Three-Wave and Four-Wave Interactions in the 4d Einstein

Gauss-Bonnet (EGB) and Lovelock Theories

Claudio Corianò,1 Mario Cretì,1, 2 Stefano Lionetti,1 and Matteo Maria Maglio3
1
Dipartimento di Matematica e Fisica, Università del Salento
and INFN Sezione di Lecce, Via Arnesano 73100 Lecce, Italy
National Center for HPC, Big Data and Quantum Computing
2
Center for Biomolecular Nanotechnologies,
Istituto Italiano di Tecnologia, Via Barsanti 14, 73010 Arnesano, Lecce, Italy
3
Institute for Theoretical Physics (ITP), University of Heidelberg
arXiv:2302.02103v1 [hep-th] 4 Feb 2023

Philosophenweg 16, 69120 Heidelberg, Germany

We derive the conformal constraints satisfied by classical vertices of a (Einstein) Gauss-Bonnet
theory around flat space, in general dimensions and at d = 4 (4d EGB). In 4d EGB they are obtained
by a singular limit of the integral of the Euler-Poincarè density. Our analysis exploits the relation
between this theory and the conformal anomaly action, which allows to uncover some interesting
features of the GB vertex at cubic and quartic level. If we introduce a conformal decomposition of
the metric, the resulting theory can be formulated in two different versions, which are regularization
dependent, a local one which is quartic in the dilaton field, and a nonlocal one, with a quadratic
dilaton. The nonlocal version is derived by a finite redefinition of the GB density with the inclusion
of a (d − 4)R2 correction, before performing the singular d → 4 limit. In the local version of the
theory, we show how the independent dynamics of the metric and of the dilaton are interwinded by a
classical trace identity. Three-gravitational wave interactions can be organised in a nontrivial way by
using directly the nonlocal 4d EGB version of the theory. This is possible thanks to the consistency
of such formulation - only up to 3-point functions - directly inherited from the conformal anomaly
(Riegert) action. The constraints satisfied by the vertices are classical, hierarchical Ward identities.
At quartic level, similar relations are derived, borrowing from the analysis of the counterterms of the
4T correlators of the conformal anomaly action, as defined by a perturbative expansion. For d 6= 4
these constraints hold also for Lovelock actions. They can be extended to higher order topological
invariants in such class of theories.

I. INTRODUCTION not that, in four dimensions, the theory is topological.

The search for modifications of Einstein’s theory of
General Relativity (GR) that may explain important
phenomenological aspects of current cosmology, such
as inflation and dark energy, follows several directions.
One of them, traditionally, contemplates the inclusion
of extra scalar fields into the theory. Such a role is
taken by a scalar that drives the metric inflation and
couples to all of matter present around the Planck
scale, finally decaying into the spectrum of particles,
parents of the Standard Model ones. Another mod-
ification is the inclusion of a cosmological constant,
which fits very well the CMB data within the ΛCDM
model, but underscores a huge hierarchy problem.
An interesting class of modified cosmologies are those
that include higher powers of the curvature of space-
time, the Riemann tensor, but in a form in which no
dimensionful coupling is present in the action and ex-
hibiting equations of motion of the second order, as
for the Einstein-Hilbert (EH) action.
A nice example of these is the Einstein Gauss-Bonnet
(EGB) theory at finite GB coupling, which would be
ideal for the study of some of these phenomena, were it
In string theory in d = 10 such quadratic corrections
get combined in the GB term only in the heterotic case
[1]. It was observed that they are deprived of double
poles, generated by the second functional derivative
√
of this term ( gE (2) )µνρσ , from the quadratic metric
fluctuations around flat space (∼ O(h2 )) [2].
We recall that topological contributions in the form of
either the Einstein-Hilbert (EH) action at d = 2
√
Z
VEH (g, d) ≡ µ dd x gR, (I.1)
 = d − 2, or the Gauss Bonnet action (GB) at d = 4,
define evanescent terms in the equations of motion of
gravity. In d = 2 the EH action itself is metric inde-
pendent.
Evanescent terms can be turned into dynamical con-
tributions by performing a singular limit on the corre-
sponding coupling constant, which are dimensionless.
This features is commonly present and held into ac-
count in the context of conformal anomaly actions,
but recently it has been reproposed in a purely clas-
sical context [3]. In the case of d = 2, the theory is
rendered dynamical by replacing the EH action by the
regulated action [4]
R d
d x (VEH (g, d) − VEH (ḡ, d))
S2 = lim , (I.2)
d→2 d−2
where one introduces a conformal decomposition of
the metric
gµν = ḡµν e2φ (I.3)
in terms of a dilaton (Weyl) factor φ and a fiducial
metric ḡ, with VEH (ḡ, d) being a subtraction that al-
lows a finite limit. The limiting procedure allows to
generate special forms of dilaton gravities, which are
closely related to conformal anomaly actions, and are
of Horndeski type. The differences among the possi-
ble realizations of such actions are related to the ways
in which the subtractions are included in the regular-
ization of such VEH (g, d)/ → 0/0 contributions as
 → 0.
The method, well-known both in the case of 2-D grav-
ity and of conformal anomaly actions [6–9], where the
same procedure is applied to the counterterms in the
dimensional regularization (DR) of the theory, has re-
cently regained significant attention, for offering, pos-
sibly, a way to evade Lovelock’ s theorem [10] in a
purely classical framework [3].
Eq. (I.2) identifies the ordinary Wess-Zumino (WZ)
form of the action. Different subtractions, defined ei-
ther in d dimensions - as in (I.2) - or at d = 4, allow to
include or exclude extra - Weyl invariant - terms. It
has been pointed out that this arbitrariness is a pos-
sible way to account for the difference between topo-
logical and non topological anomalies, or anomalies of
types A and B [5]. Both types of anomalies appear in
the trace of correlation functions involving stress en-
ergy tensors (T ), but the first are not associated with
the breaking of scale invariance.
Similarly to the d = 2 case, in the case of the 4d GB
theory, the coupling gs (d) becomes singular as d → 4,
while the integral of the Euler-Poincarè density be-
comes topological, generating again a 0/0 contribu-
tions that requires a subtraction. The result is a fi-
nite, non topological action, whose structure depends
on the subtraction [11].
The d → 4 limit is purely geometrical, but involves
additional scales, coming from the dependence of the
fields on the extra (d − 4) coordinates of the manifold,
and borrows its features from dimensional regulariza-
tion (DR) in flat Minkowski space. As a theory, it
is unrelated to the regularization of some quantum
corrections, as is the case of the conformal anomaly
action, derived by integrating out a conformal matter
sector. However, it is a well-defined variant that opens
the way to new interesting developments. It is clear
2
that such behaviour can be classified as a strongly
coupled one, geometrically induced and with a dimen-
sional reduction of the field dynamics. It can be envi-
sioned for every topological term of a general Lovelock
theory.
Lovelock’s theorem states that, at d = 4, the only
gravitational action that generates second order equa-
tions of motion is the EH action, plus a cosmological
constant
√
Z
SEH = dd x g(MP2 R + 2Λ). (I.4)
Its generalization to higher dimensions takes the form
[10]
n! [µ1
L(n) = δ · · · δνµnn ] Rµ1 µ2ν1 ν2 Rµ3 µ4ν3 ν4 · · ·
2n/2 ν1
· · · Rµn−1 µνnn−1 νn , n = 0, 2, 4, . . . . (I.5)
n = 0 identifies the cosmological constant, n = 2
the EH action and n = 4 the GB density. Given
the antisymmetrization present in its definition, this is
nonzero only in specific dimensions. Once the indices
are contracted, the resulting density and its coordi-
nate dependence can be extended to any dimension.
The reduction of the action to the topological dimen-
sion d = 4 for E (E4 ) is investigated by an embedding
of the metric into the extra (d − 4) dimensions, a pro-
cedure which is not unique [11]. The case n = 4, with
d = 4, defines the GB density
√
Z
VE (g, d) ≡µε dd x −g E, (I.6)
where µ is a renormalization scale and E is the inte-
grated Euler-Poincarè density
E = R2 − 4Rµν Rµν + Rµνρσ Rµνρσ , (I.7)
whose inclusion modifies the EH action just by bound-
ary contributions, since in an ordinary EGB theory
SEGB = SEH + gs VE , (I.8)
the GB term is evanescent at d = 4. Its contribution
to the gravitational equation of motion
1

1

Rµν − gµν R + Λ0 gµν +gs (VE (d))µν = 0, (I.9)
κ 2
explicitly given by
√

δVE 1 µν
VEµν ≡ = −g g E4 − 2Rµαβγ Rαβγ
ν
+
δgµν 2 cence, one can perform a similar singular rescaling of
 the coupling
µα ν µανβ µν
4R R α + 4R Rαβ − 2RR , (I.10)
vanishes at d = 4 if we use (I.7). In d > 4, VE is
not a boundary term, and is indeed contemplated by
Lovelock’ s theorem as a possible modification of the
EH action (see [12, 13]). We are going to investigate
the conformal constraints associated with this term in
d dimensions, that have not been investigated before.
A similar singular limit can be performed at d = 6 for
the topological invariant E6 , cubic in the curvature,
extending the strategy discussed in this work.
As we have already discussed in the introduction,
such constraints are a natural consequence of the role
played by such terms in the context of conformal
anomaly actions in every even dimensions.
As mentioned, when expanded around a flat spacetime
gµν = ηµν + κhµν , the operators of highest derivatives
( 2 ) of the GB action, contributing to the quadratic
term in the action (h 2 h), cancel out, showing that
the theory is free of ghosts. We recall that at quadratic
order, the contribution to an action containing the
Riemann tensor and its contractions, with arbitrary
combinations, is affected by a propagator with double
poles in the form
Z
√ elaborating on various previous analysis of the confor-
dd g (Rµνρσ )2 + a1 (Rµν )2 + a2 R2 =


√
Z
1 tities (CWIs) derived for the SW Z theory carry the
dd x g (a1 + 4)hµν 2 hµν + (a2 − 1)h 2 h +


4 same structure of the anomalous CWIs characterizing
O(h3 ) (I.11)
that vanish if a1 and a2 are chosen to reproduce the
Euler-Poincarè density.
For general metric background, the analysis of the be-
haviour of such actions can be performed starting from
a conformal decomposition
gµν = ḡµν e2φ (I.12)
and eliminating the dilaton using its expression in
terms of the entire metric g. The theory becomes non-
local if the dilaton is removed from the spectrum, as
suggested for the conformal anomaly action in [14].
One relies on integrable conformal decompositions,
such as the one discussed in [15], recently investigated
in connection with the perturbative hierarchy of the
conformal Ward identities for a specific 4-point func-
tion in [16]. Different nonlocal forms of such actions
are possible, which differ - rather nontrivially- by dif-
3
ferent possible inclusions of Weyl invariant terms. As
in the d = 2 case, in order to bring the dynamics of
VE down to d = 4 from d > 4, bypassing its evanes-
gs µ
gs VE → gs (d)VE gs (d) ≡ (I.13)
d−4
in order to remodulate gs VE as a 0/0 contribution.
Obviously, the definition of the 4d, (d = 4) singu-
lar limit of the GB theory, requires a specific com-
pactification, which depends on the underlying geom-
etry and is, in general, affected by extra Kaluza-Klein
modes. We are essentially performing an infinite cou-
pling limit (gs → gs (d)) on the GB term, as we ap-
proach the dimension at which the GB contribution is
topological. These variant theories should be seen as
classical modifications of the EH action that resolve
the evanescence of a certain topological term in spe-
cific even dimensions.
Compared to the ordinary Lovelock’s classification of
pure gravity actions yielding equations of motion of
second order, this procedure is essentially new. As
mentioned, it can be performed in any even dimen-
sion, starting with d = 2.
In this work we are going to provide the expression
of the nonlocal EGB action expanded up to quartic
order (4-graviton vertex) in the fluctuations around
a classical metric background. The result is derived
mal anomaly actions, adapted and simplified for 4d
EGB theories. Notice that the conformal Ward iden-
the quantum anomaly action [17], where vertices of the
VE term defined in (I.6), obtained by differentiating
this functional, are constrained by the fundamental
symmetries.
The main difference between a classical and a quan-
tum approach lays in the fact that the constraints ob-
tained by the procedure either apply to classical ver-
tices - for 4d EGB theories - or to quantum averages of
correlation functions, if the analysis is performed in a
quantum context. In this second case, the classical ac-
tion in the functional integral So (χi , g), where the χi ’s
are generic conformal field, is decomposed in terms of
the two components ḡ and φ of (I.12), corresponding
to the fiducial metric and the dilaton field. A simi-
lar decomposition can be introduced for the GB term,
with equations of motion that are constrained by the
"anomaly" of VE , defined below in (II.10), which is
Weyl variant for general d.
 A. Singular rescalings and finite subtractions
The singular rescaling of the coupling can be applied
to any topological term, such as E4 , E6 and so on. In
practice, the method is sufficient in order to regulate
the 0/0 limit of the d = 4 action, though the result,
as we have mentioned, depends on the geometry of
the compactification. In practical terms, the resulting
actions are usually simplified, by neglecting the de-
pendence of the metric on the extra coordinates while
they reduced them to d = 4. This is a procedure that,
even if not stated explicitly in the literature, is essen-
tially based on dimensional reduction (DRed), in the
form described in [11]. Conformal anomaly effective
actions, to which 4d GB models are related, are de-
rived by a similar procedure, applied to the GB (VE )
and the Weyl tensor squared (VC 2 ) counterterms. The
latter, in the case of a 4d EGB, is unnecessary.
The topological evanescence of the VE contribution is
lifted by the procedure, but some ambiguities are en-
countered, due to the non-unique choice of the back-
ground metric against which VE is calculated, which
remains an indetermination of the method [11].
The derivation of the geometric effective action de-
pends on the specific choice of the fiducial metric and
of the subtraction term, here identified in the form of
a Wess-Zumino action [18] via the conformal decom-
position, although other subtractions are possible. A
discussion of this point can be found in [11].
The regularization of the action that results from
(I.13) is not uniquely defined, since the DRed pro-
cedure is naturally affected by an integration cutoff.
The singular limit of the GB term is investigated by a
Weyl rescaling of this term in d 6= 4, which introduces
a dilaton in the spectrum, and the  = d − 4 → 0 ex-
pansion is performed afterwards, accompanied by the
DRed procedure.
B. Content of this work
The goal of our work is to identify the constraints sat-
isfied by the classical vertices of the theory, obtained
by removing the dilaton from the spectrum, and re-
sorting to a nonlocal description of the 4d GB theory.
The local version of such theory, which is given by
dilaton gravity, therefore, is replaced by a nonlocal
theory when a finite (classical) renormalization of the
GB interaction, proportional to an R2 term, is added
to the usual GB action, which is allowed by the sin-
gular limit .
We identify the constraints satisfied by the vertices of
such nonlocal action once its expression is expanded
around flat space. These correspond to a set of Ward
4
identities which are naturally satisfied by the Weyl-
variant part of a renormalized anomaly action, due
to the similarity between such action, which describes
the Weyl/conformal anomaly, and the 4d EGB theory.
We provide a description of such vertices up to quar-
tic order in the fluctuations around a flat background.
The conformal constraints satisfied by the vertex de-
rived from VE (g, d) are valid in d dimension due to the
fact that Weyl variation of this term given by (II.10)
is exactly linear in (d − 4).
At d = 4 the evanescence of the term, as already
pointed out, is removed by the inclusion of a subtrac-
tion, corresponding to a classical renormalization, and
the conformal constraints remain valid once we replace
VE (g, d) by the regulated vertex V̂E0 , which describes
the WZ form of the action.
Section 4 contains a first principle discussion of the
constraints on the equations of motion found once
we perform a conformal decomposition, showing that
they are a rigorous consequence of symmetry (II.1) .
The latter is broken by the subtraction term intro-
duced in the definition of V̂E0 . Also in this case, the
corresponding constraint, given in (IV.1), is naturally
borrowed from the case of the conformal anomaly ac-
tions [11].
II. THE LOCAL EGB THEORY AND THE
NONLOCAL ACTION
The correctly regulated theory takes the form of a
Wess-Zumino (WZ) action, which depends on the reg-
ularization procedure and the treatment of the dila-
tonic field (φ). The 0/0 regularization follows closely
the 2D case, where the Einstein-Hilbert term is also
topological, and the limit is performed by redefining
the coupling as α → α/(d − 2).
In general, dilaton effective actions may contain solu-
tions with the conformal factor that needs to be sta-
bilized around a certain scale f . Such a scale is the
conformal breaking scale. The scale (f ) is required in
order to redefine the dimensionless conformal factor
of the metric g in the conformal decomposition (II.1).
The local shift symmetry, which allows to identify a
fiducial metric and the dilaton field, via the transfor-
mation
φ → φ − σ, ḡµν → ḡµν e2σ (II.1)
with σ = σ(x), is indeed broken by the regularization
of the Lagrangian in the 0/0 limit. This issue is not
present in the nonlocal action, since φ can be elimi-
nated in terms of the entire metric, but, as we have
already mentioned, one needs to perform an additional
finite renormalization of the action [14] in order to re-
duce the equations of motion for φ to a linear form. that differs from (II.3) by Weyl invariant terms
A EGB theory is not uniquely defined in such a singu- SEGB = SE.H + SW Z . (II.9)
lar limit, due to several issues, related to the selection
of the background metric and to the regularization The different EGB actions that can be generated in
procedure that it is invoked. The VE (GB) term can the d → 4 limit are all associated with the treatement
be expanded around d = 4 in several ways. One pos- of the VE term, a procedure that should be completely
sibility is defined by the ordinary DR-like procedure defined in DR and with the choice of a specific fidu-
cial metric ḡ. This would correspond to the choice of
a specific scheme, as usually done in Minkowski space.
1 µ Note that contracting (I.10) with 2g µν gives the rela-
VE (g, 4) + VE0 (g, 4) + O(ε2 ) ,


VE (g, d) =
ε  tion
(II.2)
√ √
Z
in terms of a single metric g, implicitly defining the δ
2gµν dd y −gE(y) =  gE(x), (II.10)
GB part of the EGB action in the form δgµν
1 which is at the core of (II.7), since the subtraction
VE0 = (VE (g, d) − VE (g, 4)) . (II.3)
 term VE (ḡ, d) is Weyl independent. Such subtraction
is essential in order to generate a 0/0 limit of the topo-
Note that the subtraction term VE (g, 4) obviously
logical term and obtain, henceforth, a finite action.
does not contribute to the dynamics, for being topo-
This Weyl variation is an exact property of the VE
logical, and amounts just to a constant being added
terms, therefore valid to all orders in . We will come
to the action, since
back to it in a next section, when discussing its impli-
Z
√ cation in the context of Lovelock theory.
VE (4) = d4 x gE = 4πχ0 (M), (II.4)
(II.10) is identified using the scaling relation
where χ0 (M ) is the Euler-Poincarè characteristic of a √ √ (d−4)φ

manifold M . Therefore, the evanescence of the GB −gE = ḡe ¯ µ J¯µ (ḡ, φ)+
Ē + (d − 3)∇
contribution VE is related to the fact that at d = 4 its 
variation is zero, together with all the classical vertices (d − 3)(d − 4)K̄(ḡ, φ) , (II.11)
derived from its functional differentiation
δ n VE (4) where we have defined
VEµ1 ν1 ...µn νn = . (II.5)
δgµ1 ν1 . . . δgµn νn
J¯µ (ḡ, φ) = 8R̄µν ∇
¯ ν φ − 4R̄∇¯ µ φ+
The finiteness of the contributions generated by the 4(d − 2)(∇ ¯ µφ ¯ φ − ∇ ¯ µ∇
¯ ν φ∇
¯ νφ + ∇
¯ µ φ∇
¯ λ φ∇
¯ λ φ),
renormalized vertices 1/VEµ1 ν1 ...µn νn (d) is therefore (II.12)
related to the O() behaviour of (II.5) as  → 0, as we
will discuss next. This generates a finite EGB theory
of the form K̄(ḡ, φ) = 4R̄µν ∇¯ µ φ∇¯ ν φ − 2R̄∇
¯ λ φ∇
¯ λ φ+
4(d − 2) ¯ φ∇ ¯ λ φ∇¯ λ φ + (d − 1)(d − 2)(∇ ¯ λ φ∇
¯ λ φ)2 ,
(II.13)
SEGB = SEH + VE0 , (II.6)
that allows to perform the expansion in  of the form
where VE0 is bound to satisfy the constraint

δ 0 √
√
Z
V = gE, (II.7) ¯ M J¯M +
δφ E VE (g, d) = dd x ḡ Ē + ∇

d √ √
Z Z
as recognized in the conformal anomaly effective ac- ¯ ¯M
+  dd x ḡK. (II.14)


 d x ḡφ Ē + ∇M J
tion. Alternatively, the finite action could be defined
in the Wess-Zumino (WZ) form
The scheme dependence of the regularization comes
as a last step, when the integrals present in (II.35)
1 are reduced to d = 4 from general d dimensions. This
SW Z ≡ V̂E0 = lim VE (ḡe2φ , d) − VE (ḡ, d) (II.8)


d→4  can can be obtained by introducing a cutoff (L) in the

5
extra dimensions in the form A. The 4d EGB + R2 theory

We briefly summarize the steps that take to the non-

local action. One may proceed by introducing a finite
1 R
1 √ renormalization/extension of the topological term, in
d−4 VE (g, d) (Lµ) d4 x −g 4 Ē +
=  order to derive a different version of SW Z , which is
R √ h
¯ µ φ∇¯ ν φ) + quadratic in φ, rather than quartic. This is obtained
+ (Lµ) d4 x −g φ4 Ē − (4Gµν (∇
i by extending the topological term at O() in the form
2(∇λ φ∇λ φ)2 + 4 φ∇λ φ∇λ φ) , (II.15) [14]

where all the terms in the integrands are 4-

dimensional and L is a space cutoff in the d − 4 ex-
tra dimensions. L is the volume of the extra space.
Taking the  → 0 limit and the conformal separation R2
Eext = E4 +  , (II.17)
gµν = ḡµν e2φ for the fiducial metric, we finally derive 2(d − 1)2
the expressions
and the singular limit performed on the functional
1
V̂E0 (g, φ) ≡ SW Z= (VE (g, d) − VE (ḡ, d)) = √
Z
 ṼE = dd x gEext . (II.18)
√ h
Z
¯ µ φ∇
d4 x −g φ4 E − (4Gµν (∇ ¯ ν φ) + 2(∇λ φ∇λ φ)2 +
i The effective action is then defined similarly to (IV.1),
4 φ∇λ φ∇λ φ) . (II.16) with SW Z in (II.16) now redefined by the inclusion of
(II.17)
It is easy to show that the use of the regularization
1 
in the form given above, by subtracting VE (ḡ, d) in S̃W Z = ṼE (ḡe2φ , d) − ṼE (ḡ, d) (II.19)
d dimensions - rather than at d = 4 - is eliminating 
some Weyl invariant terms [11]. The local action given induced by this additional finite modification of the
above is quartic in φ, and its structure depends on the action. A direct computation, using the rescaling for-
chosen fiducial metric. mula for R2
In summary, it is possible to define a consistent proce-
√ 2 √ φ

dure for the extraction of the effective action at d = 4, gR = ḡe R̄ − 2(d − 1) ¯ φ−
from the singular limit of a topological term. The
approach can be performed in d dimensions by a 1) 2 
rescaling of the topological density using (II.11), with ¯ λ φ∇
(d − 1)(d − 2)∇ ¯ λφ (II.20)
a metric which is d-dimensional. This implies, obvi-
ously, that the dilaton field carries dependence on the
gives, after an expansion at O()
extra dimensions. At the last stage, 2) we dimension-
ally reduce the fields, by allowing only the zero mode
4 √
 
d √
Z Z
1 1 ¯ ¯
of φ to survive the compactification procedure, while d x gEext = d x ḡ Ē + ∇ · J +
 
the extra components of the metric are assumed to be
√
Z  
flat. The cutoff L in the size of the extra dimensions d4 x ḡφ Ē + ∇ ¯ · J¯
is introduced in order to guarantee the convergence
of the integral VE in the d → 4 limit. Finally, 3) we 4 √
Z  
1
subtract the same term expressed only in terns of the + d x ḡ K̄ + R̄ − 2(d − 1) ¯ φ−
2(d − 1)2
fiducial metric, performing the limit. 2 
As shown above, the subtraction can be performed (d − 1)(d − 2)∇ ¯ λ φ∇¯ λφ .
either as in (II.3) or as in (II.8), the difference be-
tween the two being given by Weyl invariant terms, (II.21)
described in [11]. The SW Z action, which identifies a
contribution that we have also called V̂E0 in (II.8), will The expression can be simplified by some integration
define our starting action. by parts and the omission of boundary terms. Explic-

6
itly, one uses Using the currents
¯ · J¯ = −8R̄µν ∇ ¯ µ φ∇
¯ ν φ + 4R̄∇
¯ µ φ∇¯ µ φ− √
 
φ∇ ¯ + 4√g∆4 φ(x),
J(x) = J(x) ¯
J(x)
2
≡ ḡ Ē − ¯ R̄ ,
4(d − 2)∇ ¯ µ φ∇
¯ µ φ ¯ φ + 4(d − 2)∇ ¯ µ∇¯ ν φ∇¯ ν φ∇
¯ µφ 3
 
¯ µ∇
− 4(d − 2)(∇ ¯ µ φ)2 + b.t. (II.22) √ 2
J(x) ≡ g E − R (II.30)
3
and
and the quartic Green function of ∆4
¯µ ¯ν ¯ µφ = − 1 ∇
¯ ν φ∇
∇ ∇ φ∇ ¯ µ φ∇
¯ µ φ ¯ φ + b.t. (II.23) √
2 ( −g ∆4 )x D4 (x, y) = δ 4 (x, y). (II.31)

We can invert (II.30), obtaining

(where b.t. indicates the boundary terms). This gives Z
the modified relation 1 ¯
φ(x) = d4 y D4 (x, y)(J(y) − J(y)). (II.32)
4
The expression of φ in terms of the entire metric g is
√ √ what defines a conformal decomposition of the metric,
Z
2
δφ dd x gEext =  g(Eext − R) (II.24) which in this case is integrable, in the sense that we
d−1
can express φ covariantly. S̃W Z can be obtained by
which can be used in (II.18) to give solving the equation

δ S̃W Z
= J, (II.33)
δφ
2
 
δ 1 √ 2 R
ṼE (gµν , d) = g E − R+ . clearly identified in the form
δφ  3 2(d − 1)2
(II.25)
¯ + 2φ(√g∆4 )φ .
Z
S̃W Z = d4 x Jφ


Therefore from (II.19) we derive the Weyl variation (II.34)
δS W Z
 
√ 2
=α g E− R At this stage it is just matter of inserting the expres-
δφ 3 sion of φ given by (II.32) into this equation to obtain
√
 
2 ¯ ¯ the WZ action in the form
= α ḡ Ē − R̄ + 4∆4 φ . (II.26)
3
S̃W Z = Sanom (g) − Sanom (ḡ), (II.35)

Note that the redefinition of the GB density (E → with

Eext ) allows to re-obtain a rescaling of the combina- 1
Z
tion E − 2/3 R as in d = 4, where ∆4 is the fourth Sanom (g) = d4 xd4 yJ(x)D4 (x, y)J(y), (II.36)
order self-adjoint operator. This is conformal invari- 8
ant and a similar expression for Sanom (ḡ). Using the ex-
2 1 plicit expression of φ, and including the contribution
∆4 = ∇2 +2 Rµν ∇µ ∇ν − R + (∇µ R)∇µ (II.27) from the rescaled C 2 term, we finally find the nonlocal
3 3
and covariant anomaly effective action as
and satisfies the relation
4 √
Z  
√ √ 1 2
¯ 4 χ,
−g ∆4 χ = −ḡ ∆ (II.28) Sanom (g) = d x −gx E − R
8 3 x
 0 
√
Z 
if χ is invariant (i.e. has scaling equal to zero) under b 2
d4 x0 −gx0 D4 (x, x0 ) E− R .
a Weyl transformation. This brings the action S̃W Z 2 3 x0
to the form (II.37)

√
Z   
2 ¯ 4φ .
S̃W Z = α d4 x −ḡ E − ¯ R φ + 2 φ∆
3
(II.29)
The elimination of φ can be performed quite directly.

7
 III. CONSTRAINTS ON VE IN d vation of the conformal current
DIMENSIONS FOR LOVELOCK GRAVITY
√
Z  
dd x g ∇µ ξν(K) VEµν = 0, (III.2)

The presence in 4d EGB of a topological term has im- and, analogously, we can write
portant implications concerning the structure of such
√
Z
classical contributions. This point can be understood dd x g ∇µ (ξν VEµνµ1 ν1 ) = 0. (III.3)
more clearly by discussing the role of the term gs (d)VE
in the context of the conformal anomaly effective ac-
tion [11]. This action is naturally derived from a path The identity
integral, once we integrate out a conformal sector, and
one can show that the counterterm sector - that in this 2δµν δ λρ ξλ ∂ρ φ = 2δµν e2φ g λρ ξλ ∂ρ φ =
case involves also the square of the Weyl tensor - sep- 2δµν e2φ ξ λ ∂λ φ (III.4)
arately satisfies anomalous CWIs.
The derivation of such identities follows a direct pat- can be used in the covariant derivatives of (III.1) to
tern, that consists in writing down the conformal obtain
anomaly action in a background metric endowed with
conformal Killing (CKVs) vectors. In this section we ∂µ ξν − 2ξµ ∂ν φ + ∂ν ξµ − 2ξν ∂µ φ + 2δµν δ λρ ξλ ∂ρ φ =
illustrate the derivation of these constraints that de- 2
∂λ ξ λ e2φ δµν + 2δµν e2φ ξ λ ∂λ φ. (III.5)
fine, in the context of the 4d GB theory, the appli- d
cation of this method. This point can be understood
geometrically in the following way. This can be written in the form
In a local free falling frame of a curved spacetime (i.e.
in tangent space), we require that a certain action is ∂µ ξν − 2ξµ ∂ν φ + ∂ν ξµ − 2ξν ∂µ φ =
endowed with a conformal symmetry, enlarging the e2φ [∂µ (e−2φ ξν ) + ∂ν (e−2φ ξµ )] =
usual local Poincarè symmetry of Einstein’s theory. e2φ (δνλ ∂µ + δµλ ∂ν )ξ λ , (III.6)
Such conformal symmetry of each local frame can be
gauged in the form of a general metric that allows Substituting (III.4) and (III.6) in (III.5) we get (III.1)
CKVs. In the case of a conformal anomaly action, with the ordinary derivative replacing the covariant
the contribution coming from VE , as already men- ones
tioned, is paired with VC 2 . The latter is the only
effective counterterm needed in order to remove the 2
∂µ ξν + ∂ν ξµ = δµν (∂ · ξ) . (III.7)
singularity of the quantum corrections at d = 4. The d
CWIs of the complete effective action get splitted into
Writing explicitly the action of the covariant deriva-
three separate contributions: those derived from the
tive in (III.2) and taking the flat limit, then we obtain
renormalized quantum corrections and those associ-
the constraint
ated with VE and VC 2 . All the three contributions
satisfy separate conservation WIs and CWIs. Those Z  
d µνµ1 ν1 µνµ1 ν1
corresponding to the finite renormalized quantum cor- 0= d x ∂µ ξν VE + ξν ∂µ VE . (III.8)
rections, once we perform the flat spacetime limit, are
ordinary, while the other two hierarchies, related to We recall that ξν satisfies the conformal Killing equa-
VE and VC 2 , are anomalous. The constraints on these tion in flat space and by using (III.3), the equation
functionals come from their response once we perform above can be re-written in the form
a variation respect to the conformal factor φ. Z  
We detail the derivation of this point. d µνµ1 ν1 1
µνµ1 ν1
0= d x ξν ∂µ VE + ∂ · ξ δµν VE .
We recall that the CKVs are solutions of the equation d
(III.9)

If we use in this previous expression the conservation

2 and the trace identities for VEµνµ1 ν1 , that are explicitly
∇µ ξν + ∇ν ξµ = ∇λ ξ λ gµν . (III.1)
d
To derive the CWI’s in the flat limit, we need to re-
quire that the background metric allows CKVs that
leave the action invariant. We start from the conser-

8
given by - and we have for n ≥ 3
 n    
ν )
∂µ VEµνµ1 ν1 (x, x1 ) = δµ(µ1 δλ1 ∂ ν δ(x − x1 )− α ∂ 2 κα ∂
X
κ
2xj d + xj −xj δ
j=1
∂xαj ∂xαj

2δ ν(µ1 δµν1 ) ∂λ δ(x − x1 ) VEλµ (x), VEµ1 ν1 ...µn νn (x1 , . . . , xn )+
n  
µ1 ν1 µ
δµν VEµνµ1 ν1 (x, x1 ) = 2(d − 4) −g(x)E(x)
p X
(x1 )− 2 δ κµj xj α − δακ xj j
2δ(x − x1 )VEµ1 ν1 (x) (III.10) j=1
µ ν ...ν α...µn νn
VE 1 1 j (x1 , . . . , xj , . . . , xn )+
together with the explicit expression of the Killing vec- n  
(C) κ νj
X
κνj
tor ξν for the special conformal transformations, we 2 δ xj α − δ α xj
derive the relation j=1
µ ν ...µ α...µn νn
ξµ(C) κ = 2xκ xµ − x2 δµκ VE 1 1 j (x1 , . . . , xj , . . . , xn ) =
(III.11) Z 
p
∂ · ξ (C) κ = 2d xκ 2n+1 (d − 4) dd x xκ −g(x)
µ1 ν1 ...µn νn
where κ = 1, . . . , d. By using (III.11) in the integral
E(x) (x1 , . . . , xn ). (III.15)
(III.9), we derive the relation
Z 
The dilatation CWI is obtained by the choice of the
dd x 2xκ xν − x2 δνκ ∂µ VEµνµ1 ν1


0=
CKV of the form

+2 xκ δµν VEµνµ1 ν1 (III.12) ξµ(D) (x) = xµ , ∂ · ξ (D) = d, (III.16)

and equation (III.9), in the general case becomes

that can be cast into the form
Z   Z 
0= d κ 2 κ

(µ1 ν1 ) ν
d x 2x xν − x δν δµ δλ ∂ δ(x − x1 )− 0 = d x xµ ∂ν VEµνµ1 ν1 ...µn νn (x, x1 , . . . , xn )+
d

  
µνµ1 ν1 ...µn νn
2δ ν(µ1 ν1 ) λµ κ µ1 ν1
δµ ∂λ δ(x − x1 ) VE (x) − 4 x δ(x − x1 )VE (x) . δ V
µν E (x, x1 , . . . , x n ) . (III.17)

(III.13) Taking into account the conservation and trace iden-

tities satisfied by VEµ1 ν1 ...µn νn we obtain the final ex-
Integrating by parts, we obtain the expression
pression
 
∂ ∂
2d xκ1 + 2xκ1 xµ1 µ + x21 VEµ1 ν1 +
 
n
∂x1 ∂x1κ X ∂  µ1 ν1 ...µn νn
n d + xα
j α VE (x1 , . . . , xn ) =
∂x
 
j
2 x1λ δ µ1 κ − xµ1 1 δλκ VEλν1 (x1 ) j=1
Z hp iµ1 ν1 ...µn νn
  2 (d − 4) dd x
n
−g(x) E(x) (x1 , . . . , xn ).
+ 2 x1λ δ ν1 κ − xν11 δλκ VEµ1 λ (x1 ) =
(III.18)
Z
p µ1 ν1
4(d − 4) dx xκ −g(x)E(x) (x1 ) (III.14) This constraint is non trivial starting from n = 3. In
momentum space these equations are written as
which is the special CWIs satisfied by VEµ1 ν1 . This  
n−1
relation is, in this case, trivially satisfied just because X ∂
VEµ1 ν1 , in the flat limit, vanishes. Non trivial con-
d − pj α  VEµ1 ν1 ...µn νn (p1 , . . . , p̄n ) =
α

j=1
∂pj
straints are generated when we consider the contribu- √ µ1 ν1 ...µn νn
tions - starting from the three graviton vertex onwards 2n (d − 4) −g E (p1 , . . . , p̄n ), (III.19)

9
and expanding the covariant derivative, we obtain the re-
! lation
n−1
X ∂2 ∂
pκj
α − 2pαj α VEµ1 ν1 ...µn νn (p1 , . . . , p̄n ) ∂ V µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4 (x , x , x , x ) =
j=1
∂pj ∂pjα ∂pj ∂pjκ ν1 E 1 2 3 4
  µ1
δΓλν1 (x1 )

n−1
!
X
κµj ∂ κ ∂
− 2
+2 δ − δα µj δgµ2 ν2 (x2 ) g=δ
j=1
∂pj α ∂pj 
µ ν ...ν α...µn νn VEλν1 µ2 ν2 µ3 ν3 µ4 ν4 (x1 , x2 , x3 , x4 ) + (23) + (24)
VE 1 1 j (pa , . . . , pj , . . . p̄n )
δ 2 Γµλν1 1 (x1 )
n−1
!  !
κνj ∂ κ ∂
X
+2 δ − δα νj − 4 VEλν1 (x1 , x4 )+
j=1
∂pj α ∂pj δgµ2 ν2 (x2 )δgµ3 ν3 (x3 )
g=δ
µ1 ν1 ...µj α...µn νn 
VE (pa , . . . , pj , . . . p̄n )
(24) + (34) , (III.22)
n+1
= −2 (d − 4)
√
  
∂ µ1 ν1 ...µn νn where
−g E (p1 , . . . , pn ) ,
∂pnκ pn =p̄n  µ1
δΓλν1 (x1 )
 
(III.20) 1 µ1 (µi νi )
= δ δν1 ∂λ δx1 xi +
δgµi νi (xi ) g=δ 2
Pn−1
where p̄n = − i=1 pi for the conservation of the to-

µ1 (µi νi ) (µi νi ) µ1
tal momentum. These constraints are directly sat- δ δλ ∂ν1 δx1 xi − δλ δν1 ∂ δx1 xi
isfied in d 6= 4 dimensions and are therefore typi-
cal of Lovelock’s theories of gravity in generic dimen- δ 2 Γµλν1 1 (x1 )
!
sions. Notice that the equations above are modified =
δgµi νi (xi )δgµj νj (xj )
by an overall dimensionful factor both on their lhs and g=δ
rhs, for dimensional reasons, but leaving (III.20) un- 
δx1 xi µ1 (µi νi ) (µj νj )
altered. In the d → 4 limit the analysis of these equa- − δ δ δ δν1 ∂λ δx1 xj +
2
tions requires an accurate study of the degeneracy of 
such such structures, similarly to the one performed ν ) (µ
δ(µj δλj ∂ν1 δx1 xj − δλ j δνν1j ) ∂ δx1 xj +(ij), (III.23)
in [16].
In the context of anomaly actions, their reduction to
d → 4 requires the inclusion of a Weyl invariant sector, that in momentum space becomes
provided by the finite quantum corrections, missing
p1 ν1 VEµ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4 (p1 , p2 , p3 , p̄4 ) =
here. h
= 4 B µ1λν1µ2 ν2 µ3 ν3 (p2 , p3 )VEλν1 µ4 ν4 (p1 + p2 + p3 , p̄4 )
i
+ (34) + (24)
h
+ 2 C µ1λν1µ2 ν2 (p2 )VEλν1 µ3 ν3 µ4 ν4 (p1 + p2 , p3 , p̄4 )+
i
(23) + (24) , (III.24)

where
A. Conservation identity at d = 4 and trace
identities at d 6= 4 1
B µ1λν1µ2 ν2 µ3 ν3 (p2 , p3 ) ≡ − δ µ1 (µ2 δ ν2 ) ×
 2 
ν ) (µ
× δ(µ3 δνν13 ) p3 λ + δ(µ3 δλ3 p3 ν1 − δλ 3 δνν13 ) p3  + (23),
The same vertex satisfies a hierarchy of of conservation (III.25)
identity at d = 4 in flat space, starting from a curved
background. Trace Ward identities, instead, are valid
for this vertex at d 6= 4, as we are going to show next.
Indeed, from

∇µ VEµν (x)g = 0, (III.21)

10
and As we are going to illustrate in a final section, in
that case we need to amend the action by Weyl
1 ν )
C µ1λν1µ2 ν2 (p2 ) ≡ δ µ1 (µ2 δνν12 ) p2 λ + δ µ1 (µ2 δλ2 p2 ν1 invariant contributions, that are necessary in order
2  to derive the exact expression of the hierarchy. This,
(µ
−δλ 2 δνν12 ) pµ2 1 , (III.26) in principle, requires an analysis of the 4T correlator,
correcting the predictions derived from the nonlocal
correspond, respectively, to the second and first func- actions, with extra Weyl-invariant terms. Such terms
tional derivatives of the Christoffel connection. can be identified by an approach that has been
A similar analysis can be performed for a trace iden- already discussed for a simpler correlator, the T T JJ
tity. In this case, we simply functionally differentiate [16]. While this is an important point that we hope to
the anomalous Weyl variation (II.10) multiple times discuss elsewhere, it is possible to obtain the correct
and transform the expression to momentum space, ob- hierarchies satisfied by the 4-point vertices of a 4d
taining GB theory by resorting directly to a previous analysis
of the counterterms of the same vertex. This study
δµ1 ν1 VEµ1 ν1 ...µn νn (p1 , . . . , pn ) = has been presented in [17].
√ µ2 ν2 ...µn νn
2n−1 (d − 4) −gE (p2 , . . . , pn )

− 2 VEµ2 ν2 ...µn νn (p1 + p2 , p3 , . . . , pn )+

VEµ2 ν2 ...µn νn (p2 , p1 + p3 , . . . , pn ) + · · · + A. The dilaton gravity action


VEµ2 ν2 ...µn νn (p2 , p3 , . . . , p1 + pn ) . (III.27)
Coming to the local dilaton-gravity form of the limit-
This constraint is satisfied by all the vertices extracted ing theory, the two fields can be treated as indepen-
from the E4 term, and remain valid once this contri- dent, but their equations will be linked by the con-
bution is included in a Lovelock action when d 6= 4. straints coming from the anomalous variation of the
Euler Poincarè density. For this reason, (II.16) de-
fines a dilaton gravity theory in which the trace of the
equations of motion of the gravity metric ḡ and that
IV. CLASSICAL CONSTRAINTS ON THE
EQUATIONS OF MOTION IN THE LOCAL
of the conformal factor are related in the form
√
 
ACTION δ δ (W Z)
2ḡµν − SGB = −α ḡ Ē. (IV.1)
δḡµν δφ
The constraints derived in the previous sections, as
already mentioned, are obtained by performing the The derivation of this relation is discussed in [11] for
flat spacetime limit of the metric variations, without conformal anomaly actions. Note that the regulated
resorting to a conformal decomposition of the metric action (II.8) is, separately, a functional of ḡ and φ and
itself. one can use the relations
More general constraints are obtained if we, instead, δ δ √
perform a conformal decomposition and vary the fidu- 2gµν VE (g) = VE (g) =  gE, (IV.2)
δgµν δφ
cial metric and the conformal factor independently.
The separation is consistent with the fact that the
subtractions included in the definition of the WZ δ δ √
action introduce a conformal scale. This separation is 2ḡµν VE (ḡ) = 2gµν VE (ḡ), =  ḡ Ē (IV.3)
δḡµν δgµν
scale invariant, in the sense that the resulting action
is free of any dimensionful constant. These types (using ḡµν = gµν e−2φ ) and
of actions are typical of dilaton gravities and can
be modified by the addition of extra scale-invariant δ
potentials. VE (ḡ) = 0 (IV.4)
δφ
In this section we investigate the consistency of the
equations of motion, discussing their conservation in to obtain (IV.1).
the local version of the theory. The same consistency It is convenient to define the two tensors
will be missing once we move to the nonlocal theory,
obtained by eliminating the dilaton, using the Riegert 2 δSW Z
T µν = √ (IV.5)
decomposition (II.32). ḡ δḡµν

11
and From now to the rest we will omit the "bar" above all
the tensor, derivatives ecc., assuming that they are all
1 δSW Z
Tφ = √ . (IV.6) evaluated respect to the fiducial metric ḡ, treated as
ḡ δφ an independent field. The first variation gives

The relation can also be obtained by a direct compu- δ
∇λ φ SW Z = ∇λ φ E + 8Gµν ∇µ ∇ν φ+
tation using (II.16) δφ
8 φ∇µ φ∇µ φ + 16∇λ ∇µ φ∇λ φ∇µ φ
√

√ ¯ µ φ∇
¯ νφ
ḡTφ = g Ē + 8Ḡµν ∇

− 8Rµν ∇µ φ∇ν φ + 8( φ)2 − 8∇µ ∇µ φ∇µ ∇ν φ .
+ 8 ¯ φ∇ ¯ µ φ∇
¯ µ φ + 16∇¯ λ∇ ¯ µ φ∇
¯ λ φ∇
¯ µφ
(IV.12)
− 8R̄µν ∇¯ µ φ∇
¯ ν φ + 8( ¯ φ)2 −
The second variation is given by
¯ ν φ ≡ √ḡḡµν T µν

¯ µ∇
8∇ ¯ ν φ∇¯ µ∇ (IV.7)
δ
− 2gµλ ∇ν SW Z =
δgµν
where the last equality follows from the trace of (I.10).
√
Z
δ
Rescalings in the conformal decomposition are typi- − 2gµλ ∇ν d4 x −g (φE)
cally of the form δgµν

− ∇λ φ 8Gµν ∇µ ∇ν φ + 8 φ∇µ φ∇µ φ+

2
Rµνρσ = e−4φ R̄µνρσ
2 ¯ µν −
− 8R̄µν ∆
¯ λ φ∇
¯ λ φ + 4(d − 2)∆ ¯ 2µν + 16∇λ ∇µ φ∇λ φ∇µ φ − 8Rµν ∇µ φ∇ν φ
4R̄∇ 
¯ 2 + 8(d − 1)∆¯∇ ¯ λ φ∇¯ λφ + 2 µ ν
4∆ + 8( φ) − 8∇µ ∇µ φ∇ ∇ φ +

2d(d − 1)(∇¯ λ φ∇¯ λ φ)2 (IV.8) 8Rλµνρ ∇ν φ(∇µ ∇ρ φ − ∇ρ ∇µ φ). (IV.13)

and similar ones. They can be found in [11]. The Adding (IV.12) and (IV.13), recalling that
action is diffeomorphism invariant since φ transforms [∇µ , ∇ν ]φ = 0, we finally get
as a scalar under changes of coordinates. We can use  
δ δ
this invariance to derive the equation satisfied by the ∇λ φ − 2gµλ ∇ν SW Z = ∇λ φE−
stress energy tensor, by varying the action with re- δφ δgµν
√
Z
spect to the fiducial metric. The Lie derivatives for δ
2gµλ ∇ν d4 x −g (φE). (IV.14)
the scalar field φ and the fiducial metric ḡµν are δgµν
(
¯ λφ
δξ φ = ξ λ ∇ The explicit form of the second term in the equation
¯ ν ξ λ + ḡνλ ∇
¯ µξλ . (IV.9) above is expressed by
δξ ḡµν = ḡµλ ∇
√
Z
δ
giving the variation − 2gµλ ∇ν d4 x −g (φE) = −∇λ φE−
δgµν
4Rνµαβ φ∇µ Rλναβ − 2Rνµαβ φ∇λ Rνµαβ
Z  
d δ δ
δSW Z = d x δξ φ + δξ ḡµν SW Z
δφ δḡµν + 4∇µ ∇ν φ(∇ν Rλµ − ∇µ Rλν )−
Z  
¯ λ φ δ + (ḡµλ ∇ ¯ µξλ) δ
α µ ν µαβ ν
= dd x ξ λ ∇ ¯ ν ξ λ + ḡνλ ∇ SW Z 4(Rλµνα + Rλνµα )∇ ∇ ∇ φ + 4Rνµαβ Rλ ∇ φ
δφ δḡµν + 2φ[∇ λ , ∇ ν ]∇ ν
R + 2R λν φ∇ ν
R. (IV.15)
Z  
d λ ¯ δ ¯ δ
= d xξ ∇λ φ − 2ḡµλ ∇ν SW Z . (IV.10) The simplification of this expression requires some in-
δφ δḡµν
termediate steps. A lengthy but direct computation
The condition to be imposed to get the invariance un- shows that the Bianchi identity
der diffeomorphism for a generic functional such as
SW Z is Rµνρσ + Rµρσν + Rµσνρ = 0 (IV.16)

is still satisfied for a conformal decomposition.

 
∇¯ λ φ δ − 2ḡµλ ∇¯ν δ SW Z = 0. (IV.11)
δφ δḡµν Thanks to the above equation, using the symmetry

12
of the Riemann tensor and
1 √
0
Z  
[∇ν , ∇µ ]V ρ = Rρ σµν V σ , (IV.17) (3)
Snonl = −
b
d x R(1)
4
−g 2 (1)


R
1
+ (1)
18
it is quite easy to show (IV.11). A similar computation b0
Z  
(1) 1 1 µν (1) 1
 
4 (1)µν (1)
can be performed to derive (IV.1) from (II.16). d x ∂µ R R − η R ∂ν R
9 3
It is quite obvious that the singular procedure that Z
1 1 (1)
 
takes to a regulated 4d EGB action is consistent. This − d4 x b0 E(2) R +
analysis becomes rather nontrivial as the dilaton is 6
b0 1 √
(1) (1) b0
Z Z
removed from the spectrum. As already mentioned, in
d4 x R(1) −g R + d4 x R(2) R(1)
that case the nonlocal action needs to be amended by 9 9
extra terms. This will occur at the level of the classical (V.5)
4-graviton vertex. The 3-graviton vertex, instead, can
be handled directly with the nonlocal action (II.37). where the suffixes (1), (2) denote the order of the ex-
Results of this analysis are given below. pansion in the fluctuations around flat space (gµν =
δµν + hµν ). From (V.5) we can extract the expres-
sions of the classical 3-wave gravitational interactions
in this effective theory by differentiating three times
V. 3- AND 4-WAVE INTERACTION IN THE with respect to hµν .
NONLOCAL 4d GB THEORY

The nonlocal structure of the 4d EGB theory results A. 3-wave interactions in momentum space
from an iterative solution of the equations of motion
in which the dilaton is expressed in terms of the full The GB term in the equations of motion induces in-
original metric g, as shown in [19]. One can rewrite teractions of higher orders exhibiting specific features,
the nonlocal action in the form that we are going to identify in this and in the next
1R 4 √
h section. Obviously, cubic and quartic interactions in
Sanom (g, ϕ) ≡ − d x −g ( ϕ)2 − the nonlocal 4d EGB theory (V.5) share close similar-
2
1

i ities with those identified in the nonlocal conformal
µν
2 R − 3 Rg µν (∇µ ϕ)(∇ν ϕ) anomaly action. At cubic level, the most convenient
1 R 4 √ way to organize such contributions is to transform the
h
i
+ d x −g E − 23 R ϕ, (V.1) expressions to momentum space. For this purpose, we
2
define
that can be varied with respect to φ, giving Z
(1)  (1) µ1 ν1
√ √ hE R
i d4 x e−ip·x Rµανβ (x) ≡ Rµανβ (p) h̃µ1 ν1 (p)
−g ∆4 ϕ = −g − . (V.2)
2 3 (V.6)
Three-wave interactions can be derived by expanding for the linear expansion of the Ricci tensor
perturbatively in the metric fluctuations in the form 1
n o
(1)
Rµανβ = −∂α ∂β hµν −∂µ ∂ν hαβ +∂α ∂ν hβµ +∂β ∂µ hαν ,
(0) (1) (2) 2
gµν = gµν + gµν + gµν + · · · ≡ (V.7)
(2)
ηµν + hµν + hµν + . . . (V.3a) which in momentum space becomes
(0) (1) (2)
ϕ=ϕ +ϕ +ϕ + ... (V.3b)  (1) µ1 ν1 1
n
Rµανβ (p) = δα(µ1 δ ν1)β pµ pν + δµ(µ1 δνν1 ) pα pβ −
2
The expansion above should be interpreted as a col- o
lection of terms generated by setting δβ (µ1 δµν1 ) pα pν − δα(µ1 δνν1 ) pβ pµ . (V.8)

gµν = δµν + κhµν (V.4) We will be needing the identities

(1) µ1 ν1 µ2 ν2
having reinstated the coupling expansion κ, with h Rµανβ R(1)µανβ

(p1 , p2 ) ≡
of mass-dimension one, and collecting all the higher  (1) µ1 ν1  (1)µανβ µ2 ν2
order terms in the functional expansion of (V.1) of Rµανβ (p1 ) R (p2 ) =
the order h2 , h3 and so on. A similar expansion holds
(p1 · p2 )2 η µ1 (µ2 η ν2 )ν1 − 2 (p1 · p2 ) p1(µ2 η ν2 )(ν1 p2µ1 ) +
for ϕ if we redefine ϕ(1) = κϕ̄(1) , ϕ(2) = κ2 ϕ̄(2) and so
on. At cubic level the vertex is given by pµ1 2 pν12 pµ2 1 pν21 (V.9)

13
and
 (1) (1)µν µ1 ν1 µ2 ν2  (1) µ1 ν1 µ2 ν2
(p1 ) R(1)µν

Rµν R (p1 , p2 ) ≡ Rµν (p2 ) =
1 2 µ1 ν1 µ2 ν2
 
p p p η − 2 p2(µ1 η ν1 )(ν2 p2µ2 ) +
4 1 2 2
1 2 µ2 ν2 µ1 ν1
 
p2 p1 p1 η − 2 p1(µ1 η ν1 )(ν2 p1µ2 ) +
4
1 2 2 µ1 (µ2 ν2 )ν1
p p η η +
4 1 2
1 1 (µ ν ) (µ ν )
(p1 · p2 )2 η µ1 ν1 η µ2 ν2 + p1 1 p21 p1 2 p22 +
4 2
1

(µ1 ν1 )(ν2 µ2 ) (µ ν )
(p1 · p2 ) p1 η p2 − η µ1 ν1 p1 2 p22 −
2

(µ1 ν )
η µ2 ν2 p1 p21 . (V.10)

We also use the relation

 (1) 2 µ1 ν1 µ2 ν2
(R ) (p1 , p2 ) ≡ FIG. 1 Mixing in three-point functions
 (1) µ1 ν1  (1) µ2 ν2
R (p1 ) R (p2 ) =
p21 p22 π µ1 ν1 (p1 ) π µ2 ν2 (p2 ), (V.11)

and re-express the third order classical vertices of the

GB action and its contribution (V.5) to the three-
point correlator in momentum space, in the form
8
n
S3µ1 ν1 µ2 ν2 µ3 ν3 (p1 , p2 , p3 ) = α π µ1 ν1 (p1 )
3 prove the identity
h iµ2 ν2 µ3 ν3 o
(2)
E (p2 , p3 ) + (cyclic) 
δα1 β1 S3α1 β1 µ2 ν2 µ3 ν3 (p1 , p2 , p3 ) =

16α
n
p3 =−(p1 +p2 )
− π µ1 ν1 (p1 ) Qµ2 ν2 (p1 , p2 , p3 ) π µ3 ν3 (p3 ) µ2 ν2 µ3 ν3
9
8α E (2)

(p2 , p3 ), (V.16)
16α µ1 ν1
o
+ (cyclic) + π (p1 ) π µ2 ν2 (p2 ) π µ3 ν3 (p3 )
27 together with the conservation identities
n o
p23 p1 · p2 + (cyclic) , (V.12)
p2µ2 Qµ2 ν2 (p1 , p2 , p3 ) = 0 (V.17a)
µ2 ν2 µ3 ν3
p2µ2 E (2)

where we have defined (p2 , p3 ) = 0. (V.17b)
pµ pν Double tracing (V.12) of the nonlocal theory one ob-
π µν (p) ≡ δ µν − , (V.13)
p2 tains

δα1 β1 δα3 β3 S3α1 β1 µ2 ν2 α3 β3 (p1 , p2 , p3 )p3 =−(p1 +p2 ) =


and
µ2 ν2 α3 β3
Qµ2 ν2 (p1 , p2 , p3 ) ≡ p1µ [Rµν ]µ2 ν2 (p2 ) p3ν = 8α δα3 β3 E (2)

(p2 , p3 )
1 µ2 ν2
n 
(µ ν ) = 16α Q (p1 , p2 , p3 )p3 =−(p1 +p2 ) +
(p1 · p2 )(p2 · p3 ) δ µ2 ν2 + p22 p1 2 p32 −
2
8α p22 p21 + p1 · p2 π µ2 ν2 (p2 ).


(µ ν ) (µ ν )
o (V.18)
(p2 · p3 ) p1 2 p22 − (p1 · p2 ) p2 2 p32 . (V.14)
Notice that the expression above is purely polynomial
We have defined with since p22 π µ2 ν2 (p2 ) is a local term. Thus in the first line
 (2) µi νi µj νj  (1) µi νi µj νj of (V.12) we may substitute (V.16) and in the second
E = Rµανβ R(1)µανβ line use (V.18) to eliminate the Qµ2 ν2 terms and its
 (1) (1)µν µi νi µj νj  (1)
2 µi νi µj νj three cyclic permutations. The triple trace gives
− 4 Rµν R + R (V.15)
δα1 β1 δα2 β2 δα3 β3 S3α1 β1 α2 β2 α3 β3 (p1 , p2 , p3 )p =−(p +p ) =

the second functional derivative of the topological den- 3 1 2
sity in flat space, after Fourier transform. One can  2 2
16α p1 p2 − (p1 · p2 ) . 2

(V.19)

14
We use these expressions to derive the structure of the
3-wave interaction in the form
1 µ1 ν1
S3µ1 ν1 µ2 ν2 µ3 ν3 = π (p1 ) δα1 β1 S3α1 β1 µ2 ν2 µ3 ν3 +
3
1 µ2 ν2
π (p2 ) δα2 β2 S3µ1 ν1 α2 β2 µ3 ν3
3
1
+ π µ3 ν3 (p3 ) δα3 β3 S3µ1 ν1 µ2 ν2 α3 β3 −
3
1 µ1 ν1
π (p1 ) π µ3 ν3 (p3 ) δα1 β1 δα3 β3 S3α1 β1 µ2 ν2 α3 β3 −
9
1 µ2 ν2
π (p2 )π µ3 ν3 (p3 ) δα2 β2 δα3 β3 S3µ1 ν1 α2 β2 α3 β3 −
9
1 µ1 ν1
π (p1 )π µ2 ν2 (p2 )δα1 β1 δα2 β2 S3α1 β1 α2 β2 µ3 ν3 +
9
1 µ1 ν1
π (p1 )π2µ2 ν2 (p2 )π µ3 ν3 (p3 ) δα1 β1 δα2 β2 δα3 β3
27
FIG. 2 Mixing in four-point functions
S3α1 β1 α2 β2 α3 β3 . (V.20)

The nonlocal EGB theory has a structure that at tri-

linear level in the fluctuations, similarly to the case 1
of the nonlocal anomaly actions, can be depicted as S = lim [VE (g, d) − VE (g, 4)] =
d→4 (d − 4)
in Fig. 1. The vertex of the 3-wave is organized in Z Z 
1 d d−4 4
terms of longitudinal insertions of massless states on lim d xµ E4 − d x E4 . (VI.1)
d→4 (d − 4)
each of the weavy lines, in a sequence of single, double
and triple insertions. The dark blob at the center de-
Notice that the subtraction introduced above differs
notes the polynomial contributions coming from the
by Weyl invariant terms from (II.8) [5, 11] and it re-
functional derivatives of the Euler-Poincarè density.
produces, in the flat spacetime limit, the usual reg-
Each of the π µν projectors introduces a massless pole
ularization approach of DR with ḡ → δ, as applied,
in momentum space, identified from the expression
ordinarily, to momentum space. The subtraction is
1 µν needed in order to account for the 0/0 evanescent
π µν = π̂ µν = δ µν p2 − pµ pν


π̂ (V.21) contributions that, as we have already mentioned,
p2
are present in the differentiation of VE with open in-
which induce nonlocal corrections on each of the exter- dices. The derivation of the conformal constraints,
nal gravitational metric fluctuations hµν . This picture in this case, can be adapted from [17], where it has
gets modifed when we move to the case of 4-wave in- been shown that these constraints can be directly ex-
teractions. In that case the nonlocal action and hence- tracted from the counterterms introduced to regulate
forth Fig. 2 does not provide the correct expression of the quantum corrections coming from a conformal sec-
the vertices, and one has to resort to a perturbative tor. One may organize the contribution to the 4-point
expansion. While the specific features of these vertices vertex (4-wave interaction) in terms of a sequence of
can always be obtained by the brute force expansion anomaly poles and of a traceless contribution ("0-
of the VE term around flat space to fourth order, there trace") derived from a consistent longitudinal/trans-
are some uncommon features that are typical of the verse decomposition of the 4-graviton interaction
expansion, as in the cubic case. The vertex can be
reshuffled in an interesting way around flat space, as S µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4 (p1 , p2 , p3 , p̄4 ) =
µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4
we are now going to show. Spole (p1 , p2 , p3 , p̄4 )+
µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4
S0−trace (p1 , p2 , p3 , p̄4 ). (VI.2)

the expression of S0−trace is given below in (VI.4). It

is worth mentioning that the subtraction in (VI.1) is
VI. 4-WAVE INTERACTIONS IN THE
SUBTRACTED VERTEX
necessary in order to have a finite functional varia-
tion with respect to the metric fluctuation in the limit
d → 4, as also pointed out in [20, 21]. This fact is
Also in this case, as for conformal anomaly actions, reflected in the presence of an evanescent 0/0 piece,
we perform a subtraction of the form due to Lovelock tensor identities, that can be elim-

15
inated once the subtraction (VI.1) is taken into ac- tives as
count. There is a difference in the handling expres-
sions such as (VI.1), due to this subtle behaviour, p1 β1 S4α1 β1 α2 β2 α3 β3 α4 β4 (p1 , p2 , p3 , p̄4 ) =
if we perform functional derivatives with respect to
h
= 4 B α1λ1 βα1 2 β2 α3 β3 (p2 , p3 )
the metric either with uncontracted or contracted in-
dices. For instance, a differentiation with respect to
i
S2λ1 β1 α4 β4 (p1 + p2 + p3 , p̄4 ) + (34) + (24)
the φ commutes with the limit in (VI.1), and indeed h
reproduces the anomaly contribution in (IV.2), but a + 2 C α1λ1 βα1 2 β2 (p2 )
differentiation with open indices, followed by the flat i
Minkowski limit g → δ, needs special care. If we in- S3λ1 β1 α3 β3 α4 β4 (p1 + p2 , p3 , p̄4 ) + (23) + (24) . (VI.5)
clude the subtraction given in (VI.1) we can follow
in the 4d EGB the same approach implemented in the We have defined the transverse traceless projector
case of a conformal anomaly action. In (VI.2) the zero
trace part has the property 1 µ ν  1
Πµν
αβ = πα πβ + πβµ παν − π µν παβ , (VI.6)
µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4 2 d−1
δµi νi S0−trace (p1 , p2 , p3 , p̄4 ) = 0,
i = 1, 2, 3, 4, (VI.3) and the tensor
pµ pν
 
1 pα
The explicit expression of S0−trace requires the inclu- Iαµν = 2 pµ δαν + pν δαµ − (δ µν + (d − 2) 2 .
sion of suitable projectors separating the tensorial in- p d−1 p
dices into transverse traceless, trace and longitudinal (VI.7)
sectors, introduced in the analysis of the CWIs for 4T
On the other hand, the pole part is then explicitly
correlators in momentum space. This contribution, in
given as
the context of the action functional (VI.1), and more
generally, of conformally anomaly actions, has been S µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4 (p1 , p2 , p3 , p̄4 )poles =
shown to be necessary in order to generate consistent
π µ1 ν1 (p1 )
conservation WIs, as also discussed in [16]. We have = δα1 β1 S4α1 β1 µ2 ν2 µ3 ν3 µ4 ν4 (p1 , p2 , p3 , p̄4 )+
3
S µ1 ν1 µ2 ν2 µ3 ν3 µ4 ν4 (p1 , p2 , p3 , p̄4 )0−trace = π µ1 ν1 (p1 ) π µ2 ν2 (p2 )
(perm.) − δα1 β1 δα2 β2 ×
3 3

= Iαµ11 ν1 (p1 )Iαµ22 ν2 (p2 )Iαµ33 ν3 (p3 ) × S4α1 β1 α2 β2 µ3 ν3 µ4 ν4 (p1 , p2 , p3 , p̄4 ) + (perm.)
Iαµ44 ν4 (p4 ) p1β1 p2β2 p3β3 p4β4 π µ1 ν1 (p1 ) π µ2 ν2 (p2 ) π µ3 ν3 (p3 )
+ δα2 β2 δα1 β1 δα2 β2 δα3 β3
 3 3 3
+ Πµα11νβ11 (p1 )Iαµ22 ν2 (p2 )Iαµ33 ν3 (p3 ) S4α1 β1 α2 β2 α3 β3 µ4 ν4 (p1 , p2 , p3 , p̄4 ) + (perm.)
 π µ1 ν1 (p1 ) π µ2 ν2 (p2 ) π µ3 ν3 (p3 ) π µ4 ν4 (p4 )
Iαµ44 ν4 (p4 ) p2β2 p3β3 p4β4 + (perm.) − ×
3 3 3 3
 × δα1 β1 δα2 β2 δα3 β3 δα4 β4 S4α1 β1 α2 β2 α3 β3 α4 β4 (p1 , p2 , p3 , p̄4 ),
+ Πµα11νβ11 (p1 )Πµα22νβ22 (p2 )Iαµ33 ν3 (p3 ) (VI.8)

µ4 ν4 where the single trace is expressed in terms of the third
Iα4 (p4 ) p3β3 p4β4 + (perm.)
functional derivative

+ Πµα11νβ11 (p1 )Πµα22νβ22 (p2 )Πµα33νβ33 (p3 ) δα1 β1 S4α1 β1 α2 β2 α3 β3 α4 β4 (p1 , p2 , p3 , p4 ) =
√ α2 β2 α3 β3 α4 β4
 8 −gE (p2 , p3 , p4 )
µ4 ν4
Iα4 (p4 ) p4β4 + (perm.) 
− 2 S3α2 β2 α3 β3 α4 β4 (p1 + p2 , p3 , p4 )+
S4α1 β1 α2 β2 α3 β3 α4 β4 (p1 , p2 , p3 , p̄4 ), (VI.4)
S3α2 β2 α3 β3 α4 β4 (p2 , p1 + p3 , p4 )+
where the contractions are expressed in terms of the 
contribution of the third and second functional deriva- S3α2 β2 α3 β3 α4 β4 (p2 , p3 , p1 + p4 ) , (VI.9)

and additional traces involve vertices of lower orders.

16
 VII. CONCLUSIONS
In this work we have investigated the nonlinear con-
straints emerging in a nonlocal 4d EGB theory for the
3- and 4-point classical interactions present in its fun-
damental action. They are hierarchical and directly
linked to to the topological properties of the vertex
VE and its subtracted expression V̂E0 , valid in d 6= 4
and d = 4 dimensions respectively.
The analysis is essentially built on several previous
studies of conformal anomaly actions, that allow to
identify some nontrivial features of this specific 4d the-
ory. Among these, the presence of bilinear mixing in
the graviton vertices and of extra, traceless contribu-
tions, which are not predicted by the nonlocal action
as we move to 4-wave interactions. The nonlocal ver-
sion of such a theory, as pointed out, is derived by a
finite renormalization of the topological density, that
allows to remove the dilaton from the spectrum by ex-
pressing it in terms of the full metric. Three-wave in-
teractions are naturally derived from the nonlocal ac-
tion, but the derivation of the hierarchical constraint
satisfied by the four-wave interactions requires a dif-
ferent approach, given the limitations of such actions
[1] M. J. Duff, B. E. W. Nilsson, and C. N. Pope,
Gauss-Bonnet From Kaluza-Klein, Phys. Lett. B 173
(1986) 69–72.
[2] B. Zwiebach, Curvature Squared Terms and String
Theories, Phys. Lett. B 156 (1985) 315–317.
[3] D. Glavan and C. Lin, Einstein-Gauss-Bonnet
Gravity in Four-Dimensional Spacetime, Phys. Rev.
Lett. 124 (2020), no. 8 081301, [arXiv:1905.0360].
[4] R. B. Mann and S. F. Ross, The D —> 2 limit of
general relativity, Class. Quant. Grav. 10 (1993)
1405–1408, [gr-qc/9208004].
[5] C. Corianò, M. Cretì, and M. M. Maglio, Broken
Scale Invariance and the Regularization of a
Conformal Sector in Gravity with Wess-Zumino
actions, arXiv:2301.0746.
[6] M. Matsumoto and Y. Nakayama, Dilaton invading
from infinitesimal extra dimension,
arXiv:2202.1353.
[7] C. Corianò, L. Delle Rose, C. Marzo, and M. Serino,
The dilaton Wess-Zumino action in six dimensions
from Weyl gauging: local anomalies and trace
relations, Class. Quant. Grav. 31 (2014) 105009,
[arXiv:1311.1804].
[8] F. M. Ferreira and I. L. Shapiro, Integration of trace
anomaly in 6D, Phys. Lett. B 772 (2017) 174–178,
[arXiv:1702.0689].
[9] H. Elvang, D. Z. Freedman, L.-Y. Hung,
M. Kiermaier, R. C. Myers, et. al., On
17
in reproducing the correct flat spacetime limit. The
constraints that we have derived are satisfied also in
the case of Lovelock actions. Indeed it is possible to
generalize them to those cases where topological in-
variants of higher orders, such as E6 , E8 and so on,
are present, extending the approach outlined in this
work. We hope to return to this point in a future
study.
Acknowledgements
We thank Riccardo Tommasi for discussions. This
work is partially supported by INFN within the In-
iziativa Specifica QFT-HEP. The work of C. C. and
M. C. is funded by the European Union, Next Gener-
ation EU, PNRR project "National Centre for HPC,
Big Data and Quantum Computing", project code
CN00000013 and by INFN iniziativa specifica QFT-
HEP. The work of M.C. is also supported by a PON
fellowship DOT1312457-3. M. M. M. is supported
by the European Research Council (ERC) under the
European Union as Horizon 2020 research and in-
novation program (grant agreement No818066) and
by Deutsche Forschungsgemeinschaft (DFG, German
Research Foundation) under Germany’s Excellence
Strategy EXC-2181/1 - 390900948 (the Heidelberg
STRUCTURES Cluster of Excellence).
renormalization group flows and the a-theorem in 6d,
JHEP 1210 (2012) 011, [arXiv:1205.3994].
[10] D. Lovelock, The Einstein tensor and its
generalizations, J. Math. Phys. 12 (1971) 498–501.
[11] C. Corianò, M. M. Maglio, and D. Theofilopoulos,
Topological corrections and conformal backreaction in
the Einstein Gauss–Bonnet/Weyl theories of gravity
at D = 4, Eur. Phys. J. C 82 (2022), no. 12 1121,
[arXiv:2203.0421].
[12] J. D. Edelstein, Lovelock theory, black holes and
holography, Springer Proc. Math. Stat. 60 (2014)
19–36, [arXiv:1303.6213].
[13] C. Charmousis, From Lovelock to Horndeski‘s
Generalized Scalar Tensor Theory, Lect. Notes Phys.
892 (2015) 25–56, [arXiv:1405.1612].
[14] P. O. Mazur and E. Mottola, Weyl cohomology and
the effective action for conformal anomalies,
Phys.Rev. D64 (2001) 104022, [hep-th/0106151].
[15] A. O. Barvinsky, A. G. Mirzabekian, and V. V.
Zhytnikov, Conformal decomposition of the effective
action and covariant curvature expansion, in 6th
Moscow Quantum Gravity, 6, 1995. gr-qc/9510037.
[16] C. Corianò, M. M. Maglio, and R. Tommasi,
Four-point functions of gravitons and conserved
currents of CFT in momentum space: testing the
nonlocal action with the TTJJ, arXiv:2212.1277.
[17] C. Corianò, M. M. Maglio, and D. Theofilopoulos,
The Conformal Anomaly Action to Fourth Order
 (4T) in d = 4 in Momentum Space,
arXiv:2103.1395.
[18] C. Corianò and M. M. Maglio, Einstein
Gauss-Bonnet theories as ordinary, Wess-Zumino
conformal anomaly actions, Phys. Lett. B 828
(2022) 137020, [arXiv:2201.0751].
[19] C. Corianò, M. M. Maglio, and E. Mottola, TTT in
CFT: Trace Identities and the Conformal Anomaly
18
Effective Action, Nucl. Phys. B942 (2019) 303–328,
[arXiv:1703.0886].
[20] A. Bzowski, P. McFadden, and K. Skenderis,
Renormalised 3-point functions of stress tensors and
conserved currents in CFT, arXiv:1711.0910.
[21] A. Bzowski, P. McFadden, and K. Skenderis,
Renormalised CFT 3-point functions of scalars,
currents and stress tensors, JHEP 11 (2018) 159,
[arXiv:1805.1210].