NDA-FP-57

March 1998

On the conformal anomaly from higher

derivative gravity in AdS/CFT
correspondence

Shin’ichi NOJIRI1 and Sergei D. ODINTSOV♠2

Department of Mathematics and Physics

National Defence Academy, Hashirimizu Yokosuka 239, JAPAN

♠ Tomsk Pedagogical University, 634041 Tomsk, RUSSIA

abstract

We follow to Witten proposal [1] in the calculation of conformal anomaly

from d + 1-dimensional higher derivative gravity via AdS/CFT correspon-
dence. It is assumed that some d-dimensional conformal field theories have
a description in terms of above d + 1-dimensional higher derivative grav-
ity which includes not only Einstein term and cosmological constant but
also curvature squared terms. The explicit expression for two-dimensional
and four-dimensional anomalies is found, it contains higher derivative cor-
rections. In particular, it is shown that not only Einstein gravity but also
theory with the Lagrangian L = aR2 + bRµν Rµν + Λ (even when a = 0 or
b = 0) is five-dimensional bulk theory for d = 4 N = 4 super Yang-Mills the-
ory in AdS/CFT correspondence. Similar d + 1 = 3 theory with (or without)
Einstein term may describe d = 2 scalar or spinor CFTs. That gives new
versions of bulk side which may be useful in different aspects. As application
of our general formalism we find next-to-leading corrections to the conformal
anomaly of N = 2 supersymmetric theory from d = 5 AdS higher derivative
gravity (low energy string effective action).
1
e-mail: [email protected], [email protected]
2
e-mail: [email protected], [email protected]

1
1 Introduction. Conformal anomaly from
dilatonic gravity.
Recently so-called AdS/CFT (anti-de Sitter space vs. conformal field theory)
correspondence attracted a lot of attention [2, 3, 1]. In particulary, the cor-
respondence between correlation functions in AdS and those of the boundary
manifold was discussed.
D dimensional anti-de Sitter space can be realized by imposing a con-
straints on D + 1 coordinates:

−x20 + x21 + · · · x2D−1 − x2D = −L2 . (1)

From this realization, it is easy to see that this space has SO(D − 1, 2)
symmetry as isometry. The algebra of SO(D − 1, 2) symmetry is the same
as the algebra of conformal transformations acting on D − 1 dimensional
Minkowski space.
In an adequate coordinate choice, the metric on the D-dimensional anti-
de Sitter space is given by

X
D−2
ds2AdS = ρ−2 dρ2 + ρ−1 ηij dxi dxj . (2)
i,j=0

Here η is the metric on the flat D − 1-dimensional Minkowski space. We

should note that there is a boundary when ρ vanishes. The topology of
the boundary is almost that of the D − 1 dimensional Minkowski space, or
more exactly, Minkowski space with a point at infinity, that is, topologically
compactified Minkowski space. On the boundary manifold, SO(D−1, 2) acts
exactly as usual conformal transformation. When we consider the surface
with fixed finite ρ, there is a correction proportional ρ only in the conformal
boost but the correction vanishes just on the boundary, that is, in the limit
that ρ vanishes.
AdS/CFT correspondence is conjectured in ref.[2]. When N p-branes
in superstring theory or so-called M-theory coincide with each other and
the coupling constant is small, the classical supergravity on AdSD=d+1=p+2 ,
which is the low energy effective theory of superstring or M-theory, is, in some
sense, dual to large N conformal field theory on M d , which is the boundary
of the AdS. For example, d = 2 case corresponds to (4,4) superconformal

2
field theory, d = 4 case corresponds to U(N) or SU(N) N = 4 super Yang
Mills theory and d = 6 case to (0,2) superconformal field theory.
The conjecture tells that partition function in d-dimensional conformal
field theory is given in terms of the classical action in d + 1-dimensional
gravity theory:
Zd (φ0 ) = e−SAdS (φ 0 )
classical (φ )
. (3)
Here φ0 is the value of the field φ on the boundary and φclassical (φ0 ) is a
field on bulk background, which is AdS, given by solving the  equations of
d classical
motion with the boundary value φ0 on M . SAdS φ (φ0 ) is the classical
gravity action on AdS. When we substitute the classical solution into the
action, the action, in general, contains infrared divergences coming from the
infinite volume of AdS. Then we need to regularize the infrared divergence.
It is known that as a result of the regularization and the renormalization
(for the introduction to the renormalization in background field method see
ref.[10]) there often appear anomalies. In ref.[1] Witten made the proposal
how to calculate the conformal anomaly from classical gravity (bulk) side.
This proposal has been worked out in detail in ref.[4] (for Einstein gravity)
where it was shown that the conformal (Weyl) anomaly may be recovered
after regularizing the above infrared divergence (from bulk side- Einstein
theory). The usual result for conformal anomaly of boundary QFT thus may
be reproduced. This is a kind of IR-UV duality.
Since the calculation of ref. [4] was done without dilaton background, in
the previous paper [5], the authors investigate the conformal anomaly in the
non-trivial dilaton background. We start from the action of d+1-dimensional
dilatonic gravity with boundary terms
"Z q
1 n o
S = d d+1
x −Ĝ R̂ + X(φ)(∇φ) ˆ + 4λ2
ˆ 2 + Y (φ)∆φ
16πG Md+1
Z q  
+ d x −ĝ 2∇
d ˆ µ nµ + α . (4)
Md

Here Md+1 is d+1 dimensional manifold, which is identified with AdSd+1 and
nµ is the unit vector normal to the boundary manifold Md . α is a parameter
which is chosen properly. The boundary terms play a role for cancellation of
the leading infrared divergences which guarantee the system depends only on
the boundary value. Note that dilatonic gravity makes the beatiful realization

3
of two-boundaries AdS/CFT correspondence [15] (for other examples of such
two-boundaries correspondence see refs.[16, 17]).
X(φ), Y (φ) can be arbitrary functions of φ. The arbitrariness is not real
but apparent. In fact, by the redefinition,
Z q
ϕ≡ dφ 2V (φ) , V (φ) ≡ X(φ) − Y 0 (φ) (5)

the action can be rewritten as

Z q n o
1
S = d x −Ĝ R̂ + 2(∇ϕ)
d+1 ˆ 2 + 4λ2
16πG
Z q  
+ d x −ĝ 2∇µ n + α + Y (φ)n ∂µ φ .
d ˆ µ µ
(6)
Md

The last term on Md finally does not contribute to anomaly.

Following the calculation in [4], the explicit expression for the conformal
anomaly has been obtained
l n o
T = R(0) + X(φ(0) )(∇φ(0) )2 + Y (φ(0) )∆φ(0) (7)
16πG
for d = 2. This result could be compared with the UV-calculation of the
conformal anomaly of matter dilaton coupled N scalars [7, 6, 9] and M
Majorana dilaton coupled spinors [18] whose action is given by
Z ( )
1 √ X
N X
M
S= d x −g f (φ)g
2 µν
∂µ χi ∂ν χi + g(φ) ψ̄i γ µ ∂µ ψi . (8)
2 i=1 i=1

The results in (7) correspond to

l 2N + M
= ,
16πG 48π ! !
l N f 00 f 02 M g 00 g 02
X(φ(0) ) = − − 2 − − 2 ,
16πG 4π 2f 4f 24π g g
0 0
l N f M g
Y (φ(0) ) = − − . (9)
16πG 4π 2f 24π g
For d = 4, the expression of the conformal anomaly from bulk dilatonic
gravity has the following form:

l3 1 1 2
T = ij
R(0)ij R(0) − R(0)
8πG 8 24

4
 1 ij 1
+ R(0) ∂i ϕ(0) ∂j ϕ(0) − R(0) g(0)ij
∂i ϕ(0) ∂j ϕ(0)
2 6 
( )
1 1 q  2 1  2
+ √ ∂i −g(0) g(0)
ij
∂j ϕ(0) + g ij ∂i ϕ(0) ∂j ϕ(0)  (.10)
4 −g(0) 3 (0)

In [8], it has been calculated the conformal anomaly coming from the mul-
tiplets of N = 4 supersymmetric U(N) or SU(N) Yang-Mills coupled with
N = 4 conformal supergravity(in a covariant way). If we choose the length
parameter l as
l3 2N 2
= (11)
16πG (4π)2
and consider the background where only gravity and the real part of the
scalar field ϕ are non-trivial and other fields vanish, IR-calculation (10) from
the bulk exactly reproduces the result of UV-calculation by Liu and Tseytlin.
Since dilaton field always appears in string theory, the conformal anomaly
gives further check of AdS/CFT correspondence.
In this paper, we consider R2 -gravity, which contains the squares of the
curvatures in the action, as an extension of Einstein gravity. This is our bulk
theory (for a general review and list of refs. on quantum higher derivative
gravity see[10]). Moreover for some combinations of higher derivative terms it
corresponds to different compactifications of superstring and heterotic string
effective actions.
The paper is organised as follows. In the next section we present general
formalism for the calculation of conformal anomaly from bulk higher deriva-
tive gravity. As an example we give the explicit result for two-dimensional
CFT anomaly as obtained from three-dimensional higher derivative gravity.
Section three is devoted to the same calculation in four-dimensional case. We
find conformal d = 4 QFT (boundary N = 4 super Yang-Mills theory) which
has description in terms of five-dimensional higher derivative gravity. The
application of results of section 3 to evaluation of next-to-leading correction
to N = 2 supersymmetric theory conformal anomaly from bulk AdS higher
derivative gravity (low energy string effective action) is done in section 4.
The last section is devoted to brief description of some perspectives.

5
2 Conformal anomaly from bulk higher
derivative gravity. General formalism and
d2 case.
As an extension of the Einstein gravity, we consider R2 gravity (bulk theory),
which contains the squares of the curvatures in the action:
Z q  
1
S = d d+1
x −Ĝ aR̂ + bR̂µν R̂
2 µν
+ cR̂µνρσ R̂ µνρσ
+ 2 R̂ − Λ
κ
+SB (12)

Here Md+1 is d + 1 dimensional manifold whose boundary is d dimensional

manifold Md . The boundary term SB is necessary to make the variational
principle to be well-defined but we do not need the explicit expression of SB
since it does not contribute to the anomaly. Note that in Einstein gravity
the boundary terms may be used to present the action as functional of fields
and their first derivatives [12]. One can also add higher derivative boundary
terms.
Note that for some values of parameters above theory can be related with
superstring or heterotic string effective action. Superstring effective action
does not include R2 terms in d + 1 = 10 dimensions. However, making
Calabi-Yau compactification in lower dimensions one can get also R2 terms
(see for example [11]). We may consider as particular example of higher
derivative gravity the effective action of heterotic string [13, 14] which we
take for torus compactification and supposing all fields except the metric
tensor to be constant.
Z ( )
√ 1 α0
Shet = d d+1
x −g R − Λ + Rµνρσ Rµνρσ . (13)
κ2 8
The cosmological constant may be non-zero, for example for covariantly con-
stant gauge fields.
The conventions of curvatures are given by

R = g µν Rµν
Rµν = −Γλµλ,κ + Γλµκ,λ − Γηµλ Γλκη + Γηµκ Γλλη
1 ην
Γηµλ = g (gµν,λ + gλν,µ − gµλ,ν ) . (14)
2

6
 First we investigate if the equations of motion for the action (12) have a
solution which describes anti de Sitter space, whose metric is given by

l2 −2 Xd
ds2 = Ĝ(0) µ ν
µν dx dx = ρ dρdρ + ρ−1 ηij dxi dxj . (15)
4 i=1

When we assume the metric in the form (15), the scalar, Ricci and Riemann
curvatures are given by
d2 + d d (0) 1  (0) (0) 
R̂(0) = − , R̂ (0)
µν = − G µν , R̂ (0)
µνρσ = − G µρ G νσ − G (0) (0)
µσ G νρ ,
l2 l2 l2
(16)
which tell that these curvatures are covariantly constant. Then in the equa-
tions of motion from the action (12), the terms containing the covariant
derivatives of the curvatures vanish and the equations have the following
forms:
 
1 (0) 1
0 = − Gζξ aR̂(0)2 + bR̂µν
(0) (0)µν
R̂ (0)
+ cR̂µνρσ R̂(0)µνρσ + 2 R̂(0) − Λ
2 κ
(0) (0) (0) (0)µ (0) (0)µνρ 1 (0)
+2aR Rζξ + 2bR̂µζ R̂ξ + 2cR̂ζµνρ R̂ξ + 2 R̂ζξ . (17)
κ
Then substituting Eqs.(16) into (17), we find
a b
0 = 4
(d + 1)d2 (d − 3) + 4 d2 (d − 3)
l l
2c d(d − 1)
+ 4 d(d − 3) − −Λ . (18)
l κ2 l2
The equation (18) can be solved with respect to l2 if

d2 (d − 1)2
− 4d(d − 3) {(d + 1)da + db + 2c} Λ ≥ 0 (19)
κ4
which can been found from the determinant in (18). Then we obtain
q
d2 (d−1)2
d(d+1)
κ2
± κ4
− 4d(d − 3) {(d + 1)da + db + 2c} Λ
l =−
2
. (20)
2Λ
The sign in front of the root in the above equation may be chosen to be
positive what corresponds to the Einstein gravity (a = b = c = 0).

7
 The existence of the solution where the sign is negative tells that there
can be anti de Sitter solution even if the cosmological constant Λ is positive
since l2 can be positive if (d − 3) {(d + 1)da + db + 2c} < 0. The solution
describes the de Sitter or anti de Sitter space, depending on the overall sign of
l2 . l2 can be negative or positive depending from the choice of the parameters
and sign in above equation.
In order to calculate the conformal anomaly, we consider the fluctuations
around the anti de Sitter space (15). As in [4, 5], we assume the metric has
the following form:

l2 −2 Xd
ds2 ≡ Ĝµν dxµ dxν = ρ dρdρ + ĝij dxi dxj
4 i=1
ĝij = ρ−1 gij . (21)

We should note that there is a redundancy in the expression of (21). In fact,

if we reparametrize the metric :

δρ = δσρ , δgij = δσgij . (22)

by a constant parameter δσ, the expression (21) is invariant. The transfor-

mation (22) is nothing but the scale transformation on Md .
In the parametrization in (21) we find the expressions for the scalar cur-
vature R̂
d2 + d 2(d − 1)ρ ij 0 3ρ2 ij kl 0 0
R̂ = − + ρR + g g ij + g g gik gjl
l2 l2 l2
4ρ2 ρ2
− 2 g ij gij00 − 2 g ij g kl gij0 gkl
0
(23)
l l
for Ricci tensor R̂µν
d 1 1
R̂ρρ = − 2
− g ij gij00 + g ik g lj gkl
0 0
gij
4ρ 2 4
2ρ 2ρ ρ
R̂ij = Rij − 2 gij00 + 2 g klgki0 0
glj − 2 gij0 g klgkl
0
l l l
2−d 1 d
− 2 gij0 + 2 gij g klgkl 0
− 2 gij
l l l ρ
R̂iρ = R̂ρi

8
 1 jk 0 1 0 1 0
= g gki,j − g kj gjk,i + g,jjk gki
2 2 2
1 kl 0 jm 1 kj 0
+ g gli g gjm,k − g,i gjk (24)
4 4
and for the Riemann tensor R̂µνρσ

R̂ρρρρ = 0
R̂ρρρi = Rρρiρ = Rρiρρ = Riρρρ = 0
R̂ijρρ = R̂ρρij = 0
R̂iρjρ = R̂ρiρj = −R̂iρρj = R̂ρijρ
1 1 0 0 1
= − 3 gij + g kl gki glj − gij00
4ρ 4ρ 2ρ
R̂ρijk = −R̂iρjk = R̂jkρi = −R̂jkiρ
1 n 0 0
= 2gij,k − 2gik,j − g lm (gim,k + gkm,i − gik,m) glj0
4ρ
o
0
+g lm (gim,j + gjm,i − gij,m) glk
1
R̂ijkl = Rijkl (25)
ρ
1 n 0

0

0
 o
− 2 2 gjl − ρgjl (gik − ρgik ) − gjk − ρgjk (gil − ρgil0 )
ρl
Here “ 0 ” expresses the derivative with respect to ρ and R, Rij , Rijkl are
scalar curvature, Ricci and Riemann tensors, respectively, on Md .
As in the previous papers [4, 5], we expand the metric gij as a power
series with respect to ρ,
gij = g(0)ij + ρg(1)ij + ρ2 g(2)ij + · · · . (26)
Substituting (26) into (23), (24), and (25), we find the following expansions
with respect to ρ:
q 
l − d −1 q ρ ij
−Ĝ = ρ 2 −g(0) 1 + g(0) g(1)ij
2 2

1 ij 1 ij kl
+ρ2 g(0) g(2)ij − g(0) g(0) g(1)ik g(1)jl
2 4
1  ij 2  
3
+ g(0) g(1)ij + O(ρ )
8

9
 q (
l − d −1 q d2 + d
−ĜR̂ = ρ 2 −g(0) − 2
2 l
!
−d2 + 3d − 4 ij
+ρ R(0) + g(0) g(1)ij
2l2

1
+ρ −g(1)ij R(0)
2 ij
+ R(0) g(0) ij
g(1)ij
2
−d2 + 7d − 24 ij
+ g(0) g(2)ij
2l2
d2 − 7d + 20 ij kl
+ g(0) g(0) g(1)ik g(1)jl
4l2 ! )
−d2 + 7d − 16  ij 2
3
+ g(0) g(1)ij + O(ρ )
8l2
q (
l − d −1 q d2 (d + 1)2
−ĜR̂ = 2
ρ 2 −g(0)
2 l2
!
2d(d + 1) d4 − 6d3 + d2 + 8d ij
+ρ − R(0) + g(0) g(1)ij
l2 2l4
2d(d + 1) ij
+ρ2 R(0) 2
+ 2
g(1)ij R(0)
l
−d + 3d − 4
2
ij
+ R(0) g(0) g(1)ij
l2
d4 − 14d3 + 33d2 + 48d ij
+ g(0) g(2)ij
2l4
−d4 + 14d3 − 25d2 − 40d ij kl
+ g(0) g(0) g(1)ik g(1)jl
4l4 ! )
d4 − 14d3 + 49d2 − 32d + 32  ij 2
3
+ g(0) g(1)ij + O(ρ )
8l4
q (
l − d −1 q d2 (d + 1)
−ĜR̂µν R̂ µν
= ρ 2 −g(0)
2 l2
!
2d d3 − 7d2 + 8d ij
+ρ − 2 R(0) + g(0) g(1)ij
l 2l4
ij 4d − 4 ij
+ρ2 R(0) R(0)ij + 2
g(1)ij R(0)
l

10
 −d + 2 ij d3 − 15d2 + 48d ij
+ R g g
(0) (0) (1)ij + g(0) g(2)ij
l2 2l4
−d3 + 19d2 − 56d + 16 ij kl
+ g(0) g(0) g(1)ik g(1)jl
4l4 ! )
d3 − 15d2 + 56d − 32  ij 2
3
+ g(0) g(1)ij + O(ρ )
8l4
q (
l − d −1 q 2d(d + 1)
−ĜR̂µνρσ R̂µνρσ = ρ 2 −g(0)
2 l4
!
4 d2 − 7d + 8 ij
+ρ − 2 R(0) + g(0) g(1)ij
l 2l4

4 2
2
+ρ R(0) ijkl
R(0)ijkl + 2 g(1)ij R(0) ij
− 2 R(0) g(0) ij
g(1)ij
l l
d2 − 15d + 48 ij
+ g(0) g(2)ij
l4
−d2 + 23d − 56 ij kl
+ g(0) g(0) g(1)ik g(1)jl
2l4 ! )
d2 − 15d + 48  ij 2
3
+ g(0) g(1)ij + O(ρ ) . (27)
4l4

We regard g(0)ij in (26) as independent field on Md . We can solve g(l)ij

(l = 1, 2, · · ·) with respect to g(0)ij using equations of motion. When substi-
tuting the expression (26) or (27) into the classical action (12), the action
diverges in general since the action contains the infinite volume integration
on Md+1 . We regularize the infrared divergence by introducing a cutoff pa-
rameter (following to Witten [1]):
Z Z Z Z   Z  

d d+1
x→ d
d x dρ , d x ··· →
d
dd x · · ·  . (28)
 Md ρ=

Then the action (12) can be expanded as a power series of :

S = S0 (g(0)ij )− 2 + S1 (g(0)ij , g(1)ij )− 2 −1

d d

1
+ · · · + Sln ln  + S d + O( 2 ) . (29)
2

The term Sln proportional to ln  appears when d =even. In (29), the terms
proportional to the inverse power of  in the regularized action are invariant

11
under the scale transformation
δg(0)µν = 2δσg(0)µν , δ = 2δσ . (30)
The invariance comes from the redanduncy (22). The subtraction of these
terms proportional to the inverse power of  does not break the invari-
ance. When d is even, however, there appears the term Sln proportional
to ln . The subtraction of the term Sln breaks the invariance under the
transformation(30). The reason is that the variation of the ln  term under
the scale transformation (30) is finite when  → 0 since ln  → ln  + ln(2δσ).
Therefore the variation should be canceled by the variation of the finite term
S d (which does not depend on )
2

δS d = − ln(2σ)Sln (31)
2

since the original total action (12) is invariant under the scale transforma-
tion. Since the action S d can be regarded as the action renormalized by the
2
subtraction of the terms which diverge when  → 0, the ln  term Sln gives
the conformal anomaly T of the renormalized theory on the boundary Md :
1Z 2 q
Sln = − d x −g(0) T . (32)
2
First we consider d = 2 case. When d = 2, substituting (26) or (27) into
the action (12) and using the regularization (28), we obtain
Z q  
3a
Sln = dx 2
−g(0) g(0)
ij
g(1)ij −
)
l3
b c 1 lΛ
− − − −
l3 l3 2lκ2 4
( )#
6a 2b 2c 1
+R(0) − − − + 2 . (33)
l l l 2κ
Since Eq.(18) gives the following equation for d = 2
12a 4b 4c 2
0=− 4
− 4 − 4 − 2 2 −Λ , (34)
l l l κ l
ij
we find the term proportional to g(0) g(1)ij in (33) vanishes. Then using (32),
we find the expression of Weyl anomaly T
( )
l
T = 12a + 4b + 4c − 2 R(0) . (35)
κ

12
When a = b = c = 0, the above equation reproduces the result of refs.[4, 5]:
l
T =−
R(0) . (36)
κ2
For the heterotic string effective action (13), we obtain
( )
α0 l
Tstring = − 2 R(0) . (37)
2 κ
The important result which follows from above consideration is the fol-
lowing: two-dimensional CFT may have description not only in terms of
three-dimensional Einstein gravity (bulk theory) but also in terms of three-
dimensional higher derivative gravity where Einstein term may be absent (but
cosmological term is present). Choosing the coefficients a, b, c in Eq.(35) so
that the correspondent conformal anomaly would give the anomaly for two-
dimensional scalar or spinor CFT we get new examples of AdS/CFT corre-
spondence. It is interesting that in this case new theory appears on the bulk
side. Note that one can consider higher derivative gravity with Einstein term
but without of cosmological term as the bulk theory.

3 Conformal anomaly for d = 4 theory

In this section, we consider d = 4 case. Substituting the expressions in (27)
into the action (12) and using the regularization in (28), we find
Z q h  
1
Sln = − dx4 −g(0) l aR(0)
2 ij
+ bR(0) ijkl
R(0)ij + cR(0) R(0)ijkl
2 !
40a 8b 4c 6 lΛ ij
+ 3
+ 3 + 3 − 2− g(0) g(2)ij
l l l lκ 2
!
40a 12b 12c l
+ + + − 2 g(1)ij R(0) ij
l l l κ
!
8a 2b 2c l
+ − − − + 2 R(0) g(0) ij
g(1)ij
l l l 2κ
!
20a 8b 10c 2 lΛ ij kl
+ + 3 + 3 + 2+ g(0) g(0) g(1)ik g(1)jl
l3 l l lκ 4
! #
6a 2b c 1 lΛ  ij 2
+ + 3 + 3− − g(0) g(1)ij (38)
l3 l l 2lκ2 8

13
Other terms proportional −1 or −2 which diverge when  → 0 can be sub-
tracted without loss of the general covariance and scale invariance. Since
Eq.(18) has the following form in d = 4,

80a 16b 8c 12
0= 4
+ 4 + 4 − 2 2 −Λ , (39)
l l l κ l
ij
we find that the term in (38) proportional to g(0) g(2)ij vanishes. The equation
given by the variation over g(1)ij is given by
ij ij ik jl ij kl
0 = AR(0) + Bg(0) R(0) + 2Cg(0) g(0) g(1)kl + 2Dg(0) g(0) g(1)kl
40a 12b 4c l
A ≡ + + − 2
l l l κ
8a 2b 2c l
B ≡ − − − + 2
l l l 2κ
20a 8b 10c 2 lΛ 40a 12b 12c 1
C ≡ + + + + = + + −
l3 l3 l3 lκ2 4 l3 l3 l3 lκ2
6a 2b c 1 lΛ 4a 1
D ≡ 3 + 3 + 3− 2
− =− 3 + 2 . (40)
l l l 2lκ 8 l lκ
Here we used (39) to re-write the expressions of C and D and to remove the
cosmological constant Λ. Multiplying g(1)ij with (40), we obtain

A + 4B
ij
g(0) g(1)ij = − R(0) . (41)
2(C + 4D)

Substituting (41) into (40), we can solve (40) with respect to g(1)ij as follows:

A AD − BC
g(1)ij = − R(0)ij + R(0) g(0)ij . (42)
2C 2C(C + 4D)

Substituting (42) into (38) (and using (39)), we find the following expression
for the anomaly T
!
A2 D − 4B 2 C − 2ABC 2
T = al + R(0)
4C(C + 4D)
!
2
A
+ bl − ij
R(0)ij R(0) ijkl
+ clR(0)ijkl R(0) . (43)
4C

14
If we put a = b = c = 0, the above expression reproduces the result in [4, 5]:

l3 l3
T =− R 2
+ ij
R(0)ij R(0) . (44)
12κ2 (0) 4κ2
For general case, by substituting (40) into (43), we obtain
!
l3 cl
T = − 2 + 5al + bl (G − F ) + (G + F ) . (45)
8κ 2

Here we used the Gauss-Bonnet invariant G and the square of the Weyl
tensor F , which are given by
2
G = R(0) − 4R(0)ij R(0)
ij ijkl
+ R(0)ijkl R(0)
1 2
F = R(0) − 2R(0)ij R(0)
ij ijkl
+ R(0)ijkl R(0) . (46)
3
This is really beautiful and compact result which shows that apparently
any variant of higher derivative AdS gravity has correspondence with some
d = 4 conformal field theory. It is very general and it maybe used in var-
ious contexts as we see in next section. It maybe also used to search new
examples of bulk/boundary correspondence with various higher derivative
supergravities on bulk side.
Consider one particular example:

l3 2N 2
c=0, − 40al − 8bl = , (47)
κ2 (4π)2
Then, the anomaly of D = 4 N = 4 super Yang-Mills is reproduced.
It is very interesting that last relation is still true even in case of ab-
sence the Einstein term there. In fact, in this case we found new example
of AdS/CFT correspondence where boundary QFT is well-known N = 4
super Yang-Mills theory and bulk side is described by higher derivative five-
dimensional gravity with cosmological term and without (or with) the Ein-
stein term. Moreover, one can consider more narrow class of theories with
a = 0 or b = 0 in initial classical action.
It should be also noted that one cannot consider pure higher derivative
gravity in above context because without cosmological and Einstein terms
the theory is scale invariant. Then it presumbly cannot be useful as bulk

15
side in AdS/CFT correspondence. In our approach the manifestation of this
fact is that for pure R2 -gravity the conformal anomaly is zero.
Finally we note that when only c and cosmological constant are not zero
among the parameters of AdS higher derivative gravity as it follows from
general Eq.(45) bulk side does not correspond to N = 4 super Yang-Mills
theory.

4 Next-to-leading corrections to N = 2 su-

perconformal theory trace anomaly from
AdS/CFT correspondence
Our main motivativation in this work was to find new higher derivative ver-
sions of bulk theory in AdS/CFT correspondence via calculation of holografic
conformal anomaly. However, our general result maybe applied perfectly well
for another formulations (or motivation): when higher derivatives terms in
AdS gravity appear as next-to- leading corrections in low-energy string effec-
tive action (after compactification). Let us demonstrate this on the example
of N = 2 SCFT (and corresponding bulk AdS higher derivative gravity).
Recently (after the first version of this paper appeared in hep-th ), the
trace anomaly of d = 4, N = 2 and N = 4 SCFT has been investigated up
to next to leading order in the 1/N expansion by Blau, Narain and Gava
[22]. For an N = 2 SCFT theory with nV vector multiplets and nH hyper-
multiplets, the trace anomaly is, by usual UV calculations, given by
1 h 1
T = − (11nV + nH )R2
24 · 16π 2 3 i
+12nV Rij R + (nH − nV )Rijkl Rijkl ,
ij
(48)

and especially, when the gauge group is Sp(N),

h 
1 1
T = − 8N 2 + 6N − R2
24 · 16π 2 3 i
+(24N 2 + 12N)Rij Rij + (6N − 1)Rijkl Rijkl . (49)

The N = 2 theory with the gauge group Sp(N) arise as the low-energy
theory on the world volume on N D3-branes sitting inside 8 D7-branes at an

16
O7-plane [23]. The string theory dual to this theory has been conjectured
to be type IIB string theory on AdS5 × X 5 where X5 = S 5 /Z2 [24], whose
low energy effective action is given by (see the corresponding derivation in
ref.[22])
Z ( )
√ N2 6N
S= 5
dx G (R − 2Λ) + Rµνρσ Rµνρσ . (50)
AdS5 4π 2 24 · 16π 2

In order to use the general result for the conformal anomaly T in (45), we
compare (50) with (24) and find
1 N2 6N
= , a=b=0, c= . (51)
κ2 4π 2 24 · 16π 2
Futhermore we need to choose
12 12N 2
Λ=− = − , (52)
κ2 4π 2
which corresponds to that the length scale is chosen to be unity in [22].
We should note, however, the length scale l in our paper has a correction.
Substituting (51) and (52) into (39), we find
1 2cκ2
= 1 +
l2 3l2
2cκ2  
= 1+ + O (cκ2 )2 . (53)
3
1
In the second line in (53), we have assumed cκ2 = 16N
is small or N is large.
Then we find the following expression for T :
 
1 c c 1  
T = − + (G − F ) + (G + F ) + O (cκ) 2
8κ2 8 2 κ2
2  
N 1 2
= 2
− R(0) ij
+ R(0)ij R(0)
16π 3
 
6N 3 2 13
+ R − R R ij
(0)ij (0) + R R ijkl
(0)ijkl (0) + O(1) . (54)
24 · 16π 2 4 (0) 4
In the first line in (54), we have substituted (53) and put a = b = 0 and in
the second line, (51), (52) and the explicit expression for G and F in (46)
are substituted. Eq. (54) exactly reproduces the result in (48) as in [22].

17
 Hence, we demonstrated how to apply our general formalism to calculate
next-to-leading order corrections to conformal anomaly from bulk AdS side.
Similarly, one can get conformal anomaly from bulk side for any other low-
energy string effective action.

5 Discussion
In summary, we studied higher derivative gravity as bulk theory in AdS/CFT
correspondence. We explicitly calculated the conformal anomaly from such
theory following to Witten proposal and assuming that they describe some
boundary QFTs. This approach opens new lines in AdS/CFT correspondence
as it presents new versions of bulk theory. For example, three-dimensional
R2 gravity with cosmological constant and with (or without) Einstein term
may describe d = 2 scalar or spinor CFT. In the same way, five-dimensional
R2 gravity with cosmological constant and with (or without) Einstein term
may be considered as bulk theory for N = 4 super Yang-Mills quantum
theory. Pure R2 gravity seems to be not good as bulk theory due to its scale
invariance (no mass scale). We also applied our general result to calculation
of next-to-leading corrections to conformal anomaly of N = 4 SCFT from
bulk AdS higher derivative gravity. Perfect agreement with results of very
recent work [22] is found.
There are also interesting possibilities to extend the results of present
work. In the recent paper [19] it was studied the conformal anomaly of
submanifold observables in AdS/CFT correspondence. Such investigation is
expected to be useful for the study of Wilson loops in large N theories in
frames of bulk-boundary correspondence[20]. It is clear that modification of
bulk theory (say of the form we discuss in this work) will also modify the
conformal anomaly on submanifolds.
In the recent paper[21] the form of Witten proposal has been applied to
study the boundary stress tensor of anti-de Sitter Einstein gravity in d+1 = 3
and d + 1 = 5 dimensions via AdS/CFT correspondence. As by-product the
conformal anomaly in d = 2 and d = 4 is also recovered. It would be
of interest to re-consider this problem for our version of higher derivative
gravity as it may lead to acceptable definition of the gravitational energy
and stress tensor even in such case.
For example, the Casimir energy of d = 4 SU(N) N = 4 super Yang-Mills

18
theory on S 3 × R which is global boundary of five-dimensional AdS is given
by
3(N 2 − 1)
ECasimir = . (55)
16r
√
Here r is the radius of S 3 . By multiplying −gtt = r/l, the Casimir mass is
given by
3(N 2 − 1)
MCasimir = , (56)
16l
which gives the non-vanishing ground state energy for AdS5 , i.e. correspond-
ing BH mass [21]. Using eq.(47) as the conversion formula to gauge theory
variables from higher derivative gravity we get AdS mass for higher derivative
gravity suggesting the way to construct stress-energy tensor in such theory.
As the final remark we would like to point out that Eq.(45) maybe used
in the constructive way in order to find new examples of bulk/boundary
relations. Let us imagine that we got some variant of low-energy string
effective action (after some compactification) as particular version of d = 5
AdS higher derivative gravity. It may also come in another way as bosonic
sector of some higher deivative AdS supergravity. Eq.(45) gives holografic
conformal anomaly (up to next-to-leading terms) from bulk side. Having such
conformal anomaly maybe often enough to suggest new boundary CFT.

Acknowledgments. We thank M. Blau, K. Narain and E. Gava for pointing

out mistake in the first version of this work.

References
[1] E. Witten, Adv.Theor.Math.Phys. 2 (1998) 505, hep-th/9802150.
[2] J.M. Maldacena, Adv.Theor.Math.Phys. 2 (1998) 231, hep-th/9711200.
[3] S. Gubser, I. Klebanov and A. Polyakov, Phys.Lett. B428 (1998) 105,
hep-th/9802109.
[4] M. Henningson and K. Skenderis, J.High Energy Phys. 9807 (1998) 023
[5] S. Nojiri and S.D. Odintsov, Phys.Lett. B444 (1998) 92.
[6] S. Nojiri and S.D. Odintsov, Phys.Rev. D57 (1998) 2363.

19
 [7] E. Elizalde, S. Naftulin and S.D. Odintsov, Phys.Rev. D49 (1994) 2852.

[8] H. Liu and A.A. Tseytlin, Nucl.Phys. B533(1998) 88.

[9] R. Bousso and S. Hawking, Phys.Rev. D56 (1997) 7788; S. Nojiri and
S.D. Odintsov, Mod.Phys.Lett. A12 (1997) 2083; T. Chiba and M. Siino,
Mod.Phys.Lett. A12 (1997) 709; S. Ichinose, Phys.Rev. D57 (1998)
6224; W. Kummer, H. Liebl and D.V. Vasilevich, Mod.Phys.Lett. A12
(1997) 2683; J.S. Dowker, Class.Quant.Grav. 15 (1998) 1881

[10] I.L. Buchbinder, S.D. Odintsov and I.L. Shapiro, Effective Action in
Quantum Gravity, IOP Publishing, Bristol and Philadelphia 1992.

[11] I. Antoniadis, E. Gava and K.S. Narain, Nucl.Phys. B393 (1992) 93; I.
Antoniadis, S. Ferrara, R. Minasian and K.S. Narain, Nucl. Phys. B507
(1997) 571.

[12] G.W. Gibbons and S.W. Hawking, Phys.Rev. D15 (1977) 2752

[13] C. Callan, D. Friedan, E. Martinec and M.J. Perry, Nucl.Phys. B262

(1985) 593; Nucl.Phys. B278 (1986) 78; D. Gross, J.A. Harvey, E. Mar-
tinec and R. Rohm, Nucl.Phys. B267 (1986) 75.

[14] A. Tseytlin, Nucl.Phys. B467 (1996) 383

[15] S. Nojiri and S.D. Odintsov, Phys.Lett.B to appear; hep-th/9812017

[16] H.J. Boonstra, K.S. Skenderis and P.K. Townsend, hep-th/9807137.

[17] K. Behrndt, hep-th/9809015.

[18] P. van Nieuwenhuizen, S. Nojiri and S.D. Odintsov, hep-th/9901119.

[19] C.R. Graham and E. Witten, hep-th/9901021.

[20] J. Maldacena,Phys.Rev.Lett. 80 (1998) 4859, S.-J. Rey and J. Yee, hep-

th 9803001.

[21] V. Balasubramanian and P. Kraus, hep-th/9902121.

[22] M. Blau, K.S. Narain and E. Gava, hep-th/9904179.

20
[23] A. Sen, Nucl.Phys. B475 (1996) 562, hep-th/9605150; T. Banks, M.R.
Douglas and N. Seiberg, hep-th/9605199; O. Aharony, C. Sonnenstein,
S. Yankielowicz and S. Theisen, hep-th/9611222; M.R. Douglas, D.A.
Lowe, J.H. Schwarz, hep-th/9612062.

[24] A. Fayyazuddin and M. Spalinski, Nucl.Phys. B535 (1998) 219, hep-

th/9805096; O. Aharony, A. Fayyazuddin and J.M. Maldacena, JHEP
9807 (1998) 013, hep-th/9806159.

21