 2!4 3! = 18,432. The
total number of routes is thus 18,432 times twice the number of edges [why?], viz.,
18,432 × 24 = 442,368. Assuming the mail carrier delivered mail 250 days a year,
it would be 1769 years before he would have to repeat a route!
10.8 Example (binary de Bruijn sequences). A binary sequence is just a sequence
of 0’s and 1’s. A binary de Bruijn sequence of degree n is a binary sequence
A = a1 a2 · · · a2n such that every binary sequence b1 · · · bn of length n occurs
exactly once as a “circular factor” of A, i.e., as a sequence ai ai+1 · · · ai+n−1 ,
where the subscripts are taken modulo 2n if necessary. For instance, some circular
factors of the sequence abcdefg are a, bcde, fgab, and defga. Note that there
are exactly 2n binary sequences of length n, so the only possible length of a
binary de Bruijn sequence of degree n is 2n [why?]. Clearly any cyclic shift
ai ai+1 · · · a2n a1 a2 · · · ai−1 of a binary de Bruijn sequence a1 a2 · · · a2n is also a
binary de Bruijn sequence, and we call two such sequences equivalent. This relation
of equivalence is obviously an equivalence relation, and every equivalence class
contains exactly one sequence beginning with n 0’s [why?]. Up to equivalence, there
is one binary de Bruijn sequence of degree two, namely, 0011. It’s easy to check
that there are two inequivalent binary de Bruijn sequences of degree three, namely,
00010111 and 00011101. However, it’s not clear at this point whether binary de
Bruijn sequences exist for all n. By a clever application of Theorems 10.2 and 10.4,
we will not only show that such sequences exist for all positive integers n, but we
will also count the number of them. It turns out that there are lots of them. For
instance, the number of inequivalent binary de Bruijn sequences of degree eight is
equal to
1329227995784915872903807060280344576,
as the reader can easily check by writing down all these sequences. De Bruijn
sequences have a number of interesting applications to the design of switching
networks and related topics.
Our method of enumerating binary de Bruijn sequences will be to set up a
correspondence between them and Eulerian tours in a certain directed graph Dn ,
the de Bruijn graph of degree n. The graph Dn has 2n−1 vertices, which we will
take to consist of the 2n−1 binary sequences of length n − 1. The edges are indexed
by binary sequences a1 a2 · · · an with initial vertex a1 a2 · · · an−1 and final vertex
a2 a3 · · · an . Thus every vertex has indegree two and outdegree two [why?], so Dn
is balanced. The number of edges of Dn is 2n . Moreover, it’s easy to see that Dn is
connected (see Lemma 10.9). The graphs D3 and D4 look as follows:
000
00 001 100
01 10
010 101
11 011 110
111
Suppose that E = e1 e2 · · · e2n is an Eulerian tour in Dn . If fin(ei ) is the binary

sequence ai,1 ai,2 · · · ai,n−1 , then replace ei in E by the last bit ai,n−1 . For instance,
the Eulerian tour (where we simply write the vertices)
000, 000, 001, 010, 101, 011, 111, 111, 110, 101, 010, 100, 001, 011, 110, 100,000
corresponds to the sequence 0101111010011000 (the last bits of the vertices above,
excluding the first vertex 000). It is easy to see that the resulting sequence β(E) =
a1,n−1 a2,n−1 · · · a2n ,n−1 is a binary de Bruijn sequence, and conversely every binary
de Bruijn sequence arises in this way. In particular, since Dn is balanced and
connected there exists at least one binary de Bruijn sequence. In order to count the
total number of such sequences, we need to compute det L0 (Dn ). One way to do this
is by a clever but messy sequence of elementary row and column operations which
transforms the determinant into triangular form. We will give instead an elegant
computation of the eigenvalues of L(Dn ) based on the following simple lemma.
10.9 Lemma. Let u and v be any two vertices of Dn . Then there is a unique
(directed) path from u to v of length n − 1.
Proof. Suppose u = a1 a2 · · · an−1 and v = b1 b2 · · · bn−1 . Then the unique path of

length n − 1 from u to v has vertices
a1 a2 · · · an−1 , a2 a3 · · · an−1 b1 , a3 a4 · · · an−1 b1 b2 , . . . ,
an−1 b1 · · · bn−2 , b1 b2 · · · bn−1 .
10.10 Theorem. The eigenvalues of L(Dn ) are 0 (with multiplicity one) and 2 (with
multiplicity 2n−1 − 1).
Proof. Let A(Dn ) denote the directed adjacency matrix of Dn , i.e., the rows and
columns are indexed by the vertices, and for n ≥ 2 we have

1, if (u, v) is an edge
Auv =
0, otherwise.
Now Lemma 10.9 is equivalent to the assertion that An−1 = J , the 2n−1 ×
2n−1 matrix of all 1’s [why?]. If the eigenvalues of A are λ1 , . . . λ2n−1 , then the
eigenvalues of J = An−1 are λn−1 n−1
1 , . . . , λ2n−1 . By Lemma 1.4, the eigenvalues of J
are 2n−1 (once) and 0 (2n−1 − 1 times). Hence the eigenvalues of A are 2ζ (once,
where ζ is an (n − 1)-st root of unity to be determined), and 0 (2n−1 − 1 times).
Since the trace of A is 2, it follows that ζ = 1, and we have found all the eigenvalues
of A.
Now L(Dn ) = 2I − A(Dn ) [why?]. Hence the eigenvalues of L are

2 − λ1 , . . . , 2 − λ2n−1 , and the proof follows from the above determination of
λ1 , . . . , λ2n−1 .
10.11 Corollary. The number B0 (n) of binary de Bruijn sequences of degree n

beginning with n 0’s is equal to 22 −n . The total number B(n) of binary de Bruijn
n−1
n−1
sequences of degree n is equal to 22 .
Proof. By the above discussion, B0 (n) is the number of Eulerian tours in Dn whose
first edge is the loop at vertex 00 · · · 0. Moreover, the outdegree of every vertex of
Dn is two. Hence by Corollary 10.5 and Theorem 10.10 we have
1 2n−1 −1
= 22 −n .
n−1
B0 (n) = 2
2n−1
Finally, B(n) is obtained from B0 (n) by multiplying by the number 2n of edges, and
the proof follows.
n
Note that the total number of binary sequences of length 2n is N = 22 . By√the
previous corollary, the number of these which are de Bruijn sequences is just N .
This suggests the following problem, which remained open until 2009. Let An be
the set of all binary sequences of length 2n . Let Bn be the set of binary de Bruijn
sequences of degree n. Find an explicit bijection
ψ : Bn × B n → An , (10.3)
thereby giving a combinatorial proof of Corollary 10.11.
Notes for Chapter 10
The characterization of Eulerian digraphs given by Theorem 10.1 is a result of Good

[52], while the fundamental connection between oriented subtrees and Eulerian
tours in a balanced digraph that was used to prove Theorem 10.2 was shown by van
Aardenne-Ehrenfest and de Bruijn [142, Thm. 5a]. This result is sometimes called
the BEST Theorem, after de Bruijn, van Aardenne-Ehrenfest, Smith, and Tutte.
However, Smith and Tutte were not involved in the original discovery. (In [118]
Smith and Tutte give a determinantal formula for the number of Eulerian tours in
a special class of balanced digraphs. Van Aardenne-Ehrenfest and de Bruijn refer
to the paper of Smith and Tutte in a footnote added in proof.) The determinantal
formula for the number of oriented subtrees of a directed graph (Theorem 10.4) is
due to Tutte [139, Thm. 3.6].
De Bruijn sequences are named from the paper [30] of de Bruijn, where they
are enumerated in the binary case. However, it was discovered by Stanley in 1975
that this work had been done earlier by Flye Sainte-Marie [42] in 1894, as reported
by de Bruijn [32]. The generalization to d-ary de Bruijn sequences (Exercise 10.2)

is due to van Ardenne-Ehrenfest and de Bruijn [142]. Some recent work in this
area appears in a special issue [132] of Discrete Mathematics. Some entertaining
applications to magic are given by Diaconis and Graham [35, Chs. 2–4]. The
bijection ψ of (10.3) is due to Bidkhori and Kishore [9].
1. Choose positive integers a1 , . . . , ap−1 . Let D = D(a1 , . . . , ap−1 ) be the digraph

defined as follows. The vertices of D are v1 , . . . , vp . For each 1 ≤ i ≤ p − 1,
there are ai edges from xi to xi+1 and ai edges from xi+1 to xi . For instance,
D(1, 3, 2) looks like
(a) Find by a direct argument (no determinants) the number τ (D, v) of oriented
subtrees with a given root v.
(b) Find the number (D, e) of Eulerian tours of D whose first edge is e.
2. Let d > 1. A d-ary sequence is a sequence whose terms belong to {0, 1, . . . , d −
1}. A d-ary de Bruijn sequence of degree n is a d-ary sequence A = a1 a2 · · · ad n
such that every d-ary sequence b1 b2 · · · bn of length n occurs exactly once as a
circular factor of A. Find the number of d-ary de Bruijn sequences of length n
which begin with n 0’s.
3. (a) Let D be a digraph and v a vertex of D. For k ≥ 1, let D [k] denote the
digraph obtained from D by replacing each directed edge e with k edges
with the same initial and final vertices as e. Express τ (D [k] , v) in terms of
τ (D, v).
(b) Suppose also that D is balanced and connected, with vertex outdegrees
d1 , . . . , dp . Let e be an edge of D, and let e be an edge of D [k] . Express
(D [k] , e ) in terms of (D, e) and the numbers d1 , . . . , dp .
4. (a) (*) Let n, d, k ≥ 1. Show that the number of d-ary sequences
a1 a2 · · · akd n +n−1 such that every d-ary sequence b1 b2 · · · bn occurs exactly
k times as a linear factor aj aj +1 · · · aj +n−1 (so 1 ≤ j ≤ kd n ) is equal to
d n−1
(kd)!
.
k!d
(b) (more difficult) (*) Show that the number of d-ary sequences a1 a2 · · · akd n
such that every d-ary sequence b1 b2 · · · bn occurs exactly k times as a
circular factor is equal to
n−1
1 (rd)! d
φ(k/r) .
k r!d
r|k
5. Let G be a regular loopless (undirected) graph of degree d with p vertices and q

edges.
(a) Find a simple relation between p, q, and d.
(b) (*) Express the largest eigenvalue of the adjacency matrix A of G in terms
of p, q, and d.
(c) Suppose also that G has no multiple edges. Express the number of closed
walks in G of length two in terms of p, q, and d.
(d) Suppose that G has no multiple edges and that the number of closed walks
in G of length is given by
6 + 2 · (−3) .
Find the number κ(G) of spanning trees of G. (Don’t forget that A may have
some eigenvalues equal to 0.) Give a purely numerical answer, not involving
p, q, or d.
(e) Let G be as in (d). How many closed walks in G walk along each edge of G
exactly once in each direction? Give a purely numerical answer.
6. (a) (difficult) Let f (p) be the number of loopless connected digraphs D on the
vertex set [p] such that D has exactly one Eulerian tour (up to cyclic shift).
For instance, f (3) = 5; two such digraphs are triangles, and three consist of
two 2-cycles with a common vertex. Show that
f (p) = (p + 2)(p + 3) · · · (2p − 1), p ≥ 2.
(b) (somewhat more difficult) Suppose now that loops are allowed, and let g(p)
denote the resulting number of digraphs. For instance, g(1) = 2, the two
graphs being a single vertex and a single vertex with a loop. Show that
√
xp (1 − x)2 − (1 + x) 1 − 6x + x 2
g(p) = .
(p − 1)! 4x
p≥1
7. Suppose that the connected digraph D has p vertices, each of outdegree d and
indegree d. Let D be the graph obtained from D by doubling each edge, i.e.,
replacing each edge u → v with two such edges. Express (D , e ) (the number
of Eulerian tours of D beginning with the edge e of D ) in terms of (D, e).
8. Let D be a digraph with p vertices, and let be a fixed positive integer. Suppose
that for every pair u, v of vertices of D, there is a unique (directed) walk of length
from u to v.
(a) (*) What are the eigenvalues of the (directed) adjacency matrix A(D)?
(b) How many loops (v, v) does D have?
(c) (*) Show that D is connected and balanced.
(d) Show that all vertices have the same indegree d and same outdegree, which
by (c) is also d. Find a simple formula relating p, d, and .
(e) How many Eulerian tours does D have?
(f) (*) (open–ended) What more can be said about D? Show that D need not be
a de Bruijn graph (the graphs used to solve #2).
9. (a) Let n ≥ 3. Show that there does not exist a sequence a1 , a2 , . . . , an! such
that the n! circular factors ai , ai+1 , . . . , ai+n−1 (subscripts taken modulo n!
if necessary) are the n! permutations of [n].
(b) Show that for all n ≥ 1 there does exist a sequence a1 , a2 , . . . , an! such that
the n! circular factors ai , ai+1 , . . . , ai+n−2 consist of the first n − 1 terms
b1 , . . . , bn−1 of all permutations b1 , b2 , . . . , bn of [n]. Such sequences are
called universal cycles for Sn . When n = 3, an example of such a universal
cycle is 123213.
(c) When n = 3, find the number of universal cycles beginning with 123.
(d) (unsolved) Find the number Un of universal cycles for Sn beginning with
1, 2, . . . , n. It is known that
U4 = 27 · 3
U5 = 233 · 38 · 53
U6 = 2190 · 349 · 533
U7 = 21217 · 3123 · 5119 · 75 · 1128 · 4335 · 7320 · 7921 · 10935 .
Moreover, U9 is divisible by p168 , where p = 59229013196333 is prime.

Most likely there is not a “nice” formula for Un . Some, but not all, of
the factorization of Un into lots of factors can be explained using the
representation theory of Sn , a topic beyond the scope of this text.
Chapter 11
Cycles, Bonds, and Electrical Networks
11.1 The Cycle Space and Bond Space
In this chapter we will deal with some interesting linear algebra related to the
structure of a directed graph. Let D = (V , E) be a digraph. A function f : E → R
is called a circulation or flow if for every vertex v ∈ V , we have

f (e) = f (e). (11.1)
e∈E e∈E
init(e)=v fin(e)=v
Thus if we think of the edges as pipes and f as measuring the flow (quantity per
unit of time) of some commodity (such as oil) through the pipes in the specified
direction (so that a negative value of f (e) means a flow of |f (e)| in the direction
opposite the direction of e), then (11.1) simply says that the amount flowing into
each vertex equals the amount flowing out. In other words, the flow is conservative.
The figure below illustrates a circulation in a digraph D.
3
−1
5 2 4 7 π
−6
3
Let C = CD denote the set of all circulations on D. Clearly if f, g ∈ C and

α, β ∈ R then αf + βg ∈ C. Hence C is a (real) vector space, called the cycle space
of D. Thus if q = #E, then CD is a subspace of the q-dimensional vector space RE
of all functions f : E → R.

https://doi.org/10.1007/978-3-319-77173-1_11
164 11 Cycles, Bonds, and Electrical Networks
What do circulations have to do with something “circulating,” and what does the
cycle space have to do with actual cycles? To see this, define a circuit or elementary
cycle in D to be a set of edges of a closed walk, ignoring the direction of the arrows,
with no repeated vertices except the first and last. Suppose that a circuit C has been
assigned an orientation (direction of travel) o. (Note that this meaning of orientation
is not the same as that appearing in Definition 9.5.)
Define a function fC : E → R (which also depends on the orientation o, though we

suppress it from the notation) by
⎧
⎨ 1, if e ∈ C and e agrees with o
fC (e) = −1, if e ∈ C and e is opposite to o
⎩
0, otherwise.
It is easy to see that fC is a circulation. Later we will see that the circulations fC
span the cycle space C, explaining the terminology “circulation” and “cycle space.”
The figure below shows a circuit C with an orientation o, and the corresponding
circulation fC .
−1 −1
−1 1
Given a function ϕ : V → R, called a potential on D, define a new function

δϕ : E → R, called the coboundary1 of ϕ, by
δϕ(e) = ϕ(v) − ϕ(u), if u = init(e) and v = fin(e).
Figure 11.1 shows a digraph D with the value ϕ(v) of some function ϕ : V → R
indicated at each vertex v, and the corresponding values δϕ(e) shown at each edge e.
One should regard δ as an operator which takes an element ϕ of the vector space
RV of all functions V → R and produces an element of the vector space RE of all
functions E → R. It is immediate from the definition of δ that δ is linear, i.e.,
δ(aϕ1 + bϕ2 ) = a · δϕ1 + b · δϕ2 ,
1 The term “coboundary” arises from algebraic topology, but we will not explain the connection
here.
11.1 The Cycle Space and Bond Space 165
Fig. 11.1 A function 2 −1

−3
(potential) and its coboundary
4
−3 3 1 3 0
3
3
5 5 0
for all ϕ1 , ϕ2 ∈ RV and a, b ∈ R. Thus δ is simply a certain linear transformation

δ : RV → RE between two finite-dimensional vector spaces.
A function g : E → R is called a potential difference on D if g = δϕ for some
ϕ : V → R. (Later we will see the connection with electrical networks that accounts
for the terminology “potential difference.”) Let B = BD be the set of all potential
differences on D. Thus B is just the image of the linear transformation δ and is
hence a real vector space, called the bond space of D.
Let us explain the reason behind the terminology “bond space.” A bond in a
digraph D is a set B of edges such that (a) removing B from D disconnects some
(undirected) component of D (i.e., removing B creates a digraph which has more
connected components, as an undirected graph, than D), and (b) no proper subset
of B has this property. A subset of edges satisfying (a) is called a cutset, so a bond
is just a minimal cutset. Suppose, for example, that D is given as follows (with no
arrows drawn since they are irrelevant to the definition of bond):
e
b c d
Then the bonds are the six subsets ab, de, acd, bce, ace, and bcd.
Let B be a bond. Suppose B disconnects the component (V , E ) into two pieces
(a bond always disconnects some component into exactly two pieces [why?]) with
vertex set S in one piece and S̄ in the other. Thus S ∪ S̄ = V and S ∩ S̄ = ∅. Define
[S, S̄] = {e ∈ E : exactly one vertex of e lies in S and one lies in S̄}.
Clearly B = [S, S̄]. It is often convenient to use the notation [S, S̄] for a bond.
Given a bond B = [S, S̄] of D, define a function gB : E → R by
⎧
⎨ 1, if init(e) ∈ S̄, fin(e) ∈ S
gB (e) = −1, if init(e) ∈ S, fin(e) ∈ S̄
⎩
0, otherwise.
Note that gB really depends not just on B, but also on whether we write B as [S, S̄]
or [S̄, S]. Writing B in the reverse way simply changes the sign of gB . Whenever
we deal with gB we will assume that some choice B = [S, S̄] has been made.
Now note that gB = δϕ, where

1, if v ∈ S
ϕ(v) =
0, if v ∈ S.
Hence gB ∈ B, the bond space of D. We will later see that B is in fact spanned by
the functions gB , explaining the terminology “bond space.”
11.1 Example. In the digraph below, open (white) vertices indicate an element of
S and closed (black) vertices an element of S̄ for a certain bond B = [S, S̄]. The
elements of B are drawn with solid lines. The edges are labelled by the values of
gB , and the vertices by the function ϕ for which gB = δϕ.
1 0
−1
0
0 0 1 −1 0
0 1
1 1
Recall that in Definition 9.5 we defined the incidence matrix M(G) of a loopless
undirected graph G with respect to an orientation o. We may just as well think of G
together with its orientation o as a directed graph. We also will allow loops. Thus if
D = (V , E) is any (finite) digraph, define the incidence matrix M = M(D) to be
the p × q matrix whose rows are indexed by V and columns by E, as follows. The
entry in row v ∈ V and column e ∈ E is denoted mv (e) and is given by
⎧
⎨ −1, if v = init(e) and e is not a loop
mv (e) = 1, if v = fin(e) and e is not a loop
⎩
0, otherwise.
For instance, if D is given by

1
1 2 5
3
4
2
11.1 The Cycle Space and Bond Space 167
then
⎡ ⎤
1 1 −1 0 0
⎣
M(D) = −1 −1 0 1 0⎦.
0 0 1 −1 0
11.2 Theorem. The row space of M(D) is the bond space BD . Equivalently, the
functions mv : E → R, where v ranges over all vertices of D, span BD .
Proof. Let g = δϕ be a potential difference on D, so
g(e) = ϕ(fin(e)) − ϕ(init(e))

= ϕ(v)mv (e).
v∈V

Thus g = v∈V ϕ(v)m v , so g belongs to the row space of M.
Conversely, if g = v∈V ψ(v)mv is in the row space of M, where ψ : V → R,
then g = δψ ∈ B.
We now define a scalar product (or inner product) on the space RE by

f, g = f (e)g(e),
e∈E
for any f, g ∈ RE . If we think of the numbers f (e) and g(e) as the coordinates of
f and g with respect to the basis E, then f, g is just the usual dot product of f
and g. Because we have a scalar product, we have a notion of what it means for f
and g to be orthogonal, viz., f, g = 0. If V is any subspace of RE , then define the
orthogonal complement V ⊥ of V by
V ⊥ = {f ∈ RE : f, g = 0 for all g ∈ V}.
Recall from linear algebra that
dim V + dim V ⊥ = dim RE = #E. (11.2)

⊥
Furthermore, V ⊥ = V. Let us also note that since we are working over R, we
have V ∩ V ⊥ = {0}. Thus RE = V ⊕ V ⊥ (direct sum).
Intuitively there is a kind of “duality” between elementary cycles and bonds.
Cycles “hold vertices together,” while bonds “tear them apart.” The precise state-
ment of this duality is given by the next result.
11.3 Theorem. The cycle and bond spaces of D are related by C = B ⊥ .
(Equivalently, B = C ⊥ .)
Proof. Let f : E → R. Then f is a circulation if and only if

mv (e)f (e) = 0
e∈E
for all v ∈ V [why?]. But this is exactly the condition that f ∈ B ⊥ .
11.2 Bases for the Cycle Space and Bond Space
We want to examine the incidence matrix M(D) in more detail. In particular, we

would like to determine which rows and columns of M(D) are linearly independent,
and which span the row and column spaces. As a corollary, we will determine the
dimension of the spaces B and C. We begin by defining the support $f $ of f : E →
R to be the set of edges e ∈ E for which f (e) = 0.
11.4 Lemma. If 0 = f ∈ C, then $f $ contains an undirected circuit.
Proof. If not, then $f $ has a vertex of degree one [why?], which is clearly
impossible.
11.5 Lemma. If 0 = g ∈ B, then $g$ contains a bond.
Proof. Let 0 = g ∈ B, so g = δϕ for some ϕ : V → R. Choose a vertex v which is

incident to an edge of $g$, and set
U = {u ∈ V : ϕ(u) = ϕ(v)}.
Let Ū = V − U . Note that Ū = ∅, since otherwise ϕ is constant so g = 0. Since

g(e) = 0 for all e ∈ [U, Ū ] [why?], we have that $g$ contains the cutset [U, Ū ].
Since a bond is by definition a minimal cutset, it follows that $g$ contains a bond.
A matrix B is called a basis matrix of B if the rows of B form a basis for B.

Similarly define a basis matrix C of C.
Recall the notation of Theorem 9.4: let A be a matrix with at least as many
columns as rows, whose columns are indexed by the elements of a set T . If S ⊆ T ,
then A[S] denotes the submatrix of A consisting of the columns indexed by the
elements of S. In particular, A[e] (short for A[{e}]) denotes the column of A indexed
by e. We come to our first significant result about bases for the vector spaces B
and C.
11.6 Theorem. Let B be a basis matrix of B, and C a basis matrix of C. (Thus the
columns of B and C are indexed by the edges e ∈ E of D.) Let S ⊆ E, Then:
11.2 Bases for the Cycle Space and Bond Space 169
(i) The columns of B[S] are linearly independent if and only if S is acyclic (i.e.,
contains no circuit as an undirected graph).
(ii) The columns of C[S] are linearly independent if and only if S contains no bond.
Proof. The columns of B[S] are linearly dependent if and only if there exists a
function f : E → R such that
f (e) = 0 for some e ∈ S
f (e) = 0 for all e ∈ S

f (e)B[e] = 0, the column vector of 0 s. (11.3)
e∈E
The last condition is equivalent to f, mv = 0 for all v ∈ V , i.e., f is a circulation.

Thus the columns of B[S] are linearly dependent if and only if there exists a nonzero
circulation f such that $f $ ⊆ S. By Lemma 11.4, $f $ (and therefore S) contains a
circuit. Conversely, if S contains a circuit C then 0 = fC ∈ C and $fC $ = C ⊆ S,
so fC defines a linear dependence relation (11.3) among the columns. Hence the
columns of B[S] are linearly independent if and only if S is acyclic, proving (i).
(Part (i) can also be deduced from Lemma 9.7.)
The proof of (ii) is similar and is left as an exercise.
11.7 Corollary. Let D = (V , E) be a digraph with p vertices, q edges, and k
connected components (as an undirected graph). Then
dim B = p − k
dim C = q − p + k.
Proof. For any matrix X, the rank of X is equal to the maximum number of linearly
independent columns. Now let B be a basis matrix of B. By Theorem 11.6(i), the
rank of B is then the maximum size (number of elements) of an acyclic subset of E.
In each connected component Di of D, the largest acyclic subsets are the spanning
trees, whose number of edges is p(Di ) − 1, where p(Di ) is the number of vertices
of Di . Hence

k
rank B = (p(Di ) − 1)
i=1
= p − k.
Since dim B + dim C = dim RE = q by (11.2) and Theorem 11.3, we have
dim C = q − (p − k) = q − p + k.
(It is also possible to determine dim C by a direct argument similar to our

determination of dim B.)
The number q − p + k (which should be thought of as the number of independent

cycles in D) is called the cyclomatic number of D (or of its undirected version G,
since the direction of the edges has no effect).
Our next goal is to describe explicit bases of C and B. Recall that a forest is
an undirected graph without circuits, or equivalently, a disjoint union of trees. We
extend the definition of forest to directed graphs by ignoring the arrows, i.e., a
directed graph is a forest if it has no circuits as an undirected graph. Equivalently
[why?], dim C = 0.
Pick a maximal forest T of D = (V , E). Thus T restricted to each component of
D is a spanning tree. If e is an edge of D not in T , then it is easy to see that T ∪ e
contains a unique circuit Ce .
11.8 Theorem. Let T be as above. Then the set S of circulations fCe , as e ranges
over all edges of D not in T , is a basis for the cycle space C.
Proof. The circulations fCe are linearly independent, since for each e ∈ E(D) −
E(T ) only fCe doesn’t vanish on e. Moreover,
#S = #E(D) − #E(T ) = q − p + k = dim C,
so S is a basis.
11.9 Example. Let D be the digraph shown below, with the edges a, b, c of T
shown by dotted lines.
b
c f
a d
Orient each circuit Ct in the direction of the added edge, i.e., fCt (t) = 1. Then the
basis matrix C of C corresponding to the basis fCd , fCe , fCf is given by
⎡ ⎤
0 −1 −1 1 0 0
C = ⎣ −1 −1 −1 0 1 0 ⎦ . (11.4)
0 0 −1 0 0 1
We next want to find a basis for the bond space B analogous to that of
Theorem 11.8.
11.10 Lemma. Let T be a maximal forest of D = (V , E). Let T ∗ = D − E(T )
(the digraph obtained from D by removing the edges of T ), called a cotree if D is
connected. Let e be an edge of T . Then E(T ∗ ) ∪ e contains a unique bond.
Proof. Removing E(T ∗ ) from D leaves a maximal forest T , so removing one
further edge e disconnects some component of D. Hence E(T ∗ ) ∪ e contains a bond
B. It remains to show that B is unique. Removing e from T breaks some component
11.2 Bases for the Cycle Space and Bond Space 171
of T into two connected graphs T1 and T2 with vertex sets S and S̄. It follows [why?]
that we must have B = [S, S̄], so B is unique.
Let T be a maximal forest of the digraph D, and let e be an edge of T . By the

previous lemma, E(T ∗ )∪e contains a unique bond Be . Let gBe be the corresponding
element of the bond space B, chosen for definiteness so that gBe (e) = 1.
11.11 Theorem. The set of functions gBe , as e ranges over all edges of T , is a basis
for the bond space B.
Proof. The functions gBe are linearly independent, since only gBe is nonzero on
e ∈ E(T ). Since
#E(T ) = p − k = dim B,
it follows that the gBe ’s are a basis for B.
11.12 Example. Let D and T be as in the previous diagram. Thus a basis for B is
given by the functions gBa , gBb , gBc . The corresponding basis matrix is given by
⎡ ⎤
100010
B = ⎣0 1 0 1 1 0⎦.
001111
Note that the rows of B are orthogonal to the rows of the matrix C of (11.4), in
accordance with Theorem 11.3. Equivalently, BC t = 0, the 3 × 3 zero matrix.
(In general, BC t will have q − p + k rows and p − k columns. Here it is just a
coincidence that these two numbers are equal.)
The basis matrices C T and B T of C and B obtained from a maximal forest T have
an important property. A real m × n matrix A with m ≤ n is said to be unimodular
if every m × m submatrix has determinant 0, 1, or −1. For instance, the adjacency
matrix M(D) of a digraph D is unimodular, as proved in Lemma 9.7 (by showing
that the expansion of the determinant of a full submatrix has at most one nonzero
term).
11.13 Theorem. Let T be a maximal forest of D. Then the basis matrices C T of C
and B T of B are unimodular.
Proof. First consider the case C T . Let P be a full submatrix of C (so P has q −p+k
rows and columns). Assume det P = 0. We need to show det P = ±1. Since
det P = 0, it follows from Theorem 11.6(ii) that P = C T [T1∗ ] for the complement
T1∗ of some maximal forest T1 . Note that the rows of the matrix C T [T1∗ ] are indexed
by T ∗ and the columns by T1∗ . Similarly the rows of the basis matrix C T1 are indexed
by T1∗ and the columns by E (the set of all edges of D). Hence it makes sense to
define the matrix product
Z = C T [T1∗ ]C T1 ,
a matrix whose rows are indexed by T ∗ and columns by E.

Note that the matrix Z is a basis matrix for the cycle space C since its rows are
linear combinations of the rows of the basis matrix CT1 , and it has full rank since the
matrix C T [T1∗ ] is invertible. Now C T1 [T1∗ ] = IT1∗ (the identity matrix indexed by
T1∗ ), so Z[T1∗ ] = C T [T1∗ ]. Thus Z agrees with the basis matrix C T in columns T1∗ .
Hence the rows of Z − C T are circulations supported on a subset of T1 . Since T1 is
acyclic, it follows from Lemma 11.4 that the only such circulation is identically 0,
so Z = C T .
We have just shown that
C T [T1∗ ]C T1 = C T .
Restricting both sides to T ∗ , we obtain
C T [T1∗ ]C T1 [T ∗ ] = C T [T ∗ ] = IT ∗ .
Taking determinants yields
det(C T [T1∗ ]) det(C T1 [T ∗ ]) = 1.
Since all the matrices we have been considering have integer entries, the above
determinants are integers. Hence
det C T [T1∗ ] = ±1,
A similar proof works for B T .
11.3 Electrical Networks
We will give a brief indication of the connection between the above discussion and
the theory of electrical networks. Let D be a digraph, which for convenience we
assume is connected and loopless. Suppose that at each edge e there is a voltage
(potential difference) Ve from init e to fin e, and a current Ie in the direction of e (so
a negative current Ie indicates a current of |Ie | in the direction opposite to e). Think
of V and I as functions on the edges, i.e., as elements of the vector space RE . There
are three fundamental laws relating the quantities Ve and Ie .
Kirchhoff’s First Law. I ∈ CD . In other words, the current flowing into a vertex
equals the current flowing out. In symbols,
11.3 Electrical Networks 173

Ie = Ie ,
e e
init e=v fin e=v
for all vertices v ∈ V .

Kirchhoff’s Second Law. V ∈ CD ⊥ = B. In other words, the sum of the voltages
around any circuit (called loops by electrical engineers), taking into account
orientations, is 0.
Ohm’s Law. If edge e has resistance Re > 0, then Ve = Ie Re .
The central problem of electrical network theory dealing with the above three
laws2 is the following: which of the 3q quantities Ve , Ie , Re need to be specified
to uniquely determine all the others, and how can we find or stipulate the solution
in a fast and elegant way? We will be concerned here only with a special case,
perhaps the most important special case in practical applications. Namely, suppose
we apply a voltage Vq = 0 at edge eq , with resistances R1 , . . . , Rq−1 at the other
edges e1 , . . . , eq−1 . Let Vi , Ii be the voltage and current at edge ei . We would like
to express each Vi and Ii in terms of Vq and R1 , . . . , Rq−1 . By “physical intuition”
there should be a unique solution, since we can actually build a network meeting the
specifications of the problem. Note that if we have quantities Vi , Ii , Ri satisfying the
three network laws above, then for any scalar α the quantities αVi , αIi , Ri are also
a solution. This means that we might as well assume that Vq = 1, since we can
always multiply all voltages and currents afterwards by whatever value we want Vq
to be.
When we apply a voltage Vq the current will flow along eq from the lower
potential (say at vertex u) to the higher (at vertex v). Since by convention the
direction of current flow is from the higher potential to the lower, if we orient eq
from u to v, then Vq < 0 and Iq > 0. Thus Vq /Iq < 0 so we should define the
total resistance R(D) of the network D, together with the distinguished edge e, by
R(D) = −Vq /Iq . If we replace the entire network D except for the edge eq by a
single edge from u to v, then the resistance of this edge will be R(D).
As an illustration of a simple method of computing the total resistance of a
network, the following diagram illustrates the notion of a series connection D1 +D2
and a parallel connection D1 $ D2 of two networks D1 and D2 with a distinguished
edge e at which a voltage is applied.
A e B e e A B e
A
D1 D2 D1 || D2
D1 + D 2
2 Ofcourse the situation becomes much more complicated when one introduces dynamic network
elements like capacitors, alternating current, etc.
It is well-known and easy to deduce from the three network laws that
R(D1 + D2 ) = R(D1 ) + R(D2 )
1 1 1
= + . (11.5)
R(D1 $ D2 ) R(D1 ) R(D2 )
A network that is built up from a single edge by a sequence of series and parallel
connections is called a series–parallel network. An example is the following, with
the distinguished edge e shown by a broken line from bottom to top.
The simplest network which is not a series-parallel network is called the Wheatstone
bridge and is illustrated below. (The direction of the arrows has been chosen
arbitrarily.) We will use this network as our main example in the discussion that
follows.
1 2
3
6
4 5
We now return to an arbitrary connected loopless digraph D, with currents Ii ,

voltages Vi , and resistances Ri at the edges ei . Recall that we are fixing Vq = 1
and R1 , . . . , Rq−1 . Let T be a spanning tree of D. Since I is a current if and only
if it is orthogonal to the bond space B (Theorem 11.3 and Kirchhoff’s First Law), it
follows that any basis for B defines a complete and minimal set of linear relations
satisfied by the Ii ’s (namely, the relation that I is orthogonal to the basis elements).
In particular, the basis matrix B T defines such a set of relations. For example, if D
is the Wheatstone bridge shown above and if T = {e1 , e2 , e5 }, then we obtain the
following relations by adding the edges e1 , e2 , e5 of T in turn to T ∗ :
I1 − I3 − I4 = 0
I2 + I3 + I4 + I6 = 0 (11.6)
I4 + I5 + I6 = 0.
11.3 Electrical Networks 175
These three (= p − 1) equations give all the relations satisfied by the Ii ’s alone, and
the equations are linearly independent.
Similarly if V is a voltage then it is orthogonal to the cycle space C. Thus any
basis for C defines a complete and minimal set of linear relations satisfied by the Vi ’s
(namely, the relation that V is orthogonal to the basis elements). In particular, the
basis matrix CT defines such a set of relations. Continuing our example, we obtain
the following relations by adding the edges e3 , e4 , e6 of T ∗ in turn to T .
V1 − V2 + V3 = 0
V1 − V2 + V4 − V5 = 0 (11.7)
V2 + V5 = 1,
These three (= q − p + k) equations give all the relations satisfied by the Vi ’s alone,
and the equations are linearly independent.
In addition, Ohm’s Law gives the q − 1 equations Vi = Ri Ii , 1 ≤ i ≤ q − 1.
We have a total of (p − k) + (q − p + k) + (q − 1) = 2q − 1 equations in the
2q − 1 unknowns Ii (1 ≤ i ≤ q) and Vi (1 ≤ i ≤ q − 1). Moreover, it is easy to see
that these 2q − 1 equations are linearly independent, using the fact that we already
know that just the equations involving the Ii ’s alone are linearly independent, and
similarly the Vi ’s. Hence this system of 2q − 1 equations in 2q − 1 unknowns has a
unique solution. We have now reduced the problem to straightforward linear algebra.
However, it is possible to describe the solution explicitly. We will be content here
with giving a formula just for the total resistance R(D) = −Vq /Iq = 1/Iq . (Recall
that we take Vq < 0.)
Write the 2q − 1 equations in the form of a (2q − 1) × 2q matrix K. The columns
of the matrix are indexed by I1 , I2 , . . . , Iq , V1 , V2 . . . , Vq . The last column Vq of
the matrix keeps track of the constant terms of the equations. The rows of K are
given first by the equations among the Ii ’s, then the Vi ’s, and finally Ohm’s Law.
For our example of the Wheatstone bridge, we obtain the matrix
I1 I2 I3 I4 I5 I6 V1 V2 V3 V4 V5 V6
1 0 −1 −1 0 0 0 0 0 0 0 0
0 1 1 1 0 1 0 0 0 0 0 0
0 0 0 1 1 1 0 0 0 0 0 0
0 0 0 0 0 0 1 −1 1 0 0 0
K= 0 0 0 0 0 0 1 −1 0 1 −1 0
0 0 0 0 0 0 0 −1 0 0 −1 −1
R1 0 0 0 0 0 −1 0 0 0 0 0
0 R2 0 0 0 0 0 −1 0 0 0 0
0 0 R3 0 0 0 0 0 −1 0 0 0
0 0 0 R4 0 0 0 0 0 −1 0 0
0 0 0 0 R5 0 0 0 0 0 −1 0
We want to solve for Iq by Cramer’s rule. Call the submatrix consisting of all but
the last column X. Let Y be the result of replacing the Iq column of X by the last
column of K. Cramer’s rule then asserts that
det Y
Iq = .
det X
We evaluate det X by taking a Laplace expansion along the first p − 1 rows. In other
words,

det X = ± det(X[[p − 1], S]) · det(X[[p − 1]c , S̄]), (11.8)
S
where (a) S indexes all (p − 1)-element subsets of the columns, (b) X[[p − 1], S]
denotes the submatrix of X consisting of entries in the first p − 1 rows and in
the columns S, and (c) X[[p − 1]c , S̄] denotes the submatrix of X consisting of
entries in the last 2q − p rows and in the columns other than S. In order for
det(X[[p −1], S]) = 0, we must choose S = {Ii1 , . . . , Iip−1 }, where {ei1 , . . . , eip−1 }
is a spanning tree T1 (by Theorem 11.6(i)). In this case, det(X[[p − 1], S]) = ±1 by
Theorem 11.13. If Iq ∈ S, then the Iq column of X[[p − 1]c , S̄] will be zero. Hence
to get a nonzero term in (11.8), we must have eq ∈ S. The matrix X[[p − 1]c , S̄]
will have one nonzero entry in each of the first q − p + 1 columns, namely, the
resistances Rj where ej is not an edge of T1 . This accounts for q − p + 1 entries
from the last q − 1 rows of X[[p − 1]c , S̄]. The remaining p − 2 of the last q − 1
rows have available only one nonzero entry each, namely, a −1 in the columns
indexed by Vj where ej is an edge of T1 other than eq . Hence we need to choose
q − p + 1 remaining entries from rows p through q and columns indexed by Vj for
ej not an edge of T1 . By Theorems 11.6(ii) and 11.13, this remaining submatrix has
determinant ±1. It follows that

det(X[[p − 1], S]) · det(X[[p − 1]c , S̄]) = ± Rj .
ej ∈E(T1 )
Hence by (11.8), we get

⎛ ⎞

det X = ±⎝ Rj ⎠ , (11.9)
T1 ej ∈E(T1 )
where T1 ranges over all spanning trees of D containing eq . A careful analysis of the
signs (omitted here) shows that all signs in (11.9) are positive, so we finally arrive
at the remarkable formula

det X = Rj .
spanning trees T1 ej ∈E(T1 )
containing eq
11.4 Planar Graphs (Sketch) 177
For example, if D is the Wheatstone bridge as above, and if we abbreviate R1 = a,

R2 = b, R3 = c, R4 = d, R5 = e, then
det X = abc + abd + abe + ace + ade + bcd + bde + cde.
Now suppose we replace column Iq in X by column Vq in the matrix K,

obtaining the matrix Y . There is a unique nonzero entry in the new column, so it
must be chosen in any nonzero term in the expansion of det Y . The argument now
goes just as it did for det X, except we have to choose S to correspond to a spanning
tree T1 that doesn’t contain eq . We therefore obtain

det Y = Rj .
not containing eq ej =eq
For example, for the Wheatstone bridge we get
det Y = ac + ad + ae + bc + bd + be + cd + ce.
Recall that Iq = det(Y )/ det(X) and that the total resistance of the network is
1/Iq . Putting everything together gives our main result on electrical networks.
11.14 Theorem. In the situation described above, the total resistance of the network
is given by

Rj
1 containing eq
R(D) = = .
Iq Rj
not containing eq ej =eq
11.15 Corollary. If the resistances R1 , . . . , Rq−1 are all equal to one, then the total
resistance of the network is given by
1 number of spanning trees containing eq

R(D) = = .
Iq number of spanning trees not containing eq
In particular, if R1 = · · · = Rq−1 = 1, then the total resistance, when reduced to

lowest terms a/b, has the curious property that the number κ(D) of spanning trees
of D is divisible by a + b.
11.4 Planar Graphs (Sketch)
A graph G is planar if it can be drawn in the plane R2 without crossing edges. A

drawing of G in this way is called a planar embedding. An example of a planar
embedding is shown in Figure 11.2. In this section we state the basic results on the
bond and cycle spaces of a planar graph. The proofs are relatively straightforward
and are omitted.
If the vertices and edges of a planar embedding of G are removed from R2 , then
we obtain a disjoint union of open sets, called the faces (or regions) of G. (More
precisely, these open sets are the faces of the planar embedding of G. Often we
will not bother to distinguish between a planar graph and a planar embedding if no
confusion should result.) Let R = R(G) be the set of faces of G, and as usual V (G)
and E(G) denote the set of vertices and edges of G, respectively.
NOTE. If G is a simple (no loops or multiple edges) planar embedding, then it
can be shown that there exists a planar embedding of the same graph with edges
as straight lines and with faces (regarding as the sequence of vertices and edges
obtained by walking around the boundaries of the faces) preserved.
The dual G∗ of the planar embedded graph G has vertex set R(G) and edge set
E ∗ (G) = {e∗ : e ∈ E(G)}. If e is an edge of G, then let r and r be the faces on its
two sides. (Possibly r = r ; there are five such edges in Figure 11.2.) Then define e∗
to connect r and r . We can always draw G∗ to be planar, letting e and e∗ intersect
once. If G is connected then every face of G∗ contains exactly one (nonisolated)
vertex of G and G∗∗ ∼ = G. For any planar embedded graph G, the dual G∗ is
∼
connected. Then G = G∗∗ if and only if G is connected. In general, we always have
Fig. 11.2 A planar

embedding
Fig. 11.3 A planar embedding and its dual

11.4 Planar Graphs (Sketch) 179
G∗ ∼= G∗∗∗ . Figure 11.3 shows the dual G∗ to the graph G of Figure 11.2, with the
vertices of G∗ drawn as open circles and the edges as broken lines.
11.16 Example. Let G consist of two disjoint edges. Then G∗ has one vertex and
two loops, while G∗∗ is a three-vertex path. The unbounded face of G∗ contains two
vertices of G, and G∗∗ ∼
= G.
Orient the edges of the planar graph G in any way to get a digraph D. Let r be
an interior (i.e., bounded) face of D. An outside edge of r is an edge e such that r
lies on one side of the edge, and a different face lies on the other side. The outside
edges of any interior face r define a circulation (shown as solid edges in the diagram
below), and these circulations (as r ranges over all interior faces of D) form a basis
for the cycle space CG of G.
Given the orientation D of G, orient the edges of G∗ as follows: as we walk

along e in the direction of its orientation, e∗ points to our right.
11.17 Theorem. Let f : E(G) → R. Define f ∗ : E(G∗ ) → R by f ∗ (e∗ ) = f (e).

Then
f ∈ BG ⇔ f ∗ ∈ CG∗
f ∈ CG ⇔ f ∗ ∈ BG∗ .
11.18 Proposition. The set S is the set of edges of a spanning tree T of G if and
only if S ∗ = {e∗ : e ∈ S} is the set of edges of a cotree T ∗ of G∗ .
11.19 Corollary. κ(G) = κ(G∗ )
For nonplanar graphs there is still a notion of a “dual” object, but it is no longer a
graph but rather something called a matroid. Matroid theory is a flourishing subject
which may be regarded as a combinatorial abstraction of linear algebra.
11.5 Squaring the Square
A squared rectangle is a rectangle partitioned into finitely many (but more than one)
squares. A squared rectangle is perfect if all the squares are of different sizes. The
earliest perfect squared rectangle was found in 1936; its size is 33 × 32 and consists
of nine squares, as shown in Figure 11.4.
The question then arose: does there exist a perfect squared square? A single
example was found by Sprague in 1939; it has 55 squares. Then Brooks, Smith,
Stone, and Tutte developed a network theory approach which we now explain.
The Smith diagram D of a squared rectangle R is a directed graph whose vertices
are the horizontal line segments of R and whose edges are the squares of R, directed
from top to bottom. The top vertex (corresponding to the top edge of R) and the
bottom vertex (corresponding to the bottom edge) are called poles. Label each edge
by the side length of the square to which it corresponds. Figure 11.5 shows the Smith
diagram of the (perfect) squared rectangle in Figure 11.4.
The following result concerning Smith diagrams is straightforward to verify.
11.20 Theorem. (a) If we set Ie and Ve equal to the label of edge e, then
Kirchhoff’s two laws hold (so Re = 1) except at the poles.
(b) The Smith diagram is planar and can be drawn without separation of poles.
Joining the poles by an edge from the bottom to the top gives a 3-connected
graph, i.e., a connected graph that remains connected when one or two vertices
are removed.
Call the 3-connected graph of Theorem 11.20 the extended Smith diagram of the
a ×b squared rectangle. If we impose a current Ie1 = b on the new edge e1 (directed
Fig. 11.4 A squared

rectangle
8 9
15 1
7
10
4
18
14
11.5 Squaring the Square 181
Fig. 11.5 A Smith diagram pole
9
8
1
15
7
10
4
18
14
pole
from bottom to top) between poles, and a voltage Ve1 = −a, then Kirchhoff’s two
laws hold at all vertices. The diagram below shows the extended Smith diagram
corresponding to Figure 11.5, with the new edge e1 labelled by the current Ie1 .
pole
8 9
15 1
7
10
32 4
18
14
pole
We therefore have a recipe for searching for perfect squared rectangles and
squares: start listing all three-connected planar graphs. Then choose an edge e1 to
apply a voltage V1 . Put a resistance Re = 1 at the remaining edges e. Solve for Ie
(= Ve ) to get a squared rectangle, and hope that one of these will be a square. One
example found by Brooks et al. was a 112 × 75 rectangle with 14 squares. It was
given to Brooks’ mother as a jigsaw puzzle, and she found a different solution !
We therefore have found a squared square (though not perfect):
Δ 75 x 75
112 x 112 Γ
Building on this idea, Brooks et al. finally found two 422×593 perfect rectangles
with thirteen squares, all 26 squares being of different sizes. Putting them together
as above gives a perfect squared square. This example has two defects: (a) it contains
a smaller perfect squared rectangle (and is therefore not simple), and (b) it contains
a “cross” (four squares meeting a point). They eventually found a perfect squared
square with 69 squares without either of these defects. It is now known (thanks to
computers) that the smallest order (number of squares) of a perfect squared square
is 21. It is unique and happens to be simple and crossfree. See the figure below. It
is known that the number (up to symmetry) of simple perfect squared squares of
order n for 21 ≤ n ≤ 35 is 1, 8, 12, 26, 160, 441, 1152, 3001, 7901, 20566, 54541,
144161, 378197, 990981, 2578081.
27
35
50
8
19
15 17 11
2 6
9 7 24
25 18
29
16
4
37 42
33
The theory of cycle spaces and bond spaces developed here had its origins with the
pioneering work of Kirchhoff [75] in 1847. The proof given here of Theorem 11.13
is due to Tutte [140] in 1965. A nice account of the history of squaring the square
due to Tutte appears in a Scientific American column by Martin Gardner [51]. See
also [141] for another article by Tutte. A further survey article on this topic is
by Kazarinoff and Weitzenkamp [73]. For some interesting connections between
random walks and electrical networks, see Doyle and Snell [36].
1. (a) Let Cn be the graph of the n-cube. Find the dimension of the bond space
and cycle space of Cn . Does there exist a circulation (with respect to some
orientation of Cn ) supported on three edges?
(b) Show that the cycle space CCn (with respect to some orientation of Cn ) is
spanned by circulations fC , where C is a circuit of length four.
2. We have defined real-valued flows f : E → R, but we could just as easily
allow the values to be in any abelian group A, i.e., f : E → A such that (11.1)
is satisfied.
(a) A bridge or isthmus in a graph G (directed or undirected) is an edge e
whose removal disconnects the connected component (as an undirected
graph in the case where G is directed) to which it belongs. For instance,
G is a forest if and only if every edge is a bridge. If e is an isthmus of the
digraph D and f : E → A is a flow, then show that f (e) = 0.
(b) Let G be an undirected (finite) graph and o an orientation. For a positive
integer n, define CG (n) to be the number of flows f : G → Zn (with
respect to the orientation o, where Zn denotes the integers modulo n) which
never take the value 0. Show that CG (n) depends only on G and n, not on o.
(c) Let G be a finite bridgeless graph, i.e., a graph with no bridges. Show that
CG (n) is a polynomial function of n whose degree is the dimension of the
cycle space of G. (If G has an bridge, then CG (n) = 0 by part (a) of this
exercise.)
(d) Let Gp denote the cycle of length p ≥ 3. Find CGp (n).
(e) (*) Let d = deg CG (n). Show that for 0 ≤ i ≤ d, the coefficient of ni in
the polynomial (−1)d CG (−n) is positive.
(f) (difficult) Let Kp denote the complete graph with p vertices. Show that
⎛ ⎞n
p−1 xp k x k
(−1)( 2) CK (n)
p = 1 − ⎝ (1 − n)(2) (−n)−k ⎠
p! k!
p≥0 k≥0
x3
= x − (n − 1)
3!
x4
−(n − 1)(n − 2)(n − 3)
4!
x5
+(n − 1)(n5 − 9n4 + 36n3 − 79n2 + 96n − 51) + ··· .
5!
(g) Find some positive integer j with the following property: for any finite
bridgeless graph G we have CG (j ) > 0.
(h) (very difficult) Show that if G is finite bridgeless graph, then CG (6) > 0.
(i) (unsolved) Show that if G is finite bridgeless graph, then CG (5) > 0. This
inequality is a famous conjecture of Tutte, known as the nowhere-zero 5-
flow conjecture.
3. What digraphs have the property that every nonempty set of edges is a cutset?
4. What is the size of the largest set of edges of the complete graph Kp that doesn’t
contain a bond? How many such sets are there?
5. (*) The cycle space CD and bond space BD of a finite digraph D were defined
over R. However, the definition works over any field K, even those of positive
characteristic. Show that the dimensions of these two spaces remain the same
for any K, i.e., dim CD = q − p + k and dim BD = p − k.
6. (a) Let Kp be the complete graph on the vertex set [p]. Suppose that a voltage
Ve = 1 is applied to an edge e, and all other edges have a resistance of
1. Without doing any computation, find the current If along any edge f
disjoint from e (i.e., e and f have no common vertices).
(b) (*) Using (a), find the total resistance of the network of (a).
(c) Check that your answer to (b) agrees with Corollary 11.15.
7. (a) A graph G is edge-transitive if its automorphism group Aut(G) is transitive
on the edges, i.e., for any two edges e, e of G, there is an automorphism φ
which takes e to e . For instance, the cube graph Cn is edge-transitive. Is an
edge-transitive graph also vertex-transitive? What about conversely? If we
consider only simple graphs (no loops or multiple edges), does that affect
the answers?
(b) Suppose that G is edge-transitive and has p vertices and q edges. A one
volt battery is placed on one edge e, and all other edges have a resistance
of one ohm. Express the total resistance Re = −Ve /Ie of the network in
terms of p and q.
8. Let D be a loopless connected digraph with q edges. Let T be a spanning tree
of D. Let C be the basis matrix for the cycle space C of D obtained from T , and
similarly B for the bond space (as described in Theorems 11.8 and 11.11).
(a) (*) Show that det CC t = det BB t = κ(D), the number of spanning trees
of D (ignoring the directions of the edges).
(b) (*) Let

C
Z= ,
B
a q × q matrix. Show that det Z = ±κ(D).

9. (difficult) Let M be an m × n real unimodular matrix such that every two
columns of M are linearly independent. Show that n ≤ m2
10. Let D be a planar electrical network with edges e1 , . . . , eq . Place resistances
Ri = 1 at ei , 1 ≤ i ≤ q − 1, and place a voltage Vq = 1 at eq . Let D ∗ be the
dual network, with the same resistances Ri at ei∗ and voltage Vq∗ at eq∗ . What is
the connection between the total resistances R(D) and R(D ∗ )?
11. Let D be the extended Smith diagram of a squared rectangle, with the current
I and voltage V as defined in the text. What is the “geometric” significance of
the fact that I, V = 0?
12. Let D be the extended Smith diagram of an a × b squared rectangle. Show that
the number of spanning trees of D is divisible by a + b.
Chapter 12
A Glimpse of Combinatorial
Commutative Algebra
12.1 Simplicial Complexes
In this chapter we will discuss a profound connection between commutative rings

and some combinatorial properties of simplicial complexes. The deepest and most
interesting results in this area require a background in algebraic topology and
homological algebra beyond the scope of this book. However, we will be able to
prove a highly nontrivial combinatorial result that relies on commutative algebra
(i.e., the theory of commutative rings and modules over them) in an essential way.
This result is our Theorem 12.25, the characterization of f -vectors of shellable
simplicial complexes. Of course we must first define these terms and then set up
the necessary machinery.
Let V = {x1 , . . . , xn } be a finite set, called a vertex set. An abstract simplicial
complex on V , or just simplicial complex for short, is a collection of subsets of
V satisfying the following two conditions (of which the second is the significant
one):
1. {xi } ∈ for 1 ≤ i ≤ n
2. If F ∈ and G ⊆ F , then G ∈ .
An element F of is called a face. A maximal face F , i.e., a face that is not
contained in any larger face, is called a facet. The dimension of F is #F − 1. In
particular, the empty set ∅ is a face of dimension −1, unless = ∅ (see the remark
below concerning empty simplicial complexes and empty faces). An i-dimensional
face is called an i-face. Soon we will see the geometric reason for our definition of
dimension.
12.1 Remark. There is a small subtlety about the definition of simplicial complex
that can lead to confusion. Namely, one must distinguish between the empty
simplicial complex = ∅ which has no faces whatsoever, and the simplicial
complex = {∅} whose only face is the empty set ∅.

https://doi.org/10.1007/978-3-319-77173-1_12
188 12 A Glimpse of Combinatorial Commutative Algebra
If is any finite collection of finite sets, then denotes the smallest simplicial
complex containing the elements of . Thus
= {F : F ⊆ G for some G ∈ }.
In presenting examples we will often abbreviate a set such as {1, 2, 3} as simply

123. Thus for instance 123, 14, 24 denotes the simplicial complex with faces
∅, 1, 2, 3, 4, 12, 13, 23, 14, 24, 123.
It is worthwhile to understand simplicial complexes geometrically, though such

understanding is not really germane to our main results here. Let us first review some
basic definitions. A convex set in Rd is a subset S of Rd such that if u, v ∈ S, then the
line segment joining u and v is also in S. Equivalently, λu + (1 − λ)v ∈ S for all real
numbers 0 ≤ λ ≤ 1. Clearly the intersection of convex sets is convex. The convex
hull of any subset S of Rd , denoted conv(S), is defined to be the intersection of all
convex sets containing S. It is therefore the smallest convex set in Rd containing S.
A set {v0 , v1 , . . . , vj } ⊂ Rd is affinely independent
if the following
condition
holds: if α0 , α1 , . . . , αj are real numbers for which αi vi = 0 and αi = 0, then
α0 = α1 = · · · = αj = 0. Equivalently, define an affine subspace of Rd to be the
translate of a linear subspace, i.e., a set
A = {v ∈ Rd : v · y (1) = α1 , . . . , v · y (k) = αk },
where y (1) , . . . , y (k) ∈ Rd (with each y (i) = 0) and α1 , . . . , αk ∈ R are fixed, and
where v ·y denotes the usual dot product in Rd . The dimension of A is the dimension
of the linear subspace
{v ∈ Rd : v · y (1) = 0, . . . , v · y (k) = 0}.
The affine span of a subset S of Rd , denoted aff(S), is the intersection of all affine
subspaces containing S. It is easy to see that aff(S) is itself an affine subspace. It
is then true that a set of k + 1 points of Rd is affinely independent if and only if
its affine span has dimension k, the maximum possible. In particular, the largest
number of points of an affinely independent subset of Rd is d + 1.
A simplex (plural simplices) σ in Rd is the convex hull of an affinely independent
subset of Rd . The dimension of a simplex σ is the dimension of its affine span.
Equivalently, if σ is the convex hull of j + 1 affinely independent points, then
dim σ = j . If S is affinely independent and σ = conv(S), then a face of σ is a
set conv(T ) for some T ⊆ S. In particular, taking T = ∅ shows that ∅ is a face of
σ . A face τ of dimension +1zero
(i.e., τ is a single point) is called a vertex of . If
dim σ = j , then σ has ji+1 i-dimensional faces. For instance, a zero-dimensional
simplex is a point, a one-dimensional simplex is a line segment, a two-dimensional
simplex is a triangle, a three-dimensional simplex is a tetrahedron, etc.
12.1 Simplicial Complexes 189
A (finite) geometric simplicial complex is a finite set of simplices in Rd such

that the following two conditions hold:
1. If σ ∈ and τ is a face of σ , then τ ∈ .
2. If σ, τ ∈ , then σ ∩ τ is a common face (possibly empty) of σ and τ .
1
We sometimes identify with the union σ ∈ σ of its simplices. In this situation
is just a subset of Rd , but it is understood that it has been described as a union of
certain simplices.
There is an obvious abstract simplicial complex that we can associate with a
geometric simplicial complex. Namely, the vertex set V of consists of the vertices
of , and a set F of vertices of is a face of if F is the set of vertices of some
simplex σ ∈ . We then say that (regarded as a union of its simplices) is a
geometric realization of , denoted = ||. Note that if F is a face of with
k + 1 vertices, then it corresponds to a k-dimensional simplex in , explaining why
we defined dim F = #F − 1.
NOTE. In some situations it is useful for to have a unique (canonical) geomet-
ric realization. We can do this as follows. Suppose that has n vertices v1 , . . . , vn .
Let δi be the ith unit coordinate vector in Rn . For each face F = {vi1 , . . . , vik } ∈ ,
define the simplex σF = conv(δi1 , . . . , δik ). The linear independence of the δi ’s
guarantees that σF is indeed a simplex and that σF ∩ σG = σF ∩G . Hence the set
= {σF : F ∈ } is a geometric realization of , so we could define as the
geometric realization of (unique once we have labelled the vertices v1 , . . . , vn ).
However, for our purposes we don’t need this uniqueness.
12.2 Remark (for those with some knowledge of topology). The geometric
realization || is a topological space X (a topological subspace of some Rd ). We
say that is a triangulation of X.
12.3 Example. Let = 123, 234, 235, 36, 56, 57, 8. A geometric realization
of is shown in Figure 12.1, projected from three dimensions. Note that since
three triangles share the edge 23, any geometric realization in Rd requires d ≥
3. It is a result of Karl Menger, though irrelevant for us, that any d-dimensional
simplicial complex can be realized in R2d+1 , and that this result is best possible, i.e.,
the dimension 2d + 1 cannot in general be decreased. In fact, the simplicial complex
Fig. 12.1 A geometric 4

realization 7
5
2
8
1 3 6
whose facets are all the (d + 1)-element subsets of a (2d + 3)-element set cannot be
realized in R2d . For example, when d = 1 we get that the complete graph K5 cannot
be embedded in the plane (without crossing edges), a famous result in graph theory
known to Euler at least implicitly, since he showed in 1750 that f1 ≤ 3f0 − 6 for
any planar graph (where fi is defined below). The first person to realize explicitly
that K5 is not planar seems to be A. F. Möbius in 1840, who stated the result in the
form of a puzzle.
12.4 Example. Let V = {1, 1̄, 2, 2̄, 3, 3̄} and
= 123, 1̄23, 12̄3, 123̄, 1̄2̄3, 1̄23̄, 12̄3̄, 1̄2̄3̄.
Then the boundary of an octahedron is a geometric realization of . See Figure 12.2.

Thus we can also say, following Remark 12.2, that is a triangulation of the 2-
sphere (two-dimensional sphere).
We now come to the combinatorial information about simplicial complexes that
is our primary interest in this chapter. For i ≥ −1, let fi be the number of i-
dimensional faces of . Thus f−1 = 1 unless = ∅, and f0 = #V , the number of
vertices of . If dim = d − 1, then the vector
f () = (f0 , f1 , . . . , fd−1 )
is called the f -vector of . Thus the simplicial complex of Figure 12.1 has f -
vector (8, 10, 3), while that of Figure 12.2 has f -vector (6, 12, 8).
An important general problem is to characterize the f -vector of various classes
of simplicial complexes. The first class to come to mind is all simplicial complexes.
Fig. 12.2 The boundary of

an octahedron
In other words, what vectors (f0 , f1 , . . . , fd−1 ) of positive integers are f -vectors
of (d − 1)-dimensional simplicial complexes? Although this result is not directly
related to the upcoming connection with commutative algebra, we will discuss it
because of its general interest and its analogy to the upcoming Theorem 12.28.
We first make some strange-looking definitions and then explain their connection
with f -vectors.
12.5 Proposition. Given positive integers n and j , there exist unique integers
nj > nj −1 > · · · > n1 ≥ 0
such that

nj nj −1 n1
n= + + ··· + . (12.1)
j j −1 1
Proof. The proof is based on the following simple combinatorial identity. Let 1 ≤
i ≤ m. Then

m m−1 m−i+1 m+1
+ + ··· + +1= . (12.2)
i i−1 1 i
This identity can easily be proved by induction on i, for instance. It also has a
simple combinatorial interpretation. Namely, the right-hand side is the number of
i-element subsets S of the set [m + 1] = {1, 2, . . . , m + 1}.Thenumber of such
subsets for which the least missing element is s +1 is equal to m−s
i−s . Summing over
all 0 ≤ s ≤ i completes the proof of (12.2).
We now prove the proposition by induction on j . For j = 1 we have n = n1 ,
m
while n = 1 for m = n. Hence the proposition is true for j = 1.
Assume the proposition
for j − 1. Given n, j , define mj to be the largest integer
for which n ≥ mjj . Hence if the proposition is true for n and j , then nj ≤ mj . But
by (12.2),

mj − 1 mj − 2 mj − j mj
+ + ··· + = − 1 < n.
j j −1 1 j
Since the above sum is the largest possible number of the form (12.1) beginning

with mjj−1 , we have nj ≥ mj . Hence nj = mj . By induction there is a unique way
to write

nj nj −1 nj −2 n1
n− = + + ··· + ,
j j −1 j −2 1
where nj −1 > nj −2 > · · · > n1 ≥ 0. Thus we need only check that nj −1 < nj . If
on the contrary nj −1 ≥ nj , then

nj nj −1 nj nj nj + 1
+ ≥ + = ,
j j −1 j j −1 j
contradicting the maximality of nj and completing the proof.

The representation of n in the form (12.1) is called the j -binomial expansion of

n. Given this formula, define

nj nj −1 n1
n(j ) = + + ··· + .
j +1 j 2
In other words, add 1 to the bottom ofall the
binomial
coefficients in the
j-binomial

expansion of n. For instance, 51 = 74 + 53 + 42 + 01 , so 51(4) = 75 + 54 + 43 +
0
2 = 30. For notational simplicity we sometimes suppress the binomial coefficients

equal to 0, e.g., 51 = 74 + 53 + 42 . Note that a binomial coefficient nii = 0 in
the j -binomial expansion of n if and only if ni = i − 1, in which case nr = r − 1
for all 1 ≤ r ≤ i.
We can now state a famous theorem of Schützenberger and Kruskal–Katona,
often called the Kruskal–Katona theorem.
12.6 Theorem. A vector (f0 , f1 , . . . , fd−1 ) ∈ Pd is the f -vector of a ((d − 1)-
dimensional) simplicial complex if and only if
(i+1)
fi+1 ≤ fi , 0 ≤ i ≤ d − 2. (12.3)
As an example, the fact that 51(4) = 30 means that in any simplicial complex
with f3 = 51 we must have f4 ≤ 30, and that this result is best possible.
Theorem 12.6 says qualitatively the intuitively clear result that given fi , the number
fi+1 cannot be too big. However, the precise quantitative result given by this
theorem is by no means intuitively obvious. Let us try to provide some intuition
and at the same time convey some idea of the proof.
Let α = (a1 , . . . , aj ) and β = (b1 , . . . , bj ) be two sequences of nonnegative
integers of the same length j . We say that α is less than β in reverse lexicographic
R
order (or reverse lex order for short), denoted α < β, if for some 0 ≤ i ≤ j − 1
we have
aj = bj , aj −1 = bj −1 , . . . , aj −i+1 = bj −i+1 , and aj −i < bj −i . (12.4)
Equivalently, if we regard the nonnegative integers as the letters of an alphabet in

their usual order, then the reverse sequences (aj , . . . , a1 ) and (bj , . . . , b1 ) are in
dictionary (lexicographic) order. If S and T are two j -element subsets of N, then
R R
we say that S < T if S < T , where A denotes the sequence of elements of the set
A written in increasing order. If we abbreviate a set like {2, 4, 7} as 247, then the
one-element subsets of N in reverse lex order are
R R R R R R R
0 < 1 < 2 < 3 < 4 < 5 < 6 < ··· .
The two-element subsets are

R R R R R R R
01 < 02 < 12 < 03 < 13 < 23 < 04 < · · · .
The three-element subsets are

R R R R R R R
012 < 013 < 023 < 123 < 014 < 024 < 124 < 034
R R R R
< 134 < 234 < 015 < · · · .
The next result explains the connection between the j -binomial expansion and
reverse lex order on j -element subsets of N.
12.7 Theorem. Let S0 , S1 , . . . be the sequence of j -element subsets of N in reverse
lex order. Suppose that Sn = {a1 , . . . , aj } with a1 < · · · < aj . Then

aj aj −1 a1
n= + + ··· + ,
j j −1 1
and this formula gives the j -binomial expansion of n.

Before beginning the proof, here is an example. What is the 1985th term (calling
the first term S0 the 0th term) S1985 of the reverse lex order on four-element subsets
of N? We have

16 11
1985 = + .
4 3
Hence S1985 = {16, 11, 1, 0}.
Proof. It suffices to show that the

number
−1ofj -elementsubsets
of N that are smaller
than Sn in reverse lex order is ajj + ajj−1 + · · · + a11 . We claim that for each
1 ≤ k ≤ n the number of such subsets that agree with Sn in their largest j − k
elements but differ in their (j − k + 1)st largest one is akk . In fact, these are just
the union of the k-element subsets of {0, 1, . . . , ak − 1} with {ak+1 , ak+2 , . . . , aj },
so the proof follows.
Note that this argument assumes nothing about j -binomial expansions. Since
for each n there is just one Sn , the above proof in fact yields a new proof of
Proposition 12.5.
Now suppose that f = (f0 , . . . , fd−1 ) ∈ Pd . Define a collection f of subsets
of N to consist of the empty set ∅ together with the first fi of the (i + 1)-element
subsets of N in reverse lex order. For example, if f = (6, 8, 5, 2), then (writing as
usual {1, 2, 3} = 123, etc.)
f = {∅, 0, 1, 2, 3, 4, 5, 01, 02, 12, 03, 13, 23, 04, 14,
012, 013, 023, 123, 014, 0123, 0124}.
Note that for this example, f is not a simplicial complex.

(i+1)
12.8 Theorem. The set f is a simplicial complex if and only if fi+1 ≤ fi for
0 ≤ i ≤ d − 2.
Proof. Let us use the notation [0, m] = {0, 1, . . . , m} and for any set S,

S
= {T ⊆ S : #T = k}.
k
i+1 ni
Let fi = ni+1 + i + · · · + n11 be the (i + 1)-binomial expansion of fi . By the
definition of reverse lex order, we see that the set X of the first fi (i + 1)-elements
of N in reverse lex order is given by

[0, ni+1 − 1] [0, ni − 1]
X= {ni+1 } ∪
i+1 i

[0, ni−1 − 1]
{ni+1 , ni } ∪ ··· .
i−1
The set of (i + 2)-elements subsets F of N all of whose (i + 1)-element subsets

belong to X is given by

[0, ni+1 − 1] [0, ni − 1]
X= {ni+1 } ∪
i+2 i+1

[0, ni−1 − 1]
{ni+1 , ni } ∪ ··· .
i
(i+1)
These are just the first fi (i + 2)-element subsets of N in reverse lex order, and
the proof follows.
Theorem 12.8 establishes the “if” direction of the Kruskal–Katona theorem

(Theorem 12.6), i.e., condition (12.3) is sufficient for the existence of a simplicial
complex with f -vector f = (f0 , f1 , . . . , fd−1 ). We have in fact constructed a
“canonical” simplicial complex f with this f -vector. Such a simplicial complex
is called compressed.
The difficult part of the Kruskal–Katona theorem is the “only if” direction. We
need to show that every simplicial complex has the same f -vector as some
compressed simplicial complex f . This is proved by transforming to f by
a sequence of steps preserving the simplicial complex property and preserving
the f -vector. It is not necessary to understand this argument (or in fact even the
statement of the Kruskal–Katona theorem) in order to understand the main result of
this chapter (Theorem 12.25) and its proof, so we will omit it here. See the “Notes
for Chapter 13” below for a reference to a readable proof.
12.9 Example. Is f = (5, 7, 5) an f -vector? Of course we could simply check

whether the Kruskal–Katona conditions (12.3) hold. Alternatively, we can construct
f and check whether it is a simplicial complex. In fact,
f = {∅, 0, 1, 2, 3, 4, 01, 02, 12, 03, 13, 23, 04, 012, 013, 023, 123, 014}.
This is not a simplicial complex since 14 is a subset of 014 ∈ f , but 14 ∈ f .

Hence (5, 7, 5) is not an f -vector. In fact, we have 7 = 42 + 11 and 7(2) = 43 +
1
2 = 4 < 5.
We next want to characterize the f -vectors of a certain class of simplicial
complexes, called shellable simplicial complexes. The result will be very similar
to the Kruskal–Katona theorem, but the proof is vastly different. It will use tools
from commutative algebra. First we will define shellable simplicial complexes and
state the characterization of their f -vectors. We will then develop the algebraic tools
necessary for the proof. Finally we will discuss a connection with an analogue of
the Kruskal–Katona theorem.
We say that a simplicial complex is pure if every facet (maximal face) has the
same dimension. For instance, the simplicial complex of Figure 12.1 is not pure;
it has facets of dimensions zero, one and two. A subcomplex of a simplicial
complex is a subset of that is itself a simplicial complex. (We don’t require
that has the same vertex set as .)
12.10 Definition. A (d − 1)-dimensional simplicial complex is shellable if is
pure and there exists an ordering F1 , F2 , . . . , Ft of its facets (so t = fd−1 ) such that
the following property holds. For 0 ≤ j ≤ t let j = F1 , . . . , Fj , the subcomplex
of generated by F1 , . . . , Fj . In particular, 0 = ∅. Now let 1 ≤ j ≤ t. Then
we require that the set of faces of Fj (i.e., the set of all subsets of Fj ) has a unique
minimal element Gj not belonging to j −1 . Call the sequence F1 , . . . , Ft a shelling
order or just shelling of , and call Gj the restriction of Fj (with respect to the
shelling F1 , . . . , Ft ).
NOTE. Let be a pure (d − 1)-dimensional simplicial complex. It is easy to
see (Exercise 2) that a facet ordering F1 , . . . Ft is a shelling if and only if for all
2 ≤ i ≤ t, the subcomplex F1 , . . . , Fi−1 ∩ Fi (i.e., the set of faces of Fi that
already belong to F1 , . . . , Fi−1 ) is a pure simplicial complex of dimension d − 2.
More informally, Fi attaches along some nonempty union of its facets.
It takes some time looking at examples to develop a feeling for shellings. First
note that since 0 = ∅, the empty set is the unique minimal element of F1 not in
0 . Thus we always have G1 = ∅.
12.11 Example. (a) Consider the one-dimensional simplicial complex of Fig-
ure 12.3a. The ordering 1,2,3 of the facets is a shelling order, with G1 = ∅,
G2 = {c} (abbreviated as c), G3 = d. For instance, when we attach facet 2
to facet 1, we create the two new faces c and bc. The unique minimal element
(with respect to inclusion) of the two sets c and bc is G2 = c. Another shelling
order is 2,1,3, with G1 = ∅, G2 = a, G3 = d. In fact, there are exactly four

shelling orders: 123, 213, 231, 312. For instance, 132 is not a shelling order.
When we adjoin 3 to 1 we create the new faces c, d, cd, and now we have two
minimal elements c and d.
(b) One shelling order of the simplicial complex of Figure 12.3b is 1,2,3,4, with
G1 = ∅, G2 = c, G3 = d, G4 = ad.
(c) As essentially explained in (a) above, the simplicial complex c of Figure 12.3c
is not shellable.
(d) The simplicial complex d of Figure 12.3d is also not shellable. Otherwise by
symmetry we can assume 1, 2 is a shelling order. But when we adjoin facet 2
to 1, we introduce the new faces c, d, cd, de, and cde. There are two minimal
new faces: c and d.
There is in fact a close connection between c and d . Given any simplicial
complex with vertex set V , define the cone over , denoted C(), to be the
simplicial complex with vertex set V ∪ {v}, where v is a new vertex not in V ,
and with faces
C() = ∪ {{v} ∪ F : F ∈ }.
Then d = C(c ). Moreover, it is not hard to see (Exercise 7) that a simplicial

complex is shellable if and only if C() is shellable.
12.12 Example. For a somewhat more complicated example of a shellable sim-

plicial complex, let be the simplicial complex realized by the boundary of an
1
a b
1 2 3
4 2
a b c d
d c
3
(a) (b)
a c a c
e
1 2 1 2
b d b d
(c) (d)
Fig. 12.3 Some simplicial complexes
octahedron. Figure 12.4 shows a shelling of . This figure is a projection of the

octahedron into the plane. All eight triangular regions, including the unbounded
outside region with vertices d, e, f represent faces of . The sets Gi of minimal
new faces are as follows:
G1 = ∅, G2 = d, G3 = e, G4 = f
G5 = de, G6 = ef, G7 = df, G8 = def.
12.13 Example. Figure 12.5 shows a more subtle example of a nonshellable

simplicial complex . It has nine triangular facets. There is no “local” obstruction
to shellabilty. That is, we cannot look at just a small part of and conclude that it
is nonshellable. We will explain why is nonshellable in Corollary 12.16 (see the
paragraph after its proof). In general, however, there is no simple way to tell whether
a simplicial complex is shellable.
We now want to discuss a connection between f -vectors and shellability.
Suppose that F1 , . . . , Ft is shelling of a (d − 1)-dimensional simplicial complex
. Let Gj and j −1 have the meaning in Definition 12.10. Suppose that #Gj = m.
Thus some m-element subset S of Fj is the unique minimal face of Fj not belonging

to j −1 . This set S is contained in d−m i (m + i)-element subsets T of Fj , since
#Fj = d. Therefore, knowing the number of elements of Gj tells us exactly how
many new faces of each dimension we have adjoined to at the j th shelling step.
There is an elegant and very useful way of organizing the above information.
Given the f -vector (f0 , f1 , . . . , fd−1 ) of a (d − 1)-dimensional simplicial complex
, define numbers h0 , h1 , . . . , hd by the formula

d
d
fi−1 (x − 1)d−i = hi x d−i , (12.5)
i=0 i=0
Fig. 12.4 A shelling order of d

the octahedron
8
2
5
a b
7
1
4 3
c
6
f e
Fig. 12.5 A nonshellable

simplicial complex
where as usual f−1 = 1 unless = ∅. We call the vector
h() = (h0 , h1 , . . . , hd )
the h-vector of . It is clear from (12.5) that the f -vector and h-vector contain
equivalent information—f () determines h() and vice versa.
12.14 Example.
(a) The f -vector of the simplicial complex of Figure 12.3a is (3, 2). We compute
that
(x − 1)2 + 3(x − 1) + 2 = x 2 + x.
Hence h() = (1, 1, 0).

(b) For Figure 12.3c we have f () = (4, 2) and
(x − 1)2 + 4(x − 1) + 2 = x 2 + 2x − 1.
Hence h() = (1, 2, −1).

(c) For Figure 12.2 (the boundary of an octahedron) we have f () = (6, 12, 8)
and
(x − 1)3 + 6(x − 1)2 + 12(x − 1) + 8 = x 3 + 3x 2 + 3x + 1.
Hence h() = (1, 3, 3, 1).

(d) For a more general example, let be generated by a single d-element face F ,
i.e., consists of all subsets of F . A geometric realization of is a (d − 1)-
dimensional simplex. Now

d d d d
f () = , , ,..., ,
1 2 3 d
and
d
d
(x − 1)d−i = x d ,
i
i=0
by the binomial theorem. Hence h() = (1, 0, 0, . . . , 0).

There are some elementary properties of the h-vector worth noting:
• By taking coefficients of x d on both sides of (12.5), we see that h0 = 1 unless
= ∅.
• Taking coefficients of x d−1 shows that h1 = f0 − d.
• If we set x = 1 in (12.5), then we obtain1
fd−1 = h0 + h1 + · · · + hd . (12.6)
The left-hand side fd−1 is the number of (d − 1)-faces of . It would be nice

if hi were the number of such faces with some property (depending on i).
Example 12.14(b) shows that we can have hi < 0, so in general hi does not
have such a nice combinatorial interpretation. However, for shellable simplicial
complexes hi has a simple interpretation given by Theorem 12.15 below.
• (for readers with some knowledge of topology) Putting x = 0 on both sides of
(12.5) shows that
hd = (−1)d−1 (−f−1 + f0 − f1 + f2 − · · · + (−1)d−1 fd−1 ). (12.7)
If X is any topological space that possesses a finite triangulation, then for any
triangulation , say with f -vector (f0 , . . . , fd−1 ), the alternating sum −f−1 +
f0 − f1 + · · · + (−1)d−1 fd−1 is independent of the triangulation and is known
as the reduced Euler characteristic of X, denoted χ̃ (X). We also write χ̃ ( ) =
χ̃ (X) for any triangulation of X. We say that χ̃( ) is a topological invariant of
since it depends only on the geometric realization | | as a topological space.
Equation (12.7) therefore shows that
hd = (−1)d−1 χ̃( ). (12.8)
Recall also that the ordinary Euler characteristic χ (X) is given by f0 − f1 +

f2 − · · · + (−1)d−1 fd−1 for any triangulation as above. Thus, if = ∅, then
1 Sincewe have (x − 1)0 = 1 in the term indexed by i = d on the left-hand side of (12.5), we
need to interpret 00 = 1 when we set x = 1. Although 00 is an indeterminate form in calculus, in
combinatorics it usually makes sense to set 00 = 1.
χ̃ (X) = χ (X) − 1
since f−1 = 1. Hence the difference between the reduced and ordinary Euler
characteristics depends on whether or not we regard ∅ as a face.
We now come to the relationship between shellings and h-vectors.
12.15 Theorem. Let F1 , . . . , Ft be a shelling of the simplicial complex , with
restrictions G1 , . . . , Gt . Then

d
t
hi x i = x #Gj .
i=0 j =1
In other words, hi is the number of restrictions with i elements (independent of the

choice of shelling).
Proof. We noted after Example 12.13 that when we adjoin a facet Fj to a shelling

with restriction Gj satisfying #Gj = m, then we adjoin d−m new faces with
d i
m+i elements. Hence the contribution to the polynomial i=0 fi−1 (x −1)d−i from
d−m
adjoining Fj (using the symmetry d−mi = d−m−i and the binomial theorem) is
given by

d−m
d −m
d − m
d−m
(x − 1) d−(m+i)
= (x − 1)i
i i
i=0 i=0
= x d−m ,
12.16 Corollary. A necessary condition for a (pure) (d − 1)-dimensional simplicial

complex to be shellable is that hi () ≥ 0 for all 0 ≤ i ≤ d. Moreover, if
triangulates a topological space X, then a necessary condition for shellability is
(−1)d−1 χ̃(X) ≥ 0.
Proof. Assume that is shellable. By Theorem 12.15 we have hi () ≥ 0 for all
0 ≤ i ≤ d. The second assertion then follows from (12.8).
Corollary 12.16 explains why the simplicial complex of Figure 12.5 is not
shellable. We have
h3 () = (−1)2 (−1 + 9 − 18 + 9) = −1.
The geometric realization of is a cylinder (or more accurately, homeomorphic to

a cylinder). Since h3 () = −1, it follows that any triangulation of a cylinder
12.2 The Face Ring 201
X satisfies h3 ( ) = −1 = χ̃ (X). Similarly, the two-dimensional torus T satisfies

(−1)2 χ̃ (T ) = −1, so no triangulation of T can be shellable.
The condition of Corollary 12.16 is necessary but not sufficient for shellability.
For instance, the disjoint union of two cycles (a one-dimensional simplicial
complex) satisfies hd = 1 but isn’t shellable. (See Exercise 28.) For some more
subtle examples, see Exercises 9 and 11.
12.2 The Face Ring
Our goal is a complete characterization of the f -vector of a shellable simplicial

complex, analogous to the characterization of the f -vector of all simplicial com-
plexes given by the Kruskal–Katona theorem (Theorem 12.6). The main tool will be
a certain commutative ring associated with a simplicial complex on the vertex set
V = {x1 , . . . , xn }. To keep the presentation as simple as possible, we will develop
the necessary ring theory to prove the main result of this chapter (Theorem 12.25),
but no more. Most of our definitions, results, and proofs can be extended to a
far greater context. We make a brief remark on one of these generalizations in
Remark 12.26.
Let K be a field. Any infinite field will do for our purposes. Think of the elements
of the vertex set V as indeterminates. Let K[x1 , . . . , xn ] or K[V ] denote the
polynomial ring in the indeterminates x1 , . . . , xn . For any subset S of {x1 , . . . , xn },
write

xS = xi . (12.9)
xi ∈S
Let I denote the ideal of K[V ] generated by all monomials xS such that S ∈ .
We call such a set S a nonface of . If S is a nonface and T ⊃ S, then clearly T
is a nonface. Hence I is generated by the minimal nonfaces of , that is, those
nonfaces for which no proper subset is a nonface. A minimal nonface is also called
a missing face.
12.17 Example. For the simplicial complexes of Figure 12.3 we have the following
minimal generators of I , i.e., the monomials corresponding to missing faces: (a)
ac, ad, bd, (b) ac, bd, (c) ac, ad, bc, bd, (d) ac, ad, bc, bd. Note that (c) and (d)
have the same missing faces. This is because (d) is a cone over (c). The cone vertex
e is attached to every face F of (c) (i.e., {e} ∪ F is a face of (d)), so e belongs to no
missing face.
For Figure 12.1, the missing faces all have two elements except for {3, 5, 6}. For
the octahedron of Figure 12.2, the missing faces are (writing as usual 11 for {1, 1 },
etc.) 11 , 22 , and 33 .
The quotient ring K[] := K[V ]/I is called the face ring (also called the
Stanley–Reisner ring) of . It is the fundamental algebraic object of this chapter.
If face rings are to be useful in characterizing f -vectors, we need to connect

the two together. For this aim, define the support supp(u) of a monomial u =
x1a1 · · · xnan by
supp(u) = {xi : ai > 0}.
Note that a K-basis for the ideal I consists of all monomials u satisfying supp(u) ∈
[why?]. Hence a K-basis for K[] consists of all monomials u satisfying
supp(u) ∈ , including (unless = ∅) the monomial 1, whose support is ∅. (More
precisely, we mean the images of these monomials in K[] under the quotient map
K[V ] → K[], but in such situations we identify elements of K[V ] with their
images in K[].) For i ≥ 0 define K[]i to be the span of all monomials u of
degree i satisfying supp(u) ∈ . Then
K[] = K[]0 ⊕ K[]1 ⊕ · · · (vector space direct sum).
We define the Hilbert series of K[] to be the power series

L(K[], λ) = (dimK K[]i ) λi ,
i≥0
where λ is an indeterminate. Thus L(K[], λ) is some kind of measurement of the

“size” of K[].
12.18 Theorem. If dim = d − 1 and h() = (h0 , h1 , . . . , hd ), then
h 0 + h1 λ + · · · + hd λ d
L(K[], λ) = . (12.10)
(1 − λ)d
Proof. We have seen that a K-basis for K[] consists of monomials whose support
is a face F of . Let MF be the set of monomials with support F . Then
⎛ ⎞

λdeg(u) = ⎝ λ ai ⎠
u∈MF xi ∈F ai ≥1
λ#F
= . (12.11)
(1 − λ)#F
In particular, when F = ∅ the two sides of (12.11) are equal to 1. Summing over all
F ∈ gives
λ#F
L(K[], λ) =
(1 − λ)#F
F ∈

d
λi
= fi−1
(1 − λ)i
i=0

d
fi−1 λi (1 − λ)d−i
i=0
= .
(1 − λ)d
Now

d
d d−i
1
fi−1 λi (1 − λ)d−i = λd fi−1 −1
λ
i=0 i=0

d
= λd hi λ−(d−i) (by (12.5))
i=0

d
= hi λi ,
i=0
The integer d = 1 + dim is called the Krull dimension of K[], denoted

dim K[]. Do not confuse the vector space dimension dimK with the Krull
dimension dim! By (12.6) and (12.10) dim K[] is the order to which 1 is a pole
of L(K[], λ), i.e., the least integer k for which (1 − λ)k L(K[], λ) does not have
a singularity at λ = 1. It is known (but not needed here) that dim K[] is also the
most number of elements of K[] that are algebraically independent over K, and is
also the length of the longest chain p0 ⊂ p1 ⊂ · · · ⊂ p of prime ideals of K[].
There is a special situation in which the hi ’s have a direct algebraic interpretation.
In any commutative ring R, recall that an element u is called a non-zero-divisor (or
NZD) if whenever y ∈ R and uy = 0, then y = 0. Now let θ∈ K[]1 be an NZD.
(Note that θ ∈ K[]1 means that θ is a linear combination ni=1 αi xi (αi ∈ K) of
the vertices x1 , . . . , xn of .) Since θ is an NZD, we have that for i ≥ 0 the map
K[]i → K[]i+1 defined by y → θy is injective (one-to-one). Hence
dimK θ K[]i = dimK K[]i . (12.12)
Let (θ ) denote the ideal of K[] generated by θ . Since θ is homogeneous, the

quotient ring K[]/(θ ) has the vector space grading
K[]/(θ ) = (K[]/(θ ))0 ⊕ (K[]/(θ ))1 ⊕ · · · , (12.13)
where (K[]/(θ ))i is the image of K[]i under the quotient homomorphism
K[] → K[]/(θ ).

If A(λ) = i≥0 ai λ and B(λ) =
i i
i≥0 bi λ are two power series with real
coefficients, write A(λ) ≤ B(λ) to mean ai ≤ bi for all i.
12.19 Lemma. Let θ ∈ K[]1 . Then
L(K[]/(θ ), λ)
L(K[], λ) ≤ , (12.14)
1−λ
with equality if and only if θ in an NZD.
Proof. If θ is an NZD, then by (12.12) we have
H (K[]/(θ ), i + 1) = H (K[], i + 1) − H (K[], i).
Multiplying both sides by λi+1 and summing on i ≥ −1 gives
L(K[]/(θ ), λ) = L(K[], λ) − λL(K[], λ),
so
L(K[]/(θ ), λ)
L(K[], λ) = .
1−λ
If θ is not an NZD, then we always have
dimK θ K[]i ≤ dimK K[]i ,
and for at least one i strict inequality holds. This is easily seen to imply that strict
inequality holds in (12.14).
By iteration of Lemma 12.19 we have the following result.

12.20 Theorem. Let θ1 , . . . , θj ∈ K[]1 . Then
L(K[]/(θ1 , . . . , θj ), λ)
L(K[], λ) ≤ ,
(1 − λ)j
with equality if and only θi is an NZD in the ring K[]/(θ1 , . . . , θi−1 ) for 1 ≤ i ≤
j − 1.
If θ1 , . . . , θj ∈ K[]1 has the property that θi is an NZD in the ring
K[]/(θ1 , . . . , θi−1 ) for 1 ≤ i ≤ j − 1, then we say that θ1 , . . . , θj is a regular
sequence. The number of elements of the largest regular sequence in K[]1 is called
the depth of K[], denoted depth K[]. Let us remark that it can be shown that all
maximal regular sequences have the same number of elements, though we do not
need this fact here.
It is easy to see that a regular sequence θ1 , . . . , θj ∈ K[]1 is algebraically
independent over K (Exercise 22). In other words, there does not exist a polynomial
0 = P (t1 , . . . , tk ) ∈ K[t1 , . . . , tk ] for which P (θ1 , . . . , θk ) = 0 in K[]. Let us

point out that if the sequence θ1 , . . . , θj ∈ K[] is algebraically independent and
moreover each θi is an NZD in K[], then these conditions are not sufficient for
θ1 , . . . , θj to be a regular sequence. For instance, if has vertices a, b, c and the
single edge ab, then a − c and b − c are algebraically independent NZDs. However,
in the ring K[]/(a − c) we have c = 0 but (b − c)c = 0. In fact, we have
depth K[] = 1, e.g., by Exercise 25. For another example, let have vertices
a, b, c and edges ab, bc, and assume that char(K) = 2. Now a + b and a − b are
algebraically independent NZDs but not a regular sequence since in K[]/(a + b)
we have c = 0 and c(a − b) = 0. Unlike the previous example, this time we have
depth K[] = 2. A regular sequence of length two is given by, for instance, a −c, b.
Suppose that θ1 , . . . , θd ∈ K[]1 is a regular sequence, where as usual d =
dim K[] = dim + 1. Let R = K[]/(θ1 , . . . , θd ). Thus R inherits a grading
R = R0 ⊕ R1 ⊕ · · · from K[]. Let h() = (h0 , h1 , . . . , hd ) as usual. By
Theorem 12.18 and the definition of regular sequence we have
L(R, λ) = h0 + h1 λ + · · · + hd λd , (12.15)
apolynomial in λ. Hence R is a finite-dimensional vector space, and dimK R =

hi = fd−1 . Clearly R1 cannot contain an NZD ψ, since, e.g., if u is a nonzero
element of R of maximal degree (which must exist since dimK R < ∞), then ψu =
0. Hence depth K[] = d = dim K[], the maximum possible. This motivates the
following key definition.
12.21 Definition. Assume that K is an infinite field. We say that the simplicial
complex is Cohen–Macaulay (with respect to the field K) and that the ring K[]
is a Cohen–Macaulay ring if dim K[] = depth K[].
NOTE. Note that the above definition assumes that K is infinite. There is a more
algebraic definition of Cohen–Macaulay that coincides with our definition when K
is infinite but not always when K is finite. For our purposes it doesn’t hurt to assume
that K is infinite.
It follows from (12.15) that a Cohen–Macaulay simplicial complex satisfies
hi () ≥ 0. However, two basic problems remain, as follows.
• What simplicial complexes are Cohen–Macaulay?
• What more can be said about the h-vector (or f -vector) of a Cohen–Macaulay
simplicial complex?
The definitive answer to the first question is beyond the scope of this book, but
for the benefit of readers with some knowledge of algebraic topology we discuss the
answer in Remark 12.26. What we will prove is that shellable simplicial complexes
are indeed Cohen–Macaulay. Regarding the second question, we will obtain a
complete characterization of the h-vector of a Cohen–Macaulay simplicial complex,
which will also characterize h-vectors of shellable simplicial complexes. This char-
acterization is a multiset analogue of the Kruskal–Katona theorem (Theorem 12.6).
Let us first consider the second question. A multicomplex on a set V is a

multiset analogue of a simplicial complex whose vertex set is contained in V . More
precisely, is a collection of multisets (sets with repeated elements, as discussed on
page 1), such that every element of is contained in V , and if M ∈ and N ⊆ M,
then N ∈ . We will assume from now on that the underlying set V is finite. For
example (writing 112 for {1, 1, 2}, etc.), = {∅, 1, 2, 3, 11, 12, 112, 1112} is not a
multicomplex, since 1112 ∈ and 111 ⊆ 1112, but 111 ∈ .
If is a multicomplex with ei elements of size i, then we call the sequence
e( ) = (e0 , e1 , . . . ) the e-vector of . Any integer vector (e0 , e1 , . . . ) which is the
e-vector of some multicomplex is called an e-vector. Our e-vectors are also called
M-vectors after F. S. Macaulay and O-sequences, where O stands for “order ideal
of monomials,” defined below.
NOTE ON TERMINOLOGY. It might seem more natural to let fi be the number of
elements of of size i + 1, and define (f0 , f1 , . . . ) to be the f -vector of in
complete analogy with simplicial complexes. Historically, the indexing of f -vectors
is explained by fi being the number of faces of dimension (rather than cardinality)
i. For multicomplexes, we have no need for the concept of the dimension of a face
F (and if we did, the “best” definition would be that dim F is one less than the
number of distinct elements of F ). Counting elements of multicomplexes by their
cardinality is more natural for almost all purposes. In the literature our ei is often
replaced with hi , and our e-vector is called an h-vector. This is because e-vectors
of multicomplexes do sometimes coincide with h-vectors of simplicial complexes
(e.g., Theorem 12.25). Moreover, e-vectors of multicomplexes coincide with the
sequence of Hilbert function values of standard graded K-algebras (not defined here,
though K[] and its quotients considered here are special cases), so one can think
that h stands for “Hilbert.” To avoid any possible confusion we will use the new
notation ei and terminology e-vector.
We now discuss a “multiplicative equivalent” of multicomplexes. If u and v are
monomials in the variables x1 , . . . , xn , we say that u divides v (written u | v) if
there is a monomial w for which uw = v. Equivalently, if u = x1a1 · · · xnan and
v = x1b1 · · · xnbn , then u | v if and only if ai ≤ bi for all i. An order ideal of monomials
in the variables x1 , . . . , xn is a collection o of monomials in these variables such
that if v ∈ o and u | v, then u ∈ o. Equivalently, associate with the monomial
u = x1a1 · · · xnan the multiset Mu = {1a1 , . . . , nan } (i.e., i has multiplicity ai ). Then
a set M of monomials is an order ideal of monomials if and only if the collection
{Mu : u ∈ M} is a multicomplex.
While we need the next result only for quotients of face rings by a regular
sequence, it involves no extra work to prove it in much greater generality. For this
purpose, we define a homogeneous ideal of the polynomial ring K[x1 , . . . , xn ] to be
an ideal I generated by homogeneous polynomials.
12.22 Theorem. Let I be a homogeneous ideal of K[x1 , . . . , xn ], and let P =
K[x1 , . . . , xn ]/I . Then P has a K-basis that is an order ideal of monomials in the
variables x1 , . . . , xn .
Proof. We define reverse lex order on monomials of some fixed degree m as

follows: define
R
x1a1 · · · xnan < x1b1 · · · xnbn ,

where ai = bi , if
R
(1a1 , . . . , nan ) < (1b1 , . . . , nbn ),
R
where i j denotes a sequence i’s of length j , and < is defined by 12.4. For instance,
the reverse lex order on monomials of degree three in the variables x1 , x2 , x3 is
R R R R R R R R R
x13 < x12 x2 < x1 x22 < x23 < x12 x3 < x1 x2 x3 < x22 x3 < x1 x32 < x2 x32 < x33 .
For each m ≥ 0 let Pm denote the span (over the field K) of the homogeneous
polynomials in P of degree m. Because I is generated by homogeneous polynomials
we have the vector space direct sum
P = P0 ⊕ P1 ⊕ P2 ⊕ · · · ,
a direct generalization of (12.13).

For each degree m, let Bm be the least K-basis for Pm in reverse lex order. In
other words, first choose the least monomial u1 of degree m in reverse lex order that
is nonzero in P . Then choose the least monomial u2 of degree m in reverse lex order
such that {u1 , u2 } are linearly independent, etc. We eventually obtain a K-basis for
Pm by this process since every linearly independent subset of a vector space can be
extended to a basis.
We claim that the set B0 ∪ B1 ∪ B2 ∪ · · · is a basis for P which is an order ideal
of monomials. Suppose not. Let v ∈ Bj , u | v, but u ∈ Bi where i = deg u. Then u
R
is a linear combination of monomials w < u, say

u= αw w.
R
w<u
R v R
Multiply both sides by v
u. It is easy to see that if w < u then w · u < v. Thus
wv R
we have expressed v as a linear combination of monomials u < v, contradicting
v ∈ Bj .
12.23 Corollary. Let be a Cohen–Macaulay simplicial complex. Then the h-

vector of is an e-vector.
Proof. Let θ1 , . . . , θd be a regular sequence in K[]1 , and let
R = R0 ⊕ R1 ⊕ · · · = K[]/(θ1 , . . . , θd ).
According to the definition of a Cohen–Macaulay simplicial complex and (12.15)

we have hi () = dimK Ri . Thus if Di is any K-basis for Ri , then #Di = hi . By
Theorem 12.22 there is a K-basis Bi for each i such that B0 ∪B1 ∪· · ·∪Bd is an order
ideal of monomials. Thus if has e-vector (e0 , e1 , . . . ), then ei = #Bi = hi , and
the proof follows.
The next step is to find some simplicial complexes to which we can apply
Corollary 12.23. The following 0 theorem is the primary algebraic result of this
chapter. Recall the notation xS = xi ∈S xi of (12.9).
12.24 Theorem. If is a shellable simplicial complex on the vertex set V =
{x1 , . . . , xn }, then the face ring K[] is Cohen–Macaulay for any infinite field
K. Moreover, if F1 , . . . , Ft is a shelling of with restrictions G1 , . . . , Gt , then
xG1 , . . . , xGt is a K-basis for R = K[]/(θ1 , . . . , θd ) for any regular sequence
θ1 , . . . , θd ∈ K[]1 , and such a regular sequence always exists.
Proof. Let B = {xG1 , . . . , xGt }. Let θ1 , . . . , θd ∈ K[]1 satisfy the following

property.
(P) The restriction of θ1 , . . . , θd to any facet (or face) F spans the K-vector space
KF with basis F . In other words, if we define
ψi = θi |xj =0 if xj ∈F ,
then ψ1 , . . . , ψd span KF .
Note that K being infinite guarantees that there is enough “room” to find such
θ1 , . . . , θd . This is the reason why we require K to be infinite. The argument below
will show in particular that if θ1 , . . . , θd ∈ K[]1 satisfies Property (P), then
θ1 , . . . , θd is a regular sequence.
Now let R = K[]/(θ1 , . . . , θd ). By Theorems 12.15 and 12.20 (in the case
j = d) it follows that if B spans R (as a vector space over K) then θ1 , . . . , θd is
regular, and B is a K-basis for R. Thus we need to show that B spans R.
The proof is by induction on t.
First assume that t = 1. Then is just a simplex and K[] = K[x1 , . . . , xd ].
Moreover, any K-basis θ1 , . . . , θd is a regular sequence and K[]/(θ1 , . . . , θd ) =
K. The Hilbert series of the field K is just 1. Finally, if F is the unique facet of ,
then F (regarded as a one-term sequence) is a shelling of with G1 = ∅. Since
x∅ = 1 is a basis for K, the theorem is true for t = 1.
Now assume the theorem for t − 1, and let F1 , . . . , Ft be a shelling of .
Claim. xi xGt = 0 in R for all 1 ≤ i ≤ n.
Case 1. Suppose that xi ∈ Ft . By definition of shelling the new faces F obtained
by adjoining Ft to the shelling are given by Gt ⊆ F ⊆ Ft . Thus {xi } ∪ Gt cannot
be a new face, so {xi } ∪ Gt ∈ . Hence xi xGt = 0 in K[], so also in R.
Case 2. Suppose that xi ∈ Ft . Set
K[Ft ] = K[]/(xj : xj ∈ Ft ) = K[xj : xj ∈ Ft ],

a polynomial ring in the vertices of Ft . By property (P), the restrictions ψ1 , . . . , ψd

of θ1 , . . . , θd to F span the space KF . Hence there exists a linear combination of
θ1 , . . . , θd of the form

η = xi + αj xj , αj ∈ K.
xj ∈Ft
Then in the ring R we have
xi xGt = (xi − η)xGt (since η = 0 in R)

⎛ ⎞

= −⎝ αj xj ⎠ xGt
xj ∈Ft
= 0 (by Case 1).
This completes the proof of the claim.

Now let R = R/(xGt ) and t−1 = F1 , . . . , Ft−1 . By definition of Gt we have
K[t−1 ] = K[]/(xGt ).
Condition (P) still holds for K[t−1 ] (since the facets of t−1 are also facets of ).
Moreover,
R = K[t−1 ]/(θ1 , . . . , θd ).
By the induction hypothesis, xG1 , . . . , xGt−1 span R . By the claim, the ideal
(xGt ) of R is a vector space of dimension at most one. Hence xG1 , . . . , xGt span R,
and the proof follows for any sequence (necessarily regular by the argument above)
θ1 , . . . , θd satisfying (P).
It remains to show that every regular sequence θ1 , . . . , θd ∈ K[]1 satisfies (P).
This result is an easy exercise; a somewhat more general result is given by the “only
if” part of Exercise 24.
NOTE. Note the structure of the previous proof. We pick θ1 , . . . , θd ∈ K[]1

satisfying Property (P). Let F1 , . . . , Ft be a shelling of , and set R =
K[]/(θ1 , . . . , θd ). As we successively quotient R by the monomials xG1 , . . . , xGt ,
the vector space dimension drops by at most one, and we end up with the ring 0.
Hence dimK R ≤ t = fd−1 (). On the other hand, by Theorems 12.18 and 12.20
we have dimK R ≥ hi = t. Hence xG1 , . . . , xGt must be a K-basis for R.
We are finally ready for the main theorem of this chapter.
12.25 Theorem. Let h = (h0 , h1 , . . . , hd ) be a sequence of integers. The following
three conditions are equivalent.
(a) There exists a (d − 1)-dimensional Cohen–Macaulay simplicial complex (over

any infinite field) with h() = h.
(b) There exists a (d − 1)-dimensional shellable simplicial complex with
h() = h.
(c) The sequence h is an e-vector.
Proof. (b)⇒(a) Immediate from Theorem 12.24.

(a)⇒(c) This is Corollary 12.23.
(c)⇒(b) Given the e-vector h, we need to construct a shellable simplicial
complex whose h-vector is h. We will identify a (finite) multiset M of positive
integers with the increasing sequence of its elements. Given 0 ≤ i ≤ d, let
α1 , α2 , . . . , αhi be the first hi terms of the reverse lex order on i-element multisets
of positive integers. For instance, when i = 3 and h3 = 8, we have
(α1 , . . . , α8 ) = (111, 112, 122, 222, 113, 123, 223, 133). (12.16)
If αj = a1 a2 · · · ai , define
βj = 1, 2, 3, . . . , d − i, a1 + d − i + 1, a2 + d − i + 2, . . . , ai + d. (12.17)
For the example of (12.16) and d = 5 we have
(β1 , . . . , β8 ) = (12456, 12457, 12467, 12567, 12458, 12468, 12568, 12478).
Now let σ be the concatenation β1 , β2 , . . . , βd of the sequences β1 , . . . , βd . For

instance, if h = (1, 4, 2, 1) (so d = 3), then we get (where we separate the different
βj ’s with semicolons, and we write in boldface the terms a1 + d − i + 1, a2 + d −
i + 2, . . . , ai + d of each βj )
σ = (123; 124, 125, 126, 127; 134, 135; 234). (12.18)
We leave as an exercise (Exercise 29) to show that σ is a shelling of a (d − 1)-

dimensional simplicial complex on the vertex set {1, 2, . . . , h1 + d}. Moreover,
if we write σ = (F1 , . . . , Fm ) (where m = hi = fd−1 ()) and if Fk is given by
the sequence on the right-hand side of (12.17), then the restriction Gk is given by
Gk = {a1 + d − i + 1, a2 + d − i + 2, . . . , ai + d}. This being the case, exactly hi
restrictions Gk have i elements, so indeed h() = h.
As an example, the sequence σ of (12.18) is a shelling (F1 , . . . , F8 ) of a

simplicial complex with vertices 1, . . . , 7. The restrictions G1 , . . . , G8 are given
by ∅, 4, 5, 6, 7, 34, 35, 234 (the elements in boldface).
12.26 Remark. We mentioned earlier that the complete characterization of Cohen–
Macaulay simplicial complexes is beyond the scope of this book. For readers
familiar with some algebraic topology (only the rudiments of simplicial homology
are needed), we will state without proof the theorem of Gerald Reisner that provides
this characterization. For this purpose, if F ∈ , then define the link of F , denoted
lk (F ), by
lk (F ) = {G ∈ : F ∩ G = ∅, F ∪ G ∈ }.
It is clear that lk (F ) is a subcomplex of . In particular, lk (∅) = . For any

simplicial complex we write H i (; K) for the ith reduced homology group of
over the field K.
12.27 Theorem. Let K be an infinite field. The following two conditions on a
simplicial complex are equivalent.
• is Cohen–Macaulay with respect to K.
i (lk (F ); K) = 0 for all
• For every F ∈ (including F = ∅), we have H
i = dim lk (F ).
It can be shown from Reisner’s theorem that for fixed K (or actually, for fixed
characteristic of K), Cohen–Macaulayness is a topological property, i.e., it depends
only on the geometric realization of (as a topological space). For instance, all
triangulations of spheres and balls (of any dimension) are Cohen–Macaulay over
any (infinite) field. A triangulation of the real projective plane is Cohen–Macaulay
with respect to K if and only if char(K) = 2.
Theorem 12.25 gives an elegant characterization of h-vectors (and hence f -
vectors) of shellable simplicial complexes, but one ingredient is still missing—a
“nice” characterization of e-vectors. Since an e-vector is a multiset analogue of an
f -vector of a simplicial complex, it is not unreasonable to expect a characterization
of e-vectors similar to the Kruskal–Katona theorem (Theorem 12.6) for ordinary
f -vectors. We conclude this chapter by discussing such a characterization.
Given positive integers n and j , let

nj nj −1 n1
n= + + ··· +
j j −1 1
be the j -binomial expansion of n (equation (12.1)). Now define

nj + 1 nj −1 + 1 n1 + 1
nj = + + ··· + .
j +1 j 2
Thus instead of adding 1 to the bottom of each binomial coefficient as we did when
we defined n(j ) , now we add 1 to the bottom and top. We now have the following
exact analogue of the Kruskal–Katona theorem.
12.28 Theorem. A vector (e0 , e1 , . . . , ed ) ∈ Pd+1 is an e-vector if and only if
e0 = 1 and
i
ei+1 ≤ ei , 0 ≤ i ≤ d − 1. (12.19)
The proof is analogous to that of the Kruskal–Katona theorem. Namely, we

identify a finite multiset M on N with the increasing sequence of its elements.
For instance, the multiset {0, 0, 2, 3, 3, 3} becomes the sequence 002333, and the
sequence of 3-element multisets on N in reverse lex order begins
000 001 011 111 002 012 112 022 122 222 003 · · · .
It can easily be checked that if a1 a2 · · · aj is the nth term (beginning with term
0) in the reverse lex ordering of j -element multisets on N, then a1 , a2 + 1, a3 +
2, . . . , aj + j − 1 is the nth term in the reverse lex ordering of j -element subsets of
N. Hence Theorem 12.7 applies equally well to multisets on N.
We next have the following multiset analogue of Theorem 12.8. The proof is
completely analogous to that of Theorem 12.8.
12.29 Theorem. Given e = (e0 , e1 , . . . , ed ) ∈ Pd with e0 = 1, let e consist of
the union over all i ≥ 1, together with ∅, of the first ei of the i-element multisets on
i
N in reverse lex order. The set e is a multicomplex if and only if ei+1 ≤ ei for
1 ≤ i ≤ d − 1.
Theorem 12.29 proves the “if” direction of Theorem 12.28. The proof of the
“only if” direction is similar to that of the Kruskal–Katona theorem. A multicomplex
e for some e-vector e is called compressed. Given any multicomplex , we
transform it by a sequence of simple operations into a compressed multicomplex,
at all steps preserving the e-vector. We omit the details, which are somewhat more
complicated than in the simplicial complex case.
12.30 Example. Is (e0 , e1 , e2 , e3 ) = (1, 4, 5, 7) an e-vector? The first ei multisets
on N in reverse lex order for 1 ≤ i ≤ 3 are given by
0 1 2 3
00 01 11 02 12 .
000 001 011 111 002 012 112
These multisets (together with ∅) form a multicomplex, so (1, 4, 5, 7) is an e-vector.

On the other hand, (1, 4, 5, 8) is not an e-vector. We need to add the
multiset
022,
but 22 does not appear. We can also check this by writing 5 = 32 + 21 . Then

52 = 43 + 32 = 7, so if in an e-vector we have e2 = 5 then e3 ≤ 7.
The embedding theorem of Menger discussed in Example 12.3 appears (in much
greater generality) in Menger [93]. The statement that the simplicial complex whose
facets are all (d + 1)-element subsets of a (2d + 3)-element set cannot be realized
in R2d is due to A. Flores and E. R. van Kampen. For a modern treatment, see
Matoušek [91].
The Kruskal–Katona theorem (Theorem 12.6) was first stated by M.-P. Schützen-
berger in a rather obscure journal [117]. The first published proofs were by Kruskal
[79] and later independently by Katona [72]. A nice survey of this area is given by
Greene and Kleitman [56], including a good presentation of a proof of the Kruskal–
Katona theorem due to Clements and Lindström [25].
The first indication of a connection between commutative algebra and combi-
natorial properties of simplicial complexes appears in a paper of Melvin Hochster
[67]. The face ring of a simplicial complex first appeared in the Ph.D. thesis of
Gerald Reisner (published version in [110]), which was supervised by Hochster,
and independently in two papers of Stanley [121, 122]. For an exposition of the
connections between combinatorics and commutative algebra, see Stanley [124].
The concept of shelling goes back to nineteenth century geometers, but perhaps
the first substantial result on shellings is due to Bruggesser and Mani [15]. The
characterization of h-vectors of shellable simplicial complexes (Theorem 12.25) is
a special case of a result of Stanley [121]. Our proof here is based on that of Kind
and Kleinschmidt [74].
The characterization of e-vectors (Theorem 12.28) is due to Macaulay [87], who
gave a very complicated proof as part of his characterization of Hilbert series of
graded algebras. It is interesting that Macaulay’s theorem preceded the Kruskal–
Katona theorem, though the latter is somewhat easier to prove. Simpler proofs
of Macaulay’s theorem were later given by Sperner [120], Whipple [144], and
Clements and Lindström [25], among others.
Cohen–Macaulay rings are named after Cohen [26] and Macaulay [86], who
were interested in them primarily because of their connection with “unmixedness”
theorems. For a modern treatment, see the text of W. Bruns and J. Herzog [19].
1. A simplicial complex has 2,528 three-dimensional faces. Show that it has at

most 6,454 four-dimensional faces, and that this result is best possible. (“Best
possible” means that there exists some with f3 = 2,528 and f4 = 6,454.)
2. Prove the assertion of the Note following Definition 12.10. That is, let be a
pure (d − 1)-dimensional simplicial complex. Then a facet ordering F1 , . . . Ft
is a shelling if and only if for all 2 ≤ i ≤ t, the subcomplex F1 , . . . , Fi1 ∩Fi
is a pure simplicial complex of dimension d − 2.
3. (a) Find the number of shellings of a path of length n, i.e., the simplicial
complex with n + 1 vertices and n edges forming a path.
(b) Find the number of shellings of a cycle of length n.
4. Find explicitly every simplicial complex with the property that every ordering
of its facets is a shelling.
5. Suppose that F1 , F2 , . . . , Ft is a shelling of a simplicial complex . Is it always
the case that the reverse order Ft , Ft−1 , . . . , F1 is also a shelling of ?
6. Show that if a simplicial complex is shellable and F ∈ , then the link

lk (F ) (as defined in Remark 12.26) is also shellable.
7. Prove the assertion of Example 12.11(d) that a simplicial complex is
shellable if and only if the cone C() is shellable.
8. A matroid complex is a simplicial complex on the vertex set V such that for
any W ⊆ V , the restriction W of to W , i.e.,
W = {F ∈ : F ⊆ W },
is pure.
(a) Let V = {x1 , . . . , xn } be a set of distinct nonzero vectors in some vector
space over a field. Define
= {F ⊆ V : F is linearly independent}.
Clearly is a simplicial complex on V . Show that is a matroid complex.

(b) Show that a matroid complex is shellable.
9. (*) (for those who know a little topology) Let X be the topological space
obtained by identifying the three sides (edges) of a solid triangle as shown
in Figure 12.6. (The edges are identified in the direction of the arrows.) This
space is called the topological dunce hat. Show that if is a simplicial complex
whose geometric realization is homeomorphic to then is not shellable.
10. (*) Give an example of a shellable simplicial complex with more than one
facet, such that there is a unique facet F that comes last in every shelling.
11. (a) Show that every triangulation of a two-dimensional ball X (i.e., the
geometric realization of is homeomorphic to X) is shellable.
(b) (very difficult) Find a triangulation of a three-dimensional ball that is not
shellable.
(c) (even more difficult) Find a triangulation of a three-dimensional sphere that
is not shellable.
12. A partial shelling of a pure (d − 1)-dimensional complex is a sequence
F1 , . . . , Fr of some subset of the facets such that this sequence is a shelling
order for the simplicial complex F1 , . . . , Fr which they generate. Clearly
if F1 , . . . , Ft is a shelling of , then F1 , . . . , Fr is a partial shelling for all
1 ≤ r ≤ t.
(a) (*) Give an example of a shellable simplicial complex that has a partial
shelling that cannot be extended to a shelling.
(b) Let n,k be the simplicial complex on an n-element vertex set V whose
facets consist of all k-element subsets of V . Show that n,k is shellable.
(c) (unsolved) Can every partial shelling of n,k be extended to a shelling?
13. Let n be the largest integer, if it exists, with the following property: if is any
shellable simplicial complex with at most n facets, then for any facet F there
exists a shelling beginning with F . If no such n exists, then set n = ∞.
Fig. 12.6 The topological

dunce hat
(a) (easy) Show that n ≥ 2.

(b) (*) Show that n ≥ 9.
(c) (*) Show that n < ∞.
14. Let (f0 , f1 , . . . , fd−1 ) be the f -vector of a (d − 1)-dimensional simplicial
complex . We will illustrate a certain procedure with the example (6, 12, 8)
(the f -vector of an octahedron). Write down the numbers f0 , f1 , . . . , fd−1 on
a diagonal, and put 1 to the left of f0 :
1 6
12
8
Think of the 1 as being preceded by a string of 0’s. Turn the array into
a difference table by writing below each pair of consecutive numbers their
difference:
1 6
1 5 12
1 4 7 8
Now write down one further row of differences:
1 6
1 5 12
1 4 7 8
1 3 3 1
Show that this bottom row is the h-vector of .

15. Find the f -vector and h-vector of the simplicial complex whose geometric
realization is the boundary of an icosahedron.
16. (a) Let d be the simplicial complex on the vertex set V = {x1 , . . . , xd ,
y1 , . . . , yd } whose faces are those subsets of V that do not contain both
xi and yi , for any 1 ≤ i ≤ d. Compute the h-vector of d .
CULTURAL NOTE. Let δi be the ith unit coordinate vector in Rd . A

geometric realization of d consists of the boundary of the convex hull Cd
of the vectors ±δi , 1 ≤ i ≤ d. The polytope C is called the d-dimensional
cross-polytope and is a d-dimensional generalization of an octahedron, the
case d = 3.
(b) Show that d is shellable.
17. Give an example of two simplicial complexes 1 and 2 such that the
geometric realizations of 1 and 2 are homeomorphic, the h-vector of 1
is nonnegative, and some hi (2 ) < 0. What is the smallest possible dimension
of 1 and 2 ?
18. (difficult from first principles) (*) Let be a triangulation of a (d − 1)-
dimensional sphere, and let h() = (h0 , h1 , . . . , hd ). Show that hi = hd−i for
0 ≤ i ≤ d. This result is called the Dehn–Sommerville equations for spheres.
19. Let h = (h0 , h1 , . . . , hd ) and k = (k0 , k1 , . . . , kd ) be e-vectors. Define
h ∧ k = (min{h0 , k0 }, min{h1 , k1 }, . . . , min{hd , kd })

h ∨ k = (max{h0 , k0 }, max{h1 , k1 }, . . . , max{hd , kd }).
Show that h ∧ k and h ∨ k are e-vectors.

20. (*) Let be the simplicial complex of Figure 12.7. Each triangle, including
the “outer triangle” with vertices a, b, c is a face, so f () = (6, 12, 8) and
triangulates a 2-sphere. Find the minimal nonfaces (or missing faces) of .
21. (*) Suppose that and are simplicial complexes whose face rings K[ ] and
K[] are isomorphic as K-algebras (or even as rings). Show that and are
isomorphic.
22. Show that a regular sequence θ1 , . . . , θj ∈ K[]1 is algebraically independent
over K.
Fig. 12.7 A triangulation of a

the 2-sphere
e f
b c
23. (a) Let θ1 , . . . , θj ∈ K[]1 be a regular sequence. Show that any permutation
of this sequence is also a regular sequence.
(b) Show that each θi is an NZD in K[].
24. Let be any (d − 1)-dimensional simplicial complex, and let θ1 , . . . , θd ∈
K[]1 . Show that the quotient ring R = K[]/(θ1 , . . . , θd ) is a finite-
dimensional vector space over K if and only if θ1 , . . . , θd satisfy Property (P).
25. Show that the face ring K[] of a simplicial complex has depth one if and
only if is disconnected. Deduce that a disconnected simplicial complex of
dimension at least one is not Cohen–Macaulay.
26. Let and be simplicial complexes on disjoint vertex sets V and W ,
respectively. Define the join ∗ to be the simplicial complex on the vertex
set V ∪ W with faces F ∪ G, where F ∈ and G ∈ . (If consists of a single
point, then ∗ is the cone over . If consists of two disjoint points, then
∗ is the suspension of .)
(a) Compute the h-vector h( ∗ ) in terms of h( ) and h().
(b) Show that if and are Cohen–Macaulay, then so is ∗ .
(c) Generalizing Exercise 7, show that if and are shellable, then so is ∗.
27. Let be a (d − 1)-dimensional Cohen–Macaulay simplicial complex with h-
vector (h0 , h1 , . . . , hd ). Suppose that hi = 1 for some 1 ≤ i ≤ d − 1. What
are the possible values of hd ?
28. Let be a one-dimensional simplicial complex. Show that the following three
conditions are equivalent: (a) is Cohen–Macaulay, (b) is shellable, and (c)
is connected.
29. Complete the proof of Theorem 12.25 by showing that the sequence σ is a
shelling of with the stated restrictions Gk .
30. (*) Let be a four-dimensional shellable simplicial complex with f0 = 13,
f1 = 67, and f2 = 204. What is the most number of facets that can have?
What if we drop the hypothesis of shellability?
31. (*) Let be a (d − 1)-dimensional Cohen–Macaulay simplicial complex with
h-vector (h0 , h1 , . . . , hd ). Let be a (d − 1)-dimensional Cohen–Macaulay
subcomplex of with h-vector (h0 , h1 , . . . , hd ). Show that hi ≤ hi for all
0 ≤ i ≤ d.
32. (*) Let be a (d − 1)-dimensional simplicial complex on the vertex set
V . We say that is balanced if we can write V as a disjoint union V =
· 2 ∪· · · · ∪V
V1 ∪V · d such that for every F ∈ and every 1 ≤ i ≤ d we have
#(F ∩ Vi ) ≤ 1. In particular, if is pure, then #(F ∩ Vi ) = 1 when F is a facet.
(Sometimes is required to be pure in the definition of balanced.) Suppose
that (h0 , h1 , . . . , hd ) is the h-vector of a Cohen–Macaulay balanced simplicial
complex . Show that (h1 , h2 , . . . , hd ) is the f -vector of a balanced simplicial
complex. You may want to use the result of Exercise 24.
Chapter 13
Miscellaneous Gems of Algebraic
Combinatorics
13.1 The 100 Prisoners
An evil warden is in charge of 100 prisoners (all with different names). He puts a
row of 100 boxes in a room. Inside each box is the name of a different prisoner. The
prisoners enter the room one at a time. Each prisoner must open 50 of the boxes, one
at a time. If any of the prisoners does not see his or her own name, then they are all
killed. The prisoners may have a discussion before the first prisoner enters the room
with the boxes, but after that there is no further communication. A prisoner may
not leave a message of any kind for another prisoner. In particular, all the boxes are
shut once a prisoner leaves the room. If all the prisoners choose 50 boxes at random,
then each has a success probability of 1/2, so the probability that they are not killed
is 2−100 , not such good odds. Is there a strategy that will increase the chances of
success? What is the best strategy?
It’s not hard to see that the prisoners can achieve a success probability of greater
than 2−100 . For instance, suppose that the first prisoner opens the first 50 boxes and
the second prisoner opens the last 50. If the first prisoner succeeds (with probability
1/2), then the first prisoner’s name is guaranteed not to be in one of the boxes opened
by the second prisoner, so the second prisoner’s probability of success is 50/99. Each
pair of prisoners can do this strategy, increasing the overall success probability to
(25/99)50 , still an extremely low number. Can they do significantly better? The key
to understanding this problem is the realization that the prisoners do not have to
decide in advance on which boxes they will open. A prisoner can decide which box
to open next based on what he has seen in the boxes previously opened.
13.1 Theorem. There exists a strategy with a success probability of

100
1
1− = 0.3118278207 · · · .
j
j =51

https://doi.org/10.1007/978-3-319-77173-1_13
220 13 Miscellaneous Gems of Algebraic Combinatorics
Proof. The prisoners assign themselves the numbers 1, 2, . . . , 100 by whatever

method they prefer. Each prisoner is assigned a different number. The prisoners
memorize everyone’s number. They regard the boxes, which are lined up in a row,
as being numbered 1, 2, . . . , 100 from left to right. A prisoner with number k first
goes to box k. If the prisoner sees his name, then he breathes a temporary sigh of
relief, and the next prisoner enters. Otherwise the first prisoner will see the name
of some other prisoner, say with number n1 . He then opens box n1 and repeats the
procedure, so whenever he opens a box B that doesn’t contain his own name, the
next box that he opens has the number of the prisoner whose name appears in box B.
What is the probability of success of this strategy? Suppose that box i contains
the name of the prisoner numbered π(i). Thus π is a permutation of 1, 2, . . . , 100.
The boxes opened by prisoner i are those containing the names of prisoners with
numbers π(i), π 2 (i), π 3 (i), etc. If k is the length of the cycle of π containing i,
then the prisoner will see his name after opening the kth box. This will happen
whenever k ≤ 50. Thus all prisoners see their names if and only if every cycle of
π has length at most 50. If π does not have this property, then it has exactly one
cycle of length r > 50. There are 100 r ways to choose the elements of the cycle and
(r − 1)! ways to arrange them in a cycle. There are then (100 − r)! ways to arrange
the other elements of π . Thus the number of permutations π ∈ S100 with a cycle of
length r > 50 is

100 100!
(r − 1)!(100 − r)! = .
r r
(There are more clever ways to see this.) Hence the probability of success, i.e., the
probability that π has no cycle of length more than 50, is
1 100!
100 100
1
1− =1− ,
100! r r
r=51 r=51
as claimed.
If we apply the above argument to 2n prisoners rather than 100, then we get a
success probability of

2n
1
2n
1 1
n
1− =1− + .
r r r
r=n+1 r=1 r=1
From calculus we know that there is a constant γ = 0.577215665 · · · , known as

Euler’s constant, for which
n
1
lim − log n = γ .
n→∞ r
r=1
13.2 Oddtown 221
Thus if there are n prisoners, then it follows that as n → ∞, the success probability
of the prisoners is
lim (1 − log 2n + log n) = 1 − log 2 = 0.3068528194 · · · .

n→∞
It seems quite amazing that no matter how many prisoners there are, they can always
achieve a success probability of over 30%!
NOTE. It can be shown that the above strategy is in fact optimal, i.e., no strategy
achieves a higher probability of success. The proof, however, is not so easy.
13.2 Oddtown
The village of Oddtown has a population of n people. Inhabitants of Oddtown like

to form clubs. Every club has an odd number of members, and every pair of clubs
share an even number of members (possibly none).
13.2 Theorem. There are at most n clubs.
Proof. Let k be the number of clubs. Define a matrix M = (Mij ) over the two-
element field F2 as follows. The rows of M are indexed by the clubs Ci and the
columns by the inhabitants xj of Oddtown. Set
$
1, xj ∈ Ci
Mij =
0, otherwise.
The matrix M is called the incidence matrix corresponding to the clubs and their
members.
In general, let S be a subset of [n], and let χS ∈ Zn be the characteristic vector
of S, i.e., χS = (a1 , . . . , an ) where

1, i ∈ S
ai =
0, i ∈ S.
If T is another subset of [n], then the key observation is that the scalar (dot) product
of χS and χT is given by χS · χT = #(S ∩ T ). Hence if we now work over F2 , then

1, if #(S ∩ T ) is odd
χS · χ T = (13.1)
0, if #(S ∩ T ) is even.
Let A = MM t , a k × k matrix. By (13.1) and the assumption that every club

has an odd number of members, we see that main diagonal elements of A are 1.
Similarly the off-diagonal elements of A are 0, so A = Ik , the k × k identity matrix.
Hence rank(A) = k.
Recall that if B is a k × m matrix and C is an m × n matrix (over some field),

then rank(BC) ≤ rank(B) (as well as rank(BC) ≤ rank(C)), since for any matrix
D, rank(D) = dim image(D). Hence, since M has n columns,
n ≥ rank(M) ≥ rank(MM t ) = rank(A) = k.

While Theorem 13.2 can be proved without linear algebra, the proof is not easy.
13.3 Complete Bipartite Partitions of Kn
Figure 13.1 shows the six edges of the complete graph K4 partitioned (according to
the edge label) into the edge sets of the three complete bipartite graphs K3,1 , K2,1 ,
and K1,1 . Clearly we can extend this construction, achieving a partition of the edges
E(Kn ) of Kn into the edge sets of n − 1 complete bipartite graphs. Specifically,
let E1 be the set of edges incident to a fixed vertex v. Thus E1 is the edge set of a
complete bipartite graph Kn−1,1 . Remove E1 from E(Kn ) and proceed by induction,
obtaining a partition of E(Kn ) into the edges of Kn−1,1 , Kn−2,1 , . . . , K1,1 .
The question thus arises as to whether E(Kn ) can be partitioned into fewer than
n − 1 edge sets of complete bipartite graphs.
13.3 Theorem. If E(Kn ) is the disjoint union of the edge sets of m complete
bipartite graphs, then m ≥ n − 1.
·
Proof. Take the vertex set of Kn to be [n]. Let E(Kn ) = E(B1 )∪E(B · · · · ∪·
1 )∪
E(Bm ) (disjoint union), where Bk is a complete bipartite graph with vertex
bipartition (Xk , Yk ) (so Xk ∩ Yk = ∅). For 1 ≤ k ≤ n, define an n × n matrix
Ak by

1, i ∈ Xk , j ∈ Yk
(Ak )ij =
0, otherwise.

All nonzero rows of Ak are equal, so rank Ak = 1. Let S = m k=1 Ak . For i = j ,
exactly one of the 2m numbers (Ak )ij and (Ak )j i , 1 ≤ k ≤ m, is equal to 1, since
Fig. 13.1 A decomposition

of the edges of K4 into three
complete bipartite graphs
3
1 2
1 2
1
13.4 The Nonuniform Fisher Inequality 223
every edge ij of Kn appears in one E(Bk ) with either i ∈ Xk and j ∈ Yk , or else

j ∈ Xk and i ∈ Yk . Hence
S + S t = J − I,
where as usual J is the n × n all 1’s matrix, and I is the n × n identity matrix.
Claim. If T is any real matrix satisfying T + T t = J − I , then rank T ≥ n − 1.
Suppose to the contrary that rank T ≤ n − 2. Then T has (at least) two linearly
independent eigenvectors x, y such that T x = T y = 0 [why?]. Since J has rank
one, the space x, y spanned by x and y contains a nonzero vector z satisfying
J z = 0 [why?]. Then from T + T t = J − I and T z = 0 we get −z = T t z. Take
the dot product with zt on the left. We get
−|z|2 = zt T t z
= (zt T t z)t (since a 1 × 1 matrix is symmetric)
= zt T z (since in general (AB)t = B t At )
= 0 (since T z = 0),
contradicting z = 0. Hence the claim is proved, so in particular rank S ≥ n

− 1. But
in general rank(A + B) ≤ rank A + rank B [why?]. Therefore from S = m k=1 Ak
and rank Ak = 1 we get rank S ≤ m. It follows that m ≥ n − 1, completing the
proof.
13.4 The Nonuniform Fisher Inequality
A balanced incomplete block design (BIBD) with parameters (v, k, λ, r, b) is a v-

element set X and a collection A of k-element subsets (blocks), with #A = b, such
that any two points x, y ∈ X lie in exactly λ blocks, and each point is in exactly r
blocks. We also assume that k < v, which is the reason for the word “incomplete.”
We can draw a BIBD as a bipartite graph with vertex bipartition (X, A). There is an
edge from x ∈ X to A ∈ A if x ∈ A. Thus the degree of each vertex x ∈ X is r,
and the degree of each vertex A ∈ A is k. It follows that vr = kb (the total number
of edges of the graph). We can also count the number of two-element sets of edges
of A. On the one hand, since each vertex in A
that are incident to the same vertex
has degree k this number is b k2 . On the other hand, each pair of points in X are

mutually adjacent to λ points in A, so we get λ v2 = b k2 . A little manipulation
v k
shows that the two equalities vr = kb and λ 2 = b 2 are equivalent to
vr = kb, λ(v − 1) = r(k − 1),
the usual form in which they are written.

R.A. Fisher showed in 1940 that b ≥ v. This inequality was generalized by

R.C. Bose in 1949. The most convenient way to state Bose’s inequalities, known as
the nonuniform Fisher inequality, is to reverse the roles of points and blocks. Thus
consider the elements x of X to be sets whose elements are the blocks A ∈ A that
contain them. In other words, we have a collection C1 , . . . , Cv of r-element sets
whose union contains b points x1 , . . . , xb . Each point is in exactly k of the sets.
Finally, #(Ci ∩ Cj ) = λ for all i = j .
13.4 Theorem. Let C1 , . . . , Cv be distinct subsets of a b-element set X such that
for all i = j we have #(Ci ∩ Cj ) = λ for some 1 ≤ λ < b (independent of i and j ).
Then v ≤ b.
Proof. Case 1. Some #Ci = λ. Then all other Cj ’s contain Ci and are disjoint
otherwise, so
1 +
v ≤ 3456 b−λ ≤ b.
3 45 6
from Ci from all Cj =Ci
Case 2. All #Ci > λ. Let γi = #Ci − λ > 0. Let M be the incidence matrix of the
set system C1 , . . . , Cv , i.e., the rows of M correspond to the Ci ’s and the columns
to the elements x1 , . . . , xb of X, with

1, xj ∈ Ci
M ij =
0, xj ∈ Ci .
Let A = MM t . The hypotheses imply that A = λJ + G, where J as usual is the

all 1’s matrix (of size v), and G is the diagonal matrix diag(γ1 , . . . , γv ).
Claim. rank(A) = v (i.e., A is invertible). We would then have
v = rank(A) ≤ rank(M) ≤ b,
the last inequality because M has b columns.

As in the proof of Theorem 4.7, a real symmetric matrix B is positive semidefi-
nite if it has nonnegative eigenvalues. Equivalently, by basic linear algebra, uBut ≥
0 for all row vectors u of length v. Moreover B is positive definite (and so has
positive eigenvalues) if uBut > 0 for all u = 0.
Now we easily compute that
u(λJ + G)ut = λ(u1 + · · · + uv )2 + γ1 u21 + · · · + γv u2v > 0
for all u = 0. Thus A = λJ + G is positive definite and hence of full rank v.

13.5 Odd Neighborhood Covers 225
Fig. 13.2 The 3 × 4 grid

graph
13.5 Odd Neighborhood Covers
Consider an m × n grid graph. The case m = 3, n = 4 is shown in Figure 13.2. At

each vertex are a turned on light bulb and also a switch that changes the state of its
bulb and those of its neighbors (adjacent vertices). Can all the lights be turned off?
This problem was open for many years until in 1989 K. Sutner, then a graduate
student, showed using automata theory that the answer is yes for any (finite) graph!
More explicitly, let G be a finite graph with a turned on light bulb at each vertex.
At each vertex is a switch that changes the state of that vertex and all its neighbors.
Then it is possible to turn off all the lights. We will give a modification of a simpler
proof due to Y. Caro based on linear algebra.
Without loss of generality we may assume that G is simple. If v ∈ V (G), then
the neighborhood N (v) of v is the set consisting of v and all vertices adjacent to v.
A little thought shows that we need to prove the following result.
13.5 Theorem. There exists a subset S ⊆ V = V (G) such that #(S ∩ N (v)) is odd
for all v ∈ V . (It follows that switching at the vertices v ∈ S turns all the lights off.)
Proof. Let V (G) = {v1 , . . . , vp }. Let A be the adjacency matrix of G over the field
p
F2 , and let y = (1, 1, . . . , 1) ∈ F2 . Write row(A + I ) for the row space of the
matrix A + I , and let γv denote the row of A + I indexed by v ∈ V . Note that
switching at S turns all the lights off if and only if v∈S γv = y. Hence we need to
show that y ∈ row(A + I ) [why?].
Let us recall from linear algebra some standard facts about orthogonal subspaces.
Let K be a field, and for u, v ∈ K n let u · v be the usual dot product (2.1) of u and v,
so u · v ∈ K. If W is a subspace of K n , then define the orthogonal subspace W ⊥ by
W ⊥ = {u ∈ K n : u · v = 0 for all v ∈ W }.
(In Chapter 11 we discussed the case K = R.) Let d = dim W . Since W ⊥ is the
set of solutions to d linearly independent homogeneous linear equations [why?], we
have
dim W + dim W ⊥ = n. (13.2)

Note that by definition of ⊥ we have W ⊆ (W ⊥ )⊥ . By (13.2) and the equation

obtained from it by replacing W with W ⊥ , we get dim W = dim (W ⊥ )⊥ . Hence
(W ⊥ )⊥ = W. (13.3)
NOTE. Though irrelevant here, let us point out that if K ⊆ R then W ∩ W ⊥ =

{0}, but that this fact need not hold in characteristic p = 0. Over C we should define
u · v = u1 v̄1 + · · · + un v̄n , where ¯ denotes complex conjugation, in order to get the
most sensible theory.
Now by (13.3) the vector y = (1, 1, . . . , 1) (or any vector in Fn2 ) lies in the
row space of A + I if and only if it is orthogonal to every vector in row(A + I )⊥ =
ker(A+I ). Thus we need to show that if (A+I )v t = 0, then v ·y = 0. Equivalently,
if yv t = 0 then (A+I )v t = 0. Note that (a) yv t = 0 means that v has an odd number
of 1’s, and (b) (A + I )v t is the sum of the rows of A + I indexed by the positions
of the 1’s in v. Thus we need to show that A + I does not have an odd number of
rows summing to 0.
Suppose that v1 , . . . , vk are vertices indexing rows of A + I summing to 0. Let
H be the subgraph induced by v1 , . . . , vk , i.e., H consists of the vertices v1 , . . . , vk
and all edges of G between two of these vertices. Let bij be the (i, j )-entry of A+I .

Since ki=1 bij = 0 for 1 ≤ j ≤ n, and each bii = 1, it follows that every vertex of
H has odd degree. Since [why?]

deg v = 2 · #E(H ),
v∈V (H )
we have that k = #V (H ) is even, completing the proof.
13.6 Circulant Hadamard Matrices
For our next “gem of algebraic combinatorics,” we will provide some variety by
leaving the realm of linear algebra and looking at some simple algebraic number
theory.
An n × n matrix H is a Hadamard matrix if its entries are ±1 and its rows are
orthogonal. Equivalently, its entries are ±1 and H H t = nI . In particular [why?],
det H = ±nn/2 . (13.4)
It is easy to see that if H is an n × n Hadamard matrix then n = 1, n = 2, or n = 4m

for some integer m ≥ 1. (See Exercise 12.19.) It is conjectured that the converse is
true, i.e., for every such n there exists an n × n Hadamard matrix.
An n × n matrix A = (bij ) is a circulant or circulant matrix if it has the form
bij = ai−j for some a0 , a1 , . . . , an−1 , where the subscript i − j is taken modulo n.
13.6 Circulant Hadamard Matrices 227
For instance,
⎡ ⎤
a b c d
⎢d a b c⎥
A=⎢
⎣c
⎥
d a b⎦
b c d a
is a circulant. Let A = (ai−j ) be an n × n circulant, and let ζ = e2π i/n , a primitive

nth root of unity. It is straightforward to compute that for 0 ≤ j < n the column
vector [1, ζ j , ζ 2j , . . . , ζ (n−1)j ]t is an eigenvector of A with eigenvalue a0 + ζ j a1 +
ζ 2j a2 + · · · + ζ (n−1)j an−1 . Hence

n−1
det A = (a0 + ζ j a1 + ζ 2j a2 + · · · + ζ (n−1)j an−1 ). (13.5)
j =0
Note that the matrix

⎡ ⎤
−1 1 1 1
⎢ 1 −1 1 1 ⎥
⎢ ⎥
⎣ 1 1 −1 1 ⎦
1 1 1 −1
is both a Hadamard matrix and a circulant.

Conjecture. Let H be an n × n circulant Hadamard matrix. Then n = 1 or n = 4.
The first significant work on this conjecture is due to R.J. Turyn. He showed that
there does not exist a circulant Hadamard matrix of order 8m, and he also excluded
certain other orders of the form 4(2m + 1). Turyn’s proofs use the machinery of
algebraic number theory. Here we will give a proof for the special case n = 2k ,
k ≥ 3, where the algebraic number theory can be “dumbed down” to elementary
commutative algebra and field theory. (Only in Theorem 13.14 do we use a little
Galois theory, which can be avoided with a bit more work.) It would be interesting
to find similar proofs for other values of n.
13.6 Theorem. There does not exist a circulant Hadamard matrix H of order 2k ,
k ≥ 3.
NOTE. It is curious that the numbers 2k (k ≥ 3) are the easiest multiples of 4
to show are not the orders of circulant Hadamard matrices, while on the other hand
the numbers 2k (k ≥ 1) are the easiest numbers to show are the orders of Hadamard
matrices. To see that 2k is the order of a Hadamard matrix H , first note that the case
k = 1 is trivial. It is routine to show that if H1 is a Hadamard matrix of order a and
H2 is a Hadamard matrix of order b, then the tensor (or Kronecker) product A ⊗ B
is a Hadamard matrix of order ab. It follows that there exists a Hadamard matrix of
order 2k , k ≥ 1.
k
From now on we assume n = 2k and ζ = e2π i/2 . Clearly ζ is a zero of the
k−1
polynomial pk (x) = x 2 + 1. We will be working in the ring Z[ζ ], the smallest
subring of C containing Z and ζ . Write Q(ζ ) for the quotient field of Z[ζ ], i.e., the
field obtained by adjoining ζ to Q.
13.7 Lemma. The polynomial pk (x) is irreducible over Q.
Proof. If pk (x) is reducible then so is pk (x + 1). A standard fact about polynomial

factorization is Gauss’ lemma, namely, an integral polynomial that factors over Q
also factors over Z. If p(x), q(x) ∈ Z[x], write p(x) ≡ q(x) (mod 2) to mean that
the coefficients of p(x) − q(x) are even. Now [why?]
k−1 k−1
pk (x + 1) ≡ (x + 1)2 + 1 ≡ x2 (mod 2).
Hence any factorization of pk (x + 1) over Z into two factors of degree at least one
has the form pk (x + 1) = (x r + 2a)(x s + 2b), where r + s = 2k−1 and a, b are
polynomials of degrees less than r and s, respectively. Hence the constant term of
pk (x + 1) is divisible by 4, a contradiction.
It follows by elementary field theory that every element u ∈ Z[ζ ] can be uniquely
written in the form
u = b0 + b1 ζ + b2 ζ 2 + · · · + bn/2−1 ζ n/2−1 , bi ∈ Z.
The basis for our proof of Theorem 13.6 is the two different ways to compute
det H given by (13.4) and (13.5), yielding the formula

n−1
k−1
(a0 + ζ j a1 + ζ 2j a2 + · · · + ζ (n−1)j an−1 ) = ±nn/2 = ±2k2 . (13.6)
j =0
k−1
Thus we have a factorization in Z[ζ ] of 2k2 . Algebraic number theory is
concerned with factorization of algebraic integers (and ideals) in algebraic number
fields, so we have a vast amount of machinery available to show that no factor-
ization (13.6) is possible (under the assumption that each aj = ±1). Compare
Kummer’s famous approach toward Fermat’s Last Theorem (which led to his
creation
0 of algebraic number theory), in which he considered the equation x n +y n =
z as τ n =1 (x + τy) = z when n is odd.
n n
We are continuing to assume that H = (aj −i ) is an n × n circulant Hadamard

matrix. We will denote the eigenvalues of H by
γj = a0 + a1 ζ j + a2 ζ 2j + · · · + an−1 ζ (n−1)j , 0 ≤ j ≤ n − 1.
13.8 Lemma. For 0 ≤ j ≤ n − 1 we have

√
|γj | = n.
Thus √
all the factors appearing on the left-hand side of (13.6) have absolute
value n.
First proof (naive). Let Hi denote the ith row of H , let · denote the usual dot
product, and let ¯ denote complex conjugation. Then
γj γ̄j = (a0 + a1 ζ j + · · · + an−1 ζ (n−1)j )(a0 + a1 ζ −j + · · · + an−1 ζ −(n−1)j )

= H1 · H1 + (H1 · H2 )ζ j + (H1 · H3 )ζ 2j + · · · + (H1 · Hn )ζ (n−1)j .
By the Hadamard property we have H1 · H1 = n, while H1 · Hk = 0 for 2 ≤ k ≤ n,

Second proof (algebraic). The matrix √1n H is a real orthogonal matrix. By linear
algebra, all its eigenvalues
√ have absolute value 1. Hence all eigenvalues γj of H
have absolute value n.
13.9 Lemma. We have
2 = (1 − ζ )n/2 u, (13.7)
where u is a unit in Z[ζ ].
Proof. Put x = 1 in

n−1
x n/2 + 1 = (x − ζ j )
j =0
j odd
0
to get 2 = j (1 − ζ
j ). Since
1 − ζ j = (1 − ζ )(1 + ζ + · · · + ζ j −1 ),
it suffices to show that 1+ζ +· · ·+ζ j −1 is a unit when j is odd. Let j j¯ ≡ 1 (mod n).
Note that j¯ exists since j and n are relatively prime. Then
1−ζ
(1 + ζ + · · · + ζ j −1 )−1 =
1 − ζj
¯
1 − (ζ j )j
=
1 − ζj
¯
= 1 + ζ j + ζ 2j + · · · + ζ (j −1)j ∈ Z[ζ ],
as desired.
13.10 Lemma. We have Z[ζ ]/(1 − ζ ) ∼

= F2 .
Proof. Let R = Z[ζ ]/(1 − ζ ). The integer 2 is not a unit in Z[ζ ], e.g., because 1/2
is not an algebraic integer (the zero of a monic polynomial f (x) ∈ Z[x]). Thus by
Lemma 13.9, 1 − ζ is also not a unit. Hence R = 0 (where 0 is short for {0}).
For all j we have ζ j = 1 in R since ζ = 1 in R. Hence all elements of R can be
written as ordinary integers m. But 0 = 2 in R by Lemma 13.9, so the only elements
of R are 0 and 1.
13.11 Lemma. For all 0 ≤ j ≤ n − 1 there is an integer hj ≥ 0 such that
a0 + a1 ζ j + a2 ζ 2j + · · · + an−1 ζ (n−1)j = vj (1 − ζ )hj ,
where vj is a unit in Z[ζ ].

Proof. Since 2 is a multiple of 1 − ζ by Lemma 13.9, we have by (13.6) that

n−1
(a0 + a1 ζ j + a2 ζ 2j + · · · + an−1 ζ (n−1)j ) = 0
j =0
in Z[ζ ]/(1 − ζ ). Since Z[ζ ]/(1 − ζ ) is an integral domain by Lemma 13.10, some
factor a0 + a1 ζ j + · · · + an−1 ζ (n−1)j is divisible by 1 − ζ . Divide this factor and the
right-hand side of (13.6) by 1 − ζ , and iterate the procedure. We continue to divide a
factor of the left-hand side and the right-hand side by 1 − ζ until the right-hand side
becomes the unit u. Hence each factor of the original product has the form v(1−ζ )h ,
where v is a unit.
13.12 Corollary. Either γ0 /γ1 ∈ Z[ζ ] or γ1 /γ0 ∈ Z[ζ ]. (In fact, both γ0 /γ1 ∈ Z[ζ ]
and γ1 /γ0 ∈ Z[ζ ], as will soon become apparent, but we don’t need this fact here.)
Proof. By the previous lemma, each γj has the form vj (1 − ζ )hj . If h0 ≥ h1 then
γ0 /γ1 ∈ Z[ζ ]; otherwise γ1 /γ0 ∈ Z[ζ ].
We now need to appeal to a result of Kronecker on elements of Z[ζ ] of absolute
value one. For completeness we include a proof of this result, beginning with a
lemma. Recall that if θ is an algebraic number (the zero of an irreducible polynomial
f (x) ∈ Q[x]), then a conjugate of θ is any zero of f (x).
13.13 Lemma. Let θ be an algebraic integer such that θ and all its conjugates have
absolute value one. Then θ is a root of unity.
Proof. Suppose the contrary. Let deg θ = d, i.e., [Q(θ ) : Q] := dimQ Q(θ ) =
d. Now θ , θ 2 , θ 3 , . . . are all distinct and hence infinitely many of them have the
property that no two are conjugate. Each θ j ∈ Z[θ ] and so is the root of a monic
integral polynomial of degree at most d, since the set of algebraic integers forms a
ring. If θ1 , θ2 , . . . , θd are the conjugates of θ , then all the conjugates of θ j are among
j j j
θ1 , θ2 , . . . , θd . Hence each θ j satisfies the hypothesis that all its conjugates have
absolute value 1 (and θ j is an algebraic integer). Thus

the rth elementary symmetric
function er in θ j and its conjugates has at most dr terms, each of absolute value

1, so |er | ≤ dr . Moreover, er ∈ Z since θ j is an algebraic integer. It follows
that there are only finitely many possible polynomials that can be an irreducible
monic polynomial with a zero equal to some θ j , contradicting the fact that there are
infinitely many θ j ’s for which no two are conjugate.
13.14 Theorem (Kronecker). Let τ be any root of unity and α ∈ Z[τ ] with |α| = 1.
Then α is a root of unity.
Proof. Since α ∈ Z[τ ], we see that α is an algebraic integer. We use the basic fact
from Galois theory that the Galois group of the extension field Q(τ )/Q is abelian.
Let β be a conjugate of α, so β = w(α) for some automorphism w of Q(τ ). Apply
w to the equation α ᾱ = 1. Since complex conjugation is an automorphism of Q(τ )
it commutes with w, so we obtain β β̄ = 1. Hence all the conjugates of α have
absolute value one, so α is a root of unity by the previous lemma.
For our next result, we need the standard algebraic fact that if τ = e2π i/m ,
a primitive mth root of unity, then [Q(τ ) : Q] = φ(m) (the Euler φ-function).
Equivalently, the unique monic polynomial m (x) whose zeros are the primitive
mth roots of unity is irreducible. This polynomial is by definition given by

m (x) = (x − τ j )
1≤j ≤m
gcd(j,m)=1
and is called a cyclotomic polynomial. Lemma 13.7 is the case m = n (= 2k ).

13.15 Lemma. If τ ∈ Z[ζ ] is a root of unity, then τ = ζ r for some r ∈ Z.
Proof. Suppose not. It is easy to see that then either τ is a primitive 2m th root of
unity for some m > k, or else τ s is a primitive pth root of unity for some odd prime
p and some s ≥ 1. In the former case
[Q(τ ) : Q] = φ(2m ) = 2m−1 > 2k−1 = φ(2k ) = [Q(ζ ) : Q],
a contradiction. In the latter case, τ s ζ is a primitive pn-th root of unity, so
[Q(τ s ζ ) : Q] = φ(pn) = (p − 1)φ(n) > φ(n) = [Q(ζ ) : Q],
again a contradiction.
We now have all the ingredients to complete the proof of Theorem 13.6. Note
that we have yet to use the hypothesis that ai = ±1. By Lemma 13.8 we have
|γ1 /γ0 | = |γ0 /γ1 | = 1.

Hence by Corollary 13.12, Theorem 13.14 and Lemma 13.15 we have γ0 = ζ −r γ1

for some r. Expand γ0 and ζ −r γ1 uniquely as integer linear combinations of
n
1, ζ, ζ 2 , . . . , ζ 2 −1 :
√
γ0 = a0 + a1 + · · · + an−1 = ± n
ζ −r γ1 = ζ −r ((a0 − an/2 ) + (a1 − an/2+1 )ζ + · · · )

= (ar − an/2+r ) + (ar+1 − an/2+r+1 )ζ + · · · .
√
Equating coefficients of ζ 0 yields ± n = ar − an/2+r . Since each ai = ±1, we
must have n ≤ 4, completing the proof.
13.7 P -Recursive Functions
A function f : N → C is called polynomially recursive, or P -recursive for short, if

there exist polynomials P0 (n), . . . , Pd (n) ∈ C[n], with Pd (n) = 0, such that
Pd (n)f (n + d) + Pd−1 (n)f (n + d − 1) + · · · + P0 (n)f (n) = 0 (13.8)
for all n ≥ 0.
For instance, the Fibonacci sequence Fn is P -recursive since Fn+2 − Fn+1 −
Fn = 0 for all n ≥ 0. Here d = 2 and P2 (n) = 1, P1 (n) = P0 (n) = −1.
This situation is quite special since the polynomials Pi (n) are constants. Another
P -recursive function is f (n) = n!, since f (n + 1) − (n + 1)f (n) = 0 for all n ≥ 0.
Let P denote the set of all P -recursive functions f : N → C. Our goal in this
section is to prove that P is a C-algebra, which amounts to showing that for any
f, g ∈ P and α, β ∈ C, we have
αf + βg ∈ P, fg ∈ P.
There is one technical problem that needs to be dealt with before proceeding to the
proof. We would like to conclude from (13.8) that
1
f (n + d) = − (Pd−1 (n)f (n + d − 1) + · · · + P0 (n)f (n)). (13.9)
Pd (n)
This formula, however, is problematical when Pd (n) = 0. This can happen only for
finitely many n, so (13.9) is valid for n sufficiently large. Thus we want to deal with
functions f (n) only for n sufficiently large. To this end, define f ∼ g if f (n) =
g(n) for all but finitely many n. Clearly ∼ is an equivalence relation; the equivalence
classes are called germs at ∞ of functions f : N → C. The germ containing f is
denoted [f ]. Write G for the set of all germs.
13.7 P -Recursive Functions 233
13.16 Lemma. (a) If f is P -recursive and f ∼ g, then g is P -recursive. In

other words, the property of P -recursiveness is compatible with the equivalence
relation ∼.
(b) Write CN for the complex vector space of all functions f : N → C. Let α, β ∈ C
and f1 , f2 , g1 , g2 ∈ CN . If f1 ∼ f2 and g1 ∼ g2 , then αf1 + βg1 ∼ αf2 + βg2
and f1 g1 ∼ f2 g2 . In other words, linear combinations and multiplication are
compatible with the equivalence relation ∼. Thus the set G has the structure of
a C-algebra, i.e., a complex vector space and a ring (with obvious compatibility
properties such as (αf )g = f (αg) = α(fg)).
Proof. (a) Suppose that f (n) = g(n) for all 0 n0 > n0 . Let (13.8) be the recurrence
satisfied by f . Multiply both sides by nj =0 (n − j ). We then get a recurrence
relation satisfied by g. Hence g is P -recursive.
(b) This is clear.
Let C[n] denote the ring of complex polynomials in n. Let C(n) denote the
quotient field of C[n], i.e., the field of all rational functions P (n)/Q(n), where
P , Q ∈ C[n]. Suppose that f ∈ CN and R ∈ C(n). Then f (n)R(n) is defined
for n sufficiently large (i.e., when the denominator of R(n) is nonzero). Thus we
can define the germ [f (n)R(n)] ∈ G to be the germ of any function that agrees
with f (n)R(n) for n sufficiently large. It is easy to see that this definition of scalar
multiplication makes G into a vector space over the field C(n). We now come to the
key characterization of P -recursive functions (or their germs).
13.17 Lemma. A function f ∈ CN is P -recursive if and only if the vector space
Vf over C(n) spanned by the germs [f (n)], [f (n + 1)], [f (n + 2)], . . . is finite-
dimensional.
Proof. Suppose that f (n) satisfies (13.8). Let Vf be the vector space over C(n)
spanned by [f (n)], [f (n + 1)], [f (n + 2)], . . . , [f (n + d − 1)], so dimC(n) Vf ≤ d.
Equation (13.9) shows that [f (n + d)] ∈ Vf . Substitute n + 1 for n in (13.9). We get
that [f (n+d+1)] is in the span (over C(n)) of [f (n+1)], [f (n+2)], . . . , [f (n+d)].
Since these d germs are all in Vf , we get that [f (n + d + 1)] ∈ Vf . Continuing
in this way, we get by induction on k that f (n + d + k) ∈ Vf for all k ≥ 0, so
Vf = Vf . Thus Vf is finite-dimensional.
Conversely, assume that dimC(n) Vf < ∞. Then for some d, the germs [f (n)],
[f (n + 1)], . . . , [f (n + d)] are linearly dependent over C(n). Write down this linear
dependence relation and clear denominators to get a recurrence (13.8) satisfied by
f . Hence f is P -recursive.
We now have all the ingredients necessary for the main result of this section.
13.18 Theorem. Let f, g ∈ P and α, β ∈ C. Then:
(a) αf + βg ∈ P
(b) f g ∈ P.
Proof. (a) By Lemma 13.17 it suffices to show that dim Vαf +βg < ∞. Now
by definition, the sum Vf + Vg is the vector space consisting of all linear
combinations γ [u] + δ[v], where [u] ∈ Vf and [v] ∈ Vg and γ , δ ∈ C(n).
In particular, Vf + Vg contains all the germs α[f (n + k)] + β[g(n + k)] =
[αf (n + k) + βg(n + k)], k ≥ 0. Hence
Vαf +βg ⊆ Vf + Vg .
Now if V and W are subspaces of some vector space, then V + W is spanned

by the union of a basis for V and basis for W . In particular, if V and W are
finite-dimensional, then dim(V + W ) ≤ dim V + dim W . Hence
dim Vαf +βg ≤ dim(Vf + Vg ) ≤ dim Vf + dim Vg < ∞,
(b) The proof is analogous to (a), except that instead of the sum V + W we need
the tensor product V ⊗K W over the field K. Recall from linear algebra that
V ⊗K W may be thought of (somewhat naively) as the vector space spanned by
all symbols v ⊗ w, where v ∈ V and w ∈ W , subject to the conditions
(v1 + v2 ) ⊗ w = v1 ⊗ w + v2 ⊗ w
v ⊗ (w1 + w2 ) = v ⊗ w1 + v ⊗ w2
αv ⊗ w = v ⊗ αw = α(v ⊗ w),
where α is a scalar. A standard and simple consequence is that if V has the

basis {v1 , . . . , vm } and W has the basis {w1 , . . . , wn }, then V ⊗K W has the
basis vi ⊗ wj , for 1 ≤ i ≤ m and 1 ≤ j ≤ n. In particular,
dim(V ⊗K W ) = (dim V )(dim W ).
Recall the basic “universality” property of the tensor product V ⊗W = V ⊗K W :

there is a bilinear map : V × W → V ⊗ W such that for any vector space Y
and bilinear map : V × W → Y , there is a unique linear map ϕ : V ⊗ W →
Y for which = ϕ. In particular, there is a unique linear transformation
ϕ : Vf ⊗C(n) Vg → G satisfying
ϕ
[f (n + i)] ⊗ g[(n + j )] → [f (n + i)g(n + j )].
The image of ϕ contains all germs [f (n + i)g(n + i)], so Vfg ⊆ image(ϕ).

Thus
dim Vfg ≤ dim(Vf ⊗C(n) Vg ) = (dim Vf )(dim Vg ) < ∞,

13.8 Affine Monoids 235
13.8 Affine Monoids
In this section we give a simple application of commutative algebra to a topic on the

interface of algebra (but not commutative algebra per se) and combinatorics. It is
not difficult to prove our main result (Theorem 13.21) without using commutative
algebra, but it is nonetheless of interest to see the connection with algebra. In
Chapter 12 we gave a more substantial application of commutative algebra which
has no known more elementary proof.
Recall that a semigroup is a set with an associative binary operation. A monoid
is a semigroup with an identity element. For example, the set
Nd = {(a1 , . . . , ad ) : ai ∈ N},
with the usual operation of componentwise addition, is a monoid. The identity

element is (0, 0, . . . , 0). This monoid is in fact commutative since α + β = β + α
for all α, β ∈ Nd . We will be concerned here with submonoids M of Nd , that is,
subsets of Nd which are closed under addition and which contain (0, 0, . . . , 0).
A monoid M is finitely generated if there exists a finite subset G of M such
that every element of M is a (finite) sum (allowing any number of repetitions) of
elements of G. Thus Nd is finitely generated, since we can take G to be the set of
d unit coordinate vectors (vectors whose components are all 0 except for one 1).
However, it is not true in general that every submonoid of Nd is finitely generated.
For instance, let
M = {(i, j ) ∈ N2 : i > 0} ∪ {(0, 0)}. (13.10)
Then a subset G of M generates M if and only if G contains the infinitely many

elements (1, j ), j ≥ 0. Let us mention that, on the other hand, every submonoid of
N is finitely generated (Exercise 13.30).
We will be concerned here with special submonoids of Nd . First note that the set
of all solutions (a1 , . . . , ad ) in nonnegative integers to a system of homogeneous
linear equations with integer coefficients in the unknowns x1 , . . . , xd is a submonoid
of Nd .
13.19 Example. (a) Let M be the monoid of all solutions (a1 , a2 , a3 , a4 ) ∈ N4 to
the equations
x1 + x2 − x3 − x4 = 0
x1 − 2x4 = 0.
The reader should try to show that M is generated by (2, 0, 1, 1) and (0, 1, 1, 0).
(b) Let M be the monoid of all solutions (a1 , . . . , a6 ) ∈ N6 of the equation
x1 + x2 + x3 − x4 − x5 − x6 = 0.
Can the reader see why the smallest set of generators for M has nine elements?
(c) Let M be the monoid of all solutions (a1 , a2 , a3 ) ∈ N3 of the equations
2x1 + x2 − x3 = 0
−x1 + x2 + 2x3 = 0.
Then M = {(0, 0, 0)}. A quick way to see this is to add the two equations:
x1 + 2x2 + x3 = 0.
In complete analogy to the concept of isomorphism of groups, rings, etc., define
two monoids M and N to be isomorphic if there is a bijection f : M → N such that
f (uv) = f (u)f (v) for all u, v ∈ M (where we have written the monoid operation
multiplicatively). Naturally the map f is called an isomorphism.
Define a submonoid M of Nd to be linear if it is the set of solutions (a1 , . . . , ad )
in nonnegative integers to a system of homogeneous linear equations with integer
coefficients in the unknowns x1 , . . . , xd .
We say that an arbitrary monoid M is normal if it is isomorphic to a linear
monoid. For the reason behind this terminology, see Exercise 13.43. The following
simple result shows that some monoids more general than linear monoids are in fact
normal.
13.20 Theorem. Let M be the set of solutions (a1 , . . . , ad ) ∈ Nd to a set
of homogeneous linear equations with integer coefficients, homogeneous linear
congruences modulo a positive integer, and homogeneous linear inequalities with
integer coefficients. Then M is normal.
As an example of a monoid covered by the above theorem, we could take all
vectors (a1 , . . . , a7 ) ∈ N7 satisfying the five conditions
3x1 − 4x2 − x3 + 7x4 + 5x6 − 4x7 = 0

x1 + x3 − 9x6 − 4x7 = 0
x1 + x2 − 3x6 ≡ 0 (mod 5)
4x2 − x3 + 2x4 + 2x5 − 3x7 ≥ 0
x3 − 3x5 ≥ 0.
Proof. Let the given congruences be Li (x1 , . . . , xd ) ≡ 0 (mod ni ), 1 ≤ i ≤ r, so

each Li (x1 , . . . , xd ) has the form

d
Li (x1 , . . . , xd ) = aim xm , aim ∈ Z.
m=1
We can replace each aim with a nonnegative integer bim satisfying aim ≡
bim (mod ni ) without affecting the solutions to our system of equations, congru-
ences, and inequalities. Let

d
Li (x1 , . . . , xd ) = bim xm .
m=1
13.8 Affine Monoids 237
Introduce a new variable yi , and replace the congruence Li (x1 , . . . , xd ) ≡

0 (mod ni ) with the equation
Li (x1 , . . . , xd ) − ni yi = 0. (13.11)
Thus if (x1 , . . . , xd , y) = (a1 , . . . , ad , b) is a solution to (13.11) in integers with

each ai ≥ 0, then also b ≥ 0. Similarly for each inequality Mj (x1 , . . . , xd ) ≥ 0,
introduce a new variable zj (called a slack variable), and replace the inequality with
the equation
Mj (x1 , . . . , xd ) − zj = 0.
It is clear [why?] that the monoid of solutions in nonnegative integers to the original
set of equations, congruences, and inequalities is isomorphic to the monoid of
solutions in nonnegative integers to the new system of equations only, so the proof
follows.
We now come to the main result of this section. For the proof, recall that a
commutative ring R with 1 is said to be noetherian if there does not exist an infinite
strictly ascending chain I1 ⊂ I2 ⊂ · · · of ideals of R. (This condition is called
the ascending chain condition or ACC.) Equivalently, every ideal of R is finitely-
generated. A fundamental (and not so difficult to prove) result of commutative
algebra, called the Hilbert basis theorem, asserts that if R is noetherian, then so
is the polynomial ring R[x]. In particular, if K is a field (and hence noetherian
since its only ideals are {0} and K), then so is the polynomial ring K[x1 , . . . , xd ].
Moreover, it is not hard to show that if I is an ideal of a noetherian ring and is a
set of generators of I , then some finite subset of generates I .
13.21 Theorem. If M is a normal monoid, then M is finitely-generated.
Proof. We may assume that M is linear, since if a monoid N is finitely generated

then any monoid isomorphic to N is also finitely generated. Suppose that M ⊆ Nd .
Let K be a field, and let R = K[x1 , . . . , xd ], the polynomial ring over K in the
indeterminates x1 , . . . , xd . If α = (α1 , . . . , αd ) ∈ Nd , then we use the “multivariate
notation”
x α = x1α1 · · · xdαd .
Let I be the ideal of R generated by all monomials x α where α ∈ M. By the Hilbert

basis theorem I is finitely generated as an ideal of R; in fact, some finite set of
monomials x α , α ∈ R, generates I . However, this does not imply that M is finitely
generated as a monoid. (If it did, then every submonoid of Nd would be finitely
generated, but we have seen in (13.10) that this is not the case.) The key property
(whose proof is trivial) that we need of linear submonoids M of Nd is the following.
(P) If α, β ∈ M and β − α ∈ Nd , then β − α ∈ M.
We claim that M is generated by those elements α for which x α ∈ . For any

β = (β1 , . . . , βd ) ∈ M write
s(β) = β1 + · · · + βd .
We prove by induction on s(β) that any β ∈ M is a nonnegative linear combination

of elements of .
The assertion is clearly true for s(β) = 0, since then β = (0, 0, . . . , 0). Now let
β ∈ M with s(β) > 0, and assume the induction hypothesis for every γ ∈ M with
s(γ ) < s(β). Since β ∈ M we have x β ∈ I , so we can write

xβ = fα (x)x α , (13.12)
xα ∈
where fα (x) ∈ R. In order for the term x β to appear on the right-hand side of
(13.12), some fα (x) must have a term x γ (with some nonzero coefficient) such that
β = α + γ . By Property (P) above, we have γ ∈ M. Note that s(γ ) < s(β), so by
the induction hypothesis γ is a nonnegative integer linear combination of elements
of . Since β = α + γ where α ∈ , the same is true for β. Hence the proof follows
by induction.
The 100 prisoners problem was first considered by Miltersen. It appeared in a paper
with Gál [50]. Further information on the history of this problem, together with a
proof of optimality of the prisoners’ strategy, is given by Curtin and Warshauer [27].
The Oddtown theorem is due to Berlekamp [7]. Theorem 13.3 on decomposing
Kn into complete bipartite subgraphs is due to Graham and Pollak [54, 55]. For
Fisher’s original proof of the inequality v ≤ b for BIBD’s and Bose’s nonuniform
generalization, see [41] and [12]. Sutner’s original proof of the odd neighborhood
theorem (Theorem 13.5) appears in [133], while the simpler proof of Caro may
be found in [20]. The odd neighborhood problem is also known as the Lights
Out Puzzle. For a host of other applications of linear algebra along the lines of
Sections 13.2–13.5, see the unpublished manuscript [4] of Babai and Frankl, and
the book [92] of Matoušek.
The circulant Hadamard matrix conjecture was first mentioned in print by Ryser
[113, p. 134], though its precise origin is obscure. The work of Turyn mentioned
in the text appears in [137, 138]. Some more recent progress is due to Leung and
Schmidt [81].
While P -recursive functions and their cousins the D-finite series of Exer-
cise 12.26 were known to nineteenth century analysts, the first systematic treatment
of them did not appear until the paper of Stanley [125] in 1980, which includes a
statement and proof of Theorem 13.18. For an exposition, see Stanley [131, §6.4].
Theorem 13.21 is known as Gordan’s lemma, named after the German math-
ematician Paul Gordan (1837–1912). See [53]. For further information on sub-
monoids of Nd , see Bruns and Gubeladze [18, Chapter 2].
1. Suppose that we have 2n prisoners and the same evil warden as in Section 13.1.
Let 0 < α < 1. Now the prisoners open 2αn of the boxes (more precisely, the
closest integer to 2αn). For what value of α will the strategy used in the proof
of Theorem 13.1 yield a 50% chance of success in the limit as n → ∞?
2. Suppose that we have the same 100 prisoners and evil warden as in Section 13.1.
This time, however, each prisoner must open 99 boxes. If any prisoner sees his
or her name, then they are all killed. Find the best strategy for the prisoners and
the resulting probability p of success. Note that 10−200 ≤ p ≤ 10−2 , the upper
bound because the first prisoner has a success probability of 1/100. (Unlike the
situation in Section 13.1, once the best strategy is found for the present problem
the proof of optimality is easy.)
3. (a) This time the evil warden puts a red hat or a blue hat on the head of each
of the 100 prisoners. Each prisoner sees all the hats except for his own.
The prisoners simultaneously guess the color of their hat. If any prisoner
guesses wrong, then all are killed. What strategy minimizes the probability
that all are killed?
(b) Now the prisoners have hats as before, but only the prisoners who guess
wrong are killed. What is the largest integer m such that there is some
strategy guaranteeing that at least m prisoners survive?
4. (*) Our poor 100 prisoners have distinct real numbers written on their fore-
heads. They can see every number but their own. They each choose (indepen-
dently, without communication) a red or blue hat and put it on their heads. The
warden lines them up in increasing order of the numbers on their foreheads. If
any two consecutive prisoners have the same color hat, then all are killed. What
is the best strategy for success?
5. (a) (*) Suppose that n people live in Reverse Oddtown. Every club contains an
even number of persons, and any two clubs share an odd number of persons.
Show that no more than n clubs can be formed.
(b) (rather difficult) Show that if n is even, then at most n − 1 clubs can be
formed.
6. a. Suppose that n people live in Eventown. Every club contains an even number
of persons, every two clubs share an even number of persons, and no two
clubs have identical membership. Show that the maximum number of clubs
is 2n/2 .
b. (rather difficult) Suppose that fewer than 2n/2 clubs have been formed
using the Eventown rules of (a). Show that another club can be formed
without breaking the rules.
7. (Bipartite Oddtown) A town of n citizens has m red clubs R1 , . . . , Rm and m

blue clubs B1 , . . . , Bm . Assume that #(Ri ∩ Bi ) is odd for every i, and that
#(Ri ∩ Bj ) is even for every i = j . Prove that m ≤ n.
8. Triptown has n persons and some clubs. Any two distinct clubs share a number
of members that is a multiple of 3. Let j be the number of clubs C for which
#C is not a multiple of 3. Show that j ≤ n.
9. Strangetown has n people and k clubs C1 , . . . , Ck . Each club has an even
number of members, and any two clubs have an even number of members in
common, with the following exception: Ci and Ck+1−i have an odd number of
members in common for 1 ≤ i ≤ k. (If k = 2m + 1 and i = m + 1, then this
means that club Cm+1 has an odd number of members.) Show that there are at
most n clubs.
10. Weirdtown has n people and k clubs C1 , . . . , Ck . Each club has an even number
of members, and any two clubs have an even number of members in common,
with the following exception: if 1 ≤ i ≤ k − 1, then Ci and Ci+1 have an odd
number of members in common. As a function of n, what is the largest possible
value of k?
11. (a) An n × n real matrix is skew-symmetric if At = −A. Let A be such a
matrix with n odd. Show that det A = 0. (This is a standard result from
linear algebra.)
(b) Let G be a simple graph with an odd number of vertices. The deleted
neighborhood N (v) of a vertex v is the set of all vertices adjacent to v.
Show that there is a nonempty subset S of the vertices such that S intersects
every deleted neighborhood N (v) in an even number of elements.
12. (*) Let Mn denote the vector space of all real n × n matrices, so dim Mn = n2 .
Let V be a subspace of Mn such that every eigenvalue of every matrix A ∈ V
is real. Show that dim V ≤ n+1 2 .
13. Let the edge set E(Kn ) of the complete graph Kn be a union of the edge sets of
complete bipartite graphs B1 , . . . , Bm such that every edge of Kn is covered an
odd number of times. Show that m ≥ (n − 1)/2. (The minimum value of m is
not known.)
14. A complete tripartite graph with vertex tripartition (X1 , X2 , X3 ) is the graph
on the disjoint vertex sets Xi with an edge between any two vertices not in
the same set Xi . (We allow one of the Xi to be empty, so for instance K2 is a
complete tripartite graph.) Thus if #Xi = pi then the complete tripartite graph
has p1 p2 + p1 p3 + p2 p3 edges. Suppose that the edge set E(Kn ) is partitioned
into m disjoint edge sets of complete tripartite graphs. What is the minimum
value of m?
15. (*) Let A1 , . . . , An be distinct subsets of an n-set X. Give a linear algebra proof
that for some x ∈ X, all the sets Ai − x (short for Ai − {x}) are distinct. (There
is a fairly simple combinatorial proof, but that is not what is asked for.) NOTE
ON NOTATION : Ai − x = Ai if x ∈ Ai .
16. Show that the number of “switching sets” S in Theorem 13.5 has the form 2n
for some n ≥ 0.
17. (a) Let G be a simple graph with p vertices such that exactly 2p−1 subsets of
the vertices are switching sets, i.e., they turn off all the light bulbs in the
scenario of Section 13.5. Show that G is a complete graph Kp . Give a proof
based on linear algebra.
(b) Describe the 2p−1 switching sets for Kp .
(c) (more difficult) Same as above, but with exactly 2p−2 switching sets. Show
that G is a disjoint union of two complete graphs.
18. Given v = (v1 , . . . , vn ) ∈ Zn , let f (v) be the number of values of 1 ≤ i ≤ n
for which the sum vi + vi+1 + vi+2 is even, where the subscripts are taken
modulo n so that they always lie in the set {1, 2, . . . , n}. For
which positive
integers n is the following true: for all 0 ≤ k ≤ n, exactly nk vectors among
the 2n vectors v ∈ {0, 1}n satisfy f (v) = k?
19. (*) Show that a Hadamard matrix H has order 1, 2, or n = 4m for some integer
m ≥ 1.
20. For what values of n do there exist n + 1 vertices of an n-dimensional cube
such that any two of them are the same distance apart? For instance, it’s clearly
impossible for n = 2, while for n = 3 the vertices can be 000, 110, 101, 011.
Your answer should involve Hadamard matrices.
21. (a) Show that if H is an n × n Hadamard matrix all of whose rows have the
same number of 1’s, then n is a square.
(b) Show also that all columns of H have the same number of 1’s.
22. (*) Clearly 2n and n! are P -recursive functions, so by Theorem 13.18 so is
f (n) = 2n + n!. Find a recurrence of the form (13.8) satisfied by f (n).
3
23. (a) (difficult) Let f (n) = nk=0 nk . Show that
(n+2)2 f (n+2)−(7n2 +21n+16)f (n+1)−8(n+1)2 f (n) = 0, n ≥ 0.

d
(b) (*) Fix a positive integer d. Show that the function fd (n) = nk=0 nk is
P -recursive.
24. (*) (difficult) Let f (n) be the number of paths from (0, 0) to (n, n), where
each step in the path is (1, 0), (0, 1), or (1, 1). For instance, f (1, 1) = 3,
corresponding to the paths (where we abbreviate (1, 0) as 10, etc.) 00 → 10 →
11, 00 → 01 → 11, and 00 → 11. Show that
(n + 2)f (n + 2) − 3(2n + 3)f (n + 1) + (n + 1)f (n) = 0, n ≥ 0.
25. (a) Let α1 , . . . , αk be distinct nonzero complex numbers, and let Q1 (n),
. . . , Qk (n) be distinct nonzero complex polynomials. Define
f (n) = Q1 (n)α1n + Q2 (n)α2n + · · · + Qk (n)αkn , n ≥ 0.
Show that f (n) is P -recursive.

(b) (difficult) Show that the least degree d of a recurrence (13.8) satisfied by
f (n) is equal to k.

26. (a) Let C[[x]] denote the ring of all power series n≥0 an x over C. It
n
is easyto see that C[[x]] is an integral domain. A Laurent series is a

series n∈Z bn x n , i.e., any integer exponents are allowed. Show that the
of C[[x]], denoted C((x)), consists of all Laurent series of
quotient field
the form n≥n0 bn x n for some n0 ∈ Z, i.e., all Laurent series with finitely
many negative exponents. Equivalently, C((x)) is obtained from C[[x]] by
inverting the single element x. Show also that C((x)) contains the field
C(x) of all rational functions P (x)/Q(x), where P , Q ∈ C[x], in the sense
that there is a Laurent series F (x) ∈ C((x)) satisfying Q(x)F (x) = P (x).
(b) A Laurent series y ∈ C((x)) is called D-finite (short for differentiably
finite) if there exist polynomials p0 (x), . . . , pd (x), not all 0, such that
pd (x)y (d) + · · · + p1 (x)y + p0 (x)y = 0,
where y (d) denotes

the dth derivative of y with respect to x. Show that a
power series n≥0 f (n)x n is D-finite if and only if the function f (n) is
P -recursive.
(c) Show that y ∈ C((x)) is D-finite if and only if the vector space over C(x)
spanned by y, y , y , . . . is finite-dimensional (whence the terminology
“D-finite”).
(d) Show that the set D of D-finite Laurent series f ∈ C((x)) forms a
subalgebra of C((x)), i.e., D is closed under complex linear combinations
and under product.
(e) (more difficult) Show that D is not a subfield of C((x)).
27. (*) A Laurent series y ∈ C((x)) is called algebraic if there exist polynomials
p0 (x), p1 (x), . . . , pd (x) ∈ C[x], not all 0, such that
pd (x)y d + pd−1 (x)y d−1 + · · · + p1 (x)y + p0 (x) = 0. (13.13)
Show that an algebraic series y is D-finite.

28. (very difficult) Show that a nonzero Laurent series f (x) ∈ C((x)) and its
reciprocal 1/f (x) are both D-finite if and only if f (x)/f (x) is algebraic.
29. Fix k ≥ 1 and write x = (x1 , . . . , xk ). Suppose that

F (x) = an1 ,...,nk x1n1 · · · xknk
n1 ,...,nk ≥0
is a power series in x with complex

coefficients an1 ,...,nk . Define the diagonal
of F (x) to be the power series n≥0 an,n,...,n x n in the single variable x.
(a) (difficult) Suppose that F (x) represents a rational function. In other words,
there are polynomials P (x), Q(x) ∈ C[x] such that Q(x)F (x) = P (x).
Show that if k = 2 then the diagonal of F (x) is algebraic.
(b) Show√that the diagonal of the rational function 1/(1 − x − y − xy) is equal
to 1/ 1 − 6x + x 2 .
(c) (difficult) Let
1 (k + m + n)! k m n
F (x, y, z) = = x y z .
1−x−y−z k! m! n!
k,m,n≥0

Show that the diagonal series n≥0 (3n)!n!3
x n is not algebraic.
(d) (very difficult) Show that the diagonal of any power series over C in finitely
many variables that represents a rational function is D-finite.
30. Show that every submonoid of N (under the operation of addition) is finitely-
generated.
31. Let a and b be relatively prime positive integers, and let M be the submonoid
of N generated by a and b.
(a) Show that
1 − x ab
xj = .
(1 − x a )(1 − x b )
j ∈M
(b) Show that if j ≥ ab − a − b + 1 then j ∈ M.

(c) Show that ab − a − b ∈ M.
(d) Let 0 ≤ i ≤ ab − a − b. Show that exactly one of i and ab − a − b − i
belongs to M. Note that this generalizes (c) since 0 ∈ M.
32. Does the monoid N2 have a submonoid isomorphic to N3 ?
33. (a) Show that any submonoid of Nd has a unique minimal generating set G,
i.e., every set of generators contains G.
(b) In contrast to (a), give an example of a commutative monoid M with more
than one element with the following property: if G generates M and S is
any finite subset of G, then G − S generates M.
34. Let n ≥ 1. Show that there exists a submonoid of N whose minimal generating
set G (as defined in the previous problem) has exactly n elements.
35. (moderately difficult) (*) Write x = (x1 , . . . , xd ). Let L1 (x), . . . , Lm (x) be
homogeneous linear forms with integer coefficients. Show that the following
two conditions are equivalent:
• The only solution x = (a1 , . . . , ad ) ∈ Nd of the equations L1 (x) =
0, . . . , Lm (x) = 0 is (0, 0, . . . , 0).
• Some integer linear combination of L1 (x), . . . , Lm (x) has every coefficient
positive.
An example is given by Example 13.19(c).
36. (*) Let d ≥ 1. Does there exist an infinite increasing chain M1 ⊂ M2 ⊂ · · · of
submonoids of Nd ?
be the submonoid of N consisting of all vectors (a1 , . . . , ad ) ∈ N such
37. Let M d d
that ai is even. Describe the unique minimal set G of generators of M.

38. (a) Show that every normal submonoid of N is linear.

(b) Given an example of a finitely-generated submonoid M of N2 such that
M is normal, but it fails to satisfy Property (P) (from the proof of
Theorem 13.21).
39. Let S be a set of monomials in the variables x1 , . . . , xd . Suppose that no
monomial in S is divisible by another monomial is S. Show that S is finite.
Try to give a simple proof using commutative algebra.
40. Prove the assertion preceding Theorem 13.21 that if R is a noetherian ring
(commutative with 1) and generates an ideal I of R, then some finite subset
of generates R.
41. Give a proof of Theorem 13.21 that does not use commutative algebra.
42. (difficult) Show that a finitely generated commutative monoid M is normal if
and only if the following condition holds: if α, β, γ ∈ M, n ∈ P, and nα + β =
nγ , then β = nδ for some δ ∈ M. Note that this result gives an intrinsic
characterization of normal monoids, i.e., a characterization defined solely in
terms of M and not how it is embedded in a larger monoid.
43. Let R be a commutative integral domain (with 1). Recall from commutative
algebra that R is normal if it is integrally closed in its quotient field R̄, i.e.,
if α ∈ R̄ satisfies a monic polynomial equation with coefficients in R, then
α ∈ R. For instance, the ring R = K[x 2 , x 3 , x 4 , . . . ] (where K is a field)
3
is not normal since x = xx 2 ∈ R̄, x ∈ R, and x is a zero of the polynomial
f (t) = t 2 − x 2 ∈ R[t]. Let M be a finitely generated submonoid of Nd , and let
R be the subalgebra of K[x1 , . . . , xd ] generated (or, in this situation, spanned)
by the monomials x α , α ∈ M. Show that R is normal if and only if M is a
normal monoid. This result explains the terminology “normal monoid.”
Hints and Comments for Some Exercises
Chapter 1
1.5 Consider A(Hn )2 and use Exercise 1.3.
1.6 (a) First count the number of sequences Vi0 , Vi1 , . . . , Vi for which there exists
a closed walk with vertices v0 , v1 , . . . , v = v0 (in that order) such that
vj ∈ Vij .
1.11 Consider the rank of A( ), and also consider A( )2 . The answer is very
simple and does not involve irrational numbers.
1.12 (b) Consider A(G)2 .
Chapter 2
2.2 Give an argument analogous to the proof of Theorem 8.8.9.
2.3 This result is known as the “middle levels conjecture” and was proved by
Torsten Mütze in 2016.
2.5 (c) Mimic the proof for the graph Cn , using the definition

χu , χv = χu (w)χv (w),
w∈Zn
where an overhead bar denotes complex conjugation.
Chapter 3
3.4 You may find Example 3.1 useful.
3.8 It is easier not to use linear algebra.
3.9 See previous hint.

https://doi.org/10.1007/978-3-319-77173-1
246 Hints and Comments for Some Exercises
3.11 First show (easy) that if we start at a vertex v and take n steps (using our
random walk model), then the probability that we traverse a fixed closed walk
W is equal to the probability that we traverse W in reverse order.
3.13 See hint for Exercise 3.8.
Chapter 4
4.4 (b) One way to do this is to count in two ways the number of k-tuples
(v1 , . . . , vk ) of linearly independent elements from Fnq : (1) first choose v1 , then
v2 , etc., and (2) first choose the subspace W spanned by v1 , . . . , vk , and then
choose v1 , v2 , etc.
4.4 (c) The easiest way is to use (b).
4.7 (b) This result was proved by R. P. Dilworth in 1950 and is known as Dilworth’s
theorem.
Chapter 5
5.5 (a) Show that Nn ∼
= Bn /G for asuitable
group G.
5.9 (a) Use Corollary 2.4 with n = p2 .
5.13 Use Exercise 5.12.
Chapter 6
6.2 (b) Not really a hint, but the result is equivalent [why?] to the case r = m,
s = n, t = 2, and x = 1 of Exercise 36.
6.3 Consider μ = (8, 8, 4, 4).
6.5 First consider the case where S has ζ elements equal to 0 (so ζ = 0 or 1), ν
elements that are negative, and π elements that are positive, so ν + ζ + π =
2m + 1.
Chapter 7
7.3 Hint. Dark red and burgundy are the same color!
For those who are interested, the cycle indicator ZG of the full symmetry
group G of the dodecahedron acting on the vertices is
1 20
z1 + 15z210 + 20z12 z36 + 24z54 + 15z14 z28 + z210 + 20z2 z63 + 24z10
2
.
120
The number of inequivalent vertex colorings using two distinct colors is
ZG (2, 2, . . . , 2) = 9436.
7.19 (a) Use Pólya’s theorem.
Hints and Comments for Some Exercises 247
Chapter 8
8.3 Encode a maximal chain by an object that we already know how to enumerate.
8.7 Partially order by diagram inclusion the set of all partitions whose diagrams
can be covered by nonoverlapping dominos, thereby obtaining a subposet Y2
of Young’s lattice Y . Show that Y2 ∼ = Y × Y.
8.14 Use induction on n.
8.17 (a) One way to do this is to use the generating function n≥0 ZSn (z1 , z2 , . . . )
x n for the cycle indicator of Sn (Theorem 7.13). Another method is to find a
recurrence for B(n + 1) in terms of B(0), . . . , B(n) and then convert this
recurrence into a generating function
8.18 Consider the generating function
t kqn
G(q, t) = κ(n → n + k → n)
(k!)2
k,n≥0
and use (8.25).

8.20 (b) Consider the square of the adjacency matrix of Yj −1,j .
8.24 Use Exercise 8.14.
8.33 Use an inclusion–exclusion argument (a topic not covered in the text).
Chapter 9
9.1 There is a simple proof based on the formula κ(Kp ) = pp−2 , avoiding the
Matrix-Tree Theorem.
9.3 Consider the matrices L − nI and L − (n − m)I .
9.2 (c) Use the fact that the rows of L sum to 0, and compute the trace of L.
9.7 (b) Use Exercise 1.3.
9.8 (a) For the most elegant proof, use the fact that commuting p × p matrices
A and B can be simultaneously triangularized, i.e., there exists an invertible
matrix X such that both XAX−1 and XBX−1 are upper triangular.
9.8 (d) Use Exercise 9.10(a).
9.9 Let G∗ be the full dual graph of G, i.e., the vertices of G∗ are the faces of G,
including the outside face. For every edge e of G separating two faces R and S
of G, there is an edge e∗ of G∗ connecting the vertices R and S. Thus G∗ will
have some multiple edges, and #E(G) = #E(G∗ ). First show combinatorially
that κ(G) = κ(G∗ ). (See Corollary 11.19.)
9.10 (a) Use the Binet–Cauchy theorem.
9.12 (a) The Laplacian matrix L = L(G) acts on the space RV (G), the real vector
space with basis V (G). Consider the subspace W of RV (G) spanned by the
elements v + ϕ(v), v ∈ V (G).
9.13 (a) Let s(n, q, r) be the number of n × n symmetric matrices of rank r over
Fq . Find a recurrence satisfied by s(n, q, r), and verify that this recurrence is
satisfied by
248 Hints and Comments for Some Exercises
⎧
⎪
⎪
t
q 2i
2t−1
⎪
⎪ · (q n−i − 1), 0 ≤ r = 2t ≤ n
⎪
⎨ q 2i − 1
i=1 i=0
s(n, q, r) =
⎪
⎪
t
q 2i
2t
⎪
⎪ · (q n−i − 1), 0 ≤ r = 2t + 1 ≤ n.
⎪
⎩ q 2i − 1
i=1 i=0
9.14 Any of the three proofs of the Appendix to Chapter 9 can be carried over to
the present exercise.
Chapter 10
10.4 (a) Use a suitable generalization of the deBruijn graph.
(b) Use techniques similar to those in Theorem 7.7.10.
10.5 (b) Use the Perron–Frobenius theorem (Theorem 3.3).
10.8 (a) Consider A .
10.8 (f) There is an example with nine vertices that is not a de Bruijn graph.
10.8 (c) Let E be the (column) eigenvector of A(D) corresponding to the largest
eigenvalue. Consider AE and At E, where t denotes transpose.
Chapter 11
11.2(d) Let e be an edge of G that is not an isthmus. Let G − e denote G with the
edge e removed, and let G/e denote G with the edge e contracted to a point
(so G/e has one less vertex than G). Find a simple recurrence satisfied by
CG (n) in terms of CG−e (n) and CG/e (n), and then use induction.
11.5 Use the unimodularity of the basis matrices C T and B T .
11.6 Use the formula (11.5) for the resistance of a parallel connection.
11.8 (a) Mimic the proof of Theorem 9.8 (the Matrix-Tree Theorem).
11.8 (b) Consider ZZ t .
Chapter 12
12.9. Compute the reduced Euler characteristic of X and consider the last step of
a possible shelling.
12.10. Modify the dunce hat of the previous exercise.
12.12. There are two examples with f -vector (6, 14, 9) and none smaller.
12.13. (b) Consider the hint for Problem 12.12.
(c) Use Exercise 12.10.
12.18. The following property of triangulations of spheres is sufficient for
solving this exercise. For every F ∈ we have
χ̃(lk (F )) = (−1)dim lk (F ) .

Hints and Comments for Some Exercises 249
Here χ̃ denotes reduced Euler characteristic, and lk denotes link, as defined

in Remark 12.26.
12.20. There are five minimal nonfaces.
12.21. One approach is to consider minimal ideals I of K[ ], say, for which
K[ ]/I is a polynomial ring, i.e., isomorphic to K[y1 , . . . , yr ] for inde-
terminates y1 , . . . , yr .
12.30. Answer: 588. Without shellability, the answer is 589.
12.31. Let I be the ideal of K[] generated by all monomials x F , where F ∈ .
Clearly K[ ] is isomorphic to K[]/I (as a K-algebra). Let θ1 , . . . , θd ∈
K[]1 satisfy Property (P), and consider the natural map
f : K[]/(θ1 , . . . , θd ) → (K[]/I )/(θ1 , . . . , θd ).

12.32. Consider θi ∈ K[]1 defined by θi = xj ∈Vi xj . Also consider the product
xj θi in the ring K[]/(θ1 , . . . , θd ) for all 1 ≤ i ≤ d and xj ∈ Vi .
Chapter 13
13.4 The best strategy involves the concept of odd and even permutations.
13.5 For the easiest solution, don’t use linear algebra but rather use the original
Oddtown theorem.
13.12 What are the eigenvalues of skew-symmetric matrices?
13.15 Consider the incidence matrix M of the sets and their elements. Consider two
cases: det M = 0 and det M = 0.
13.19 Consider the first three rows of H . Another method is to use row operations
to factor a large power of 2 from the determinant.
13.22 It is easiest to proceed directly
and not use the proof of Theorem 13.18.
xn
13.23 (b) Consider the power series n≥0 n! d.
13.24 Use Exercise 13.29(b).
13.27 Differentiate with respect to x (13.13) satisfied by y.
13.35 This result is known as Stiemke’s theorem (after E. Stiemke, 1892–1915) and
is a forerunner of the duality theorem of linear programming.
13.36 The answer depends on d.
Bibliography
1. I. Anderson, Combinatorics of Finite Sets (Oxford University Press, Oxford, 1987); Corrected
republication by Dover, New York, 2002
2. G.E. Andrews, The Theory of Partitions (Addison-Wesley, Reading, 1976)
3. G.E. Andrews, K. Eriksson, Integer Partitions (Cambridge University Press, Cambridge,
2004)
4. L. Babai, P. Frankl, Linear Algebra Methods in Combinatorics, preliminary version 2 (1992),
216 pp.
5. E.A. Beem, Craige and Irene Schensted don’t have a car in the world, in Maine Times (March
12, 1982), pp. 20–21
6. E.A. Bender, D.E. Knuth, Enumeration of plane partitions. J. Comb. Theor. 13, 40–54 (1972)
7. E.R. Berlekamp, On subsets with intersections of even cardinality. Can. Math. Bull. 12,
363–366 (1969)
8. O. Bernardi, On the spanning trees of the hypercube and other products of graphs. Electron.
J. Comb. 19(P52), 16 pp (2012)
9. H. Bidkhori, S. Kishore, Counting the spanning trees of a directed line graph. Preprint
[arXiv:0910.3442]
10. E.D. Bolker, The finite Radon transform, in Integral Geometry, Brunswick, Maine, 1984.
Contemporary Mathematics, vol. 63 (American Mathematical Society, Providence, 1987),
pp. 27–50
11. C.W. Borchardt, Ueber eine der Interpolation entsprechende Darstellung der Eliminations–
Resultante. J. Reine Angew. Math. (Crelle’s J.) 57, 111–121 (1860)
12. R.C. Bose, A note on Fisher’s inequality for balanced incomplete block designs. Ann. Math.
Stat. 619–620 (1949)
13. F. Brenti, Log-concave and unimodal sequences in algebra, combinatorics, and geometry: an
update, in Jerusalem Combinatorics ’93. Contemporary Mathematics, vol. 178 (American
Mathematical Society, Providence, 1994)
14. A.E. Brouwer, W.H. Haemers, Spectra of Graphs (Springer, New York, 2012)
15. H. Bruggesser, P. Mani, Shellable decompositions of cells and spheres. Math. Scand. 29,
197–205 (1971)
16. W. Burnside, Theory of Groups of Finite Order (Cambridge University Press, Cambridge,
1897)
17. W. Burnside, Theory of Groups of Finite Order, 2nd edn. (Cambridge University Press,
Cambridge, 1911); Reprinted by Dover, New York, 1955
18. W. Bruns, J. Gubeladze, Polytopes, Rings, and K-Theory (Springer, Dordrecht, 2009)

https://doi.org/10.1007/978-3-319-77173-1
252 Bibliography
19. W. Bruns, J. Herzog, Cohen-Macaulay Rings, revised edn. Cambridge Studies in Advanced
Mathematics, vol. 39 (Cambridge University Press, Cambridge/New York, 1998)
20. Y. Caro, Simple proofs to three parity theorems. Ars Combin. 42, 175–180 (1996)
21. N. Caspard, B. Leclerc, B. Monjardet, Finite Ordered Sets. Encyclopedia of Mathematics and
Its Applications, vol. 144 (Cambridge University Press, Cambridge, 2012)
22. A.L. Cauchy, Mémoire sur diverses propriétés remarquables des substitutions régulaires ou
irrégulaires, et des systèmes de substitutions conjuguées (suite). C. R. Acad. Sci. Paris 21,
972–987 (1845); Oeuvres Ser. 1 9, 371–387
23. A. Cayley, Note sur une formule pour la réversion des séries. J. Reine Angew. Math. (Crelle’s
J.) 52, 276–284 (1856)
24. A. Cayley, A theorem on trees. Q. J. Math. 23, 376–378 (1889); Collected Papers, vol. 13
(Cambridge University Press, Cambridge, 1897), pp. 26-28
25. G.F. Clements, B. Lindström, A generalization of a combinatorial theorem due to Macaulay.
J. Comb. Theory 7, 230–238 (1969)
26. I.S. Cohen, On the structure and ideal theory of complete local rings. Trans. Am. Math. Soc.
59, 54–106 (1946)
27. E. Curtin, M. Warshauer, The locker puzzle. Math. Intelligencer 28, 28–31 (2006)
28. D.M. Cvetković, M. Doob, H. Sachs, Spectra of Graphs: Theory and Applications, 3rd edn.
(Johann Ambrosius Barth, Heidelberg/Leipzig, 1995)
29. D.M. Cvetković, P. Rowlinson, S. Simić, in An Introduction to the Theory of Graph
Spectra. London Mathematical Society. Student Texts, vol. 75 (Cambridge University Press,
Cambridge, 2010)
30. N.G. de Bruijn, A combinatorial problem. Proc. Koninklijke Nederlandse Akademie v.
Wetenschappen 49, 758–764 (1946); Indagationes Math. 8, 461–467 (1946)
31. N.G. de Bruijn, Pólya’s theory of counting, in Applied Combinatorial Mathematics, ed. by
E.F. Beckenbach (Wiley, New York, 1964); Reprinted by Krieger, Malabar, FL, 1981
32. N.G. de Bruijn, Acknowledgement of priority to C. Flye Sainte-Marie on the counting of
circular arrangements of 2n zeros and ones that show each n-letter word exactly once,
Technische Hogeschool Eindhoven, T.H.-Report 75-WSK-06, 1975
33. M.R. DeDeo, E. Velasquez, The Radon transform on Zkn . SIAM J. Discrete Math. 18, 472–478
(electronic) (2004/2005)
34. P. Diaconis, R.L. Graham, The Radon transform on Zk2 . Pac. J. Math. 118, 323–345 (1985)
35. P. Diaconis, R.L. Graham, Magical Mathematics (Princeton University Press, Princeton,
2012)
36. P.G. Doyle, J.L. Snell, Random Walks and Electrical Networks. Carus Mathematical Mono-
graphs (Book 22) (Mathematical Association of America, Washington, DC, 1984); Version
3.02 at arXiv:math/0001057
37. E.B. Dynkin, Some properties of the weight system of a linear representation of a semisimple
Lie group (in Russian). Dokl. Akad. Nauk SSSR (N.S) 71, 221–224 (1950)
38. E.B. Dynkin, The maximal subgroups of the classical groups. Am. Math. Soc. Transl. Ser. 2
6, 245–378 (1957); Translated from Trudy Moskov. Mat. Obsc. 1, 39–166
39. K. Engel, Sperner Theory. Encyclopedia of Mathematics and Its Applications, vol. 65
(Cambridge University Press, Cambridge, 1997)
40. P. Fishburn, Interval Orders and Interval Graphs: A Study of Partially Ordered Sets (Wiley,
New York, 1985)
41. R.A. Fisher, An examination of the different possible solutions of a problem in incomplete
blocks. Ann. Eugen. 10, 52–75 (1940)
42. C. Flye Sainte-Marie, Solution to question nr. 48. l’Intermédiaire des Mathématiciens 1,
107–110 (1894)
43. S. Fomin, Duality of graded graphs. J. Algebr. Combin. 3, 357–404 (1994)
44. S. Fomin, Schensted algorithms for dual graded graphs. J. Algebr. Combin. 4, 5–45 (1995)
45. J.S. Frame, G. de B. Robinson, R.M. Thrall, The hook graphs of Sn . Can. J. Math. 6 316–324
(1954)
Bibliography 253
46. D.S. Franzblau, D. Zeilberger, A bijective proof of the hook-length formula. J. Algorithms 3,
317–343 (1982)
47. F.G. Frobenius, Über die Congruenz nach einem aus zwei endlichen Gruppen gebildeten
Doppelmodul. J. Reine Angew. Math. (Crelle’s J.) 101, 273–299 (1887); Reprinted in
Gesammelte Abhandlungen, vol. 2 (Springer, Heidelberg, 1988), pp. 304–330
48. F.G. Frobenius, Über die Charaktere der symmetrischen Gruppe, Sitzungsber. Kön. Preuss.
Akad. Wissen. Berlin (1900), pp. 516–534; Gesammelte Abh. III, ed. by J.-P. Serre (Springer,
Berlin, 1968), pp. 148–166
49. W.E. Fulton, Young Tableaux. Student Texts, vol. 35 (London Mathematical Soci-
ety/Cambridge University Press, Cambridge, 1997)
50. A. Gál, P.B. Miltersen, The cell probe complexity of succinct data structures, in Proceedings
of the 30th International Colloquium on Automata, Languages and Programming (ICALP)
(2003), pp. 332–344
51. M. Gardner, Squaring the square, in The 2nd Scientific American Book of Mathematical
Puzzles and Diversions (Simon and Schuster, New York, 1961)
52. I.J. Good, Normally recurring decimals. J. Lond. Math. Soc. 21, 167–169 (1947)
53. P. Gordan, Über die Auflosung linearer Gleichungen mit reellen Coefficienten. Math. Ann. 6,
23–28 (1873)
54. R.L. Graham, H.O. Pollak, On the addressing problem for loop switching. Bell Syst. Tech.
J. 50, 2495–2519 (1971)
55. R.L. Graham, H.O. Pollak, On embedding graphs in squashed cubes, in Lecture Notes in
Mathematics, vol. 303 (Springer, New York, 1973), pp. 99-110
56. C. Greene, D.J. Kleitman, Proof techniques in the theory of finite sets, in Studies in
Combinatorics, ed. by G.-C. Rota. M.A.A. Studies in Mathematics, vol. 17 (Mathematical
Association of America, Washington, DC, 1978), pp. 22–79
57. C. Greene, A. Nijenhuis, H.S. Wilf, A probabilistic proof of a formula for the number of
Young tableaux of a given shape. Adv. Math. 31, 104–109 (1979)
58. J.I. Hall, E.M. Palmer, R.W. Robinson, Redfield’s lost paper in a modern context. J. Graph
Theor. 8, 225–240 (1984)
59. F. Harary, E.M. Palmer, Graphical Enumeration (Academic, New York, 1973)
60. F. Harary, R.W. Robinson, The rediscovery of Redfield’s papers. J. Graph Theory 8, 191–192
(1984)
61. G.H. Hardy, J.E. Littlewood, G. Pólya, Inequalities, 2nd edn. (Cambridge University Press,
Cambridge, 1952)
62. L.H. Harper, Morphisms for the strong Sperner property of Stanley and Griggs. Linear
Multilinear Algebra 16, 323–337 (1984)
63. T.W. Hawkins, The origins of the theory of group characters. Arch. Hist. Exact Sci. 7,
142–170 (1970/1971)
64. T.W. Hawkins, Hypercomplex numbers, Lie groups, and the creation of group representation
theory. Arch. Hist. Exact Sci. 8, 243–287 (1971/1972)
65. T.W. Hawkins, New light on Frobenius’ creation of the theory of group characters. Arch. Hist.
Exact Sci. 12, 217–243 (1974)
66. A.P. Hillman, R.M. Grassl, Reverse plane partitions and tableaux hook numbers. J. Comb.
Theor. A 21, 216–221 (1976)
67. M. Hochster, Rings of invariants of tori, Cohen-Macaulay rings generated by monomials, and
polytopes. Ann. Math. 96, 318–337 (1972)
68. R.A. Horn, C.R. Johnson, Matrix Analysis (Cambridge University Press, Cambridge, 1985)
69. J.W.B. Hughes, Lie algebraic proofs of some theorems on partitions, in Number Theory and
Algebra, ed. by H. Zassenhaus (Academic, New York, 1977), pp. 135–155
70. A. Hurwitz, Über die Anzahl der Riemannschen Flächen mit gegebenen Verzweigungspunk-
ten. Math. Ann. 55, 53–66 (1902)
71. A. Joyal, Une théorie combinatoire des séries formelles. Adv. Math. 42, 1–82 (1981)
72. G.O.H. Katona, A theorem of finite sets, in Proc. Tihany Conf. 1966, Budapest (1968)
254 Bibliography
73. N.D. Kazarinoff, R. Weitzenkamp, Squaring rectangles and squares. Am. Math. Mon. 80,
877–888 (1973)
74. B. Kind, P. Kleinschmidt, Schälbare Cohen-Macaulay-Komplexe und ihre Parametrisierung.
Math. Z. 167, 173–179 (1979)
75. G. Kirchhoff, Über die Auflösung der Gleichungen, auf welche man bei der Untersuchung
der Linearen Vertheilung galvanischer Ströme geführt wird. Ann. Phys. Chem. 72, 497–508
(1847)
76. C. Krattenthaler, Bijective proofs of the hook formulas for the number of standard Young
tableaux, ordinary and shifted. Electronic J. Combin. 2, R13, 9 pp. (1995)
77. D.E. Knuth, Permutations, matrices, and generalized Young tableaux. Pac. J. Math. 34,
709–727 (1970)
78. I. Krasikov, Y. Roditty, Balance equations for reconstruction problems. Arch. Math. (Basel)
48, 458–464 (1987)
79. J. Kruskal, The number of simplices in a complex, in Mathematical Optimization Techniques
(University of California Press, Berkeley/Los Angeles, 1963), pp. 251–278
80. J.P.S. Kung, Radon transforms in combinatorics and lattice theory, in Combinatorics and
Ordered Sets, Arcata, CA, 1985. Contemporary Mathematics, vol. 57 (American Mathemati-
cal Society, Providence, 1986), pp. 33–74
81. K.H. Leung, B. Schmidt, New restrictions on possible orders of circulant Hadamard matrices.
Designs Codes Cryptogr. 64, 143–151 (2012)
82. E.K. Lloyd, J. Howard Redfield: 1879–1944. J. Graph Theor. 8, 195–203 (1984)
83. L. Lovász, A note on the line reconstruction problem. J. Comb. Theor. Ser. B 13, 309–310
(1972)
84. L. Lovász, Random walks on graphs: a survey, in Combinatorics. Paul Erdǒs is Eighty, vol. 2,
Bolyai Society Mathematical Studies, vol. 2 (Keszthely, Hungary, 1993), pp. 1–46
85. D. Lubell, A short proof of Sperner’s lemma. J. Comb. Theor. 1, 299 (1966)
86. F.S. Macaulay, The Algebraic Theory of Modular Systems (Cambridge University Press,
Cambridge, 1916); Reprinted by Cambridge University Press, Cambridge/New York, 1994
87. F.S. Macaulay, Some properties of enumeration in the theory of modular systems. Proc. Lond.
Math. Soc. 26, 531–555 (1927)
88. P.A. MacMahon, Memoir on the theory of the partitions of numbers — Part I. Philos. Trans.
R. Soc. Lond. A 187, 619–673 (1897); Collected Works, vol. 1, ed. by G.E. Andrews (MIT,
Cambridge, 1978), pp. 1026–1080
89. P.A. MacMahon, Memoir on the theory of the partitions of numbers — Part IV. Philos. Trans.
R. Soc. Lond. A 209, 153–175 (1909); Collected Works, vol. 1, ed. by G.E. Andrews (MIT,
Cambridge, 1978), pp. 1292–1314
90. P.A. MacMahon, Combinatory Analysis, vols. 1, 2 (Cambridge University Press, Cambridge,
1915/1916); Reprinted in one volume by Chelsea, New York, 1960
91. J. Matoušek, Using the Borsuk-Ulam Theorem: Lectures on Topological Methods in Combi-
natorics and Geometry. Universitext (Springer, Berlin, 2003)
92. J. Matoušek, Thirty-Three Miniatures (American Mathematical Society, Providence, 2010)
93. K. Menger, Untersuchungen über allgemeine Metrik. Math. Ann. 100, 75–163 (1928)
94. J.W. Moon, Counting Labelled Trees. Canadian Mathematical Monographs, vol. 1
(Canadian Mathematical Congress, 1970)
95. V. Müller, The edge reconstruction hypothesis is true for graphs with more than n log2 n
edges. J. Comb. Theor. Ser. B 22, 281–283 (1977)
96. P.M. Neumann, A lemma that is not Burnside’s. Math. Sci. 4, 133–141 (1979)
97. J.-C. Novelli, I. Pak, A.V. Stoyanovskii, A new proof of the hook-length formula. Discrete
Math. Theor. Comput. Sci. 1, 053–067 (1997)
98. K.M. O’Hara, Unimodality of Gaussian coefficients: a constructive proof. J. Comb. Theor.
Ser. A 53, 29–52 (1990)
99. W.V. Parker, The matrices AB and BA. Am. Math. Mon. 60, 316 (1953); Reprinted in
Selected Papers on Algebra, ed. by S. Montgomery et al. (Mathematical Association of
America, Washington), pp. 331–332
Bibliography 255
100. J. Pitman, Coalescent random forests. J. Comb. Theor. Ser. A 85, 165–193 (1999)
101. G. Pólya, Kombinatorische Anzahlbestimmungen für Gruppen, Graphen und chemische
Verbindungen. Acta Math. 68, 145–254 (1937)
102. G. Pólya, R.C. Read, Combinatorial Enumeration of Groups, Graphs, and Chemical Com-
pounds (Springer, New York, 1987)
103. M. Pouzet, Application d’une propriété combinatoire des parties d’un ensemble aux groupes
et aux relations. Math. Zeit. 150, 117–134 (1976)
104. M. Pouzet, I.G. Rosenberg, Sperner properties for groups and relations. Eur. J. Combin. 7,
349–370 (1986)
105. R.A. Proctor, A solution of two difficult combinatorial problems with linear algebra. Am.
Math. Mon. 89, 721–734 (1982)
106. H. Prüfer, Neuer Beweis eines Satzes über Permutationen. Arch. Math. Phys. 27, 742–744
(1918)
107. J. Radon, Über die Bestimmung von Funktionen durch ihre Integralwerte längs gewisser
Mannigfaltigkeiten. Berichte über die Verhandlungen der Königlich Sächsischen Gesellschaft
der Wissenschaften zu Leipzig 69, 262–277 (1917); Translation by P.C. Parks, On the
determination of functions from their integral values along certain manifolds. IEEE Trans.
Med. Imaging 5, 170–176 (1986)
108. J.H. Redfield, The theory of group reduced distributions. Am. J. Math. 49, 433–455 (1927)
109. J.H. Redfield, Enumeration by frame group and range groups. J. Graph Theor. 8, 205–223
(1984)
110. G. Reisner, Cohen-Macaulay quotients of polynomial rings. Adv. Math. 21, 30–49 (1976)
111. J.B. Remmel, Bijective proofs of formulae for the number of standard Young tableaux. Linear
Multilinear Algebra 11, 45–100 (1982)
112. G. de B. Robinson, On the representations of Sn . Am. J. Math. 60, 745–760 (1938)
113. H.J. Ryser, Combinatorial Mathematics (Mathematical Association of America, Washington,
1963)
114. B.E. Sagan, The Symmetric Group, 2nd edn. (Springer, New York, 2001)
115. C.E. Schensted, Longest increasing and decreasing subsequences. Can. J. Math. 13, 179–191
(1961)
116. J. Schmid, A remark on characteristic polynomials. Am. Math. Mon. 77, 998–999 (1970);
Reprinted in Selected Papers on Algebra, ed. by S. Montgomery et al. (Mathematical
Association of America, Washington), pp. 332-333
117. M.-P. Schützenberger, A characteristic property of certain polynomials of E.F. Moore and
C.E. Shannon. RLE Quarterly Progress Report, No. 55 (Research Laboratory of Electronics,
M.I.T., 1959), pp. 117–118
118. C.A.B. Smith, W.T. Tutte, On unicursal paths in a network of degree 4. Am. Math. Mon. 48,
233–237 (1941)
119. E. Sperner, Ein Satz über Untermengen einer endlichen Menge. Math. Z. 27(1), 544–548
(1928)
120. E. Sperner, Über einen kombinatorischen Satz von Macaulay und seine Anwendung auf die
Theorie der Polynomideale. Abh. Math. Sem. Univ. Hamburg 7, 149–163 (1930)
121. R. Stanley, Cohen-Macaulay rings and constructible polytopes. Bull. Am. Math. Soc. 81,
133–135 (1975)
122. R. Stanley, The upper bound conjecture and Cohen-Macaulay rings. Stud. Appl. Math. 54,
135–142 (1975)
123. R. Stanley, Weyl groups, the hard Lefschetz theorem, and the Sperner property. SIAM J.
Algebr. Discrete Meth. 1, 168–184 (1980)
124. R. Stanley, Combinatorics and Commutative Algebra, 2nd edn. Progress in Mathematics, vol.
41 (Birkhäuser, Boston/Basel/Berlin, 1996)
125. R. Stanley, Differentiably finite power series. Eur. J. Combin. 1, 175–188 (1980)
126. R. Stanley, Quotients of Peck posets. Order 1, 29–34 (1984)
127. R. Stanley, Differential posets. J. Am. Math. Soc. 1, 919–961 (1988)
256 Bibliography
128. R. Stanley, Unimodal and log-concave sequences in algebra, combinatorics, and geometry,
in Graph Theory and Its Applications: East and West. Annals of the New York Academy of
Sciences, vol. 576 (1989), pp. 500–535
129. R. Stanley, Variations on differential posets, in Invariant Theory and Tableaux, ed. by D.
Stanton. The IMA Volumes in Mathematics and Its Applications, vol. 19 (Springer, New
York, 1990), pp. 145–165
130. R. Stanley, Enumerative Combinatorics, vol. 1, 2nd edn. (Cambridge University Press,
Cambridge, 2012)
131. R. Stanley, Enumerative Combinatorics, vol. 2 (Cambridge University Press, New York,
1999)
132. B. Stevens, G. Hurlbert, B. Jackson (eds.), Special issue on “Generalizations of de Bruijn
cycles and Gray codes”. Discrete Math. 309 (2009)
133. K. Sutner, Linear cellular automata and the Garden-of-Eden. Math. Intelligencer 11, 49–53
(1989)
134. J.J. Sylvester, On the change of systems of independent variables. Q. J. Math. 1, 42–56 (1857);
Collected Mathematical Papers, vol. 2 (Cambridge, 1908), pp. 65–85
135. J.J. Sylvester, Proof of the hitherto undemonstrated fundamental theorem of invariants. Philos.
Mag. 5, 178–188 (1878); Collected Mathematical Papers, vol. 3 (Chelsea, New York, 1973),
pp. 117–126
136. W.T. Trotter, Combinatorics and Partially Ordered Sets: Dimension Theory. Johns Hopkins
Studies in the Mathematical Sciences, vol. 6 (Johns Hopkins University Press, Baltimore,
1992)
137. R.J. Turyn, Character sums and difference sets. Pac. J. Math. 15, 319–346 (1965)
138. R.J. Turyn, Sequences with small correlation, in Error Correcting Codes, ed. by H.B. Mann
(Wiley, New York, 1969), pp. 195–228
139. W.T. Tutte, The dissection of equilateral triangles into equilateral triangles. Proc. Camb.
Philos. Soc. 44, 463–482 (1948)
140. W.T. Tutte, Lectures on matroids. J. Res. Natl. Bur. Stand. Sect. B 69, 1–47 (1965)
141. W.T. Tutte, The quest of the perfect square. Am. Math. Mon. 72, 29–35 (1965)
142. T. van Aardenne-Ehrenfest, N.G. de Bruijn, Circuits and trees in oriented linear graphs. Simon
Stevin (Bull. Belgian Math. Soc.) 28, 203–217 (1951)
143. M.A.A. van Leeuwen, The Robinson-Schensted and Schützenberger algorithms, Part 1: new
combinatorial proofs, Preprint no. AM-R9208 1992, Centrum voor Wiskunde en Informatica,
1992
144. F. Whipple, On a theorem due to F.S. Macaulay. J. Lond. Math. Soc. 28, 431–437 (1928)
145. E.M. Wright, Burnside’s lemma: a historical note. J. Comb. Theor. B 30, 89–90 (1981)
146. A. Young, Qualitative substitutional analysis (third paper). Proc. Lond. Math. Soc. (2) 28,
255–292 (1927)
147. D. Zeilberger, Kathy O’Hara’s constructive proof of the unimodality of the Gaussian
polynomials. Am. Math. Mon. 96, 590–602 (1989)
Index
132922799578491587290380706 B
0280344576, 157 Babai, Lászlo, 238
4483130665195087, 32, 41 balanced, see simplicial complex, balanced
balanced digraph, 152
balanced incomplete block design, 223
A basis matrix, 168
van Aardenne-Ehrenfest, Tanya, 159, 160 Bender, Edward Anton, 125
ACC, 237 Berlekamp, Elwyn Ralph, 238
access time, 23 Bernardi, Olivier, 143, 147
acts on (by a group), 43 BEST theorem, 159
acyclic (set of edges), 169 BIBD, 223
adjacency matrix, 1 Bidkhori, Hoda, 160
directed, 158 binary de Bruijn sequence, see de Bruijn
adjacent (vertices), 1 sequence
adjoint (operator), 36 binary sequence, 156
affine span, 188 Binet–Cauchy theorem, 136, 137
affine subspace, 188 binomial moment, 29
dimension, 188 bipartite graph, 8
affinely independent, 188 bipartition, 8
algebraic (Laurent series), 242 block (of a block design), 223
algebraic integer, 230 block design, 223
Anderson, Ian, 40 Bolker, Ethan David, 18
Andrews, George W. Eyre, 71 bond, 165
antichain, 33 bond space, 165
antisymmetry, 31 Boolean algebra, 31
ascending chain condition, 237 Borchardt, Carl Wilhelm, 147
automorphism Bose, Raj Chandra, 224, 238
of a graph, 98 Brenti, Francesco, 53
of a poset, 44 bridge, 183
automorphism group Bruggesser, Heinz, 213
of a graph, 98 de Bruijn, Nicolaas Govert, 97, 159
of a poset, 44 de Bruijn sequence, 156

https://doi.org/10.1007/978-3-319-77173-1
258 Index
de Bruijn graph, 157 convex hull, 188

Bruns, Winfried, 213, 239 convex set, 188
bump cotree, 170
in RSK algorithm for CSPP, 117 covers (in a poset), 31
in RSK algorithm for SYT, 113 cross (in a squared square), 182
Burnside’s lemma, 80, 97 cross-polytope, 216
Burnside, William, 97 CSPP, 117
cube (graph), 11
Curtin, Eugene, 238
C Cvetković, Dragoš M., 7
C-algebra, 232 cycle enumerator, 83
Caro, Yair, 238 cycle index polynomial, 83
Caspard, Nathalie, 40 cycle indicator, 97
Cauchy, Augustin Louis, 97 of a group of permutations, 83
Cauchy–Binet theorem, 136, 137 of a permutation, 82
Cauchy–Frobenius lemma, 80, 97 cycle space, 163
Cayley, Arthur, 147 cyclic group, 19
chain (in a poset), 32 cyclomatic number, 170
characteristic polynomial, 8 cyclotomic polynomial, 231
characteristic vector (of a set), 221
circuit, 164
circulant (matrix), 226 D
circulation, 163 d-ary de Bruijn sequence, 160
Clements, George F., 213 d-ary sequence, 160
closed walk, 4 Dedekind, Julius Wilhelm Richard, 17
coboundary, 164 DeDeo, Michelle Rose, 18
Cohen, Irvin Sol, 213 degree (of a vertex), 8, 21
Cohen–Macaulay face ring, see ring, face, Dehn–Sommerville equations, 216
Cohen–Macaulay deleted neighborhood (of a vertex), 240
Cohen–Macaulay simplicial complex, see depth, 204
simplicial complex, Cohen–Macaulay D-finite, 242
Collatz, Lothar, 7 Diaconis, Persi Warren, 18, 160
coloring, 76 diagonal (of a power series), 242
column-strict plane partition, 117 diagram (of a partition), 58
commutative algebra, 187 differentiably finite, 242
commutative diagram, 48 digraph, 151
complementary graph, 148 of a permutation, 144
complete bipartite graph, 8, 147 dihedral necklace, 86
complete graph, 4 Dilworth’s theorem, 246
complete tripartite graph, 240 Dilworth, Robert Palmer, 246
complete p-partite graph, 8 dimension
complexity (of a graph), 135 Krull, 203
compressed multicomplex, see multicomplex, of a face, 187
compressed direct product (of posets), 73
compressed simplicial complex, see simplicial directed graph, 151
complex, compressed divides (as a relation on monomials), 206
cone, 196, 217 Doob, Michael, 7
conjugate doubly-rooted tree, 144
of an algebraic number, 230 down (linear transformation), 36
partition, 58 Doyle, Peter Grant, 183
connected dual (of a planar graph), 178
digraph, 151 dunce hat, topological, 214
graph, 21, 135 Dynkin, Eugene (Evgenii) Borisovitsh, 71
Index 259
E Franzblau, Deborah Sharon, 124

edge reconstruction conjecture, 50 Frobenius, Ferdinand Georg, 17, 97, 124, 125
edge set (of a graph), 1 Fulton, William Edgar, 125
eigenvalues of a graph, 3 Fundamental Theorem of Algebra, 52
elementary cycle, 164 f -vector, 190
Engel, Konrad, 40
equivalent colorings, 75, 77
Erdős–Moser conjecture, 71 G
weak, 71 Gál, Anna, 238
Eriksson, Kimmo, 71 Gardner, Martin, 183
Euler characteristic, 199 Gauss’ lemma, 228
reduced, 199 Gaussian coefficient, 60
Euler phi-function, 85 generalized ballot sequence, 124
Euler’s constant, 220 generating function, 6, 79
Euler, Leonhard, 190 geometric realization, 189
Eulerian cycle (in a graph), 151 G-equivalent, 44
Eulerian digraph, 152 colorings, 77
Eulerian graph, 151 germ, 232
Eulerian tour Good, Irving John, 159
in a digraph, 152 Gordan’s lemma, 239
in a graph, 151 Gordan, Paul Albert, 239
e-vector (of a multicomplex), 206 graded poset, 32
extended Smith diagram, 180 Graham, Ronald Lewis, 18, 160, 238
graph, 1
Grassl, Richard, 124
Greene, Curtis, 124, 213
F group determinant, 17
face group reduction function, 97
missing, 201 Gubeladze, Joseph Soso, 239
of a simplex, 188
of a simplicial complex, 187
face (of a planar embedding), 178 H
face ring, see ring, face Hadamard matrix, 226
facet, 187 Hamiltonian cycle, 18
faithful action, 43 Hamming weight, 15
Fermat’s Last Theorem, 228 Harary, Frank, 97
Ferrers diagram, 58 Hardy, Godfrey Harold, 53
Fibonacci number, 131, 147, 232 Harper, Lawrence Hueston, 53
final vertex (of an edge) Hasse diagram, 31
in a digraph, 151 Hasse walk, 103
in an orientation, 139 Hawkins, Thomas W., 18
finitely generated (monoid), 235 Herzog, Jürgen, 213
Fishburn, Peter, 40 Hilbert basis theorem, 237
Fisher, Ronald Aylmer, 224, 238 Hilbert series, 202
Flores, A., 212 Hillman, Abraham, 124
flow, 163 hitting time, 23
Flye Sainte-Marie, Camille, 159 Hochster, Melvin, 213
Fomin, Sergey Vladimirovich, 124 homogeneous ideal, 206
forest, 145, 170 hook length formula, 105, 124
Forsyth, Andrew Russell, 125 Horn, Roger Alan, 27
Frame, James Sutherland, 124 Hughes, J. W. B., 71
Frame–Robinson–Thrall, 104 Hurwitz, Adolf, 97
Frankl, Peter, 238 h-vector (of a simplicial complex), 198
260 Index
I Laurent series, 242

i-face, 187 Leclerc, Bruno, 40
incidence matrix van Leeuwen, Marc A. A., 124, 125
Oddtown, 221 length
of a digraph, 166 of a chain, 32
of a graph, 139 of a necklace, 54, 85
incident, 1 of a walk, 2
indegree (of a vertex), 152 Leung, Ka Hin, 238
induced subgraph, 226 level (of a ranked poset), 32
initial vertex (of an edge) Lights Out Puzzle, 238
in a digraph, 151 Lindström, Bernt, 213
in an orientation, 139 linear monoid, 236
internal zero, 51 link (of a face of a simplicial complex), 211
inverse bump (in RSK algorithm), 115 Littlewood, Dudley Ernest, 124
isolated vertex, 23 Littlewood, John Edensor, 53
isomorphic Littlewood–Richardson rule, 124
graphs, 49 log-concave, 51
posets, 32 polynomial, 55
isomorphic monoids, 236 logarithmically concave, 51
isomorphism (of monoids), 236 loop
isomorphism class (of simple graphs), 49 in a digraph, 152
isthmus, 183 in a graph, 1
Lovász, László, 27, 53
Lubell, David, 33, 38, 40
J
Johnson, Charles Royal, 27
join (of simplicial complexes), 217 M
Joyal, André, 144, 147 Macaulay, Francis Sowerby, 206, 213
MacMahon, Percy Alexander, 124, 125
mail carrier, 156
K Mani, Peter, 213
Katona, Gyula O. H., 192, 213 Markov chain, 21
Kazarinoff, Nicholas D., 183 Matoušek, Jiří, 212, 238
Kind, Bernd, 213 matrix
Kirchhoff’s laws, 172 irreducible, 23
Kirchhoff, Gustav Robert, 147, 183 nonnegative, 23
Kishore, Shaunak, 160 permutation, 23
Kleinschmidt, Peter, 213 matrix analysis, 27
Kleitman, Daniel J., 213 Matrix-Tree Theorem, 141, 147
Knuth, Donald Ervin, 110, 125 matroid, 180
Krasikov, Ilia, 53 matroid complex, 214
Krattenthaler, Christian Friedrich, 124 maximal chain, 32
Kronecker, Leopold, 230 Menger, Karl, 189, 212
Krull dimension, see dimension, Krull middle levels conjecture, 18, 245
Kruskal, Joseph Bernard, 192, 213 Miltersen, Peter Bro, 238
Kruskal–Katona theorem, 192 minimal nonface, see nonface, minimal
Kummer, Ernst Eduard, 228 missing face, see face, missing
Kung, Joseph PeeSin, 18 Möbius function, 99
Möbius, August Ferdinand, 190
Monjardet, Bernard, 40
L monoid, 235
laplacian matrix, 139 commutative, 235
lattice, 57 Moon, John W., 147
lattice permutation, 124 Müller, Vladimír, 53
Index 261
multicomplex, 206 Parker, William Vann, 72

compressed, 212 part (of a partition of n), 57
multiple edge, 1 partial shelling, see shelling, partial
multiset, 1 partially ordered set, 31
multivariate notation, 237 partition
Mütze, Torsten, 18, 245 of a set X, 44
M-vector, 206 of an integer n, 57
Pascal’s triangle, 33, 62
q-analogue, 62
N path (in a graph), 135
n-cycle, 19 closed, 18
necklace, 54, 85 perfect squared rectangle, 180
neighborhood (of a vertex), 225 Perron–Frobenius theorem, 23
Newton, Isaac, 52, 53 physical intuition, 173
Nijenhuis, Albert, 124 Pitman, James William, 145, 147
N-matrix, 118 planar embedding, 177
no internal zero, 51 planar graph, 177
noetherian ring, 237 plane partition, 115
nonface, 201 history of, 125
minimal, 201 planted forest, 145
nonuniform Fisher inequality, 224 pole (of a Smith diagram), 180
non-zero-divisor, 203 Pollak, Henry Otto, 238
normal (ring), 244 Pólya, George (György), 53, 75, 97
normal monoid, 236 Pólya theory, 75
Novelli, Jean-Christophe, 124 polynomially recursive function, 232
nowhere-zero 5-flow conjecture, 184 poset, 31
NZD, 203 positive definite, 224
positive semidefinite, 37, 224
potential, 164
O potential difference, 165
octahedron, 190 Pouzet, Maurice André, 40, 53
Oddtown, 221 P -recursive function, 232
O’Hara, Kathleen Marie, 65, 71 primitive necklace, 98
Ohm’s law, 173 probability matrix, 21, 27
orbit, 44 Proctor, Robert Alan, 72
order (of a poset), 32 Prüfer sequence, 143
order (of a squared square), 182 Prüfer, Ernst Paul Heinz, 143, 147
order ideal of monomials, 206 pure, see simplicial complex, pure
order-matching, 34
explicit for Bn , 39
order-raising operator, 35 Q
orientation (of a graph), 139 q-binomial coefficient, 41, 60
oriented tree, 153 quantum order-matching, 36
orthogonal complement, 167 quotient poset, 45
orthogonal Lie algebra, 71
orthogonal subspace, 225
O-sequence, 206 R
outdegree (of a vertex), 152 Radon transform, 13
Radon, Johann Karl August, 18
rank
P of a Boolean algebra, 31
Pak, Igor M., 124 of a graded poset, 32
Palmer, Edgar Milan, 97 of a poset element, 32
parallel connection, 173 rank-generating function, 33
262 Index
rank-symmetric, 33 Simić, Slobodan, 7

rank-unimodal, 33 simple (squared square), 182
reciprocity theorem, 100 simple graph, 1
Redfield, John Howard, 75, 97 simple group, 99
reduced Euler characteristic, see Euler simplex, 188
characteristic, reduced dimension, 188
reduced incidence matrix, 140 simplices, 188
reflexivity, 31 simplicial complex, 187
region (of a planar embedding), 178 abstract, 187
regular graph, 22 balanced, 217
regular sequence, 204 Cohen–Macaulay, 205
Reisner, Gerald Allen, 210, 213 compressed, 194
Remmel, Jeffrey Brian, 124 geometric, 189
restriction pure, 195
of a facet in a shelling, 195 shellable, 195
of a simplicial complex, 214 Sinogowitz, Ulrich, 7
reverse lex order, 192 skew-symmetric (matrix), 240
on monomials, 207 slack variable, 237
reverse lexicographic order, 192 Smith diagram, 180
ring Smith, Cedric Austen Bardell, 159
face, 201 Snell, James Laurie, 183
Cohen–Macaulay, 205 solid partition, 123
Stanley–Reisner, 201 spanning subgraph, 49, 135
Robinson, Gilbert de Beauregard, 110, 124, spectral graph theory, 7
125 Sperner poset, 33
Robinson–Schensted correspondence, 110 Sperner property, 33
Roditty, Yehuda, 53 Sperner’s theorem, 33
Rolle’s theorem, 52 Sperner, Emanuel, 40, 213
root (of a tree), 145 squared rectangle, 180
Rosenberg, Ivo G., 40, 53 stabilizer, 46
row insertion, 113, 117 standard Young tableau, 104
Rowlinson, Peter, 7 Stanley, Richard Peter, 40, 53, 72, 97, 124, 125,
RSK algorithm, 110, 124 159, 213, 238
Ryser, Herbert John, 238 Stanley–Reisner ring, see ring, Stanley–Reisner
stationary distribution, 28
von Staudt, Karl Georg Christian, 147
S Stiemke’s theorem, 249
Sachs, Horst, 7 Stiemke, Erich, 249
Sagan, Bruce Eli, 125 Stirling number, signless of the first kind, 88
Schensted, Craige Eugene, 110, 124, 125 Stoyanovskii, Alexander V., 124
Schmid, Josef, 72 strongly connected digraph, 151
Schmidt, Bernard, 238 strongly log-concave, 51
Schützenberger, Marcel-Paul, 125, 192, 213 subcomplex (of a simplicial complex), 195
semidefinite, 37, 224 submonoid, 235
semigroup, 235 sum (of vector spaces), 234
series connection, 173 support (of a function), 168
series-parallel network, 174 support (of a monomial), 202
shape (of a CSPP), 117 suspension, 217
shellable, see simplicial complex, Sutner, Klaus, 225, 238
shellable switching (at a vertex), 54
shelling, 195 switching reconstructible, 55
partial, 214 Sylvester, James Joseph, 65, 71, 147
shelling order, 195 symmetric chain decomposition, 40
shifted Ferrers diagram, 120 symmetric function, 79
Index 263
symmetric plane partition, 133 V

symmetric sequence, 33 valid λ-word, 106
SYT, 104 van Kampen, Egbert Rudolf, 212
Velasquez, Elinor Laura, 18
vertex (of a geometric simplicial complex),
T 188
tensor product (of vector spaces), 234 vertex bipartition, 8
Thrall, Robert McDowell, 124 vertex reconstruction conjecture, 50
topological dunce hat, see dunce hat, vertex set, 1
topological vertex set (of a simplicial complex), 187
topological invariant, 199
total resistance, 173
totient function, 85 W
tour (in a digraph), 152 walk, 2
trace (of a plane partition), 133 in a digraph, 151
transitive (group action), 55, 80 Warshauer, Max, 238
transitivity, 31 weakly connected digraph, 151
transport, 48 weakly switching-reconstructible, 55
transposition, 99 weight
tree, 135 of a binary vector, 15
triangulation, 189 of a necklace, 54
Trotter, William Thomas, Jr., 40 Weitzenkamp, Roger, 183
Turyn, Richard Joseph, 227, 238 Wheatstone bridge, 174
Tutte, William Thomas, 159, 183, 184 Whipple, Francis John Welsh, 213
two-line array, 118 Wilf, Herbert Saul, 124
type wreath product, 63
of a Hasse walk, 104
of a permutation, 82
Y
Young diagram, 58
U Young, Alfred, 57, 124
unimodal sequence, 33 Young’s lattice, 57
unimodular (matrix), 171
universal cycle for Sn , 162
universality (of tensor products), 234 Z
unmixedness, 213 Zeilberger, Doron, 71, 124
up (linear transformation), 35 Zyklus, 83

Algebraic Combinatorics: Richard P. Stanley

Uploaded by

Copyright:

Available Formats

Algebraic Combinatorics: Richard P. Stanley

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

Algebraic Combinatorics: Richard P. Stanley

Uploaded by

Copyright:

Available Formats

Undergraduate Texts in Mathematics

Colin Adams, Williams College

Undergraduate Texts in Mathematics are generally aimed at third- and fourth-

More information about this series at http://www.springer.com/series/666

ISSN 0172-6056 ISSN 2197-5604 (electronic)

Library of Congress Control Number: 2018941202

Mathematics Subject Classification (2010): Primary 05Exx; Secondary 15-01

© Springer Science+Business Media New York 2013

Printed on acid-free paper

Cambridge, MA, USA Richard P. Stanley

This book is intended primarily as a one-semester undergraduate text for a course in

Cambridge, MA, USA Richard P. Stanley

Preface to the Second Edition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

Updated Preface to the First Edition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi

13 Miscellaneous Gems of Algebraic Combinatorics . . . . . . . . . . . . . . . . . . . . . . 219

Hints and Comments for Some Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

[n] The set {1, 2, . . . , n} for n ∈ N (so [0] = ∅)

Zn The group of integers modulo n

R[x] The ring of polynomials in the variable x with coefficients

YX For sets X and Y , the set of all functions f : X → Y

Fq The finite field with q elements

#S or |S| Cardinality (number of elements) of the finite set S

2S The set of all subsets of the set S

KS The vector space with basis S over the field K

Bn The poset of all subsets of [n], ordered by inclusion

ρ(x) The rank of the element x in a graded poset

[x n ]F (x) Coefficient of x n in the polynomial or power series F (x)

δij The Kronecker delta, which equals 1 if i = j and 0 otherwise

|L| The sum of the parts (entries) of L, if L is any array of

(λ) Length (number of parts) of the partition λ

p(n) Number of partitions of the integer n ≥ 0

ker ϕ The kernel of a linear transformation or group homomorphism

Sn Symmetric group of all permutations of 1, 2, . . . , n

ι The identity permutation of a set X, i.e., ι(x) = x for all x ∈ X

© Springer International Publishing AG, part of Springer Nature 2018 1

A walk in G of length from vertex u to vertex v is a sequence v1 , e1 , v2 , e2 , . . . ,

Proof. This is an immediate consequence of the definition of matrix multiplication.

where diag(λ1 , . . . , λp ) denotes the diagonal matrix with diagonal entries

(A(G) )ij = c1 λ1 + · · · + cp λp . (1.1)

In fact, if U = (urs ) is a real orthogonal matrix such that U −1 AU =

Proof. We have [why?]

fG () = λ1 + · · · + λp .

1.5 Proposition. The eigenvalues of the complete graph Kp are as follows: an

Proof. We have A(Kp ) = J − I , where I denotes the p × p identity matrix. If the

(Note that this is also the number of sequences (i1 , . . . , i ) of numbers 1, 2, . . . , p

A combinatorial proof of Corollary 1.6 is quite tricky (Exercise 1.1). Our

Again by the binomial theorem we have

Taking the (i, j )-entry of each side when i = j yields

(A(Kp ) )ii − (A(Kp ) )ij = (−1) .

We could also have deduced (1.5) from Corollary 1.6 using

α1 + · · · + αr = β1 + · · · + βs . (1.6)

Then r = s and the α’s are just a permutation of the β’s.

Notes for Chapter 1

The connection between graph eigenvalues and the enumeration of walks is

Exercises for Chapter 1

9. Let G be a (finite) graph with vertices v1 , . . . , vp and eigenvalues λ1 , . . . , λp .

(a) Show that ck (i, i) ≥ 0.

Find the eigenvalues of G in terms of those of G.

Let us now consider a more interesting example of a graph G, one whose

Now = {δ1 , . . . , δn } and δi · u is 1 if u has a one in its ith coordinate and is 0