, , ,

Representations of Artin's braid groups

ℬ_{n}

are of great importance to mathematicians [B] , and physicists recently [W] . Certain representations of the braid groups have been proposed as the fractional statistics of anyons [W] , and used in the topological models for quantum computing [FKLW] . Therefore it is interesting to identify the images of such braid group representations.

In this paper we analyze a particular representation of the braid groups afforded by a unitary solution of the braid relation, i.e. a flipped R-matrix

R = \frac{1}{\sqrt{2}} (\begin{matrix} 1 & 0 & 0 & 1 \\ 0 & 1 & - 1 & 0 \\ 0 & 1 & 1 & 0 \\ - 1 & 0 & 0 & 1 \end{matrix})

that satisfies the Yang-Baxter equation

(YBE) $(R \otimes I_{2}) (I_{2} \otimes R) (R \otimes I_{2}) = (I_{2} \otimes R) (R \otimes I_{2}) (I_{2} \otimes R)$

where

I_{2}

is the

2 \times 2

identity matrix. All solutions to the YBE of the form

R : V \otimes V \to V \otimes V

with

V

2-dimensional have been listed in [H] . Dye [D] found all unitary solutions of this form to the braid relations based on this list. The importance of this particular braid operator

R

was pointed out in the work of Kauffman and Lomonaco [KL] , and the connection of

R

with quantum computing was explored there which is another reason for our interest.

As is well-known, any (invertible) matrix satisfying the YBE gives rise to representations of

ℬ_{n}

for any

n

. The representation

(π_{n}, (C^{2})^{\otimes n})

corresponding to the matrix

R

above is unitary and defined as follows:

π_{n} (σ_{i}) = I_{2}^{\otimes i - 1} \otimes R \otimes I_{2}^{\otimes n - i - 1},

where

σ_{i}

is the

i

-th braid generator. The images of the braid groups

ℬ_{n}

under this representation are finite groups, and the image matrix of each braid generator

σ_{i}

has only two distinct eigenvalues. It follows that the image group of an irreducible constituent of

π_{n}

is generated by the conjugacy class of a braid generator with two distinct eigenvalues whose ratio is not -1, i.e. has the so-called 2-eigenvalue property defined in [FLW] .

Such representations are completely classified [FLW] , so in principle the image groups of the irreducible constituents of

π_{n}

can be identified by using the complete list in [FLW] Theorem 1.6. But as we will see that

π_{n}

is reducible, hence first we need to find the irreducible constituents of

π_{n}

; then we need to distinguish a few different cases in the complete list for the images of the irreducible constituents, so instead we choose to solve the problem in an elementary and self-contained way. We decompose these representations

π_{n}

(for all

n

) into their irreducible constituents and describe the images of

ℬ_{n}

under

π_{n}

as abstract groups. We find that the images of the pure braid groups are (nearly) extra-special 2-groups

E_{n - 1}^{- 1}

. The images of the full braid groups

ℬ_{n}

are extensions of the (nearly) extra-special 2-groups

E_{n - 1}^{- 1}

by the symmetric groups

S_{n}

, and the restrictions of the representations

π_{n}

to the subgroup of pure braids are isotypic copies of the odd representations of

E_{n - 1}^{- 1}

As already discussed in [KL] we can define link invariants using the representations

π_{n}

. By observing that

π_{n}

is related to the Jones representation of the braid groups at the 4-th root of unity, we improve slightly some earlier results of Jones about the images of the Jones representation of the braid groups at the 4-th root of unity [J1] . As a consequence we point out that the resulting link invariants are essentially the Jones polynomial at a 4-th root of unity, hence really the Arf invariant of a link (see references in [J2] ). The slight improvement of Jones's result comes from two subtle points about the Jones representations. Firstly, in the Jones representation of the braid group, there is some freedom in choosing phases so it is convenient to state the results projectively, i.e. modulo scalars, while not losing any significance mathematically. We choose to work out the images in full generality (as opposed to projectively) as this is desirable in physics for the applications to the fractional statistics of quantum Hall fluid [R] . This changes the images of the pure braid groups from the elementary abelian groups

Z_{2}^{n - 1}

to the (nearly) extra-special 2-groups

E_{n - 1}^{1}

. Secondly, when the number of strands of the braid groups is even, there are two irreducible sectors of the Jones representation [J1] . Jones found the projective images for each sector, but we determine the images of the two sectors together. This brings up a subtlety about the centers of the (nearly) extraspecial 2-groups in those cases, which disappears when the two irreducible sectors are treated separately, and projectively.

These results lead to several questions for future research currently being worked out by the authors. What are the closed images of the braid groups under the representations afforded by the other

R

-matrices listed in [D] and what are the associated link invariants? What are the other extraspecial

p

-groups that appear as homomorphic images of the pure braid groups?

2 Preliminaries

2.1 Definitions and computations

Definition 2.1. Artin's braid group

ℬ_{n}

n

strands has presentation in generators

σ_{1}, \dots, σ_{n - 1}

satisfying relations:

(B1) $σ_{i} σ_{j} = σ_{j} σ_{i}$ if $| i - j | \geq 2$ .
(B2) $σ_{i} σ_{i + 1} σ_{i} = σ_{i + 1} σ_{i} σ_{i + 1}$ for $1 \leq i \leq n - 2$

For tensor products of matrices we use the convention “left into right,” that is, if

X = (\begin{matrix} w & x \\ y & z \end{matrix})

and

A = (\begin{matrix} a & b \\ c & d \end{matrix})

then

X \otimes A = (\begin{matrix} a X & b X \\ c X & d X \end{matrix})

Various matrices and quantities will be needed throughout, so we define them here:

(1) $I_{m}$ is the $m \times m$ identity matrix.
(2) $R = \frac{1}{\sqrt{2}} (\begin{matrix} 1 & 0 & 0 & 1 \\ 0 & 1 & - 1 & 0 \\ 0 & 1 & 1 & 0 \\ - 1 & 0 & 0 & 1 \end{matrix})$
(3) $s = (\begin{matrix} 0 & 1 \\ - 1 & 0 \end{matrix})$
(4) $σ_{x} = (\begin{matrix} 0 & 1 \\ 1 & 0 \end{matrix})$
(5) $P_{s} = (\begin{matrix} 1 & \sqrt{- 1} \\ \sqrt{- 1} & 1 \end{matrix})$
(6) $P_{σ_{x}} = (\begin{matrix} 1 & - 1 \\ 1 & 1 \end{matrix})$
(7) $P_{n} = (P_{s} \otimes P_{σ_{x}})^{\otimes ⌊ n / 2 ⌋} \otimes I_{2}^{\otimes (n - 2 ⌊ n / 2 ⌋)}$ where $⌊ a ⌋$ is the integer part of $a$ .
(8) $σ_{z} = (\begin{matrix} 1 & 0 \\ 0 & - 1 \end{matrix})$
(9) $g_{i} = π_{n} ((σ_{i})^{2}) = I_{2}^{\otimes (i - 1)} \otimes R^{2} \otimes I_{2}^{\otimes (n - i - 1)}$ (observe we ignore the dependence of $g_{i}$ on $n$ ; the value of $n$ will always be clear from the context).
(10) $ζ = \frac{1}{\sqrt{2}} (1 + \sqrt{- 1})$ .
(11) $d = (\begin{matrix} ζ & 0 \\ 0 & \bar{ζ} \end{matrix})$
(12) $D = (\begin{matrix} ζ & 0 & 0 & 0 \\ 0 & \bar{ζ} & 0 & 0 \\ 0 & 0 & \bar{ζ} & 0 \\ 0 & 0 & ζ \end{matrix})$
(13) $M = \frac{1}{\sqrt{2}} (\begin{matrix} 1 & 1 \\ - 1 & 1 \end{matrix})$

We will also need a few simple computations involving these matrices so we record them in the following:

Lemma 2.2. The matrices defined above satisfy:

(a) $R^{2} = s \otimes σ_{x}$
(b) $R = \frac{1}{\sqrt{2}} (R^{2} + I_{4})$ , $R^{- 1} = \frac{1}{\sqrt{2}} (R^{- 2} + I_{4})$
(c) $(R^{2} \otimes I_{2}) (I_{2} \otimes R^{2}) = - (I_{2} \otimes R^{2}) (R^{2} \otimes I_{2})$
(d) $g_{i} g_{i + 1} = - g_{i + 1} g_{i}$
(e) $(R^{- 1} \otimes I_{2}) (I_{2} \otimes R^{2}) (R \otimes I_{2}) = (I_{2} \otimes R^{2}) (R^{2} \otimes I_{2})$
(f ) $π_{n} (σ_{i}^{- 1}) g_{i \pm 1} π_{n} (σ_{i}) = g_{i \pm 1} g_{i}$
(g) $g_{i} g_{j} = g_{j} g_{i}$ and $π_{n} (σ_{i}) g_{j} = g_{j} π_{n} (σ_{i})$ if $| i - j | \geq 2$
(h) $R^{4} = - I_{4}$ , $(g_{i})^{2} = - I_{2^{n}}$ .
(i) $(P_{s})^{- 1} s P_{s} = \sqrt{- 1} σ_{z}$ , and $(P_{s})^{- 1} σ_{x} P_{s} = σ_{x}$
(j) $(P_{σ_{x}})^{- 1} σ_{x} P_{σ_{x}} = σ_{z}$ , $(P_{σ_{x}})^{- 1} s P_{σ_{x}} = s$
(k) $(P_{n})^{- 1} g_{2 i + 1} P_{n} = \sqrt{- 1} (I_{2}^{\otimes 2 i} \otimes σ_{z}^{\otimes 2} \otimes I_{2}^{\otimes n - 2 i - 2})$ , and $(P_{n})^{- 1} g_{2 i} P_{n} = g_{2 i}$ .

Proof. The first assertions (a) and (b) are straightforward computations.
Having checked that $s$ and $σ_{x}$ anti-commute (c) follows, and (d) is immediate from (c). Using (b) and the observation $R^{- 2} = - R^{2}$ we express the left-hand side of the equality in (e) in terms of $R^{2}$ and then use (c) to derive the right-hand side. Assertion (f ) is immediate from (e). Assertion (g) is a consequence of the “far commutation” relations satisfied by the braid group, and (h) follows from (b) and the definition of $g_{i}$ . The matrix $P_{s}$ (resp. $P_{σ_{x}}$ ) is a change of bases matrix that diagonalizes $s$ (resp. $σ_{x}$ ) and commutes with $σ_{x}$ (resp. $s$ ). This is the statement (j), and (k) follows directly from this fact and the definition of $P_{n}$ . □

2.2 Restriction to $P_{n}$

The homomorphism

ℬ_{n}

to the symmetric group on

n

letters

S_{n}

given by

σ_{i} \to (i, i + 1)

has kernel

P_{n}

the so-called pure braid group.

P_{n}

is generated by all conjugates of the squares of the generators of

ℬ_{n}

(σ_{i})^{2}

. Actually a more economical presentation of

P_{n}

can be found (see e.g. [B] ), but we shall not need it here. To exploit this relationship between

ℬ_{n}

and

P_{n}

we shall restrict

π_{n}

to the subgroup

P_{n}

. For convenience of notation we introduce the following notation:

Definition 2.3.

H_{n} : = π_{n} (P_{n})

and

G_{n} : = π_{n} (ℬ_{n})

We can describe

H_{n}

very succinctly:

Lemma 2.4.

H_{n}

is generated by

g_{1}, \dots, g_{n - 1}

Proof. Observe that $H_{n}$ is generated by all conjugates of $g_{i}$ , so that $H_{n}$ is the smallest normal subgroup of $G_{n}$ containing the subgroup $〈 g_{1}, \dots, g_{n - 1} 〉$ generated by the $g_{i}$ . But by Lemma 2.2 (f ),(g) $〈 g_{1}, \dots, g_{n - 1} 〉$ is normal in $G_{n}$ so $〈 g_{1}, \dots, g_{n - 1} 〉 = H_{n}$ . □

Remark 2.5. Combining this with Lemma 2.2 (a) we have a very powerful tool for studying the representation

π_{n}

ℬ_{n}

. After decomposing the representation

π_{n}

restricted to

P_{n}

into its irreducible components and computing the corresponding images of the

(σ_{i})^{2}

, we can immediately determine the decomposition of the images of the

σ_{i}

under

π_{n}

π_{n} (σ_{i}) = \frac{1}{\sqrt{2}} (g_{i} + I_{2^{n}})

Once we understand

H_{n}

as an abstract group and decompose its defining representation (as it is presented to us as a matrix group) we will need to consider the group

G_{n} / H_{n}

. We can immediately see that

G_{n} / H_{n}

is a homomorphic image of

S_{n}

π_{n}

induces a surjective homomorphism

{\bar{π}}_{n} : ℬ_{n} / P_{n} \to G_{n} / H_{n}

and

ℬ_{n} / P_{n} \sim = S_{n}

. We would like to know if

{\bar{π}}_{n}

is an isomorphism, so we must determine if

K e r ({\bar{π}}_{n})

is trivial. Observing that the kernel of

{\bar{π}}_{n}

is (isomorphic to) a normal subgroup of

S_{n}

we need only check that the kernel is not

S_{n}

A_{n}

or the normal subgroup of

S_{4}

isomorphic to

Z_{2} \times Z_{2}

. For

n \geq 4

it is sufficient to check that the element

(12) (34)

is not in the kernel, while for

n = 3

we should check that

(123)

is not in the kernel. Since

H_{n}

is a normal subgroup of

G_{n}

, we also have a homomorphism

ϑ : G_{n} \to A u t (H_{n})

where

A u t (H_{n})

is the automorphism group of

H_{n}

and

ϑ (s)

is conjugation by

s \in G_{n}

. Restricting to

H_{n}

we see that

ϑ (H_{n}) = I n n (H_{n}) \subset A u t (H_{n})

the normal subgroup of inner automorphisms of

H_{n}

and so we have the induced homomorphism

\bar{ϑ} : G_{n} / H_{n} \to A u t (H_{n}) / I n n (H_{n})

Since

K e r ({\bar{π}}_{n}) \subset K e r (\bar{ϑ} \circ {\bar{π}}_{n})

if we can show the composition has trivial kernel then

{\bar{π}}_{n}

must be an isomorphism. By Lemma 2.2 (d) the generators

g_{i}

H_{n}

commute or anti-commute, so the elements of

I n n (H_{n})

act by sign changes.

So if we can show that the automorphisms corresponding to

(12) (34)

(for

n \geq 4

) and

(123)

are not simply sign changes, we will have shown that

{\bar{π}}_{n}

is an isomorphism. The corresponding elements of

ℬ_{n}

are

(σ_{1} σ_{3})

and

(σ_{2} σ_{1})

and we use Lemma 2.2 (f ) to compute that under

\bar{ϑ} {\bar{π}}_{n}

the element

(σ_{1} σ_{3})

maps

g_{2}

g_{2} g_{1} g_{3}

, and

(σ_{2} σ_{1})

maps

g_{2}

g_{2} g_{1} g_{2} = g_{1}

. We check directly that

g_{1} g_{3} \neq \pm 1

using Lemma 2.2 (a),(i) and (j), so

{\bar{π}}_{n}

is an isomorphism for

n \geq 3

. In the case

n = 2

we see that

G_{2}

is the group generated by the matrix

R

which is isomorphic to

Z_{4}

, so combining, we have:

Theorem 2.6. We have an exact sequence:

1 \to H_{n} \overset{\subset}{\to} G_{n} \overset{ϑ}{\to} S_{n} \to 1

for all

n \geq 2

. In other words,

G_{n}

is an extension of

H_{n}

S_{n}

3 Extraspecial 2-groups and related groups

Definition 3.1. The group

E_{m}^{ν}

is the abstract group generated by

x_{1}, \dots, x_{m}

with relations:

\begin{matrix} x_{i}^{2} = ν, 1 \leq i \leq m \end{matrix}

(1)

\begin{matrix} x_{i} x_{j} = x_{j} x_{i}, | i - j | \geq 2 \end{matrix}

(2)

\begin{matrix} x_{i + 1} x_{i} = - x_{i} x_{i + 1}, 1 \leq i \leq m \end{matrix}

(3)

where

- 1

is an order two central element, and

ν = \pm 1

These groups appear classically and have important connections with Clifford algebras. The case

ν = - 1

appears in Exercise 3.9 in the text by Fulton and Harris [FH] , and other cases appeared in [G] . The necessary facts about these groups are found in various places, but are elementary so we reprove them here for the reader's convenience.

3.1 Properties of $E_{m}^{ν}$

Any element in

E_{m}^{ν}

can be expressed in the normal form:

\pm x_{1}^{α_{1}} \dots x_{m}^{α_{m}}

where

α_{i} \in Z_{2}

. The following lemma will show that it is unique.

Lemma 3.2. Denote by

Z (E_{m}^{ν})

the center of

E_{m}^{ν}

. We have:

(a) $Z (E_{m}^{ν}) = {\begin{matrix} {\pm 1} & m even \\ {\pm 1, \pm x_{1} x_{3} \dots x_{m}} & m odd \end{matrix}$
(b) $E_{m}^{ν} / {\pm 1} \sim = (Z_{2})^{m}$
(c) Any $x \in E_{m}^{ν} \ Z (E_{m}^{ν})$ is conjugate to $- x$ .
(d) Any nontrivial normal subgroup of $E_{m}^{ν}$ intersects $Z (E_{m}^{ν})$ nontrivially.
(e) For $m = 2 k - 1$ odd, $Z (E_{2 k - 1}^{ν}) \sim = {\begin{matrix} Z_{2} \times Z_{2} & if ν = 1 or k even \\ Z_{4} & if ν = - 1 and k odd \end{matrix}$
(f ) The normal form $\pm x_{1}^{α_{1}} \dots x_{m}^{α_{m}}$ is unique.

Proof. Using the above-mentioned normal form we may assume, without loss of generality, that

z = x_{1}^{α_{1}} \dots x_{2 k}^{α_{m}} \in Z (E_{m}^{ν})

since if

z

is central, so is

- z

. By the commutation/anti-commutation relations in

E_{m}^{ν}

we have

x_{i} z = (- 1)^{α_{i - 1} + α_{i + 1}} z x_{i} = z x_{i}

for all

i

where we take

α_{- 1} = α_{m + 1} = 0

Thus we get the system of equations over

Z_{2}

\begin{matrix} α_{2} = 0 \end{matrix}

\begin{matrix} α_{m - 1} = 0 \end{matrix}

\begin{matrix} α_{i} + α_{i + 2} = 0 (m o d 2), 1 \leq i \leq m - 2 \end{matrix}

m

is even then the system has only the trivial solution

α = 0

, but if

m

is odd there are two solutions

0

and

(1, 0, 1, \dots, 0, 1)

, that is, all the

α_{2 i} = 0

and

α_{2 i + 1} = 1

. Thus we have (a). It is clear from the relations in

E_{m}^{ν}

that

E_{m}^{ν} / {\pm 1}

is presented by

m

commuting generators of order 2, i.e

(Z_{2})^{m}

. To prove (c) observe that any non-central element

x \in E_{m}^{ν}

must anti-commute with some

x_{i}

. So (d) follows from (c) as any nontrivial normal subgroup

N

must either be central or contain

{x, - x}

for some non-central element

x

so that

- 1 \in N

as well. For (e) we compute the order of the central element

x_{1} x_{3} \dots x_{2 k - 1}

and find that it is 2 or 4, which gives us the two cases.

Assertion (f ) follows from a simple counting argument as

| E_{m}^{ν} | = 2^{m + 1}

by (b). □

Definition 3.3. A group

G

of order

2^{m + 1}

is an extraspecial 2-group if (see [G] ):

(1) The center $Z (G)$ and the commutator subgroup $G^{'}$ coincide and are isomorphic to $Z_{2}$ .
(2) $G / Z (G) \sim = (Z_{2})^{m}$ .

It is immediate from the anti-commutation relations that the commutator subgroup of

E_{m}^{ν}

{\pm 1}

, and for

m = 2 k

the other conditions were verified in Lemma 3.2 so we have:

Proposition 3.4.

E_{2 k}^{ν}

is an extraspecial 2-group.

Remark 3.5. Since the group

E_{2 k + 1}^{ν}

contain

E_{2 k}^{ν}

, we call the groups

E_{m}^{ν}

nearly extraspecial 2-groups for any

m

(so they include extraspecial 2-groups). This should not be confused with almost extraspecial 2-groups found in the literature which are central products of extraspecial 2-groups with

Z_{4}

. The cases where the center of

E_{m}^{ν}

is isomorphic to

Z_{4}

are almost extraspecial, but when the center is

Z_{2} \times Z_{2}

they are not (see [G] ).

3.2 Representations of $E_{m}^{ν}$

We wish to construct the irreducible representations of

E_{m}^{ν}

. There are 4 cases corresponding to the parity of

m

and the choice of

ν

. For the reader's convenience we recall the following standard facts from the character theory of finite groups (see any standard text, e.g. [FH] ):

Proposition 3.6. Let

G

be a finite group, and

I r r (G) = {χ_{i}}_{i \in ℐ}

the set of irreducible characters of

G

, corresponding to irreducible representations

V_{i}

(a) $| I r r (G) |$ is equal to the number of conjugacy classes of elements of $G$ .
(b) $| G | = \sum_{ℐ} (dim V_{i})^{2}$
(c) For $χ_{i}$ , $χ_{j} \in I r r (G)$ $\sum_{g \in G} χ_{i} (g) \bar{χ_{j} (g)} = {\begin{matrix} 0 & if V_{i} ≁ = V_{j} \\ | G | & if V_{i} ∼ = V_{j} \end{matrix}$ .
(d) If $g$ and $h$ are not conjugate then $\sum_{ℐ} χ_{i} (h) \bar{χ_{i} (g)} = 0$ .

3.2.1 $E_{2 k}^{- 1}$

To determine the number of irreducible representations we count conjugacy classes. The center

{\pm 1}

gives us two singleton classes, and Lemma 3.2 (c) shows that the non-singleton conjugacy classes are given by

[\pm x_{1}^{α_{1}} \dots x_{2 k}^{α_{2 k}}]

for any

α \in (Z_{2})^{2 k} \ {0}

. So we have

2 + (2^{2 k} - 1) = 2^{2 k} + 1

inequivalent irreducible representations. Let

I r r (E_{2 k}^{- 1}) = {V_{1}, \dots, V_{2^{2 k} + 1}}

denote a set of inequivalent irreducible representations of

E_{2 k}^{- 1}

. By Lemma 3.2 we can induce 1-dimensional representations of

E_{2 k}^{- 1}

from any representation of

(Z)^{2 k}

by letting the center act trivially. Thus we have

2^{2 k}

1-dimensional representations (say,

V_{2}, \dots, V_{2^{2 k} + 1}

) leaving only one representation,

V_{1}

to determine. Using the class equation

2^{2 k + 1} = | E_{2 k}^{- 1} | = (dim V_{1})^{2} + \sum_{2}^{2^{2 k} + 1} (dim V_{i})^{2} = (dim V_{1})^{2} + 2^{2 k}

we find that

dim V_{1} = 2^{k}

. The 1-dimensional representations are equal to their characters so for

2 \leq i \leq 2^{2 k} + 1

we have

χ_{i} (1) = χ_{i} (- 1) = 1

, and

χ_{i} ([\pm x_{j}]) = \pm 1

for all possible choices of sign. From Proposition 3.6 (c),(d) we find that

χ_{1} (1) = - χ_{1} (- 1) = 2^{k}

, and

χ_{1} ([\pm x_{1}^{α_{1}} \dots x_{2 k}^{α_{2 k}}]) = 0

. We can construct the representation

(ρ_{1}, V_{1})

as follows (recall the definition of the matrices

s

and

σ_{z}

from section 2.1 ):

\begin{matrix} ρ_{1} (x_{1}) = \sqrt{- 1} (σ_{z} \otimes I_{2}^{\otimes k - 1}) \end{matrix}

\begin{matrix} ρ_{1} (x_{2}) = s \otimes I_{2}^{\otimes k - 1} \end{matrix}

\begin{matrix} ⋮ \end{matrix}

\begin{matrix} ρ_{1} (x_{2 i}) = I_{2}^{\otimes i - 1} \otimes s \otimes I_{2}^{\otimes k - i} \end{matrix}

\begin{matrix} ρ_{1} (x_{2 i + 1}) = \sqrt{- 1} (I_{2}^{\otimes i - 1} \otimes σ_{z} \otimes σ_{z} \otimes I_{2}^{\otimes k - i - 1}) \end{matrix}

\begin{matrix} ⋮ \end{matrix}

\begin{matrix} ρ_{1} (x_{2 k}) = I_{2}^{\otimes k - 1} \otimes s \end{matrix}

(σ_{z})^{2} = I_{2}

s^{2} = - I_{2}

and

σ_{z} s σ_{z} = - s

we see that this is indeed a representation of

E_{2 k}^{- 1}

, and since

t r (s) = t r (σ_{z}) = 0

it follows from the orthogonality of characters that this is the irreducible

2^{k}

-dimensional representation of

E_{2 k}^{- 1}

3.2.2 $E_{2 k - 1}^{- 1}$

We now construct the irreducible representations of

E_{2 k - 1}^{- 1}

. Denote by

z

the central element

x_{1} x_{3} \dots x_{2 k - 1}

for convenience. Using Lemma 3.2 we find that there are

2^{2 k - 1} + 2

distinct conjugacy classes in

E_{2 k - 1}^{- 1}

and therefore we may label the inequivalent classes of irreducible representations by

I r r (E_{2 k - 1}^{- 1}) = {W_{1}, \dots, W_{(2^{2 k - 1} + 2)}}

. We get

2^{2 k - 1}

distinct 1-dimensional representations from

(Z_{2})^{2 k - 1}

by composing with the projection onto

E_{2 k} / {\pm 1}

, and them by

W_{3}, \dots, W_{(2^{2 k - 1} + 2)}

. We compute their characters

ψ_{i}

for

3 \leq i \leq 2^{2 k - 1} + 2

as in the

E_{2 k}^{- 1}

case

ψ_{i} (1) = ψ_{i} (- 1) = 1

and

ψ_{i} (\pm x_{j}) = \pm 1

which determines their values on all classes (observe that

ψ_{i} (z) = ψ_{i} (- z)

for nontrivial central elements

\pm z

for these 1-dimensional representations).

From Proposition 3.6 (b) we get

dim W_{1} + dim W_{2} = 2^{k}

for the remaining two irreducible representations. Since

dim W_{i} | 2^{2 k}

we see that in fact,

dim W_{1} = dim W_{2} = 2^{k - 1}

. Using Proposition 3.6 (c) we find that the characters

ψ_{1}

and

ψ_{2}

vanish on all equivalence classes except for the central classes:

[1]

[- 1]

[z]

and

[- z]

. Observing that the restrictions of

W_{1}

and

W_{2}

to the subgroup

E_{2 k - 2}^{- 1} \subset E_{2 k - 1}^{- 1}

must both be the unique non-trivial irreducible

2^{k - 1}

-dimensional representation we find that

ψ_{1} (- 1) = ψ_{2} (- 1) = - 2^{k - 1}

Proposition 3.6 (c),(d) then implies first that

ψ_{1} (z) = ψ_{2} (- z) = - ψ_{1} (- z) = - ψ_{2} (z)

, and then using this and the orthogonality of

ψ_{1}

and

ψ_{2}

to see that

∥ ψ_{1} (z) ∥ = 2^{k - 1}

. Restricting to

Z (E_{2 k - 1}^{- 1})

and recalling that

Z_{2} \times Z_{2}

has only real characters while the non-trivial characters of

Z_{4}

have pure complex values on its generators we determine the value of

ψ_{1} (z)

up to a choice of sign coming from switching

W_{1}

and

W_{2}

. For the purpose of simplifying notation later we include a sign depending on the value of

k (m o d 4)

and define:

\begin{matrix} ψ_{1} (x) = {\begin{matrix} \pm 2^{k - 1} & for x = \pm 1 \\ \pm (- 1)^{(k / 2} (2^{k - 1}) & for x = \pm z \\ 0 & otherwise \end{matrix} \end{matrix}

(4)

and

\begin{matrix} ψ_{2} (x) = {\begin{matrix} \pm 2^{k - 1} & for x = \pm 1 \\ \mp (- 1)^{(k / 2} (2^{k - 1}) & for x = \pm z \\ 0 & otherwise \end{matrix} \end{matrix}

(5)

Next we give explicit matrix realizations of

W_{1}

and

W_{2}

. Since

Z (E_{2 k - 1}^{- 1})

must act non-trivially (although not necessarily faithfully) on

W_{1}

and

W_{2}

we use the inclusion

E_{2 k - 1}^{- 1} \subset E_{2 k}^{- 1}

to observe:

I n d_{E_{2 k - 1}^{- 1}}^{E_{2 k}^{- 1}} (W_{1}) = I n d_{E_{2 k - 1}^{- 1}}^{E_{2 k}^{- 1}} (W_{2}) = V_{1}

where

V_{1}

is the

2^{k}

-dimensional irreducible representation of

E_{2 k}^{- 1}

given in 3.2 . Thus by Frobenius reciprocity (and a dimension count) we have that

\begin{matrix} R e s_{E_{2 k - 1}^{- 1}}^{E_{2 k}^{- 1}} (V_{1}) = W_{1} \oplus W_{2} \end{matrix}

(6)

From this we get explicit realizations

(λ_{1}, W_{1})

and

(λ_{2}, W_{2})

. (N.b. the only difference of

λ_{1}

and

λ_{2}

on the generators is that the image of

x_{2 k - 1}

differs in sign.)

\begin{matrix} λ_{1} (x_{1}) = λ_{2} (x_{1}) = \sqrt{- 1} σ_{z} \otimes I_{2}^{\otimes k - 2} \end{matrix}

\begin{matrix} λ_{1} (x_{2}) = λ_{2} (x_{2}) = s \otimes I_{2}^{\otimes k - 2} \end{matrix}

\begin{matrix} ⋮ \end{matrix}

\begin{matrix} λ_{1} (x_{2 i}) = λ_{2} (x_{2 i}) = I_{2}^{\otimes i - 1} \otimes s \otimes I_{2}^{\otimes k - i - 1} \end{matrix}

\begin{matrix} λ_{1} (x_{2 i + 1}) = λ_{2} (x_{2 i + 1}) = \sqrt{- 1} I_{2}^{\otimes i - 1} \otimes σ_{z} \otimes σ_{z} \otimes I_{2}^{\otimes k - i - 2} \end{matrix}

\begin{matrix} ⋮ \end{matrix}

\begin{matrix} λ_{1} (x_{2 k - 2}) = λ_{2} (x_{2 k - 2}) = I_{2}^{\otimes k - 2} \otimes s \end{matrix}

\begin{matrix} λ_{1} (x_{2 k - 1}) = - λ_{2} (x_{2 k - 1}) = \sqrt{- 1} I_{2}^{\otimes k - 2} \otimes σ_{z} \end{matrix}

One easily checks that these indeed define irreducible representations of

E_{2 k - 1}^{- 1}

just as in the

m = 2 k

case. It is perhaps worth computing the traces of the images of the central element

z

under

λ_{1}

and

λ_{2}

. We have:

λ_{1} (z) = - λ_{2} (z) = (\sqrt{- 1})^{k} ((σ_{z})^{2} \otimes \dots \otimes (σ_{z})^{2}) = (\sqrt{- 1})^{k} I_{2^{k - 1}}

so that:

t r (λ_{1} (z)) = - t r (λ_{2} (z)) = {\begin{matrix} 2^{k - 1} & if k \equiv 0 (m o d 4) \\ - 2^{k - 1} & if k \equiv 2 (m o d 4) \\ \sqrt{- 1} (2^{k - 1}) & if k \equiv 1 (m o d 4) \\ - \sqrt{- 1} (2^{k - 1}) & if k \equiv 3 (m o d 4) \end{matrix}

The traces of the images of

\pm 1

are also easily computed, and comparing these values with the above formulas 4 and 5 , we check that the characters of

λ_{1}

and

λ_{2}

are

ψ_{1}

and

ψ_{2}

respectively.

3.2.3 $E_{m}^{1}$

Suppose that

(ρ, V)

is any representation of

E_{m}^{- 1}

defined on generators

ρ (x_{i}) = A_{i}

for some set of matrices

{A_{i}}_{1 \leq i \leq m}

. Denote by

x_{1}^{'}, \dots, x_{m}^{'}

the generators of

E_{m}^{1}

and define

ρ^{'} (x_{i}^{'}) = \sqrt{- 1} A_{i}

. Then since

(A_{i})^{2} = - I d_{V}

we have

(ρ^{'} (x_{i}^{'}))^{2} = I d_{V}

and

(ρ^{'}, V)

defines a representation of

E_{m}^{1}

(observe that the commutation/anti-commutation relations are homogeneous and hence also satisfied). Obviously this process is reversible, so that all representations of

E_{m}^{1}

are obtained in this way. If we define representations

λ_{1}^{'}

and

λ_{2}^{'}

E_{2 k - 1}^{1}

corresponding to the two

2^{k - 1}

-dimensional representations of

E_{2 k - 1}^{- 1}

then we find that the characters

ψ_{1}^{'}

and

ψ_{2}^{'}

always have real values on the central elements

\pm z^{'} = \pm x_{1}^{'} x_{3}^{'} \dots x_{2 k - 1}^{'}

as they should–since according to Lemma 3.2 the center of

E_{2 k - 1}^{1}

is always isomorphic to

Z_{2} \times Z_{2}

4 Applications

In this section we describe the abstract structure of the groups

G_{n}

and

H_{n}

and decompose the representation

π_{n} : ℬ_{n} \to (C^{2})^{\otimes n}

into its irreducible constituents. We then extend these ideas to the re-normalized representation of

ℬ_{n}

that factors over the Temperley-Lieb algebra.

4.1 $H_{n}$ and $G_{n}$ as abstract groups

Theorem 4.1.

H_{n} \sim = E_{n - 1}^{- 1}

Proof. To verify that the map

φ : E_{n - 1}^{- 1} \to H_{n}

defined by

x_{i} \to g_{i}

extends to a (surjective) group homomorphism one just checks that the

g_{i}

satisfy the defining relations of

E_{n - 1}^{- 1}

. Since

K e r (φ)

is normal it must be trivial or intersect

Z (E_{m}^{- 1})

by Lemma 3.2 (d). We check that

φ (- 1) = φ (x_{1}^{2}) = g_{1}^{2} = - I_{2^{n}}

- 1 \notin K e r (φ)

and we have proved the theorem for

n - 1

even. If

n - 1 = 2 k - 1

is odd, we must also check that

\pm z \notin K e r (φ)

where

z

is the nontrivial central element defined in Lemma 3.2 . For this we must use Lemma 2.2 (k) which shows that there is a change of basis which diagonalizes the odd-indexed

g_{2 i + 1}

while fixing the even indexed

g_{2 i}

. We compute the image of

z

in this basis:

\begin{matrix} (P_{n})^{- 1} φ (\pm z) P_{n} = (P_{n})^{- 1} (\pm g_{1} g_{3} \dots g_{2 k - 1}) P_{n} = \pm (\sqrt{- 1})^{k} (σ_{z}^{\otimes 2 k}) \end{matrix}

which is a diagonal matrix of trace

0

, so not the identity. □

Combining with Theorem 2.6 we have:

Theorem 4.2. The image of

ℬ_{n}

under the representation

π_{n}

is an extension of

E_{n - 1}^{- 1}

S_{n}

4.2 Decomposition of $π_{n}$

By Theorem 4.1 we have

E_{n - 1}^{- 1} \sim = H_{n}

as an abstract group so the (defining) representation

(π_{n}, (C^{2})^{\otimes n})

H_{n}

induces a representation

φ_{n} : = π_{n} \circ φ

E_{n - 1}^{- 1}

4.2.1 $n$ odd

Assume that

n = 2 k + 1

is odd. Then we may decompose

(C^{2})^{\otimes 2 k + 1} \sim = \oplus_{i} m_{i} V_{i}

as representations of

E_{2 k}^{- 1}

for some multiplicities

m_{i}

. Let

χ

be the character of

φ_{2 k + 1}

. Since

φ_{2 k + 1} (- 1) = (I_{2} \otimes \dots \otimes g_{i}^{2} \dots \otimes I_{2}) = - I_{2^{n}}

we see that

χ (- 1) = - 2^{2 k + 1}

and

χ (1) = 2^{2 k + 1}

. By Proposition 3.6 we can compute the multiplicities

m_{i}

of the irreducible components

V_{i}

m_{i} = \frac{1}{2^{2 k + 1}} \sum_{x \in E_{2 k}^{- 1}} χ_{i} (x) \bar{χ (x)}

The character

χ_{1}

of the

2^{k}

-dimensional representation

V_{1}

vanishes on the non-central elements of

E_{2 k}^{- 1}

so we compute the multiplicity

m_{1} = {\frac{2^{k} \cdot 2^{2 k + 1} + 2^{k} \cdot 2^{2 k + 1}}{2}}^{2 k + 1} = 2^{k + 1}

V_{1}

appears

2^{k + 1}

times. But

dim V_{1} = 2^{k}

dim (2^{k + 1} V_{1}) = 2^{2 k + 1} = dim (C^{2})^{\otimes 2 k + 1}

, so in fact

π_{2 k + 1}

decomposes diagonally as

2^{k + 1}

copies of the unique

2^{k}

-dimensional representation

(ρ_{1}, V_{1})

E_{2 k}^{- 1}

4.2.2 $n$ even

Suppose

n = 2 k

is even. We have already established (see 6 in Section 3.2 ) that the restriction of the irreducible

2^{k}

-dimensional representation

V_{1}

E_{2 k}^{- 1}

E_{2 k - 1}^{- 1}

decomposes as the direct sum

W_{1} \oplus W_{2}

of the two inequivalent irreducible

2^{k - 1}

dimensional representations

W_{1}

and

W_{1}

. So the

2^{2 k}

-dimensional representation

φ_{2 k}

decomposes diagonally as the direct sum of

2^{k}

copies of each of

(λ_{1}, W_{1})

and

(λ_{2}, W_{2})

. One could also use the characters

ψ_{i}

to determine these multiplicities.

Remark 4.3. As

π_{n} (P_{n}) = φ_{n} (E_{n - 1}^{- 1})

, all of the arguments above hold mutatis mutandis for decomposing

π_{n}

restricted to

P_{n}

4.2.3 Extension to $ℬ_{n}$

With the explicit formulas for the representations

ρ_{1}

λ_{1}

and

λ_{2}

in hand, we easily compute the extensions

{\hat{ρ}}_{1}

{\hat{λ}}_{1}

and

{\hat{λ}}_{2}

ℬ_{n}

using Lemma 2.2 (a).

Using the matrices

d

M

and

D

from Section 2.1 we give the explicit matrices for the

2^{k}

-dimensional irreducible representation

{\hat{ρ}}_{1}

with

n = 2 k + 1

noting that the

{\hat{λ}}_{1} \oplus {\hat{λ}}_{2}

is just the restriction of

{\hat{ρ}}_{1}

\begin{matrix} {\hat{ρ}}_{1} (σ_{1}) = d \otimes I_{2}^{\otimes k - 1} \end{matrix}

\begin{matrix} ⋮ \end{matrix}

\begin{matrix} {\hat{ρ}}_{1} (σ_{2 i}) = I_{2}^{\otimes i - 1} \otimes M \otimes I_{2}^{\otimes k - i} \end{matrix}

\begin{matrix} {\hat{ρ}}_{1} (σ_{2 i + 1}) = I_{2}^{\otimes i - 1} \otimes D \otimes I_{2}^{\otimes k - i - 1} \end{matrix}

\begin{matrix} ⋮ \end{matrix}

\begin{matrix} {\hat{ρ}}_{1} (σ_{2 k}) = I_{2}^{\otimes k - 1} \otimes M \end{matrix}

The decomposition of

π_{n}

remains the same, so summarizing we have:

Theorem 4.4. The representation

π_{n}

ℬ_{n}

decomposes as

(C^{2})^{\otimes n} \sim = {\begin{matrix} (C^{2})^{\otimes (n + 1) / 2} \otimes V_{1} & n odd \\ (C^{2})^{\otimes n / 2} \otimes (W_{1} \oplus W_{2}) & n even \end{matrix}

5 Jones representation and Jones polynomial

The Jones representation of the braid groups

ℬ_{n}

are defined using the Temperley-Lieb algebras

T L_{n} (q)

. Jones representation

ρ_{r}

in the following means the unitary representation of the braid groups at

q = e^{2 π i / r}

factoring through the semisimple Temperley-Lieb algebras, which are quotients of the Hecke algebras in [J1] . The specific formulas that we use are the ones in [FLW] .

Definition 5.1. Let

q = \sqrt{- 1}

. The Temperley-Lieb algebra

T L_{n} (q)

is defined as the (semisimple) quotient of the braid group algebra

C [ℬ_{n}]

by (the ideal generated by) the relations:

TL1: $(σ_{i} + 1) (σ_{i} - q) = 0$
TL2: $σ_{i} σ_{i + 1} σ_{i} + σ_{i} σ_{i + 1} + σ_{i + 1} σ_{i} + σ_{i} + σ_{i + 1} + 1 = 0$
TL3: $(σ_{i} - σ_{i + 1})^{2} = \sqrt{- 1}$ (i.e. Jones-Wenzl projector $p_{3} = 0$ )

Observing that the Yang-Baxter operator

R

satisfies

(R - ζ I_{4}) (R - \bar{ζ} I_{4}) = 0

we can define a new matrix

R^{'} = - \bar{ζ} R

that satisfies

(R^{'} - \sqrt{- 1} I_{4}) (R^{'} + I_{4}) = 0

. Since the equation (YBE) is homogeneous, (YBE) is satisfied by

R^{'}

also.

It is a (mildly tedious) computation to verify that the matrices

A_{1} = (R^{'} \otimes I_{2})

and

A_{2} = (I_{2} \otimes R^{'})

satisfy

A_{1} A_{2} A_{1} + A_{1} A_{2} + A_{2} A_{1} + A_{1} + A_{2} + I_{4} = 0

, and

(A_{1} - A_{2})^{2} = \sqrt{- 1} I_{4}

. Thus the representation

π_{n}^{'}

ℬ_{n}

afforded us by

R^{'}

(or

C ℬ_{n}

if we prefer) factors over the Temperley-Lieb algebra

T L_{n} (\sqrt{- 1})

We can easily extend what we have learned about the representation

π_{n}

ℬ_{n}

to this slight variation by observing the effect of renormalizing

R

. We record the result in the following (compare to [J1] ):

Corollary 5.2. Denote by

H_{n}^{'} = π_{n}^{'} (P_{n})

and

G_{n}^{'} = π_{n}^{'} (ℬ_{n})

. Then we have

H_{n}^{'} \sim = E_{n - 1}^{1}

, and

G_{n}^{'} / H_{n}^{'} \sim = S_{n}

Proof. This follows easily from the observation that renormalizing $R$ by $- \bar{ζ}$ has the effect of multiplying the generators $g_{i}$ of $H_{n}$ by $- \sqrt{- 1}$ . Doing the same to the generators of the group $E_{n - 1}^{- 1}$ just gives us a presentation of the group $E_{n - 1}^{1}$ , and the same arguments as in the original representation $π_{n}$ go through verbatim. □

To relate

π_{n}^{'}

to the Jones representation

ρ_{4}

ℬ_{n}

, we recall some facts about the Jones representation. The Temperley-Lieb algebras at a 4-th root of unity are complex Clifford algebras and are isomorphic to the matrix algebra of

2^{n - 1} \times 2^{n - 1}

matrices if

n

is odd, and the direct sum of two matrix algebras of

2^{\frac{n}{2} - 1} \times 2^{\frac{n}{2} - 1}

matrices if

n

is even [J2] . (Note here

n

is the number of strands in the geometric realization of

ℬ_{n}

, and differs by 1 from Jones' notation in [J1] .) So the Jones representation

ρ_{4}

consists of a single irreducible sector if

n

is odd, and the direct sum of two irreducible sectors if

n

is even. Comparing with the comments in Subsection 3.2.3 we can also determine the decomposition of the representation

π_{n}^{'}

as before. It follows that the restriction of the Jones representation

ρ_{4}

P_{n}

for

n

even is the odd representation

V_{1}

of the extra-special 2-group

E_{n - 1}^{1}

, and for

n

odd,

W_{1} \oplus W_{2}

as in Theorem 4.4. The images

ρ_{4} (ℬ_{n})

fit into the following exact sequence:

1 \to E_{n - 1}^{1} \to ρ_{4} (ℬ_{n}) \to S_{n} \to 1 .

Projectively, we have

1 \to Z_{2}^{n - 1} \to ρ_{4} (ℬ_{n}) \to S_{n} \to 1 .

The symmetric group

S_{n}

acts on the coordinates of

Z_{2}^{n}

, hence

Z_{2}^{n - 1}

when

n

is even. This action splits the exact sequence. But when

n

is odd, this sequence does not split as is shown in [J1] .

The Jones polynomial of a link at

\sqrt{- 1}

is given by the following formula [FLW] :

J_{4} (\hat{σ}) = (- 1)^{n - 1 + \frac{e (σ)}{4}} \cdot (\sqrt{2})^{- \frac{1 + (- 1)^{n}}{2}} \cdot Trace (ρ_{4} (σ)),

where

e (σ)

is the sum of all exponents of the standard braid generators appearing in

σ

, and

\hat{σ}

is the closure of

σ

. We can also define link invariants using the flipped R-matrix

R

. The conditions for enhancement

(μ_{i}, α, β)

is given in [T] Theorem 2.3.1. Working through the conditions, we found two link invariants:

T_{R} (\hat{σ}, α) = α^{n - e (σ)} \cdot (\sqrt{2})^{- n} \cdot Trace (π_{n} (σ))

, where

α = \pm 1

Comparing with the Jones polynomial we get the relation:

T_{R} (\hat{σ}, α) = (- 1)^{n - 1 + e (σ)} \cdot α^{n - e (σ)} \cdot \sqrt{2} \cdot J_{4} (\hat{σ}) .

As we know that Jones polynomial

J_{4} (\hat{σ})

(\sqrt{2})^{c (\hat{σ}) - 1} \cdot (- 1)^{Arf (\hat{σ})}

Arf (\hat{σ})

is defined and

0

otherwise, where

c (\hat{σ})

is the number of components of the link

\hat{σ}

[J2] . It follows that

T_{R} (\hat{σ}, α)

computes essentially the Arf invariant of a link.

References

J. Birman, Braids, links, and mapping class groups. Annals of Mathematics Studies, No. 82. Princeton University Press, Princeton, N.J.; University of Tokyo Press, Tokyo, 1974.
H. Dye, Unitary solutions to the Yang-Baxter equation in dimension four. Quantum information processing 2 (2002) nos. 1-2, 117–150 (2003).
M. Freedman, M. Larsen, Z. Wang, The two-eigenvalue problem and density of Jones representation of braid groups. Comm. Math. Phys. 228 (2002), 177-199, arXiv: math.GT/0103200.
M. Freedman, A. Kitaev, M. Larsen, Z. Wang, Topological quantum computation. Bull. Amer. Math. Soc. (N.S.) 40 (2003), no. 1, 31–38.
W. Fulton, J. Harris, Representation theory, A first course. Graduate Texts in Mathematics, 129. Readings in Mathematics. Springer-Verlag, New York, 1991
R. Griess, Automorphisms of extra special groups and nonvanishing degree two cohomology. Pacific J. Math. 48 no. 2 (1973) 403–422.
J. Hietarinta, All solutions to the constant quantum Yang-Baxter equation in two dimensions. Phys. Lett. A, 165 (1992), 2452-52.
V. F. R. Jones, Braid groups, Hecke algebras and type $I I_{1}$ factors. Geometric methods in operator algebras (Kyoto, 1983), 242–273, Pitman Res. Notes Math. Ser., 123, Longman Sci. Tech., Harlow, 1986.
———, Hecke algebra representations of braid groups and link polynomials. Ann. Math. 126 (1987), 335–388.
L. Kauffmann, S. Lomonaco Jr., Braiding operators are universal quantum gates. New J. Phys. 6 (2004), 134.1-134.40 (electronic), arXiv: quant-ph/0401090.
N. Read, Non-abelian braid statisitcs versus projective permutation statistics. J. Math. Phys. 44 (2003), no. 2, 558–563, hep-th/0201240.
V. Turaev, The Yang-Baxter equation and invaraints of links. Invent. Math. 92 (1988), 527-553.
F. Wilczek, Fractional statistics and anyon superconductivity. World Scentific, 1990.

E-mail address : jefranko@indiana.edu Department of Mathematics, Indiana University, Bloomington, IN 47405, U.S.A. E-mail address : errowell@indiana.edu Department of Mathematics, Indiana University, Bloomington, IN 47405, U.S.A. E-mail address : zhewang@indiana.edu Department of Mathematics, Indiana University, Bloomington, IN 47405, U.S.A.

Jennifer Franko, Eric C. Rowell,

Zhenghan Wang