Classical and Quantum
Gravity
LETTER TO THE EDITOR You may also like
- Institute for communications!
Asymptotically anti-de Sitter space-times - Higher memory effects and the post-
Newtonian calculation of their
To cite this article: A Ashtekar and A Magnon 1984 Class. Quantum Grav. 1 L39 gravitational-wave signals
S Siddhant, Alexander M Grant and David
A Nichols
- Electronic ink displays promise
Peter Gymnne
View the article online for updates and enhancements.
This content was downloaded from IP address 138.204.73.181 on 25/02/2025 at 22:38
Class. Quantum Grav. 1 (1984) L39-L44. Printed in Great Britain
LETTER TO THE EDITOR
Asymptotically anti-de Sitter space-times
Abhay AshtekartS and Anne MagnonP
t Physique Thtorique, IHP, 11 Rue P et M Curie, 75231 Paris, France
and
$ Physics Department, Syracuse University, Syracuse, NY 13210, USA
5 Dtpartement de MathCmatiques, Universitt de Clermont-Fd, 63 I70 Aubikre, France
Received 1 May 1984
Abstract. The structure of the gravitational field at infinity of asymptotically anti-de Sitter
space-times is analysed in detail using conformal techniques. It is found that the situation
differs from that in the case of asymptotically Minkowskian space-times in a number of
respects, primarily because 9 is now time-like rather than null. In particular, the asymptotic
symmetry group is quite different from the BMS group, and there is no analogue of the
Bondi news. The analysis also introduces definitions of ‘conserved’ quantities at 9 in terms
of the curvature tensor which are free of the ambiguities present in the previous definitions
based on the deviation of the physical metric from an anti-de Sitter background.
Recently, there has been considerable interest in the effects of a negative cosmological
constant in Einstein’s equation. The general scenario is that, in presence of this term,
the anti-de Sitter space-time replaces the Minkowski space as the ground state of the
theory. Much attention has been given to issues such as the stability of this ground
state (Abbot and Deser 1982, Boucher et a1 1983), the appropriate boundary conditions
satisfied by the excited states of the theory, i.e. asymptotically anti-de Sitter spaces
(Breitenlohner and Friedman 1982, Hawking 1983), the positivity of energy of such
space-times (Gibbons et a1 1983, Choquet-Bruhat 1983), etc. However, these discussions
are incomplete in certain ways and the definitions of ‘conserved’ quantities used in
some of them are ambiguous because of the use of a background anti-de Sitter
space-time. These ambiguities are like-but much worse than-the supertranslation
ambiguities in the definition of angular momentum in the asymptotically Minkowskian
(or AM) contextst$. Since these quantities play an important role in the discussion of
issues mentioned above, it is desirable to have an intrinsic, background independent
definition for them analogous to the definition of the Bondi four-momentum (Penrose
1963). More generally, what is missing in the literature is a systematic analysis of the
asymptotic structure of these space-times along the lines of Penrose’s treatment (1963)
of AM space-times. The purpose of this letter is to fill this gap.
t Alfred P Sloan Research Fellow. Supported in part by the NSF contract PHY 8310041 with Syracuse
University.
$ For example, one says effectively that the mass information is contained in the deviation of the given
physical metric from an anti-de Sitter background. If ds2 = -(1 -2M/r-fAr2) d t 2 +. . . , the mass is given
by M . However, since Ar2 diverges as r + CO, this definition of mass is ambiguous: M changes if r is replaced,
for example, by P = r( 1 + a / r’).
0264-9381/84/040039 +06$02.25 0 1984 The Institute of Physics L39
L40 Letter to the Editor
Dejinition 1. A space-time? (k,&) will be said to be weakly asymptotically anti-de
Sitter if there exists a ?anifold with boundary, M, equipped with a metric gab and a
diffeomorphism from M onto M - dM, such that:
(i) there exists a function R on M such that g a b = R 2 & ,on k ;
(ii) 4 = dM is topologically S 2 x R, and, on 4, R = 0;
(iii) &, satisfies &-&ab + h & b = 8rGfab, with A < O , where R - ~ F ' ~admits
~ a
smooth limit to 4.
Remarks. (i) Note that, if one sets A = 0, the definition reduces to the usual definition
of asymptotic flatness at null infinity (see e.g. Ashtekar and Schmidt 1980). Hence one
can meaningfully compare the consequences of the two definitions.
(ii) It is straightforward to verify that the Schwarzschild anti-de Sitter space-time
satisfies definition 1 if one sets, e.g., R = l / r .
(iii) The condition on the fall-off of stress-energy required by the definition is quite
weak. For example, if the matter con$sts only of Yang-Mills fields, it is natural ,to
require that the field strength Fabe Fabe admit a smooth limit to 4. Then Rn-4Tab
admits a smooth limit as well.
Using the three conditions of the definition, one can establish a number of results
by a straightforward use of local differential geometry.
( 1 ) Set n, =Van.Then nana P -+A, where, from now on, P will denote 'equals at
points of 4 to'. Since A is negative, it follows that n a is space-like, i.e. that 4 is time-like
(Penrose 1964).
(2) Conformal freedom. Given a conformal completion ( M , g a b ) of (k,$?ab) which
satisfies definition 1, ( M , g a b ) also satisfies it iff g a b = W 2 g , b where w is a smooth nowhere
vanishing function on M. Using this conformal freedom, one can set V a n bC 0. The
restricted conformal freedom is now R + w R where 2,,w P 0. Note, however, that,
since n a is a space-like normal to 4, the condition 2'"w 2 0 does not restrict the
functional form of w within 4 in any way. This is in contrast with the situation in the
AM context where the condition VanbP 0 forces w to be a function of 0 and q only
on 4. Although the condition Vanb P 0 simplifies the calculations that follow, it is not
essential since all our final results are conformally inyariant.
(3) The physical space Bianchi identities yield on M (equation ( 9 )in Geroch (1977))
RV[aSb]c+ Cabcdnd = V[aR-'Lb]c-fln-2gc[aLb]dnd, (1)
where
, b - 8rGf)&b.
Lab = 8 ~ G f ~ ' g ++(A (2)
Using the fact that R-'fab admits a smooth limit to 9,it now follows that
Cabcdnd P 0. (3)
Since n d is not null on 4, and since (3) implies that the electric and magnetic parts
of Cabcd must vanish on 4, we have
Cabcd e 0. (4)
Note that (4) has been obtained as a local result; it holds even if 4 is not topologically
S2x R. In the AM case, the proof of (4)-but not of (3)-uses global arguments which
require that 4 be S 2 X R (Penrose 1964, Geroch 1977).
t All our manifolds and fields are C". The metric has signature - + + + and the curvature tensors are
defined by V L , V b l K ,= f R a b C d K dR,,
, = Rob:.
Letter to the Editor L41
(4) The Bianchi identities in ( M , g a b ) imply
VdCabcd + v [ a s b ] c =0 (5)
which in turn implies
Kabcdnd + v [ a s b ] c 0
on 4, where Kabcd = ~ ' c , b , , . (6) and (1) now yield
lim
-9 f l - ' ( v [ a f l - ' L b ] c -a-'gc[aLb]dnd) = 0. (7)
By suitable projections of this equation, one obtains
lim
+.9 fabnaqbc= o (8)
where qbc is the petric induced on the C! =constant surfaces by &b. Thus, although
the limit of fl-3Tabto 4 is not zero in gen!ral, the field equations and the Bianchi
identities imply that certain components of TRbfall off faster than a'.Finally, (1) and
( 6 ) imply
VdKabcd = -fl-'V[afl-'Lb]c +flP4gc[aLb]dnd (9)
on M, whence
D,E"G
~ -4 lim
-9 ~ - ~ f ~ ~ n ~ q > (10)
where D is the intrinsic derivative operator on (4,q a b ) induced by g a b and where
Eab = - ( 3 / A ) K a m b n n m n n is the electric part of the asymptotic Weyl curvature. (Recall
that nn = -$A.) Equation (8) implies that the limit of the right side of (10) exists.
( 5 ) 'Conserved' quantities. Given any conformal Killing field 6" on (9,q a b ) and
a cross section C of 4, we define a 'conserved' quantity (analogous to the Bondi
four-momentum in the AM case):
Qc[C]:= - ( 8 ~ ) - ' ( - 3 / A ) ' ' ' fc EabSa dSb. (1 1)
(Note that 6" is a conformal Killing field just on (9,q a b ) , not necessarily in the interior.)
This definition has a number of desirable properties. First, the expression on the right
side of (1 1) is conformally invariant. Second, the numerical factors have been so
chosen that if one uses for 5" the static Killing field (with norm -(1-2M/r - f A r 2 ) )
in the Schwarzschild anti-de Sitter space-time, one obtains Q c [ C ]= M for any C.
Next, the flux F,(A) of Ql across any three-dimensional region A of 4 is given by
F,(A) = (16.n)-'
J* lim(fl-4f.i.,.fiaqcb)[bd 3 4
where fi is the unit normal (with respect to g a b ) to =constant surfaces. The 'local
flux', i.e. the expression under the integral sign on the right, is by itself conformally
invaritnt. Furthermore, the integral is precisely the limit of the usual flux constructed
from Tabin the physical space-time across a time-like volume element as the element
approaches A within 4. The surprising thing is that, in absence of matter, the flux
vanishes and Q c [ C ] is absolutely conserved. This is in striking contrast with the
situation in the AM space-times; there is no analogue of the Bondi news leakage!
L42 Letter to the Editor
Thus, although definition 1 reduces to the usual definition of asymptotic flatness when
A tends to zero, the limit is, so to say, discontinuous: for any non-zero value of A,
there is no analogue of (the physical effects associated with) ‘pure’ gravitational
radiation?. This is in striking contrast with the zero rest mass fields, such as Yang-Mills,
propagating on an asymptotically anti-de Sitter background: these fields do get radiated
away across 9, the loss of their ‘energy’ being given by the right side of (12). Finally,
one can also define ‘dual’ conserved quantities, * Q t [ C ] by , replacing Eabin (11) by
the magnetic part, Bab := -(3/A)*Kambnnmnn, of the asymptotic Weyl curvature. These
quantities are absolutely conserved even in presence of matter fields. (At 4 or io of
AM space-times, these dual quantities vanish identically if gab is smooth, without any
‘wire singularities’. The proof, however, does not extend to the class of space-times
satisfying definition 1 ; a priori, there is no reason why these dual quantities should
vanish in this case.)
In spite of the fact that definition 1 leads to a rich structure, it is quite unsatisfactory
in one respect. To see this, let us first examine the ‘universal structure’ of 9,i.e. the
structure which is common to all weakly asymptotically anti-de Sitter space-times: 4
is topologically S 2 X R and time-like. The metric-or rather the conformal class of
metrics-on 9 is not a part of the universal structure; the metric q a b on 4 of a
space-time satisfying definition 1 need not be conformally related to the metric on 4
of another such space-time. (In the AM context, on the other hand, 9 is equipped
with just one class of degenerate conformal metrics. This comes about because 4,
being null, is ruled by the integral curves of na, and because the space of generators,
being topologically S 2 , has a unique conformal class of metrics. In the present case,
since n n is not tangential to 4, 4 does not have a fibre bundle structure, and S 2 x R
admits many conformal structures.) Since the universal structure is so limited, the
asymptotic symmetry group is very large: it is the diffeomorphism group of 4. Hence,
in general, there is no guarantee that (9,q a b ) would admit any conformal Killing field
whatsoever. Thus, for the class of space-times under consideration, the number of
‘conserved’ quantities (1 1) can vary from one space-time to another depending on the
number of conformal Killing fields on 4, i.e. by the form of the Bach tensor of q a b $ .
To improve this situation, one may strengthen the asymptotic conditions and reduce
the group. The maximum number of conformal Killing fields that a three-manifold 4
with a non-degenerate metric q a b can admit is ten and this number is reached if and
only if the Bach tensor
of qab vanishes. Furthermore, in this case, the conformal Killing Lie algebra is
isomorphic with the anti-de Sitter Lie algebra. Thus, it is natural to require, in addition
to the conditions in definition 1, that %&c of (4,q a b ) should vanish, i.e. that (4,q a b )
be conformally flat. What does this condition look like in the four-dimensional
language? It is straightforward to show that
t Since 9 is time-like, one’s first reaction may be that QJC]would not be conserved absolutely: radiation
may go out or come in (Abbot and Deser 1982, Gibbons et a1 1983). However, as we have seen, this is not
the case even when the condition Bo, & 0 (see below) is not imposed.
$ For this class, in general, one cannot use the Witten procedure (Gibbons et a1 1983) to construct conserved
quantities because one does not know what boundary conditions to use to solve the Witten equation.
Letter to the Editor L43
Hence we have
8abc 90 e Bab & 0 (15)
conformal Lie algebra of (9,q a b ) is the 0(2,3) Lie algebra. (That the third statement
implies the first two was emphasised by Hawking (1983) to argue that his ‘reflective
boundary condition’ BobP 0 is natural.) Since (15 ) implies that qab must be conformally
flat, the conformal class of gab can now be included in the universal structure. This
reduces the asymptotic symmetry group to the ten-dimensional anti-de Sitter group
(or covering thereof). This is analogous to the PoincarC reduction of the BMS group
at 9 (and the Spi group at io) in the AM context (see e.g., Ashtekar and Magnon 1984).
In the AM context, the condition BnbC 0 is too strong at 9 because it forces the Bondi
news to vanish; in particular, it is not stable against gravitational perturbations with,
say, initial data of compact support. In the present case, it seems reasonable since
there is no analogue of Bondi news in any case. Furthermore, now one does expect
the condition to be stable under gravitational perturbations: since the linearised field
equation now involves an effective ‘mass-term’ p 2 = - 2 4 one expects that these
perturbations will not affect the structure at 9. Finally, the condition fits in very well
with twistorial ideas and resolves certain issues concerning boundary conditions for
gauged supergravity (Hawking 1983). Therefore, following Hawking, we introduce:
Dejnition 2. A space-time (k,&b) will be said to be asymptotically anti-de Sitter if,
in addition to definition 1, one has:
(iv) The conformal group of (9,qab) is the anti-de Sitter group (or covering group
thereof).
For this class of space-times, we have ten ‘conserved’ quantities given by (1 1)
whose fluxes are completely determined by the matter terms (12). If the matter consists
of Yang-Mills fields only, it is natural to require that
B: := (-3/h)‘”F:bnb (16)
should also vanish on 9. This is a reflective boundary condition because it implies
that the right side of (12), which is otherwise arbitrary, should vanish, i.e. that there
is no leakage through 9. Thus, now QJC] of (1 1) are absolutely conserved. Further-
more, if B: P 0, the electric charges
can be defined at 9 without the ‘supertranslation ambiguities’ pointed out by Tafel
and Trautman (1983) in the AM context, and are absolutely conserved (Ashtekar 1984).
We will conclude by giving explicit expressions of the ten conformal Killing fields
which are useful in some calculations. First of all, definition 2 implies that 9 admits
a global chart U, 8, cp such that the metric qob is conformally related to q:b given by
dS2 - 13A. du2+de2 +sin2 8 dq2.
(18)
Clearly, qb: admits four Killing fields, a time translation? U”=aa/au, and three
rotations, EabCDbuDcLY, where LY is any linear combination of the first three spherical
t Hence, by condition (iv) in definition 2, it follows that the ‘generators’ of (9,q“,) are complete.
L44 Letter to the Editor
harmonics. The other six conformal Killing fields of q:b are
a [ Acos(-fA)'"u - (-3/A)'"B ~in(-$A)"~u]U"
- [ ( - f A ) ' / ' A sin(-fA)'12u + B c ~ s ( - f A ) ' / ~ u ] D " a (19)
where A and B are arbitrary constants. The Lie algebra of these ten vector fields does
not admit an Abelian four-dimensional subalgebra. Hence, there is no canonical
'translation subgroup' in the asymptotic symmetry group. Nonetheless, since 9 is
S2x R, time-like and conformally flat, it is analogous to the hyperboloid of unit
space-like directions in the tangent space of io of the AM space-times and using the
Spi framework as a guide, one can select four preferred vector fields out of these ten.
They are given by
U", [ a cos( - $ A ) ' / 2 ~ ]U" - ( -$A)'/'[sin( --fA)'/'u]D"a (20)
where, as before, a is a linear combination of K,'s. In the limit A + 0, the metric q t b
takes the canonical AM degenerate form and the vector fields (20) reduce to the BMS
translations. Therefore, one may regard these fields as the anti-de Sitter translations
and the corresponding conserved quantities Q)[C] as the components of the anti-de
Sitter four-momentum.
References
Abbot L F and Deser S 1982 Nucl. Phys. B 195 76
Ashtekar A 1984 in Roc. Oregon Con$ on Mass and Asymptotic Structure ofspace-time ed F Flaherty (Berlin:
Springer)
Ashtekar A and Magnon A 1984 J. Math. Phys. in press
Ashtekar A and Schmidt B 1980 J. Math. Phys. 21 862
Boucher W, Gibbons G W a n d Horowitz G T 1983 Preprint
Breitenlohner P and Friedman D Z 1982 Ann. Phys. 144 249
Choquet-Bruhat Y 1983 Les Houches Lectures (Preprint)
Geroch R 1977 in Asymptotic structure ofspace-time ed P Esposito and L Witten (New York: Plenum)
Gibbons G W, Hawking S W, Horowitz G T and Perry M J 1983 Commun. Math. Phys. 88 295
Hawking S W 1983 Phys. Left. B 126 175
Penrose R 1963 Phys. Rev. Lett. 10 66
- 1964 Proc. R. Soc. A 284 159
Tafel J and Trautman A 1983 J. Math. Phys. 24 1087
