a

r
X
i
v
:
1
2
0
9
.
3
5
1
1
v
1


[
g
r
-
q
c
]


1
6

S
e
p

2
0
1
2
The effective eld theory treatment of quantum
gravity
John F. Donoghue
Department of Physics, University of Massachusetts, Amherst, MA 01003, U.S.A.
Abstract. This is a pedagogical introduction to the treatment of quantum general relativity as an
effective eld theory. It starts with an overview of the methods of effective eld theory and includes
an explicit example. Quantum general relativity matches this framework and I discuss gravitational
examples as well as the limits of the effective eld theory. I also discuss the insights from effective
eld theory on the gravitational effects on running couplings in the perturbative regime.
Keywords: Effective eld theory, general relativity, quantum gravity
PACS: 04.60.-n, 11.10.-z
INTRODUCTION
Several years ago, I wrote some lectures[1] on the application of effective eld theory
(EFT) to general relativity. Because those lectures were given at an advanced school
on effective eld theory, most of the students were well versed about the effective eld
theory side, and the point was to show how general relativity matches standard effective
eld theory practice. In contrast, the present lectures were delivered at a school where
the students primarily knew general relativity, but knew less about effective eld theory.
So the written version here will focus more on the effective eld theory side, and can
be viewed as complementary to previous lectures. There will be some repetition and
updating of portions of the previous manuscript, but the two can potentially be used
together
1
.
There has been a great deal of good work combining general relativity and eld theory.
In the early days the focus was on the high energy behavior of the theory, especially on
the divergences. However, this is the portion of gravity theory that we are most unsure
about - the conventional expectation is that gravity needs to be modied beyond the
Planck scale. What effective eld theory does is to shift the focus to low energy where
we reliably know the content of general relativity. It provides a well-dened framework
for organizing quantum calculations and understanding which aspects are reliable and
which are not. In this sense, it may provide the maximum content of quantum general
relativity until/unless we are able to solve the mystery of Planck scale physics.
Effective eld theory has added something important to the understanding of quantum
gravity. One can nd thousands of statements in the literature to the effect that general
relativity and quantum mechanics are incompatible. These are completely outdated and
1
The reader can also consult my original paper on the subject[2] or the review by Cliff Burgess[3], and
of course Steven Weinbergs thoughts on effective eld theory[4] are always interesting.
no longer relevant. Effective eld theory shows that general relativity and quantum
mechanics work together perfectly normally over a range of scales and curvatures,
including those relevant for the world that we see around us. However, effective eld
theories are only valid over some range of scales. General relativity certainly does have
problematic issues at extreme scales. There are important problems which the effective
eld theory does not solve because they are beyond its range of validity. However, this
means that the issue of quantum gravity is not what we thought it to be. Rather than
a fundamental incompatibility of quantum mechanics and gravity, we are in the more
familiar situation of needing a more complete theory beyond the range of their combined
applicability. The usual marriage of general relativity and quantum mechanics is ne at
ordinary energies, but we now seek to uncover the modications that must be present in
more extreme conditions. This is the modern view of the problem of quantum gravity,
and it represents progress over the outdated view of the past.
EFFECTIVE FIELD THEORY IN GENERAL
Let us start by asking howany quantummechanical calculation can be reliable. Quantum
perturbation theory instructs us to sum over all intermediate states of all energies.
However, because physics is an experimental science, we do not know all the states that
exist at high energy and we do not know what the interactions are there either. So why
doesnt this lack of knowledge get in the way of us making good predictions? It is not
because energy denominators cut off the high energy portion - in fact phase space favors
high energy. It is not because the transition matrix elements are small. In fact, one can
argue that all processes are sensitive to the highest energy at some order in perturbation
theory.
The answer is related to the uncertainty principle. The effects of the highest energies
correspond to such short distances that they are effectively local when viewed at low
energy. As such they are identical to some local term in the Lagrangian that we use
to dene our low energy theory. These terms in the Lagrangian come with coupling
constants or masses as coefcients, and then the effect of the highest energy goes into
measured value of these parameters. This is a well-known story for divergences, but is
true for nite effects also, and would be important for quantum predictions even if the
ultimate theory had no divergences.
In contrast, low energy effects of massless (or very light) particles are not local. Ex-
amples include the photon propagator or two photon exchange potentials. Intermediate
states that go on-shell clearly propagate long distances. Sometimes loop diagrams of
light particles can have both short and long distance contributions within the same dia-
gram. We will see that we can separate these, because the short distant part looks like
a local effect and we can catalog all the local effects. The long distance portions of
processes can be separated from the short distance physics.
Effective eld theory is then the procedure for describing the long-distance physics
of the light particles that are active at low energy. It can be applied usefully in both
cases where we know the full theory or when we do not. In both situations the effects of
heavy particles and interactions (known or unknown) are described by a local effective
Lagrangian, and we treat the light particles with a full eld theoretic treatment, including
loops, renormalization, etc.
CONSTRUCTING AN EXPLICIT EFFECTIVE FIELD THEORY -
THE SIGMA MODEL
Perhaps the best way to understand the procedures of effective eld theory is to make an
explicit construction of one from a renormalizable theory
2
. This can be done using the
linear sigma model. Moreover the resulting effective eld theory has a form with many
similarities to general relativity.
Let us start with the bosonic sector of the linear sigma model
3
L =
1
2

+
1
2

+
1
2

2
_

2
+
2
_

4
_

2
+
2
_
2
. (1)
with four elds ,
1
,
2
,
3
and an obvious invariance of rotations among these
elds. This will be referred to as "the full theory". The invariance can be described
by SU(2)
l
SU(2)
R
symmetry, with the transformation being
+i V
L
( +i )V

R
(2)
where V
L,R
are 22 elements of SU(2)
L,R
, and
i
are the Pauli SU(2) matrices.
This theory is the well-known example of spontaneous symmetry breaking with the
ground state being obtained for < >= v =
_

2
/. The eld picks up a mass
m
2

= 2
2
, and the elds are massless Goldstone bosons. After the shift = v + ,
the theory becomes
L =
1
2

+
1
2
_

+m
2

v (
2
+
2
)

4
_

2
+
2
_
2
(3)
The only dimensional parameter in this theory is v or equivalently =

v or
m

2, all of which carry the dimension of a mass. For readers concerned with
gravity, you should think of this as the Planck mass of the theory. Consider it to be
very large, well beyond any energy that one can reach with using just the massless
elds.
At low energy, well below the Planck mass, all that you will see are the pions and
you will seek a Lagrangian describing their interactions without involving the . This is
2
The treatment of this section follows that of our book[5], to which the reader is referred for more
information.
3
The full linear sigma model also includes coupling to a fermion doublet, but that aspect is not needed
for our purpose.

0
a)

0
b)
FIGURE 1. Diagrams for scattering
not obvious looking at Eq. 3, but can be done. That result is
L
eff
=
v
2
4
Tr
_

_
(4)
where
U = exp[
i
v
] . (5)
This of course looks very different. It is a non-renormalizable Lagrangian where the
exponential generates non-linear interactions to all orders of the pion eld. It also comes
with two derivatives acting on the elds, so that all matrix elements are proportional
to the energy or momentum of the elds squared. In contrast, the original sigma model
has non-derivative interactions such as the
4
coupling. As we will see in more detail,
these features of the effective theory are shared with general relativity.
Example of the equivalence
Before I describe the derivation of the effective theory, let us look at an explicit
calculation in order to see the nature of the approximation involved. If we calculate a
process such as
+

0
scattering, in the full theory one nds both a direct pion
coupling and also a diagram that involves exchange which arises from the trilinear
coupling in Eq. 3, as shown in Fig. 1.
This results in a cancelation such that the matrix element is
iM

0 =2i +(2iv)
2
i
q
2
m
2

=2i
_
1+
2v
2
q
2
2v
2
_
(6)
= +i
_
q
2
v
2
+
q
4
v
2
m
2

+. . .
_
. (7)
Here q
2
= (p
+
p

+
)
2
=t is the momentum transfer
4
and I remind the reader that m
2

=
2v
2
. Despite the full theory having only polynomial interactions without derivatives,
the result ends up proportional to q
2
. Here we also see the nature of the energy expansion.
The correction to the leading term is suppressed by two powers of the sigma mass (i.e.
the Planck mass).
In the effective theory one proceeds by expanding the exponential in Eq. 4 to order
4
and taking the matrix element. Because the Lagrangian has two derivatives, the result
will automatically be of order q
2
. We nd
iM

0 =
iq
2
v
2
(8)
which of course is exactly the same as the rst term in the scattering amplitude of the full
theory. Moreover, exploration of other matrix elements will showthat all amplitudes will
agree for the full theory and the effective theory at this order in the energy expansion.
This is a very non-trivial fact. We will see how one obtains the next order term (and
more) soon.
Construction of the effective theory
Lets build the effective theory starting with the original linear sigma model. Instead
of the redenition of the eld = v + , let us consider the renaming
+i = (v +

)exp[
i

v
] = v +

+i

+. . . (9)
This change of elds from (,) to (

) is simply a renaming of the elds which

by general principles of eld theory will not change the physical amplitudes. Using the
new elds (although quickly dropping the primes for notational convenience) the full
original linear sigma model can be rewritten without any approximation as
L
eff
=
1
4
(v +)
2
Tr
_

_
+L() (10)
where U is the exponential of the pion elds of Eq. 5 and
L() =
1
2
_

+m
2

v
3

4
. (11)
This alternate form also describes a heavy eld with a set of self interactions and
with couplings to the massless pions. The symmetry of the original theory +i
V
L
( +i )V

R
is still manifest as U V
L
UV

R
,

. This form looks non-

renormalizable, but is not - it is really just the same theory as the original form.
4
For future use, the reader is reminded of the Mandelstam variables s = (p
+
+ p
0
)
2
, t = (p
+
p

+
)
2
=
q
2
, u = (p
0
p

+
)
2
.

FIGURE 2. The tree-level effect of the exchange of a heavy scalar, . The represents a vertex
involving pions, as described in the text.
To transition to the effective eld theory, we note that the lives at "the Planck scale"
and therefore is inaccessible at low energy. Only virtual effects of the can have any
inuence on low energy physics. By inspection we see that the coupling of the heavy
particle to the pions involves two derivatives (i.e. powers of the energy), so that exchange
of the will involve at least four powers of the energy. (We will treat loops soon.) So
it appears fairly obvious that the leading effect - accurate to two derivative order - is
simply to neglect the exchange of the , directly leading to the effective Lagrangian of
Eq. 4.
To conrm and extend this, it is useful to calculate explicitly the effect of exchange.
The effect of this is pictured schematically in Fig 2, where the represents the "current"
vTr
_

_
/2. This diagram yields the modication to the effective Lagrangian
L
eff
=
v
2
4
Tr
_

_
+
v
2
8m
2

[Tr
_

_
]
2
(12)
As expected, the modication involves four derivatives. Moreover, if we take the matrix
element for the scattering amplitude calculated in the previous section, we recover
exactly the q
4
term found in the full theory, Eq. 7. Using the effective Lagrangian
framework we are able to match, order by order in the energy expansion, the results
of the full theory.
Loops and renormalization
The real power of effective eld theory comes when we treat it as a full eld theory -
including loop diagrams - rather than just a set of effective Lagrangians. The idea is that
we "integrate out" the heavy eld, leaving a result depending on the light elds only. In
path integral notation this means
Z =
_
[d][d]e
i
_
d
4
xL(,)
=
_
[d]e
i
_
d
4
xL
e f f
()
(13)

a)

b)
FIGURE 3. a) A nite box diagram which occurs in the full theory, b) A bubble diagram which occurs
in the effective theory in the situation when the propagator of the heavy has been shrunken to a point.
In practice, it means that we treat loop diagrams involving the as well as tree diagrams.
But in loops also, the heavy particle does not propagate far, so even the effects of heavy
particle in loops can be represented by local Lagrangians.
The key to determining the lowenergy effective eld theory is to match the predictions
of the full theory to that of the effective theory. To be sure, the effective eld theory
has an incorrect high energy behavior. This can be seen even within specic diagrams.
For example consider the diagram of Fig. 3a which occurs within the full theory. This
diagram is nite, although pretty complicated. When treated as an effective theory we
approximate the sigma propagator as a constant, shrinking the effect of sigma exchange
to a local vertex. This leads to the diagram of Fig. 3b, which is clearly divergent
5
. So
new divergences will occur in the effective eld theory treatment which are not present
in the full theory.
However, the low energy behavior of Fig. 3a and Fig 3b will be similar. When the
loop momentum is very small, the approximation of sigma exchange as a local vertex
will be appropriate, and the pions will propagate long distances. This will be manifest in
non-analytic behaviors such as ln(s) dependence at low energy
6
. If the reader has not
yet struggled with the complexity of box diagrams, it would be a good exercise to look
up the box diagram for Fig 3a[10] and verify that the ln(s) behavior is identical (in the
low energy limit) to that of the much simpler diagram Fig 3b.
In order to deal with the general divergences of the effective theory we need to have
the most general local Lagrangian. With the symmetry of the theory being U V
L
UV

R
the general form with up to four derivatives is
L
e f f
=
M
2
P
4
Tr[

] +
1
_
Tr[

]
_
2
+
2
Tr[

]Tr[

]
(14)
5
The Taylor expansion of the propagator will lead to increasingly divergent terms also.
6
The i which follows from ln(s) = lns i for s > 0 accounts for the discontinuity from on-shell
intermediate states.
where
i
are constant coefcients. We saw at tree level that

1
=
v
2
8m
2

,
2
= 0 (15)
The divergences of the effective theory will be local and will go into the renormalization
of these coefcients.
The comparison of the full theory and the effective theory can be carried out directly
for the reaction
+

0
. The dimensionally regularized result for the full theory
is quite complicated - it is given in [7]. However, the expansion of the full theory at low
energy in terms of renormalized parameters is relatively simple[8]
M
f ull
=
t
v
2
+
_
1
m
2

v
2

11
96
2
v
4
_
t
2

1
144
2
v
4
[s(s u) +u(us)]

1
96
2
v
4
_
3t
2
ln
t
m
2

+s(s u)ln
s
m
2

+u(us)ln
u
m
2

_
(16)
One calculates the same reaction in the effective theory, which clearly does not know
about the existence of the . The result [8, 9] has a very similar form,
M
e f f
=
t
v
2
+
_
8
r
1
+2
r
2
+
5
192
2
_
t
2
v
4
(17)
+
_
2
r
2
+
7
576
2
_
[s(s u) +u(us)]/v
4
(18)

1
96
2
v
4
_
3t
2
ln
t

2
+s(s u)ln
s

2
+u(us)ln
u

2
_
(19)
where we have dened the renormalized parameters

r
1
=
1
+
1
384
2
_
2
4d
+ln4
_
(20)

r
2
=
2
+
1
192
2
_
2
4d
+ln4
_
(21)
At this stage we can match the two theories, providing identical scattering amplitudes to
this order, through the choice

r
1
=
v
2
8m
2

+
1
384
2
_
ln
m
2

2

35
6
_
(22)

r
2
=
1
192
2
_
ln
m
2

2

11
6
_
(23)
Loops have modied the result of tree level matching by a nite amount. We have not
only obtained a more precise matching, we also have generated important kinematic
dependence, particularly the logarithms, in the scattering amplitude. The logarithms do
not involve the chiral coefcients
i
because the logs follow from the long-distance
portion of loops while the chiral coefcients are explicitly short-distance. This is part
of the evidence for a separation of long and short distances.
We have seen that we can renormalize a "nonrenormalizable" theory. The divergences
are local and can be absorbed into parameters of a local Lagrangian. Moreover, the pre-
dictions of the full theory can be reproduced even when using only the light degrees of
freedom, as long as one chooses the coefcients of the effective lagrangian appropri-
ately. This holds for all observables. This can be demonstrated using a background eld
method [8, 5, 11]. Once the matching is done, other processes can be calculated using
the effective theory without the need to match again for each process. The total effect of
the heavy particle, both tree diagrams and loops, has been reduced to a few numbers in
the Lagrangian, which we have deduced from matching conditions to a given order in
an expansion in the energy.
Power counting and the energy expansion
We saw that the result of the one loop calculation involved effects that carried two
extra factors of the momenta q
2
. It is relatively easy to realize that this is a general
result. The only dimensional parameter in the effective theory is v, and one can count
the powers of v that enter into loop diagrams. These enter in the denominator because
the exponential is expanded in /v. In dimensional regularlization there is no scale to
provide any competing powers in the renormalization procedure
7
. Therefore the external
momenta q
2
, s, t, u must enter to compensate for the powers of 1/v
n
, leading to an
expansion in q
2
/v
2
or more appropriately q
2
/(4v)
2
.
This energy expansion is a second key feature of EFT techniques. Predictions are
ordered in an expansion of powers of the light energy scales over the high energy
scale, the Planck scale, of the full theory. For the chiral theory, Weinberg provided
a compact theorem showing how higher loop diagrams always lead to higher energy
dependence[6]. A similar power counting occurs in general relativity.
The power counting allows an efcient approach to loops in gravity also because
general relativity has a power counting behavior similar to the chiral case described
here. The coupling constant of gravity, Newtons constant G, carries the dimension of
the inverse Planck mass squared, G 1/M
2
P
. This means that loops carry extra powers
of G (just like the extra factors of 1/v
2
of the chiral case) and this is compensated for by
factors of the lowenergy scale in the numerator, leading to an efcient energy expansion.
7
There is an arbitrary scale which enters only logarithmically.
Effective eld theory in action
Lest the above description sound somewhat formal, I should point out that the chiral
effective eld theory is actively and widely used in phenomenological applications for
QCD. Of course, the linear sigma model is not the same as QCD. However, if the mass
of the up and down quarks were zero, QCD would also have an exact SU(2)
L
SU(2)
R
symmetry which is dynamically broken, with massless pions as the Goldstone bosons.
The effective Lagrangian of the theory would have the same symmetry as the sigma
model and so the general structure of the effective Lagrangian would be the same. The
identication of the vector and axial currents allows the identication of v with the
pion decay constant v = F

= 92.4 MeV measured in the decay . The chiral

coefcients
1
,
2
are relatively different from those of the sigma model. In QCD they
are difcult to calculate from rst principles, but as measured by experiment they are of
order 10
3
and their relative sizes reveal the inuence of the vector meson (770) rather
than of a scalar [12].
The chiral symmetry of QCD also has small explicit breaking from the up and
down quark masses, leading to a non-zero pion mass. This symmetry breaking can also
be treated perturbatively as an element in the energy expansion. The result is chiral
perturbation theory, a lively interplay of energies, masses, loop diagrams, experimental
measurements that makes for a ne demonstration of the practical power of effective
eld theory. The reader is referred to our book [5] for a more complete pedagogic
development of the subject.
There also are very many more uses of effective eld theory throughout physics. Some
of these have slightly different issues and techniques, and the above is not a complete
introduction to all aspects of effective eld theory. However, all EFTs share the common
features of 1) using only the active degrees of freedom and interactions relevant for the
energy that one is working at, and 2) organizing the results in some form of an energy
expansion. The example of the linear sigma model is particularly well suited for the
introduction to the gravitational theory.
Summary of general procedures
In preparation for the discussion of the effective eld theory for general relativity, let
me summarize the techniques for an EFT in the situation when we do not know the full
theory. In contrast with the linear sigma model, we then cannot match the EFT to the
full theory. However EFT techniques can still be useful and predictive.
1) One starts by identifying the low energy particles and the symmetries governing
their interaction.
2) This information allows one to construct the most general effective Lagrangian
using only these particles and consistent with the symmetry. This Lagrangian can be
ordered in an energy expansion in terms of increasing dimensions of the operators
involved. In contrast with the stringent constraints of "renormalizable eld theory",
one allows operators with dimension greater than four. Renormalizable theories can be
thought of as a subclass of EFTs where the operators of dimension greater than four are
not (yet) needed.
3) The operator(s) of lowest dimension provide the leading interactions at low energy.
One can quantize the elds and identify propagators in the usual way.
4) These interactions can now be used in a full eld theoretic treatment. Loop dia-
grams can be computed. Because the divergences will be local they can be absorbed into
the renormalization of terms in the local effective Lagrangian.
5) Because, by assumption, the full theory is not known, the coefcients of various
terms in the effective Lagrangian cannot be predicted. However, in principle they can be
measured experimentally. In this case, these coefcients are not predictions of the effec-
tive eld theory treatment - they represent information about the full theory. However,
once measured they can be used in multiple processes.
6) Predictions can be made. The goal is to obtain predictions from long distance (low
energy) physics. Because we know that short distance physics is local, it is equivalent
to the local terms in the Lagrangian, and everything left over is predictive. This most
visible involves nonanalytic terms in momentum space, such as the logarithmic terms
in the scattering amplitude above. These are obviously not from a local Lagrangian and
correspond to long distance propagation. The logs can also pick up imaginary parts when
their arguments are negative - the imaginary parts are indicative of on-shell intermediate
states. Analytic terms, such as the s
2
/16
2
v
2
terms in the scattering amplitude above,
can also be predictive when they can be separated from the Lagrangian parameters, such
as
i
, through measurement of the parameters in other reactions.
GRAVITY AS AN EFFECTIVE FIELD THEORY
In the case of gravity, we assume that there is some well dened full theory of gravity
that yields general relativity as the low energy limit. We do not need to know what it is,
but we take as experimental fact that it is well behaved, with no signicant instabilities
or run-away solutions when dealing with low curvature situations such as our own. Other
particles and interactions - the Standard Model - can be added, but are not crucial to the
EFT treatment. Basically, we assume that there is a limit of the full theory that looks like
the world that we see around us.
If the low energy limit is general relativity, we know that the relevant degrees of
freedom are massless gravitons, which are excitations of the metric. The symmetry
is general coordinate invariance. In constructing the energy expansion of the effective
Lagrangian, one must pay attention to the number of derivatives. The connection

=
g

2
_

(24)
has one derivative of the metric, while the curvatures such as
R

(25)
have two. The various contractions of the Riemann tensor are coordinate invariant. These
features determine the nature of the energy expansion of the the action for general
relativity
S
grav
=
_
d
4
x

g
_
+
2

2
R+c
1
R
2
+c
2
R

+. . . +L
matter
_
(26)
Here the terms have zero, two and four derivatives respectively.
Following our EFT script, we turn to experiment to determine the parameters of this
Lagrangian. The rst term, the cosmological constant, appears to be non-zero but it
is so tiny that it is not relevant on ordinary scales. The EFT treatment does not say
anything novel about the smallness of the cosmological constant - it is treated simply as
an experimental fact. The next term is the Einstein action, with coefcient determined
from Newtons constant
2
= 32G. This is the usual starting place for a treatment
of general relativity. The curvature squared terms yield effects that are tiny on normal
scales if the coefcients c
1,2
are of order unity. In fact these are bounded by experiment
[13] to be less than 10
+74
- the ridiculous weakness of this constraint illustrates just
how irrelevant these terms are for normal physics. So we see that general coordinate
invariance allows a simple energy expansion.
At times people worry that the presence of curvature squared terms in the action will
lead to instabilities or pathological behavior. Such potential problems have been shown
to only occur at scales beyond the Planck scale [14] where yet higher order terms are
also equally important. This is not a aw of the effective eld theory, which holds only
below the Planck scale. Given the assumption of a well-behaved full theory of gravity,
there is no aspect of the effective theory that needs to display a pathology.
Quantization and renormalization
The quantization of general relativity is rather like that of Yang-Mills theory. There
are subtle features connected with the gauge invariance, so that only physical degrees
of freedom count in loops. Feynman[15], and then DeWitt[16], did this successfully in
the 1960s, introducing gauge xing and then ghost elds to cancel off the unphysical
graviton states. The background eld method employed by tHooft and Veltman[17]
was also a beautiful step forward. It allows the expansion about a background metric
( g

) and explicitly preserves the symmetries of general relativity. It is then clear that
quantization does not spoil general covariance and that all quantum effects respect this
symmetry. The uctuation of the metric around the background is the graviton
g

(x) = g

(x) +h

(x) (27)
and the action can be expanded in powers of h

(x) (with corresponding powers of ).

The Feynman rules after gauge xing and the addition of ghosts have been given in
several places[1, 2, 17] and need not be repeated here. They are unremarkable aside
from the complexity of the tensor indices involved.
Renormalization also proceeds straightforwardly. As advertised, the divergences are
local, with the one loop effect being equivalent to[17]
L =
1
16
2
2
4d
_
1
120
R
2
+
7
20
R

_
(28)
The divergence can then be easily absorbed into renormalized values of the coefcients
c
1,2
. The fact that these occur at the order of four derivatives can be seen by counting
powers of and is completely equivalent to the energy expansion of the linear sigma
model described above.
Pure gravity is one-loop nite. This is because the equations of motion for pure gravity
(not including the cosmological constant) is R

= 0, so that both of the divergent

counter terms vanish when treated as a perturbation. This is an interesting and useful
fact, although in the real world it does not imply any special niteness to the theory
because in the presence of matter the counterterms are physically relevant.
Predictions of the effective eld theory
Perhaps the most elementary prediction of quantum general relativity is the scattering
of two gravitons. This was worked out to one-loop by Dunbar and Norridge [18]. The
form is
A(++; ++) =
i
4

2
s
3
tu
_
1+

2
s t u
4(4)
2

2
(1)(1+)
(12)

_
_
2

_
ln(u)
st
+
ln(t)
su
+
ln(s)
tu
_
+
1
s
2
f
_
t
s
,
u
s
_
+2
_
ln(u)ln(s)
su
+
ln(t)ln(s)
tu
+
ln(t)ln(s)
ts
_
_
_
_
_
A(++; ) = i

4
30720
2
_
s
2
+t
2
+u
2
_
A(++; +) =
1
3
A(++; ) (29)
where
f
_
t
s
,
u
s
_
=
(t +2u)(2t +u)
_
2t
4
+2t
3
ut
2
u
2
+2tu
3
+2u
4
_
s
6
_
ln
2
t
u
+
2
_
+
(t u)
_
341t
4
+1609t
3
u+2566t
2
u
2
+1609tu
3
+341u
4
_
30s
5
ln
t
u
+
1922t
4
+9143t
3
u+14622t
2
u
2
+9143tu
3
+1922u
4
180s
4
, (30)
and where the +, refer to the graviton helicities. Here A(++; ++) shows the na-
ture of the energy expansion for general relativity most clearly. It is the only amplitude
with a tree level matrix element - the others all vanish at tree level. The tree amplitude
is corrected by terms at the next order in the energy expansion, i.e. by factors of order

2
(Energy)
2
relative to the leading contribution. Note also the nonanalytic logarithms.
Another interesting feature, despite the presence of 1/ terms in the formulas, is that
these results are nite without any unknown parameters. Because the counterterms van-
ish in pure gravity, as noted above, the scattering amplitudes cannot depend on the coef-
cients of the higher order terms. The 1/ terms have been shown to be totally infrared
in origin, and are canceled as usual by the inclusion of gravitational bremsstrahlung [19],
as would be expected from general principles[20]. This result is a beautiful low energy
theorem of quantum gravity. No matter what the ultimate ultraviolet completion of the
gravitational theory, the scattering process must have this form and only this form, with
no free parameters, as long as the full theory limits to general relativity at low energy.
The leading quantum correction to the gravitational potential is also a low energy
theorem independent of the ultimate high energy theory. The result for the potential of
gravitational scattering of two heavy masses is[21, 22]
V(r) =
GMm
r
_
1+3
G(M+m)
rc
2
+
41
10
G h
r
2
c
3
_
. (31)
where the last term is the quantum correction and the term preceding that is the classical
post-Newtonian correction. Let me explain how this is calculated and why it is reliable.
We already know the form of the quantum correction from dimensional analysis, as the
unique dimensionless parameter linear in h and linear in G is G h/r
2
c
3
. The classical
post-Newtonian correction is also a well-known dimensionless combination, without h.
Fourier transforming tells us that the corresponding results in momentum space are
1
r

1
q
2
,
1
r
2

1
q
2

_
q
2
,
1
r
3

1
q
2
q
2
lnq
2
,
3
(x)
1
q
2
q
2
(32)
So here we see the nature of the energy expansion. The leading potential comes from
one-graviton exchange with the 1/q
2
coming from the massless propagator. The correc-
tions come from loop diagrams - all the one-loop diagrams that can contribute to the
scattering of two masses. The kinematic dependence of the loops then brings in nonana-
lytic corrections of the form Gm
_
q
2
, Gq
2
lnq
2
, as well as analytic terms Gq
2
. As shown
above in the chiral scattering result, the analytic terms include the effects of the next or-
der Lagrangian (the c
i
coefcients from Eq. 26) and the divergences. However, since
1
q
2
q
2
constant, the Fourier transforms of the analytic term appears as a delta func-
tion in position space, and these terms do not lead to any long-distance modications
of the potential. The power-law corrections come from the non-analytic terms, with the
logarithm generating the quantum correction. This standard reasoning of effective eld
theory allows us to know in advance that the quantum correction will be nite and free
from unknown parameters.
The actual calculation is pretty standard, although it is notationally complicated be-
cause of all the tensor indices for graviton couplings. After a series of partial calcula-
tions, possible mistakes and alternative denitions[23], the result appears solid with two
groups in agreement
8
[21, 22]. However, the result itself is interesting mainly because
of the understanding of why it is calculable. In this calculation we see the compatibility
of general relativity and quantum mechanics at low energy, and have separated off the
8
I also know of some unpublished conrmations of the same result.
unknown high-energy portion of the theory. The magnitude of the correction is far too
small to be seen - a correction of 10
40
at a distance of one fermi. There are no free
parameters that we can adjust to change this fact. However, in some ways this can be
pitched as a positive. When we do perturbation theory, the calculations are the most reli-
able when the corrections are small. The gravitational quantum correction is the smallest
perturbative correction of all our fundamental theories. So instead of general relativity
being the worst quantum theory as is normally advertised, perhaps it should be consid-
ered the best!
It is also worth commenting on the classical correction in Eq. 31
9
. This comes from
a one-loop calculation, which surprises some who know the supposed theorem that the
loop expansion is an expansion in h. However, that theorem is not really a theorem, and
this is one of several counterexamples[26]. The usual proof neglects to note that there
also can be a factor of h within the Lagrangian, as mass terms carry a factor of m/ h. This
is manifest within the calculation in the square-root nonanalytic term as an extra factor
_
m
2
/k
2
h
2
.
This is perhaps a good place to note that a powerful classical EFT treatment of gravity
has been developed by Goldberger, Rothstein and their collaborators[27]. This requires a
further development of EFT methods to separate out the relevant components within the
graviton eld itself. Their work provides a systematic treatment of the classical bound
state and gravitational radiation problems. It is now a useful component of the gravity
wave community, presently ahead of conventional methods in the prediction of gravity
waves of binaries with spin.
Another calculation that shows how EFT methods work is the quantum correction to
the Reissner-Nordstom and Kerr-Newman metrics (for charged objects without and with
spin), which takes the form[28] in harmonic gauge
g
00
= 1
2Gm
r
+
G
r
2

8G h
3mr
3
+. . .
g
0i
= (
2G
r
3

G
mr
4
+
2G h
m
2
r
5
)(

Sr)
i
g
i j
=
i j

i j
2Gm
r
+G
r
i
r
j
r
2
+
4G h
3mr
3
_
r
i
r
j
r
2

i j
_
+. . . (33)
Here the quantum correction comes from photon loops, and gravity is treated classically.
But the techniques are the same. One nds that the classical correction again comes
from the square-root non-analytic term in momentum space. Physically we can identify
exactly what causes this correction. It comes from the electric eld which surrounds
the charged object. The electric eld is non-local, falling with r
2
and it carries energy.
Gravity couples to this energy and precisely[28] reproduces the classical correction in
the metric. At tree-level in eld theory one sees only the point charged particle, and
the photon loop diagram then is needed to describe the energy in the electric eld
surrounding the charged object. The quantumcorrection again comes from the logarithm
in momentum space.
9
The classical correction was previously known to be calculable using eld theory methods [24, 25].
A calculation of Hawking radiation that appears solidly in the spirit of effective eld
theory is by Burgess and Hambli[29]. They study a scalar propagator in a low curvature
region outside the black hole and regularize with a high energy cutoff. The ux from
the black hole can calculated from the propagator and appears insensitive to the cutoff.
However, EFT cannot address questions about the end state of black hole evaporation -
this concerns a situation beyond the region of validity of the effective eld theory.
ISSUES IN THE GRAVITATIONAL EFFECTIVE FIELD THEORY
Effective eld theories are expected to have limits. While it is logically conceivable that
we could nd some innovative method that allows one to extend general relativity to
all energies, this is generally viewed as unlikely. More typically, effective eld theories
get modied by new particles and new interactions as one goes to high energy. The
archetypical example is QCD, where the pions of the effective theory get replaced by
the quarks and gluons of QCD at high energy. The most common expectation is that
something similar happens for general relativity. String theory would be a consistent
example for general relativity, although there may be other ways to form an ultraviolet
completion of the gravitational interactions. It would be lovely to understand the nature
of this high energy modication. However, since physics is an experimental science and
the Planck scale appears to be frustratingly out of reach of experiment, it is not clear that
we will have reliable evidence for the nature of the ultimate theory in the foreseeable
future. The competition between divergent proposals for the ultimate theory, which in
principle would be a scientic discussion if experiment could keep up, may remain
unresolved in our lifetimes without the input of experiment.
However, it does remain interesting to explore the limits of the effective theory.
Potentially these limits can help us understand when new physics must enter. One
obvious case is at the Planck energy. Scattering amplitudes are all proportional to powers
of GE
2
, with
M =M
0
[1+aGE
2
+bGE
2
lnE
2
+....] (34)
This clearly leads to problems at the Planck scale, where the energy expansion breaks
down.
However, there may also be other limits to the validity of the effective theory. I
have argued elsewhere[2, 30] that the extreme infrared of general relativity may pose
problems that other effective theories do not face. This is because gravitational effects
may build up - the integrated curvature may be large even in the local curvature is not.
Such effects are manifest as horizons and singularities. Horizons by themselves are not
expected to be problematic - in a local neighborhood nothing special need be evident.
But the long distance relations of horizons to spatial innity clearly brings in problematic
features for black holes.
Moreover, singularities pose problems for the effective theory in the long distance
limit. The singularity itself is not the problem - we know that the effective theory breaks
down when the curvature is large. But even for the long distance theory, propagating past
a singularity poses some difculties. We do not know the fate of the modes that ow into
the singularity. Perhaps this can be solved by treating the location of the singularity as an
external source. In chiral perturbation theory, this is done for baryons, which arguably
appear as solitons in the effective theory which are found only beyond the limits of the
effective theory, yet also serve as a source for the Goldstone bosons. Perhaps singular
regions can be treated as localized sources of gravitons.
However, even if singularities can be isolated and tamed, there are issues of the rela-
tion of localized regions of small curvature, where the effective theory is demonstrably
valid, to far distant regions where the integrated curvature is large. This conict is at the
heart of black hole paradoxes, where the horizon is well behave locally but is problem-
atic when dened by its relation to spatial innity.
GRAVITATIONAL CORRECTIONS AND RUNNING COUPLING
CONSTANTS
Let me also address what effective eld theory has to say about the gravitational correc-
tions to the running of coupling constants, including that of gravity itself. This subject
has had a confusing recent history, and EFT is useful in sorting out the issues
10
.
Despite having a pre-history[32], recent activity stems from the work of Robinson
and Wilczek[33], who suggested that the beta function of a gauge theory could have the
form
(e, E) =
b
0
(4)
2
e
3
+a
0
e
2
E
2
(35)
and calculated a
0
to be negative. While this correction is tiny for most energies, the neg-
ative sign suggests that all couplings could be asymptotically free if naively extrapolated
past the Planck scale. Subsequent work by several authors, all using dimensional regu-
larization, found that the gravitational correction to the running coupling vanishes[34].
Further work, including some of the same authors, using variations of cutoff () regu-
larization then found that it does run[35], where the cutoff plays the role of the energy,
i.e. G
2
. Papers trying to clarify this muddle include[36, 37, 38, 39]. My treatment
here most naturally follows the ones of my collaborators and myself[36, 37]
In renormalizable eld theories, the use of a running coupling is both useful and
universal. It is useful because it sums up a set of quantum corrections which potentially
could have large logarithmic factors. It is universal because the logarithmic factors are
tied to the renormalization of the charge, and hence enter the same way in all processes.
Specically, when one employs dimensional regularization the charge renormalization
and the ln
2
factors always enter the same way, because of the expansion of (
2
)

/.
For example, in perturbation theory photon exchange at high energy involves
e
2
q
2
=
e
2
0
q
2
_
1+

3
2
_
2
4d
ln
q
2

2
+...
__
(36)
The ln
2
dependence follows the renormalization of the charge, and dimensionally the
lnq
2
dependence has to accompany the ln
2
dependence. When this is turned into a
10
The comments of this section continue to assume that the cosmological constant is small enough to
be neglected. In the presence of a cosmological constant the story is different and there can be genuine
contributions to running couplings[31]
running coupling, the energy dependence always appears in a universal fashion because
the charge is renormalized the same way in all processes. In addition, with logarithmic
running the running coupling constant is crossing symmetric since lnq
2
has the same
value, up to an imaginary part due to on-shell intermediate states, for q
2
either positive
or negative. This lets the running coupling constant be applicable in both the direct
channel and in the crossed channel for a given type of reaction.
For gravitational corrections in the perturbative regime, effective eld theory explains
why each of the features (useful, universal, crossing symmetric) no longer holds. Let me
discuss them individually.
We have seen that gravitational corrections are expansions in Gq
2
with potentially ex-
tra logarithms. Because in different reactions q
2
can take on either sign, the gravitational
correction will not have the same sign under crossing. For example if we calculated the
gravitational correction to a process such as f +

f f

+

f

, for two different avors

f , f

, within QED, one would nd a correction of the form

M
e
2
s
[1+bGs] (37)
with s = (p
1
+p
2
)
2
> 0 and b is a constant to be calculated. However, if one studies the
related crossed process f + f

f + f

one would have

M
e
2
t
[1+bGt] (38)
with t = (p
1
p
3
)
2
< 0 and b is the same constant. If one tried to absorb the quantum
correction into a running coupling, it would be an increasing function of energy in the
one process and a decreasing function of the energy in the other process. Moreover,
other processes such as the single avor case f +

f f +

f , would involve both s and t
variables within the same reaction. A universal energy dependent running coupling with
quadratic four-momentum dependence cannot account for the features of crossing.
Effective eld theory also explains why gravitational corrections are not a universal
factor correcting the coupling. We have seen in the direct calculations above and in the
discussion of the energy expansion that gravitational corrections do not renormalize the
original operator, but generate divergences that are two factors of the energy (derivatives)
higher. This of course is due to the dimensional coupling constant. For gravity itself,
after starting with the Einstein action R we found that divergences were proportional to
R
2
and R

. For the QED case, the higher order operators could be

2
A

. (39)
These operators are related to each other by the equations of motion, but there are
two points associated with this observation. One is that, in contrast to renormalizable
eld theories, we are not renormalizing the original coupling so that we do not have
any expectation of universality. But in addition, different processes involve different
combinations of the higher order operators, so that the renormalization that takes place
is intrinsically different for different processes. Different reactions will involve different
factors. It is for this reason that use of the renormalization group in effective eld theory,
which is in fact a well studied subject[6, 40], does not involve the running of the basic
coupling, but instead is limited to predicting logarithmic factors associated with the
higher order operators.
So we have seen that effective eld theory, in the region where we have control
over the calculation, does not lead to gravitational corrections to the original running
couplings. We might still ask if there might nevertheless be some useful denition that
plays such a role. This logically could occur if one repackaged some of the effects of the
higher order operators as if they were the original operator - i.e. some sort of truncation
of the operator basis. However, here one can simply calculate gravitational corrections
to various processes to see if some useful denition emerges. The answer is decidedly
negative[37]. The numerical factors involved in different reactions vary by large factors
and come with both signs.
In the perturbative regime there is no useful and universal denition of running G(q
2
)
nor gravitational corrections to running e(q
2
), and there is a kinematic crossing obstacle
to any conceivable proposal.
Far too often in the literature, a dimensional cutoff () is employed to incorrectly
conclude that it contributes to a running coupling. For example, one can use cutoff and
nd that the original coupling is renormalized. For example for gravity or QED
G = G
0
[1+cG
0

2
], e
2
= e
2
0
[1+dG
2
] (40)
for some constants c, d. If one then says the magic phrase Wilsonian, one might think
that this then denes a running coupling G() or e(). However, this is not a running
coupling - it is just renormalization. The quadratic dependence disappears into the
renormalized value of G or e at low energy[37, 38]. Once you measure this value there
is no remaining quadratic dependence, and there is no energy dependence that tracks
the G
2
behavior.
These observations pose questions for the sub-eld of Asymptotic Safety[41, 42]. In
this area, on considers a Euclidean coupling g =Gk
2
E
, where k
E
is a Euclidean momenta,
and attempts to nd a ultraviolet xed point for g. Since the operator basis expands to
operators of increasingly high number of derivatives, this involves a truncation of the
operator basis. Given such a Euclidean truncation, one can by construction dene a
running coupling, and one does nd a UV xed point well beyond the Planck scale.
However, the question remains whether this implies anything useful for real processes
in Lorentzian gravity. One would think that one should nd evidence of such a running
coupling in the perturbative region where we have control over the calculation. However,
we have seen the reasons why this does not occur. The conict certainly deserves further
study.
SUMMARY
Effective eld theory is a well-developed framework for isolating the quantum effects
of low energy particles and interactions, even if these interactions by themselves fall
outside of a complete renormalizable eld theory. It works well with general relativity
applied over ordinary curvatures and energy scales. The effective eld theory has limits
to its validity, most notably it is limited to scales below the Planck energy, and does not
resolve all of the issues of quantum gravity. However, effective eld theory has shown
that general relativity and quantum mechanics do in fact go together ne at ordinary
scales where both are valid. GR behaves like an ordinary eld theory over those scales.
This is important progress. We still have work to do in order to understand gravity and
the other interactions at extreme scales.
ACKNOWLEDGMENTS
This material was presented at the Sixth International School on Field Theory and Grav-
itation, Petropolis Brazil, April 2012. I thank the organizers and my fellow participants
for an interesting and informative school. This work is supported in part by the U.S.
National Science Foundation grant PHY-0855119
REFERENCES
1. J. F. Donoghue, Introduction to the effective eld theory description of gravity, gr-qc/9512024, in
Proceedings of the Advanced School on Effective Field Theories , Almuecar, Spain 1995, ed. by
Fernando Cornet and Marie Jose Herrero (World Scientic, 1997).
2. J. F. Donoghue, General relativity as an effective eld theory: The leading quantum corrections,
Phys. Rev. D 50, 3874 (1994) [gr-qc/9405057].
3. C. P. Burgess, Quantumgravity in everyday life: General relativity as an effective eld theory, Living
Rev. Rel. 7, 5 (2004) [gr-qc/0311082].
4. S. Weinberg, Effective Field Theory, Past and Future, PoS CD 09, 001 (2009) [arXiv:0908.1964
[hep-th]].
5. J. F. Donoghue, E. Golowich and B. R. Holstein, Dynamics of the Standard Model, Camb. Monogr.
Part. Phys. Nucl. Phys. Cosmol. 2, 1 (1992).
6. S. Weinberg, Phenomenological Lagrangians, Physica A 96, 327 (1979).
7. A. V. Manohar and V. Mateu, Dispersion Relation Bounds for pi pi Scattering, Phys. Rev. D 77,
094019 (2008) [arXiv:0801.3222 [hep-ph]].
8. J. Gasser and H. Leutwyler, Chiral Perturbation Theory to One Loop, Annals Phys. 158, 142 (1984).
9. H. Lehmann, Chiral invariance and effective range expansion for pion pion scattering, Phys. Lett. B
41, 529 (1972).
10. A. Denner, U. Nierste and R. Scharf, A Compact expression for the scalar one loop four point
function, Nucl. Phys. B 367, 637 (1991).
11. J. Gasser and H. Leutwyler, Chiral Perturbation Theory: Expansions in the Mass of the Strange
Quark, Nucl. Phys. B 250, 465 (1985).
12. J. F. Donoghue, C. Ramirez and G. Valencia, The Spectrum of QCD and Chiral Lagrangians of the
Strong and Weak Interactions, Phys. Rev. D 39, 1947 (1989).
G. Ecker, J. Gasser, A. Pich and E. de Rafael, The Role of Resonances in Chiral Perturbation Theory,
Nucl. Phys. B 321, 311 (1989).
13. K. S. Stelle, Classical Gravity with Higher Derivatives, Gen. Rel. Grav. 9, 353 (1978).
14. J. Z. Simon, Higher Derivative Lagrangians, Nonlocality, Problems And Solutions, Phys. Rev. D
41, 3720 (1990).
J. Z. Simon, The Stability of at space, semiclassical gravity, and higher derivatives, Phys. Rev. D
43, 3308 (1991).
15. R. P. Feynman, Quantum theory of gravitation, Acta Phys. Polon. 24, 697 (1963).
16. B. S. DeWitt, Quantum Theory of Gravity. 2. The Manifestly Covariant Theory, Phys. Rev. 162,
1195 (1967).
17. G. t Hooft and M. J. G. Veltman, One loop divergencies in the theory of gravitation, Annales
Poincare Phys. Theor. A 20, 69 (1974).
18. D. C. Dunbar and P. S. Norridge, Calculation of graviton scattering amplitudes using string based
methods, Nucl. Phys. B 433, 181 (1995) [hep-th/9408014].
19. J. F. Donoghue and T. Torma, On the power counting of loop diagrams in general relativity, Phys.
Rev. D 54, 4963 (1996) [hep-th/9602121].
20. S. Weinberg, Infrared photons and gravitons, Phys. Rev. 140, B516 (1965).
21. N. E. J. Bjerrum-Bohr, J. F. Donoghue, B. R. Holstein, Quantum gravitational corrections to the
nonrelativistic scattering potential of two masses, Phys. Rev. D67, 084033 (2003). [hep-th/0211072].
22. I. B. Khriplovich, G. G. Kirilin, Quantum power correction to the Newton law, J. Exp. Theor. Phys.
95, 981-986 (2002). [gr-qc/0207118].
23. A. F. Radkowski, "Some aspects of the source description of gravitation" Ann. of Phys. 56, 319 1970.
M. J. Duff, Quantum corrections to the schwarzschild solution, Phys. Rev. D 9, 1837 (1974).
J. F. Donoghue, Leading quantum correction to the Newtonian potential, Phys. Rev. Lett. 72, 2996
(1994) [gr-qc/9310024].
H. W. Hamber and S. Liu, On the quantum corrections to the Newtonian potential, Phys. Lett. B
357, 51 (1995) [hep-th/9505182].
I. J. Muzinich and S. Vokos, Long range forces in quantum gravity, Phys. Rev. D 52, 3472 (1995)
[hep-th/9501083].
A. A. Akhundov, S. Bellucci and A. Shiekh, Gravitational interaction to one loop in effective quantum
gravity, Phys. Lett. B 395, 16 (1997) [gr-qc/9611018].
24. Y. Iwasaki, Quantum theory of gravitation vs. classical theory. - fourth-order potential, Prog. Theor.
Phys. 46, 1587 (1971).
25. S. N. Gupta and S. F. Radford, Quantum eld theoretical electromagnetic and gravitational two
particle potentials, Phys. Rev. D 21, 2213 (1980).
26. B. R. Holstein and J. F. Donoghue, Classical physics and quantum loops, Phys. Rev. Lett. 93,
201602 (2004) [hep-th/0405239].
27. W. D. Goldberger and I. Z. Rothstein, An Effective eld theory of gravity for extended objects,
Phys. Rev. D 73, 104029 (2006) [hep-th/0409156].
W. D. Goldberger and A. Ross, Gravitational radiative corrections from effective eld theory, Phys.
Rev. D 81, 124015 (2010) [arXiv:0912.4254 [gr-qc]].
R. A. Porto and I. Z. Rothstein, Next to Leading Order Spin(1)Spin(1) Effects in the Motion of
Inspiralling Compact Binaries, Phys. Rev. D 78, 044013 (2008) [Erratum-ibid. D 81, 029905 (2010)]
[arXiv:0804.0260 [gr-qc]].
28. J. F. Donoghue, B. R. Holstein, B. Garbrecht and T. Konstandin, Quantum corrections to the
Reissner-Nordstrom and Kerr-Newman metrics, Phys. Lett. B 529, 132 (2002) [hep-th/0112237].
29. N. Hambli and C. P. Burgess, Hawking radiation and ultraviolet regulators, Phys. Rev. D 53, 5717
(1996) [hep-th/9510159].
30. J. F. Donoghue, When effective eld theories fail, PoS EFT 09, 001 (2009) [arXiv:0909.0021 [hep-
ph]].
31. D. J. Toms, Quantum gravity, gauge coupling constants, and the cosmological constant, Phys. Rev.
D 80, 064040 (2009) [arXiv:0908.3100 [hep-th]].
32. E. Kiritsis, C. Kounnas, Infrared regularization of superstring theory and the one loop calculation of
coupling constants, Nucl. Phys. B442, 472-493 (1995). [hep-th/9501020].
33. S. P. Robinson and F. Wilczek, Gravitational correction to running of gauge couplings, Phys. Rev.
Lett. 96, 231601 (2006) [arXiv:hep-th/0509050].
34. A. R. Pietrykowski, Gauge dependence of gravitational correction to running of gauge couplings,
Phys. Rev. Lett. 98, 061801 (2007) [arXiv:hep-th/0606208].
D. J. Toms, Quantum gravity and charge renormalization, Phys. Rev. D 76, 045015 (2007)
[arXiv:0708.2990 [hep-th]].
D. Ebert, J. Plefka and A. Rodigast, Absence of gravitational contributions to the running Yang-Mills
coupling, Phys. Lett. B 660, 579 (2008) [arXiv:0710.1002 [hep-th]].
Y. Tang and Y. L. Wu, Gravitational Contributions to the Running of Gauge Couplings,
arXiv:0807.0331 [hep-ph].
A. Rodigast and T. Schuster, Gravitational Corrections to Yukawa and
4
Interactions, Phys. Rev.
Lett. 104, 081301 (2010) [arXiv:0908.2422 [hep-th]].
35. D. J. Toms, Quantum gravitational contributions to quantum electrodynamics, Nature 468, 56-59
(2010). [arXiv:1010.0793 [hep-th]].
H. -J. He, X. -F. Wang, Z. -Z. Xianyu, Gauge-Invariant Quantum Gravity Corrections to Gauge
Couplings via Vilkovisky-DeWitt Method and Gravity Assisted Gauge Unication, [arXiv:1008.1839
[hep-th].
J. E. Daum, U. Harst and M. Reuter, Running Gauge Coupling in Asymptotically Safe Quantum
Gravity, JHEP 1001, 084 (2010) [arXiv:0910.4938 [hep-th]].
O. Zanusso, L. Zambelli, G. P. Vacca and R. Percacci, Gravitational corrections to Yukawa systems,
Phys. Lett. B 689, 90 (2010) [arXiv:0904.0938 [hep-th]].
36. M. M. Anber, J. F. Donoghue and M. El-Houssieny, Running couplings and operator mixing in the
gravitational corrections to coupling constants, Phys. Rev. D 83, 124003 (2011) [arXiv:1011.3229
[hep-th]].
37. M. M. Anber and J. F. Donoghue, On the running of the gravitational constant, Phys. Rev. D 85,
104016 (2012) [arXiv:1111.2875 [hep-th]].
38. D. J. Toms, Quadratic divergences and quantum gravitational contributions to gauge coupling con-
stants, Phys. Rev. D 84, 084016 (2011).
39. J. Ellis and N. E. Mavromatos, On the Interpretation of Gravitational Corrections to Gauge Cou-
plings, Phys. Lett. B 711, 139 (2012) [arXiv:1012.4353 [hep-th]].
40. M. Buchler and G. Colangelo, Renormalization group equations for effective eld theories, Eur.
Phys. J. C 32, 427 (2003) [arXiv:hep-ph/0309049].
M. Bissegger and A. Fuhrer, A renormalizable effective theory for leading logarithms in ChPT, Eur.
Phys. J. C 51, 75 (2007) [arXiv:hep-ph/0701132].
J. Bijnens, Chiral Perturbation Theory Beyond One Loop, Prog. Part. Nucl. Phys. 58, 521 (2007)
[arXiv:hep-ph/0604043].
J. Bijnens and L. Carloni, Leading Logarithms in the Massive O(N) Nonlinear Sigma Model, Nucl.
Phys. B 827, 237 (2010) [arXiv:0909.5086 [hep-ph]].
N. Kivel, M. V. Polyakov and A. Vladimirov, Chiral Logarithms in the Massless Limit Tamed, Phys.
Rev. Lett. 101, 262001 (2008) [arXiv:0809.3236 [hep-ph]].
J. Koschinski, M. V. Polyakov and A. A. Vladimirov, Leading Infrared Logarithms from Unitarity,
Analyticity and Crossing, Phys. Rev. D 82, 014014 (2010) [arXiv:1004.2197 [hep-ph]].
41. Weinberg, S., Ultraviolet divergencies in quantum theories of gravitation, in Hawking, .W., and
Israel, W., eds., General Relativity: An Einstein Centenary Survey, (Cambridge University Press,
Cambridge, U.K.; New York, U.S.A., 1979).
42. M. Niedermaier, M. Reuter, The asymptotic safety scenario in quantum gravity, Living Rev. Rel. 9,
5 (2006).