Mathematics Subject Classification Key Words and Phrases.: G G Ns G G G Ns G

PRESENTATIONS OF DEHN QUANDLES
NEERAJ K. DHANWANI, HITESH RAUNDAL, AND MAHENDER SINGH

arXiv:2202.02531v1 [math.GR] 5 Feb 2022
Abstract. In this paper, we give two approaches to write explicit presentations for Dehn quandles
using presentations of their underlying groups. The first approach gives finite presentations for Dehn
quandles of a class of Garside groups and Gaussian groups. The second approach is for general
Dehn quandles when the centralisers of generators of their underlying groups are known. We give
several examples including Dehn quandles of spherical Artin groups, surface groups and mapping
class groups of orientable surfaces to illustrate the results.
1. Introduction
A quandle is an algebraic system with a binary operation that encodes the three Reidemeister
moves of planar diagrams of links in the 3-space. Besides being fundamental to knot theory as
observed first in [21, 22, 27], these objects appear in the study of mapping class groups [38],
set-theoretic solutions of the quantum Yang-Baxter equation and Yetter-Drinfeld Modules [12],
Riemannian symmetric spaces [25] and Hopf algebras [2], to name a few. Many classical topological,
combinatorial and geometric knot invariants such as the knot group [21, 27], the knot coloring, the
Conway polynomial, the Alexander polynomial [21] and the volume of the complement in the 3-
sphere of a hyperbolic knot [20] can be retrieved from the knot quandle. Though quandles give
strong invariants of knots, the isomorphism problem for them is hard. This has motivated search
for newer properties, constructions and invariants of quandles themselves.
Understanding of presentations of quandles is a fundamental problem and determining a pre-
sentation is challenging in general. Even for the simplest quandles arising from groups, such as
conjugation quandles of infinite abelian groups, the number of generators and relations turn out
to be infinite. In this paper, we give two approaches to write explicit presentations for a class of
quandles called Dehn quandles introduced recently in [11].
The notion of a Dehn quandle of a group is motivated by two classes of examples. The first
being the free quandle on a set S, which is simply the union of conjugacy classes of elements of S
in the free group on S, equipped with the quandle operation of conjugation. The second class of
examples is given by surfaces. Let Mg be the mapping class group of a closed orientable surface Sg
of genus g ≥ 1 and Dgns the set of isotopy classes of non-separating simple closed curves on Sg . It
is known that Mg is generated by Dehn twists along finitely many simple closed curves from Dgns
[13, Theorem 4.1]. The binary operation
α ∗ β = Tβ (α),
where α, β ∈ Dgns and Tβ is the Dehn twist along β, turns Dgns into a quandle called the Dehn
quandle of the surface Sg . It turns out that Dgns can be seen as a subquandle of the conjugation
quandle Conj(Mg ) of the mapping class group Mg of Sg , by identifying the isotopy class of a simple
closed curve with the isotopy class of the corresponding Dehn twist. These quandles originally
Mathematics Subject Classification 2020. Primary 20M32, 57K12; Secondary 57K20, 20F36.
Key words and phrases. Artin group, Dehn quandle, Garside group, Gaussian group, mapping class group, surface
group.
1
2 NEERAJ K. DHANWANI, HITESH RAUNDAL, AND MAHENDER SINGH
appeared in the work of Zablow [38, 39]. He derived a homology theory based on Dehn quandles
of surfaces [40] and showed that isomorphism classes of Lefschetz fibrations over a disk correspond
to quandle homology classes in dimension two. Further, in [6], the Dehn quandle structure of the
torus has been extended to a quandle structure on the set of its measured geodesic foliations and
the quandle homology of this extended quandle has been explored. On the other hand, [24, 36, 37]
considered a quandle structure on the set of isotopy classes of simple closed arcs on an orientable
surface with at least two punctures, and called it the quandle of cords. In the case of a disk with n
punctures, this is simply the Dehn quandle of the braid group Bn with respect to its standard Artin
generators. A presentation for the quandle of cords of the real plane and the 2-sphere has been
given in [24]. Beyond these special cases, we have not seen explicit presentations of Dehn quandles
of surfaces in the literature. Dehn quandles of groups with respect to their subsets include many
well-known constructions of quandles from groups including conjugation quandles, free quandles,
Coxeter quandles [1], Dehn quandles of closed orientable surfaces, quandles of cords of orientable
surfaces, knot quandles of prime knots, core quandles of groups and generalized Alexander quandles
of groups with respect to fixed-point free automorphisms, to name a few. See [11] for details.
The paper is organized as follows. We give two approaches to write explicit presentations of Dehn
quandles using presentations of their underlying groups. In Section 2, using Garside theory, we
give finite presentations of Dehn quandles of groups of fractions of Garside monoids and Gaussian
monoids (Theorem 2.20 and Theorem 2.28). We give examples of presentations for Dehn quandles
using this approach including those of spherical Artin quandles. In Section 3, we prove a general
result giving a presentation of the Dehn quandle of a group with respect to a generating set when
the centraliser of each generator is known (Theorem 3.1). As a consequence, it follows that is
G = hS | Ri is a finitely presented group such that the centraliser of each generator from S is
finitely generated, then the Dehn quandle D(S G ) is finitely presented (Corollary 3.2). Although
Theorem 3.1 is general, finding generating sets for centralisers of elements in interesting classes
of groups like Garside groups and Artin groups is usually challenging. We give presentations for
Dehn quandles of surface groups, braid groups and mapping class groups of orientable surfaces to
illustrate the result. Examples of braid groups show that presentations of Garside quandles given
by Theorem 3.1 usually have larger number of relations than the ones given by Theorem 2.20.
2. Presentations of Gaussian quandles and Garside quandles

We refer the reader to [21, 22, 27] for basic facts on quandles. Throughout the paper, we consider
only right distributive quandles and follow conventions used in [11]. If (X, ∗) is such a quandle with
a generating set S, then by [35, Lemma 4.4.8], any element of X can be written in a left-associated
product of the form
((· · · ((a0 ∗e1 a1 ) ∗e2 a2 ) ∗e3 · · ·) ∗en−1 an−1 ) ∗en an
for ai ∈ S and ei ∈ {1, −1}. For simplicity of notation, we write such an element as
a0 ∗e1 a1 ∗e2 · · · ∗en an .
Let G be a group, S a non-empty subset of G and S G the set of all conjugates of elements of
S in G. The Dehn quandle D(S G ) of G with respect to S is the set S G equipped with the binary
operation of conjugation, that is,
x ∗ y = yxy −1
for x, y ∈ D(S G ). We refer to our recent work [11] for examples and basic results on Dehn quandles.
PRESENTATIONS OF DEHN QUANDLES 3
Understanding of presentations of quandles is fundamental to the development of quandle theory.

In this section, we give presentations of Dehn quandles of certain Gaussian groups and Garside
groups, which we shall refer as Gaussian quandles and Garside quandles, respectively. The notion
of a Gaussian group and a Garside group was first introduced by Dehornoy and Paris [10] and
developed further in the works of Dehornoy [9] and Picantin [30, 31, 32]. The definition of a Garside
group which many authors use first appeared in [9]. Many equivalent definitions of a Garside group
can be seen in the literature. For example, see [4, 8, 14, 17]. Note that a Garside group is referred
as a small Gaussian group or a thin Gaussian group in [10, 30, 31, 32], and the notion of a Garside
group in [10] is a special kind of a Garside group (see the comment after the definition of a Garside
monoid). The theory of Garside groups is largely inspired by the work of Garside [16] in which he
treated braid groups, and by the work of Brieskorn and Saito [5] generalizing the work of Garside
to Artin groups of spherical type.
2.1. Preliminaries on Gaussian and Garside monoids, and groups. An element a of a

monoid M is called an atom if a 6= 1 and a is indecomposable (i.e. if a = bc, then either b = 1 or
c = 1). A monoid M is called atomic if the atoms of M are finite in number, they generate M , and
if for every x ∈ M , there exists an integer Nx such that x cannot be written as a product of more
than Nx atoms. Using [10, Proposition 2.1(iii)], we can say that every atomic monoid is conical, i.e.
there are no invertible elements other than the identity element of the monoid. In a monoid M , an
element x is a left divisor of y, or y is a right multiple of x, if there exists an element z satisfying
y = xz. Right divisors and left multiples are defined analogously. We denote by x ≤L y if x is a
left divisor of y, and x ≤R y if x is a right divisor of y. In an atomic monoid M , the left and right
divisibility relations (i.e. the relations ≤L and ≤R ) are respectively left and right invariant partial
orders on M (see [10, Proposition 2.3]).
A right Gaussian monoid is a monoid M satisfying the following properties:
• M is atomic;
• M is left and right cancellative;
• (M, ≤L ) is a lattice (i.e. the left g.c.d. and the right l.c.m. exist and they are unique for
every x, y ∈ M , see the definitions below).
A left Gaussian monoid is defined similarly. A Gaussian monoid is both left as well as right
Gaussian monoid. Note that the preceding definition of a right Gaussian monoid is different from
that of [10]. We have placed an extra assumption of right cancellativity to insure that such a
monoid embeds in its group of fractions (see [7, Theorem 1.23] for Ore’s criterion).
For elements x and y in a right Gaussian monoid, the left greatest common divisor (left g.c.d.)
of x and y is a common left divisor z of x and y such that every common left divisor of x and y
is a left divisor of z. The right least common multiple (right l.c.m.) of x and y is a common right
multiple z of x and y such that every common right multiple of x and y is a right multiple of z.
Similarly, in a left Gaussian monoid, we can define the right g.c.d. and the left l.c.m. of elements.
For elements x and y in a right Gaussian monoid, denote the left g.c.d. and the right l.c.m. of
x and y respectively by x ∧ y and x ∨ y. The right residue of x in y (denoted by x\y) is the unique
element z satisfying x ∨ y = xz. Thus, we have
x ∨ y = x(x\y) = y(y\x) = y ∨ x.
A Garside element of a monoid M is an element ∆ ∈ M such that the left divisors of ∆ coincide
with the right divisors of ∆, they are finite in number, and they generate M . We say that a
Gaussian monoid M is a Garside monoid if it contains a Garside element. By [10, Proposition 2.2],
the set of atoms in M is contained in every generating subset of M . It follows that atoms are also
divisors of ∆. In [10, 30, 31, 32], a Garside monoid is referred as a small Gaussian monoid or a
thin Gaussian monoid. One can look at [9, Proposition 2.1] for a necessary and sufficient condition
on a monoid to be a Garside monoid. Note that the notion of a Garside monoid in [10] is slightly
restricted in the sense that the left l.c.m. and the right l.c.m. of atoms in a Garside monoid should
coincide and it is a Garside element in the sense defined above.
The group of fractions of a monoid M is the group which has the same presentation as that of
the monoid M . In other words, if a monoid M = hS | Ri is the quotient of the free monoid on S
modulo the relations in R, then its group of fractions is the quotient of the free group on S modulo
the relations in R.
A Gaussian group (respectively, a Garside group) is the group of fractions of a Gaussian monoid
(respectively, of a Garside monoid). Similarly, we can also define a right Gaussian group and a left
Gaussian group.
An Ore monoid is one that embeds in its group of fractions. Ore’s criterion says that if a monoid
M is left and right cancellative, and if any two elements of M have a common right multiple, then
M embeds in its group of fractions [7, Theorem 1.23]. Note that a Gaussian monoid (respectively,
a Garside monoid) satisfies Ore’s conditions, and thus embeds in the corresponding Gaussian group
(respectively, in the corresponding Garside group). The same is true for right Gaussian monoids
as well as left Gaussian monoids.
Let M be a right Gaussian monoid, A the set of atoms in M and G the group of fractions of
M . Then the Dehn quandle D(AG ) will be referred as a right Gaussian quandle. The terms left
Gaussian quandles, Gaussian quandles and Garside quandles are defined similarly. For example, in
case of a spherical Artin group, the Artin quandle as defined in [11, Section 3] is a Garside quandle.
Here, note that, by [5] and [10, Example 1], a spherical Artin group is a Garside group.
Let S be any finite set. We set the following notations:
• S −1 - a set which is in one to one correspondence with S (for each s ∈ S, the corresponding
element in S −1 is denoted by s−1 which denotes the inverse of s in the free group on S).
• S ∗ - the set of words on S together with the empty word (i.e. it is nothing but the free
monoid on S and the empty word corresponds the identity element of the free monoid).
Note that, by the notation itself, the set (S ∪ S −1 )∗ is nothing but the free monoid on S ∪ S −1 . The
empty word represents the identity element in S ∗ (or, in (S ∪ S −1 )∗ ) and we denote it by ǫ or by
1, as per convenience of notation. For a word x = sδ11 sδ22 · · · sδnn in (S ∪ S −1 )∗ , we denote the word
n −δn−1 −δ1
s−δ
n sn−1 · · · s1 by x−1 , where si ∈ S and δi ∈ {1, −1}. Note that, for any x ∈ (S ∪ S −1 )∗ , the
elements xx−1 and x−1 x are different elements of (S ∪ S −1 )∗ , and they differ from the empty word.
A word x ∈ (S ∪ S −1 )∗ is called positive if x ∈ S ∗ . For a positive word x ∈ S ∗ , the length of x
(denoted by ℓ(x)) is the number of letters in x (i.e. ℓ(x) = n if x = s1 s2 · · · sn , where si ∈ S). Note
that the length of a positive word is zero if and only if it is the empty word. A word w ∈ (S ∪ S −1 )∗
is called symmetric if it is of the form x−1 sδ x for some x ∈ (S ∪ S −1 )∗ , s ∈ S and δ ∈ {−1, 1}.
The word w is called positive symmetric if δ = 1 and it is called negative symmetric if δ = −1. We
denote the set of positive symmetric words on S ∪ S −1 by S (S∪S ) , i.e.,
−1 ∗
S (S∪S
−1 )∗
= {x−1 sx | s ∈ S and x ∈ (S ∪ S −1 )∗ }.
Let M be a monoid generated by a finite set S and G the group of fractions of M . For words x
and y in (S ∪ S −1 )∗ , we write x =G y if x and y represent the same element in G. Similarly, we
write x =M y if x and y are positive words and they represent the same element in M . Note that
if M is an Ore monoid, then two positive words x and y are equivalent in M if and only if they are
equivalent in G. In particular, this is true for right Gaussian monoids, left Gaussian monoids and
Garside monoids, since such monoids are Ore monoids.
For a finite set S, a complement on S is a map f : S × S → S ∗ such that f (s, t) is the empty
word if and only if s = t. A complement f on S is said to be homogeneous if ℓ(f (s, t)) = ℓ(f (t, s))
for every (s, t) ∈ S × S. A presentation of a monoid (or, of a group) is called a complemented
presentation if it is of the form
hS | sf (s, t) = tf (t, s) for s, t ∈ Si or hS | g(s, t)t = g(t, s)s for s, t ∈ Si ,
where f and g are complements on S. A complemented presentation is called homogeneous if
the associated complement is homogeneous. In general, a finite presentation hS | Ri of a monoid
is homogeneous if ℓ(x) = ℓ(y) for every relation x = y in R, where x, y ∈ S ∗ . A monoid is
homogeneous if it has a finite homogeneous presentation. For example, Artin monoids (equipped
with Artin presentations) are homogeneous.
Let M be a right Gaussian monoid (or, a Garside monoid) and S a finite generating set for M .
A right l.c.m. selector on S in M is a complement f on S such that f (s, t) represents the element
s\t in M for every (s, t) ∈ S × S. By [10, Theorem 4.1], M has the complemented presentation
hS | sf (s, t) = tf (t, s) for s, t ∈ Si ,
where f is the right l.c.m. selector on S in M . The corresponding right Gaussian group (respec-
tively, the corresponding Garside group) is the quotient of the free monoid (S ∪ S −1 )∗ by relations
ss−1 = 1, s−1 s = 1 and sf (s, t) = tf (t, s) for s, t ∈ S.
Let M be a left Gaussian monoid (or, a Garside monoid) and S a finite generating set for M . A
left l.c.m. selector on S in M is a complement g on S such that g(s, t) represents the element s/t
in M for every (s, t) ∈ S × S. The monoid M has the complemented presentation
hS | g(s, t)t = g(t, s)s for s, t ∈ Si ,
where g is the left l.c.m. selector on S in M . The corresponding left Gaussian group (respectively,
the corresponding Garside group) has the same complemented presentation as given above.
One can look at [10, sections 3 and 4], for necessary and sufficient conditions on a monoid
with complimented presentations to be a Gaussian monoid (or, a right Gaussian monoid, or a left
Gaussian monoid). Also, see [8, Criterion 5.9], [9, propositions 5.4, 5.12 and 6.14] and [10, Corollary
3.11 and Theorem 4.2], for necessary and sufficient conditions on a monoid with complimented
presentations to be a Garside monoid.
2.2. Presentations of right Gaussian quandles and Garside quandles of certain types.
For this subsection, we set the following notations:
• M - a right Gaussian monoid;
• S - a finite generating set for M ;
• A - the set of atoms in M ;
• N - a Garside monoid;
• T - a finite generating set for N ;
• B - the set of atoms in N ;
• ∆ - a Garside element in N ;
• (M, S), (M, A), (N, T ), (N, B), (N, ∆) and (N, T, ∆) - pairs and a triple of objects with
the meaning above.
For a pair (M, S), we assume throughout this subsection that elements in S are pairwise distinct
in M . The same is assumed in case of a pair (N, T ).
For any pair (N, ∆), by [9, lemmas 2.2 and 2.3], the map
x 7→ (x\∆)\∆
is a permutation of the set of divisors of ∆ and it extends to an automorphism of N , which further
extends to an automorphism φ of the Garside group H corresponding to N .
Remark 2.1. By [9, Lemma 2.2], we have x∆ = ∆φ(x) for all x ∈ H. Since φ(N ) = N and
φ(B) = B, we have N ∆ = ∆N and B∆ = ∆B.
Consider the following conditions on a pair (M, S):
(i) (s\t)s ∈ S whenever (s, t) ∈ S × S and s ≤L t.
(ii) s\t ≤L s ∨ t for all (s, t) ∈ S × S.
(iii) (s\t)\(s ∨ t) ∈ S for all (s, t) ∈ S × S.
(iv) M has a finite homogeneous presentation hS | Ri.
(v) For each s ∈ S and each x ∈ M , there exists y = y(s, x) in M such that xy ≤L sxy.
Recall that the set of atoms in M is contained in every generating subset of M . Thus, if the
set of atoms does not satisfy condition (ii), then there does not exist any generating set for M
satisfying condition (ii).
Remark 2.2. Note that, in an atomic monoid, an atom x is a divisor of an atom y if and only if
x = y. Thus, a pair (M, A) of a right Gaussian monoid M and the set A of atoms in M satisfies
condition (i) trivially.
Lemma 2.3. If a pair (M, S) satisfies conditions (ii) and (iv), then it satisfies conditions (i) and
(iii). Moreover, in this case S must be the set of atoms in M .
Proof. Since (M, S) satisfies condition (iv), we have ℓ(x) = ℓ(y) for any words x, y ∈ S ∗ with
x =M y and ℓ(z) = ℓ(x) + ℓ(y) for any words x, y, z ∈ S ∗ with z =M xy. Also, it is easy to see that
S must be the set of atoms in M . By Remark 2.2, (M, S) satisfies condition (i). Since S satisfies
condition (ii), we have
(s\t)((s\t)\(s ∨ t)) = s ∨ t = s(s\t)
for all s, t ∈ S. Let s and t be any elements in S. Suppose x and y be positive words representing
s\t and (s\t)\(s ∨ t), respectively. Then xy =M sx and hence ℓ(x) + ℓ(y) = ℓ(s) + ℓ(x). This implies
that ℓ(y) = ℓ(s) = 1. Let y = s1 s2 · · · sn for some si ∈ S. Since ℓ(y) = 1, we must have n = 1.
Thus, y ∈ S and consequently (s\t)\(s ∨ t) ∈ S. Hence, S satisfies condition (iii).
We define the following terms:
• A pair (M, S) is of
– type R1 if it satisfies conditions (i), (ii), (iii) and (v).
– type R2 if it satisfies conditions (i), (ii) and (iii), and if there exist a triple (N, T, ∆)
with T ∆ = ∆T and an epimorphism π : (N, T ) ։ (M, S).
• A pair (N, T ) is of type R3 if it satisfies conditions (i), (ii) and (iii), and there exists a
Garside element ∆ ∈ N such that T ∆ = ∆T .
• A pair (M, A) is of
– type R4 if it satisfies conditions (ii), (iii) and (v).
– type R5 if it satisfies conditions (ii), (iv) and (v).
– type R6 if it satisfies conditions (ii) and (iii), and if there exist a triple (N, T, ∆) with
T ∆ = ∆T and an epimorphism π : (N, T ) ։ (M, A).
– type R7 if it satisfies conditions (ii) and (iv), and if there exist a triple (N, T, ∆) with
• A pair (N, B) is of
– type R8 if it satisfies conditions (ii) and (iii).
– type R9 if it satisfies conditions (ii) and (iv).
For i = 4, 5, 6, 7, we say a right Gaussian monoid M is of type Ri if the pair (M, A) is of type
Ri , where A is the set of atoms in M . In this case, we also say the right Gaussian group G
corresponding to M and the right Gaussian quandle D(AG ) are of type Ri . Similarly, for i = 8, 9,
we say a Garside monoid N is of type type Ri if the pair (N, B) is of type Ri , where B is the set
of atoms in N . In this case, we also say the Garside group H corresponding to N and the Garside
quandle D(B H ) are of type Ri .
Many classes of Garside monoids are of type R8 and R9 . For example, braid monoids, and
more generally, Artin monoids of spherical type are Garside monoids of type R9 . Nnote that, by
[5, 16] and [10, Example 1], such monoids are Garside monoids. The next two propositions give a
machinery to produce families of Garside monoids of types R8 and R9 .
Proposition 2.4. [10, Proposition 5.2] Consider a finite set S = {x1 , x2 , . . . , xn }, n positive words
u1 , u2 , . . . , un in S ∗ , and a permutation δ of {1, 2, . . . , n}. We assume that:
(1) There exists a map ν : S → Z≥0 which when extended to S ∗ by setting ν(ǫ) = 0 and ν(uv) =
ν(u) + ν(v), satisfies

ν x1 u1 xδ(1) = ν x2 u2 xδ(2) = · · · = ν xn un xδ(n) .
(2) For each index k, there exists an index j satisfying
xk uk = uj xδ(j) .
Let M be the monoid defined by the presentation

x1 , x2 , . . . , xn | x1 u1 xδ(1) = x2 u2 xδ(2) = · · · = xn un xδ(n) .
Then M is a Garside monoid.
In Proposition 2.4, one can check that the map k 7→ j is a permutation of {1, 2, . . . , n}. It
follows from the proof of the proposition that S is the set of atoms in M . If the words xi ui xδ(i)
and xj uj xδ(j) both represent a right l.c.m. of xi and xj for all i and j, then the set S satisfies
conditions (ii) and (iii). Consequently, M is a Garside monoid of type R8 .
Example 2.5. [10, Example 4] Consider a set S = {x1 , x2 , . . . , xn } and integers p1 , p2 , . . . , pn
strictly greater than 1. Take δ to be the identity permutation and ui = xpi i −2 for each i. Choose n
positive integers k1 , . . . , kn satisfying k1 p1 = · · · = kn pn and set ν(xi1 xi2 · · · xir ) = ki1 + · · · + kir .
Then, by Proposition 2.4, the monoid
hx1 , x2 , . . . , xn | xp11 = x2p2 · · · = xpnn i
is a Garside monoid. Moreover, it is of type R8 . Note that monoids with presentations hx, y | xp =
y q i have torus knot groups as their groups of fractions.
Example 2.6. [10, Example 5] For p letters x1 , x2 , . . . , xp and a positive integer m ≥ 2, let
prod(x1 , x2 , . . . , xp ; m) denotes the word
x1 x2 · · · xp x1 x2 · · · .
| {z }
m terms
Now, let S = {x1 , x2 , . . . , xn }, where n ≥ 2. By Proposition 2.4, the monoid
hx1 , x2 , . . . , xn | prod(x1 , x2 , . . . , xn ; m) = prod(x2 , x3 , . . . , xn , x1 ; m) = · · · = prod(xn , x1 , x2 , . . . , xn−1 ; m)i
is a Garside monoid. Further, it is of type R9 .

As a special case of this example, the monoid
hx1 , x2 , . . . , xn | x1 x2 · · · xn = x2 x3 · · · xn x1 = · · · = xn x1 · · · xn−1 i
has the group of fractions as the fundamental group of the complement of n lines through the origin
in C2 [34]. And, the monoid
hx1 , x2 , . . . , xn | x1 x2 = x2 x3 = · · · = xn x1 i
has the Artin group of type I2 (n) as its group of fractions.
Proposition 2.7. [10, Proposition 5.3] Consider Garside monoids M1 , M2 , . . . , Mn and positive
integers p1 , p2 , . . . , pn for n ≥ 2. Let ∆i denote the minimal Garside element of Mi . For each i,
we assume that:
(1) There is a map νi : Mi → Z≥0 satisfying νi (a) > 0 for all a ∈ Mi with a 6= 1 and νi (ab) =
νi (a) + νi (b) for all a, b ∈ Mi .
(2) If Mi has only one atom, that is, if Mi is isomorphic to Z+ , then pi ≥ 2.
Let M be the quotient of the free product M1 AM2 A · · · AMn of monoids modulo the congruence ≡
p
generated by ∆pi i = ∆j j for all i, j, that is,
M = (M1 AM2 A · · · AMn )/≡ .
Then M is a Garside monoid.
It is easy to see that if the Garside monoids M1 , M2 , . . . , Mn in Proposition 2.7 are of type
R8 , and if Si is the set of atoms in Mi satisfying conditions (ii) and (iii), then the monoid M =
(M1 AM2 A · · · AMn )/≡ is a Garside monoid of type R8 with the set of atoms S = S1 ⊔ S2 ⊔ · · · ⊔ Sn
satisfying conditions (ii) and (iii). It follows from the proof of the proposition that S is, in fact,
the set of atoms in M .
Example 2.8. [10, Example 6] Mixing presentations in examples 2.5 and 2.6 gives new examples.
For example,

x1 , x2 , y1 , y2 , y3 | x21 = x52 = y1 y2 y3 y1 = y2 y3 y1 y2 = y3 y1 y2 y3
is a Garside monoid by Proposition 2.7. Further, it is of type R8 .
Example 2.9. [10, Example 7] Applying Proposition 2.7 to Artin monoids of type B3 and A3
shows that the monoid
D
x1 , x2 , x3 , y 1 , y 2 , y 3 | x1 x2 x1 x2 = x2 x1 x2 x1 , x1 x3 = x3 x1 , x2 x3 x2 = x3 x2 x3 ,
E
y1 y2 y1 = y2 y1 y2 , y1 y3 = y3 y1 , y2 y3 y2 = y3 y2 y3 , (x1 x2 x3 )6 = (y1 y2 y3 y1 y2 y1 )3
is a Garside monoid of type R9 .

There are many examples of Garside monoids in the literature, for example, [4, 10, 14, 16, 17, 30].
One can choose Garside monoids of type R8 and/or of type R9 from these examples. Then using
Proposition 2.7 one can produce more Garside monoids of type R8 and/or of type R9 . In [30],
the cross product of monoids has been defined, and it has been proved that the cross product of
Garside monoids is a Garside monoid. This allows us to construct more Garside monoids of type
R8 and/or of type R9 once we have some families of such monoids.
Proposition 2.10. Pairs of types R2 through R9 are of type R1 .
Proof. For a right Gaussian monoid M , by [10, Proposition 2.2], the set of atoms in M is a finite
generating set for M . With this in mind, together with Remark 2.1, Remark 2.2 and Lemma 2.3,
we can conclude that:
• Pairs of types R4 and R5 are of type R1 .
See Figure 1 for relations among the pairs Ri , where an edge Ri → Rj means that a pair of type
Ri is of type Rj .
R1
R2 R4
R3 R6 R5
R8 R7
R9
Figure 1. A diagram for pairs of types Ri
Since every Garside monoid is of course a right Gaussian monoid, a pair (N, T ) can be viewed
as a pair (M, S) of a right Gaussian monoid with M = N and a finite generating set S = T for M .
Thus, by taking M = N , S = T , π = IdN (where IdN denotes the identity map of N ), we conclude
that a pair of type R3 is of type R2 . Now, it is enough to prove that a pair of type R2 is of type
R1 .
Let (M, S) be a pair of type R2 . Then, there exist a triple (N, T, ∆) with T ∆ = ∆T and an
epimorphism π : (N, T ) ։ (M, S). In order to prove that the pair (M, S) is of type R1 , we only have
to show that it satisfies condition (v). Let s ∈ S and x ∈ M be any elements. Since π is surjective
and it maps T onto S, there exist s′ ∈ T and x′ ∈ N such that π(s′ ) = s and π(x′ ) = x. Since
divisors of ∆ generate N , we can write x′ = d1 d2 · · · dn for some divisors d1 , d2 , . . . , dn of ∆. Here,
note that di ’s are not necessarily pairwise distinct. Recall that the map φ given by x 7→ (x\∆)\∆
is a permutation of the set of divisors of ∆. It is easy to see that d∆ = ∆φ(d) for a divisor d of ∆.
Since φn−i (di ) is a divisor (in particular, a left divisor) of ∆, we have φn−i (di ) φn−i (di )\∆ = ∆.
Let us set ei = φn−i (di )\∆ for i = 1, 2, . . . , n. Then, for each i = 1, 2, . . . , n, we have
di ∆n−i ei = ∆n−i φn−i (di )ei (since di ∆ = ∆φ(di ))

n−i n−i n−i

=∆ φ (di ) φ (di )\∆ (since ei = φn−i (di )\∆)

= ∆n−i+1 (since φn−i (di ) φn−i (di )\∆ = ∆).
Let y ′ = en en−1 · · · e1 . Then, using ∆n−i+1 = di ∆n−i ei once for each i = 1, 2, . . . , n, we get
∆n = d1 ∆n−1 e1
= d1 d2 ∆n−2 e2 e1
..
.
= d1 d2 · · · dn ∆0 en en−1 · · · e1
= x′ y ′ .
We see that
s ′ x′ y ′ = s ′ ∆ n (since x′ y ′ = ∆n )
= ∆ n t′ for some t′ ∈ T (since T ∆ = ∆T )
= x′ y ′ t ′ (since ∆n = x′ y ′ ).
Let y = π(y ′ ) and t = π(t′ ). Then
sxy = π(s′ x′ y ′ ) (since π is a morphism of monoids)
′ ′ ′
= π(x y t ) (since s′ x′ y ′ = x′ y ′ t′ )
= xyt,
and hence xy ≤L sxy. Thus, the pair (M, S) satisfies condition (v), and hence is of type R1 .
Suppose that (M, S) is a pair that satisfies condition (i). Then, we have a map
α : {(s, t) ∈ S × S | s ≤L t} → S
defined by
(2.2.1) (s, t) 7→ (s\t)s.
Note that α(s, s) = s for all s ∈ S.
Remark 2.11. Let s, t ∈ S be such that s ≤L t. Then t = s(s\t) and α(s, t) = (s\t)s. Suppose
that f is a right l.c.m. selector on S in M . Since f (s, t) represents the element s\t, we have
t =M sf (s, t) and α(s, t) =M f (s, t)s. Thus, s−1 ts =G f (s, t)s =G α(s, t), where G is the right
Gaussian group corresponding to M .
Suppose that (M, S) is a pair that satisfies conditions (ii) and (iii). Then, we have a map
β : S × S → S defined by
(2.2.2) (s, t) 7→ (s\t)\(s ∨ t).
Note that β(s, s) = s for all s ∈ S.
Remark 2.12. Let s and t be elements in S. Since (M, S) satisfies condition (ii), we have s ∨ t =
(s\t) ((s\t)\(s ∨ t)). In other words, s(s\t) = (s\t) ((s\t)\(s ∨ t)). Suppose that f is a right l.c.m.
selector on S in M . Since f (s, t) represents the element s\t and β(s, t) = (s\t)\(s ∨ t), we have
sf (s, t) =M f (s, t)β(s, t). Thus, f (s, t)−1 sf (s, t) =G β(s, t), where G is the right Gaussian group
corresponding to M .
Suppose that (M, S) is a pair that satisfies conditions (i), (ii) and (iii). Let α and β be the maps
defined by (2.2.1) and (2.2.2), respectively. Let G be the right Gaussian group corresponding to
M and f a right l.c.m. selector on S in M . Then, the right f -equivalence on S (S∪S ) is defined
−1 ∗
to be the equivalence relation generated by the following transformations:

(Rα ) x−1 s−1 tsx ←→ x−1 α(s, t)x

where s, t ∈ S with s ≤L t and x ∈ (S ∪ S −1 )∗ .
(Rf,β ) x−1 f (s, t)−1 sf (s, t)x ←→ x−1 β(s, t)x
where s, t ∈ S and x ∈ (S ∪ S −1 )∗ .
(RG ) x−1 sx ←→ y −1 sy
where s ∈ S and x, y ∈ (S ∪ S −1 )∗ with x =G y.
For x, y ∈ S (S∪S , we denote x ≃fR y if x and y are right f -equivalent.

−1 )∗
Lemma 2.13. Let x, y ∈ S (S∪S and z ∈ (S ∪ S −1 )∗ . If x ≃fR y, then z −1 xz ≃fR z −1 yz.

−1 )∗
Proof. Let x, y ∈ S (S∪S ) and z ∈ (S ∪ S −1 )∗ such that x ≃fR y. It is clear that if x and y differ
−1 ∗
by a single transformation Rα or Rf,β or RG , then by the transformation itself, z −1 xz ≃fR z −1 yz.

Since the right f -equivalence on S (S∪S ) is generated by transformations Rα , Rf,β and RG , we
−1 ∗
have z −1 xz ≃fR z −1 yz in the general case as well.

The next result is motivated by [23, Theorem 1.4].
Theorem 2.14. Suppose that (M, S) is a pair of type R1 . Let G be the right Gaussian group
corresponding to M and f a right l.c.m. selector on S in M . Then, two words in S (S∪S ) are
−1 ∗
right f -equivalent if and only if they represent the same element in G.

We require some lemmas to prove the preceding theorem.
Lemma 2.15. Let (M, S) be a pair that satisfies conditions (i), (ii) and (iii). Let G be the right
Gaussian group corresponding to M and f a right l.c.m. selector on S in M . Let s, t ∈ S and
x ∈ S ∗ be such that x−1 sx =G t. Then x−1 sx ≃fR t.
Proof. Since M is atomic, by [10, Proposition 2.1], there exists a map ν : M → Z≥0 such that
ν(a) + ν(b) ≤ ν(ab) for all a, b ∈ M , and ν(a) = 0 if and only if a = 1. For w ∈ S ∗ , let w̄ denote
the element in M represented by w. We prove the lemma by induction on ν(x̄). Suppose that
x−1 sx =G t and ν(x̄) = 0. Then x̄ = 1, and it follows from properties of ν that x is the empty word.
This implies that s =M t, i.e. s and t are the same elements of S. Thus x−1 sx ≃fR s ≃fR t. Suppose
that the lemma is true when ν(x̄) ≤ n. Let x ∈ S ∗ be such that ν(x̄) = n + 1 and x−1 sx =G t.
Let x = uy for u ∈ S and y ∈ S ∗ . Then x−1 sx =G y −1 u−1 suy, and hence y −1 u−1 suy =G t. Let us
consider the following cases:
(1) u ≤L s: By transformation Rα , we have
y −1 u−1 suy ≃fR y −1 α(u, s)y,
where α is the map defined by (2.2.1). Since x = uy and u 6= 1, it follows from properties of ν
that ν(ȳ) < ν(x̄). In other words, ν(ȳ) ≤ n. Using Remark 2.11, we get
y −1 α(u, s)y =G y −1 u−1 suy =G t.
By induction hypothesis, we have
y −1 α(u, s)y ≃fR t,
and hence
x−1 sx ≃fR y −1 u−1 suy ≃fR y −1 α(u, s)y ≃fR t.
(2) u L s: Since y −1 u−1 suy =G t and M embeds in G, we have suy =M uyt. Thus, the words
suy and uyt both represent the common right multiple s̄ūȳ = ūȳ t̄ of s̄ and ū in M . Since
sf (s, u) represents the right l.c.m. of s̄ and ū, we have suy =M sf (s, u)z for some z ∈ S ∗ . By
cancellation on the left, we get uy =M f (s, u)z. Thus
y −1 u−1 suy ≃fR z −1 f (s, u)−1 sf (s, u)z (by transformation RG )
≃fR z −1 β(s, u)z (by transformation Rf,β ),
where β is the map defined by (2.2.2). Since u(u\s) = s(s\u) and u L s, we have s\u 6= 1. In
other words, f (s, u) does not represent the identity element of M . Since x =M uy =M f (s, u)z,
it follows from properties of ν that ν(z̄) < ν(x̄). Thus, we have ν(z̄) ≤ n. Using Remark 2.12,
we get
z −1 β(s, u)z =G y −1 u−1 suy =G t.
By induction hypothesis, we have
z −1 β(s, u)z ≃fR t.
Hence, we obtain
x−1 sx ≃fR y −1 u−1 suy ≃fR z −1 β(s, u)z ≃fR t,
and the proof is complete.
Lemma 2.16. Let (M, S) be a pair that satisfies conditions (i), (ii) and (iii). Let s ∈ S and
x, y ∈ S ∗ be such that sx =M xy. Then, there exists t ∈ S such that y =M t.
Proof. By [10, Proposition 2.1], we have a map ν : M → Z≥0 such that ν(a) + ν(b) ≤ ν(ab) for all
a, b ∈ M , and ν(a) = 0 if and only if a = 1. For w ∈ S ∗ , let w̄ denote the element in M represented
by w. We prove the lemma by induction on ν(x̄). Suppose sx =M xy and ν(x̄) = 0. Then x̄ = 1,
and it follows from by the properties of ν that x is the empty word. This implies that y =M s, and
the lemma holds. Assume that the lemma is true when ν(x̄) ≤ n. Suppose s ∈ S and x, y ∈ S ∗ be
such that ν(x̄) = n + 1 and sx =M xy. Let x = uz for u ∈ S and z ∈ S ∗ . Then suz =M uzy. We
now consider two cases below:
(1) u ≤L s: Since x = uz and u 6= 1, it follows from properties of ν that ν(z̄) < ν(x̄). In other
words, ν(z̄) ≤ n. Let α be the map defined by (2.2.1). Then, Remark 2.11 gives su =M uα(u, s).
Since suz =M uzy, we have uα(u, s)z =M uzy. Cancellation on the left gives α(u, s)z =M zy.
By induction hypothesis, there exists t ∈ S such that y =M t.
(2) u L s: Since suz =M uzy, the words suz and uzy both represent the common right multiple
s̄ūz̄ = ūz̄ ȳ of s̄ and ū in M . Let f be a right l.c.m. selector on S in M . Since sf (s, u) represents
the right l.c.m. of s̄ and ū, we have suz =M sf (s, u)w for some w ∈ S ∗ . Cancellation on the
left gives uz =M f (s, u)w. Since u L s, we have s\u 6= 1. Thus, f (s, u) does not represent
the identity element of M . Since x =M uz =M f (s, u)w, it follows from properties of ν that
ν(w̄) < ν(x̄), and hence ν(w̄) ≤ n. Let β be the map defined by (2.2.2). Then
f (s, u)β(s, u)w =M sf (s, u)w (by Remark 2.12)
=M suz (since f (s, u)w =M uz)
=M uzy (since suz =M uzy)
=M f (s, u)wy (since uz =M f (s, u)w).
Left cancellation gives β(s, u)w =M wy. Thus, by induction hypothesis, there exists t ∈ S such
that y =M t, and the proof is complete.
Lemma 2.17. Let (M, S) be a pair of type R1 . Then, for each s ∈ S and each x ∈ S ∗ , there exist
t = t(s, x) in S and y = y(s, x) in S ∗ such that sxy =M xyt.
Proof. Let s ∈ S and x ∈ S ∗ . For w ∈ S ∗ , let w̄ denote the element in M represented by w.
By condition (v), there exist y = y(s, x) and z = z(s, x) in S ∗ such that s̄x̄ȳ = x̄ȳz̄. Thus,
we have sxy =M xyz. By Lemma 2.16, there exists t ∈ S such that z =M t. Hence, we have
sxy =M xyt.
Lemma 2.18. Let (M, S) be a pair of right Gaussian monoid M and a finite generating set S for
M , and G the group of fractions of M . Then, for each element x ∈ (S ∪ S −1 )∗ , there exist y, z ∈ S ∗
such that x =G yz −1 .
Proof. Let f be a right l.c.m. selector on S in M . Then, by [10, Theorem 4.2], (S, f ) satisfies
f
condition IIIR , and hence RR (x) exists for any x ∈ (S ∪ S −1 )∗ . In other words, for every x ∈
(S ∪ S ) , there exist y, z ∈ S ∗ such that x =G yz −1 . See the definitions just after Proposition
−1 ∗
f
3.2 in [10] for definitions of RR (x) and condition IIIR .
Lemma 2.19. Let (M, S) be a pair of type R1 , G the right Gaussian group corresponding to M
and f a right l.c.m. selector on S in M . Let s, t ∈ S and x ∈ (S ∪ S −1 )∗ be such that x−1 sx =G t.
Then x−1 sx ≃fR t.
Proof. Let s, t ∈ S and x ∈ (S ∪ S −1 )∗ be such that x−1 sx =G t. By Lemma 2.18, there exist
y, z ∈ S ∗ such that x =G yz −1 . Further, by Lemma 2.17, there exist u ∈ S and w ∈ S ∗ such that
syw =M ywu. Thus w−1 y −1 syw =G u, and hence by Lemma 2.15,
w−1 y −1 syw ≃fR u.
Since x−1 sx =G t and x =G yz −1 , we have y −1 sy =G z −1 tz. Thus w−1 z −1 tzw =G w−1 y −1 syw =G
u. Again, by Lemma 2.15, we have
w−1 z −1 tzw ≃fR u,
and consequently
w−1 y −1 syw ≃fR w−1 z −1 tzw.
Now, using transformation RG and Lemma 2.13 alternatively, we get
x−1 sx ≃fR zww−1 y −1 syww−1 z −1 (since x =G yww−1 z −1 )
≃fR zww−1 z −1 tzww−1 z −1 (since w−1 y −1 syw ≃fR w−1 z −1 tzw)
≃fR t (since zww−1 z −1 =G ǫ, where ǫ is the empty word).
This proves the lemma.
Proof of Theorem 2.14. It follows from remarks 2.11 and 2.12 that if two words in S (S∪S ) are
−1 ∗
right f -equivalent, then they represent the same element in G. For the converse part, let x, y ∈
(S ∪ S −1 )∗ and s, t ∈ S be such that x−1 sx =G y −1 ty. Then yx−1 sxy −1 =G t. By Lemma
2.19, we have yx−1 sxy −1 ≃fR t. Thus, by Lemma 2.13 together with transformation RG , we get
y −1 ty ≃fR y −1 yx−1 sxy −1 y ≃fR x−1 sx.
The following main result is inspired by [24, Theorem 3.1] for the quandle of cords of the plane.
Theorem 2.20. Suppose that (M, S) is a pair of type R1 . Let α and β be maps defined by (2.2.1)
and (2.2.2), respectively. Let G be the right Gaussian group corresponding to M and f a right l.c.m.
selector on S in M . For (s, t) ∈ S × S with s 6= t, let f1 (s, t), f2 (s, t), . . . , fnst (s, t) be elements in S
such that f (s, t) = f1 (s, t)f2 (s, t) · · · fnst (s, t). Then, the Dehn quandle D(S G ) has a presentation
with S as the set of generators and defining relations as follows:
(1) α(s, t) ∗ s = t if s ≤L t,
(2) β(s, t) ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) = s,
(3) u ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) ∗ s = u ∗ fnts (t, s) ∗ fnts −1 (t, s) ∗ · · · ∗ f1 (t, s) ∗ t
for s, t, u ∈ S with s 6= t.
Proving Theorem 2.20 requires some preparation in the form of following lemmas.
Lemma 2.21. Let (M, S) be a pair that satisfies conditions (i), (ii) and (iii). Let α and β be maps
defined by (2.2.1) and (2.2.2), respectively. Let f be a right l.c.m. selector on S in M . Then,
the right f -equivalence on S (S∪S ) is same as the equivalence relation generated by the following
−1 ∗
transformations:
(Tα ) xsα(s, t)s−1 x−1 ←→ xtx−1 ,
where s, t ∈ S with s ≤L t and x ∈ (S ∪ S −1 )∗ .
(Tf,β ) xf (s, t)β(s, t)f (s, t)−1 x−1 ←→ xsx−1 ,
where s, t ∈ S and x ∈ (S ∪ S −1 )∗ .
(T0 ) xsx−1 ←→ ysy −1 ,
where s ∈ S and x, y ∈ (S ∪ S −1 )∗ are words differing by a single use of a relation of the form
tt−1 = ǫ or t−1 t = ǫ for t ∈ S. Here ǫ denotes the empty word.
(Tf ) xsx−1 ←→ ysy −1 ,
where s ∈ S and x, y ∈ (S ∪ S −1 )∗ are words differing by a single use of a relation of the form
uf (u, v) = vf (v, u) for u, v ∈ S.
Proof. For x, y ∈ S (S∪S ) , we write x ≃fT y if x and y are equivalent under the equivalence relation
−1 ∗
generated by Tα , Tf,β , T0 and Tf . By [10, Theorem 4.1], M has a complemented presentation

hS | sf (s, t) = tf (t, s) for s, t ∈ Si .
Let G be the right Gaussian group corresponding to M . Then, G is the quotient of the free monoid
(S ∪ S −1 )∗ by relations ss−1 = 1, s−1 s = 1 and sf (s, t) = tf (t, s) for s, t ∈ S. To prove the
lemma, we need to show that each single transformation Rα , Rf,β and RG can be obtained from a
combination of transformations Tα , Tf,β , T0 and Tf , and vice-versa.
Let s ∈ S and x, y ∈ (S ∪ S −1 )∗ . If x =G y, then there is a sequence x = x1 , x2 , . . . , xn =
y of words such that xi and xi+1 differ by a single use of a relation of the form uu−1 = ǫ or
u−1 u = ǫ or uf (u, v) = vf (v, u) for u, v ∈ S. In this case, by virtue of transformations T0
and Tf , we have xi sx−1
i ≃fT xi+1 sx−1
i+1 for all i. Consequently, xsx
−1 ≃f ysy −1 . This suggests
T
that, for the equivalence relation ≃fT , transformations T0 and Tf can be replaced by the following
transformation:
(TG ) xsx−1 ←→ ysy −1 ,
where s ∈ S and x, y ∈ (S ∪ S −1 )∗ with x =G y. Note that if w ∈ (S ∪ S −1 )∗ , then w−1 ∈ (S ∪ S −1 )∗

and (w−1 )−1 = w. Further, x−1 =G y −1 if and only if x =G y. Thus, transformations TG and RG
are the same. It is now sufficient to prove the following statements:
(1) If x, y ∈ S (S∪S ) are words differing by a single use of Rα or Rf,β , then x ≃fT y.
−1 ∗
(2) If x, y ∈ S (S∪S ) are words differing by a single use of Tα or Tf,β , then x ≃fR y.
−1 ∗
With analogy to Lemma 2.13, we can say that if y, z ∈ S (S∪S ) and x ∈ (S ∪ S −1 )∗ such that
−1 ∗
y ≃fT z, then xyx−1 ≃fT xzx−1 . Since x−1 ∈ (S ∪ S −1 )∗ and (x−1 )−1 = x, we also have x−1 yx ≃fT
x−1 zx whenever y ≃fT z. This together with the following statements will prove statement (1).
(2.2.3) s−1 ts ≃fT α(s, t) for s, t ∈ S with s ≤L t.
(2.2.4) f (s, t)−1 sf (s, t) ≃fT β(s, t) for s, t ∈ S.
Thus, to prove statement (1), it is enough to prove (2.2.3) and (2.2.4). Using transformations Tα
and Tf,β , we have the following:
(2.2.5) sα(s, t)s−1 ≃fT t for s, t ∈ S with s ≤L t.
(2.2.6) f (s, t)β(s, t)f (s, t)−1 ≃fT s for s, t ∈ S.
Let x ∈ (S ∪ S −1 )∗ and y, z ∈ S (S∪S be such that xyx−1 ≃fT z. Then
−1 )∗
x−1 zx ≃fT x−1 xyx−1 x (since z ≃fT xyx−1 )

≃fT y (by transformation TG ).
Thus, x−1 zx ≃fT y whenever xyx−1 ≃fT z. This together with (2.2.5) and (2.2.6) implies (2.2.3) and
(2.2.4), respectively. This proves the statement (1). Statement (2) can be proved similarly.
The following lemma is a direct consequence of quandle axioms. See, for example, [35, Lemma
4.4.8].
Lemma 2.22. Let Q be quandle. Then, any left associated product
u0 ∗δ1 u1 ∗δ2 · · · ∗δm um ∗η1 v1 ∗η2 v2 ∗η3 · · · ∗ηn vn
can be written as
(u0 ∗η1 v1 ∗η2 v2 ∗η3 · · · ∗ηn vn ) ∗δ1 (u1 ∗η1 v1 ∗η2 v2 ∗η3 · · · ∗ηn vn ) ∗δ2 · · · ∗δm (um ∗η1 v1 ∗η2 v2 ∗η3 · · · ∗ηn vn )
where ui , vj ∈ Q and δi , ηj ∈ {−1, 1}.
Proof of Theorem 2.20. Since S generates G, it follows from [11, Proposition 3.2] that S also gen-
erates the Dehn quandle D(S G ). Note that relations (1), (2) and (3) (in the statement of the
theorem) can be written respectively as follows:
(2.2.7) sα(s, t)s−1 = t if s ≤L t,
(2.2.8) f (s, t)β(s, t)f (s, t)−1 = s,
(2.2.9) sf (s, t)uf (s, t)−1 s−1 = tf (t, s)uf (t, s)−1 t−1
for s, t, u ∈ S with s 6= t. By remarks 2.11 and 2.12, relations (2.2.7) and (2.2.8) are in fact
relations in D(S G ), and hence so are relations (1) and (2). Also, since sf (s, t) =G tf (t, s) for any
(s, t) ∈ S × S, (2.2.9) is a relation in D(S G ), and hence so is a relation as in (3). By Theorem
2.14 together with Lemma 2.21, it is sufficient to prove that relations in D(S G ) corresponding to
transformations Tα , Tf,β , T0 and Tf can be obtained by relations (1), (2) and (3). Note that a
relation corresponding to T0 is a trivial relation. Also, a relation corresponding to Tα or Tf,β is
trivial if s = t, and a relation corresponding to Tf is trivial if u = v. The remaining relations can
be written as follows:
(2.2.10) α(s, t) ∗ s ∗δ1 u1 ∗δ2 u2 ∗δ3 · · · ∗δm um = t ∗δ1 u1 ∗δ2 u2 ∗δ3 · · · ∗δm um if s ≤L t,
δ1 δ2 δ3 δm
(2.2.11) β(s, t) ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) ∗ u1 ∗ u2 ∗ ···∗ um
= s ∗δ1 u1 ∗δ2 u2 ∗δ3 · · · ∗δm um ,
(2.2.12) u0 ∗δ1 u1 ∗δ2 · · · ∗δm um ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) ∗ s ∗ǫ1 v1 ∗ǫ2 v2 ∗ǫ3 · · · ∗ǫn vn
= u0 ∗δ1 u1 ∗δ2 · · · ∗δm um ∗ fnts (t, s) ∗ fnts −1 (t, s) ∗ · · · ∗ f1 (t, s) ∗ t ∗ǫ1 v1 ∗ǫ2 v2 ∗ǫ3 · · · ∗ǫn vn
for s, t, ui , vi ∈ S with s 6= t and δi , ǫi ∈ {−1, 1}. Here, note that (2.2.10), (2.2.11) and (2.2.12)
are non-trivial relations corresponding to transformations Tα , Tf,β and Tf , respectively. Relations
(2.2.10) and (2.2.11) can be obtained by relations (1) and (2), respectively. By right cancellation,
(2.2.12) takes the form
(2.2.13) u0 ∗δ1 u1 ∗δ2 · · · ∗δm um ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) ∗ s
= u0 ∗δ1 u1 ∗δ2 · · · ∗δm um ∗ fnts (t, s) ∗ fnts −1 (t, s) ∗ · · · ∗ f1 (t, s) ∗ t.
It follows from Lemma 2.22 that (2.2.13) can be written as
(2.2.14) v0 ∗δ1 v1 ∗δ2 · · · ∗δm vm = w0 ∗δ1 w1 ∗δ2 · · · ∗δm wm ,
where
vi = ui ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) ∗ s
and
wi = ui ∗ fnts (t, s) ∗ fnts −1 (t, s) ∗ · · · ∗ f1 (t, s) ∗ t.
It follows from relations as in (3) that vi = wi for each i. Thus (2.2.14) can be obtained by relations
as in (3), and hence so are (2.2.13) and (2.2.12). This completes the proof of the theorem.
By Proposition 2.10, theorems 2.14 and 2.20 are also true for pairs of types R2 through R9 . For
i ∈ {4, 5, 6, 7}, let (M, A) be a pair of type Ri . Then, for a, b ∈ A, a ≤L b if and only if a = b (see
Remark 2.2). Thus, we have the following corollary of Theorem 2.20.
Corollary 2.23. For i ∈ {4, 5, 6, 7}, let M be a right Gaussian monoid of type Ri . Let A be
the set of atoms in M and β : A × A → A the map defined by (a, b) 7→ (a\b)\(a ∨ b). Let G
be the right Gaussian group corresponding to M and f a right l.c.m. selector on A in M . For
(a, b) ∈ A × A with a 6= b, let f1 (a, b), f2 (a, b), . . . , fnab (a, b) be elements in A such that f (a, b) =
f1 (a, b)f2 (a, b) · · · fnab (a, b). Then, the right Gaussian quandle D(AG ) has a presentation with A as
its set of generators and defining relations as follows:
(1) β(a, b) ∗ fnab (a, b) ∗ fnab −1 (a, b) ∗ · · · ∗ f1 (a, b) = a,
(2) c ∗ fnab (a, b) ∗ fnab −1 (a, b) ∗ · · · ∗ f1 (a, b) ∗ a = c ∗ fnba (b, a) ∗ fnba −1 (b, a) ∗ · · · ∗ f1 (b, a) ∗ b
for a, b, c ∈ A with a 6= b.
For i ∈ {8, 9}, let (N, B) be a pair of type Ri . Then, for a, b ∈ B, a ≤L b if and only if a = b.
Thus, we have the following corollary.
Corollary 2.24. For i ∈ {8, 9}, let N be a Garside monoid of type Ri . Let B be the set of atoms
in N and β : B × B → B the map defined by (a, b) 7→ (a\b)\(a ∨ b). Let H be the Garside group
corresponding to N and f a right l.c.m. selector on B in N . For (a, b) ∈ B × B with a 6= b,
let f1 (a, b), f2 (a, b), . . . , fnab (a, b) be elements in B such that f (a, b) = f1 (a, b)f2 (a, b) · · · fnab (a, b).
Then, the Garside quandle D(B H ) has a presentation with B as its set of generators and defining
relations as:
(1) β(a, b) ∗ fnab (a, b) ∗ fnab −1 (a, b) ∗ · · · ∗ f1 (a, b) = a,
(2) c ∗ fnab (a, b) ∗ fnab −1 (a, b) ∗ · · · ∗ f1 (a, b) ∗ a = c ∗ fnba (b, a) ∗ fnba −1 (b, a) ∗ · · · ∗ f1 (b, a) ∗ b
for a, b, c ∈ A with a 6= b.
2.3. Presentations of left Gaussian quandles and Garside quandles of certain types.
This subsection is a left analogue of the preceding section and we present it without details. We
set the following notations:
• M - a left Gaussian monoid;
• S - a finite generating set for M ;
• A - the set of atoms in M ;
• N - a Garside monoid;
• T - a finite generating set for N ;
• B - the set of atoms in N ;
• ∆ - a Garside element in N ;
• (M, S), (M, A), (N, T ), (N, B), (N, ∆) and (N, T, ∆) - pairs and a triple of objects with
the meaning above.
For a pair (M, S), we assume throughout this subsection that elements in S are pairwise distinct
in M . The same is assumed in case of a pair (N, T ). For elements x and y in a left Gaussian
monoid, denote the right g.c.d. and the left l.c.m. of x and y respectively by x∧ e y and x∨
e y. The
e y = zx. Thus, we have
left residue of x in y (denoted by y/x) is the unique element z satisfying x∨
e y = (y/x)x = (x/y)y .
x∨
Consider the following conditions on a pair (M, S):
(vi) t(s/t) ∈ S whenever (s, t) ∈ S × S and t ≤R s.
(vii) s/t ≤R s∨ e t for all (s, t) ∈ S × S.
e t)/(s/t) ∈ S for all (s, t) ∈ S × S.
(viii) (s∨
(ix) M has a finite homogeneous presentation hS | Ri.
(x) For each s ∈ S and each x ∈ M , there exists y = y(s, x) in M such that yx ≤R yxs.
Define the following terms:
• A pair (M, S) is of
– type L1 if it satisfies conditions (vi), (vii), (viii) and (x).
– type L2 if it satisfies conditions (vi), (vii) and (viii), and if there exist a triple (N, T, ∆)
with T ∆ = ∆T and an epimorphism π : (N, T ) ։ (M, S).
• A pair (N, T ) is of type L3 if it satisfies conditions (vi), (vii) and (viii), and there exists a
Garside element ∆ ∈ N such that T ∆ = ∆T .
• A pair (M, A) is of
– type L4 if it satisfies conditions (vii), (viii) and (x).
– type L5 if it satisfies conditions (vii), (ix) and (x).
– type L6 if it satisfies conditions (vii) and (viii), and if there exist a triple (N, T, ∆) with
– type L7 if it satisfies conditions (vii) and (ix), and if there exist a triple (N, T, ∆) with
• A pair (N, B) is of
– type L8 if it satisfies conditions (vii) and (viii).
– type L9 if it satisfies conditions (vii) and (ix).

For i = 4, 5, 6, 7, we say a left Gaussian monoid M is of type Li if the pair (M, A) is of type Li .
In this case, we also say that the left Gaussian group G corresponding to M and the left Gaussian
quandle D(AG ) are of type Li . Similarly, for i = 8, 9, we say a Garside monoid N is of type type Li
if the pair (N, B) is of type Li . In this case, we also say that the Garside group H corresponding
to N and the Garside quandle D(B H ) are of type Li .
The next proposition is analogous to Proposition 2.10.
Proposition 2.25. Pairs of types L2 through L9 are of type L1 .
Let (M, S) be a pair that satisfies condition (vi). Then, we have a map
α : {(s, t) ∈ S × S | t ≤R s} → S
defined by
(2.3.1) (s, t) 7→ t(s/t).
Note that α(s, s) = s for all s ∈ S.
Remark 2.26. Let s, t ∈ S be such that t ≤R s. Then s = (s/t)t and α(s, t) = t(s/t). Suppose f
be a left l.c.m. selector on S in M . Since f (s, t) represents the element s/t, we have s =M f (s, t)t
and α(s, t) =M tf (s, t). Thus tst−1 =G tf (s, t) =G α(s, t), where G is the left Gaussian group
corresponding to M .
Let (M, S) be a pair that satisfies conditions (vii) and (viii). Then, we have a map β : S × S → S
defined by
(2.3.2) e t)/(s/t).
(s, t) 7→ (s∨
Note that β(s, s) = s for all s ∈ S.

Remark 2.27. Let s and t be elements in S. Then s∨ e t = (s∨
e t)/(s/t) (s/t). In other words,
(s/t)t = (s∨ e t)/(s/t) (s/t). Suppose f be a left l.c.m. selector on S in M . Since f (s, t)
represents the element s/t and β(s, t) = (s∨ e t)/(s/t), we have f (s, t)t =M β(s, t)f (s, t). Thus
−1
f (s, t)tf (s, t) =G β(s, t), where G is the left Gaussian group corresponding to M .
The next theorem is analogous to Theorem 2.20.
Theorem 2.28. Suppose (M, S) be a pair of type L1 . Let α and β be maps defined by (2.3.1) and
(2.3.2), respectively. Suppose G be the left Gaussian group corresponding to M and f a left l.c.m.
selector on S in M . For (s, t) ∈ S × S with s 6= t, let f1 (s, t), f2 (s, t), . . . , fnst (s, t) be elements in S
such that f (s, t) = f1 (s, t)f2 (s, t) · · · fnst (s, t). Then, the Dehn quandle D(S G ) has a presentation
with S as its set of generators and defining relations as:
(1) s ∗ t = α(s, t) if t ≤R s,
(2) t ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) = β(s, t),
(3) u ∗ t ∗ fnst (s, t) ∗ fnst −1 (s, t) ∗ · · · ∗ f1 (s, t) = u ∗ s ∗ fnts (t, s) ∗ fnts −1 (t, s) ∗ · · · ∗ f1 (t, s)
for s, t, u ∈ S with s 6= t.
In view of Proposition 2.25, Theorem 2.28 is also true for pairs of types L2 through L9 . Further,
we can deduce corollaries of Theorem 2.28 analogous to corollaries 2.23 and 2.24.
2.4. Examples of presentations of Garside quandles. Theorem 2.20 and Corollary 2.24 can
be used to write presentations of certain Garside quandles.
Example 2.29. Let A be an Artin group with Artin presentation
(2.4.1) hS | (st)mst = (ts)mts for s, t ∈ S with s 6= ti ,
where S is a finite set, (uv)k denotes the word uvuvu · · · of length k, mst = mts ≥ 2 for all s 6= t
and mst = ∞ if there is no relation. Let M be the monoid with the same presentation as above,
i.e. M is the quotient of the free monoid S ∗ by relations (st)mst = (ts)mts for all s 6= t. We refer
to such a monoid as an Artin monoid and call it of spherical type if the corresponding Artin group
is of spherical type. Further, we call the Dehn quandle D(S A ) as an Artin quandle and say that it
is of spherical type if the Artin group A is of spherical type. Note that, in these definitions, S is
an Artin generating set for A.
For the further discussion, we assume that M is a spherical Artin monoid. It follows from [5]
(see also [10, Example 1]) that M is a Garside monoid. Note that, for s 6= t, the words (st)mst
and (ts)mts both represent the right l.c.m. of s and t. It is easy to see that the pair (M, S)
satisfies conditions (ii) and (iv). By Lemma 2.3, the set S is the set of atoms in M. Thus, M is
a Garside monoid of type R9 , and hence the Artin quandle D(S A ) is a Garside quandle of type
R9 . Again by Lemma 2.3, the pair (M, S) satisfies condition (iii), and thus we have the map
β : S × S → S defined by (s, t) 7→ (s\t)\(s ∨ t). Let f : S × S → S ∗ be the map defined by
f (s, t) = ǫ if s = t and f (s, t) = (ts)mst −1 if s 6= t, where ǫ is the empty word. For s 6= t, since
s(ts)mst −1 = (st)mst represents s(s\t) = s ∨ t, the word (ts)mst −1 must represent s\t. Thus, f is
a right l.c.m. selector on S in M. For s 6= t and i = 0, 1, . . . , mst , let fi (s, t) = s if i is even and
fi (s, t) = t if i is odd. Then, for any s 6= t and any i = 1, 2, . . . , mst , we have fi (s, t) = fi−1 (t, s)
and f (s, t) = f1 (s, t)f2 (s, t) · · · fmst −1 (s, t). Let s 6= t be elements in S. Then
f (s, t)fmst (s, t) = (ts)mst −1 fmst (s, t)

= (ts)mts −1 fmts −1 (t, s)
= tf1 (t, s)f2 (t, s) · · · fmts −1 (t, s)
= tf (t, s)
= t(st)mts −1
= (ts)mts .
Since (M, S) satisfies condition (ii), we have (s\t) ((s\t)\(s ∨ t)) = s ∨ t. Since f (s, t)fmst (s, t) =
(ts)mts represents (s\t) ((s\t)\(s ∨ t)) = s ∨ t and f (s, t) represents s\t, it follows that fmst (s, t)
represents (s\t)\(s ∨ t). Thus, we have β(s, t) = fmst (s, t). By Corollary 2.24, the Artin quandle
D(S A ) has a presentation with S as its set of generators and defining relations as follows:
(2.4.2) fmst (s, t) ∗ fmst −1 (s, t) ∗ · · · ∗ f1 (s, t) = f0 (s, t) and

(2.4.3) u ∗ fmst −1 (s, t) ∗ fmst −2 (s, t) ∗ · · · ∗ f1 (s, t) ∗ f0 (s, t)
= u ∗ fmts −1 (t, s) ∗ fmts −2 (t, s) ∗ · · · ∗ f1 (t, s) ∗ f0 (t, s)
for s, t, u ∈ S with s 6= t (note that s = f0 (s, t) and t = f0 (t, s)). We shall show that (2.4.3) can be
obtained by (2.4.2). Let s, t, u ∈ S with s 6= t. Then
u ∗ fmst −1 (s, t) ∗ fmst −2 (s, t) ∗ · · · ∗ f1 (s, t) ∗ f0 (s, t)
= (u ∗ fmst −1 (s, t) ∗ fmst −2 (s, t) ∗ · · · ∗ f1 (s, t)) ∗ (fmst (s, t) ∗ fmst −1 (s, t) ∗ · · · ∗ f1 (s, t))
= u ∗ fmst (s, t) ∗ fmst −1 (s, t) ∗ · · · ∗ f1 (s, t)
= u ∗ fmts −1 (t, s) ∗ fmts −2 (t, s) ∗ · · · ∗ f1 (t, s) ∗ f0 (t, s),
where the first equality is by (2.4.2), the second is by Lemma 2.22 and third is by relations fi (s, t) =
fi−1 (t, s). This shows that (2.4.3) is redundant. Hence, if A is spherical, then the Artin quandle
D(S A ) has a presentation
D
(2.4.4) D(S A ) = S | (s ∗ t)mst = s if mst is even, and
E
(t ∗ s)mst = s if mst is odd for all s 6= t ,
where (u ∗ v)k denotes the left associated product u ∗ v ∗ u ∗ v ∗ u · · · of length k.

Example 2.30. Let M be the Garside monoid as in Example 2.5 and G its group of fractions.
The set S = {x1 , x2 , . . . , xn } is the set of atoms in M (see the proof of Proposition 2.4, i.e. [10,
Proposition 5.2]). It can be seen that, for each i 6= j, the word xpi i represents xi ∨xj . Let f (xi , xi ) = ǫ
for i = 1, 2, . . . , n and f (xi , xj ) = xpi i −1 for i 6= j, where ǫ is the empty word. Then, f is a right
l.c.m. selector on S in M . Let β(xi , xj ) = xi for any i and j. Then xi f (xi , xj ) = f (xi , xj )β(xi , xj )
which, of course, represents xi ∨ xj . In other words, β(xi , xj ) = (xi \xj )\(xi ∨ xj ). Then, by
Corollary 2.24, the Garside quandle D(S G ) has a presentation
D(S G ) = hx1 , x2 , . . . , xn | xi ∗pj xj = xi for all i 6= ji.
Example 2.31. Let M be the Garside monoid as in Example 2.6 and G its group of fractions. The
set S = {x1 , x2 , . . . , xn } is the set of atoms in M . For an integer i, let xi = xj for 1 ≤ j ≤ n such
that i ≡ j (mod n). One can see that, for each i 6= j, the word xi xi+1 · · · xi+m−1 represents xi ∨ xj .
Let f (xi , xi ) = ǫ for i = 1, 2, . . . , n and f (xi , xj ) = xi+1 xi+2 · · · xi+m−1 for i 6= j, where ǫ is the
empty word. Then, f is a right l.c.m. selector on S in M . Let β(xi , xi ) = xi for i = 1, 2, . . . , n and
β(xi , xj ) = xi+m for i 6= j. Then xi f (xi , xj ) = f (xi , xj )β(xi , xj ), i.e. β(xi , xj ) = (xi \xj )\(xi ∨ xj ).
Then, by Corollary 2.24, D(S G ) is generated by S and has defining relations
xi+m ∗ xi+m−1 ∗ · · · ∗ xi+1 = xi ,
xi ∗ xj+m−1 ∗ xj+m−2 ∗ · · · ∗ xj = xi ∗ xk+m−1 ∗ xk+m−2 ∗ · · · ∗ xk
for all i, j, k = 1, 2, . . . , n with j 6= k. Note that relations of second type are redundant. For, we
have
xi ∗ xj+m−1 ∗ xj+m−2 ∗ · · · ∗ xj = (xi ∗ xj+m−1 ∗ xj+m−2 ∗ · · · ∗ xj+1 ) ∗ (xj+m ∗ xj+m−1 ∗ · · · ∗ xj+1 )
= xi ∗ xj+m ∗ xj+m−1 ∗ · · · ∗ xj+1
for all i, j = 1, 2, . . . , n, where the first equality follows from relations of first type and the second
follows from Lemma 2.22. Iterating the process yields relations of second type, and hence the final
presentation is
D(S G ) = hx1 , x2 , . . . , xn | xi+m ∗ xi+m−1 ∗ · · · ∗ xi+1 = xi for i = 1, 2, . . . , ni.
Example 2.32. Let M be the Garside monoid as in Example 2.8 and G be its group of fractions.
The set S = {x1 , x2 , y1 , y2 , y3 } is the set of atoms in M . Let p1 = 2 and p2 = 5. For an integer
i, let yi = yj for 1 ≤ j ≤ 3 such that i ≡ j (mod 3). Using Corollary 2.24 and after reducing the
relations, we get a presentation
D
D(S G ) = x1 , x2 , y1 , y2 , y3 | yk+4 ∗ yk+3 ∗ yk+2 ∗ yk+1 = yk , xi ∗pj xj = xi ,
E
xi ∗ y1 ∗ y3 ∗ y2 ∗ y1 = xi , yk+4 ∗pi xi = yk for i, j = 1, 2 with i 6= j and k = 1, 2, 3 .
Example 2.33. Let M be the Garside monoid as in Example 2.9 and G its group of fractions. The
set S = {x1 , x2 , x3 , y1 , y2 , y3 } is the set of atoms in M . Let ∆1 and ∆2 be minimal Garside elements
in underling Artin monoids M1 and M2 of type B3 and A3 , respectively. Then, the words (x1 x2 x3 )3 ,
(x2 x3 x1 )3 and (x3 x1 x2 )3 all represent ∆1 , and the words y1 y2 y3 y1 y2 y1 , (y1 y3 y2 )2 , (y2 y1 y3 )2 and
(y3 y1 y2 )2 all represent ∆2 (see [4, examples 3 and 4] for the minimal Garside elements in Artin
monoids of spherical type). Note that M = (M1 AM2 )/≡, where ≡ is the equivalence relation on
M1 AM2 generated by ∆21 = ∆32 . One can see that both ∆21 and ∆32 represent xi ∨ yj in M for
i, j = 1, 2, 3. For an integer i, let xi = xj and yi = yj for 1 ≤ j ≤ 3 such that i ≡ j (mod 3).
Define f (xi , yj ) = xi+1 xi+2 · · · xi+17 , f (yk , xj ) = yk+17 yk+16 · · · yk+1 and f (y3 , xj ) = y1 y2 · · · y17 for
i, j = 1, 2, 3 and k = 1, 2. Then f can be extended to a right l.c.m. selector on S in M . For
i = 1, 2, let ψi be the permutation of the set of divisors of ∆i given by x 7→ x\∆i , and φi be the
automorphism of Mi defined by φi (x) = ψi2 (x) for a divisor x of ∆i (see [9, lemmas 2.2 and 2.3]).
One can verify that xψi (x) = ∆i = ψi (x)φi (x) and x∆i = ∆i φi (x) for a divisor x of ∆i . Note that
φi maps atoms to atoms. Let z1 = y3 , z2 = y2 and z3 = y1 . It is easy to see that φ1 (xi ) = xi and
φ2 (yi ) = zi for i = 1, 2, 3. Thus
xi ψ1 (xi )∆1 = ∆21 = ψ1 (xi )∆1 xi
and
yi ψ2 (yi )∆22 = ∆32 = ψ2 (yi )∆22 zi
for i = 1, 2, 3. Define β(xi , yj ) = xi and β(yj , xi ) = zj for i, j = 1, 2, 3. Then β(xi , yj ) =
(ψ1 (xi )∆1 )\∆21 = (xi \yj )\(xi ∨ yj ) and β(yj , xi ) = (ψ2 (yj )∆22 )\∆32 = (yj \xi )\(yj ∨ xi ) for all
i, j = 1, 2, 3. Note that β can be extended to the map S × S → S defined as in (2.2.2). Using
Corollary 2.24 and after reducing the relations, we see that D(S G ) has a presentation
D
x1 , x2 , x3 , y 1 , y 2 , y 3 | x1 ∗ x2 ∗ x1 ∗ x2 = x1 ∗ x3 = x1 , x2 ∗ x1 ∗ x2 ∗ x1 = x3 ∗ x2 ∗ x3 = x2 ,
x3 ∗ x1 = x2 ∗ x3 ∗ x2 = x3 , y 2 ∗ y 1 ∗ y 2 = y 1 ∗ y 3 = y 1 , y 1 ∗ y 2 ∗ y 1 = y 3 ∗ y 2 ∗ y 3 = y 2 ,
y3 ∗ y1 = y2 ∗ y3 ∗ y2 = y3 , xi+18 ∗ xi+17 ∗ · · · ∗ xi+1 = xi , y1 ∗ y2 ∗ · · · ∗ y12 = y1 ,
y2 ∗ y3 ∗ · · · ∗ y19 = y2 , y12 ∗ y11 ∗ · · · ∗ y1 = y3 , xi ∗ y2 ∗ y3 ∗ · · · ∗ y19 = xi ,
y1 ∗ x18 ∗ x17 ∗ · · · ∗ x1 = y1 ∗ y2 ∗ · · · ∗ y7 , y2 ∗ x18 ∗ x17 ∗ · · · ∗ x1 = y2 ,
E
y3 ∗ x18 ∗ x17 ∗ · · · ∗ x1 = y1 for i = 1, 2, 3 .
3. Presentations of Dehn quandles of groups

In this section, we prove a general result giving presentations of Dehn quandles of groups when the
centraliser of each generator is known. Although the result is general, determining generating sets
for centralisers of elements in interesting classes of groups like Garside groups and Artin groups
is usually challenging. See, for example, [4, 14, 15, 17, 19, 28, 32] for related works. Further,
presentations obtained for Garside quandles using Theorem 3.1 usually have larger number of
relations than the one given by Theorem 2.20. We shall see many examples later in this section.
3.1. Presentations of Dehn quandles. The following theorem gives a presentation of the Dehn
quandle of a group G with respect to a generating set S when a generating set for the centraliser
of each element in S is known.
Theorem 3.1. Let G be a group with a presentation hS | Ri. For s ∈ S, let As be a generating
set for the centraliser CG (s) of s in G. Let T = {(s, t) ∈ S × S | s and t are conjugate in G}, and
f : T → G a map such that
• f (t, s)sf (t, s)−1 = t,
• f (t, s) = f (s, t)−1 ,
• f (u, t)f (t, s) = f (u, s)
for all (s, t), (t, u) ∈ T . Then, the Dehn quandle D(S G ) has a presentation with the generating set
S and defining relations as follows:
(1) For each s ∈ S and each relation r = sδr11 sδr22 · · · sδrkk in R, where sri ∈ S and δi ∈ {−1, 1}, we
have
s ∗δk srk ∗δk−1 srk−1 ∗δk−2 · · · ∗δ1 sr1 = s.
(2) For each s ∈ S and each w = sǫw11 sǫw22 · · · sǫwll in As , where swi ∈ S and ǫi ∈ {−1, 1}, we have
s ∗ǫl swl ∗ǫl−1 swl−1 ∗ǫl−2 · · · ∗ǫ1 sw1 = s.
(3) For each (s, t) ∈ T such that f (t, s) = f1 (t, s)µ1 f2 (t, s)µ2 · · · fn (t, s)µn , where fi (t, s) ∈ S and
µi ∈ {−1, 1}, we have
s ∗µn fn (t, s) ∗µn−1 fn−1 (t, s) ∗µn−2 · · · ∗µ1 f1 (t, s) = t.
Proof. Let ψ : S → S be a bijection of S onto another set S, where, for brevity, we denote ψ(s)
by s. Let (Q, ⋆) be a quandle that has a presentation with the set of generators S and defining
relations as in (1), (2) and (3) written in terms of elements of S. We write elements of Q in bold
to differentiate them from elements of D(S G ). We claim that D(S G ) ∼ = Q.
Let φ : S → S given by φ(s) = s be the inverse of ψ. The map φ induces a quandle homomorphism
φ : Q → D(S G ), which is clearly surjective. It suffices to show that φ is injective.
We first claim that if x = sηx11 sηx22 · · · sηxnn and y = sθy11 sθy22 · · · sθym
m
, where sxi , syj ∈ S and ηi , θj ∈
{1, −1}, represent the same element of G, then
s ⋆ηn sxn ⋆ηn−1 sxn−1 ⋆ηn−2 · · · ⋆η1 sx1 = s ⋆θm sym ⋆θm−1 sym−1 ⋆θm−2 · · · ⋆θ1 sy1
for all s ∈ S.
Since x = y in G, there is a sequence of elements x = g0 , g1 , g2 , . . . , gℓ = y in G such that gi
and gi+1 differ by a single relation from R ∪ R′ , where R′ = {s−1 s, ss−1 | s ∈ S} is the set of
trivial relations. Thus, it is enough to consider the case when x and y differ by a single relation
η
i.e. when x = sηx11 sηx22 · · · sηxnn and y = sηx11 sηx22 · · · sηxii rsxi+1 ηn ′ ′
i+1 · · · sxn for some r ∈ R ∪ R . If r ∈ R , then
the claim holds by the second quandle axiom (i.e. bijectivity of right multiplication). Suppose that
r = sδr11 sδr22 · · · sδrkk is an element of R. Then, Lemma 2.22 gives
s ⋆ηn sxn ⋆ηn−1 · · · ⋆ηi+1 sxi+1 ⋆δk srk ⋆δk−1 srk−1 ⋆δk−2 · · · ⋆δ1 sr1 ⋆ηi sxi ⋆ηi−1 · · · ⋆η1 sx1
= (s ⋆δk srk ⋆δk−1 srk−1 ⋆δk−2 · · · ⋆δ1 sr1 ) ⋆ηn (sxn ⋆δk srk ⋆δk−1 srk−1 ⋆δk−2 · · · ⋆δ1 sr1 ) ⋆ηn−1 · · · ⋆ηi+1
(sxi+1 ⋆δk srk ⋆δk−1 srk−1 ⋆δk−2 · · · ⋆δ1 sr1 ) ⋆ηi sxi · · · ⋆η1 sx1
= s ⋆ηn sxn ⋆ηn−1 · · · ⋆ηi+1 sxi+1 ⋆ηi sxi ⋆ηi−1 · · · ⋆η1 sx1 , by relation of type (1).
This proves the claim.

Now, to prove injectivity of φ, take two elements s ⋆αp sap ⋆αp−1 sap−1 ⋆αp−2 · · · ⋆α1 sa1 and
t ⋆βq sbq ⋆βq−1 sbq−1 ⋆βq−2 · · · ⋆β1 sb1 in Q such that
(3.1.1) s ∗αp sap ∗αp−1 sap−1 ∗αp−2 · · · ∗α1 sa1 = t ∗βq sbq ∗βq−1 sbq−1 ∗βq−2 · · · ∗β1 sb1
in D(S G ). Rewriting (3.1.1) gives
(3.1.2) s ∗αp sap ∗αp−1 sap−1 ∗αp−2 · · · ∗α1 sa1 ∗−β1 sb1 ∗−β2 sb2 ∗−β3 · · · ∗−βq sbq = t.
Since s and t are conjugate in G, we have (s, t) ∈ T . Thus, we have
(3.1.3) s ∗µn fn (t, s) ∗µn−1 fn−1 (t, s) ∗µn−2 · · · ∗µ1 f1 (t, s) = t,
where f (t, s) ∈ G such that f (t, s) = f1 (t, s)µ1 f2 (t, s)µ2 · · · fn (t, s)µn for fi (t, s) ∈ S and µi ∈
{−1, 1}. Using (3.1.2) and (3.1.3), we have
s = s ∗αp sap ∗αp−1 sap−1 ∗αp−2 · · · ∗α1 sa1 ∗−β1 sb1 ∗−β2 sb2 ∗−β3 · · · ∗−βq sbq
(3.1.4) ∗−µ1 f1 (t, s) ∗−µ2 f2 (t, s) ∗−µ3 · · · ∗−µn fn (t, s).
Writing (3.1.4) in terms of conjugation in G implies that the element
−β −β α
fn (t, s)−µn fn−1 (t, s)−µn−1 · · · f1 (t, s)−µ1 sbq q sbq−1q−1 · · · s−β 1 α1 α2 p
b1 sa1 sa2 · · · sap
commutes with s, and hence equals to an element, say, w of CG (s). Without loss of generality, we
can assume that w ∈ As . If not, then w will be a product of elements of As and the argument will
be similar. Let w = s∂w11 s∂w22 · · · s∂wll written in terms of generators S of G. Then the equality
−β −β α
s∂w11 s∂w22 · · · s∂wll = fn (t, s)−µn fn−1 (t, s)−µn−1 · · · f1 (t, s)−µ1 sbq q sbq−1q−1 · · · s−β 1 α1 α2 p
b1 sa1 sa2 · · · sap
of elements in G can be rewritten as

β α
(3.1.5) sβb11 sβb22 · · · sbqq f1 (t, s)µ1 f2 (t, s)µ2 · · · fn (t, s)µn s∂w11 s∂w22 · · · s∂wll = sαa11 sαa22 · · · sapp .
Our earlier proved claim gives
s ⋆αp sap ⋆αp−1 sap−1 ⋆αp−2 · · · ⋆α1 sa1
= s ⋆∂l swl ⋆∂l−1 swl−1 ⋆∂l−2 · · · ⋆∂1 sw1 ⋆µn fn (t, s) ⋆µn−1 fn−1 (t, s) ⋆µn−2 · · · ⋆µ1 f1 (t, s)
⋆βq sbq ⋆βq−1 sbq−1 ⋆βq−2 · · · ⋆β1 sb1
= s ⋆µn fn (t, s) ⋆µn−1 fn−1 (t, s) ⋆µn−2 · · · ⋆µ1 f1 (t, s) ⋆βq sbq ⋆βq−1 sbq−1 ⋆βq−2 · · · ⋆β1 sb1 ,
by relation of type (2)
= t ⋆βq sbq ⋆βq−1 sbq−1 ⋆βq−2 · · · ⋆β1 sb1 , by relation of type (3).
This completes the proof of the theorem.
Corollary 3.2. If G = hS | Ri is a finitely presented group such that the centraliser of each
generator from S is finitely generated, then the Dehn quandle D(S G ) is finitely presented.
It follows from [15, Theorem 12] that the centraliser of an element in a Garside group is finitely
generated. Thus, the corollary holds for Garside groups, which includes spherical Artin groups.
Remark 3.3. One can reduce the number of relations in Theorem 3.1 as follows:
(1) Relations of type (2) need to be checked only for those generators in S that represent
distinct conjugacy classes in G.
(2) We write S = ⊔i Xi as a disjoint union of sets where Xi consists of elements that are
conjugate to each other. We equip each Xi with a partial order such that the corresponding
poset graph is a tree. Then, for each i, relations of type (3) need to be checked only for
elements in Xi × Xi that are adjacent in the corresponding poset graph.
3.2. Examples of presentations of Dehn quandles. As applications of Theorem 3.1, we give

several examples of presentations of Dehn quandles.
Example 3.4. Let Core (Zn ) be the core quandle of the cyclic group of order n (also called the
dihedral quandle of order n). Then, the quandle operation in Core (Zn ) is given by x ∗ y = 2y − x
mod n. It is shown in [11, Proposition 3.11] that the core quandle of any group is a Dehn quandle
of some group. Let Dn = hs1 , s2 | s21 = 1, s22 = 1, (s1 s2 )n = 1i be the Coxeter presentation of the
dihedral group of order 2n, and set S = {s1 , s2 }. It follows from a direct computation (see also [3,
Lemma 6.3]) that Core(Zn ) ∼= D(S Dn ).
First consider the case when n is odd. In this case, s1 and s2 are conjugate, and CDn (si ) =
n−1
hsi i. Set s1 < s2 and f (s2 , s1 ) = (s1 s2 ) 2 . Let (s1 ∗ s2 )n denote the left associated product
s1 ∗ s2 ∗ s1 ∗ s2 ∗ · · · of length n. Then Theorem 3.1 gives the following set of relations:
(3.2.1) si ∗ sj ∗ sj = si for i, j = 1, 2,
(3.2.2) si ∗ s2 ∗ s1 ∗ s2 ∗ s1 ∗ · · · ∗ s2 ∗ s1 = si for i = 1, 2,
| {z }
2n terms
(3.2.3) (s1 ∗ s2 )n = s2 .
Note that (3.2.1) for i = j are trivial relations. Since n is odd, using (3.2.1), the relation (3.2.3)
can be turned into the relation (s2 ∗ s1 )n = s1 . Now, using (3.2.3) and (s2 ∗ s1 )n = s1 , we can
recover (3.2.2). Thus, the final presentation is
∼ hs1 , s2 | s2 ∗ s1 ∗ s1 = s2 , s1 ∗ s2 ∗ s2 = s1 , (s1 ∗ s2 )n = s2 i.
Core(Zn ) =
Suppose now that n is even. In this case, s1 and s2 lie in different conjugacy classes. Also,
n
CDn (si ) = hsi , (s1 s2 ) 2 i. Again, Theorem 3.1 gives the following set of relations:
(3.2.4) si ∗ sj ∗ sj = si for i, j = 1, 2,
(3.2.5) si ∗ s2 ∗ s1 ∗ s2 ∗ s1 ∗ · · · ∗ s2 ∗ s1 = si for i = 1, 2,
| {z }
2n terms
(3.2.6) s1 ∗ s2 ∗ s1 ∗ s2 ∗ s1 ∗ · · · ∗ s2 ∗ s1 = s1 ,
| {z }
n terms
(3.2.7) s2 ∗ s2 ∗ s1 ∗ s2 ∗ s1 ∗ · · · ∗ s2 ∗ s1 = s2 .
| {z }
n terms
Again we ignore the trivial relations. Further, using (3.2.6) and (3.2.7), we can recover relations in
(3.2.5). Hence, the final presentation is
Core(Zn ) ∼
= hs1 , s2 | s2 ∗ s1 ∗ s1 = s2 , s1 ∗ s2 ∗ s2 = s1 , (s1 ∗ s2 )n = s1 , (s2 ∗ s1 )n = s2 i.
Example 3.5. The fundamental group of a closed orientable surface Sg of genus g ≥ 2 has a
presentation
π1 (Sg ) = ha1 , b1 , a2 , b2 , . . . , ag , bg | [a1 , b1 ][a2 , b2 ] · · · [ag , bg ] = 1i.
Setting S = {a1 , b1 , a2 , b2 , . . . , ag , bg }, we note that no two elements of S are conjugate to each

other. Further, it is known that Cπ1 (Sg ) (ai ) = hai i and Cπ1 (Sg ) (bi ) = hbi i [13, Section 1.1.3]. Thus,
by Theorem 3.1, we get
D
D(S π1 (Sg ) ) = a1 , b1 , a2 , b2 , . . . , ag , bg | ai ∗ bg ∗ ag ∗−1 bg ∗−1 ag ∗ · · · ∗ b1 ∗ a1 ∗−1 b1 ∗−1 a1 = ai ,
E
(3.2.8) bi ∗ bg ∗ ag ∗−1 bg ∗−1 ag ∗ · · · ∗ b1 ∗ a1 ∗−1 b1 ∗−1 a1 = bi for all i .
We note that the presentation of the enveloping group of D(S π1 (Sg ) ) given in [11, Theorem 3.18]
can be recovered using (3.2.8) and [35, Theorem 5.1.7].
Example 3.6. For n ≥ 3, recall the Artin presentation of the braid group
Bn = hσ1 , σ2 , . . . , σn−1 | σi σj = σj σi for |i − j|≥ 2, σi σi+1 σi = σi+1 σi σi+1 for all ii.
Let us set S = {σ1 , σ2 , . . . , σn−1 } and the total ordering σ1 < σ2 < · · · < σn−1 on S. Note that all
the elements of S are conjugate to each other. Choose f (σi+1 , σi ) = σi σi+1 for each 1 ≤ i ≤ n − 2.
It follows from [19, Theorem 4] that CBn (σ1 ) is generated by X ∪ Y , where
X = {σ1 , σ3 , σ4 , . . . , σn−1 , σ2 σ1 σ1 σ2 }
and
Y = {σ12 , (σ2 σ1 )(σ3 σ2 ) · · · (σr+1 σr )(σr σr+1 ) · · · (σ2 σ3 )(σ1 σ2 ) for 1 ≤ r ≤ n − 2}.
It follows from the discussion after [19, Theorem 4] that generators from Y can be written in terms
of generators from X. In fact, setting xr = (σ2 σ1 )(σ3 σ2 ) · · · (σr+1 σr )(σr σr+1 ) · · · (σ2 σ3 )(σ1 σ2 ), one
−1
can show that xr = xr−1 σr+1 σr · · · σ4 σ3 x1 σ3−1 σ4−1 · · · σr−1 σr+1 for each r ≥ 2. Thus, we obtain
(3.2.9) CBn (σ1 ) = hσ1 , σ3 , σ4 , . . . , σn−1 , σ2 σ1 σ1 σ2 i.
First we consider n = 3 to clarify the general idea. We have f (σ2 , σ1 ) = σ1 σ2 and CB3 (σ1 ) =
hσ1 , σ2 σ1 σ1 σ2 i. Thus, by Theorem 3.1, the defining relations in D(S B3 ) are given by
(3.2.10) σi ∗ σ1 ∗ σ2 ∗ σ1 ∗−1 σ2 ∗−1 σ1 ∗−1 σ2 = σi for i = 1, 2,
(3.2.11) σ1 ∗ σ2 ∗ σ1 ∗ σ1 ∗ σ2 = σ1 ,
(3.2.12) σ1 ∗ σ2 ∗ σ1 = σ2 .
By using (3.2.12) in (3.2.11), we get
(3.2.13) σ2 ∗ σ1 ∗ σ2 = σ1 .
It follows from relations (3.2.12) and (3.2.13) that (3.2.10) is redundant. Thus, we obtain
D(S B3 ) = hσ1 , σ2 | σ1 ∗ σ2 ∗ σ1 = σ2 , σ2 ∗ σ1 ∗ σ2 = σ1 i.
Now, assume that n ≥ 4. In this case, defining relations for D(S Bn ) are as follows:
(3.2.14) σk ∗ σi ∗ σi+1 ∗ σi = σk for 1 ≤ k ≤ n − 1 and 1 ≤ i ≤ n − 2,
−1 −1 −1
∗ σi+1 ∗ σi ∗ σi+1
(3.2.15) σk ∗ σi ∗ σj ∗−1 σi ∗−1 σj = σk for 1 ≤ k ≤ n − 1 and 3 ≤ i + 2 ≤ j ≤ n − 1,
(3.2.16) σ1 ∗ σj = σ1 for 3 ≤ j ≤ n − 1,
(3.2.17) σ1 ∗ σ2 ∗ σ1 ∗ σ1 ∗ σ2 = σ1 ,
(3.2.18) σi ∗ σi+1 ∗ σi = σi+1 for 1 ≤ i ≤ n − 2.
Using relations in (3.2.18), we can recover relations in (3.2.14), and the relation (3.2.17) can be
rewritten as σ2 ∗ σ1 ∗ σ2 = σ1 . Taking k = i − 1 and using induction on n, we can rewrite relations
in (3.2.15) as σi ∗ σj = σi for 3 ≤ i + 2 ≤ j ≤ n − 1. The remaining relations in (3.2.15) can be
recovered using these relations. Thus, the presentation D(S Bn ) is
D
(3.2.19) D(S Bn ) = σ1 , . . . , σn−1 | σi ∗ σi+1 ∗ σi = σi+1 for 1 ≤ i ≤ n − 2,
E
σ2 ∗ σ1 ∗ σ2 = σ1 , σi ∗ σj = σi for 3 ≤ i + 2 ≤ j ≤ n − 1 .
Remark 3.7. Using induction on n, one can show that the presentation for D(S Bn ) given by
(2.4.4) can also be reduced to the one given in (3.2.19).
Example 3.8. Recall the presentation of an Artin group from (2.4.1). We say that A is a right
angled Artin group if each mij is either 2 or ∞. Setting S = {x1 , x2 , . . . , xn } to be the Artin
generating set of A, we see that no two elements from S are conjugate to each other. Further,
CA (xi ) = hxj | mij = 2i. Thus, by Theorem 3.1, D(S A ) has a presentation with generating set S
and defining relations as
(3.2.20) xk ∗ xi ∗ xj ∗−1 xi ∗−1 xj = xk for 1 ≤ k ≤ n and mij = 2,
(3.2.21) xi ∗ xj = xi for mij = 2.
Clearly, relations in (3.2.20) can be recovered from relations in (3.2.21). Thus, the presentation is
(3.2.22) D(S A ) = hx1 , x2 , . . . , xn | xi ∗ xj = xi whenever mij = 2i.
In view of examples 2.29 and 3.8, we propose the following.
Conjecture 3.9. If A = hS | (st)mst = (ts)mts for s, t ∈ S with s 6= ti is an Artin group, then its
Dehn quandle D(S A ) has a presentation
D(S A ) = hS | (s ∗ t)mst = s if mst is even, and (t ∗ s)mst = s if mst is odd for all s 6= ti.
Next, we consider Dehn quandles of surfaces. Let Sg,p be an orientable surface of genus g with
p marked points. The mapping class group Mg,p of Sg,p is defined as the set of isotopy classes
of orientation preserving self-homeomorphisms of Sg,p which permute the set of marked points.
A simple closed curve α on Sg,p is said to be non-separating if Sg,p \ α is connected, and called
separating otherwise. Separating and non-separating simple closed arcs are defined analogously.
ns be the set of isotopy classes of all non-separating simple closed curves and simple closed
Let Dg,p
arcs in Sg,p . For each isotopy class y ∈ Dg,pns of a non-separating simple closed curve, let T be the
y
isotopy class of the right hand Dehn twist along any non-separating simple closed curve representing
y. Similarly, if y is the isotopy class of a non-separating simple closed arc, then we define Hy to be
the isotopy class of the anti-clockwise half twist about the punctures joined by any arc representing
y. This defines an injective map
ns
τ : Dg,p ֒→ Mg,p
by setting
(
Tx if x is the isotopy class of a non-separating simple closed curve,
τ (x) =
Hx if x is the isotopy class of a non-separating simple closed arc.
ns , defining x ∗ y = τ (y)(x) turns D ns into a quandle (see [11] for details), called the
For x, y ∈ Dg,p g,p
Dehn quandle of the surface Sg,p . Note that the set of all Dehn twists along non-separating simple
closed curves forms one conjugacy class in Mg,p , whereas the set of all half twists along simple
closed arcs forms another conjugacy class. For each g, p ≥ 0, the group Mg,p is generated by finitely
many Dehn twists about non-separating simple closed curves and half twists about simple closed
ns → D(S Mg,p )
arcs [13, Corollary 4.15]. Thus, if S is such a generating set for Mg,p , then τ : Dg,p
becomes an isomorphism of quandles since
τ (x ∗ y) = τ (τ (y)(x)) = τ (y)τ (x)τ (y)−1 = τ (x) ∗ τ (y).
Hence, Dg,pns has the structure of the Dehn quandle of the group M
g,p with respect to S.
The construction of Dehn quandle of Sg,p for p = 0 first appeared in the work of Zablow [38, 39].
Further, [24, 37] considered the quandle structure on the set of isotopy classes of simple closed arcs
in Sg,p for p ≥ 2, and called it quandle of cords. In general, the quandle of cords is a subquandle
ns . In the case of a disk with n marked points, this quandle can be identified with the Dehn
of Dg,p
quandle of the braid group Bn with respect to its standard set of generators, that is, half twists
along the cords. The reader may refer to [11, Section 5] for generalities on Dehn quandles of
surfaces, and [13] for basic facts about mapping class groups.
Example 3.10. Recall from [13, Section 5.1.3] that a presentation of the mapping class group M1
of the closed orientable surface of genus one (namely, the torus) is given by
M1 = hTa , Tb | Ta Tb Ta = Tb Ta Tb , (Ta Tb )6 = 1i,
where the curves a, b are as shown in Figure 2(A).
b b1 b2
c1
a a1 a2
(a) Genus one (b) Genus two
Figure 2. Humphries curves on a surface
Let us set S = {Ta , Tb } and Ta < Tb . Note that Ta and Tb are conjugates of each other and we
can take f (Tb , Ta ) = Ta Tb . Note that Ψ : M1 → SL(2, Z) given by

1 1 1 0
Ψ(Ta ) = and Ψ(Tb ) =
0 1 −1 1
is an isomorphism of groups. An elementary check gives CSL(2,Z) (Ψ(Ta )) = hΨ(Ta ), −Idi and
−Id = (Ψ(Ta )Ψ(Tb ))3 . Thus, we have CM1 (Ta ) = hTa , Tb Ta Ta Tb i. By Theorem 3.1, a presentation
of D1 has S as its generating set and defining relations as follows:
(3.2.23) Ta ∗ Ta ∗ Tb ∗ Ta ∗−1 Tb ∗−1 Ta ∗−1 Tb = Ta ,

(3.2.24) Tb ∗ Ta ∗ Tb ∗ Ta ∗−1 Tb ∗−1 Ta ∗−1 Tb = Tb ,
(3.2.25) Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta = Ta ,
(3.2.26) Tb ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta ∗ Tb ∗ Ta = Tb ,
(3.2.27) Ta ∗ Tb ∗ Ta ∗ Ta ∗ Tb = Ta ,
(3.2.28) Ta ∗ Tb ∗ Ta = Tb .
Using the relation (3.2.28), we can write the relation (3.2.27) as Tb ∗ Ta ∗ Tb = Ta . Now, using this
new relation together with (3.2.28), one can recover relations (3.2.23) through (3.2.26). Thus, we
obtain the presentation
(3.2.29) D1ns ∼
= hTa , Tb | Ta ∗ Tb ∗ Ta = Tb , Tb ∗ Ta ∗ Tb = Ta i.
Remark 3.11. Notice that, the right hand side of (3.2.29) is the presentation of the knot quandle
of the trefoil. This recovers the main result of [29, Theorem 3.1].
Example 3.12. Next, we give a presentation of the Dehn quandle D2ns of the closed orientable
surface S2 of genus two. To simplify the notation, we set α1 := Ta1 , α2 := Tb1 , α3 := Tc1 , α4 := Tb2
and α5 := Ta2 , where the curves are as in Figure 2(B). Then, by [26, Section 4], a presentation of
the mapping class group M2 of S2 is
D
M2 = α1 , α2 , α3 , α4 , α5 | [αi , αj ] = 1 for |i − j|≥ 2, αi αi+1 αi = αi+1 αi αi+1 for 1 ≤ i ≤ 4,
(3.2.30) (α1 α2 α3 α4 α5 )6 = 1, (α1 α2 α3 α4 α5 α5 α4 α3 α2 α1 )2 = 1,
E
[α1 , (α1 α2 α3 α4 α5 α5 α4 α3 α2 α1 )] = 1 .
Let S = {α1 , α2 , α3 , α4 , α5 } and consider the total order α1 < α2 < α3 < α4 < α5 on S. Note
that all Dehn twists along non-separating simple closed curves are conjugate in M2 . We choose f
such that f (α2 , α1 ) = α1 α2 , f (α3 , α2 ) = α2 α3 , f (α4 , α3 ) = α3 α4 , and f (α5 , α4 ) = α4 α5 . Using [18,
Theorem 1] we get a generating set for the mapping class group of a genus one surface with two
boundary components. Then embedding this surface into a closed surface of genus two and using
[33, Remark 1.19] gives
CM2 (α1 ) = hι, α1 , α3 , α4 , α5 i,
where ι is the hyperelliptic involution. Note that ι = α1 α2 α3 α4 α5 α5 α4 α3 α2 α1 = α5 α4 α3 α2 α1 α1 α2 α3 α4 α5

(see [26, Section 4]). Thus, by Theorem 3.1, D2ns is generated by S and has defining relations as
follows. Relations of type (1) are

(3.2.31) αi ∗ αj ∗ αj+1 ∗ αj ∗−1 αj+1 ∗−1 αj ∗−1 αj+1 = αi for 1 ≤ i ≤ 5
and 1 ≤ j ≤ 4,
(3.2.32) αi ∗ α1 ∗ αj ∗−1 α1 ∗−1 αj = αi for 1 ≤ i ≤ 5
and j = 3, 4, 5,
(3.2.33) αi ∗ α2 ∗ αj ∗−1 α2 ∗−1 αj = αi for 1 ≤ i ≤ 5,
and j = 4, 5,
(3.2.34) αi ∗ α3 ∗ α5 ∗−1 α3 ∗−1 α5 = αi for 1 ≤ i ≤ 5,
(3.2.35) αi ∗ α5 ∗ α4 ∗ α3 ∗ α2 ∗ α1 ∗ · · · ∗ α5 ∗ α4 ∗ α3 ∗ α2 ∗ α1 = αi for 1 ≤ i ≤ 5,
| {z }
6 times
(3.2.36) αi ∗ α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 ∗ α5 ∗ α4 ∗ α3 ∗ α2 = αi for 1 ≤ i ≤ 5,
∗α1 ∗ α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 ∗ α5 ∗ α4 ∗ α3 ∗ α2 ∗ α1
(3.2.37) αi ∗ α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 ∗ α5 ∗ α4 ∗ α3 ∗ α2 ∗ α1 ∗ α1 = αi for 1 ≤ i ≤ 5.
∗−1 α1 ∗−1 α2 ∗−1 α3 ∗−1 α4 ∗−1 α5 ∗−1 α5 ∗−1 α4 ∗−1 α3 ∗−1 α2 ∗−1 α1 ∗−1 α1
Relations of type (2) are
(3.2.38) α1 ∗ α5 ∗ α4 ∗ α3 ∗ α2 ∗ α1 ∗ α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 = α1 ,
(3.2.39) α1 ∗ αi = α1 for i = 3, 4, 5,
and relations of type (3) are
(3.2.40) αi ∗ αi+1 ∗ αi = αi+1 for 1 ≤ i ≤ 4.
Note that relations (3.2.31) can be recovered using relations (3.2.40). Along similar lines, we
can recover relations (3.2.32) using relations (3.2.39). We can rewrite the relation (3.2.38) as
α2 ∗ α1 ∗ α2 = α1 using relations (3.2.39). Similarly, we can rewrite relations (3.2.33) and (3.2.34)
as α2 ∗ α4 = α2 , α2 ∗ α5 = α2 and α3 ∗ α5 = α3 , respectively, using relations (3.2.39) and (3.2.40).
One can see that relations (3.2.35) can be recovered using relations (3.2.38), (3.2.39), (3.2.40) and
relations α2 ∗ α4 = α2 , α2 ∗ α5 = α2 and α3 ∗ α5 = α3 . For i = 2, the relation (3.2.37) can be
rewritten as α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 ∗ α5 ∗ α4 ∗ α3 ∗ α2 = α1 , which further gives relations (3.2.36).
Thus, D2ns has a presentation
D
D2ns = α1 , α2 , α3 , α4 , α5 | α1 ∗ α3 = α1 , α1 ∗ α4 = α1 , α1 ∗ α5 = α1 ,
(3.2.41) α2 ∗ α4 = α2 , α2 ∗ α5 = α2 , α3 ∗ α5 = α3 , α1 ∗ α2 ∗ α1 = α2 ,
α2 ∗ α1 ∗ α2 = α1 , α2 ∗ α3 ∗ α2 = α3 , α3 ∗ α4 ∗ α3 = α4 , α4 ∗ α5 ∗ α4 = α5 ,
E
α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 ∗ α5 ∗ α4 ∗ α3 ∗ α2 = α1 .
3.3. An alternate approach to presentation of D2ns . We conclude by giving an alternate proof

of the presentation of D2ns by identifying it with D0,6 ns . It follows from [13, Section 5.1.3] that a
presentation of the mapping class group of S0,6 is

D
M0,6 = σ1 , σ2 , σ3 , σ4 , σ5 | σi σj = σj σi for |i − j|≥ 2, σi σi+1 σi = σi+1 σi σi+1 for all i,
E
(σ1 σ2 σ3 σ4 σ5 )6 = 1, σ1 σ2 σ3 σ4 σ5 σ5 σ4 σ3 σ2 σ1 = 1 .
Recall that, the enveloping group Env(X) of a quandle (X, ∗) is defined as

Env(X) = hex , x ∈ X | ex∗y = ey ex e−1
y , x, y ∈ Xi.
The natural map η : X → Env(X) given by η(x) = ex is a homomorphism of quandles when we

view Env(X) with the conjugation quandle structure. A presentation of the Dehn quandle D0,6 ns of
S0,6 is given in [24, Theorem 3.2] as

D
ns
D0,6 = σ1 , σ2 , σ3 , σ4 , σ5 | σi ∗ σj = σi for |i − j|≥ 2, σi ∗ σj ∗ σi = σj for |i − j|= 1,
E
(3.3.1) σ1 ∗ σ2 ∗ σ3 ∗ σ4 ∗ σ5 = σ5 ∗ σ4 ∗ σ3 ∗ σ2 ∗ σ1 .
Further, a presentation of the enveloping group of D0,6 ns follows from [24, Theorem 3.7] (also from
[35, Theorem 5.1.7]) and given as

D
ns
(3.3.2) Env(D0,6 ) = eσ1 , eσ2 , eσ3 , eσ4 , eσ5 | eσi eσj = eσj eσi for |i − j|≥ 2,
eσi eσi+1 eσi = eσi+1 eσi eσi+1 for all i,
E
[eσj , eσ1 eσ2 eσ3 eσ4 eσ5 eσ5 eσ4 eσ3 eσ2 eσ1 ] = 1 for all j .
Note that D0,6 ns is referred as the quandle of cords of S

0,6 in [24]. It is easy to see that commutativity
of eσ1 eσ2 eσ3 eσ4 eσ5 eσ5 eσ4 eσ3 eσ2 eσ1 with eσj for 2 ≤ j ≤ 5 follows from braid relations, and hence
the last relation in (3.3.2) is needed only for j = 1. The following theorem together with (3.3.1)
gives a presentation of D2ns .
Proposition 3.13. D2ns ∼ ns .
= D0,6
ns → Env(D ns ), given by η(x) = e , induces a quandle homomorphism
Proof. The map η : D0,6 0,6 x
ns )
Env(D0,6
′ ns
η : D0,6 → D(η(S) ns as in (3.3.1). By [11, Section
), where S is the generating set of D0,6
3.2], the map
ns
Φ : Env(D0,6 ) → M0,6
given by Φ(ex ) = x is a surjective group homomorphism, and hence induces a surjective quandle
ns
homomorphism Φ′ : D(η(S)Env(D0,6 ) ) → D0,6ns . Since η ′ Φ′ and Φ′ η ′ are both identity maps, Φ′ is a
quandle isomorphism. Now, recall the presentation of M2 from (3.2.30). It follows that we can
factorise Φ = Φ2 Φ1 , where Φ1 : Env(D0,6ns ) → M and Φ : M → M
2 2 2 0,6 are surjective group
ns
homomorphisms. These maps induce surjective quandle homomorphisms Φ′1 : D(η(S)Env(D0,6 ) ) →
D2ns and Φ′2 : D2ns → D0,6
ns such that Φ′ = Φ′ Φ′ . Since Φ′ is an isomorphism, it follows that both
2 1
Φ2 , Φ1 are isomorphisms, and hence D2ns ∼
′ ′
= D0,6ns .
Remark 3.14. Note that, the additional braid and far commutativity induced relations in (3.3.1)
can be recovered using relations from (3.2.41). Further, the long relation α1 ∗ α2 ∗ α3 ∗ α4 ∗ α5 ∗ α5 ∗
α4 ∗α3 ∗α2 = α1 in (3.2.41) is equivalent to the long relation σ1 ∗σ2 ∗σ3 ∗σ4 ∗σ5 = σ5 ∗σ4 ∗σ3 ∗σ2 ∗σ1
in (3.3.1).
Acknowledgement. Neeraj K. Dhanwani thanks IISER Mohali for the institute post doctoral
fellowship. Hitesh Raundal acknowledges financial support from the SERB grant SB/ SJF/2019-
20. Mahender Singh is supported by Swarna Jayanti Fellowship grants DST/SJF/MSA-02/2018-19
and SB/SJF/2019-20.
References
[1] T. Akita, The adjoint group of a Coxeter quandle, Kyoto J. Math. 60 (2020), 1245–1260.
[2] N. Andruskiewitsch and M. Graña, From racks to pointed Hopf algebras, Adv. Math. 178 (2003), no. 2, 177–243.
[3] V. Bardakov and T. Nasybullov, Embeddings of quandles into groups, J. Algebra Appl. 19 (2020), no. 7, 2050136,
20 pp.
[4] J. Birman, V. Gebhardt and J. González-Meneses, Conjugacy in Garside groups I: cyclings, powers and rigidity,
Groups Geom. Dyn. 1 (2007), 221–279.
[5] E. Brieskorn and K. Saito, Artin-Gruppen und Coxeter-Gruppen, Invent. Math. 17 (1972), 245–271.
[6] R. Chamanara, J. Hu and J. Zablow, Extending the Dehn quandle to shears and foliations on the torus, Fund.
Math. 225 (2014), no. 1, 1–22.
[7] A. H. Clifford and G. B. Preston, The algebraic theory of semigroups, Math. Surveys No. 7 (Part I), Amer.
Math. Soc., Providence RI, 1961.
[8] J. Crisp and L. Paris, Representations of the braid group by automorphisms of groups, invariants of links, and
Garside groups, Pacific J. Math. 221 (2005), 1–27.
[9] P. Dehornoy, Groupes de Garside, Ann. Sci. Ścole Norm. Sup. (4) 35 (2) (2002), 267–306.
[10] P. Dehornoy and L. Paris, Gaussian groups and Garside groups, two generalisations of Artin groups, Proc.
London Math. Soc. 79 (3) (1999), 569–604.
[11] N. K. Dhanwani, H. Raundal and M. Singh, Dehn quandle of groups and orientable surfaces,
arXiv:2106.00290v2.
[12] M. Eisermann, Yang-Baxter deformations of quandles and racks, Algebr. Geom. Topol. 5 (2005), 537–562.
[13] B. Farb and D. Margalit, A primer on mapping class groups, Princeton Mathematical Series, 49. Princeton
University Press, Princeton, NJ, 2012. xiv+472 pp.
[14] N. Franco and J. González-Meneses, Conjugacy problem for braid groups and Garside groups, J. Algebra 266
(2003), 112–132.
[15] N. Franco and J. González-Meneses, Computation of centralizers in braid groups and Garside groups, Proceed-
ings of the International Conference on Algebraic Geometry and Singularities (Spanish) (Sevilla, 2001). Rev.
Mat. Iberoamericana 19 (2003), no. 2, 367–384.
[16] F. A. Garside, The braid group and other groups, Quart. J. Math. Oxford 20 (1969), 235–254.
[17] V. Gebhardt, A new approach to the conjugacy problem in Garside groups, J. Algebra 292 (2005), 282–302.
[18] S. Gervais, A finite presentation of the mapping class group of a punctured surface. Topology 40 (2001), no. 4,
703–725.
[19] G. G. Gurzo, Systems of generators for the normalizers of certain elements of the braid group, Math. USSR
Izv. 24 (1985), 439–478.
[20] A. Inoue and Y. Kabaya, Quandle homology and complex volume, Geom. Dedicata 171 (2014), 265–292.
[21] D. Joyce, An algebraic approach to symmetry with applications to knot theory, Ph.D. Thesis, University of
Pennsylvania, 1979, 127 pp.
[22] D. Joyce, A classifying invariant of knots, the knot quandle, J. Pure Appl. Algebra 23 (1982), 37–65.
[23] S. Kamada, An observation of surface braids via chart description, J. Knot Theory Ramifications 5 (1996), no.
4, 517–529.
[24] S. Kamada and Y. Matsumoto, Certain racks associated with the braid groups, Knots in Hellas ’98 (Delphi),
118–130, Ser. Knots Everything, 24, World Sci. Publ., River Edge, NJ, 2000.
[25] O. Loos, Reflexion spaces and homogeneous symmetric spaces, Bull. Amer. Math. Soc. 73 (1967), 250–253.
[26] D. Margalit and R. Winarski, Braids groups and mapping class groups: The Birman–Hilden theory, Bull. Lond.
Math. Soc. 53 (2021), no. 3, 643–659.
[27] S. V. Matveev, Distributive groupoids in knot theory, in Russian: Mat. Sb. (N.S.) 119 (1) (1982) 78–88, trans-
lated in Math. USSR Sb. 47 (1) (1984), 73–83.
[28] J. González-Meneses and B. Wiest, On the structure of the centraliser of a braid, Ann. Sci. École Norm. Sup.
(4) 37 (2004), no. 5, 729–757.
[29] M. Niebrzydowski and J. H. Przytycki, The quandle of the trefoil knot as the Dehn quandle of the torus, Osaka
J. Math. 46 (2009), 645–659.
[30] M. Picantin, Petits groupes gaussiens, PhD Thesis, Université de Caen, 2000, 131 pp.
[31] M. Picantin, The center of thin Gaussian groups, J. Algebra 245 (1) (2001), 92–122.
[32] M. Picantin, The conjugacy problem in small Gaussian groups, Comm. Algebra 29 (3) (2001), 1021–1038.
[33] A. Putman, Lectures on the Torelli group, www3.nd.edu/ andyp/teaching/2014SpringMath541/TorelliBook.pdf.
[34] R. Randell, The fundamental group of the complement of a union of complex hyperplanes, Invent. Math. 69
(1982) 103–108.
[35] S. K. Winker, Quandles, knots invariants and the n-fold branched cover, Ph.D. Thesis, University of Illinois at
Chicago, 1984, 198 pp.
[36] D. N. Yetter, Quandles and Lefschetz fibrations, arXiv:math/0201270.
[37] D. N. Yetter, Quandles and monodromy, J. Knot Theory Ramifications 12 (2003), no. 4, 523–541.
[38] J. Zablow, Loops, waves, and an “algebra” for Heegaard splittings, Ph.D. Thesis, City University of New York,
1999, 64 pp.
[39] J. Zablow, Loops and disks in surfaces and handlebodies, J. Knot Theory Ramifications 12 (2003), no. 2,
203–223.
[40] J. Zablow, On relations and homology of the Dehn quandle, Algebr. Geom. Topol. 8 (2008), no. 1, 19–51.
Department of Mathematical Sciences, Indian Institute of Science Education and Research (IISER)
Mohali, Sector 81, S. A. S. Nagar, P. O. Manauli, Punjab 140306, India.
Email address: neerajk.dhanwani@gmail.com
Email address: hiteshrndl@gmail.com
Email address: mahender@iisermohali.ac.in

Mathematics Subject Classification Key Words and Phrases.: G G Ns G G G Ns G

Uploaded by

Document Information

Original Description:

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

Mathematics Subject Classification Key Words and Phrases.: G G Ns G G G Ns G

Uploaded by

Copyright:

Available Formats

PRESENTATIONS OF DEHN QUANDLES

NEERAJ K. DHANWANI, HITESH RAUNDAL, AND MAHENDER SINGH

2. Presentations of Gaussian quandles and Garside quandles

Understanding of presentations of quandles is fundamental to the development of quandle theory.

2.1. Preliminaries on Gaussian and Garside monoids, and groups. An element a of a

is a Garside monoid. Further, it is of type R9 .

is a Garside monoid of type R9 .

Figure 1. A diagram for pairs of types Ri

di ∆n−i ei = ∆n−i φn−i (di )ei (since di ∆ = ∆φ(di ))

to be the equivalence relation generated by the following transformations:

(Rα ) x−1 s−1 tsx ←→ x−1 α(s, t)x

For x, y ∈ S (S∪S , we denote x ≃fR y if x and y are right f -equivalent.

Lemma 2.13. Let x, y ∈ S (S∪S and z ∈ (S ∪ S −1 )∗ . If x ≃fR y, then z −1 xz ≃fR z −1 yz.

by a single transformation Rα or Rf,β or RG , then by the transformation itself, z −1 xz ≃fR z −1 yz.

have z −1 xz ≃fR z −1 yz in the general case as well.

right f -equivalent if and only if they represent the same element in G.

generated by Tα , Tf,β , T0 and Tf . By [10, Theorem 4.1], M has a complemented presentation

where s ∈ S and x, y ∈ (S ∪ S −1 )∗ with x =G y. Note that if w ∈ (S ∪ S −1 )∗ , then w−1 ∈ (S ∪ S −1 )∗

x−1 zx ≃fT x−1 xyx−1 x (since z ≃fT xyx−1 )

– type L9 if it satisfies conditions (vii) and (ix).

Example 2.29. Let A be an Artin group with Artin presentation

(2.4.1) hS | (st)mst = (ts)mts for s, t ∈ S with s 6= ti ,

f (s, t)fmst (s, t) = (ts)mst −1 fmst (s, t)

(2.4.2) fmst (s, t) ∗ fmst −1 (s, t) ∗ · · · ∗ f1 (s, t) = f0 (s, t) and

where (u ∗ v)k denotes the left associated product u ∗ v ∗ u ∗ v ∗ u · · · of length k.

3. Presentations of Dehn quandles of groups

This proves the claim.

of elements in G can be rewritten as

3.2. Examples of presentations of Dehn quandles. As applications of Theorem 3.1, we give

Setting S = {a1 , b1 , a2 , b2 , . . . , ag , bg }, we note that no two elements of S are conjugate to each

τ (x ∗ y) = τ (τ (y)(x)) = τ (y)τ (x)τ (y)−1 = τ (x) ∗ τ (y).

M1 = hTa , Tb | Ta Tb Ta = Tb Ta Tb , (Ta Tb )6 = 1i,

where the curves a, b are as shown in Figure 2(A).

(a) Genus one (b) Genus two

Figure 2. Humphries curves on a surface

of D1 has S as its generating set and defining relations as follows:

(3.2.23) Ta ∗ Ta ∗ Tb ∗ Ta ∗−1 Tb ∗−1 Ta ∗−1 Tb = Ta ,

CM2 (α1 ) = hι, α1 , α3 , α4 , α5 i,

where ι is the hyperelliptic involution. Note that ι = α1 α2 α3 α4 α5 α5 α4 α3 α2 α1 = α5 α4 α3 α2 α1 α1 α2 α3 α4 α5

follows. Relations of type (1) are

3.3. An alternate approach to presentation of D2ns . We conclude by giving an alternate proof

presentation of the mapping class group of S0,6 is

Recall that, the enveloping group Env(X) of a quandle (X, ∗) is defined as

The natural map η : X → Env(X) given by η(x) = ex is a homomorphism of quandles when we

S0,6 is given in [24, Theorem 3.2] as

[35, Theorem 5.1.7]) and given as

Note that D0,6 ns is referred as the quandle of cords of S

You might also like