Stability of spherically symmetric spacetimes in metric f(R) gravity

Kimmo Kainulainen
*
and Daniel Sunhede
+
Department of Physics, P.O. Box 35 (YFL), FL-40014 University of Jyvaskyla, Finland and Helsinki Institute of Physics and
Department of Physical Sciences, P.O. Box 64, FI-00014 University of Helsinki, Finland
(Received 6 March 2008; published 8 September 2008)
We consider stability properties of spherically symmetric spacetimes of stars in metric ](R) gravity. We
stress that these not only depend on the particular model, but also on the specic physical conguration.
Typically congurations giving the desired
PPN
- 1 are strongly constrained, while those corresponding
to
PPN
- 1,2 are less affected. Furthermore, even when the former are found strictly stable in time, the
domain of acceptable static spherical solutions typically shrinks to a point in the phase space. Unless a
physical reason to prefer such a particular conguration can be found, this poses a naturalness problem for
the currently known metric ](R) models for accelerating expansion of the Universe.
DOI: 10.1103/PhysRevD.78.063511 PACS numbers: 04.50.Kd, 95.35.+d, 98.80.k
I. INTRODUCTION
The observation that the expansion of the Universe
appears to be accelerating [1,2] has provoked discussion
of a number of models for extended gravity involving
nonlinear interactions in the Ricci scalar R:
S
1
2k
Z
d
4
x

g
_
jR +](R)| +S
m
. (1)
Here k 8rG, S
m
is the usual matter action and ](R)
describes the new physics in the gravity sector; setting
](R) 2 corresponds to the canonical Einstein-
Hilbert action in general relativity (GR) with a cosmologi-
cal constant . The idea is that if cosmological data could
be tted by the use of some nontrivial function ](R), one
might avoid the theoretical difculties and ne-tuning
issues related to a pure cosmological constant. However,
it has been shown that when understood as a metric theory,
the action (1) can lead to predictions that are not consistent
with Solar System measurements [35]. While observa-
tions require a parameter }
PPN
1} & 10
4
[6] in the
parametrized post-Newtonian (PPN) formalism, the value
predicted in metric ](R) theories is typically
PPN
- 1,2.
This is certainly the case [35] for the rst simple ](R)
models suggested in the literature [7,8]. It is however
difcult to make a completely generic prediction of this
result and there have been many arguments both for [3
5,9] and against [1015] metric ](R) gravity failing Solar
System tests. In particular, more complicated ](R) func-
tions have since been suggested which claim to yield

PPN
- 1 [1114].
In this paper we set up the conditions which the function
](R) must fulll, so that a solution to the eld equations
which is compatible with Solar System observations exists,
in particular, with
PPN
- 1. However, we will also argue
that the mere existence of such a solution does not imply
that a model is consistent with observations. Since metric
](R) gravity is a fourth-order theory, spacetime geometry
and matter are not in as strict a correspondence as in
general relativity; depending on the boundary conditions
on the metric, a given matter distribution can be consistent
with different static spacetimes and with different values of

PPN
. Moreover, no physical principle tells us that only the
boundary conditions corresponding to
PPN
- 1 solutions
should be acceptable. The question is then, which solutions
are the most natural ones? How plausible is it that the
collapse of a protostellar dust cloud leads to the formation
of the spacetime observed in the Solar System? To answer
these questions one would ideally like to study the full
dynamical collapse, and given a domain of reasonable
initial conditions, determine the attractor in the congura-
tion space of possible solutions. This computation is be-
yond the scope of this paper however. We will instead
approach the problem by studying how the time stability
argument constrains the phase space of congurations with
the desired properties.
The conditions that a generic metric ](R) model should
satisfy in order to yield acceptable solutions are: rst, the
Ricci scalar should closely follow the trace of the energy-
momentum tensor inside a changing matter distribution,
where at the same time the dimensionless quantities ],R
and E ],R should remain much smaller than 1 at
regions of high density. Second, the effective mass termm
2
R
for a perturbation in the Ricci scalar should be positive in
order to assure that the GR-like
PPN
- 1 congurations
are stable in time. Third, the mass m
2
R
should remain small
so that a nite domain of static, GR-like congurations
exist. This is guaranteed if m
2
R
& 1,r
2
c
, where r
c
is the
radius of the Sun. If this last condition is not fullled, the
domain of GR-like congurations shrinks to essentially a
point in the phase space, while a continuum of equally
good, but observationally excluded, solutions still exists. In
such a case the credibility of the theory requires an argu-
ment as to why the particular GR-like conguration should
be preferred. None of the models so far proposed in the
*
Kimmo.Kainulainen@phys.jyu.
+
Daniel.Sunhede@phys.jyu.
PHYSICAL REVIEW D 78, 063511 (2008)
1550-7998,2008,78(6),063511(8) 063511-1 2008 The American Physical Society
literature, including Refs. [1214], satisfy all of these
constraints, and we also failed to construct a model that
would. Largely this failure comes from the difculty to
keep both the function E and m
2
R
1,(3E
,R
) small simul-
taneously when the Ricci scalar follows the matter distri-
bution, R - k.
It should be noted that the above considerations only
apply for the desired GR-like
PPN
- 1 congurations
when a model is tuned to mimic a (very small) cosmologi-
cal constant. Metric models with a true cosmological con-
stant plus some additional sufciently small ](R)
correction can be perfectly ne. Thus the above arguments
do not exclude generic ](R) modications, such as might
arise from quantum corrections, to the Einstein-Hilbert
action. One should also note that it is in general easy to
construct stable attractor solutions yielding
PPN
- 1,2 in
metric ](R) theory. It is the precision data from the Solar
System which makes these solutions unacceptable.
The paper is organized as follows. We start by reviewing
the Solar System constraints in Sec. II. We consider the
Dolgov-Kawasaki time instability [16] in Sec. III and
discuss the corresponding stability criterion for spherically
symmetric congurations. Section IV considers static con-
gurations and the possibility for metric ](R) gravity to
follow stable, GR-like solutions that are compatible with
Solar System constraints. We nd that the condition for
nding a nite domain of boundary conditions giving rise
to a GR-like metric is nearly orthogonal to the time stabil-
ity condition. Finally, Sec. V contains our conclusions and
discussion.
II. SOLAR SYSTEM CONSTRAINTS AND THE
PALATINI TRACK
Let us begin by reviewing the main constraints from the
Solar System observations on static solutions in metric
](R) gravity. Varying the action (1) with respect to the
metric gives the equation of motion:
(1+E)R
:

1
2
(R+])g
:
\

\
:
E+g
:
hEkT
:
,
(2)
where E ]
,R
],R and h g
:
\

\
:
. Taking the
trace of this equation one nds:
hE
1
3
(R(1 E) +2])
1
3
kT. (3)
If E -0 and ] -R this equation reduces to the standard
algebraic GR relation between the Ricci scalar and the
trace of the energy-momentum tensor T. In a generic
metric ](R) theory R is a dynamical variable however,
and the theory may exhibit an instability which we will
discuss in the next section. Assuming a static, spherically
symmetric metric g
:
,
Js
2
g
:
x

x
:
e
A(r)
di
2
+e
B(r)
dr
2
+r
2
d
2
, (4)
the full eld equations. (2) reduce to the following source
equations for the metric functions A and B in the weak eld
limit (to rst order in small quantities):
(rB)
/
- kr
2

1
1
3

1 +3E
1 +E

R
k
1 +
E
2
+
]
2R
1 +E

rA
/
, (5)
A
/
-
1
1 +

B
r

r
2(1 +E)

ER ] +
4
r
E
/

, (6)
where rE
/
,2(1 +E), a prime refers to a derivative
with respect to r, and we have neglected pressure so that
T - . The parameter and the terms in the square
brackets highlight the deviation from general relativity.
The value of
PPN
- B,A far away from a gravitational
source depends on the continuous evolution of A and B
throughout the Sun. It is particularly sensitive to the evo-
lution through the core where the density is the highest.
Hence, to obtain
PPN
- 1 and the correct gravitational
strength in the Solar System, the only solution, not obvi-
ously dependent on an enormous amount of ne-tuning
[17], is that the extra terms in Eqs. (5) and (6) must remain
small throughout the interior of the Sun. Now, if the extra
terms can be neglected in the B
/
equation, one nds that
B & 10
6
throughout the interior of the Sun [5]. It then
becomes clear from the A
/
equation that ],R, E, and rE
/
need to be very small compared to 1. However, to make the
correction vanish in the B
/
equation one in addition needs
to require that the Ricci scalar traces the matter density
R,k - 1. So, barring perhaps some fantastic ne-
tunings, the only possibility is that one nds a congura-
tion for which:
E <1, ],R <1, and R - k. (7)
Note that this limit was also discussed in Ref. [12]. Here we
see that the above conditions are a necessary requirement
for fullling the local gravity constraints.
In GR the equation R kT is of course exact (re-
member that we are neglecting pressure throughout so that
T - ), but this is in general very difcult to arrange in a
metric ](R) model. The problem lies in the dynamical
nature of the Ricci scalar in metric ](R) gravity. To see
this, consider the static trace Eq. (3) in the weak eld limit:
E
//
+
2
r
E
/

1
3
(R k ER +2])
1
3
((E) k),
(8)
where we have again assumed that pressure is negligible.
Assume now that E E
0
at the center of the Sun. If the
nonlinear term (E) is small compared to k, then the
solution for E becomes:
E(r)
Z
r
0
dr
/
2Gm(r
/
)
3r
/2
+E
0
, (9)
KIMMO KAINULAINEN AND DANIEL SUNHEDE PHYSICAL REVIEW D 78, 063511 (2008)
063511-2
where m(r)
R
r
0
dr
/
4rr
/2
. In the case of the Sun this
implies that E evolves only little: }E(r) E
0
} & 10
6
[5].
The solution (9) is in general not compatible with R - k
as required by the conditions (7).
Let us now set (E
0
) k
0
at r 0. If the gradients
somehow remain small throughout the evolution, then the
solution follows the Palatini trace equation:
(E) R(1 E) +2] 8rG. (10)
For small E and ],R this evolution would be consistent
with the condition R - k and, since we are following the
Palatini track, give
PPN
- 1 (see e.g. [18]). Whether such
a solution actually exists is more difcult to prove.
However, one can study under which conditions such a
solution, if it exists, would be an attractor and whether it
would also be sufciently stable in time.
Finally, let us note that if (E) <k so that E is given
by Eq. (9), small E and ],R result in a different class of
solutions with R,k <1. In this case the eld Eqs. (5)
and (6) reduce to
(rB)
/
-
2
3
kr
2
, (11)
A
/
-
B
r
2E
/
. (12)
It is easy to show that together with Eq. (9) these give

PPN
- 1,2.
III. TIME STABILITY
Let us rst consider the time stability of spherically
symmetric congurations in generic ](R) models.
Perturbing around some arbitrary conguration, R(r) -
~
R(r, i) R(r) +uR(r, i), and expanding to rst order in
uR, uR
/
and
_
uR, where the prime refers to a derivative with
respect to r and the dot to a derivative with respect to i, one
can write the trace Eq. (3) in the following form in the
weak eld limit:
(
2
i


\
2
)uR m
2
R
uR +2
E
,RR
E
,R
R
/
uR
/
+

\
2
R
+
1
3E
,R
+
E
,RR
E
,R
(R
/
)
2
, (13)
where kT R(1 E) 2] and
m
2
R

1
3E
,R
(1 E s), (14)
with
suR R(
~
E E) 2(
~
] ]) +

1
~
E
,R
E
,R

+3E
,RR

~
E
,RR
E
,RR

~
E
,R
E
,R

(R
/
)
2
. (15)
Here a tilde is used to denote that a quantity is perturbed,
i.e. it is a function of
~
R(r, i) as opposed to the background
value R(r). Assuming that the conguration R(r) we are
perturbing around is a solution to the static equation, the
second line in Eq. (13) drops out. Moreover, for most cases
the gradient term proportional to uR
/
is completely negli-
gible inside a stellar object and can be dropped as well. See
Fig. 1 for some examples. The behavior of the perturbation
around a static, spherically symmetric solution is thus
governed by the equation
(
2
i


\
2
)uR m
2
R
uR. (16)
Note that the mass m
2
R
only depends on the background
value of the Ricci scalar R(r).
Tables I and II show the components of m
2
R
in some
particular models, and the corresponding parameter values
used in all gures are displayed in Table III. As discussed
FIG. 1 (color online). The gradient term proportional to uR
/
in
Eq. (16), normalized to m
2
R
,r
c
, for various ](R) models (R
k):
4
,R (lower, solid blue),
4
,R +nR
2
,
2
(dashed
green), Hu and Sawicki (dot-dashed red) [12], and
nRlog(R,
2
) (upper, solid black). The actual density prole
used in all gures corresponds to the known density prole of the
Sun with a central density of 150 g,cm
3
and with a roughly
exponential dependence on r. We have also superimposed a
constant dark matter distribution on the prole of the Sun with

DM
0.3 GeV,cm
3
.
FIG. 2 (color online). The parameter E given as a function of
the radius for various ](R) models (R k):
4
,R (lower,
solid blue),
4
,R +nR
2
,
2
(dashed green), Hu and Sawicki
(dot-dashed red) [12], and nRlog(R,
2
) (upper, solid black).
STABILITY OF SPHERICALLY SYMMETRIC SPACETIMES . . . PHYSICAL REVIEW D 78, 063511 (2008)
063511-3
in the previous section, E needs to be small compared to
one for GR-like congurations (see Fig. 2, we will discuss
this constraint further below). When this is the case one
typically nds that also s <1, so that m
2
R
- 1,3E
,R
. It
then follows that if E
,R
<0, then m
2
R
<0 and the coef-
cient of uR is negative for the conguration in question, so
that system exhibits an instability. This is the instability
rst found by Dolgov and Kawasaki in the context of an
](R)
4
,R model [16] (see Ref. [19] for a more
general case). It is important to note that the instability
depends not only on the model, but also on the particular
conguration. Certain congurations in a given model are
more stable than others and the instability may even vanish
in some cases.
The nature of the instability is most transparent in the
special case with constant curvature. Then m
R
is a constant
and one can obtain an exact solution for uR(r, i).
Expanding uR in Fourier modes, one nds that a mode
with wave vector

k has the time dependence
uR
k
(

k, i) e
|

k
2
+m
2
R
_
i
, (17)
so that for negative m
2
R
1,3E
,R
, all modes with k <}m
R
}
are unstable. This does not necessarily rule out a model
however. If for example }m
R
} H
0
, then the instability
time is much longer than the lifetime of the Solar System
and the model is safe. Moreover, whenever }m
R
}
1
is much
larger than the size of the physical system under consid-
eration, only modes corresponding to scales much larger
than the system are unstable and this can not alter its local
geometry.
Now, assume that we have a GR-like solution, such that
R kT - 8rG. One then has
R

2
10
29

g,cm
3

, (18)
where we have used - 0.73k
crit
. Hence, for a pure

4
,R model the mass squared is on the order of
m
2
R
(10
26
s)
2

g,cm
3

3
. (19)
This system is violently unstable at normal densities for all
scales larger than 10
18
m, if is xed to account for
the present accelerating expansion of the Universe.
As was pointed out by Dick [20] and later discussed by
Nojiri and Odintsov [21], adding a conformal term
nR
2
,
2
can stabilize this system; for ](R)
4
,R +
nR
2
,
2
the previous approximation for the mass reads:
m
2
R

R
3
6
4

1
1 nR
3
,
6

2
n
, (20)
where the last step assumes that nR
3
,
6
>1. If n 1 this
may be true even for R
2
so that one always nds a
very small positive mass m
2
R

2
,n (10
18
s)
2
.
However, the above stabilization mechanism runs into
problems with the conditions in Eq. (7). Indeed, for n 1
and R - k >
2
one has
E

4
R
2
+n
R

2
>1, (21)
so that the conguration would clearly not be GR-like [22].
The problem is that changing E modies the effective
TABLE II. The parameter s for different ](R) models, where D (1
n1
n+1
,c
2

R)
(1 +1,c
2

R
n
) and a bar indicates that a quantity is dimensionless and measured in units of ,
e.g.

R R,
2
. In the third model above, originally suggested in Ref. [12], we have only kept
the leading terms in s.
](R) s

4
R
3(1 +
kT
R
) +
5
4
R
2
+3j3E
,R
|(
R
/
R
)
2

4
R
+n
R
2

2

2n
2
R+2n,

R
2
3(1+ k

T ,

R),

R
3
+5,

R
5
n1,

R
3
3
4n1,

R
3

R
3
(n1,

R
3
)
2
j3E
,R
|(
R
/
R
)
2

2
c
1
(R,
2
)
n
c
2
(R,
2
)
n
+1

n+2
D
(1 +
k

T

R

2c
1
c
2

R(1+1,c
2

R
n
)
2
)
n+2
D
2
j3E
,R
|(
R
/
R
)
2
nRlog
R

2
1
kT
R
3nlog
R

2
+j3E
,R
|(
R
/
R
)
2
TABLE I. The function E and the dominant term of the mass squared m
2
R
(1 E s),3E
,R
for different ](R) models, where

R R,
2
.
](R) E 1,3E
,R

4
R

4
R
2

R
3
6
4

4
R
+n
R
2

2
1

R
2
+2n

R

2
6(n1,

R
3
)

2 c
1
(R,
2
)
n
c
2
(R,
2
)
n
+1

c
1
n
c
2
2

R
n+1
(1+1,c
2

R
n
)
2

2
c
2
2

R
n+2
3c
1
n(n+1)
(1+1,c
2

R
n
)
3
1
n1
n+1
,c
2

R
n
nRlog
R

2
n(1 +log
R

2
)
R
3n
KIMMO KAINULAINEN AND DANIEL SUNHEDE PHYSICAL REVIEW D 78, 063511 (2008)
063511-4
strength of the gravitational constant G
eff
G,(1 +E),
which controls the buildup of the gravitational potential
inside the star. In fact, for n 1 the effect is so strong that
it would weaken the gravitational force so much as to
prohibit the growth of any density contrasts much above
the critical density. This argument can be turned around to
a constraint: in order for the function E to remain small
inside the densest objects we have reasonably accurate
information on, the neutron stars, one has to have
nk
nucl
,
2
<1, where
nucl
is the nuclear density [23].
Since
2
k
crit
we nd that
n &

crit

nucl
10
45
. (22)
This is quite a stringent constraint, but it does not rule out
the model based on the required time stability. Indeed, for
example, with n 10
47
one has n(k)
3
,
6
1 when
10
16

crit
. For any density higher than this value, the
system is stable in time with a very large positive mass
squared given by the formula in Eq. (20): m
2
R

2
,n
(10
6
s)
2
. There is a caveat to this argument however,
since for these parameters the gradient term proportional to
uR
/
becomes very large inside the Sun (see Fig. 1) and the
simplied Eq. (16) can no longer be trusted.
The complete mass squared function for the model with
a ne-tuned conformal nR
2
term, using the exact expres-
sion (14), is shown in Fig. 3 (dashed green curve). The
lower panel displays the absolute value }m
2
R
} and the upper
panel the sign of m
2
R
. In the model at hand the mass would
remain large and positive throughout the entire interior of
the Sun, which is the necessary condition for time stability.
The GR-like conguration does become unstable at low
densities, but this would not necessarily change the value
of
PPN
in the Solar System.
Figure 3 also displays m
2
R
for several other models (for
parameter values used in each model see Table III): the
solid blue line represents the simple
4
,R model, which
has a very large negative mass inside the Sun, and the dash-
dotted red curve shows the mass function in a model by Hu
and Sawicki (HS) [12]. The HS model fullls the condi-
tions (7) by construction, and its very large positive mass
guarantees time stability. The fact that m
2
R
becomes nega-
tive around r 6r
c
in the HS model is caused by the s
term in the complete expression (14), but this does not
necessarily have any effect on
PPN
. Moreover, this feature
is sensitive to the particular form of the exterior density
prole (where we have neglected, for example, the Solar
wind) and it is not important for our main results. Overall
one sees that the expression m
2
R
- 1,3E
,R
is a very good
approximation for the rst three models described in
Table I, except for a small region around r 6r
c
where
the s term may come into play. For the logarithmic model
this approximation is only good for very low densities and
we will discuss this in more detail in Sec. IV.
We can summarize this section as follows: the stability
of a static, spherically symmetric GR-like conguration
with R - k is predominantly governed by the mass term
FIG. 3 (color online). The mass squared m
2
R
in units 1,r
2
c
for
various ](R) models (R k):
4
,R (upper, solid blue),

4
,R +nR
2
,
2
(dashed green), Hu and Sawicki (dot-dashed
red) [12], and nRlog(R,
2
) (lower, solid black). The horizontal
solid gray line corresponds to the limit 1,r
2
c
. For the
4
,R and
nRlog(R,
2
) models we have also plotted 1,3E
,R
with dotted
blue and dotted black lines, respectively. The upper panel dis-
plays the corresponding sign of the mass squared where we have
excluded the
4
,R model for which m
2
R
is strictly negative.
TABLE III. Chosen parameter values for the different ](R) models in Figs. 13, where 0.73k
crit
. For the third model,
originally suggested in Ref. [12], we have used values given in the original publication. Note that a value n 4, which was also
discussed in [12], would result in an even larger value of m
2
R
in this scenario.
](R) Parameter values

4
R

2
4,

3
_

4
R
+n
R
2

2

2
4,

3
_
, n 10
47

2
c
1
(R,
2
)
n
c
2
(R,
2
)
n
+1

2
(8315 Mpc)
1
, n 1, c
1
,c
2
6 0.76,0.24, c
1
,c
2
2
10
6
41
n+1
,n
nRlog
R

2

2
4e
(1n),n
, n 1, log10
32
STABILITY OF SPHERICALLY SYMMETRIC SPACETIMES . . . PHYSICAL REVIEW D 78, 063511 (2008)
063511-5
m
2
R
1,3E
,R
. If E
,R
<0, all perturbations with wave-
lengths larger than 1,}m
R
} will be unstable so that for a
large mass }m
R
}, the curvature inside a stellar object will
evolve rapidly before some nonperturbative effect stabil-
izes the system. Hence, in order for a model to be stable,
m
2
R
should be positive throughout the Sun for GR-like
congurations [24]. Both the model with a ne-tuned
conformal nR
2
term (apart from the caveat mentioned
above) and the HS scenario do satisfy all constraints dis-
cussed so far.
IV. AN UPPER BOUND ON m
2
R
?
As mentioned in the Introduction, a given matter distri-
bution can be consistent with many different static geome-
tries, depending on how the boundary conditions are
dened at the center of the star. One always requires that
the exterior metric is asymptotically at and so different
solutions are characterized by different values of
PPN
.
There appears to be no a priori preference of one solution
to another and indeed the question is: what is the most
probable conguration to arise through gravitational col-
lapse? Lacking a dynamical calculation we are restricted
here to study how special the GR-like solutions are in the
phase space.
Consider a static solution R(r) R
T
(r) +u(r) where
R
T
is the solution to the Palatini trace equation and
u,R
T
<1 so that R remains very close to the Palatini
track. Note that the function u(r) is not a true perturbation
since R
T
(r) is not a solution to the complete metric trace
Eq. (3). However, one can easily obtain the equation gov-
erning u via Eq. (13), giving

\
2
u
1
3E
,R
(1 E s)u 2
E
,RR
E
,R
R
/
T
u
/

\
2
R
T

E
,RR
E
,R
(R
/
T
)
2
, (23)
where E and its derivatives are functions of the back-
ground value R
T
(r). Similarly, the perturbed quantities
in the denition for s are functions of R(r) R
T
(r) +
u(r).
In analogy with the above analysis for uR(r, i), let us
consider a constant density object so that R
T
const,
giving

\
2
u m
2
R
u. (24)
The mass term m
2
R
is of course the same mass that appears
in the equation for uR(r, i). Now, for m
2
R
<0, the solution
for u(r) is decaying so that the Palatini track acts as an
attractor for the solution R(r). This is exactly the behavior
that was demonstrated by solving the full eld equations in
Ref. [5] for a ](R)
4
,R model. However, if m
2
R
>0,
the solution for u(r) will also contain a growing compo-
nent:
u(r)
C
1
r
e
+m
R
r
+
C
2
r
e
m
R
r
. (25)
The ne-tuning problem we have to face is manifest from
this equation: the time stability argument of the preceding
section requires that m
2
R
>0, so that Eq. (25) with its
growing mode is what describes the deviation away from
GR-like solutions.
From Sec. III we already know that setting up a GR-like
conguration requires ne-tuning at the center of the star.
However, what Eq. (25) implies is worse: starting fromr
0 we could always choose a boundary condition (E
0
, E
/
0
)
that kills the growing mode, but as any perturbation around
this solution would be exponentially enhanced, the bound-
ary condition must be set with an incredible precision when
m
2
R
is large. Numerically such a solution can always be
found by use of a differentiation method that kills the
growing mode as was done in Ref. [12]. However, different
choices of boundary conditions would lead to other physi-
cally, but not observationally, acceptable spacetimes. For
example, if one starts from a point a little off from the GR
track, u initially grows exponentially pulling the solution
away from R - k. Then the nonlinear terms typically
become negligible in Eq. (8) and R(E) starts to approach
the solution of Eq. (9). For R,k <1 this limit corre-
sponds to the evolution of A and B given by Eqs. (11) and
(12), which leads to
PPN
- 1,2. Thus, for a large m
2
R
the
nearly singular static GR-like solution is surrounded by a
continuum of equally acceptable congurations, however
with observationally excluded values for
PPN
.
Hence, given no physical reason to prefer a given set of
boundary conditions, it would appear more natural to ex-
pect that the metric around a generic star would correspond
to
PPN
different from one. As stated above, to make a
denitive statement would require solving the dynamical
problem of collapse, but this is beyond the scope of the
present work. Nevertheless we believe that we have iden-
tied a potential problem for metric ](R) gravity models:
for an ](R) model to be credible, it is not sufcient to
provide a mere proof of existence of a GR-like solution,
but one should also give an argument as to why this
particular solution is preferred.
The situation would be ameliorated if the growing mode
is not excluded, but the length scale dictating the growth of
the perturbations, 1,m
R
, is small enough. Roughly one
should have
m
2
R
&
1
r
2
c
, (26)
throughout the Sun. However, both the HS scenario and the
ne-tuned ](R)
4
,R +nR
2
,
2
model fail this con-
straint by a large margin, as can be seen from Fig. 3. This is
also the case for the model in Ref. [11], where a stabilizing
conformal term creates a behavior very similar to the nR
2
model discussed here. The same argument also applies to
more recent models introduced in Refs. [13,14]. These
KIMMO KAINULAINEN AND DANIEL SUNHEDE PHYSICAL REVIEW D 78, 063511 (2008)
063511-6
scenarios behave very similar to the HS model at late
times, but were designed to also account for ination at
very high energies. For example, for the model suggested
in Ref. [14],
](R)
nR
m+l
R
n
1 +R
l
, (27)
where the authors chose m l n for simplicity, and n
2; one can show that the mass squared is given by [14]
m
2
R
+
R
I
3n(n 1)

R
I
R

n1
. (28)
Here R
I
(10
15
GeV)
2
is set to the scale of ination, and
so this mass is enormous in comparison with the bound
(26) inside the Sun.
The generic problem is that a small value of m
2
R

1,3E
,R
requires a large value for E
,R
. However, at the
same time one also needs E <1 in order to obtain a
reasonable gravitational potential. This tension is what
makes it difcult to nd a suitable function ](R). Let us
illustrate the problem further by trying to construct an
explicit model by rst making sure that the toughest re-
quirement is satised. At the center of Sun R -
k 10
31
10
3
,r
2
c
, which is much smaller than the
upper limit on m
2
R
. Thus we can take
m
2
R
+R. (29)
Using this together with the formula E
,R
1,m
2
R
, we can
construct a candidate model:
](R) nRlog
R

2
, (30)
where
2
4e
(1n),n
in order to obtain the desired
accelerating expansion of the Universe at present.
Furthermore, demanding that E <1 in the interior of the
Sun yields n & 0.01 so that
2
* e
100
[25]. So, curi-
ously enough the Solar System constraints would force this
model to create the desired accelerating expansion without
an extremely small energy scale

_
.
Unfortunately, there is a aw in the above argumenta-
tion, since we implicitly assumed that s <1 by assuming
m
2
R
1,3E
,R
. This assumption was ne for the discussion
in the previous sections, but it fails here. Indeed, when E
,R
is large, the gradient term proportional to j3E
,R
|(R
/
,R)
2
in
s may also become large (see Table II). We can estimate
the size of this term using an exponential density prole for
the Sun
0
,(1 +e
((rr
c
)
), where ( 10r
1
c
:

R
/
R

2
-

2
(
2
100
1
r
2
c
. (31)
Now, since 3E
,R
* r
2
c
one gets j3E
,R
|(R
/
,R)
2
>1 and our
simple estimate for the mass fails. A more careful estimate
in the model (30) nds that the mass is dominated by the
gradient term and one has
m
2
R
- m
2
R
}
grad

R
/
R

2
100
1
r
2
c
. (32)
This mass squared is actually negative, so the ne-tuning
problem is no longer an issue. However, we have recreated
a time instability corresponding to the characteristic length
scale (
1
of the system. This behavior is clearly visible in
Fig. 3 where we have plotted both the full mass squared m
2
R
(solid black) and the bare function 1,3E
,R
(dotted black)
for the nRlog(R,
2
) model.
V. SUMMARYAND DISCUSSION
We have shown in this paper that attempts to nd stable
static solutions with
PPN
- 1 in metric ](R) models,
designed to also account for the accelerating expansion
of the Universe, lead to a string of constraints on the model
parameters. One must nd a conguration for which si-
multaneously E ],R and ],R remain small com-
pared to one in the interior of the star, where the strength
of the gravitational eld is built up, while the Ricci scalar
traces the matter distribution: R - k. (See also
Ref. [12]). In addition, for the conguration to be stable
in time, the effective mass termm
2
R
for a perturbation in the
Ricci scalar needs to be either positive [16], or if negative,
}m
R
}
1
must be much larger than the size of the physical
system under consideration. Furthermore, we showed that
unless m
2
R
& 1,r
2
c
, the domain of GR-like static congu-
rations shrinks to essentially a point in the phase space,
while, for example, a continuum of solutions correspond-
ing to
PPN
- 1,2 exists.
Hence, in particular, for large positive m
2
R
, it would
appear more natural to expect that the metric around a
generic star would correspond to
PPN
different from one.
To make a more denitive statement one should solve the
dynamical gravitational collapse of a protostellar dust
cloud, which is beyond the scope of this paper however.
Moreover, to a degree the problem with the boundary
conditions would merely be translated to setting the initial
conditions for the collapse. Nevertheless, we believe that
we have identied a potential problem in that to make a
given metric ](R) model credible, one should give an
argument as to why the GR-like congurations should be
preferred. Otherwise, if the PPN and stability constraints
are supplemented by our ne-tuning argument, it seems
unlikely that any ](R) model can pass the testunless one
gives up the hope that the theory is also responsible for the
accelerating expansion of the Universe. This is because the
condition m
2
R
& 1,r
2
c
, combined with m
2
R
1,3E
,R
, im-
plies that E
,R
needs to be large. However, since at the same
time one needs E <1, a tension is created that makes
nding a suitable function ](R) very difcult.
Let us nally note that while completely GR-like con-
gurations are hard to construct in ](R) models, it does not
mean that such theories would be somehow fundamentally
ill. In fact many metric ](R) theories could describe gravi-
tational physics quite well in most situations; it is the very
STABILITY OF SPHERICALLY SYMMETRIC SPACETIMES . . . PHYSICAL REVIEW D 78, 063511 (2008)
063511-7
precise information on
PPN
from Solar System experi-
ments which eventually forces one to set R - k. Indeed,
if one looks even at the simplest model with ](R)

4
,R, one nds that setting R essentially to any other
value than k gives
PPN
- 1,2 [5]. Moreover, whenever
one has R
2
, one nds m
2
R

2
, so that these congu-
rations are effectively free of any stability problems. Thus,
one sees that the Dolgov-Kawasaki instability and ne-
tuning problems depend not only on the theory, but also on
the particular given conguration within a given model.
Even in the simplest
4
,R model one can construct
sufciently stable spacetimes for stellar objects that are
consistent with the accelerating expansion of the Universe;
these are only excluded from describing reality by the very
precise PPN constraints.
ACKNOWLEDGMENTS
This work was partly (D. S.) supported by a grant from
the Finnish Cultural Foundation.
[1] P. Astier et al., Astron. Astrophys. 447, 31 (2006).
[2] D. N. Spergel et al. (WMAP Collaboration), Astrophys. J.
Suppl. Ser. 170, 377 (2007).
[3] T. Chiba, Phys. Lett. B 575, 1 (2003).
[4] A. L. Erickcek, T. L. Smith, and M. Kamionkowski, Phys.
Rev. D 74, 121501(R) (2006); T. Chiba, T. L. Smith, and
A. L. Erickcek, Phys. Rev. D 75, 124014 (2007).
[5] K. Kainulainen, J. Piilonen, V. Reijonen, and D. Sunhede,
Phys. Rev. D 76, 024020 (2007).
[6] C. M. Will, Living Rev. Relativity 9, 3 (2005).
[7] D. N. Vollick, Phys. Rev. D 68, 063510 (2003).
[8] S. M. Carroll, V. Duvvuri, M. Trodden, and M. S. Turner,
Phys. Rev. D 70, 043528 (2004).
[9] G. J. Olmo, Phys. Rev. Lett. 95, 261102 (2005); Phys. Rev.
D 72, 083505 (2005); X. H. Jin, D. J. Liu, and X. Z. Li,
arXiv:astro-ph/0610854; I. Navarro and K. Van Acoleyen,
J. Cosmol. Astropart. Phys. 02 (2007) 022; G. J. Olmo,
Phys. Rev. D 75, 023511 (2007); K. Henttunen, T.
Multamaki, and I. Vilja, Phys. Rev. D 77, 024040 (2008).
[10] M. C. B. Abdalla, S. Nojiri, and S. D. Odintsov, Classical
Quantum Gravity 22, L35 (2005); T. P. Sotiriou, Gen.
Relativ. Gravit. 38, 1407 (2006); J. A. R. Cembranos,
Phys. Rev. D 73, 064029 (2006); S. Capozziello and A.
Troisi, Phys. Rev. D 72, 044022 (2005); S. Capozziello, A.
Stabile, and A. Troisi, Mod. Phys. Lett. A 21, 2291 (2006);
V. Faraoni, Phys. Rev. D 74, 023529 (2006); T. Faulkner,
M. Tegmark, E. F. Bunn, and Y. Mao, Phys. Rev. D 76,
063505 (2007).
[11] P. J. Zhang, Phys. Rev. D 76, 024007 (2007).
[12] W. Hu and I. Sawicki, Phys. Rev. D 76, 064004 (2007).
[13] S. Nojiri and S. D. Odintsov, Phys. Lett. B 657, 238
(2007).
[14] S. Nojiri and S. D. Odintsov, Phys. Rev. D 77, 026007
(2008).
[15] T. Clifton, Phys. Rev. D 77, 024041 (2008); see also
Ref. [26] for related work.
[16] A. D. Dolgov and M. Kawasaki, Phys. Lett. B 573, 1
(2003).
[17] By this we mean that it is in principle possible that an
evolution of B,A quite different from the one in GR could
actually lead to the same value for the exterior of a star.
However, away from the GR track the outcome becomes
sensitive to the form of the density prole, which leads to
even more uncertainties as to how the actual dynamical
gravitational collapse would proceed.
[18] G. Allemandi, M. Francaviglia, M. L. Ruggiero, and A.
Tartaglia, Gen. Relativ. Gravit. 37, 1891 (2005); K.
Kainulainen, V. Reijonen, and D. Sunhede, Phys. Rev. D
76, 043503 (2007).
[19] V. Faraoni, Phys. Rev. D 74, 104017 (2006).
[20] R. Dick, Gen. Relativ. Gravit. 36, 217 (2004).
[21] S. Nojiri and S. D. Odintsov, Phys. Rev. D 68, 123512
(2003).
[22] In general, E >1 implies that the leading approximation
in Eq. (20) is no longer valid. However, it still holds for
](R)
4
,R +nR
2
,
2
, since the contribution of the
conformal term in E will cancel with the leading term in s
where additional terms remain small compared to one.
[23] The loophole to this argument is if one only follows R -
k inside the Sun, but that the Ricci scalar is allowed to
signicantly deviate from this relation inside neutron stars.
In such a case one could maintain E <1 for much larger
values of n, resulting in a smaller m
2
R
for the Sun. This
scenario seems very contrived however.
[24] Congurations with a small enough negative m
2
R
can be
accepted as well. The minimum requirement is then that at
least all perturbations with wavelengths smaller than the
size of the physical system under consideration should
remain stable.
[25] However, note that since E n(log(R,
2
) +1)
nlog(R,4) 1 +2n, there is some tension between
getting E <1 inside both the Sun and a neutron star.
Although this may indeed prove relevant in further con-
siderations, it is not of importance for the discussion at
hand.
[26] T. Clifton and J. D. Barrow, Phys. Rev. D 72, 103005
(2005).
KIMMO KAINULAINEN AND DANIEL SUNHEDE PHYSICAL REVIEW D 78, 063511 (2008)
063511-8