Masters PDF
Masters PDF
Masters PDF
by
LIAM THOMAS WATSON
B.Sc. The University of British Columbia
A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF
THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
in
THE FACULTY OF GRADUATE STUDIES
Department of Mathematics
We accept this thesis as conforming
to the required standard
.....................................
.....................................
(Signature)
Department of Mathematics
The University of British Columbia
Vancouver, Canada
Date
Abstract
Recent work of Eliahou, Kauffmann and Thistlethwaite suggests the use of
braid actions to alter a link diagram without changing the Jones polynomial.
This technique produces non-trivial links (of two or more components) having
the same Jones polynomial as the unlink. In this paper, examples of distinct
knots that can not be distinguished by the Jones polynomial are constructed
by way of braid actions. Moreover, it is shown in general that pairs of knots
obtained in this way are not Conway mutants, hence this technique provides
new perspective on the Jones polynomial, with a view to an important (and
unanswered) question: Does the Jones polynomial detect the unknot?
ii
Table of Contents
Abstract ii
List of Figures v
Acknowledgement vii
Chapter 1. Introduction 1
Chapter 3. Polynomials 8
3.1 The Jones Polynomial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 The Alexander Polynomial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 The HOMFLY Polynomial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
iii
Table of Contents iv
Bibliography 72
iv
List of Figures
2.1 Diagrams of the Trefoil Knot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 The Hopf Link . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.3 The braid generator σi and its inverse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.4 The link β̄ formed from the closure of β. . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
v
List of Figures vi
vi
Acknowledgement
It’s a strange process, writing down the details to something you’ve been work-
ing on for some time. And while it is a particularly solitary experience, the
final product is not accomplished on you’re own. To this end, there are many
people that should be recognised as part of this thesis.
First, I would like to thank my supervisor Dale Rolfsen for ongoing patience,
guidance and instruction. It has been a privilege and a pleasure to learn from
Dale throughout my undergraduate and masters degrees, and I am grateful for
his generous support, both intellectual and financial. I have learned so much.
Of course, I am indebted to all of my teachers as this work comprises much of
what I have learned so far. However, I would like to single out Bill Casselman,
as the figures required for this thesis could not have been produced without
his guidance.
To my parents, Peter and Katherine Watson, family and friends, I am so
fortunate to have a support network that allows me to pursue mathematics.
In particular, I would like to thank Erin Despard for continued encouragement,
support and perspective, everyday.
vii
Chapter 1
Introduction
1
Chapter 1. Introduction 2
S1 ,→ R3 ⊂ S3 .
In such a diagram the indicates that the one section of the knot (the
broken line) has passed behind another (the solid line) to form a crossing.
In general there will not be any distinction made between the knot K and a
diagram representing it. That is, we allow a given diagram to represent a knot
3
Chapter 2. Knots, Links and Braids 4
and denote the diagram by K also. It should be pointed out, however, that
there are many diagrams for any given knot. Indeed, K and K 0 are equivalent
knots (denoted K ∼ K 0 ) if they are related by isotopy in S3 . Therefore, the
diagrams for K and K 0 may be very different.
An n-component link is a collection of knots. That is, a link is a disjoint union
of embedded circles
an
S1i ,→ R3 ⊂ S3
i=1
where each
S1i ,→ R3 ⊂ S3
is a knot. Of course, a 1-component link is simply a knot, and a non-trivial
link can have individual components that are unknotted.
∼ ∼ (R1 )
∼ ∼ (R2 )
∼ (R3 )
Theorem (Reidemeister). Two link diagrams represent the same link iff the
diagrams are related by planar isotopy, and the Reidemeister moves.
Chapter 2. Knots, Links and Braids 5
While writhe is not a link invariant, it does give rise to the following definition.
The linking number is a link invariant. Note that, for the Hopf link of figure
2.2, there are two distinct orientations. One orientation has linking number 1,
the other linking number −1 and hence there are two distinct oriented Hopf
links.
2.2 Braids
There are many equivalent definitions of braids (see [6], [10], [27]). In this
setting it is natural to start from a geometric point of view.
Let E ⊂ R3 denote the yz-plane and let E 0 denote its image shifted by 1 in
the x direction. Consider the the collection of points
and denote by
P 0 = {(1, 0, 1), . . . , (1, 0, n)}
the image of P in E 0 .
Chapter 2. Knots, Links and Braids 6
αi : [0, 1] ,→ [0, 1] × E ⊂ R3
such that
(a) αi (0) = i ∈ P
(b) αi (1) ∈ P 0
where the generators correspond to a crossing formed between the i and i+1
strand as in figure 2.3.
Just as group elements are formed by words in the generators, a braid diagram
for a given element can be constructed by concatenation of braids of the form
shown in figure 2.3.
If E and E 0 of an n-strand braid are identified so that P = P 0 pointwise, the
result is a collection of embeddings of S 1 in R3 and we obtain a link.
Given any braid β we can form a link β̄ by taking the closure in this way. It is
a theorem of Alexander [1] that every link arises as the closure of some braid.
Given a link diagram L, it is always possible to construct a braid β such that
β̄ = L. Two such constructions (there are many) are due to Morton [26] and
Vogel [34].
Chapter 2. Knots, Links and Braids 7
σi σi−1 = 1 = σi−1 σi
σi σj σi = σ j σi σj
Theorem (Markov). Two links β¯1 and β¯2 are equivalent iff β1 ∼M β2 .
Chapter 3
Polynomials
~ ∈ Z. Now
so that w(L)
~
−a−3w(L) hLi ∈ Z[a, a−1 ]
is invariant under R1 and gives rise to an invariant of oriented links.
8
Chapter 3. Polynomials 9
Note that it will often be convenient to work with t = a −4 , and the polynomial
obtained through this substitution will be referred to as the Jones Polynomial
also.
As we shall see, there are many examples of distinct links having the same
Jones Polynomial. However, the following is still unknown:
where y + = i+
? (y).
1 // π X
1
e // π1 X // H1 (X, Z) // 0
α ~
// lk α, K
so that the infinite cyclic group H1 (X, Z) = hti gives the covering translations
of Xe & X. Now H1 (X̃, Z), although typically not finitely generated as an
abelian group, is finitely generated as a Z[t, t −1 ]-module by the {ai }. The
variable t corresponds to the hti-action taking X i to Xi+1 .
Chapter 3. Polynomials 11
Definition 3.4. H1 (X̃, Z) is called the Alexander module and has module
presentation V − tV > , where V is the Seifert matrix. This is a knot invariant.
Question 3.6. Is there a similar topological interpretation for the Jones poly-
nomial?
On the other hand, it is easy to generate knots K such that ∆ K (t) = 1 (see
for example, [32]).
.
Theorem 3.7. For any knot K, ∆K (t) = ∆K (t−1 ).
proof. Setting t = 1 and using the standard (symplectic) basis for H 1 (F, Z)
Chapter 3. Polynomials 12
gives
V − V > = lk(ai , a+
j ) − lk(a j , a +
i )
= lk(ai , a+
j ) − lk(a −
j , a i )
= lk(ai , a+
j ) − lk(a i , a −
j )
0 1
−1 0
0 1
−1 0
=
..
.
0 1
−1 0
therefore
∆K (1) = ± det(V − V > ) = ±1.
where cm−r = ±cr , and the same choice of sign is made for each r. Now m
must be even, since m odd gives rise to ∆ K (1) even, contradicting theorem
3.8. Further, if cm−r = −cr then cm/2 = 0 and
m
X
∆K (1) = bk = 0,
k=0
We will see the form given in corollary 3.9 in the next section. Together with
the normalization ∆K (1) = 1, it is sometimes referred to as the Alexander-
Conway polynomial as it has a recursive definition, originally noticed by Alexan-
der [2] and later exploited by Conway [8].
It should be noted that there are generalizations of this construction to invari-
ants of oriented links that have been omitted. Nevertheless, we shall see that
the recursive definition of ∆K (t) is defined for all oriented links.
Moreover, this basis allows us to present any element of H n+1 in the form
x1 + x 2 σn x3
xσn = σn x
tr : Hn −→ Z[q ±1 , z]
σi 7−→ z
where m0i , m00i ∈ Hn . In this case we will make use of the following:
Chapter 3. Polynomials 15
(2)
tr(µ1 σn µ2 σn µ3 )
= tr(µ1 µ2 σn µ3 σn ) applying (1)
= tr(µ1 µ2 σn µ03 σn−1 µ003 σn )
2
= z tr(µ1 µ2 µ03 σn−1 µ003 ) as above
= z(q − 1) tr(µ1 µ2 µ03 σn−1 µ003 ) + zq tr(µ1 µ2 µ03 µ003 ) using (iii)
= z(q − 1) tr(µ03 σn−1 µ003 µ1 µ2 ) + zq tr(µ03 µ003 µ1 µ2 ) induction
2
= z tr(µ03 σn−1 µ003 µ1 µ2 ) using (iii)
= tr(σn µ03 σn−1 µ003 σn µ1 µ2 ) as above
= tr(σn µ3 σn µ1 µ2 )
= tr(µ3 µ1 σn µ2 σn ) applying (1)
Chapter 3. Polynomials 16
The aim is to use the trace function to define a link invariant. In particular
we would like to make use of this trace on braids in the composite
Bn −→ Hn −→ Z[q ±1 , z].
w
To do this, we introduce a change of variables λ = qz where
1−q λq(1 − q)
z=− and w=−
1 − λq 1 − λq
so that
1−q+z
λ= .
zq
Definition 3.12. The HOMFLY polynomial is given by
n−1
1 − λq √ e
Xβ̄ (q, λ) = − √ λ tr(β)
λ(1 − q)
Note that the closure of the identity braid in B n gives the n component unlink
···
| {z }
n
Chapter 3. Polynomials 17
σi2 = (q − 1)σi + q
σi = (q − 1) + qσi−1
qσi−1 = σi + 1 − q
σi−1 = q −1 σi + q −1 − 1
Chapter 3. Polynomials 18
hence
tr(σi−1 ) = tr(q −1 σi + q −1 − 1)
= q −1 tr(σi ) + q −1 − 1
−1 1−q
=q − + q −1 − 1
1 − λq
−1 1−q
=q − +1 −1
1 − λq
−1 −1 + q + 1 − λq
=q −1
1 − λq
1 − λ − 1 + λq
=
1 − λq
1−q
= −λ .
1 − λq
Thus
n
1 − λq √ e−1
X −1 (q, λ) = −√ λ tr(βσ1−1 )
βσn λ(1 − q)
1 1 − λq 1−q
=√ −√ −λ Xβ̄ (q, λ)
λ λ(1 − q) 1 − λq
= Xβ̄ (q, λ)
Both the single variable polynomials may be retrieved from the HOMFLY
polynomial via the substitutions
VL (t) = XL (t, t)
.
∆L (t) = XL t, t−1 .
β = γ 1 σi γ2
1
a similar construction is possible for σi−1
Chapter 3. Polynomials 19
L ∼ L0 = γσi
where γ = γ1 γ2 . Let
L+ = γσi2 and L− = γ̄.
The relation (iii) gives
γσi2 = (q − 1)γσi + qγ
so that
tr(γσi2 ) − q tr(γ) = (q − 1) tr(γσi ).
Let e = e(γ) be the exponent sum of γ. Then
1 p p
p p Xγσ2 − q λXγ̄
q λ i
n−1
1 1 − λq √ e+2
=p p −√ λ tr(γσi2 )
q λ λ(1 − q)
n−1
p p 1 − λq √ 2
− q λ −√ λ tr(γ)
λ(1 − q)
√ e+1
λ n−1
1 − λq
= √ −√ tr(γσi2 ) − q tr(γ)
q λ(1 − q)
√ e+1
λ n−1
1 − λq
= √ −√ ((q − 1) tr(γσi ))
q λ(1 − q)
n−1
√ 1 1 − λq √ e+1
= q−√ − √ λ tr(γσ1 )
q λ(1 − q)
√ 1
= q − √ Xγσ1
q
we can define
PL (t, x) = XL (q, λ)
where PL (t, x) ∈ Z[t±1 , x±1 ] is computed recursively from the axioms
P (t, x) = 1 (3.5)
In this setting, L+ , L− and L0 are diagrams that are identical except for in a
small region where they differ as in
L+ L− L0
in the skein definition of the HOMFLY polynomial. This agrees with (3.4)
under the substitutions
p p √ 1
t= q λ and x = q− √ .
q
As indicated earlier, both the Jones polynomial and the Alexander polynomial
may be computed recursively as they each satisfy a skein relation by specifying
√ 1
VL (t) = PL t, t − √
t
. √ 1
∆L (t) = PL 1, t − √ .
t
Chapter 4
Tangles and Linear Maps
21
Chapter 4. Tangles and Linear Maps 22
0= and ∞ = .
These tangles are fundamental in the sense that they form a basis for presenting
the bracket of a given tangle T . That is
T = x0 + x∞
In this way, the Kauffman bracket divides Conway tangles into equivalence
classes completely determined by br(T ). For example,
−1
=a +a
and
br = a a−1 .
Definition 4.3. The Kauffman bracket skein module S br is the Z[a, a−1 ]-
module generated by isotopy classes of Conway tangles, modulo equivalence
given by axioms (3.2) and (3.3) defining the Kauffman bracket.
The tangles {0, ∞} provide a module basis for S br so that the elements T ∈ S br
may be represented by br(T ).
Suppose the tangle T is contained in some link L. Then writing L = L(T ) and
considering T ∈ S br gives rise to a Z[a, a−1 ]-linear map
f : S br −→ Z[a, a−1 ] (4.1)
hL(0)i
T 7−→ br(T )
hL(∞)i
where
L(0) = L and L(∞) = L .
This map is simply an evaluation map computing the bracket of L(T ) since
hL(0)i
hL(T )i = br(T ) = f (T ).
hL(∞)i
TN =
Now returning to the link L(T ), recall that the external wiring L r T is itself
a tangle. Again, all crossings and closed loops may be eliminated using the
bracket axioms so that
hL r T i = br(L r T )
= hT N i hT D i br(L r T )>
Chapter 4. Tangles and Linear Maps 24
In fact, combining the linear maps (4.1) and (4.2) forms a bilinear map
where
hL(T )i = hT N i hT D i br(L r T )>
N
N
x0 + x∞
= br(L r T )>
D D
x0 + x∞
N
D
= br(T )
N
D
br(L r T )>
δ 1
= br(T ) br(L r T )>
1 δ
Definition 4.5. For Conway tangles T and U define the link J(T, U ) = (T U ) N
and call this the join of T and U .
We have that
hJ(T, U )i = F (T, U )
by definition, and further
L(T ) ∼ J(T, L r T ).
Definition 4.6. For Conway tangles T and U define the connected sum
T D #U D = (T U )D .
Chapter 4. Tangles and Linear Maps 25
Since any link L may be written as T D for some tangle T this definition gives
rise to a connected sum for links. It follows that
hL1 #L2 i = hL1 ihL2 i,
and
VL1 #L2 = VL1 VL2
provided orientations agree. A similar argument gives such an equality for the
HOMFLY polynomial, and hence the Alexander polynomial as well.
Notice that
ρ =
ρ =
ρ =
so that
hL(0)i
hL(T )i = br(T )
hL(∞)i
hL(0)i
= br(ρT )
hL(∞)i
= hL(ρT )i.
Chapter 4. Tangles and Linear Maps 26
While it may be that L(T ) L(ρT ), it is certain that this method does not
provide an answer to question 3.2: It can be shown that a Conway mutant of
the unknot is always unknotted [30]. Theorem 4.8 is in fact a corollary of a
stronger statement.
Theorem 4.9. PL(T ) = PL(ρT ) .
proof. Using the skein relation (3.6) defining the HOMFLY polynomial, it is
possible to decompose any tangle T into a linear combination of the form
T = a1 + a2 + a3
where ai ∈ Z[t±1 , x±1 ]. Therefore, these tangles provide a basis for presenting
the HOMFLY polynomial of a tangle T . Thus, we can define a Z[t ±1 , x±1 ]-
module S P generated by isotopy classes of tangles up to equivalence under
the skein relation. Moreover, if L = L(T ) then we have a Z[t ±1 , x±1 ]-linear
evaluation map
S P −→ Z[t±1 , x±1 ]
T 7−→ PL(T )
or, more generally, the bi-linear evaluation map
S P × S P −→ Z[t±1 , x±1 ]
(T, U ) 7−→ PJ(T,U ) .
Since the basis n o
, ,
is ρ-invariant, it follows that PT and PρT are equal hence
S P × S PR
RRR
RRR
R((
ρ×id Z[t±1 , x±1 ]
ll66
lll
lll
SP × SP
Chapter 4. Tangles and Linear Maps 27
commutes and
PL(T ) = PL(ρT ) .
where δ = −a−2 − a2 and the h i indicate that the rest of the tangle is left
unchanged.
Definition 4.11. The Z[a, a−1 ]-module
Sn = Mn In
and .
Chapter 4. Tangles and Linear Maps 28
proof. Simply put, we need to determine how many n-tangles there are that
have no crossings or closed loops. That is, given a disk in the plane with 2n
marked points on the boundary, how many ways can the points be connected
by non-intersecting arcs (up to isotopy)?
Clearly, C1 = 1, and as discussed earlier C2 = 2.
Now suppose n > 2 and consider a disk with 2n points on the boundary.
Starting at some chosen boundary point and numbering clockwise, the point
labeled 1 must connect to an even labeled point, say 2k. This arc divides the
disk in two: One disk having 2(k − 1) marked points, the other with 2(n − k).
Therefore
n
X
Cn = Ck−1 Cn−k
k=1
= C0 Cn−1 + C1 Cn−2 + · · · + Cn−1 C0
where C0 = 1 by convention. Now consider the generating function
∞
X
f (x) = Ci xi
i=0
and notice !
∞
X i
X
2
(f (x)) = Ck−1 Cn−k xi
i=0 k=1
so that
x(f (x))2 = f (x) − 1
and √
1 − 4x
1−
f (x) = .
2x
√
To deduce the coefficients of f (x), first consider the expansion of z about 1.
∞
X
√
z= di (z − 1)i
i=0
Chapter 4. Tangles and Linear Maps 29
where (
1 i=0
di = (−1)i−1 (2i−3)!! .
2i i!
i>0
Therefore with z = 1 − 4x
∞
X
√
1 − 4x = di (−4x)i
i=0
∞
X
=1+ (−1)i 2i 2i di xi
i=1
and
∞
1 X
f (x) = − (−1)i 2i 2i di xi
2x
i=1
∞
X
= (−1)i−1 2i 2i−1 di xi−1
i=1
so that
Ci = (−1)i 2i+1 2i di+1
(−1)i (2i − 1)!!
= (−1)i − 2i+1 2i
2i+1 (i + 1)!
2i (2i − 1)!!
=
(i + 1)!
for i > 1. Finally, the fact that
2n (2n − 1)!! (2n)!
=
(n + 1)! n!(n + 1)!
follows by induction since
2n+1 (2(n + 1) − 1)!! 2(2n + 1) 2n (2n − 1)!!
=
(n + 2)! n+2 (n + 1)!
2(n + 1)(2n + 1)
= Cn
(n + 2)(n + 1)
(2n + 2)(2n + 1) (2n)!
=
(n + 2)(n + 1) n!(n + 1)!
(2(n + 1))!
= .
(n + 1)!(n + 2)!
Chapter 4. Tangles and Linear Maps 30
Let a given n-tangle diagram be contained in the unit disk so that, of the
2n boundary points, n have positive x-coordinate while the remaining n have
negative x-coordinate. With this special position, multiplication of tangles
by concatenation, as introduced for Conway tangles, extends to all n-tangles.
When two n-tangles are in fact n-braids, we are reduced to multiplication in
Bn . With this multiplication, Sn has an algebra structure called the Temperly-
Lieb algebra [21, 22].
The n-dimensional Temperly-Lieb algebra TL n over Z[a, a−1 ] has generators
e1 , e2 , . . . , en−1 and relations
The multiplicative identity for this algebra is exactly the identity in B n , and
the generators are tangles of the form shown in figure 4.2.
For example, TL3 is generated by {e1 , e2 }, while the basis for the module S3
is the set of elements {1, e1 , e2 , e1 e2 , e2 e1 } as in (4.6).
Notice that there is a representation of the braid group via the Kauffman
Chapter 4. Tangles and Linear Maps 31
bracket given by
Bn −→ TLn
σi 7−→ a + a−1 ei
σi−1 7−→ aei + a−1 .
V L = V L0 .
Of course, it may be that L L0 and this approach has been used in attempts
to answer question 3.2 [30, 31].
Let’s first revisit Conway mutation in this context. We have, given the bilinear
evaluation map F and a 180 degree rotation ρ, the commutative diagram
S2 × S2P
PPP F
PPP
P((
ρ×id Z[a, a−1 ]
nn66
nnnn
n F
n
S2 × S 2
G : S2 × S2 −→ Z[a, a−1 ]
(T, U ) 7−→ br(T ) L br(U )> .
τ : S2 × S2 −→ S2 × S2
(T, U ) 7−→ (τ1 T, τ2 U )
S2 × S2P
PPPG
PPP
P((
τ =τ1 ×τ2 Z[a, a−1 ]
nn66
nnn
nnn G
S2 × S 2
B3 = hσ1 , σ2 |σ1 σ2 σ1 = σ2 σ1 σ2 i
where
σ1 = and σ2 = .
Chapter 4. Tangles and Linear Maps 33
S2 × B3 −→ S2
(T, β) 7−→ T β
proof. Let idB3 be the identity braid. Then for any tangle T we have
T idB3 = T
since
∼
Chapter 4. Tangles and Linear Maps 34
and
a + a −1 δ
= br(T )
a +a −1
−3
−a 0
= br(T ) −1
a a
Chapter 4. Tangles and Linear Maps 35
+a −1
a
= br(T )
a −1
+a δ
a a−1
= br(T )
0 −a−3
β : S2 × S2 −→ S2 × S2
−1
(T, U ) 7−→ (T β , U β ).
Chapter 4. Tangles and Linear Maps 36
For a link of the form L(T, U ), this leads to the definition of a new link
−1
L(T β , U β ).
Denote the evaluation matrix of L(T, U ) by
hL(0, 0)i hL(0, ∞)i
L=
hL(∞, 0)i hL(∞, ∞)i
and suppose that L ∈ GL2 (Z[a, a−1 ]), that is, det(L) 6= 0. Note that
S2 × S2P
PPPG
PPP
P((
β Z[a, a−1 ]
nn66
nnn
nnn G
S2 × S 2
Question 4.16. For a given link L(T, U ) with evaluation matrix L ∈ GL 2 (Z[a, a−1 ]),
what are the elements β ∈ B3 such that L ∈ Fix(β)?
5.1 Construction
Shortly after the discovery of the HOMFLY polynomial, Kanenobu introduced
families of distinct knots having the same HOMFLY polynomial and hence the
same Jones and Alexander polynomials as well [18]. It turns out that these
knots are members of a much larger class of knots which we will denote by
K(T, U ) for tangles T, U ∈ S2 .
proof. Since
a a−1 a−1 a
Φ(σ2 ) = and Φ(σ2−1 ) =
0 −a−3 0 −a3
37
Chapter 5. Kanenobu Knots 38
we have
a a−1 x δ a−1 0 ax + a−1 δ aδ + a−1 δ 2 a−1 0
=
0 −a−3 δ δ 2 a −a3 −a−3 δ −a−3 δ 2 a −a3
x + a−2 δ + a2 δ + δ 2 −a4 δ − a2 δ 2
=
−a−4 δ − a−2 δ 2 δ2
x + δ(a−2 + a2 + δ) δ(−a4 − a2 δ)
=
δ(−a−4 − a−2 δ) δ2
x δ(−a4 + 1 + a4 )
=
δ(−a−4 + a−4 + 1) δ2
x δ
=
δ δ2
with δ = −a−2 − a2 .
To compute the bracket of the Kanenobu knot K(T, U ), we need the evaluation
matrix
* + * +
K=
.
* + * +
Since K(0, 0) is the connected sum of two figure eight knots (the figure eight
is denoted 41 as in [32]), we can compute
D E D E
41 = σ2 σ1−1 σ2 σ1−1
and
2
h41 #41 i = a−8 − a−4 + 1 − a4 + a8
= a−16 − 2a−12 + 3a−8 − 4a−4 + 5 − 4a4 + 3a8 − 2a12 + a16 .
In addition, it can be seen from the braid closure σ 2 σ1−1 σ2 σ1−1 ∼ 41 that
w(41 ) = 0 and hence
since the three entries for K(0, ∞), K(∞, 0) and K(∞, ∞) are all equivalent to
unlinks with no crossings via applications of the Reidemeister move R 2 (recall
that R2 leaves the Kauffman bracket unchanged).
For any tangle diagram T , denote by T ? the tangle diagram obtained by switch-
ing each crossing of T . That is, for any choice of orientation
w(T ? ) = −w(T ).
This can be extended to knot diagrams K, where K ? is the diagram such that
w(K ? ) = −w(K)
Since K is of the form given in proposition 5.1, the bilinear map defined by
the braid σ2n ∈ B3
σ2n : S2 × S2 → S2 × S2
n −n
(T, U ) 7→ (T σ2 , U σ2 )
Theorem (Kanenobu). Kn,m and Kn0 ,m0 have the same HOMFLY polyno-
mial when |n−m| = |n0 −m0 |. Moreover, when (n, m) and (n0 , m0 ) are pairwise
distinct, these knots are distinct.
···
| {z }
n
is given by
n−1
t−1 − t
.
x
As this polynomial will be used often, we define
t−1 − t
P0 = .
x
Now, using the skein relation (3.6) defining the HOMFLY polynomial P (t, x),
we can compute
−t = −t−1 +x
= t−2 − t−1 x
and
t−1 =t +x
= t2 + tx .
Chapter 5. Kanenobu Knots 42
and we can conclude that K0 and K1 are distinct knots despite having the
same Jones polynomial:
Further, applying theorem 4.9, these knots cannot be Conway mutants as they
have different HOMFLY polynomials.
proof. Take U = T ? (that is, the tangle such that w(U ) = −w(T )) and define
Kanenobu knots for the pair (T, U )
V K = V K σ2 .
It remains to show that these are in fact distinct knots. To see this, we compute
the HOMFLY polynomials PK and PK σ2 .
Now with the requirement that the tangle U = T ? , there are two choices of
orientations for the tangles that are compatible with an orientation of the knot
(or possibly link, in which case a choice of orientation is made) K(T, U ). They
are
Type 1 ,
Type 2
,
= aT + bT
= aU + bU
Chapter 5. Kanenobu Knots 44
so that
−1
T σ2 = and U σ2 =
therefore
σ2 σ2−1
T ,U = a T aU , + a T bU ,
+ b T aU , + b T bU ,
= a T aU , + a T bU ,
+ b T aU , + b T bU ,
so that
= aU + bU
, = a T aU , + a T bU ,
+ b T aU , + b T bU ,
and
PK = aT aU PK0 + (aT bU + bT aU )P0 + bT bU P02
−1 −1 2
t −t t −t
= aT aU PK0 + (aT bU + bT aU ) + b T bU
x x
= a T a U P K0 + R
with R ∈ Z[t± , x± ] as before. Again, applying the action of σ 2 we have
σ2 σ2−1
T ,U = a T aU , + a T bU ,
+ b T aU , + b T bU ,
= a T aU , + a T bU ,
+ b T aU , + b T bU ,
PK 6= PK σ2
5.4 Examples
1. Consider the Kanenobu knot K(T, U ) where
T = and U =
as in figure 5.3.
so that
−a−3 0 −3
br(T ) = 0 1
a−1 a
3 3 3
−a 0 −a 0 −a 0
= 0 1
a a−1 a a−1 a a−1
−a9 0
= 0 1
a−1 − a3 + a7 a−3
−1
= a − a3 + a7 a−3
Chapter 5. Kanenobu Knots 47
and
−a−3 0 3
br(U ) = 0 1
a−1 a
−a−3 0 −a−3 0 −a−3 0
= 0 1
a−1 a a−1 a a−1 a
−7
= a − a−1 + a a3 .
Now
hence
VK(T,U ) = V n −n
K(T σ2 ,U σ2 )
Chapter 5. Kanenobu Knots 48
n −n
since w(K(T σ2 , U σ2 )) = 0 for all n ∈ Z.
−1
The Knots K(T, U ) and K(T σ2 , U σ2 ) are distinct as they have different HOM-
FLY polynomials2 :
T = and U =
so that
Again we have a family of knots (distinct from those of example 1) such that
VK(T,U ) = V n −n
K(T σ2 ,U σ2 )
n −n
since w(K(T σ2 , U σ2 )) = 0 for all n ∈ Z.
2
computed using KNOTSCAPE
Chapter 5. Kanenobu Knots 49
−1
The knots K(T, U ) and K(T σ2 , U σ2 ) of this example are distinct as can be
seen by the HOMFLY polynomials3 :
Once again, from theorem 4.9 it follows that these knots are not related by
Conway mutation.
3
computed using KNOTSCAPE
Chapter 5. Kanenobu Knots 50
5.5 Generalisation
We saw in proposition 5.1 under the action
B3 × GL2 (Z[a, a−1 ]) −→ GL2 (Z[a, a−1 ])
that
x δ −1
∈ GL2 Z[a, a ] ⊂ Fix(σ2 ).
δ δ2
As a final task for this chapter, we’ll define a family of links that generate such
evaluation matricies.
Consider a slightly different diagram of the knot K(T, U ), given in figure 5.5.
From this diagram, we are led to define a rather exotic braid closure that will
be of use. That is, for the pair of tangles (T, U ) and an appropriately chosen
braid β ∈ B6 we define a link |β| as in figure 5.6.
for each k ∈ {1, . . . , n − 1}. Note that when m = 0, this is reduces to the
natural inclusion Bn < BN . Now the group B3 ⊕ B3 arises as a subgroup of
B6 by choosing
B3 ⊕ B3 −→ B6
(α, β) 7−→ i0 (α)i3 (β).
Notice that the image i0 (α)i3 (β) contains no occurrence of the generator σ 3
and hence
i0 (α)i3 (β) = i3 (β)i0 (α)
in B6 . Now define the switch homomorphism
s : B3 −→ B3
σ1 7−→ σ2−1
σ2 7−→ σ1−1
and note that, given a 180 degree rotation ρ in the projection plane, ρ(sβ) =
β −1 .
Definition 5.5. For each α ∈ B3 define the Kanenobu braid i0 (α)i3 (sα) ∈ B6 .
where K(T, U ) = |β| for some Kanenobu braid β. Since ρ(sα) = α −1 , the link
K(∞, ∞) is of the form
so that
hK(∞, 0)i = hK(0, ∞)i = δ.
Finally, let the polynomial hK(0, 0)i = x so that
x δ
X =
δ δ2
is an element of Fix(σ2 ).
Chapter 5. Kanenobu Knots 53
where
x = − a−20 + 2a−16 − 4a−12 + 6a−8 − 7a−4
+ 9 − 7a4 + 6a8 − 4a12 + 2a16 − a20 .
Chapter 5. Kanenobu Knots 54
Figure 5.10: The knot (σ13 σ2−1 σ1 )(σ5−3 σ4 σ5−1 )
In particular,
VK(0,0) = x
and we have that
VK(σ2 ,σ−1 ) = x.
2
(−t−4 + 3 − t4 )
+ (−t−4 + t−2 + 4 − t2 + t4 − t6 )x2
+ (t−2 + 2 + t2 )x4
4
computed using KNOTSCAPE
Chapter 5. Kanenobu Knots 55
we can form the Kanenobu braid (σ12 σ2−3 σ1 )(σ5−2 σ43 σ5−1 ) ∈ B6
Figure 5.14: The Kanenobu braid (σ12 σ2−3 σ1 )(σ5−2 σ43 σ5−1 )
In particular,
VK(0,0) = x
and we have that
VK(σ2 ,σ−1 ) = x.
2
Chapter 5. Kanenobu Knots 57
Figure 5.15: The knot (σ12 σ2−3 σ1 )(σ5−2 σ43 σ5−1 )
Once again we have distinct knots that have the same Jones polynomial but
are not related by Conway mutation.
5
computed using KNOTSCAPE
Chapter 5. Kanenobu Knots 58
5.7 Observations
Definition 5.7. If a link is equivalent to a 3-braid, closed as in figure 5.18, it
is called a 2-bridge link.
For a generalised Kanenobu knot K(T, U ), the knot K(0, 0) is always of the
form L#L? where L is a 2-bridge link.
In the case where K(0, 0) ∼ K#K ? is a connected sum of 2-bridge knots with
more that 3 crossings, such a K is generated by taking the 2-bridge closure of
an element α ∈ B3 . Such a braid generates the Kanenobu braid i 0 (α)i2 (sα),
and taking the closure
|i0 (α)i2 (sα)| = K(T, U )
with U = T ? gives rise to the evaluation matrix
hK#K ? i δ
K=
δ δ2
since K#K ? = K(0, 0). Now K ∈ Fix(σ2 ), and with the specification that
U = T ? , the familly of knots
−1
K T σ2 , U σ2
V n −n = hK#K ? i.
K(T σ2 ,U σ2 )
From the diagram in figure 5.20 it can be seen that the action of σ 2 can cancel
along a band connecting the tangles.
Figure 5.20: The knot (σ1 σ2−1 σ1 )(σ5−1 σ4 σ5−1 ).
∼
so in the case that T is a rational tangle, their knot type is unaltered, while
a more general tangle results in a Conway mutant of the original diagram. In
particular, there is no change to the Jones polynomial.
In general however, the set of tangles S 2 together with the set of 2-bridge
knots (generated by B3 ) provide a range of knots (and even links) having
evaluation matricies contained in Fix(σ 2 ). In the cases discussed and the
examples produced, we have seen that the HOMFLY polynomial may be used
to distingush these knots. Thus, we conclude that this method of producing
Chapter 5. Kanenobu Knots 61
6.1 Construction
The group action of braids on tangles presented in this work was originally
discussed by Eliahou, Kauffman and Thistlethwaite [9] in the course of study
of the recently discovered links due to Thistlethwaite [33]. While it is still
unknown whether there is a non trivial knot having Jones polynomial V = 1,
Thistlethwaite’s examples allow us to answer the question for links having
more than 1 component.
Theorem (Thislethwaite). For n > 1 there are non trivial n-component
links having trivial Jones polynomial V = δ n−1 .
In the exploration of these links [9], it is shown that this is in fact a corollary
of a much stronger statement.
Theorem (Eliahou, Kauffman, Thislethwaite). For every n-component
link L there is an infinite family of (n + 1)-component links L 0 such that VL0 =
δVL .
While these assertions are discussed at length in [9], the goal of this chapter is
to present some of the examples in light of the group actions discussed in this
work.
Definition 6.1. A Thislethwaite link H(T, U ) is an external wiring of tangles
T, U ∈ S2 modeled on the Hopf link.
62
Chapter 6. Thistlethwaite Links 63
and
hH(∞, ∞)i = δ.
For the non trivial link H(0, 0), the computation of hH(0, 0)i requires a little
more work.
For this, the following switching formula (stated in [20]) will be useful.
Switching Formula. The equality
− = a4 − a−4 −
2 −2
+ a −a + − −
proof of the switching formula. First note that the double crossing
may be viewed as the braid σ2 σ1 σ3 σ2 ∈ B4 . Thus, as
σi 7−→ a + a−1 ei
gives a representation of the braid group in TLn , we can represent this element
of B4 as
(a−2 + e1 + e2 + a2 e2 e1 )(a−2 + e2 + e3 + a2 e3 e2 )
=a−4 + a−2 e1 + (2a−2 + δ + 2a2 )e2 + a−2 e3
+ a 2 e1 e3 e2 + a 2 e2 e1 e3 + a 4 e2 e1 e3 e2
+ e 1 e2 + e 1 e3 + e 2 e1 + e 2 e3 + e 3 e2 .
Chapter 6. Thistlethwaite Links 65
Therefore
−
as required.
* +
3 4 −4
=δ + a −a −
= δ 3 + (a4 − a−4 )
−(a4 − a−4 )(δ − δ 3 ) − (a2 − a−2 )(2 − 2δ 2 )
= δ 3 + (a4 − a−4 )(δ 2 − 1) δ(a4 − a−4 ) + 2(a2 − a−2 )
= δ 3 + (a4 − a−4 )(δ 2 − 1) a−6 − a−2 + a2 − a6
T = and U =
ω //
ω −1 //
Chapter 6. Thistlethwaite Links 67
−1
Therefore, the link H(T ω , U ω ) is two linked trefoils and, depending on ori-
−1
entation, w(H(T ω , U ω )) = ±8
−1
Figure 6.4: The link H(T ω , U ω )
so that
−1
hH(T ω , U ω )i = δ
and
VH(T ω ,U ω−1 ) = −a±24 δ.
However, the action of ω 2 leaves the writhe unchanged. This gives rise to a
family of 2-component Thisltlethwaite links, all having Jones polynomial δ.
Taking tangles T, U as in the previous example, the links
2n −2n
H(T ω , U ω )
have Jones polynomial δ for all n ∈ Z. Moreover, for n 6= 0 the links obtained
are non-trivial, since each component is the numerator closure of a tangle,
giving rise to a pair of 2-bridge links that are geometrically essential to a pair
of linked solid tori [32].
The first two links in this sequence (for n = 1, 2) are shown in figures 6.5 and
6.6.
Chapter 6. Thistlethwaite Links 68
(T, U ) = ,
we obtain a trivial link H(T, U ) such that w(H(T, U )) = −3. Applying the
action of ω to this link gives rise to a non-trivial link
such that
−1
hH(T, U )i = hH(T ω , U ω )i
and
−1
w H(T ω , U ω ) = −3.
The result is a non-trivial link with trivial Jones polynomial δ.
Similarly, starting with the pair of tangles
(T, U ) = ,
gives rise to another trivial link H(T, U ), in this case having w(H(T, U )) = −1.
Applying the action of ω to this link gives rise to a non-trivial link
such that
−1
hH(T, U )i = hH(T ω , U ω )i
and
−1
w H(T ω , U ω ) = −1.
Chapter 6. Thistlethwaite Links 70
(T, U ) = , ,
[3] Emil Artin. Theorie der Zöpfe. Abh. Math. Sem. Univ. Hamburg, 4:47–72,
1925.
[6] Joan S. Birman. Braids, Links and Mapping Class Groups. Number 82
in Annals of Mathematics Studies. Princeton University Press, Princeton,
1974.
[11] Roger Fenn. Hecke algebras, 2004. Notes from ‘Knots in Vancouver’.
[13] B. Hartley and T. O. Hawkes. Rings, Modules and Linear Algebra. Chap-
man and Hall Mathematics Series. Chapman and Hall, 1971.
72
Bibliography 73
[14] Jim Hoste and Jósef H. Przytycki. Tangle surgeries which preserve Jones-
type polynomials. Int. J. of Math., 8(8):1015–1027, 1997.
[18] Taizo Kanenobu. Infinitely many knots with the same polynomial invari-
ant. Proc. AMS, 97(1):158–162, 1986.
[19] Louis Kauffman. State models and the Jones polynomial. Topology,
26:395–407, 1987.
[20] Louis Kauffman and Sóstenes Lins. Temperley-Lieb recoupling theory and
invariants of 3-manifolds. Number 134 in Annals of Mathematics Studies.
Princeton University Press, Princeton, 1994.
[24] A. A. Markov. Über die freie Äquivalenz geschlossener Zöpfe. Rec. Math.
Moscou, 1:73–778, 1935.
[26] H. R. Morton. Treading knot diagrams. Mth. Proc. Camb. Phil. Soc.,
99:247–260, 1986.
Bibliography 74
[30] Dale Rolfsen. The quest for a knot with trivial Jones polynomial: Dia-
gram surgery and the Temperley-Lieb algebra. In M.E. Bozhüyük, editor,
Topics in Knot Theory, volume 399 of Nato ASI Series C, pages 195–210.
Kluwer Academic Publishers, 1993.
[32] Dale Rolfsen. Knots and Links. AMS Chelsea, Providence, third edition,
2003.