UATP-93/06

December 1993

Wilson Loops and Black Holes

in 2+1 dimensions
arXiv:gr-qc/9401017v1 17 Jan 1994

†
Cenalo Vaz
Unidade de Ciências Exactas e Humanas
Universidade do Algarve
Campus de Gambelas, P-8000 Faro, Portugal

and

††
Louis Witten
Department of Physics
University of Cincinnati
Cincinnati, OH 45221-0011, U.S.A.

Abstract

In 2+1 dimensional Chern-Simons gravity, Wilson loops in the three dimensional Anti
de Sitter group, SO(2, 2), reproduce the spinning black hole of Bañados, Teitelboim and
Zanelli (BTZ) by naturally duplicating the necessary identification of points of a four
dimensional globally SO(2, 2) invariant space in which the hole appears as an embedding.

† Internet: [email protected]
†† Internet: [email protected]
 Pure gravity, Anti-de Sitter (AdS) and de Sitter (dS) gravity and supergravity in
three dimensions can be formulated as a Chern-Simons gauge theories of Poincaré group,
ISO(2, 1), the Anti-de Sitter group, SO(2, 2), the de Sitter group, SO(3, 1), and their
various supersymmetric extensions.1,2 The Chern-Simons action is topological and in all
cases the theory is finite and solvable both on the classical and the quantum levels3 .
Moreover, the precise relationship between the Poincaré Chern-Simons theory and the
spacetime of Einstein’s gravity in three dimensions has been well established.4

Recently Bañados, Teitelboim and Zanelli5,6 (BTZ) discovered that three dimensional
AdS gravity admits a spinning black hole solution which arises by identification of points
of AdS space by a discrete subgroup of SO(2, 2). In Einstein gravity, these identifica-
tions must be made by hand and one is led to seek a more natural framework for them.
The Chern-Simons approach provides precisely such a framework. It is well known, for
example, that the multiconical structure of spacetime in pure (Poincaré) gravity arises
from multivalued gauge transformations on the coordinates of a globally trivial Minkowski
spacetime.7 The multivaluedness of the gauge transformations leads to conical identifi-
cations of points in Minkowski space. Our aim in this paper is to provide a similar
description of the BTZ black hole.

In ref. (2) gravity in three dimensions was interpreted as a gauge theory with a Chern-
Simons action

1 2
Z
IC.S. = − Tr A ∧ (dA + A ∧ A)
2 3
Z M (1)
1 1 C D
= − γBC AB ∧ (dAC + f A ∧ AE ),
2 3 DE
M

C are the structure constants of the gravity group,

where Aµ is the gauge connection, fDE
XA are the generators of its Lie algebra, γAB = 2Tr(XA XB ) plays the role of a metric on
the Lie algebra and M is an arbitrary three manifold. Consider the basis

[Jˆab , Jˆcd ] = ηad Jˆbc + ηbc Jˆad − ηac Jˆbd − ηbd Jˆac (2)

of the Lie algebra so(2, 2) of SO(2, 2). The roman indices a... ǫ {0, 1, 2, 3}, and ηab is
the metric of a four dimensional flat space, which we denote by Mq , with signature
ηab = diag(−, −, +, +). SO(2, 2) admits two Casimir operators, which in this basis are
Jab J ab and ǫabcd J ab J cd . Expanding Aµ according to

1 ab ˆ
Aµ = ω Jab , (3)
2 µ

the action in (1) can be put in the form

1 cd
Z
IC.S. = − ǫabcd ω ab ∧ (dω cd + f ω ab ∧ ω ef ), (4)
6 (ab)(ef )
M

where
cd c d c d c d c d
f(ab)(ef ) = ηaf δb δe + ηbe δa δf − ηae δb δf − ηbf δa δe (5)

and γAB = ǫabcd , ǫ0123 = +1 serves as the metric on so(2, 2). The metric has been chosen
so that (4) has a non-degenerate Poincaré limit. In a time-space decomposition, the action
in (4) can also be written in the form
Z Z  
ij
IC.S. = dt ǫabcd ǫ ωiab ∂t ωjcd − ω0ab Fijcd (6)
Σ

where i, j ǫ {1, 2} are spatial indices in M and

 
ab
Fµν = ∂µ ωνab − ∂ν ωµab + ωfaµ ωνf b − ωfaν ωµf b (7)

is the SO(2, 2) curvature tensor. From (6) follow the free field Poisson brackets,

{ωiab (x), ωjcd (y)}P.B. = − ǫabcd ǫij δ (2) (x, y) (8)

and, as ω0ab is a Lagrange multiplier, the constraints

ǫij Fijab = 0. (9)

An alternative formulation can be given in the basis P̂a , Jˆa of so(2, 2) defined by

[P̂a , P̂b ] = λǫabc J c , [P̂a , Jˆb ] = ǫabc P c, [Jˆa , Jˆb ] = ǫabc J c , (10)

where λ is the “cosmological constant” and the roman indices a... ǫ {0, 1, 2} are raised
and lowered by ηab = diag(−, +, +), if Aµ is expanded according to

Aµ = eaµ P̂a + ωµa Jˆa (11)

where eaµ is the dreibein and ωµa is the spin connection. Of the two quadratic Casimirs,
P̂ · Jˆ and P̂ 2 + Jˆ2 /λ2 , the former must be used if one is interested in a non-degenerate
invariant bilinear form in the Poincaré limit. The Chern-Simons action may then be cast
into the form
Z Z h i
ija b a b a b
IC.S = dt ηab ǫ ei ∂t ωj − e0 Fij [ω] − ω0 Fij [e] , (12)
Σ
a [e] and F a [ω] are the torsion and curvature respectively.
where Fµν µν

The two descriptions of AdS gravity may be connected by the following correspondence
between the components of the connection in the two bases:
√ √
ωµ01 = λe0µ , ωµ23 = λωµ0 ,
√ √
ωµ12 = − λe1µ , ωµ03 = λωµ1 , , (13)
√ √
ωµ13 = λe2µ , ωµ02 = λωµ2 ,

The authors in ref. (8) compute the group element associated with the BTZ solution in
the second basis. We will use the Chern-Simons theory to reproduce the identifications
required to construct the black hole as an embedding, so we will work in the first.

The generators Jˆab admit the following irreducible infinite dimensional representation
on Mq ,

Jˆab = qa ∂b − qb ∂a , (14)

where q a are coordinates in Mq . Following Witten’s2 suggestion, if pa is conjugate to

q a , spinless point sources can be coupled to the field action above by adding to (4) the
Borel-Weil-Bott action
Z
IS = dt (pa Dt q a + constraints) , (15)

in which Dt is the SO(2, 2) covariant derivative,

1 µ mn ˆ
Dt = ∂t + t ωµ Jmn , (16)
2
where tµ is tangent to the trajectory of the source in M. The q a (t) represent points in
Mq . They are functions of time, mapping points on the trajectory of the particle in M
to points in Mq so that

q a (t) = q a [xµ (t)] (17)

is a functional of points P [xµ (t)] in M. A source at the origin of M, will therefore modify
the constraints in (9) to read

ǫij Fijab = (pa q b − pb q a )δ 2 (x) = S ab δ 2 (x). (18)

where the S ab are conserved charges, by Noether’s theorem. The source action has an
obvious generalization to the case of N non-interacting particles.

The (pa , q a ), cannot be identified with the spacetime momenta and coordinates of
the particle. They transform as SO(2, 2) vectors. Consider now the coordinates q̄(P ),
constructed from the coordinates q(P ) by the following gauge transformation

ZP
q̄(P ) = g(P )q(P ) = P exp( dxµ Aµ ) q(P ), (19)
∗

where P represents path ordering. The gauge transformation

ZP
g(P ) = W∗P = P exp( dxµ Aµ ) (20)
∗

is the Wilson line from the origin, ∗ in M, to the point P . We think of the q̄(P ) as
living in an internal space, donoted hereafter by M̄q . These coordinates satisfy unusual
boundary conditions when S ab 6= 0.

To see this consider a single source located at ∗, and a trajectory γ which loops about
the source. Let the point P on γ be mapped to the point q(P ) in Mq , so that q̄(P ) in
M̄q is given by (19). Traversing the loop γ once in the counterclockwise direction, and
returning to P , we find that the same point in M̄q is also given by

q̄(P ) = W∗P Wγ q(P ) = W∗P Wγ WP ∗ q̄(P ) (21)

where we have used W∗P WP ∗ = 1 above. If W∗ is the infinitesimal loop about the origin,

Wγ = WP ∗ W∗ W∗P (22)

by a deformation of the loop γ. Thus the point in M̄q associated with q̄(P ) must be
identified with the point W∗ q̄(P ). For non-vanishing S ab this is a non-trivial matching
condition because, by the non-abelian Stokes theorem,

ab
Jˆab
W∗ = eS . (23)

If there are many sources, located at the points P1 , P2 , ...PN , the matching condition takes
the form

q̄(P ) = C̄1 C̄2 · · · C̄K q̄(P ) (24)

where each CI is an infinitesimal circle about the point PI , and C̄I is the result of parallel
transporting the source to the origin, i.e.,

ab
Jˆab
C̄I = W∗I CI WI∗ = eS̄ (25)

where

S̄ ab Jˆab = W∗I S ab Jˆab WI∗ (26)

and only those sources lying within the loop γ are included in the product on the right
hand side of (24). It is necessary of course to fix an ordering of the sources, which can be
done, for example, by increasing azimuthal angle in M. The formal matching conditions
in (24) are difficult to describe geometrically. Note, however, that the q̄(P ) defined by
(19) are sensitive only to gauge transformations at ∗. This is because, under an arbitrary
gauge transformation
W∗P → U(∗)W∗P U −1 (P )
(27)
q(P ) → U(P )q(P )

giving

q̄(P ) → U(∗)q̄(P ). (28)

where we have used (19). Note also that the derivation of these matching conditions is
independent of the nature of M, therefore the structure of M̄q is also independent of it.

The transformation (19) can be used to reconstruct spacetime. It has already been
shown to reproduce the multiconical structure of pure gravity7 . In AdS gravity it is
responsible for the appearance of a massive, spinning black hole as an embedding, not
in Mq but in M̄q . AdS space may be thought of as an embedding in Mq through the
constraint

−(q 0 )2 − (q 1 )2 + (q 2 )2 + (q 3 )2 = − l2 (29)

How one chooses to parametrize the constraint then determines which part of the manifold
one covers. The conventional parametrization covers the entire manifold. Let

(q 2 )2 + (q 3 )2 = rq2 ,
q 2 = rq cos φq , q 3 = rq sin φq
(31)
(q 0 )2 + (q 1 )2 = rq2 + l2 ,
q q
q0 = rq2 + l2 cos(tq /l), q 1 = rq2 + l2 sin(tq /l)

where rq ǫ [0, ∞) and tq /l, φ ǫ [0, 2π) are periodic coordinates. The metric ds2 =
−ηab dq a dq b is then

2 rq2 2 drq2
ds = ( 2 + 1)dtq − r 2 − rq2 dφ2q , (31)
l q
( l2 + 1)

but tq is still a periodic coordinate with period 2πl. This means that there are closed
time-like curves. To avoid them one no longer identifies tq = 0 with tq = 2πl, and arrives
at the usual AdS space which is therefore the universal covering of the space in (31). In
an alternative parametrization, let

− (q 0 )2 + (q 3 )2 = − rq2
(q 0 ) = rq cosh φq , (q 3 ) = rq sinh φq
− (q 1 )2 + (q 2 )2 = rq2 − l2 (32)
q q
q1 = rq2 − l2 sinh(tq /l), q2 = rq2 − l2 cosh(tq /l) rq > l
q q
q1 = l2 − rq2 cosh(tq /l), q2 = l2 − rq2 sinh(tq /l) rq < l

where rq , φq and tq are all non-compact coordinates. The metric ds2 = −ηab dq a dq b is then

2 rq2 2 drq2
ds = ( 2 − 1)dtq − r2 − rq2 dφ2q , (33)
l ( l2 − 1)
q

but φq is non compact, and (33) describes the universal covering of AdS space. Imposing
periodicity on φ by hand, i.e., identifying φq = 0 and φq = 2π, we recover the AdS black
hole spacetime of ref. (6) with “mass” M = 1 and zero angular momentum.
 To include angular momentum and arbitrary mass, consider a single source situated
at the origin of M, satisfying

S 03 = 2πA, S 12 = 2πB (34)

and all others zero. In a polar coordinatization of Σ, the solution of the constraint
equations reads

ωφ03 = A, ωφ12 = B (35)

and all other spatial components vanish. The time components of the connection are
Lagrange multipliers which may be selected arbitrarily. To recover the BTZ black hole
we must also pick

ωt03 = S 6= 0. (37)

With this choice of connection, the coordinates q̄(P ) defined in (19) become
 Rt Rφ Rφ 
−( dtωto3 + dφωφ03 )Jˆ03 − dφωφ12 Jˆ12
q̄(P ) = e q(P ) (38)

where φ and t, which appear in the exponent above are the angular coordinate and time
coordinate respectively in M. Equation (38) gives for the barred coordinates

q̄ 0 = cosh(St + Aφ)q 0 − sinh(St + Aφ)q 3

q̄ 1 = cosh(Bφ)q 1 − sinh(Bφ)q 2
(39)
q̄ 2 = cosh(Bφ)q 2 − sinh(Bφ)q 1
q̄ 3 = cosh(St + Aφ)q 3 − sinh(St + Aφ)q 0

Once again, imposing the AdS constraint with the same parametrization as in (32) but
on the barred coordinates one finds

− (q̄ 0 )2 + (q̄ 3 )2 = − rq2

(q̄ 0 ) = rq cosh φ̄q , (q̄ 3 ) = rq sinh φ̄q
− (q̄ 1 )2 + (q̄ 2 )2 = rq2 − l2 (40)
q q
1 2 2 2
q̄ = rq − l sinh(t̄q /l), q̄ = rq2 − l2 cosh(t̄q /l) rq > l
q q
q̄ 1 = l2 − rq2 cosh(t̄q /l), q̄ 2 = l2 − rq2 sinh(t̄q /l) rq < l.
The barred coordinates t̄q and φ̄q are related to the previous ones in (32) by

t̄q = tq − lBφ, φ̄q = φq − St − Aφ (41)

where φ and t are coordinates in M and the metric ds2 = −ηab dq̄ a dq̄ b is evidently

rq2 drq2
ds2 = ( − 1)d t̄2
q − − rq2 dφ̄2q , (42)
l2 rq2
( l2 − 1)

in terms of the barred coordinates.

When |B| < |1 − A|, an identification t ≡ tq and φ ≡ φq gives the BTZ black hole.
Let
r− r+ r−
B = , (1 − A) = , S = , (43)
l l lr+

where r− < r+ . Indeed, r− and r+ are the inner and outer horizons of the AdS black
hole,
l2
 q 
2 2 2
r± = M ± M − (J/l) , (44)
2

where M and J are its mass and angular momentum, and the metric

rq2 drq2 r+ r−
ds2 = ( 2
− 1)(dtq − r − dφ q ) 2
− r 2 − rq2 ( dφq − dt)2 , (45)
l q
( l2 − 1) l lr +

can be brought into the standard (BTZ) form

r2 J2 dr 2 J
ds2 = ( − M + )dt2
− − r 2 (dφ − dt)2 (46)
l2 4r 2 2
( rl2 − M + J2
4r 2 )
2r 2

by the coordinate transformation

r+2 − r2 

2 −
r = 2
rq2 + r−
2
l
l (47)
t = tq
r+
φ = φq .

For rq > l (46) gives the region r > r+ and for rq < l it gives the region r− < r < r+ .
 To recover the metric in the region r < r− , we consider the parametrization

− (q 0 )2 + (q 3 )2 = rq2
(q 0 ) = rq sinh φq , (q 3 ) = rq cosh φq
(48)
(q 1 )2 − (q 2 )2 = rq2 + l2
q q
q1 = rq2 + l2 cosh(tq /l), q2 = rq2 + l2 sinh(tq /l)

which gives
rq2 drq2
ds2 = − ( + 1)dt2
q − + rq2 dφ2q . (49)
l2 rq2
( l2 + 1)

When this parametrization is applied to M̄q as we did in the previous two cases, the
resulting metric can be cast into the form

rq2 drq2 2 r+ r−
ds2 = − ( + 1)(dtq − r − dφ q ) 2
− + r q ( dφ q − dt)2 , (50)
l2 2
rq
( l2 + 1) l lr +

which yields the metric (46) in the region 0 < r < r− after making the coordinate
transformation (47) with rq2 → −rq2 . The transformation is valid only in the region
0 ≤ rq2 < l2 r−
2 /(r 2 − r 2 ).
+ −

We have shown that the identifications necessary to construct the BTZ black hole
arise naturally from the Chern-Simons approach to gravity in three dimensions. Our
construction may be generalized to supergravity, which has a similar description as a
Chern-Simons theory.2,9 The advantage of a Chern-Simons description is also the ease
with which quantum effects may be described.4 These and other issues of black hole
physics in three dimensions will be discussed more fully elsewhere.

Acknowledgement

This work was supported in part by NATO under contract number CRG 920096. L.W. ac-
knowledges the partial support of the U. S. Department of Energy under contract number
DOE-FG02-84ER40153.

References
1. S Deser, R. Jackiw and G. ’t Hooft, Ann Phys. 152 (1984) 220; S. Deser and R. Jackiw,
ibid 153 (1984) 405.

2. A. Achucharro and P. K. Townsend, Phys. Letts. B180 (1986) 89; E. Witten, Nucl.
Phys. B311 (1988) 46; K. Koehler, F. Mansouri, C. Vaz and L. Witten, Mod. Phys.
Letts. A5 (1990) 935; J. Math. Phys. 32, (1990) 239.

3. E. Witten, Nucl. Phys. B323 (1989) 113; O. F. Dayi, Phys. Letts. B234 (1990) 25.

4. K. Koehler, F. Mansouri, C. Vaz and L. Witten, Nucl. Phys. B348 (1991) 373; ibid
B358 (1991) 677.

5. M. Bañados, C. Teitelboim and J. Zanelli, Phys. Rev. Letts. 69 (1992) 1894.

6. M. Bañados, C. Teitelboim and J. Zanelli, Phys. Rev. D48 (1993) 1506.

7. C. Vaz and L. Witten, Nucl. Phys. B368 (1992) 509; in Proc. of the XX International
Conference on Differential Geometric Methods in Theoretical Physics, Vol. II, eds. Sultan
Catto and Alvany Rocha, World Scientific, N.Y. 1992.

8. D. Cangemi, M. Leblanc and R. B. Mann, Phys. Rev. D48 (1993) 3606.

9. Olivier Coussarert and Marc Henneaux, Phys. Rev. Letts. 72 (1994) 183.