Professional Documents
Culture Documents
Combinatorial Problems in Braid Groups: Name: Zaffar Iqbal Year of Admission: 2004 Registration No.: 56-GCU-PHD-SMS-04
Combinatorial Problems in Braid Groups: Name: Zaffar Iqbal Year of Admission: 2004 Registration No.: 56-GCU-PHD-SMS-04
Groups
Since 1864
Submitted to
Doctor of Philosophy
in
Mathematics
By
i
DECLARATION
ii
RESEARCH COMPLETION CERTIFICATE
----------------------------- -------------------------------
Date Supervisor
Dr. Barbu Berceanu
Submitted through
iii
Dedicated to my father
Inayat Hussain
iv
Table of Contents
Table of Contents v
Acknowledgements vii
Introduction 1
1 Preliminaries 4
1.1 Braids . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Braid Groups and Braid Monoids . . . . . . . . . . . . . . . . . . . . 6
1.3 Coxeter Groups and Artin Groups . . . . . . . . . . . . . . . . . . . . 7
1.4 Notations and Reduction Rules . . . . . . . . . . . . . . . . . . . . . 9
1.5 Growth function and the Hilbert series . . . . . . . . . . . . . . . . . 11
v
4.5 Spherical Artin Monoids . . . . . . . . . . . . . . . . . . . . . . . . . 62
Bibliography 65
vi
Acknowledgements
vii
Introduction
The braid group Bn+1 admits the classical presentation given by Artin in [2] in gener-
ators x1 , x2 , . . . , xn and relations xi xj = xj xi if |i − j| ≥ 2 and xi+1 xi xi+1 = xi xi+1 xi
if 1 ≤ i ≤ n − 1. Elements of Bn+1 are words expressed in the generators and their
inverses. The monoid MB n+1 with generators y1 , y2 , . . . , yn is defined by the presen-
tation of Bn+1 . Garside [15] proved that MB n+1 ,→ Bn+1 . With this identification
we call the elements of MB n+1 positive braids (see details in Chapter 1).
In 1972, P. Deligne [14] proved that the Hilbert series of all the Artin monoids are
rational functions, particularly: the Hilbert series of the braid monoid MB n+1 is a
rational function. In 1992, P. Xu [22] found the Hilbert series for the braid monoids
MB 3 and MB 4 and she also proved that the Hilbert series of MB n+1 is a rational
function. She developed a linear system for MB n+1 of size n! and she succeeded to
reduce it further.
Chapter 1 contains the basic notions of braids and braid monoids, in particular
Garside’s results [15]. Next, Coxeter Groups, Artin groups, spherical Artin groups
(monoids), and right-angled groups (monoids) are introduced. The terminology and
notions of reduction rules or rewriting system are included. The notions of the Hilbert
function, the Hilbert series, and the growth rate are also included in this chapter. This
chapter contains the results of Bokut [7] on the non-commutative Gröbner bases or
complete presentation of the braid monoid MB n+1 . The first chapter does not contain
original results.
1
2
construct a linear system of equations for reducible words as well as for irreducible
words. The size of the system is n2 + 2n − 3 which is much smaller than the size
n−1
2n−1 + 2[ 2
]
− 2 of Xu’s system for n ≥ 7. Using this system we show that the Hilbert
series for both reducible and irreducible words are rational and the Hilbert series for
the set of positive braids coprime with Garside element is rational. We also prove
that for a given word α the Hilbert series of the words starting with α is a rational
function. The results of this chapter are contained in the paper [16] which is accepted
for publication.
While studying the growth functions for MB 3 and MB 4 , we observe that the
growth rates are 1.618 and 2.0868 (approximately) respectively. We put ourselves a
natural question: whether the growth rate is bounded above or not, as n approaches
3
infinity? In Chapter 4 we prove that the growth rate is bounded above by 4, even
for a larger class of monoids, i.e., the spherical Artin monoids. In order to find these
estimations we use right angled Artin monoids (in particular a new family of Coxeter
graphs and corresponding Artin monoid, Kn∞ ). Also we give a general theorem on
the real roots of Chebyshev type families of polynomials. The results of this chapter
are contained in the paper [3] which has been submitted.
Chapter 1
Preliminaries
1.1 Braids
braid is a set of n + 1 smooth paths connecting the chosen points on the first line
with the points on the second line (in arbitrary order), such that any plane z = t
points in the plane R2 . Assume that zk = (k, 0) for all 1 ≤ k ≤ n + 1. Define a braid
such that
• there exists a permutation χ = θ(β) ∈ Sn such that bk (1) = zχ(k) for all 1 ≤ k ≤ n+1;
For example,
4
5
z1 z2 z3
• • •
@
@
@
@
@
@ @
@ @
@ @
@ @
• @• @•
z1 z2 z3
is a 3-strand braid. The set of all (n + 1)-strand braids generates a group Bn+1 under
the operation of product. The product is concatenation of braids and rescaling the
z-coordinates. The ith element xi of this group is given by
··· ··· e
• • • •
6
there is an obvious map ψ : MB n+1 → Bn+1 given by ψ(yi ) = xi and the image of
this map is the submonoid of positive words. Garside [15] proved the basic fact:
Now we can identify the elements of MB n+1 with their images in Bn+1 and we
call these elements positive braids.
Definition 1.2.1. In the braid group Bn+1 , there is a fundamental positive braid
∆n+1 (or simply ∆) which is defined by ∆ = (x1 )(x2 x1 ) · · · (xn xn−1 · · · x1 ) . The ∆ is
z1 z2 z3 zn+1
• • • ··· •
@
@
@
@
@
@ @
@ @
@ @
@ @
@ @
@ @ @
@ @ @
@ @ @
@ @ @
• · · · @• @• @•
Fig. 1.1
Chow [12] proved that ∆2 generates the center of Bn+1 , n ≥ 2. Geometrically, Garside
braid is a half twist of the identity braid. This braid has some nice properties given
by Garside in [15].
1) ∆xi = xn+1−i ∆, 1 ≤ i ≤ n.
2) For each j (1 ≤ j ≤ n), D(∆) (the set of words equivalent to ∆) contains a word
Definition 1.3.1. Let M = (mst )s,t∈S be the Coxeter matrix of the Coxeter graph Γ.
(An )n≥1 : • • • • •
x1 x2 x3 ··· xn−1 xn
(Bn )n≥2 : • 4 • • • •
x1 x2 x3 ··· xn−1 xn
• xn−1
(Dn )n≥4 :
• • • •
x1 x2 x3 · · · xn−2
H
HH
• xn
• x4
F4 : 4
• • • •
x1 x2 x3 x4
G2 : • 6 •
x1 x2
9
(Hn )n=3,4 : 5 5
• • • • • • •
x1 x2 x3 x1 x2 x3 x4
p
I2 (p) p≥5,p6=6 : • •
x1 x2
Definition 1.3.2. If W is a Coxeter Group (with the Coxeter matrix M = (mst )s,t∈S
Definition 1.3.3. In the spherical type Coxeter graphs, if all the labels mst ≥ 3
are replaced by ∞ then the associated groups (monoids) are called right-angled Artin
groups (monoids).
Let |W | denotes the length of a positive word W . In a free monoid Fn [S] generated
by S = {x1 , x2 , . . . , xn } the total order on the set of generators given by x1 < · · · < xn
is extended to length-lexicographic order. A relation R is written in the form ai = bi
where ai is a monomial greater than bi . If ai | aj then i = j (and also ai - bj ). We
denote by ai (R) and bi (R) the terms ai and bi respectively of the given relation R.
In a monoid, a word containing the L.H.S. of a relation is called reducible word
and a word which does not contain the L.H.S. of a relation is called irreducible word.
The irreducible words are also known as canonical forms. We denote A[k] by the set
of irreducible words and by B [k] the set of reducible words in MB k .
• Let S1 and S2 (not necessarily different) be relations in MB k . Then the word of
the form a(S1 ) × a(S2 ) is called an ambiguity.
W1
We define two reduction rules (left and right reductions):
(
W, if ai (R) - W for all relations R.
L(W ) =
L(γbj δ), if W = γaj δ and if W = εak η then |γ| < |ε|,
(
W, if ai (R) - W for all relations R.
R(W ) =
L(γbj δ), if W = γaj δ and if W = εak η then |δ| < |η|.
The reduction rule γai δ → γbi δ gives a unique canonical form if all the ambiguities are
solvable (for any overlap ai = γi δ, aj = δj , δ 6= 1, reductions of (γi δ)j → bi j → · · ·
and γi (δj ) → γi bj → · · · gives the same result). The Diamond lemma (see [5] or
[13]) says that if all the ambiguities are solvable, then the set of irreducible words is
in bijection with the monoid with the presentation hx1 , x2 , . . . , xn : ai = bi , i ∈ Ii (I
could be an infinite set). A presentation of a monoid is complete if and only if all the
ambiguities are solvable.
We denote by α(i, j) = α(xi , xi−1 , . . . , xj ) an arbitrary irreducible word (possibly
empty) in xi , xi−1 , . . . , xj and α(i, i) = α(xi ). We denote the “shift” of α(i, j) by
1) xs xk = xk xs , s − k ≥ 2,
2) xi+1 xi α(i − 1, 1)β(i, j)xi+1 xi xi−1 · · · xj = xi xi+1 xi α(i − 1, 1)xi xi−1 · · · xj Σβ(i, j),
1 ≤ i ≤ n − 1, 1 ≤ j ≤ i + 1.
For a sequence {sk }k≥1 of positive numbers, we define the growth rate by:
Definition 1.5.1. We say that {sk }k≥1 has a growth rate less than γ (γ is a real
12
number) if
log s
k
lim exp < γ.
k k
This is equivalent to the fact that there exists γ0 < γ and c > 0 such that sk < c γ0k
n
βi αik has a growth rate
P
for all k. For instance, if max |αi | < γ, the sequence ck =
1≤i≤n i=1
less than γ. In this case we have also ck+1
ck+1
ck
< γ for k 0; if the sequence ck k
is convergent, its limit is called the growth ratio of the sequence {ck }. See [21] for a
general discussion of growth functions.
We have mentioned that P. Deligne [14] proved that the Hilbert series of all the
Artin monoids are rational functions, particularly:
Theorem 1.5.1. [14] The Hilbert series of the braid monoid MB n+1 is a rational
function.
Chapter 2
2.1 Introduction
Our method is to find the Garside elements of the spherical Artin monoids by induc-
tion. For this purpose we introduce and characterize relative Garside element. We
compute this relative Garside element using its characterization. Using the relative
Garside element we describe completely the Garside elements of all the spherical Artin
monoids. In last section we give the reduced complete presentation of the monoid
MB n+1 as a consequence of Theorem 1.4.1. We give another proof of Theorem 1.4.1.
The complete presentations of the monoids M(Bn ) and M I2 (p) are also given at
the end.
13
14
∆ I2 (4k) = (x1 x2 · · · x2 )2 .
| {z }
2k times
The image of ∆(Γ) in the corresponding Coxeter group W (Γ) is the longest element
of this group and it has order two [9].
We will use the following generalization of Garside lemma [15]:
Suppose we have an inclusion of Coxeter graphs Γn−1 ⊂ Γn with the set of vertices
{x1 , . . . , xn−1 }, respectively {x1 , . . . , xn }. Using the general properties of Garside
elements we have
∆(Γn−1 ) |L ∆(Γn ).
Proposition 2.2.2. The relative Garside element Tn for Γn−1 ⊂ Γn satisfies the
following properties:
15
1) Tn is square free;
2) xi |L Tn ⇔ i = n;
4) xj |R Tn ⇔ j = m.
∆n−1 xi Tn = xσn−1
−1
(i) ∆n−1 Tn = xσn−1
−1
(i) ∆n = ∆n xσn ◦σn−1
−1
(i) = ∆n−1 Tn xσn ◦σn−1
−1
(i) ,
−1
therefore xi Tn = Tn xσ(i) , where σ = σn ◦ σn−1 . The image of σ contains the elements
4) Let xj 6= xm then there exists i ∈ {1, . . . , n−1} such that j = σ(i) and xi Tn = Tn xj .
xi Tn is a divisor of ∆n .
1) x2m - Un ;
2) xi |L Un ⇔ i = n;
and this contradicts 2). If j = m, then Un = (Sxm )xm W which contradicts 1).
1) xn |L Un ;
Elements
Let Un = (xn xn−2 xn−3 · · · x1 )(xn−1 xn−2 · · · x2 )(xn xn−2 xn−3 · · · x3 )(xn−1 xn−2 · · · x4 ) · · · .
For Un we have two cases: n = even and n = odd. Let n be even, hence
1) First we prove that Un is square free and also that xn xn−1 is not a divisor of Un .
• xn−1
Dn : • • •
x1 x2 · · · xn−2
HH
H • xn
↓
An−1 : • • • •
y1 y2 · · · yn−2 yn−1
The image of Un under the map pr is ∆n−1 which is square free. Now if x2j | Un ; then
2
then pr(xn xn−1 ) = yn−1 | ∆n−1 , a contradiction.
• xn−2
Dn−1 : • • •H
(x1 < · · · < xn−2 < xn−1 )
x1 x2 · · · xn−3 HH
• xn−1
↓
• xn
Dn−1 : • (x2 < · · · < xn−2 < xn < xn−1 )
• •
x2 x3 · · · xn−2
H
HH
• xn−1
Tn−1 becomes (xn−1 xn−2 xn−3 · · · x2 )(xn xn−2 xn−3 · · · x3 )(xn−1 xn−2 xn−3 · · · x4 ) · · · . As
this is true for i ∈ {2, 3, . . . , n, n − 1}) and also x1 |L Tn−1 is impossible because Tn−1
contradiction.
(ii) If xn−2 |L Un then Lemma 2.2.1 implies that xn xn−2 xn α2 = xn xn−2 xn−3 · · · x1 Tn−1
⇒ xn α3 = Tn−1 , a contradiction.
⇒ xi β2 = xi+1 xi · · · x1 Tn−1
Tn = (xn xn−2 xn−3 · · · x1 )(xn−1 xn−2 · · · x2 )(xn xn−2 xn−3 · · · x3 )(xn−1 xn−2 · · · x4)· · ·(xn−1).
Corollary 2.3.2.
1) clearly x2 |L U2 (p);
x2 x1 x2 · · · (p factors) = x1 x2 x1 · · · (p factors).
So the length of V is greater than or equal to p − 1; hence U2 (p) has minimal length
Corollary 2.3.3.
∆ I2 (p) = x1 x2 x1 x2 · · · (p factors).
20
Example 2.3.4. We also give (without proof ) the relative Garside elements (and
∆(F4 ) = x1 x2 x1 T3 x4 T3 x4 T3 x4
x2 T3 = T3 x1 . Therefore
∆(H3 ) = x1 x2 x1 x2 x1 · x3 x2 x1 x2 x1 x3 x2 x1 x2 x3
For the reduced complete presentation we are using the following notations for the
1 ≤ j ≤ n − 2.
Theorem 1.4.1.
Corollary 2.4.1. The braid monoid MB n+1 have a reduced complete presentation
x1 , . . . , xn | R[n+1] .
Proof. (of Corollary 2.4.1 and also another proof of Theorem 1.4.1)
complete. In first step we describe how the relations R[n+1] are obtained from the
defining relations and ambiguities between them. In second step we show that all
[n+1]
the new ambiguities are solvable. For simplicity we will write R∗ instead of R∗ .
R∅,2 R∅,2 , R∅,2 R∅,1i ;1 , R∅,1i ;1 R∅,2 , R∅,1i ;1 R∅,1i ;1 , R∅,1i ;1 R∅,1j ;1 , R∅,1j ;1 R∅,1i ;1 .
(we are not mentioning the overlap in the ambiguities) First we show that the relations
x1 x2 x21 x2 , R(R∅,2 R∅,2 ) = x2 x21 x2 x1 , i.e., R∅,2 R∅,2 gives the relation R∅,12 ;1 . The
ambiguity R∅,12 ;1 R∅,2 gives R∅,13 ;1 and in general R∅,1r−1 ;1 R∅,2 gives R∅,1r ;1 which
are satisfied in MB 3 . Now we show that all other ambiguities are solvable. In the
solvable. Generally for any i and j, in the ambiguity R∅,1i ;1 R∅,1j ;1 : x2 xi1 x2 xj1 x2 x1 ;
j−1
L(R∅,1i ;1 R∅,1j ;1 ) = L(x1 x2 x21 xi−1
2 x1 x2 x1 )
and
2 j−1 j−1
R(R∅,1i ;1 R∅,1j ;1 ) = R(x2 xi+1 2 i
1 x2 x1 x2 ) = R(x1 x2 x1 x2 x1 x2 )
Therefore all the ambiguities are solvable. Hence the above reduced presentation is
complete. Now suppose the relations are true for any n. We prove them for n + 1 in
Step 1. i) Using induction we show that the relations Rα;n are satisfied in MB n+1 .
The ambiguity R∅;n Rn(n−2) gives the relations Rxn−2 ;n , Rxn−2 ;n Rn(n−2) gives Rx2n−2 ;n
and so on Rxrn−2 ;n Rn(n−2) gives Rα(n−2);n . The ambiguity Rα(n−2);n Rn(n−3) gives the
ity Rα(n−2;2);n Rn.1 gives Rα(n−2,2)x1 ;n , Rα(n−2,2)x1 ;n Rn.1 gives Rα(n−2,2)x21 ;n and so on
Rα(n−2,2)xt1 ;n Rn.1 gives the relations Rα;n which are satisfied in MB n+1 .
ii) The ambiguity Rα,n R∅;n gives Rα,xn−1 ;n−1 , Rα,xn−1 ;n−1 R∅;n gives Rα,x2n−1 ;n−1 and
so on Rα,xrn−1 ;n−1 R∅;n gives the relations Rα,β(n−1);n−1 which are satisfied in MB n+1 .
iii) Using induction again we show that the relations Rα,β;1 are satisfied in MB n+1 .
Rα,β(n−1)xn−2 ;n−1 R∅;n−1 gives Rα,β(n−1)x2n−2 ;n−2 and so on Rα,β(n−1)xsn−2 ;n−1 R∅;n−1 gives
Rα,β(n−1,n−2);n−2 . Continuing the above process, the ambiguity Rα,β(n−1,2);2 R∅;2 gives
Step 2. Here we show that all other ambiguities are solvable. In the ambiguity
L Rα;n Rγ;n = L xn−1 xn xn−1 αxn−1 γ(n − 2, 1)xn ,
R Rα;n Rγ;n = R xn xn−1 αxn−1 xn xn−1 γ(n − 2, 1)
= R xn−1 xn xn−1 αxn−1 xn γ(n − 2, 1)
= R xn−1 xn xn−1 αxn−1 γ(n − 2, 1)xn .
R Rα;n Rγ,δ;j = R xn xn−1 αxn−1 xn xn−1 γ(n − 2, 1)xn−1 · · · xj Σδ(n − 1, j)
= R xn−1 xn xn−1 αxn−1 xn γ(n − 2, 1)xn−1 · · · xj Σδ(n − 1, j)
= R xn−1 xn xn−1 αxn−1 γ(n − 2, 1)xn xn−1 · · · xj Σδ(n − 1, j)
= R xn−1 xn xn−1 α xn−1 γ(n − 2, 1) δ(n − 1, j)xn xn−1 · · · xj .
and
R Rα,β;j Rγ,δ;i
= R xn xn−1 αβ(n−1,j)xn · · · xk+1 xk−1 xk · · · xj γ(k−2,1)xk−1 · · · xi Σδ(k−1, i)
= R xn xn−1 α{β(n − 1, j)xk−1 }xn · · · xj γ(k − 2, 1)xk−1 · · · xi Σδ(k − 1, i)
24
= R xn−1 xn xn−1 α xn−1 · · · xj Σβ(n − 1, j)xk γ(k − 2, 1)xk−1 · · · xi Σδ(k − 1, i)
= R xn−1 xn xn−1 α xn−1 · · · xj Σβ(n − 1, j)γ(k − 2, 1)xk · · · xi Σδ(k − 1, i)
= R xn−1 xn xn−1 α xn−1 · · · xj Σβ(n − 1, j)γ(k − 2, 1)δ(k − 1, i)xk · · · xi .
Therefore this ambiguity is solvable too. Hence the reduced presentation of MB n+1
is complete.
Next we give the complete presentations of the monoids M(Bn ) and M I2 (p) .
We are using the following notations for the relations in M(Bn ) and for α∗ (n − 2, 1)
we will write α∗ .
x1 , . . . , xn | R[n] .
Proof. We will check here few ambiguities only. For example in the ambiguity
[n] [k]
Rn;ii Rk;jj : xn xn−1 α1 β1 (n−1, 2)α2 (i−1, 1)β2 (n−1, i+1)xn · · · xk+1 xk · · · xi · · · x1 · · · xi
xk−2 · · · x1 · · · xi α3 (k−2, 1)xk−1 · · · x1 Σβ3 (k−1, 2)x2 · · · xj α4 (j−1, 1)Σβ4 (k−1, j+1)
= R xn xn−1 α1 β1 (n−1, 2)α2 (k−1, 1)β2 (n−1, k+1)xn · · · x1 · · · xk+1 xk−1 xk xk−1 xk−2 · · ·
x1 · · · xk−1 x2k α3 (k − 2, 1)xk−1 · · · x1 Σβ3 (k − 1, 2)x2 · · · xi α4 (i − 1, 1)Σβ4 (k − 1, i + 1)
Therefore this ambiguity is solvable too. Hence the above presentation is complete.
Theorem 2.4.3. The complete presentations of the monoid M I2 (p) are
1) M I2 (2k + 1) = x1 , x2 | P2k+1 , x2 xr1 ∆ = ∆x1 xr2 , r ≥ 1 and
2) M I2 (2k) = x1 , x2 | P2k , x2 xr1 ∆ = ∆x2 xr1 , r ≥ 1 , where
Pl : x2 x1 x2 x1 · · · (l factors) = x1 x2 x1 x2 · · · (l factors).
Proof. 1) Let
and
= R x2 xr1 ∆ x1 x2 x1 · · · x2 x1 xs−1
| {z } 2
2k−1 factors
0
= R ∆x1 xr2 ∆0 xs−2 ∆ = x x · · · x and x r 0
∆ = ∆ 0 r
x
2 | 1 2{z }2 2 1
2k factors
= R ∆x1 ∆0 xr1 xs−2
2
3.1 Introduction
In this chapter first we show that the growth functions of the monoids MB 3 and MB 4
are exponential and we find the corresponding growth rates. Using the reduced com-
(see Chapter 1) we construct a linear system (other than Xu’s system) of equations for
the reducible as well as for the irreducible words to compute inductively the Hilbert
series of MB n+1 . The size of the system is n2 + 2n − 3 which is much smaller than the
n−1
size 2n−1 + 2[ 2
]
− 2 of Xu’s system for n ≥ 7. Using this system we show that the
Hilbert series of both reducible and irreducible words are rational. We give another
positive braids coprime with ∆) is rational. At the end we prove the Theorem 3.1.2,
a generalization of Theorem 1.5.1 which states that for a given word α the Hilbert
28
29
related with the prefix (beginning) of a word and ω is related with the suffix (end)
[n+1]
of the word. For example Ak(k−1)···i denotes the set of irreducible words in MB n+1
[n+1]
starting with xk xk−1 · · · xi ; Bj,k denotes the set of reducible words starting with
xn xn−1 · · · xj and ending with xn xn−1 · · · xk . As special cases we use the following
notations: if j = ∗ then the word will start with xn xn−1 and if j = n − 1 then the
word will start with xn x2n−1 . Also a special reducible word xk xk−1 xk is denoted by
[n+1]
B∅,k . We will use ia j b k c · · · for a word xai xbj xck · · · , when required.
All the above sets are graded by length, so we can introduce the Hilbert series of
[n+1] [n+1] [n+1] [n+1]
these sets. Let Qα,ω (t) denotes the Hilbert series of Bα,ω , Pα (t) of Aα and
[n+1] ` [n+1] ` [n+1] ` ` [n+1]
HM (t) of A[n+1] = {e} A1 A2 · · · An for the monoid M .
Let MB +
n+1 = MB n+1 r ∆MB n+1 be the set of positive braids coprime with ∆.
a
Bn+1 = ∆k MB +
n+1
k∈Z
the word problem in Bn+1 is reduced to the word problem in MB n+1 and the compu-
tation of ∆ exponent. There is an obvious but lengthy algorithm to solve the word
ter 2. Using this presentation, we will compute inductively the Hilbert series of
MB n+1 and MB +
n+1 and give another proof of Deligne’s theorem. As a consequence
function.
[n+1]
Let α be an irreducible word in MB n+1 and Aα be the set of all irreducible
30
words in MB n+1 starting with α then we prove the following generalization of the
Theorem 1.5.1:
[n+1]
Theorem 3.1.2. The Hilbert series of Aα is a rational function.
In Section 3.2 complete results are given for MB 3 and MB 4 , i.e., we compute the
Hilbert series for these monoids and show that the growth functions of these series
are exponential. In this section we prove Theorem 1.5.1 and the rationality of some
Firstly we present in detail the main idea of the general proof for the simplest case:
[3]
HMB (t).
[3] 1
HMB (t) = .
(1 − t)(1 − t − t2 )
Proof. To obtain this series we will compute the Hilbert series for the reducible words
[2]
as well as for the irreducible words. The Hilbert series for A1 = {x1 , x21 , x31 , . . .} is
given by
[2] t
P1 = t + t2 + t3 + · · · = .
1−t
[3]
First we compute the Hilbert series for the reducible words. As B∅,2 contains only
[3]
one word x2 x1 x2 , we have Q∅,2 = t3 . The reduced complete presentation gives the
[3] [2] [3] t5
decomposition B1,1 = {x2 x1 } × A1 × {x2 x1 }, hence Q1,1 = 1−t
.
Now we compute the Hilbert series for the irreducible words. If w is an irreducible
word then the product x1 w is also irreducible. Thus we have the decomposition
31
[3] t t [3]
P1 = + P . (3.2.1)
1−t 1−t 2
If we multiply x2 on the left with an irreducible word starting with x1 then we can
[3] [3]
have the reducible words of the form B∅,2 = {x2 x1 x2 } or B1,1 = {x2 xa1 x2 x1 }. In order
to obtain irreducible words starting with x2 x1 we have to delete the above reducible
[3]
words from {x2 } × A1 . Thus
a [3]
[3] [3] [3] [3] [3]
A21 = {x2 } × A1 r B∅,2 × A2 B1,1 × A21
2 21
[3]
and the set in the parenthesis is contained in {x2 } × A1 . Therefore
[3]
Remark 3.2.2. One can see that the coefficients ak in the series are related with
Fibonacci numbers F0 = 1, F1 = 1, F2 = 2, F3 = 3, F4 = 5, F5 = 8, . . . by the
[3]
relation ak = Fk+2 − 1.
Remark 3.2.3. As
√ √
1 −1 5−2 5 5+2 5
= + +
(1 − t)(1 − t − t2 ) 1 − t 5(1 + c1 t) 5(1 − c2 t)
√ √
5−1 5+1
where c1 = 2
and c2 = 2
. The first two terms have a negligible contribution in
In the next case, n = 3, all the technicalities of the general case appear; this is
[4] 1
HMB (t) = .
(1 − t)(1 − 2t − t2 + t3 + t4 + t5 )
Proof. As in Proposition 3.2.1, we start to compute the series for the reducible words,
[4]
i.e., B∗,∗ . Using the reduced complete presentation we have only two types of reducible
[4] [4] ` [4] [4] [2]
words ending with x3 , i.e., B∗,3 = Bφ,3 B1,3 where B1,3 = {x3 x2 } × A1 × {x3 } and
[4] [4] [4] [4] t4 t3
Bφ,3 = {x3 x2 x3 }. Hence Q∗,3 = Qφ,3 + Q1,3 = t3 + 1−t
= 1−t
.
Again using reduced complete presentation we have only two types of reducible
[4] [4] ` [4] [4] [2]
words ending with x2 , i.e., B∗,2 = B2,2 B1,2 where B2,2 = {x3 x2 } × ΣA1 × {x3 x2 }
[4] [2] [2] [4] t5 t7 t5 (1−t+t2 )
and B1,2 = {x3 x2 x1 } × A1 × ΣA1 × {x3 x2 }. Hence Q∗,2 = 1−t
+ (1−t)2
= (1−t)2
.
33
arguments we have
[4] [3] [4] ` [4]
B1,1 = {x3 x2 x1 } × A21 × {x3 x2 x1 } r B1,3 × {x2 x1 } B1,2 × {x1 } and so
21
[4] t10 [4] t7 t10 t7 (1−t+t3 )
Q1,1 = (1−t)2 (1−t−t2 )
. Hence Q∗,1 = (1−t)(1−t−t2 )
+ (1−t)2 (1−t−t2 )
= (1−t)2 (1−t−t2 )
.
[4]
In the second part we compute the series for the irreducible words, A∗ . We
[4] [3] ` [3] [4] [4] [3] ` [3] [4]
have the following decompositions: A1 = A1 A1 × A3 , A2 = A2 A2 × A3 ,
[4] [3] ` [3] [4] [4] ` [4] ` [4]
A21 = A21 A21 ×A3 and A3 = {x3 } {x3 }×A3 A32 and their corresponding
If we multiply x3 on the left with an irreducible word starting with x2 then we can
[4] [3] [3]
have the reducible words of the form B∗,3 , B∗,2 or B∗,1 . In order to obtain irreducible
[3]
words starting with x2 x1 we have to delete the above reducible words from {x3 }×A2 .
Thus
a [4] a [4]
[4] [4] [4] [4] [4] [4]
A32 = {x3 } × A2 r B∗,3 × A3 B∗,2 × A32 B∗,1 × A321
3 32 321
[3]
and the set in the parenthesis is contained in {x3 } × A2 . Therefore
neously we get
[4] t [4] t(1 + t) [4] t2
P1 = , P2 = , P21 = ,
(1 − t)T5 T5 T5
[4] t(1 − t2 − t3 − t4 ) [4] t2 (1 − t2 − t3 ) [4] t3 (1 − t − t2 )
P3 = , P32 = and P321 = .
T5 T5 T5
` [4] ` [4] ` [4]
Now A[4] = {e} A1 A2 A3 implies
[4]
Remark 3.2.5. For HMB (t) we have
1 1 .515 .269+.296t 2.638 2.56
(1−t)(1−2t−t2 +t3 +t4 +t5 )
≈ 1−t
+ 1+.85t
+ 1+.493t+.449t2
− 1−1.256t
+ 1−2.0868t
.
2.56
= 2.56 1 + 2.0868t + (2.0868)2 t2 + (2.0868)3 t3 + · · · .
1 − 2.0868t
[4] [4]
Hence ak > 2.56(2.0868)k . Thus in this case the growth function of ak is exponential
In the next lemma we construct a linear system for the reducible words in MB n+1 .
Lemma 3.2.6. The following relations hold for the reducible words in MB n+1 .
n−1
[n+1] [n] P j−1 [n+1]
1) Qn−1,1 = tn+2 Pn−1 − t Qn−1,j .
j=2
35
[n+1] [n−1]
2) Qn−2,n = t3 Pn−2 .
[n+1] [n−1] [2] [n] [2]
3) Qn−2,n−1 = t4 Pn−2 P1 − t2 Q∗,n−1 P1 .
n−1
P j−i [n+1] n−1
[n+1] [n−1] [n−i+1] [n] [n−i+1]
4) Qn−2,i = tn−i+3 Pn−2 Pn−i t Qn−2,j − tj−i+2 Q∗,j P(n−i)···(j−i+1) ,
P
−
j=i+1 j=i
i = 1, . . . , n − 2 .
Proof. Here we are using again the reduced complete presentation of MB n+1 .
[n]
1) The set {xn xn−1 } × An−1 × {xn xn−1 · · · x1 } can be written as a disjoint union of
[n+1] [n+1] [n+1]
the sets Bn−1,1 , Bn−1,2 × {x1 }, . . . , Bn−1,n−1 × {xn−2 · · · x1 }. Therefore
[n+1]
xn x2n−1 β(n − 1, 1)xn · · · x1
Bn−1,1 =
[n]
= {xn xn−1 } × An−1 × {xn xn−1 · · · x1 }
a a
[n+1] [n+1]
r Bn−1,2 × {x1 } Bn−1,3 × {x2 x1 } ···
!
a [n+1]
a a [n+1]
Bn−1,j × {xj−1 · · · x1 } ··· Bn−1,n−1 × {xn−2 · · · x1 }
implies
[n+1] n+2 [n] [n+1] 2 [n+1] j−1 [n+1] n−2 [n+1]
Qn−1,1 =t Pn−1 − tQn−1,2 +t Qn−1,3 +···+t Qn−1,j +···+t Qn−1,n−2
or
n−1
X
[n+1] n+2 [n] [n+1]
Qn−1,1 =t Pn−1 − tj−1 Qn−1,j .
j=2
[n+1] [n−1]
2) Bn−2,n = xn xn−1 xn−2 α(n − 2, 1)xn = {xn xn−1 xn−2 } × An−2 × {xn } implies
n−2
[n+1] [n−1]
Qn−2,n = t3 Pn−2
36
3)
[n+1]
Bn−2,n−1 = xn xn−1 xn−2 α(n − 2, 1)β(n − 1)xn xn−1
[n−1] [2]
= {xn xn−1 xn−2 } × An−2 × Σn−2 A1 × {xn xn−1 }
n−2
[n] [2]
r{xn } × B∗,n−1 × Σn−2 A1 × {xn xn−1 }
n−1
implies
[n+1] [n−1] [2] [n] [2]
Qn−2,n−1 = t4 Pn−2 P1 − t2 Q∗,n−1 P1 .
[n+1]
4) Bn−2,i = xn xn−1 xn−2 α(n − 2, 1)β(n − 1, i)xn · · · xi
[n−1] [n−i+1] [n+1]
= {xn xn−1 xn−2 } × An−2 × Σi−1 An−i × {xn · · · xi }r Bn−2,i+1 ×{xi }
n−2
!
a a [n+1] a a
[n+1]
··· Bn−2,j × {xj−1 · · · xi } ··· Bn−2,n−1 × {xn−2 · · · xi }
[n] [n−i+1]
r {xn } × B∗,i × Σi−1 A(n−i)···1 × {xn · · · xi }
(n−1)···i
a [n] [n−i+1]
a
{xn } × B∗,i+1 × Σi−1 A(n−i)···2 × {xn · · · xi } ···
(n−1)···(i+1)
a [n]
a
[n−i+1]
i−1
{xn } × × Σ
B∗,j × {xn · · · xi } ···
A(n−i)···(j−i+1)
(n−1)···j
!
a [n] [n−i+1]
{xn } × B∗,n−1 × Σi−1 An−i × {xn · · · xi }
n−1
implies
[n+1] n−i+3 [n−1] [n−i+1] [n+1] j−i [n+1] n−i−1 [n+1]
Qn−2,i =t Pn−2 Pn−i − ··· + t
tQn−2,i+1 + + ··· + t Qn−2,j Qn−2,n−1
[n] [n−i+1] [n] [n−i+1] [n] [n−i+1]
− t2 Q∗,i P(n−i)···1+ · · · + tj−i+2 Q∗,j P(n−i)···(j−i+1)+ · · · + tn−i+1 Q∗,n−1 Pn−i
or for i = 1, . . . , n − 2,
n−1
X n−1
X
[n+1] n−i+3 [n−1] [n−i+1] j−i [n+1] [n] [n−i+1]
Qn−2,i =t Pn−2 Pn−i − t Qn−2,j − tj−i+2 Q∗,j P(n−i)···(j−i+1) .
j=i+1 j=i
37
[n+1]
Proposition 3.2.9. The Hilbert series Q∗,∗ of the reducible words in MB n+1 is
[m] [m]
rational in P∗ and Q∗,∗ , m ≤ n.
38
Proof. We prove that the system BX = C (constructed in Lemma 3.2.6, 3.2.7 and
[n+1]
3.2.8) of n(n − 1) equations with variables Qi,j , 1 ≤ i ≤ n − 1, 1 ≤ j ≤ n, has a
unique solution. The matrix B(t) of coefficients of this system is given as:
1 t t2 · · · tn−2
0
Tn 0 · · · 0 0 1 t · · ·
tn−3 0
0 Tn · · · 0 0 0 1 · · · tn−4 0
B(t) = . where T n =
. .. ..
... ... ... .. ..
. . . . .
0 0 · · · Tn 0 0 0 · · ·
1 0
0 0 0 ··· 0 1
and the submatrices in the first, second , . . . , last row are obtained respectively for
[n+1] [n+1] [n+1]
the variables Qn−1,j , Qn−2,j , . . . , Q1,j , 1 ≤ j ≤ n. The upper triangular submatrix
[n+1] [n+1] [n+1] [n+1]
Tn is same for each of Qn−1,j , Qn−2,j , . . . , Q2,j and Q1,j . Now |B(t)| = 1 and
[n+1]
C ∈ M∗ Q(t) . Hence the system X = B −1 C ∈ M∗ Q(t) . Therefore Q∗,∗ is
[m] [m]
rational in P∗ and Q∗,∗ , m ≤ n.
[n+1]
We construct another linear system in the variables P∗ analogous to Lemma
Lemma 3.2.10. The following relations hold for the irreducible words in MB n+1 .
[n+1] [n] [n+1] [n]
1) Pk = Pk Pn + Pk , k = 1, . . . , n − 1.
[n+1] [n+1] [n+1]
2) Pn = tPn + Pn(n−1) + t.
[n+1] [n] [n+1] [n]
3) P(n−1)···i = P(n−1)···i Pn + P(n−1)···i , i = 1, . . . , n − 2.
n
[n+1] [n+1] P j−n−1 [n+1] [n+1]
4) Pn(n−1) = tPn−1 − t Q∗,j Pn···j .
j=1
n
[n+1] [n+1] [n+1] [n+1]
tj−n−1 Qk,j Pn···j , k = 1, . . . , n − 2.
P
5) Pn···k = tP(n−1)···k −
j=1
4)
a a
[n+1] [n+1] [n+1] [n+1] [n+1] [n+1]
An(n−1) = {xn } × An−1 r B∗,1 × An···1 B∗,2 × An···2 ···
n···1 n···2
!
a [n+1] [n+1]
a a
[n+1]
B∗,j × An···j ··· B∗,n × A[n+1]
n
n···j n
implies
[n+1] [n+1] [n+1] [n+1] [n+1] [n+1]
Pn(n−1) = tPn−1 − t−n Q∗,1 Pn···1 + t−n+1 Q∗,2 Pn···2 + · · ·
[n+1] [n+1] [n+1] [n+1]
+ tj−n−1 Q∗,j Pn···j + · · · + t−1 Q∗,n Pn or
n
X
[n+1] [n+1] [n+1] [n+1]
Pn(n−1) = tPn−1 − tj−n−1 Q∗,j Pn···j .
j=1
5) The decomposition
a a
[n+1] [n+1] [n+1] [n+1] [n+1] [n+1]
An···k = {xn } × A(n−1)···k r Bk,1 × An···1 Bk,2 × An···2 ···
n···1 n···2
!
a [n+1] [n+1]
a a [n+1]
Bk,j × An···j ··· Bk,n × A[n+1]
n
n···j n
n
X
[n+1] [n+1] [n+1] [n+1]
Pn···k = tPn−1···k − tj−n−1 Qk,j Pn···j , k = 1, . . . , n − 2.
j=1
[n+1]
Proposition 3.2.11. The Hilbert series P∗ of irreducible words in MB n+1 is a
[n] [n+1]
rational function in P∗ and Q∗,∗ .
variables at t = 0 is given by
In 0 Dn×(n−1)
−1, if i = n, j = 1
B(0) = 0 In−2 0
where D = (dij ) such that dij =
0, otherwise.
0 0 In−1
[n+1]
Here the submatrices in the first row are obtained for the variables Pi , 1 ≤ i ≤ n,
[n+1] [n+1]
second for P(n−1)···j , 1 ≤ j ≤ n − 2 and third for Pn···k , 1 ≤ k ≤ n − 1. As |B(0)| = 1,
we found that |B(t)| = 1 + d1 t + d2 t2 + · · · . Hence B(t) is invertible in M∗ Q(t) .
[n+1]
Now C ∈ M∗ Q(t) . Therefore the system X = B −1 C ∈ M∗ Q(t) . Hence P∗
is
[n] [n+1]
rational in P∗ and Q∗,∗ .
Now using the above results we can easily prove Theorem 1.5.1.
[n+1]
Proof. (of Theorem 1.5.1) In Proposition 3.2.11 we see that Pi , 1 ≤ i ≤ n are
is rational.
n(n+1)
Proof. (of Corollary 3.1.1) Note that |∆| = 2
for the Garside element ∆. Using
the decomposition
a a a
MB n+1 = MB +
n+1 ∆MB +
n+1 ∆2 MB +
n+1 ···
function.
Now we prove the Theorem 3.1.2, a generalization of Theorem 1.5.1 using induction
First we discuss the particular case n = 2. Now we use the induction on |α|. If
[3]
|α| = 0 then A[3] = MB 3 and the rationality of HMB (t) was proved in Proposition
3.2.1. If |α| = 1 then the claim of the theorem was also proved in Proposition 3.2.1,
[3] [3]
i.e., P1 and P2 are rational. Let |α| > 1 then we have two cases.
[3]
(1). Let the first letter of α be x1 , i.e., α = x1 β. By induction Pβ is a rational
[3] [3] [3]
function. From Proposition 3.2.1 we know that Ax1 β = {x1 } × Aβ , therefore Px1 β =
[3]
tPβ is a rational function.
function.
Proof. (of Theorem 3.1.2) We will use double induction on n and |α|. If |α| = 0 then
[n+1]
A[n+1] = MB n+1 and HMB (t) is rational, proved earlier. If |α| = 1 then the claim
[n+1]
of the theorem was also proved in Proposition 3.2.11, i.e., Pj is rational for all
Case 2. Now suppose that α contains xn . In this case we have the following subcases:
Case 2.1. If xn is not first letter of α, i.e., α = α1 (n−1, 1)xn α2 (n, 1) (α1 is nonempty)
[n+1] [n+1] [n+1] [n+1]
then Aα = {α1 } × Axn α2 and so Pα = t|α1 | Pxn α2 is rational.
Case 2.2.1. Let β = xn−1 γ(n − 2, 1)δ(n − 1, j) where δ starts with xn−1 and contains
only the letters xn−1 , xn−2 , . . . , xj . Using the reduced complete presentation we have
[n+1] [n+1]
A[n+1]
α = {xn } × Aβ r {xn xn−1 γ} × {δ} × An···j . (3.3.1)
The word δ(n − 1, j) must contain all the letters xn−1 , xn−2 , . . . , xj , hence
induction.
[n+1]
Case 2.2.2. Let β = xn−1 xn−2 γ(n − 2, 1), i.e., α = xn xn−1 xn−2 γ(n − 2, 1). Let Bα,xi
be the set of words in the L.H.S. of relations in MB n+1 starting with the given word
[n+1] [n−1]
α and ending with xn · · · xi . Then Bα,xn = {xn xn−1 } × Axn−2 γ × {xn } and hence
44
[n+1] [n−1]
Qα,xn = t3 Pxn−2 γ is rational by induction. Also
[n+1] [2]
Bα,xn−1
= {xn xn−1 } × A[n−1]
xn−2 γ × Σ
n−2
A1 × {xn xn−1 }
[2]
r{xn } × Bx[n]n−1 xn−2 γ,xn−1 × Σn−2 A1 × {xn xn−1 }.
n−1
[n+1] [n−k+1]
Bα,xk
= {xn xn−1 } × A[n−1]
xn−2 γ × Σ
k−1
An−k × {xn · · · xk } r
n−1
a a
Bx[n+1]
n xn−1 xn−2 γ,xj
× {x j−1 · · · x k }
j=k+1
n−1
!
a [n−k+1]
{xn } × Bx[n]n−1 xn−2 γ,xj × Σk−1 A(n−k)···(j−k+1) × {xn · · · xk } .
(n−1)···j
j=k
Thus
n−1
X
[n−k+1]
Q[n+1]
α,xk = tn−k+3 Px[n−1] P
n−2 γ n−k
− tj−k Qx[n+1]
n xn−1 xn−2 γ,xj
j=k+1
n−1
X [n−k+1]
− tj−k+2 Qx[n]n−1 xn−2 γ,xj P(n−k)···(j−k+1) , k = 1, . . . , n − 2
j=k
n
[n+1] [n+1] [n+1] [n+1]
tj−n−1 Qα,xj Pn···j is rational.
P
and so Pα = tPβ −
j=1
Case 2.3. Let α = xn β and β contains a unique xn such that β = γ1 (n−1, 1)xn γ2 (n−
[n+1] [n+1] [n+1] [n+1]
1, 1). If β = xn γ2 then Aα = {xn } × Aβ and Pα = tPβ is rational.
[n+1] [n+1] [n+1] [n+1]
For γ2 6= xn−1 xn−2 · · · xj we have Aα = {xn } × Aβ and Pα = tPβ is
rational and for γ2 = xn−1 xn−2 · · · xj , j ≥ 2 and γ1 (n−1, 1) = xn−1 δ(n−2, 1)ε(n−1, k),
Case 2.4. If xn is the first letter of α and α = xn β such that β contains xn at least
[n+1] [n+1]
twice, i.e., β = γ1 (n − 1, 1)xn γ2 (n − 1, 1)xn γ3 (n, 1) then Aα = {xn } × Aβ and
[n+1] [n+1]
hence Pα = tPβ is rational. The case α = xn γ1 (n − 1, 1)xn xn−1 · · · x1 cannot
4.1 Introduction
Let bk be the number of braids with three strands and k positive crossings. One can
the results in Chapter 2 we have bk = Fk+2 − 1, where {Fk } are Fibonacci numbers.
√
1+ 5
Therefore the growth function of this sequence is exponential with growth rate 2
.
A A Z
Z A A A ZZ
A A Z A A A Z
A A ZZ A A A ZZ
A A Z A A A Z
b1 = 2 b2 = 4
In this chapter we generalize this to all spherical Artin monoids (and also to some
associated right-angled Artin monoids) and we find a universal upper bound for the
To a given Coxeter graph Γ we associate the monoid Gn+ with the following pre-
46
47
We study the growth function of these monoids (or groups), with a fixed set of
Let us look at two extremal Coxeter graphs, say Nn and Cn , where Nn is the totally
disconnected graph with n vertices (all the labels are 2) and Cn is the complete graph
and all labels are ∞. The first monoid Nn+ is the free abelian monoid with Hilbert
series
∞
1 X n−1+k k
= t
(1 − t)n k=0
n − 1
with polynomial growth function. The second monoid Cn+ is the free monoid with
Hilbert series
∞
1 X
= nk tk
1 − nt k=0
and the growth function is exponential nk and this is unbounded for the sequence
(Cn )n≥1 .
In spite of the last example, we shall show that the family of spherical Artin
In [14] P. Deligne proved that all spherical Artin monoids have rational Hilbert
series and he gave a formula in terms of the number of reflections in the associated
Weyl groups.
In [22] P. Xu studied the growth function of the monoid of positive braids (the
series An ) and she found explicit formulae for the Hilbert series of A3 and A4 .
48
Theorem 4.1.1. The growth rate of all spherical Artin monoids (with classical
X∗∞ = {A∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞ ∞
∗ , B∗ , D∗ , E∗ , F4 , G2 , H3 , H4 , I2 (p)}
where all the labels ≥ 3 are replaced by ∞. For example, to the Artin monoid of
∞ ∞ ∞ ∞
• • • ··· • •
x1 x2 x3 xn−1 xn
The sequence A∞ ∞ ∞ ∞ ∞ ∞
∗ contains B∗ , F4 , G2 , H∗ , I2 (p), so the study of growth rate
and presentation
* +
yi yj = yj yi , j + 2 ≤ i ≤ n − 1
Kn∞ = y 1 , y2 , . . . , y n .
yn yk = yk yn , 1 ≤ k ≤ n − 3
Because we will not use the groups associated to Coxeter graphs with labels ∞,
Let X∗+ be the set of monoids associated to the connected spherical type Coxeter
graphs, i.e.,
X∗+ = {A+ + + + + + + + +
∗ , B∗ , D∗ , E∗ , F4 , G2 , H3 , H4 , I2 (p)}
The growth rate of X∗+ (X∗ in the classical list, but not E∗ ) is less than the growth
Growth functions for Artin groups were studied by many authors, using different
set of generators. J. Mairesse and F. Mathéus [17] proved that the dihedral series
can easily show that 4 is still an upper bound for the growth rate. R. Charney
[11] analyzed the growth function of Artin groups with the set of normal generators
{d, d−1 : d is a divisor of the Garside element ∆} and the same groups {I2 (p)} have
unbounded growth ratio p − 1. In a recent paper [18] J. Mangahas shows that the
growth rates of braid groups have a lower bound which does not depend on the (finite)
set of generators.
The main idea was to reduce the study of Artin group Xn to the study of associated
monoid Xn+ . F. A. Garside [15] and P. Deligne [14] introduced the fundamental word
Theorem 4.1.3. All spherical Artin groups Xn , as monoids generated by the set
{x1 , . . . , xn , ∇ = ∆−1
n } have growth rate less than 4.
50
In Section 4.2 we solve the word problem for Kn∞ and compute its Hilbert series
In Section 4.3 we show that Kn∞ has growth rate less than 4 using a sequence
for Chebyshev type sequences of polynomials to have only real roots in a bounded
interval.
The last section contains the proofs of the Theorems 4.1.1 and 4.1.3 and of the
Proposition 4.1.2 and also the extension of all the previous results to the spherical
For small values of k, it is easy to compute the number of positive braids with
In a presentation of a monoid we fix a total order of the generators; in all our examples
we choose the natural order y1 < y2 < · · · < yn . Such a presentation is complete if
and only if all the ambiguities are solvable (see [5], [13]).
51
is a complete presentation.
≥ j +3) and these can be simplified in two ways; after three steps we get the canonical
word yk yj yi .
words in length-lexicographic order are given by the next proposition and the defining
a b hk
yia10 yia11−1 · · · yi1k−k
1
y b0 y b1 · · · yi2k−k
1 i2 i2 −1
2
2
· · · yihp0 yihp1−1 · · · yip −k
p
p
where ij − kj < ij+1 , j = 1, . . . , p − 1 and all the exponents are positive; as a special
j∗
case, after the factor ynd∗ could also appear yn−2 .
[n] P [n] k
Now we start to compute the Hilbert series HK (t) = ck t of Kn∞ and also
k≥0
[n] P [n] k [n] [n]
HK;i (t) = ck;i t ; here ck (or ck ) and ck;i (or ck;i ) denote the number of words
k≥0
in Kn∞ of length k, respectively words (in canonical form) of length k starting with
ak tk ,
P P k
yi . In the same way we denote by bk t the Hilbert series (or generating
k≥0 k≥0
functions) of the monoids X∗+ , respectively X∗∞ .
52
Corollary 4.2.4. In the monoid Kn∞ the following relations are satisfied
n
P
a) c0 = 1, c1;i = 1, ck = ck;i (k ≥ 1).
i=1
b) ck;i (k ≥ 2) are given by the recurrence
Proof. Decompose Kn as a sum of two determinants Un and Vn with the first rows
given by [λ, 0, . . . , 0] and [−1, −1, . . . , −1] respectively. Elementary operations on the
first two rows will give Un (λ) = λKn−1 (λ) and Vn (λ) = −λKn−2 (λ). We can extend
1
The determinant of the system given by the last n equations equals tn Kn t
. The
Hilbert series of K3∞ and K4∞ are given in the next two examples.
[3] 1
HK (t) = .
1 − 3t
54
Lemma 4.2.9.
and also
Proof. This relation is true for n = 3 and n = 4 and An and Kn satisfy the same
recurrence.
[n]
Lemma 4.2.11. The Hilbert series HK;m (t), 2 ≤ m ≤ n − 1 is given by
one equal to tm−1 Am−2 ( 1t ) and a second one, given by a unipotent n − m triangular
matrix.
[n] 1
HK (t) = .
tn Kn ( 1t )
55
Now we will separate the zero roots of Kn from the others [x] is the floor function .
n+1
Kn (λ) = λ[ 2
]
Kn (λ)
1) K0 = 2, K1 = 1, K2 = λ − 2, K3 = λ − 3 ;
and
that is,
n+1
Kn (λ) = λ[ 2
]
Kn (λ).
56
For n = 2p + 1 we have
Remark 4.2.14. The results of Proposition 4.2.13 hold for the monoids A∞
n and
Dn∞ :
n
A) The polynomial An (λ) has the form An (λ) = λ[ 2 ] An (λ) where An is a polynomial
of degree [ n+1
2
]; the sequence (An )n≥−1 is defined by:
1) A−1 = 1, A0 = 1, A1 = λ − 1, A2 = λ − 2.
D) Dn (λ), the determinant associated to the monoid Dn∞ , has the form Dn (λ) =
n−1
λ[ 2
]
Dn (λ) where Dn is a polynomial of degree [ n+2
2
]; the sequence (Dn )n≥3 is defined
by:
Lemma 4.2.15. Denote by q(λ) the quadratic polynomial λ2 − 4λ. Then we have
[ p2 ]
1 p
(λ − 2)p−2i q i (λ).
P
1) K2p (λ) = 2p−1 2i
i=0
[ p2 ] [ p−1
2
]
1 p 1 p
p−2i i
(λ − 2)p−2i−1 q i (λ).
P P
2) K2p+1 (λ) = 2p 2i
(λ − 2) q (λ) + 2p
(λ − 4) 2i+1
i=0 i=0
The recurrences of the Proposition 4.2.13 and those of the Remark 4.2.14 are close to
the recurrence for classical Chebyshev polynomials, Tn (λ) = cos(n arccos λ): Tn (λ) =
2λTn−1 (λ) − Tn−2 (λ) with initial values T0 = 1, T1 = λ (see [8]). We prove, under
general hypothesis, that all the roots of such polynomials are real and contained in
a bounded common interval [a, b]; in the classical case this interval is [−1, 1], in our
Proposition 4.3.1. Let a < b be two real numbers. If the sequence of polynomials
1) deg Rn = n,
58
2) all the roots of Rn are real, distinct and contained in (a, b):
(n) (n)
a < x1 < x2 < · · · < x(n)
n < b,
Proof. We prove the proposition under the hypothesis c(i). Changing the sequence
{Rn } with {−Rn } we can suppose that R0 (a) > 0. Starting induction by n, let
(m) (n+1)
{xi }i=1,...,m be the roots of Rm . Suppose the roots {xi } separate the roots
(n) (n+1) (n)
{xi } and the signs of Rn xi and Rn+1 xi are those given in the following
table:
a x1
(n)
x2
(n)
··· xn−1
(n) (n)
xn b
Rn •
− •
+ 0 0 0 ∓ 0 ±
a (n+1)
x1
(n+1)
x2
(n+1)
x3
···
(n+1)
xn−1 xn
(n+1) (n+1)
xn+1 b
Rn+1 •
− •
+
• •
+ 0 0 0 0 ∓ 0 ± 0 ∓
a (n+2)
x1
(n+2)
x2
(n+2)
x3
(n+2)
x4
···
(n+2)
xn−1
(n+2)
xn
(n+2)
xn+1
(n+2)
xn+2 b
Rn+2 •
− •
+
•
− • •± •
+ 0 0 0 0 0 ∓ 0 0 ∓ 0 ±
a b
R0 +
+ +
a (1)
x1 b
R1 + − −
+
a (2)
x1
(2)
x2 b
R2 •
−
+ +
(n+1) (n+1)
Rn+2 xi = −Rn xi ,
59
so we have opposite signs on the lines Rn and Rn+2 . Now Rn+2 has at least one root
(n+2) (n+1) (n+2) (n+1) (n+1)
in the intervals: x1 ∈ a, x1 , xj ∈ xj−1 , xj , 2 ≤ j ≤ n + 1 and
(n+2) (n+1)
xn+2 ∈ xn+1 , b . The polynomial Rn+2 has degree n + 2, hence all the roots of
(n+2) (n+2) (n+1) (n+2)
Rn+2 are x1 , . . . , xn+2 and the pattern of signs of Rn+2 xi and Rn+1 xi
Using Lemma 4.3.2, in Proposition 4.2.13 the precise multiplicity of the zero root
of Kn is [ n+1
2
].
Remark 4.3.3. The same is true for An but not for Dn . One can prove that Dn has
at most two roots in C r R; for example, D4 , D6 , D7 have two complex roots, D5 has
Remark 4.3.4. We cannot expect to obtain Chebyshev type recurrence for the polyno-
mials corresponding to spherical Artin monoids. For instance, in the case the monoid
n+1
A+
n , Deligne’s theorem ([14], 4.14) shows that the degree of the denominator is 2
=
Theorem 4.3.5. a) All the roots of Kn are real and belong to (0, 4).
b) All the roots of Kn are real and belong to [0, 4), with zero root of multiplicity [ n+1
2
]
Proof. Let m = [ n2 ] and 0 < α1 < α2 < · · · < αm < 4 be the m roots of Kn . From
1
=
(1 − α1 t)(1 − α2 t) · · · (1 − αm t)
β1 β2 βm
= + + ··· +
1 − α1 t 1 − α2 t 1 − αm t
X∞ X m
= βi (n)αik tk
k=0 i=1
x1 ∞ x3 ∞ x4 ∞ x5 ∞ ··· xn
(En∞ )n=6,7,8 : • • • • •
∞
• x2
by the order given in the introduction is to obtain a complete presentation (in the
standard order one have to solve the ambiguity x3 x2 x1 and to add new relations: see
[5]).
The growth rate of these monoids can be reduced to the growth rate of E8∞ :
61
Lemma 4.4.1. There exist homogeneous injective morphisms E6∞ E7∞ E8∞
defined by xi 7→ xi .
Proof. (of Lemma 4.4.1 and 4.4.2): As in Lemma 4.2.1 and Proposition 4.2.2, the
presentation of E∗∞ is complete and the word problem in these monoids has a simple
solution.
Remark 4.4.3. One can consider the whole sequence (En∞ )n≥6 : the corresponding
n
polynomials are given by En (λ) = λ[ 2 ] (λ − 1)Kn−1 (λ) and they satisfy the hypothesis
and ψ : Xn∞ → Xn+ , ψ(xi ) = xi are well defined because the images of the generators
satisfy the defining relations of the domains and these images generate the range-
monoids.
Proof. (of Theorem 4.1.1: connected graphs) For the monoids E6∞ , E7∞ and E8∞ we
use the Corollary 4.5.1 and the results of Section 4 and obtain the inequalities
For the other monoids we put together the inequalities obtained before:
Here β = constant, β(n) = constant in k, and γ, δ are between the greatest root of
The last step in the proof is to show that a spherical Artin monoid with a non-
connected Coxeter diagram has growth rate less than 4. The monoid associated to
`
a disjoint union of graphs Γ1 Γ2 is the direct product of the monoids associated to
Γ1 and Γ2 .
Lemma 4.5.2. If M1 and M2 are two monoids with growth rate less than γ, then
Proof. Let us denote by m1 (k), m2 (k), p(k) the numbers of words of length k (k ≥ 1)
and two constants c1 and c2 such that for any k, m1 (k) ≤ c1 δ k and m2 (k) ≤ c2 δ k .
constant c such that (k + 1) c3 δ k < cγ0k . Now p(k) < cγ0k , hence p(k) has growth rate
less than γ.
Proof. (of Theorem 4.1.1: non-connected graphs) Take γ = 4 and apply the lemma
x1 , . . . , xn ). This map is surjective because Garside element ∆n has two basic prop-
erties:
x−1 −1
i = wi ∆n = wi ∇;
P
can be written as the product g = ∇ ki
m01 m02 · · · m0s = ∇k m = ϕ(∇k , m) and the
length of g, length of ∇k m (in the monoid Xn ) and length of (∇k , m) (in M (∇) × Xn+ )
are equal (m0i is mi with permuted generators xj ). Applying Lemma 4.5.2 we obtain
the result.
Bibliography
[1] D. J. Anick, on the homolgy of associative algebras, Trans. Amer. Math. Soc.
296(1986), pp. 641-659.
[3] B. Berceanu, Z. Iqbal, Universal upper bound for the growth of Artin monids,
arXiv:0805.2656v1 [math.GR], 2008.
[5] G. Bergman, The diamond lemma for ring theory, Adv. in Math 29 (1978), pp.
178-218.
[6] J. S. Birman: Braids links and mapping-class groups, Annals of Math. Studies.
Princeton University Press, 12(1974).
[7] L. A. Bokut, Y. Fong, W. -F. Ke, and L. -S. Shiao. Gröbner-Shirshov bases for
braid semigroup, In Advances in algebra, World Sci. Publ. 2003, pages 60-72.
65
66
[9] N. Bourbaki, Groupes et algèbres de Lie, Chapitres 4-6, Elem. Math., Hermann,
1968.
[11] R. Charney, Geodesic automation and growth functions for Artin groups of finite
type, Math. Ann. no. 2, 301(1995), 307-324.
[12] W. L. Chow, On the algebraic braid group. Ann. of Math. 49, pp. 654-658.
[14] P. Deligne, Les immeubles des groupes de tresses généralisés, Invention math.
17(1972), 273-302 .
[15] F. A. Garside, The braid groups and other groups, Quat. J. Math. Oxford, 2e
ser. 20(1969), 235-254 .
[16] Z. Iqbal, Hilbert series of positive braids (to appear in Algebra Colloquium).
[17] J. Mairesse, F. Mathéus, Growth series for Artin groups of dihedral type, Inter-
national Journal of Algebra and Computation 16, 6(2006), 1087-1107.
[19] J. Michel, A note on words in braid monoids, J. Algebra 215 (1999), no. 1,
366-377.
[20] L. Paris, Braid groups and Artin groups, arxiv: 0711.2372VI [math. GR], 2007.
67
[22] P. Xu, Growth of the positive braid semigroups, J. Pure Appl. Algebra 80, no.
2, (1992), 197-215.