a

r
X
i
v
:
m
a
t
h
/
0
3
1
0
1
4
3
v
1


[
m
a
t
h
.
Q
A
]


1
0

O
c
t

2
0
0
3
Graphical Calculus on Representations of Quantum Lie
Algebras
By
Dongseok KIM
B. S. (Kyungpook National University, Korea) 1990
M. S. (Kyungpook National University, Korea) 1992
M. A. (University of Texas at Austin) 1998
DISSERTATION
Submitted in partial satisfaction of the requirements for the degree of
DOCTOR OF PHILOSOPHY
in
MATHEMATICS
UNIVERSITY OF CALIFORNIA
DAVIS
March 2003
i
Contents
1 Introduction 1
1.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Single Clasp Expansions for Rank 2 Lie Algebras 6
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2 Single Clasp Expansion for U
q
(sl(2, C)) . . . . . . . . . . . . . . . . . 7
2.2.1 Jones-Wenzl Projector for U
q
(sl(2, C)) . . . . . . . . . . . . . 7
2.2.2 Applications of Single Clasp Expansions for U
q
(sl(2, C)) . . . . 9
2.3 Single Clasp Expansion for U
q
(sl(3, C)) . . . . . . . . . . . . . . . . . 11
2.3.1 Single Clasp Expansions of a Clasp of Weight (n, 0) and (0, n) 13
2.3.2 Single Clasp Expansions of a Non-segregated Clasp of Weight (a, b) 13
2.3.3 Double Clasp Expansion of a Segregated Clasp of Weight (a, b) 18
2.4 Single Clasp Expansion for U
q
(sp(4)) . . . . . . . . . . . . . . . . . . 31
3 Trihedron Coecients for U
q
(sl(3, C)) 37
3.1 General Shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2 Proofs of Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.1 Two Zeros . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.2.2 Proof of Theorem 3.1 . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.3 Proof of Theorem 3.2 . . . . . . . . . . . . . . . . . . . . . . 50
ii
4 A Complete Set of Relations of U
q
(sl(4, C)) 52
iii
ACKNOWLEDGEMENTS
I would like to thank my adviser, Greg Kuperberg. He has been exceptionally pa-
tient to show me how to look Mathematics in many dierent prospects. He has shared
many ideas with me, as well as guided me to the subject of representation theory.
None of these could be done without his supervision. I would like to thank Professors
Dimtry Fuchs, Joel Hass, Albert Schwartz and William Thurston for teaching me. I
would like to thank Professor Khovanov for taking time to teach me many things that
I should know and helping me to untie some of my research. I would like to thank my
friends in the math department for helping me get through the period of exhaustive
study.
I would like to thank my family. They have always supported my education. I
especially thank my wife Youngmi, for standing through our stay in Unites States.
I can not forget my parent who has been incredibly supportive for all. My children,
Richard and Hannah have been the biggest motivation for me to stay in focus.
iv
Abstract
The main theme of this thesis is the representation theory of quantum Lie algebras.
We develop graphical calculation methods. Jones-Wenzl projectors for U
q
(sl(2, C))
are very powerful tools to nd not only invariants of links but also invariants of 3-
manifolds. We nd single clasp expansions of generalized Jones-Wenzl projectors for
simple Lie algebras of rank 2. Trihedron coecients of the representation theory for
U
q
(sl(2, C)) has signicant meaning and it is called 3j symbols. Using single clasp
expansions for U
q
(sl(3, C)), we nd some trihedron coecients of the representation
theory of U
q
(sl(3, C)). We study representation theory for U
q
(sl(4, C)). We conjecture
a complete set of relations for U
q
(sl(4, C)).
v
CHAPTER 1. Introduction 1
Chapter 1
Introduction
There has been big progress in the theory bridging Lie algebras and low-dimensional
topology. These developments are based on quantum groups, braided categories and
new invariants of knots, links and 3-manifolds. After the discovery of the Jones poly-
nomial [Jon85] [Jon87], Reshetikhin and Turaev [RT90] [RT91] showed that braided
categories derived from quantum groups provide a natural generalization of the Jones
polynomial.
One of the developments is that a category of tangles with skein relations leads
to a braided category. If we decorate each component of a tangle by a module over
a simple Lie algebra, the category becomes a ribbon category. Then we can get an
invariant of links, and sometimes 3-manifolds, from a functor constructed in [Tur94].
To develop this theory further, we would like to generalize the Jones-Wenzl projectors
in the Temperley-Lieb algebra to the quantization of other simple Lie algebras. The
n-th Temperley-Lieb algebra is realized as the algebra of intertwining operators of the
U
q
(sl(2, C))-module V
n
1
, where V
1
is the two-dimensional irreducible representation
of U
q
(sl(2, C)). For each n, the algebra T
n
has an idempotent f
n
such that f
n
x =
xf
n
= (x)f
n
for all x T
n
and f
n
f
n
= f
n
, where is an augmentation. These
idempotents were rst discovered by V. Jones [Jon83] and H. Wenzl [Wen87], and
CHAPTER 1. Introduction 2
they found a recursive formula:
f
n
= f
n1
+
[n 1]
[n]
f
n1
e
n1
f
n1
.
So they are named Jones-Wenzl idempotents(Projectors). Kuperberg [Kup96]
denes a generalization of the Temperley-Lieb category to the three rank two Lie
algebras sl(3, C), sp(4) and G
2
. These generalizations are called combinatorial rank
two spiders. Also he has proved that Jones-Wenzl projectors exist for simple Lie
algebras of rank 2 and he called them clasps. We will study how they can be expanded
inductively in Chapter 2.
The skein module theory allows not only links but also graphs. The invariants
of the two simplest nontrivial trivalent graphs, the trihedron and tetrahedron, have
signicant meaning and they are called 3j and 6j symbols. So we can naturally ask
how to compute trihedron coecients for U
q
(sl(3, C)) as suggested in [Kup96]. In
Chapter 3, we will apply our clasp expansions to nd some trihedron coecients for
U
q
(sl(3, C)).
Kuperbergs generalization of the Temperley-Lieb algebra is a set of generators
and relations for each rank 2 Lie algebra [Kup96]. The generators are easy to nd,
but to get a complete set of relations is a challenging problem. In Chapter 4, we
follow Kuperbergs method to nd some relations. We conjecture a complete set of
relations of U
q
(sl(4, C)).
1.1 Preliminaries
For simple terms, we refer to [Hum72] [Kas95] and [CKT97].
Quantum integers are dened as
CHAPTER 1. Introduction 3
[n] =
q
n/2
q
n/2
q
1/2
q
1/2
[0] = 1
[n]! = [n][n 1] . . . [2][1]
_
_
n
k
_
_
=
[n]!
[k]![n k]!
Let sl(n, C) be the Lie algebra of complex n n-matrices with trace zero. Let
E
i,j
be the elementary matrix whose entries are all zero except 1 in the (i, j)-th entry.
Let E
i
= E
i,i+1
, F
i
= E
i+1,i
and H
i
= E
i,i
E
i+1,i+1
where 1 i n 1, then they
generate sl(n, C) with relations:
[H
i
, H
j
] = 0 for i, j = 1, 2, . . . , n 1
[H
i
, E
j
] =
j
(H
i
)E
j
for 1 i, j n 1
[H
i
, F
j
] =
j
(H
i
)F
j
for 1 i, j n 1
[E
i
, F
j
] =
ij
H
i
for 1 i, j n 1
[E
i
, E
j
] = 0 if |i j| 2
[F
i
, F
j
] = 0 if |i j| 2
[E
i
, [E
i
, E
j
]] = 0 if |i j| = 1
[F
i
, [F
i
, F
j
]] = 0 if |i j| = 1
where
i
is a linear form dened by

j
(H
i
) =
_

_
2, if i = j
1, if |i j| = 1
0, Otherwise
CHAPTER 1. Introduction 4
The quantum group U
q
(sl(n, C)) is an associative algebra over C(q) with genera-
tors, E
i
, F
i
, K

i
with 1 i n 1, and relations:
K
i
K
1
i
= 1 = K
1
i
K
i
for i = 1, 2, . . . , n 1
K
i
E
j
= q

j
(H
i
)
E
j
K
i
for i, j = 1, 2, . . . , n 1
K
i
F
j
= q

j
(H
i
)
F
j
K
i
for i, j = 1, 2, . . . , n 1
[E
i
, F
j
] =
ij
K
i
K
1
i
q q
1
for i, j = 1, 2, . . . , n 1
[E
i
, E
j
] = 0 if |i j| 2
[F
i
, F
j
] = 0 if |i j| 2
E
2
i
E
j
(q + q
1
)E
i
E
j
E
i
+ E
j
E
2
i
= 0 if |i j| = 1
F
2
i
F
j
(q + q
1
)F
i
F
j
F
i
+ F
j
F
2
i
= 0 if |i j| = 1
Let h be the Lie subalgebra of sl(n, C) generated by H
i
and let h be the
integral lattice of linear forms on H
i
where n1 i 1. Let , = (
1
, . . . ,
n1
),
then there is a unique universal highest weight module, a Verma Module, with highest
weight (, ). M(, ) has a unique simple quotient L(, ) which is highest module
with highest weight (, ). Then L(, ) is nite dimensional if and only if is
dominant weight. One can see that as U
q
(sl(n, C)) module
L(, )

= L(, 0) L(0, ).
So we can study L(, 0) which is denoted by L(). Then there is a theorem which
connects studies of U
q
(sl(n, C)) modules and sl(n, C) modules.
Theorem 1.1 [CKT97]
i) Any nite dimensional simple U
q
(sl(n, C)) module is of the form L(, ) where is
dominant weight and (Z/2Z)
n1
.
CHAPTER 1. Introduction 5
ii) The character ch(L()) is given by the same formula as the character of simple
sl(n) module parameterized by the same highest weight.
iii) The multiplicity of a simple module L() in the decomposition of the tensor prod-
uct L() L() of two simple modules is the same as for the decomposition of the
corresponding sl(n, C) module.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 6
Chapter 2
Single Clasp Expansions for Rank
2 Lie Algebras
2.1 Introduction
Let T
n
be the n-th Temperley-Lieb algebra with generators, 1, e
1
, e
2
, . . . , e
n1
, and
relations:
e
2
i
= (q
1
2
+ q

1
2
)e
i
e
i
e
j
= e
j
e
i
if |i j| 2
e
i
= e
i
e
i1
e
i
For each n, the algebra T
n
has an idempotent f
n
such that f
n
x = xf
n
= (x)f
n
for all x T
n
, where is an augmentation. These idempotents were rst discovered
by V. Jones [Jon83] and H. Wenzl [Wen87]. They found a recursive formula:
f
n
= f
n1
+
[n 1]
[n]
f
n1
e
n1
f
n1
as in the following gure where we use a red box to represent f
n
:
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 7
n
n
=
n 1
n 1
+
[n 1]
[n]
n 1
n 2
n 1
(2.1)
So they are named Jones-Wenzl idempotents(projectors). We will recall an al-
gebraic denition of Jones-Wenzl projectors in section 1. We refer to [Kup96] for
denitions, notation and simple calculations. We provide single clasp expansions of
generalized Jones-Wenzl projectors for U
q
(sl(3, C)) in section 2. In section 3 we study
single clasp expansions of generalized Jones-Wenzl projectors for U
q
(sp(4)).
2.2 Single Clasp Expansion for U
q
(sl(2, C))
First we recall another denition of Jones-Wenzl projectors and single clasp expan-
sions for U
q
(sl(2, C)). Then we use it to nd trihedral coecients for U
q
(sl(2, C)).
2.2.1 Jones-Wenzl Projector for U
q
(sl(2, C))
Let us give a precise denition [Kho97] of a clasp for U(sl(2, C)). Let V
i
be an
irreducible representation of highest weight i. Then i
n
: V
n
V
n
1
is dened by
i
n
(v
m
) =
_
_
n
nm
2
_
_
1

s,|s|=m
q
||s||

v
s
1
. . . v
sn
and
n
: V
n
1
V
n
is dened by

n
(v
s
1
v
s
2
. . . v
sn
) = q
||s||
+
v
|s|
where s = (s
1
, s
2
, . . . , s
n
), s
i
= 1, |s| =

s
i
and ||s||
+
=

i<j
{s
i
> s
j
},
||s||

i>j
{s
i
> s
j
} and {a > b} = 1 if a > b, and 0 otherwise.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 8
Then the composition i
n

n
is called a Jones-Wenzl projector, denoted by p
n
. It
has the following properties 1) it is an idempotent 2) p
n
e
i
= 0 = e
i
p
n
where e
i
is a
U-turn from the i-th to the i + 1-th string as in the following gures.
n
n
n = n n , n
k
n k 2
= 0
We can generalize the second property as follows: if we attach a web with a cut
path with less weight, then it is zero. Then we can axiomatize these two properties
to dene generalized Jones-Wenzl projectors for any simple Lie algebra. Kuper-
berg [Kup96] proved that Jones-Wenzl projectors exist for simple Lie algebra of rank
2 and he called clasps (sometimes they are called magic weaving elements or boxes).
Here, we will call them clasps. For U
q
(sl(2, C)), it is known that we can inductively
expand it as in equation 2.1. For advanced calculations, the single clasp expansion in
equation 2.2 is very useful and has been used in [Kho97] for some beautiful results. By
symmetry, there are four dierent positions for the single clasp expansion depending
on where the clasp of weight n 1 is located. For equation 2.2, the clasp is located
at the southwest corner, which will be considered the standard expansion, otherwise,
we will state the location of the clasp.
n
n
=
n

i=1
a
i
n 1
i 1
(2.2)
Proposition 2.1 The coecients in equation 2.2 are
a
i
=
[n + 1 i]
[n]
.
Proof: By attaching a U turn at consecutive strings to the top, we have the following
n 1 equations.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 9
a
n1
[2]a
n
= 0.
For i = 1, 2, . . . , n 2,
a
i
[2]a
i+1
+ a
i+2
= 0.
One can see that these equations are independent. By attaching the clasp of
weight n to the bottom of every web in equation 2.2, we get a
1
= 1 by the properties
of a clasp. This process is called a normalization. Then we check the answer in the
proposition satises these equations. Since these webs in equation 2.2 form a basis,
these coecients are unique. 2
2.2.2 Applications of Single Clasp Expansions for U
q
(sl(2, C))
We can easily prove the following propositions using the single clasp expansion of
U
q
(sl(2, C)). Let a + b = c + d and b = min{a, b, c, d}.
c
d
a b
=
b

k=0
a
k
c k b k
k
k
a b
c
d
d b
(2.3)
Proposition 2.2 The coecients in equation 2.3 are
a
k
=
[c]![b]![a + b k]!
[c k]![b k]![k]![a + b]!
.
Proof: We induct on a + b. If a + b = 1, it is clear. Without loss of generality,
we assume that a b. Denote the diagram corresponding to the coecient a
k
in the
right hand side of equation 2.3 by D(k). By applying a single clasp expansion for
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 10
the clasp of weight a + b, then a single clasp expansion of the clasp located at the
northeast corner, we get
c
d
a b
=
a b
a b 1
c d 1
c
d

[a][c]
[a + b][a + b 1]
a 1
c 1
b 1
d 1
a b
c
d
By induction, the right side equals
b1

k=0
[c]![b 1]![a + b 1 k]!
[c k]![b 1 k]![k]![a + b 1]!
D(k)
+
[a][c]
[a + b][a + b 1]
b1

k=0
[c 1]![b 1]![a + b 2 k]!
[c 1 k]![b 1 k]![k]![a + b 2]!
D(k + 1)
= D(0) +
b1

k=1
(
[c]![b 1]![a + b 1 k]!
[c k]![b 1 k]![k]![a + b 1]!
+
[a][c]
[a + b][a + b 1]
[c 1]![b 1]![a + b 2 (k 1)]!
[c 1 (k 1)]![b 1 (k 1)]![k 1]![a + b 2]!
)D(k)
+
[a][c]
[a + b][a + b 1]
[c 1]![b 1]![a + b 2 (b 1)]!
[c 1 (b 1)]![b 1 (b 1)]![b 1]![a + b 2]!
D(b)
= D(0) +
b1

k=1
[c]![b]![a + b k]!
[c k]![b k]![k]![a + b]!
(
[b k][a + b] + [k][a]
[b][a + b k]
)D(k)
+
[c]![b]![a]!
[c b]![0]![b]![a + b]!
D(b) =
b

k=0
[c]![b 1]![a + b k]!
[c k]![b k]![k]![a + b]!
D(k).
We use a well-known identity for quantum integers,
[m+ r][n + r] = [m][n] + [m+ n + r][r]
in the 6-th line of the above equation with n = a, m = k and r = a + b. 2
Next we look at the trihedron coecient(or 3j symbol) [Lic72] [MV94] [Tur94].
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 11
Proposition 2.3 The trihedron coecient is
j
j
i
k
= (1)
i+j+k
[i + j + k + 1]![i]![j]![k]!
[i + j]![j + k]![i + k]!
.
Proof: The idea of the proof is identical to the previous proposition. We induct on
i + j + k. If i + j + k = 1, it is just a circle, so its value is
[2] = (1)
1
[2]![1]![0]![0]!
[1]![1]![0]!
.
We apply a single clasp expansion for the clasp of weight i + k, then another single
clasp expansion, for which the clasp is located at the northeast corner. By induction,
we have
=
[j + k + 1]
[j + k]
(1)
i+j+k1
[i + j + k]![i]![j]![k 1]!
[i + j]![i + k 1]![j + k 1]!
+
[i][i]
[i + k][i + k 1]
(1)
i+j+k1
[i + j + k]![i 1]![j + 1]![k 1]!
[i + j]![i + k 2]![j + k]!
= (1)
i+j+k
[i + j + k + 1]![i]![j]![k]!
[i + j]![i + k]![j + k]!
(
[i + k][j + k]
[i + j + k + 1][k]

[i][j + 1]
[i + j + k + 1][k]
)
= (1)
i+j+k
[i + j + k + 1]![i]![j]![k]!
[i + j]![i + k]![j + k]!
2
2.3 Single Clasp Expansion for U
q
(sl(3, C))
A complete set of relations for skein theory of U
q
(sl(3, C)) as given in equations 2.4
2.5 2.6 was found in [Kup96]. There is a relation for every elliptic face, a face with less
than 6 edges. We call the relation 2.6 a rectangular relation and the rst(second) shape
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 12
in the right side of the equality is called a horizontal(vertical, respectively) splitting.
For several reasons, such as positivity and integrality, we use [2] in relation 2.5 but
one can use a quantum integer [2] and get an independent result. If one uses [2], one
can rewrite all results in here by multiplying each trivalent vertex by the complex
number i.
= [3] (2.4)
= [2] (2.5)
= + (2.6)
A clasp for U
q
(sl(3, C)) can be dened axiomatically: 1) it is an idempotent and
2) if we attach a U turn or a Y , it becomes zero. An explicit denition of clasps for
U
q
(sl(3, C)) can be found in [Kup96].
First we look at a single clasp expansion of the clasp of weight (a, 0) where the
weight (a, b) stands for a
1
+b
2
and
i
is a fundamental dominant weight of sl(3, C).
Each directed edge represents V

i
, the fundamental representation of the highest
weight
i
. We might use the notation +, for V

1
, V

2
but it should be clear.
We recall the usual partial ordering of the weight lattice of lattice of sl(3, C) as
a
1
+ b
2
(a + 1)
1
+ (b 2)
2
a
1
+ b
2
(a 2)
1
+ (b + 1)
2
.
A cut path is a path which is transverse to strings of web, and the weight of a cut
pass is (a, b) if it passes a strings decorated by V

1
and b strings decorated by V

1
.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 13
2.3.1 Single Clasp Expansions of a Clasp of Weight (n, 0) and (0, n)
A basis for the single clasp expansion is given in equation 2.7. If we attach a Y on
the top of webs in the equation 2.7, there is at least one elliptic face on which we
can apply our relations. This process gives us exactly the same equations we got for
U
q
(sl(2, C)). Thus, we can easily establish proposition 2.4. Moreover, this single clasp
expansion holds for any U
q
(sl(n, C)) where n 4 because U
q
(sl(3, C)) is naturally
embedded in U
q
(sl(n, C)). Later we will mention the importance of this fact. As same
as for U
q
(sl(2, C)), there are four dierent positions for the single clasp expansions
for U
q
(sl(3, C)) so we use the same convention we used for single clasp expansion for
U
q
(sl(2, C)).
n
. . .
n
=
n

i=1
a
i
n 1
. . .
i
. . .
1
(2.7)
Proposition 2.4 The coecients in equation 2.7 are
a
i
=
[n + 1 i]
[n]
.
Also, we can easily nd a single clasp expansion of the clasp of weight (0, b) by
reversing arrows.
2.3.2 Single Clasp Expansions of a Non-segregated Clasp of Weight (a, b)
The most interesting case is a single clasp expansion of the clasp of weight (a, b).
First of all, we nd the dimension of Inv(V
a+1

1
V
b

2
V
(b1)
1
+a
2
) in lemma 2.5.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 14
Lemma 2.5
dim(Inv(V
a+1

1
V
b

2
V
(b1)
1
+a
2
)) =
_

_
a if b = 0
(a + 1)b if b > 0
Proof: For b = 0, the result follows from proposition 2.4. We induct on b to show
the multiplicity of V
a
1
+(b1)
2
, V
(a+1)
1
+b
2
and V
(a1)
1
+(b+1)
2
in the decomposition
of V
a+1
1,0
V
b
0,1
into irreducible representations are (a +1)b, 1 and a respectively. By
a simple application of Schurs Lemma, we nd that the dimension of Inv(V
a+1

V
b

2
V
(b1)
1
+a
2
) is equal to the multiplicity of V
a
1
+(b1)
2
in the decomposition of
V
a+1

1
V
b

2
into irreducible representations. 2
Lemma 2.5 work for any A
n
where n 2 by replacing
2
by
n
. Moreover the
single clasp expansion of the clasp of weight a
1
+b
n
as in equation 2.7 is also true
for any U
q
(sl(n, C)) where n 3.
We need a set of basis webs with nice rectangular order, but we can not nd one
in the general case. Even if one nd a basis, they have many hexagonal faces which
make it very dicult to get numerical relations. So we start from an alternative,
non-segregated clasp. A non-segregated clasp is obtained from the segregated clasp
by attaching a sequence of Hs until we get the desired shape of edge orientations.
Fortunately, there is a canonical way to nd by putting H from the leftmost string of
weight
2
or until it reach to the desired position. In the following lemma, we will
show the non-segregated clasp is well dened. The following gure is an example of
a non-segregated clasp of weight (2, 3) and how to obtain it from a segregated clasp
of weight (2, 3).
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 15
+
+
+
+ +
=
+
+
+
+ +
Lemma 2.6 Non-segregated clasps are well-dened.
Proof: Let be a sequence of Hs which induce the same non-segregated clasp. We
nd the rst string from the leftmost of sign which does move to right. If there is
no such a string, then the clasp is canonical. Otherwise, there exist two consecutive
Hs which can be removed by the horizontal splitting because it had to reach the
desired position. We induct on the length of the sequence and it completes the proof.
2
We also nd that non-segregated clasps satisfy two properties of segregated clasps.
Lemma 2.7 1) Two consecutive non-segregated clasps is equal to a non-segregated
clasp.
2) If we attach a web to a non-segregated clasp and if it has a cut path of which
weight is less than the weight of the clasp, then it is zero.
Proof: Since there is two consecutive non-segregated clasps, we see that the bottom
end of the upper non-segregated clasp and the top end of the lower non-segregated
clasp are the same non-segregated. Since non-segregated clasp does not depend on
the choice of order of attaching Hs, we x the canonical one for the both side then
we an see that one is the others inverse, the inverse is just the horizontal reection.
Therefore, all the Hs in the middle cancel out and standard clasps are idempotents.
The second part is obvious because adding Hs does not change the weight of the
minimal path so we can change to a segregated clasp by adding more Hs. Then the
clasp becomes zero. 2
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 16
The following equation 2.8 is a single clasp expansion of a non-segregated clasp
of weight (a, b). Let us denote the web corresponding to the coecient a
i,j
by D
i,j
.
These webs form a basis because there are no applicable relations.
. . .
b a + 1
. . .
=
b

i=1
a

j=0
a
ij
1
i 1
j
1
i 1 j
. . . . . .
. . . . . .
(1) (2)
(2.8)
Kuperberg [Kup96] showed that for a xed boundary, interior can be lled by
a cut out from the hexagonal tiling of the plane with the given boundary. For our
cases, there are two possible llings but we use the maximal cut out of the hexagonal
tiling. We draw examples of the case i = 6, j = 5 and the rst one in equation 2.9 is
not maximal cut out and the second one is the maximal cut out which ts to the left
rectangle and the last one is the maximal cut out which ts to the right rectangle as
the number indicates in equation 2.8.
,
(1)
,
(2)
(2.9)
Theorem 2.8 The coecients in equation 2.8 are
a
i,j
=
[b i + 1]
[b]
[b + j + 1]
[a + b + 1]
.
Proof: As usual, we attach a Y or a U turn to nd one exceptional and three types
of equations as follow.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 17
[3]a
1,0
[2]a
1,1
[2]a
2,0
+ a
2,1
= 0.
Type I : For j = 0, 1, . . . , a,
a
b1,j
[2]a
b,j
= 0.
Type II : For i = 1, 2, . . . , b 2 and j = 0, 1, . . . , a.
a
i,j
[2]a
i+1,j
+ a
i+2,j
= 0.
Type III : For i = 1, 2, . . . , b and j = 0, 1, . . . , a 2.
a
i,j
[2]a
i,j+1
+ a
i,j+2
= 0.
We establish the following lemma 2.9 rst.
Lemma 2.9 Let a
1,0
= x, then the coecients in the equation 2.8 is
a
i,j
=
[b i + 1]
[b]
[b + j + 1]
[b + 1]
x.
Proof: First we can see that the right side of equation 2.8 is a basis for a single clasp
expansion because its number of the webs is equal to the dimension as in lemma 2.5
and none of these webs has any faces. Second we nd that these equations has at
least (a +1)b independent equations. Then we plug in these coecients to equations
to check that they are the right coecients. 2
Usually we normalize one basis web in the expansion to get a known value. But
we can not normalize for this expansion yet because it is not a segregated clasp. Thus
we use lemma 2.12 to nd that the coecient of a
1,a
is 1. Then, we get x =
[b+1]
[a+b+1]
and it completes the proof of the theorem. 2
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 18
2.3.3 Double Clasp Expansion of a Segregated Clasp of Weight (a, b)
Now we study a single and a double clasps expansion of a segregated clasp of weight
(a, b). We will start with an example, a single clasp expansion of a clasp of weight
(2, 2). To apply theorem 2.8, we add some Hs to change the segregated clasp to a
non-segregated clasp.
=
Then we expand the clasp. In the following equation 2.10, we will omit the
direction of edges unless there is an ambiguity.
[3]
[5]
+
[4]
[5]
+
[5]
[5]
+
[1][2]
[3][5]
+
[1][2]
[4][5]
+
[1][2]
[5][5]
(2.10)
These webs can be expanded using relations. For example, the rst one can be
expanded by equation 2.6 as
= [2]
For some small cases, we can expand this way but it will be dicult to manage
all possible expansions. An other way to look at this expansion is to use paths : since
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 19
there are ve points on the top and three points right above the clasp and these three
points have to be connected to points on the top (otherwise, we have a cut path with
weight less than (2, 1) which makes the web zero), we have two Y s or one U turn.
We rst nd all possible disjoint, monotone(except at Ys) paths connecting these
points. For the web on above example, there are eight possibilities as follows.
, , ,
, , , (2.11)
If we examine them to determine whether it will appeared in the actual expansion,
the rst two in the second row appear but the rest of them do not. One can see that
U turn can appear only once at the very top.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 20
=
[3]
[5]
_

_
[2] [2]
_

_
+
[4]
[5]
_

_
[2] + +
_

_
+
[5]
[5]
_

_
+
_

_
+
[1][2]
[3][5]
_

_
+ +
_

_
+
[1][2]
[4][5]
_

_
+ +
_

_
+
[1][2]
[5][5]
_

_
+
_

_
(2.12)
Thus, we get a single clasp expansion of a clasp of weight (2, 2) as follows.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 21
= +
[1]
[2]

[1]
[2][5]

[2]
[5]

[1]
[5]

[1]
[5]
(2.13)
By attaching (2, 1) clasps on the left top of every web in the right side of equa-
tion 2.13, we get the following double claps expansion of the clasp of weight (2, 2).
= +
[1]
[2]

[1]
[5]
(2.14)
Unfortunately, there is no particular order we can put for these basis webs for
single clasp expansions. But for the double clasp expansion, we can generalize the
euqation 2.14 as follow. In equation 2.15, the green box between two clasps is the
unique cut out from the hexagonal tiling with the given boundary as we have seen in
Figure 2.9. For equation 2.15 we assume that a b 1.
a
a b
b
=
a
a b 1
b 1
+
a b 1
a
b 1
b 2 a +
a 1 b 1
a 1 b 1
1 1
1 1
a 1
(2.15)
Theorem 2.10 The coecients in equation 2.15 are =
[b1]
[b]
, =
[a]
[b][a+b+1]
. We
assume that [0] = 0 for .
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 22
Proof: It follows from lemma 2.11 and lemma 2.12 that = a
2,a
and = a
2,a1

[2]a
1,a1
+ a
1,a
. 2
To prove two key lemmas, we generalize the idea of paths in the rst example.
First, we attach Hs as in gure 2.16 to all basis webs in equation 2.8 each of basis
web is denoted by D
ij
. After attaching Hs as in gure 2.16, the resulting web is
denoted by

D
ij
.
D
ij
b a

(2.16)
As we have seen in the example,

D
i,j
is not a basis web because it contains some
elliptic faces. If we decompose each

D
i,j
into a linear combination of some webs which
have no elliptic faces, then the union of all these resulting webs actually forms a basis.
Let us prove that these webs actually form a basis which will be denoted by D

,j
.
As vector spaces, this change, adding Hs, induces an isomorphism between two web
spaces. Its matrix representation with respect to these web bases {D
i,j
} and {D

,j
}
is an (a +1)b (a +1)b matrix whose entries are 0, 1 or [2]. In general, we will not
be able to write this matrix because there are many nonzero entries in every columns
and rows. But we know that the determinant of this matrix is [2]
ab
because each
one H contributes [2] depending on the directions.
Since

D
i,j
is not a basis web, to nd a single clasp expansion, we might have to use
relations to nd its linear expansion into a new web basis D

,j
. In general this might
not be done. If we just limit ourself to a double clasp expansion, We could use the
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 23
idea of paths as we demonstrated in the example. Let us formally dene it, a stem of
a web. Geometrically it is transversal to cut paths. From

D
i,j
, we see that there are
a+b+1 points on top but only a+b1 lines right above the clasp. Because of one of
properties of the clasp of weight (a, b1): if we have a cut path of weight which is less
than (a, b1), then the web becomes zero, we must have a+b1 vertical lines which
connect top a + b + 1 nodes to the clasp of weight (a, b 1) for non-vanishing webs
after applying relations. It is clear that these connecting lines should be mutually
disjoint, otherwise, we will have a cut path with weight less than (a, b 1). A stem of
a web is a disjoint union of lines as we described. Unfortunately some of stems do not
arise all cases because it may not be obtained by removing elliptic faces. If a stem
appears, we call it an admissible stem. For single clasp expansion, nding all these
stems will be more dicult than an expansion by relations but for the double clasp
expansion of segregated clasps, there are only few possible admissible stems whose
coecient is nonzero.
Lemma 2.11 After attaching a clasp of weight (a, b 1) to top of webs

D
i,j
from
equation 2.16, the only 3 non-vanishing shapes are those in Figure 2.17.

,

,

(2.17)
Proof: From

D
i,j
we see that there are a+b +1 lines on top and a+b 1 lines right
above the clasp. If we repeatedly use the rectangular relation as in equation 2.6, we
can push up the Y s so that there are either two Y s or one U shape at the top. It
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 24
is possible to have two adjacent Y s which appear in the second and third gures in
Figure 2.17 but a U turn can appear in only two places because of the orientation of
edges. If we attach the (a, b 1) clasp to the top of the resulting web from the left
and U or Y shape appear just below it, the web becomes zero. Therefore only these
three webs do not vanish. 2
For the next lemma, we will nd all

D
i,j
s which can be transformed to each of
the gures in Figure 2.17.
Lemma 2.12 Only

D
1,a
(

D
2,a
) can be transformed to the rst(second, respectively)
shape in Figure 2.17. Only the three webs,

D
1,a1
,

D
1,a
and

D
2,a1
can be transformed
to the last shape. Moreover, all of these transformations use only rectangular relations
as in equation 2.6 except for the transformation from

D
1,a1
to the third gure uses
one loop relation in equation 2.5.
Proof:
(2.18)
For the rst shape in gure 2.17, it is easy to see that we need to look at

D
i,a
, for
i = 1, 2, . . . , b, otherwise the last two strings can not be changed to the shape with a
U turn as shown in gure 2.18.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 25
(2.19)
Now we look at the D
i,a
where i > 1 as in gure 2.19. Since we picked where
the U turn appears already, one can nd a candidate for a stem as thick and shaded
(green in color) line from the right hand side but we can not nish because the purple
string can not be join to the bottom clasp without being zero(it will force to have a
generator caps o). So only nonzero admissible stems should be obtained from

D
1,a
.
We split the rectangle(only one in the middle) vertically(horizontal splitting vanishes
immediately) and it creates another rectangle at right top side of previous place. We
have to split vertically except in the last step, for this rectangle, as in the gure 2.20,
both splits do not vanish. The vertical split gives us the rst shape gure 2.17 and
the horizontal split gives the third shape in gure 2.17.
(2.20)
A similar argument works for the second one in gure 2.17. The third gure
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 26
in 2.17 is a little subtle. First one can see that none of the

D
i,j
work if either i > 2 or
j < a 1. Thus, we only need to check

D
1,a1
,

D
1,a
,

D
2,a1
and

D
2,a
but we already
know about

D
1,a
,

D
2,a
. The following gure 2.21 shows the nonzero admissible stem
for

D
1,a1
. As usual, we draw a stem as a union of thick and purple lines.
(2.21)
The following gure 2.22 shows the nonzero admissible stem for

D
2,a1
.
(2.22)
Note that the last gure has one loop which contributes [2]. This completes the
proof of lemma. 2
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 27
Corollary 2.13
a
a
1 =
[a + 3]
[a + 1]
a
a
a b
a
b
1 =
[b + 2][a + b + 3]
[b + 1][a + b + 2]
a b
a
b
a
=
[a + 2][a + 1]
[2]
a
b
=
[a + 1][b + 1][a + b + 2]
[2]
(2.23)
Proof: After using a double clasps expansion one can get the rst two equalities
with a simple calculation. The next two follow from the previous two by induction.
2
We will apply theorem 2.8 to derive the coecients in equations 2.24 and 2.26.
The expansion in the proposition 2.24 is known [Kup96], which is only previously
known expansion formula for a segregated clasp of weight (a, b) and it was used to
nd quantum su(3) invariants in [OY]. Our proof using single clasp expansion will
be used for trihedron coecients.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 28
a
b
a b
=
Min(a,b)

k=0
a
k
a k b k
k
a b
a
b
(2.24)
Proposition 2.14 The coecients in equation 2.24 is
a
k
= (1)
k
[a]![b]![a + b k + 1]!
[a k]![b k]![k]![a + b + 1]!
.
Proof: Let me denote that a basis web in the right side of equation 2.24 by D(k)
which corresponding to the coecient a
k
. We induct on a +b. It is clear for a = 0 or
b = 0. If a = 0 = b then we use a segregated single clasp expansion of weight (a, b) in
the middle. Even if we do not use entire single clasp expansion of segregated clasp,
once we attach (a, 0), (0, b) clasps on the top, there is only two surviving web which
are one with one U turn. One of resulting webs has some Hs as in Figure 2.25 but
if we push them down to (a, b 1) clasp, it becomes a non-segregated clasp.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 29
a
b
a b
=
a b
a b 1
a
b

[a]
[a+b+1]
a 1
b 1
. . .
a b
a
b
=
a b
a b 1
a
b

[a][a+1]
[a+b+1][a+b]
a 1 b 1
a b
a
b
(2.25)
We can nd the coecient using the same argument using stems and it is
[2]a
1,ab
+ a
a,ab+1
+
b

i=2
(a
i,ab+i2
[2]a
i,ab+i1
+ a
i,ab+i
) =
[a]
[a + b + 1]
because a
i,ab+i2
[2]a
i,ab+i1
+a
i,ab+i
= 0 for all i = 2, 3, . . . , b. Then we attach
some Hs to make the middle clasp as a non-segregated clasp of weight (a, b 1). By
using a non-segregated single clasp expansion for which clasps are located at northeast
corner and by the induction hypothesis, we have
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 30
=
b1

k=0
(1)
k
[a]![b 1]![a + b k]!
[a k]![b 1 k]![k]![a + b]!
D(k)

[a + 1][a]
[a + b + 1][a + b]
b1

k=0
(1)
k
[a 1]![b 1]![a + b 1 k]!
[a 1 k]![b 1 k]![k]![a + b a]!
D(k + 1)
= 1 D(0) +
b1

k=1
((1)
k
[a]![b 1]![a + b k]!
[a k]![b 1 k]![k]![a + b]!
+ (1)
k
[a + 1]![b 1]![a + b k]!
[a k]![b 1 k]![k 1]![a + b]!
)D(k)
(1)
b1
[a + 1][a]
[a + b + 1][a + b]
[a 1]![b 1]![a]!
[a b]![0]![b 1]![a + b 1]!
D(b)
= D(0) +
b1

k=1
(1)
k
[a]![b]![a + b + 1 k]!
[a k]![b k]![k]![a + b + 1]!
(
[b k][a + b + 1] + [k][a + 1]
[b][a + b + 1 k]
)D(k)
+ (1)
b
[a]![b 1]![a + 1]!
[a b]![0]![b 1]![a + b + 1]!
D(b)
=
b

k=0
(1)
k
[a]![b]![a + b + 1 k]!
[a k]![b k]![k]![a + b + 1]!
D(k)
2
For equation 2.26, we assume 0 a b.
a
b
a b
=
b

k=0
A(a, b, k)
k
k
b a
a
b
(2.26)
Proposition 2.15 The coecients A(a, b, k) in equation 2.26 satises the following
recurrence relation.
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 31
A(1, 1, 1) =
[2]
[3]
,
A(1, 1, 0) = 1,
A(a, a, k) =
[2a k + 1]
[2a + 1]
(A(a 1, a, k) + A(a 1, a, k 1)),
A(a, a + i, k) =
[2a + 1 +i k]
[2a + 1 +i]
A(a, a + i 1, k).
Proof: Note that we assume that A(a, b, i) = A(a, b, a+i) = 0 for all i > 0. Using
a non-segregated single clasp expansion at the clasp of weight (a, b), one standard
and one with clasp in the northeast corner, we have the result with two axioms of
clasps. Remark that these coecients are not round. 2
2.4 Single Clasp Expansion for U
q
(sp(4))
It is known [Kup96] that U
q
(sp(4)) webs are generated by a single web
with the relations
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 32
=
[6][2]
[3]
=
[6][5]
[3][2]
= 0
= [2]
2
= 0
= (2.27)
Also it is known [Kup96] that we can dene tetravalent vertex to achieve the same
end as in equation 2.28. First we will nd a single clasp expansion of clasps of weight
(n, 0) and (0, n) and then use it to nd coecients of double clasps expansion of clasps
of weight (n, 0) and (0, n). Remark that the cut weight is dened slightly dierent
way. A cut path may cut diagonally through a tetravalent vertex, and its weight is
dened as n
1
+(k +k

)
2
, where n is the number of type 1, single strands, that it
cuts, k is the number of type 2, double strands, that it cuts, and k

is the number
of tetravalent vertices that it bisects. And there is a natural partial ordering of the
B
2
weight lattice given by
a
1
+ b
2
(a 2)
1
+ (b + 1)
2
a
1
+ b
2
(a + 2)
1
+ (b 2)
2
.
We will use the following shapes to nd a single clasp expansion because there is
an ambiguity of preferred direction in the last relation 2.27. We remark that the let
side of the second equality of 2.28 is not a crossing but a vertex where four double
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 33
edges meet.
= + , = (2.28)
By combining with the weight diagram of V
n

1
and minimal cut paths, we can
nd single clasp expansion of U
q
(sp(4)) of type (n, 0) as in equation 2.29 which has
n
2
1
2
unknowns.
n
n
=
n1

i=0
n

j=i+1
a
ij
n 1
j
i
(2.29)
Since adding a U turn and a Y make the left side vanish, we will have n(n 1)
equations.
Theorem 2.16 The coecients in Figure 2.29 are
a
i,j
= [2]
ij+1
[n + 1][n j + 1][2n 2i + 2]
[n][2n + 2][n i + 1]
.
Proof: To proceed the proof of the theorem, we remark that the following useful
relations in equation 2.30 can be easily obtained from the relation in equation 2.27.
=
[6][2]
[3]
, = [2]
2
= [2]
2
[2][4]
= [2]
2
+ [2]
2
(2.30)
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 34
Using these relations, we get the following n1 equations by adding U turns from
the left to right. By capping o the generator from left to right, we have (n 1)
2
equations. There are two special equations and four dierent shapes of equation as
follows.
a
n2,n1
+
[2][6]
[3]
a
n2,n

[2][6]
[3]
a
n1,n
= 0

[2][6]
[3]
a
12
+
[2][6]
[3]
a
13
+ a
23
+ 1 +
[2][6]
[3]
b
2
[2][4]b
3
= 0
Type I : For i = 1, 2, . . . , n 3,
a
i,i+1
+
[2][6]
[3]
a
i,i+2
[2][4]a
i,i+3

[2][6]
[3]
a
i+1,i+2
+
[2][6]
[3]
a
i+1,i+3
+ a
i+2,i+3
= 0.
Type II : For i = 0, 1, . . . , n 2,
a
i,n1
[2]
2
a
i,n
= 0.
Type III : For i = 0, 1, 2, . . . , n 3, k = 2, 3, . . . , n i 1,
a
i,nk
[2]
2
a
i,nk+1
+ [2]
2
a
i,nk+2
= 0.
Type IV : For i = 3, 4, . . . , n, k = n i + 3, n i + 4, . . . , n,
[2]
2
a
nk,i
[2]
2
a
nk+1,i
+ a
nk+2,i
= 0.
Then we check the answer in the proposition satises the equations. Since these
webs in equation 2.29 form a basis, the coecients are unique. Therefore, it completes
the proof. 2
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 35
n
n
=
n 1
n 1
+ a
12
n 1
n 2
n 1
+ a
02
n 1
n 1
(2.31)
Corollary 2.17 The double clasp expansion of B
2
of type (n, 0) can be obtained as
in the equation 2.31 where a
12
, a
02
are from the Theorem 2.16.
n
n
=
n1

i=0
n

j=i+1
a
ij
n 1
j
i
(2.32)
Then we look for (0, n) case. The main idea for (n, 0) works exactly same except
we replace the base as in the equation 2.32. By capping o U turns and a lower
weight cap, we get the following coecients and we can solve them successively as in
Theorem 2.18. Also the equation 2.33 is useful to nd the following equations.
= [5] , = [2]
2
[5]
= [2][4] [2]
2
[3]
= [2][4] + [2]
4
[3] (2.33)
a
n2,n1
[5][2]
2
a
n2,n
+
[6][5]
[3][2]
a
n1,n
= 0
[3][2]
2
a
n2,n
+ [5]a
n1,n
= 0
CHAPTER 2. Single Clasp Expansions for Rank 2 Lie Algebras 36
Type I : For i = 0, 1, . . . , n 3,
a
i,i+1
[5][2]
2
a
i,i+2
+ [3][2]
4
a
i,i+3
+
[6][5]
[3][2]
a
i+1,i+2
[5][2]
2
a
i+1,i+3
+ a
i+2,i+3
= 0
Type II : For i = 0, 1, . . . , n 2,
a
i,n1
[4][2]a
i,n
= 0
Type III : For i = 0, 1, . . . , n 3 and j = i + 1, i + 2, . . . , n 2,
a
i,j
[4][2]a
i,j+1
+ [2]
4
a
i,j+2
= 0
Type IV : For i = 0, 1, . . . , n 3 and j = i + 3, i + 4, . . . , n,
[2]
4
a
i,j
[4][2]a
i+1,j
+ a
i+2,j
= 0
Type V : For i = 1, 2, . . . , n 2
[3][2]
2
a
i1,i+1
+ [2]
4
a
i1,i+2
+ [5]a
i,i+1
[3][2]
2
a
i,i+2
= 0.
Theorem 2.18 For n 2,
a
i,j
= [2]
2(1+ij)
[2n + 1 2i][2n 2j + 2]
[2n][2n + 1]
.
n
n
=
n 1
n 1
+ a
12
n 1
n 2
n 1
+ a
02
n 1
n 1
(2.34)
Corollary 2.19 The double clasp expansion of B
2
of type (0, n) can be obtained as
in the equation 2.34 where a
12
, a
02
are from the Theorem 2.18.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 37
Chapter 3
Trihedron Coecients for
U
q
(sl(3, C))
By [Lic72] [MV94] [Tur94], we dene a trivalent vertex as follow. A triple integers
(a, b, c) is admissible if a + b + c is even and |a b| c a + b. This is equivalent
to the following. For sl(2, C), dim(Inv(V
a
V
b
V
c
)) is 1 if (a, b, c) is an admissible
triple or 0 otherwise, where V
a
is an irreducible representation of highest weight a.
Given an admissible triple, we dene a trivalent vertex
a
b
c
= i
j
k
a
b
c
where i = (a + b c)/2, j = (a + c b)/2 and (b + c a)/2. Then the trihedron
coecient for U
q
(sl(2, C)) is known [Lic72] [MV94] [Tur94] as
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 38
a
b
c
=
j
j
i k
= (1)
i+j+k
[i + j + k + 1]![i]![j]![k]!
[i + j]![j + k]![i + k]!
.
In previous chapter we found a recursive formula for generalized Jones-Wenzl pro-
jectors. So we study how we generalize trihedron coecients to U
q
(sl(3, C)). But,
the denition of the trivalent vertex is a little subtle. We will prove the follow-
ing statement in lemma 3.3. Let
1
,
2
be the fundamental dominant weights of
sl(3, C) (mainly we will use sl(3, C)-modules because it is known that representa-
tion theories of sl(3, C) and U
q
(sl(3, C)) are parallel: see theorem 1.1 and the rep-
resentation theory of sl(3, C) is well known in [Hum72]). Let V
a
1
+b
2
be an irre-
ducible representation of sl(3, C) of highest weight a
1
+b
2
. Each edge is decorated
by an irreducible representation of sl(3, C), denoted by V
a
1

1
+b
1

2
, V
a
2

1
+b
2

2
and
V
a
3

1
+b
3

2
where a
i
, b
j
are nonnegative integers. Let d = Min {a
1
, a
2
, a
3
, b
1
, b
2
, b
3
}.
Then dim(Inv(V
a
1

1
+b
1

2
V
a
2

1
+b
2

2
V
a
3

1
+b
3

2
)) is d +1 if there exist non negative
integers i, j, k, l, m, n, o, p, q such that a
1
= i + p, a
2
= j + n, a
3
= k + l, b
1
= j + o,
b
2
= k + m, b
3
= i + q and 0 = le

3
i
+me

6
i
+n+ oe

6
i
+pe

3
i
q. Otherwise, it is
zero.
If dim(Inv(V
a
1

1
+b
1

2
V
a
2

1
+b
2

2
V
a
3

1
+b
3

2
)) is nonzero, we say a triple of ordered
pairs ((a
1
, b
1
), (a
2
, b
2
), (a
3
, b
3
)) is admissible. It has shown that for a xed boundary,
there are llings which are cut outs from the the hexagonal tiling of the plane [Kup96].
A general trihedron shape is given in the following gure where + = a
1
, i+j +k =
b
1
, k +l = a
2
, +j +m = b
2
, i +m = a
3
and +j +l = b
3
. The top and the bottom
part are actually the same after some modications which we will discuss later.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 39

i
j j
k
l
m
(3.1)
So we can write trihedron coecients as a (d+1)(d+1) matrix. Let us denote it
by M

(a
1
, b
1
, a
2
, b
2
, a
3
, b
3
) or M

() where = (a
1
, b
1
, a
2
, b
2
, a
3
, b
3
). Also we denotes
its (i, j) entry by
i,j
(a
1
, b
1
, a
2
, b
2
, a
3
, b
3
) or
i,j
(). It is obvious that M

() is
symmetric. Unfortunately the trihedron coecient of this shape is no longer rational
expression composed of monomials of quantum integers (if so, we say it to be round)
in a simple case ((1, 1), (1, 1), (1, 1)). So we start to look the case a
1
= 0. Then we
have found the trihedron coecients for the case = = 0 and either k = 0 or j = 0
from the general shape.
Theorem 3.1 M

(0, i + j, l, j + m, i + m, j + l) is
(1)
j
[j + l + 1][i + j + l + m+ 2]
[i + j + m+ 1]![i]![j]![m]!
[i + j]![j + m]![i + m]![2]
.
Theorem 3.2 M

(0, i + k, k + l, m, i + m, l) is
min{l,k,m}

n=0
a
n
[i + l + 1][i + l + m+ 2]
[i + n + 1][i + n + m+ 2]
[i + k + m+ 2]![k n + 1]![i + m]![m + 1]!
[i + k]![i + m+ n]![k + mn + 1]![2]
.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 40
where
a
n
= (1)
n
[k]
2n
[k + l n]![m]![k + m + l n + 1]!
[k + l]![mn]![k + m + l + 1]!
.
In section 2, we show all possible trihedron shapes and some properties of trihedron
shapes. In section 3, we prove the main theorems.
3.1 General Shapes
Now we will look at the trihedron coecients. The general shape is given in g-
ure (3.1) where the weight of clasps are a
1

1
+b
1

2
, a
2

1
+b
2

2
and a
3

1
+b
3

2
and
a
i
, b
j
are nonnegative integers.
For sl(2, C), dim(Inv(V
i
V
j
V
k
)) is 1 if (i, j, k) is an admissible triple or 0
otherwise, where V
i
be an irreducible representation of highest weight i. So there is
a unique way to ll in the triangle. But for sl(3, C), we could have more than one
ways. Thus the shape of the polygon that we are lling in might vary depending on
the weight of clasps. First we will discuss how we nd a general shape in gure (3.1).
Instead of nding a cut out from the hexagonal tiling, we nd a way to put three
clasps into the hexagonal tiling. Since it bounds a polygon and we have a set of
restrictions how clasps can be bent, we can change this problem to an elementary
geometry problem. Since all clasps are segregated, until we reverse the direction of
arrows, all possible interior angles are 60, 180 or 300. Since an Y makes the web zero,
we can exclude 60. We would not count 180 because it can be seen as a subdivision.
Thus it has to be 300 if we do not change the direction of arrows. When we change
the direction of arrows, there are also three possible interior angle either 0, 120, 240
or 360, let us denote this angle by
i
. Since U turn makes the web zero, we can also
exclude 0. Let us denote the angle between clasps by
i
. If there is a cut out bounded
by three given clasps of which all a
i
, b
j
are nonzero, we can see the following equality
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 41
from the sum of interior angles.
180(n 2) = 300(n 6) +
3

i
+
i
.
And we can simplify it to have 1440120n =

i
+
i
. Since
i
is either 120 or 240,
n can be either 6, 7 or 8. For each n, we look at the all possible combinations of
i
,
j
up to symmetries. Then we check whether it actually bounds a polygon. For example,
n = 6 there are 6 possible combinations of angles but one does not bound a polygon,
(
i
) = (0, 120, 240), (
i
) = (120, 120, 120). For n > 6, one has to use n 6 times
of 300 angles. If some of a
i
, b
j
are zero then we can play the same game to nd all
shapes, we might have 60 for some
j
. The polygon in the middle might be a triangle,
a rectangle or a pentagon but we will consider them as a special case of a hexagon.
But we can obtain all these possible shapes from the general shape by substituting
some zeros. Then we prove the following lemma. Let d = Min {a
1
, a
2
, a
3
, b
1
, b
2
, b
3
}.
Lemma 3.3 dim(Inv(V
a
1

1
+b
1

2
V
a
2

1
+b
2

2
V
a
3

1
+b
3

2
)) is d + 1 if there exist non
negative integers i, j, k, l, m, n, o, p, q such that a
1
= i + p, a
2
= j + n, a
3
= k + l,
b
1
= j + o, b
2
= k + m, b
3
= i + q and 0 = le

3
i
+me

6
i
+n+ oe

6
i
+pe

3
i
q.
Otherwise, it is zero.
Proof: One might directly nd the answer by using the decomposition of the tensor
product V
a
1

1
+b
1

2
V
a
2

1
+b
2

2
V
a
3

1
+b
3

2
into irreducible representations. But one
can see that these are conditions we can easily obtain from the gure (3.1) and the
last one is one condition that the hexagon in the middle does exist. 2
Since the parallelogram in gure (3.2) changes the directions of edges (the interior
is lled by the unique maximal hexagonal cut out), we can push the hexagon to
an equilateral triangle (possibly empty). The size of this equilateral triangle is the
minimum of dierences of lengths of three pairs of parallel edges of the hexagon.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 42
, (3.2)
But the resulting shape has mixed aspects. First it might contain some non-
segregated clasps which did not exist before. When we apply single or double clasp
expansions to these non segregated clasps, it is possible to get one by transforming
the basis by adding Hs but usually it becomes very dicult to deal with. If we keep
the hexagonal shape, it usually produces multiple non-vanishing terms in a single
clasp expansion. So we will use both shapes depend on the feasibility.
3.2 Proofs of Theorems
Since M

() is an 1 1 matrix, we will write it as a scalar through this section. For

nontrivial planar shapes with d = 0, we could have at most three zeros for a
i
, b
j
.
Before we prove the main theorems we start with cases of two zeros.
3.2.1 Two Zeros
If we have two zeros, up to symmetries, it is either one of these subcases: 1) a
1
=
b
3
= 0, 2) a
1
= a
3
= 0.
If a
1
= b
3
= 0, there exists j such that b
1
= a
2
+ j, a
3
= b
2
+ j and its shape is
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 43
b
1
a
3
j
a
2
j
b
2
(3.3)
By the same idea of the proof of proposition 2.14, we can easily get the following
proposition 3.4. Let i = a
2
and k = b
2
.
Proposition 3.4 M

(0, i + j, i, k, j + k, 0) is
[i + j + k + 2]![i + 1]![j]![k + 1]!
[i + j]![j + k]![i + k + 1]![2]
.
Proof: Let us use the notation [i, j, k] for the trihedron coecient of this shape.
We induct on k. If k = 0,
[i, j, 0] =
[i + j + 2][i + j + 1]
[2]
.
For k = 0, by the idea of the proof of proposition 2.14 and the induction hypothesis,
we have
[i, j, k] =
[j + k + 2]
[j + k]
[i.j, k 1]
[i + 1][i]
[i + k + 1][i + k]
[i 1, j + 1, k 1]
=
[i + j + k + 2]![i + 1]![j]![k + 1]!
[i + j]![i + k + 1]![j + k]!
(
[j + k + 2][i + k + 1] [j + 1][i]
[i + j + k + 2][k + 1]
)
=
[i + j + k + 2]![i + 1]![j]![k + 1]!
[i + j]![i + k + 1]![j + k]!
because [j + 1 +k + 1][i + k + 1] = [j + 1][i] + [i + j + k + 2][k + 1]. 2
If a
1
= a
3
= 0, there exists k such that b
1
= b
2
+k, a
2
= b
3
b
2
+k and its shape
is
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 44
= (3.4)
We need to prove a sequence of lemmas. First, we prove that the clasp in the
middle is not essential.
Lemma 3.5 Let n 1, then
n n
n
=
n n
n
=
n
n n
(3.5)
Proof: The idea of the proof is that if we have any Y s in the single clasp expansion
at the middle clasp, it becomes zero. The argument, we used to nd the general
shape, leads us that there does not exist a lling with boundary (0, n, 0, n, 1, n 2).
Thus, it has to vanish once we have any Y s. 2
Form the third gure in equation (3.5), we apply a single clasp expansion to the
clasp in the left. For the following equation, it should be clear without the direction
of edges.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 45
Lemma 3.6 Let n 1, then
n
n
= (1)
n
[n + 1]
n
(3.6)
Proof: We induct on n. If n = 1, the coecient is [2] = (1)
1
[2]. We apply a
single clasp expansion at the left clasp which gives us the rst equality in the following
equation (3.8).
n
=
n 1
1
+
[n1]
[n]
n 2
(3.7)
=
n 1
1
+
[n1]
[n] n 2 2
+
[n2]
[n] n 2 2
=
n 1 1
+
[n1]
[n] n 2 2
+
[n2]
[n]
n 2
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 46
where = (1)
n1
[n] and = (1)
n2
[n 1].
For the second equality, we apply a single clasp expansion for the resulting clasp
of weight (n 1)
2
for which clasps of weight (n 2)
2
are located at the northeast
corner. It is not dicult to see the following equality. The number of strings coming
from the trapezoid can be any integer between 0 and n where n is the weight of the
clasp given the left side of equality.
= .
Thus, we can put a clasp of weight (n1)
2
((n2)
2
, (n2)
2
) at the gap between
an equilateral triangle and a trapezoid at the rst (second and third, respectively)
gure in the second line of equation (3.8). Therefore, we can use induction to get the
third equality in equation (3.8). Note that the size of the third equilateral triangle in
the third line is n 2.
Last step is to count how many [2]s will be produced when we change it to
multiple of the web in the right hand side of equation (3.6). But it is fairly easy to
see that each of them has just one factor of [2] in the rst two in the third line. If
we add up the all coecients, we have
[2](1)
n1
[n] [2](1)
n2
[n1][n1]
[n]
+ (1)
n2
[n1][n2]
[n]
= (1)
n 1
[n]
([2][n]
2
[2][n 1]
2
+ [n 2][n 1])
= (1)
n 1
[n]
[n + 1][n] = (1)
n
[n + 1]
2
The lemma 3.6 can be generalized to the following lemma.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 47
Lemma 3.7
i
j
= (1)
j
[i + j + 1]
[i + 1]
(3.8)
Proof: We induct on i +j. If i = 1, j = 0, the coecient is 1 = (1)
1
[2]
[2]
. If i = 0, it
follows from the previous lemma. The rst equality can be proven with a single clasp
expansion at the left middle clasp in left gure. The second equality can be proven
the same argument we use in lemma. The rest of proof follows by induction. 2
Lemma 3.8
i
j
k
= (1)
j+1
[i + j + k + 1]![i]![j]![k]!
[i + j]![j + k]![i + k + 1]!
(3.9)
Proof: We induct on k. If k = 0, it follows from the previous lemma. If k = 0,
we use a single clasp expansion at the left middle clasp. Then we get the following
equality.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 48
By induction, the coecient is equal to
(1)
j
[i + j + k]![i]![j]![k 1]!
[i + j]![i + k]![j + k 1]!
+
[j][j]
[j + k][j + k 1]
(1)
j1
[i + j + k]![i + 1]![j 1]![k 1]!
[i + j]![i + j + 1]![j + k 2]!
= (1)
j
[i + j + k + 1]![i]![j]![k]!
[i + j]![j + k]![i + k + 1]!
(
[i + k + 1][j + k] [i + 1][j]
[i + j + k + 1][k]
)
= (1)
j
[i + j + k + 1]![i]![j]![k]!
[i + j]![j + k]![i + k + 1]!
because [i + 1 +k][j + k] = [i + 1][j] + [i + j + k + 1][k]. 2
Proposition 3.9 M

(0, b
1
, a
2
, b
2
, 0, b
3
) is
(1)
b
2
+1
[b
1
b
2
+ b
3
+ 2]![b
1
b
2
]![b
3
b
2
]![b
2
+ 1]!
[b
1
]![b
1
2b
2
+ b
3
]![b
3
]![2]
.
or M

(0, i + j, i + k, j, 0, j + k) is
(1)
j
[i + j + k + 2]![i]![j + 1]![k]!
[i + j]![j + k]![i + k]![2]
.
Corollary 3.10 M

(0, n, 0, n, 0, n) is
(1)
n
[n + 1]
2
[n + 2]
[2]
.
3.2.2 Proof of Theorem 3.1
We start to prove the following lemma to nd the trihedron coecient.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 49
Lemma 3.11
i
j
l
m
=
i
j
l
m
(3.10)
Proof: We use an expansion in equation (2.24) at the right middle clasp. Then
there are min{l, m + j} terms in the expansion. But once we have a U turn, we will
show that it becomes zero. If there is a U turn we use a single clasp expansion at
the top-left clasp of weight m + j. Then all terms vanish except one term which
has Y joining the top right clasp of weight (i + l, j + m) and the triangle in the top
center. Then, there is a sequence of Hs we can push that move the entire shape
by one string. Eventually Y has to join two strings from the triangle but we knew
from lemma 3.12 that it becomes zero. Therefore, we can free l strings from the right
middle clasp of weight (l, j + m). 2
Unfortunately this lemma 3.11 is not true if k = 0. Actually only two terms
survive but there ia a layer of Hs which makes the problem dicult in this approach.
Continuing the prrof the theorem, lemma 3.11 implies that M

(0, i + j, l, j + m, i +
m, j + l) is
[j + l + 1][i + j + l + m + 2]
[j + 1][i + j + m+ 2]
M

(0, i + j, 0, j + m, i + m, j).
Then the result follows by proposition 3.9.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 50
3.2.3 Proof of Theorem 3.2
We use the same idea of lemma 3.5 but for j = 0, all terms in this expansion do not
vanish. For next step we need to show the following lemma.
Lemma 3.12 Let = 0 if n > min{l, k}, (
[k]
n
[l+kn]!
[l+k]!
)
2
if n min{l, k}. Then
i
k
m
l
n
mn
=
i + n
k n
m
(3.11)
Proof: The clasp of weight m
2
can be pushed into the clasp of weight l
1
+
(i + m)
2
. For the clasp of weight (k + n)
1
, we use a single clasp expansion. 2
First we use the equation (2.24) at the middle clasp of weight (k +l)
1
+m
2
. By
lemma 3.12 we can transform each web to the web in the righthand side of lemma 3.12.
CHAPTER 3. Trihedron coecients for U
q
(sl(3, C)) 51
i
k
m
l
=
min{l+k,m}

n=0
a
n
i
k
m
l
n
mn
(3.12)
Since its shape can be written as [k n, i +n, m], by proposition 3.4 it has value

[i + k + m+ 2]![k n + 1]![i + m]![m + 1]!

[i + k]![i + m + n]![k + mn + 1]![2]
.
Therefore, it completes the proof.
CHAPTER 4. A Set of Complete Relations of U
q
(sl(4, C)) 52
Chapter 4
A Complete Set of Relations of
U
q
(sl(4, C))
Our webs are generated by the two shapes of trivalent vertices.
,
And the following is our conjecture for a complete set of relations for U
q
(sl(4, C)).
CHAPTER 4. A Set of Complete Relations of U
q
(sl(4, C)) 53
=
_
_
4
1
_
_
(4.1)
=
_
_
4
2
_
_
(4.2)
= [2] (4.3)
= [3] (4.4)
= (4.5)
= (4.6)
= [2] + (4.7)
= + (4.8)
= (4.9)
= + (4.10)
CHAPTER 4. A Set of Complete Relations of U
q
(sl(4, C)) 54
First of all, we compute the dimension of invariant space of all tensor products
of 4, 6 fundamental representation of U
q
(sl(4, C)). There is a general way to nd a
basis webs with a xed boundary, all of them are fundamental representation, for
U
q
(sl(2, C)) and U
q
(sl(3, C)) [KK99]. It is still unknown how we can actually nd
all basis webs of all possible boundaries of fundamental representations. But for our
case, there are only few boundaries and we can nd the dimension and even nd a
basis webs without diculty, most of basis webs do not have any faces.
Since all webs are generated by two trivalent vertices, by multiplying a complex
number, we can have a dierent set of generators. Therefore, we have two choices of
freedom to set any two independent coecients. Let a, b, c, d, e and f be unknowns in
equation 4.4, 4.6, 4.7 and 4.8. By the quantum Weyl formula, we do know the value
of the rst two equations. We use the rst choice of freedom to have the equation 4.3.
The following equality implies a = [3].
= a = [4]a
= [2] = [2]
[4][3]
[2]
= [4][3]
We use the last choice of freedom to have equation 4.5. To get the equation 4.6
(which is actually the dual of the equation 4.5), we start from the following equations.
= +
Using 4.5 and 4.6 (with an unknown variable a), we found = 0 and = [2]b. By
attaching a U turn on the top of each webs in the equation, we get [2][3] = [4]+[3].
Thus b = 1. For equation 4.7, we attach U turns on the top and right side of
CHAPTER 4. A Set of Complete Relations of U
q
(sl(4, C)) 55
each webs in the equation. Then the resulting web can be expanded as a linear
combination of a basis webs of dierent boundary. By comparing the coecients,
we get [3][3] = [4]c + d, [2][3] = c + [4]d. It is easy to nd that c = [2], d = 1. For
equation 4.8 we attach
(4.11)
to right side of each webs to get e = f = 1.
For last two equations 4.9 and 4.10, we need to start from the following equations.
= g + h + i (4.12)
= j + k + l
+ m + n + p (4.13)
By attaching U turns and H (as in 4.11) for equation 4.9, we get
[2][3] = [2][3]g + h +
[4][3]
[2]
i
[2][3] = [2][3]g +
[4][3]
[2]
h + i
[2] = [2]g + [3]h
[2] = [2]g + i
CHAPTER 4. A Set of Complete Relations of U
q
(sl(4, C)) 56
One can solve them to have g = 1 and h = i = 0. For the equation 4.10, we just
need to attach H (as in 4.11) to right top side of each basis webs in the equation 4.13.
Then we follow the same procedure to get the following six equations: j = 1, [2]
2
j
+[3]k = 1, [2]j +[3]p = [2], l = 1, m = 0 and n = 0.
Bibliography
[CKT97] Marc Rosso Christian Kassel and Vladimir Turaev, Quantum groups and
knot invariants, Panoramas et Syntheses, vol. 5, Societe Mathematique de
France, 1997.
[Hum72] James E. Humphreys, Introduction to Lie algebras and representation the-
ory, Graduate Texts in Mathematics, vol. 9, Springer-Verlag, New York-
Heidelberg-Berlin, 1972.
[Jon83] V. F. R. Jones, Index of subfactors, Invent. Math. 72 (1983), 125.
[Jon85] V. F. R. Jones, Polynomial invariants of knots via von neumann algebras,
Bull. Amer. Math. Soc. 12 (1985), 103111.
[Jon87] V. F. R. Jones, Hecke algebra representations of braid groups and link poly-
nomials, Ann. of Math. 126 (1987), 335388.
[Kas95] Christian Kassel, Quantum groups, Graduate Texts in Mathematics, vol.
155, Springer, New York-Heidelberg-Berlin, 1995.
[Kho97] Mikhail Khovanov, Canonical bases in tensor products and graphical calculus
for u
q
(sl
2
), Duke Math. Journal 87:3 (1997), 409480.
[KK99] Mikhail Khovanov and Greg Kuperberg, Web bases for sl(3) are not dual
canonical, Pacic J. Math. 188:1 (1999), 129153, arXiv:q-alg/9712046.
57
CHAPTER 4. A Set of Complete Relations of U
q
(sl(4, C)) 58
[Kup96] Greg Kuperberg, Spiders for rank 2 Lie algebras, Comm. Math. Phys. 180:1
(1996), 109151, arXiv:q-alg/9712003.
[Lic72] W. B. Raymond Lickorish, An introduction to knot theory, Graduate Texts
in Mathematics, vol. 175, Springer-Verlag, New York-Heidelberg-Berlin,
1972.
[MV94] Gregor Masbaum and Pierre Vogel, 3-valent graphs and the Kauman
bracket, Pacic J. Math. 164 (1994), 361381.
[OY] Tomotada Ohtsuki and Shinji Yamada, Quantum su(3) invariants via linear
skein theory, J. of Knot Theory and its Ramications 6:3, 373404.
[RT90] Nicolai Yu. Reshetikhin and Vladimir G. Turaev, Ribbob graphs and their
invariants derived from quantum groups, Comm. Math. Phys. 127 (1990),
126.
[RT91] Nicolai Yu. Reshetikhin and Vladimir G. Turaev, Invariants of 3-manifolds
via link polynomials and quantum groups, Invent. Math. 103 (1991), 547
597.
[Tur94] Vladimir G. Turaev, Quantum invariants of knots and 3-manifolds, W. de
Gruyter, Berlin, 1994.
[Wen87] H. Wenzl, On sequences of projections, C. R. Math. Rep. Acad. Sci. Canada
(1987), 59.