Quantum Harmonic Oscillator Algebra and Link Invariants
Quantum Harmonic Oscillator Algebra and Link Invariants
Quantum Harmonic Oscillator Algebra and Link Invariants
Invariants
arXiv:hep-th/9111005v1 4 Nov 1991
October,1991
0
1 Introduction
The connection between quantum groups [1] and link invariants was first
stablished in [2]. The simplest way to describe this connection is by associ-
ating with a given quantum group an extended YangBaxter system in the
sense of Turaev [3]. Using this procedure for the case of Uq (SU(2)) one
reobtains Jones invariant [4]. Given the finite dimensional irrep of spin j of
N
Uq (SU(2)), V (j) , a realization of the braid group Bn in End ( n V (j) ) is de-
fined in terms of the quantum Rmatrix of Uq (SU(2)) in the representation
j, as follows:
j,j
: i R(i,i+1) 1 (1)
For any word in Bn the corresponding link invariant is defined by:
k
(i j k )Rij =0
X kj
Rij j = abik (3)
j
X
(R1 )kj 1 k
ij j = a bi
j
= q H/2
a = q cj (1)2j (4)
b = [2j + 1]
In the particular case j = 1/2 the invariant (2) coincides with Jones
polynomial. The state model, in the spirit of Kauffman [5], associated
with the invariant (2) can be defined using the quantum 6jsymbols of Uq
(SU(2)). Moreover the relationship with vertex integrable models [6] follows
1
by defining the representation (1) of Bn as the very anisotropic limit of the
trigonometric solution for the YangBaxter operators.
In the last few months some new connections relating the Alexander
Conway polynomial with quantum groups have been discovered [79]. In
particular in reference [7] the quantum group realization of the Alexander
polynomial was obtained starting with the supergroup GL (1,1) which turns
out to define a free fermion model of the invariant. In a completely dif-
ferent context Date et al make contact with the Alexander polynomial in
their study of the braid group representations arising from the chiral Potts
model [9]. This result should indicate a new interplay between the Alexan-
der polynomial and quantum groups, based on the characterization of the
solution to the start triangle equation for the chiral Potts model in terms of
intertwiners for the cyclic representations of Uq (SUd (2)) at roots of unit [10].
In a previous work we have found the intertwiners R matrices for semi-
cyclic representations of Uq (SU(2)) with q = e2i/N (for definitions and
more details see references [11, 12]). In the limit of N going to infinity these
R matrices give rise to a new braid group representation realized on tensor
product of an infinite dimensional space which in fact is isomorphic to the
Hilbert space of a harmonic oscillator [12]. This braid group representation
admits a baxterization [13] which allows the introduction of new parameters
and, interesting enough, the baxterized Rmatrices satisfies the Turaev con-
ditions (3) for having a link and knot invariant which was already defined
in [12]. The aim of this paper is twofold: i)clarify the quantum group origin
of this infinite dimensional braid group representation and ii) give a proper
definition and identification of the link invariant obtained with these new R
matrices.
The main results that we obtain are:
1. The relation of the baxterized Rmatrices, coming from the semicyclic
irreps of Uq (SU(2)), with the Rmatrices of a quantum group defor-
mation of the harmonic oscillator algebra h4 .
2. The identification of the invariant of reference [12] with the inverse of
the AlexanderConway polynomial.
3. The construction of new colored braid group representations associated
with the quantum group Uq (h4 ) and their relation with the multivari-
able AlexanderConway polynomial.
These results show that the inverse of the Alexander polynomial admits a
bosonic model in close analogy with the free fermion model of the Alexander
2
invariant presented in references [7]. The fermionic version of this invariant
is based on a quantum deformation of the supergroup GL (1,1), which in
our bosonic model is replaced by the harmonic oscillator algebra h4 .
The plan of the paper is the following: in section 2 we review the infinite
dimensional representations of the braid group founded in [12]. In section 3
we define the quantum deformation of the harmonic oscillator algebra h4 and
stablish the connection between the universal R matrix of this Hopf algebra
and the infinite dimensional braid group representation introduced in section
2. In section 4 we define the link invariants and we extend the result to the
case of the multivariable AlexanderConway polynomial. In Appendix A we
perform some explicit computation of link and knots invariants, in Appendix
B we study the ClebschGordan decomposition of the irreps of Uq (h4 ), and
finally in appendix C we discuss the ribbon structure of Uq (h4 ).
EF q2 F E = 1 K 2
KE = q 2 EK (5)
2
KF = q FK
E = E 1 + K E
F = F 1+K F (6)
K = K K
3
[a, a+ ] = 1 [k, a] = [k, a+ ] = 0 (8)
The semicyclic representations in this limit are now dimensional and
they are labelled by the eigenvalues of K:
F er = (1 )er+1
Eer = r(1 + )er1 (9)
Ker = er
Denoting H these irreps, it was proved in reference [12] that there exist
an intertwiner R1 ,2 : H 1 H 2 H 2 H1 satisfying the Yang Baxter
equation. The explicit form of the intertwiner is given by
(r2 ,r1 )
Rrr11,r
+r2 ,
2
(u) = (1 + u)r1 ur2 P (1 2u2 ) (10)
1 2 (,)
with u = 11 2 and Pn (x) the Jacobi polynomials:
n
! !
1 X n+ n+
Pn(,) (x) = n (x 1)nm (x + 1)m (11)
2 m=0 m nm
R(+)R(1) = 1 (15)
The representantion of Bn as End (n H) is defined by:
4
i 1 Ri () 1 (16)
with H isomorphic to the Hilbert space of the harmonic oscillator. We will
refer to this representation of the braid group as the harmonic oscillator
representation.
A compact way to rewrite eqs (13) and (14) is in terms of the universal
R matrix defined as R = P R with P the permutation operator P (1 2 | =
1 1 (1 , 2 H). Using the creation and annihilation operators a, a+
we find that R() admit the following nice representation:
+
R(+) = (eiN 1)e2aa
+ a
R() = (1 eiN )e2a (17)
with N the number operator: N = a+ a.
The harmonic oscillator braid group representation (17) admit a baxter-
ization [13] in the following sense:
+ a
R(x, y, ) = e(xy)a (y N xN ) (18)
satisfy the Yang Baxter relation:
5
define a dimensional representation of the braid group x,y : Bn
End(Hn ). Notice that we recover the Rmatrices (13), (14) from eq (20)
in the case where x = 1, y = 1.
At a first stage it may appear that we are dealing in (20) with a two
parameter family of inequivalent braid group representations, but this is not
the case since we can prove the equivalence between the representations x,y
and x,y for any 6= 0. This follows automatically from the relation:
Proposition 2.2 The braid group representation x,y admits, for arbitrary
x, y, an extension a la Turaev.
1
p In fact it is easy to check that for the identity operator and a = b =
(y/x) the set of equations (3) are satisfied for R = R(x, y) with arbitrary
values of x and y. In these conditions the link invariant (2) becomes:
[N a+ ] = a+
[N, a] = a (23)
[a, a+ ] = E
6
and with E commuting with all the generators.
This algebra can be obtained as a particular contraction of U(2). The
two casimirs are:
c1 = EN a+ a
c2 = E 2
q E q E
[a, a+ ] =
q q 1
[N, a] = a
[N, a+ ] = a+ (24)
with the comultiplication:
a = a q E/2 + q E/2 a
a+ = a+ q E/2 + q E/2 a+ (25)
E = E 1 + 1 E
N = N 1+1N
(26)
The antipode:
(x) = x x = a, a+ , E, N (27)
and counit:
(x) = 0 (28)
1
This quantum deformation should not be confused with the qoscillators defined in
reference [14]
7
Proposition 3.1 The algebra Uq (h4 ) is a quasitriangular Ribbon Hopf al-
gebra.
is given by:
1 )(q E/2 q E/2 )aa+
R = q (EN +N E)e(qq (30)
The ribbon structure of Uq (h4 ) will be study in section C.
Each irrep of Uq (h4 ) is labelled by the values of the two Casimirs c1 , c2 (=
q E q E +
qq 1 N a a) which is equivalent to give the eigenvalues (e, n) of E and
N . A generic irrep (e, n) is defined as follows:
a|r > = [e]1/2 r|r 1 >
a+ |r > = [e]1/2 r + 1|r + 1 >
E|r > = e|r > (31)
N |r > = (r + n)|r >
where {|r >} r=0 is a orthonormal basis and the qnumber [e] is defined as
q x q x
[x] = qq1 .
We shall always consider the cases with [e] 6= 0. Evaluating the universal
R matrix (30) in the tensor product (e1 n1 ) (e2 n2 ) we define the matrix:
e1 e2 r1 r2 e1 e2
(r1 r2 )
(q q e1
)(q q e2
) 2
q 2 q e1 r1 e2 r2
8
!1/2 !1/2
r r r2 r2
(R1 )r11 r22 (e1 , e2 ) = r1 +r2 ,r1 +r2
r1 r1
e1 e2 r2 r1 e1 e2
(r1 r2 )
(q q e1
)(q q e2
) 2
(1)r2 r1 q 2 q e1 r2 +e2 r1 (33)
Rie2 e3 Ri+1
e1 e3
Rie1 e2 = Ri+1
e1 e2 e1 e3 e2 e3
Ri Ri+1 (34)
which reduces to the ordinary braid group relation if e1 = e2 = e3 .
In addition to (34) one has the following properties
P1: Hermiticity
r r
Rrr1 rr2 (e1 , e2 ) = q (e1 e2 )(r1 r1 ) Rr11r22 (e1 , e2 )
(35)
1 2
r r r r
Rr22 r11 (e2 e1 ) = (1)r1 r2 q e1 (r2 +r1 )e2 (r1 +r2 ) (R1 )r11 r22 (e1 , e2 ) (36)
9
4 The harmonic oscillator link invariants and the
Alexander polynomial
In this section we will proceed to define a regularized version of (22) which
preserves the invariance under Markov moves.
Given an irrep (e, n) of Uq (h4 ) the link invariant defined by the associated
extended YangBaxter system is:
(M I) T () = T ()
(38)
(M II) T (m ) = T () Bm1
To make sense of the infinite dimensional trace in (37) we will proceed
to define a regularization similar to the one used in references [7] [9].
The proof of this proposition goes as follows. First of all we must notice
that given the tangle obtained by cutting one strand, let say the ith strand,
c we can always find another strond i such that the tangle
of the link
c is equivalent to the original one.
obtained by cutting the ith strand of
Using this result, the invariance under Markov I of the regularized trace
(39) reduces to prove that it is independent of which strand we chose to cut,
namely:
10
T r2,m(e ()) = T r1,i,i+2..m(e ()) (40)
The identity (40) is a formal consequence of the following lemma.
Lemma 4.1
+ +
e ()(eza 11 ..) = (eza 11 ..)(11 )e ()(11 1 ) (43)
Z(unknot) = 1 (45)
The invariant Z is a polynomial in q e . In all the examples we have
considered (see Apprendix A for some non trivial cases) we have found the
result:
1
Z (q e ) = (46)
(t)
with (t = q e ) the Alexander polynomial. The result (46) is very natural
in our approach where we have considered as the starting point the Hopf
algebra Uq (h4 ) which can be thought as a bosonic version of GL (1,1). We
do not yet have a general proof of (46). Notice that although satisfies a
skein rule its inverse does not. This is of course related to the fact that the
braid group representation underlying Z is infinite dimensional and that
R(e, e) has an infinite number of different eigenvalues for generic values of
11
q e . This means that the proof of (46) cannot probably proceed through
the skein rule as is usually done for other invariants coming from quantum
groups.
The colored version of the previous invariant is easily obtained once we
have a colored braid group representation as in (34). Proceeding as the
authors of ref [9] we first renormalized the matrix R(e1 , e2 ) according to:
e1 +e2
R(e1 , e2 ) q 2 R(e1 , e2 ) (47)
such that R(e1 , e2 ) do satisfies the braid group relation while the Turaev
conditions (3) holds now with
X X
R(e, e)kj
ij = R1 (e, e)kj
ij = q
e/2 k
i (48)
j j
with this redefinition the colored version of the invariant (37) reads:
Pn n
Y
e
T () = q i=1 i T r() = ti T r() (49)
i=1
e2 e1 e1 e2
Z2 = (t1 t1
1 )t2 T r R R =1 (51)
1
which is the correct value for the two component link in the normalization
adopted above.
Here again we conjecture that the invariant (50) is the inverse of the
multivariable Alexander Conway polynomial.
12
Appendix A
Calculation of some links and knots invariants
In this appendix we shall compute some invariants of knots and links which
shall support the conjecture (46), i.e. that the invariant Z is the inverse of
the AlexanderConway polynomial.
Instead of using the normalized basis |r > defined in eqs. (31) it will be
more convenient to use the following basis for the irrep (e, n) of Uq (h4 ):
a vr = [e] r vr1
a+ vr = vr+1
(A1)
N vr = (r + n) vr
E vr = e vr
The braiding matrices (33) now read:
r r
Rr11r22 (e1 , e2 ) = r1 +r2 ,r1 +r2
!
r1 r1 r2 r1 r2
r1 r2
(t1 t1
1 ) (t1 t1
2 )
2 t1 t2 (A2)
r2
r r
(R1 )r11 r22 (e1 , e2 ) = r1 +r2 ,r1 +r2
!
r2 r2 r1 1
(r r1 ) r2 r1
(t1
2 t2 ) (t2 t1
1 )
2 2 t1 t2 (A3)
r1
where t1 = q e1 and t2 = q e2 .
It is clear from (41) that the invariant Z given in eq. (44) can also be
compute as:
13
X
T r(Ree Ree (N n 11)) = Rr11r22 (e, e)Rr11r22 (e, e)1
r1 r2 = 0
1 2
! !
X r1 1
= r1 +r2 ,1 +2 (t t1 )r1 2 +1 r2 (A5)
2 r2
r1 r2 1 2
t(r1 +r2 +1 +2 ) 1
! !
X 1 1 + r2 1
(1)2 (t t1 )2(1 r2 ) t2(1 +2 ) 1 (A6)
2 r2
r2 1 2
14
For more complex links or knots the calculation of the invariant requires
the knowledge of complicated combinatorial formulae. There is however a
way to overcome this difficulty, which consist in making an integral repre-
sentation of the functions in R and R1 ; namely:
I
dz 1+r1 +r2 r r
r1 +r2 ,r1 +r2 =
z 1 2 (A11)
2i
where the contour integral is performed around the origin. Thus we associate
to each braiding matrix R or R1 a complex variable z; having done this
we can perform all the sums over the entries so that tr(()N 11))
gets converted into a multicontour integral.
Let us exemplify this method with the computation of the invariant of
the trefoil.
Z(13 B2 ) = (1 ) t2 T rR3 (e, e)(N 11)
X
T r (R3 (e, e)) = Rrm11rm
2
2
Rr11r22 Rm
1 2
1 m2
m1
r,,m
X I dz1 dz2
= z11+r1 +r2 m1 m2 z21+1 +2 r1 r2 m1
r,l,m
2i 2i
! ! !
r1 1 m1
(tt1 )r1 +1 +m1 r2 2 m2 t(r1 +1 +m1 +r2 +2 +m2 )
m2 r2 2
(A12)
Notice that we have not introduced an integral representation for the
delta 1 +2 ,m1 +m2 , since this is implied by the first two deltas in (A12).
The sum over r2 , 2 and m2 can be done straigforwardly yielding:
I
dz1 dz2 1 1
z z
(2i)2 1 2
! r1 ! 1 !m1
X 1 + t(t t1 )z1 X z1 + z2 t(t t1 ) X [z2 + (t t1 )]
r1 t2 z2 1
t2 m1 t2 z1
I
dz1 dz2 t6
=
(2i)2 (t2 z2 t(t t1 )z1 1)(t2 z1 z2 t(t t1 ))(t2 z1 z2 t(t t1 ))
(A13)
15
The convergence of the geometric sums in r1 , 1 and m1 impose the
following restrictions on z1 , z2 and :
t2
T rR3 = (A17)
t2 + t2 1
which yields:
1
Z(Trefoil) = (A18)
1 + (t t1 )2
according to the conjecture (46).
c4 in B and d
2 2 d 1 3
We have also checked 1 2 1 2 and (1 2 ) in B3 , obtaining
the result (44).
16
Appendix B
ClebschGordan coefficients of Uq (h4 )
If (e1 , n1 ) and (e2 , n2 ) are two irreps of Uq (h4 ) with e1 , e2 and e1 +e2 different
from zero then the tensor product (e1 , n1 ) (e2 , n2 ) decomposes into irreps
as follows:
M
(e1 , n1 ) (e2 , n2 ) = (e1 + e2 , n1 + n2 + n) (B1)
n0
!1/2
X nr r re1 nr n
r e2
(1) [e1 ] 2 [e2 ] q
2 2 q 2
r
r0
(e n ),(e n ) (e n )(e n )
= (e11 +e1 2 ,n21 +n
2
2 +n)
K(e12+e2 2 ,n11 +n
1
2 +n)
17
(1)n q e1 n1 +e2 n2 (e1 +e2 )(n1 +n2 +n) (B5)
We recognize in this expression the classical casimir C1 in the exponential
q en , which is the analog of the braiding factor q j(j+1) of the quantum group
Uq (SU(4)). Similarly (1)n is a parity factor analogous to (1)j in SU(2).
18
Appendix C
Uq (h4 ) and Ribbon Hopf Algebras
Any quasitriangular A has an invertible element, usually called u, with the
property that
2 (a) = u a u1 a A (C1)
The element u and its inverse u1 can be obtained from the universal R
matrix as follows:
u = m[ id((R))]
(C2)
u1 = m[id 2 ((R))]
In the particular case of the universal Rmatrix (30) of the quantum
group Uq (h4 ) we obtain:
P (q 1 q)
u = 0 ! q E (a+ ) a q 2EN
(C3)
P (qq 1 )
u1 = 0 ! q E (a+ ) a q 2EN
The element u of the quantum deformation Uq (G) of a complex simple
Lie algebra G does not in general commute with the others elements of the
algebra, however for Uq (h4 ) we have from eqs (27) that 2 = id so that u is
central. Indeed using (C3) and (31) we obtain:
(u) = q 2E u (C5)
We are now in conditions to study wether Uq (h4 ) is a Ribbon Hopf
algebra. Following Resbetikhin and Turaev [2] we define a Ribbon Hopf
Algebra (A, R, v) as a quasitriangular Hopf algebra with a universal R matrix
and the choice of an element v such that:
19
v is central
v2 = u(u)
(v) = v (C6)
(v) = 1
(vu1 ) = vu1 vu1
In the case of Uq (h4 ) we can easily see using eq. (C5) that the element
v is given by:
v = q E u (C7)
which in the irrep (e, n) takes the value
v = e2i (C9)
where v is the value of v on the irrep of Uq (G) associated to the primary
field . If eq. (C9) would holds true for h4 it would imply that
20
References
[1] V.G. Drinfeld, Quantum Groups, Proceedings of the 1986 Interna-
tional Congress of Mathematics at Berkeley ed. A.M. Gleason (1987)
Am. Math. Soc., 1, p. 798.
M. Jimbo, Comm. Math. Phys., 102 (1986) 537.
[2] A.N. Kirillov and N.Yu. Reshetikhin, Representations of the algebra Uq
(SU(2)), qorthogonal polynomials and invariants of links. Leningrad
preprint LOMI-E-9-88.
N.Yu. Reshetikhin and V.G. Turaev. Comm. Math. Phys., 127 (1990)
1.
[3] V.G. Turaev. Invent. Math., 92 (1988) 527.
[4] V.F.R. Jones. Ann. Math., 126 (1987) 335.
[5] L.H. Kauffman. On Knots Ann. of Math. Studies N 115. Princeton
University Press 1987.
[6] Y. Akutsu, A. Kuniba and M. Wadati. Phys. Rep., 180 (1989) 427.
[7] L.H. Kauffman and H. Saleur. Free Fermions and the Conway Alexan-
der polynomial. Comm. Math. Phys. (1991) in press.
L. Rozansky and H. Saleur. YCTP-P20-91
[8] E. Date, M. Jimbo, K. Miki and T. Miwa. RIMS-729 preprint (1990).
[9] Y. Akutsu and T.K. Deguchi. Phys. Rev. Lett., 67 (1991) 777.
[10] E. Date, M. Jimbo, K. Miki and T. Miwa. RIMS-715 (1990), RIMS-706
(1990), RIMS-703 (1990).
[11] C. Gomez, M. Ruiz Altaba and G. Sierra. Phys. Lett., B265 (1991) 95.
[12] C. Gomez and G. Sierra. A New Solution to the StarTriangle Equa-
tion based on Uq (Sl(2)) at roots of unit. CERN preprint TH 6200/91.
[13] V.F.R. Jones. Int. Journ. of Modern Phys. A Vol. 6 (1991) 2035.
[14] A.J. MacFarlane. J. Phys. A: Math. Gen., 22 (1989) 4581.
L.C. Biedenharn. J. Phys. A: Math. Gen., 22 (1989) L.873.
[15] L. AlvarezGaume, C. Gomez and G. Sierra. Nucl. Phys., B330 (1990)
347.
21