a

r
X
i
v
:
1
3
1
2
.
0
7
0
5
v
1


[
g
r
-
q
c
]


3

D
e
c

2
0
1
3
Physics at the surface of a star in Eddington-inspired Born-Infeld gravity
Hyeong-Chan Kim
1,
1
School of Liberal Arts and Sciences, Korea National University of Transportation, Chungju 380-702, Korea
We study a phenomenon happening at the surface of a star in Eddington-inspired Born-Infeld
(EiBI) gravity. The star is made of particles, which are eectively described by a polytropic uid.
The EiBI theory was known to have a pathology that singularities happen at a star surface. We
suggest that the gravitational back-reaction on the particles cures the problem. Strong tidal forces
near the (surface) singularity modify the eective equation of state of the particles. The geodesic
deviation equations take the form of the Hookes law with its frequency proportional to the square
root of the curvature. Due to the oscillations, a particle collides with a probing wall measuring the
pressure more often and increases the pressure. With the modied equation of state, the surface is
no longer singular. Therefore, the EiBI gravity is saved from the pathology of surface singularity.
PACS numbers: 04.50.-h, 98.80.-k
Keywords: modied gravity, neutron star, singularity
General relativity (GR) predicts that spacetime singularities can be formed from regular initial data, e.g. in the
gravitational collapse of massive stars and in the early universe. In the presence of a singularity, the quantum
mechanical unitarity may be broken. Recently, a modied theory of gravity, the so-called Eddington-inspired Born-
Infeld (EiBI) theory was proposed in order to resolve these singularities [1]. The theory was further studied diversely [2
17]. EiBI gravity is equivalent to GR in vacuum and does not propagate any degrees of freedom other than massless
gravitons. On the other hand, the theory introduces nonlinear couplings to the matter elds [24], which resolve at
least some of the singularities appearing in Einsteins theory. In the presence of a perfect uid, the big-bang singularity
in early universe is replaced by a freezing, exponentially inating, or a bouncing behavior of the cosmological scale
factor, depending on the equation of state (EoS) of the uid [1, 5]. The gravitational collapse of noninteracting
particles does not lead to singular states in the non-relativistic limit [2, 6]. The stability of EiBI compact stars was
considered in Ref. [7], where it was shown that the standard results of stellar stability theory still hold. The structure
and physical properties of specic classes of neutron, quark and exotic stars in EiBI gravity were studied [8]. They
showed that the EiBI gravity stars are more massive than their GR counterparts. Constraints on the theory have
been considered using solar model [9] and cosmological observations [10, 12]. A tensor instability of the homogeneous
and isotropic universe was found in Ref. [13]. In the presence of a scalar eld, a regular initial state in EiBI gravity
are studied to present the ination paradigm [18], where the tensor instability was shown to be not a problem. Metric
perturbations of the homogeneous and isotropic universe based on EiBI gravity were studied in Ref. [19, 20].
With all these merits, aws of the theory were also discovered. In Ref. [11], it was shown that the EiBI theory is
reminiscent of Palatini f(R) gravity [21] and that it shares the pathology such as the curvature singularity at the
surface of polytropic stars. In Ref. [17], similar singularities were shown to exist when a phase transition happens
inside a star. These observations cast serious doubt on the viability of EiBI gravity. In this letter, we reexamine the
singularity by checking the motion of particles at the surface. We show that, in the presence of the singularity, the
(eective) EoS of particles near the surface is modied so that the singularity is removed. This happens because the
pressure acted by the particles increases due to rapid oscillations in geodesic deviation motions, which is the inuence
of the high curvature.
EiBI gravity is described by the action [1]
S
EiBI
=
1

_
d
4
x
_ _
|g

+R

()|
_
|g

|
_
+S
M
(g, ), (1)
where S
M
(g, ) is the matter action, generically denotes any matter eld, R

() is the Ricci tensor built from the

connection , |G

| = det G

, is a dimensionless parameter related with the cosmological constant by = (1)/,

is the extra EiBI parameter which has dimension of length squared, and we set 8G = 1 = c. The equation of
motions are obtained by varying the action (1) with respect to the elds g

and

respectively,

|q|
|g|
q

= g

, (2)
g

= g

+R

, (3)

Electronic address: hckim@ut.ac.kr

2
where q

is the auxiliary metric by which the connection

is dened, and q

is the matrix inverse of q

.
Because we are interested in stars in a at background, we set = 1. We also restrict our interests to the case with
positive , because the gravitational collapse does not lead to singularities [6]. We consider a spherically symmetric
star made of particles which are eectively described by a perfect uid with polytropic EoS
p = p
poly
K

, (4)
where K is a constant compensating dimensional dierence and is a positive dimensionless constant. The stress-
tensor of the uid is given by T

= ( +p)u
a
u
b
+pg
ab
. The metric and auxiliary metric for a spherically symmetric
star are
g

dx

dx

= e
(r)
dt
2
+e
(r)
dr
2
+R
2
e
(r)
d
2
, (5)
q

dx

dx

= e
(r)
dt
2
+e
(r)
dr
2
+r
2
d
2
, (6)
where (r), (r), (r), (r), and (r) are arbitrary metric functions of the radial coordinate r, d
2
= d
2
+sin
2
d
2
,
and R is the radius of the star. The problem was analyzed precisely in Refs. [7, 8, 17] recently. We follow their
denitions for most parts except for the metric function , where f(r) was used in place of R
2
e

in Ref. [8]. The

scalar curvature for the metric g

is given by
R
g
= e

+ 2

)
e

2
_

+
2
+ 2(

+ 3
2

+ 2R
2
e

, (7)
where the prime denotes the derivative with respect to r. Outside the star r > R, the metric should be the vacuum
Schwarzschild solution with
e

= e

= 1
2M
r
= e

= e

, R
2
e

= r
2
, (8)
where M is the mass of the star. The continuity equation gives
p

2
( +p) = 0. (9)
Let us describe how the curvature singularity happens at the surface of a star composed of an ideal uid described
by Eq. (4) in EiBI gravity. After solving the equation of motions (2), the derivatives of metric functions are given by
the sums of the auxiliary metric functions and sources as

a

b

b
,

+
a

a

3b

b
,

=
2
r

a

a

b

b
, (10)
where
a =
_
1 +, b =
_
1 p. (11)
After solving Eq. (3), the analytic functions (r) and (r) are determined by two rst-order dierential equations as
in Eqs. (16) and (17) in Ref. [8]. An EiBI generalization of Tolman-Oppenheimer-Volko equation in GR was also
developed in Refs. [7, 8, 17]. Because and are analytic, possible divergent contributions to the curvature scalar
in Eq. (7) come from the discontinuity of the derivatives of and p. Because
p

=
dp
d
= K
1
0 for > 1, we
notice that most singular contributions to Eq. (7) come from

, which are included in

and

. After calculating
the explicit form for

and

below, one can ensure that

2
, p
2
, p

and

do not contain divergent contribution.

The curvature scalar at the stellar surface, i.e. as r R and 0, is
R
g
e

+ 2

) = e

4
r
2

_
a

a
_

5
_
b

b
_

, (12)
where we keep only the potentially divergent parts in the rst and third equalities and ignore and p compared to
one. Assuming the continuity of (r), we may replace it to its surface value, e
(R)
= 1 2GM/R.
With the EoS (4), the continuity equation in Eq. (9) is integrated to give
=
_
1
K
exp
_
1
2
(
0
)
_

1
K
_
1/(1)
. (13)
3
For 1, p vanishes at the places where . The energy density and pressure exponentially decrease with .
For > 1, the surface of a star is dened by the place r = R where the pressure vanishes at (R) =
0
. Imposing
Eqs. (4) and (10) into Eq. (9), we obtain an analytic expression for

as

g(); g()
2(K +
1
)
4K +a
2

2
+KH()
, (14)
where H() denotes the homogeneous function
H() =
_
a
2
+ 3b
2
_
+ 3Kb
2

. (15)
Now the second derivative of the energy density becomes

g() +
2
g()
2
_
1

K + (2 )
1
K +
1

(2 )a
2

1
a
4

2
+KH

()
4K +a
2

2
+KH()
_
, (16)
where H

() denotes the derivative of H with respect to ,

H

() = a
2
+ 3b
2
+a
4
+ 3K
2
b
2

1
_
1

b
2
+
K

b
2
_
.
Near the surface of the star, we may use

= 2GM/R
2
(12GM/R)
1
and

= 4GM(RGM)/R
2
(R2GM)
2
,
which are the surface values of the corresponding outer solution. Taking the 0 limit, we have

_
0, 0 < 1
(2)
2
(2K)
2
_

32

4K

53
_
, 1 < < 2

4K+/2
+

2
(4K)(2K)
2(4K+/2)
3
, = 2

+
4
2

_
1 + (2 )H

(0)
4(2)K

2

3
_
, > 2
. (17)
Therefore, the curvature scalar diverges for 3/2 < < 2 and 2 < < 3. This implies that the surface singularity is
happening for a star composed of an ideal polytropic uid. A crucial part of this results was known in Ref. [11].
An important example of a polytropic uid belonging to this type is the non-relativistic degenerate Fermi gas with
= 5/3. This happens for dense Fermi particles when the Fermi energy exceeds by far the temperature. Examples
are the electron gas in metals and in white dwarf stars and the neutron star, whose density is so high that the neutron
gas is degenerate. We derive the EoS for the Fermi gas roughly and then discuss its validity. At low temperatures
and/or high densities we approach the uncertainty principle. In such a case, the Pauli exclusion principle will cause
the pressure to be higher than that inferred by the temperature. The complete degenerate case happens when two
electrons are occupied for each phase space volume. The average magnitude of the momentum of a Fermi particle is
roughly given by the Fermi momentum, P
F
= mv
F
=
_
3
2
n
_
1/3
, where n and m are the number density and mass
of a Fermi particle, respectively. The pressure, total momentum transfer on a unit surface area per unit time, is
p
F
P
F
nv
F
=

2
m
_
3
2
_
2/3
n
5/3
. (18)
From this equation, we take the EoS for the degenerate Fermi gas as p
F
= (/K)
5/3
because n .
Let us examine how robust the degenerated Fermi gas approximation at the surface of the star. The approximation
holds only when the temperature is smaller than the Fermi energy. Therefore, the degenerate Fermi gas approximation
does not hold for low number densities. This observation presents one reason to suspect the validity because the surface
singularity happens in the low density regime. However, if a star is very cold, the approximation will be valid even
for an extremely low energy density. Let us check how the geometry reacts on the motions of the particles composing
the uid. The EoS (18) is derived from the motions of Fermi particles in a at spacetime and its scope of usage is
extended to curved spacetimes without modication. The extension is justied by the principle of general covariance
in a comoving-locally at coordinates on the basis that the energy density and the pressure are local quantities. The
energy density must be a local quantity. However, the pressure may not be a local one because, in dening the
pressure microscopically, we need two distant objects, a colliding particle and a probing wall. Therefore, if some high
curvature eects probe (or modify) the microscopic structure, the covariant justication for the pressure may not
hold. We argue that this is the situation happening near the surface singularity. To examine the eect, let us study
4
the geodesic deviation equations. The geodesic deviation X
a
moving along the direction u
a
is given by
a
0
=
2

+
2

4
u
1
(X
0
u
1
X
1
u
0
)
R
2
e

u
3
(X
0
u
3
X
3
u
0
),
a
1
=
e

4
(2

+
2

)u
0
(X
0
u
1
X
1
u
0
)
R
2
e

4
_

u
3
(X
1
u
3
X
3
u
1
),
a
3
=
e

4
u
0
(X
0
u
3
X
3
u
0
) +
1
4
_

u
1
(X
1
u
3
X
3
u
1
), (19)
where we assume X
2
= 0 = u
2
without loss of generality because of the spherical symmetry. Keeping only the
dominant parts, we get
a
0

4
u
1
(X
0
u
1
X
1
u
0
), a
3
=

4
u
1
(X
1
u
3
X
3
u
1
),
a
1

4
u
0
(X
0
u
1
X
1
u
0
)

R
2
e

4
u
3
(u
3
X
1
u
1
X
3
), (20)
where we use

/2

> 0 for = 5/3. Consider two geodesics staying away radially so that X

=
(0, X
1
, 0, 0). Consider time evolution only with u

= (1, 0, 0, 0). Then, the geodesic deviation equation takes the form
a
1

e
(R)(R)
4
X
1
, a
0
= 0 = a
3
,
where we set the values of and to be those at the surface of the star. Similar expression can be obtained for a
3
if
we use u

= (1, 1, 0, 0) and X

= (0, 0, 0, X
3
). These are reminiscent of the Hooks law with frequency
f =
1
2
_

e
(R)(R)
4


0
4R
_
GM
R
__

3/2
K
_
, (21)
where we use Eqs. (8) and (17) for the case with 3/2 < < 2 and each term in the parenthesis is dimensionless. For
the cases with 2 < < 3, we discuss later. In the second equality in Eq. (21), we replace an order one dependent
term with
0
to denote that the calculation is approximate. We should mention that
0
is an order one dimensionless
constant, which depends on the species of particle composing the uid. Following the interpretation of the Hookes
law, any two nearby geodesics (e.g. a wall and a particle) along X
a
will cross each other f times irrespective of the
distance. Note however that this is not the oscillation of particles but that of geometry.
Now, let us calculate the pressure once more as we have done in Eq. (18). Due to the oscillations in geodesic
deviation equation, all Fermi particles inside an oscillation scale collide the wall f times. The pressure due to the
geodesic deviation motion can be estimated as
p
gd
fnP
F
, (22)
where is the characteristic scale of the oscillations. In fact, the oscillation is free from scales on the angular directions
over all the star surface. This looks like to imply that is nothing but the star radius, R. Let us try to set = R and
examine what happens. After setting P
F
= mv
F
= A
1/3
, where A is a constant, and n = p
F
/(v
F
P
F
) = mK/A
2
,
we get p
gd

7/6
. The eventual expression for the pressure will be given by the sum, p
F
+p
gd
. In low densities, p
gd
dominates the pressure over p
F
. Because p
gd
acts as if it is that of a polytropic uid with = 7/6 < 3/2, the surface
singularity disappears. However, when the curvature becomes too small for

term to be subdominant, there is no

reason for the pressure p
gd
to exist. Therefore, once again, the EoS will be given by that of the degenerate Fermi
gas (4) and the surface singularity happens again. To avoid this awkward situation, a delicate balance should be
taken between the diverging curvature eect and the modication of the EoS so that the curvature is not too small
and not too large. In the presence of such a balance, the curvature may not diverge. In addition, the characteristic
scale may not be the radius of the star but be smaller than that. It is not an easy task to exactly determine how
the characteristic size will be decreased through the balance between the gravity and the Fermi motions. However,
a good measure will be the ratio of the correlation scales between the degenerate Fermion and the gravity per unit
time, given by v
F
/c, where we restored c for comparison with v
F
. One may expect the characteristic length will be
decreased by the power of the form, = R(v
F
/c)
k
with k > 0. In this letter, we propose k = 1 for 3/2 < < 2.
Explicitly, after setting = Rv
F
/c, we get
p
gd
=
Rf
c
nv
F
P
F
=
Rf
c
p
F
= (c
2
)
3/2
; =

0
4
GM
Rc
2
_
8G
c
4
, (23)
5
where in this equation, we restore G and c with the prescription is
8G
c
4
,
M
R

GM
Rc
2
, c
2
, 1/R c/R.
Later in this work, we return to use the unit with 8G = 1 = c.
This discussion may also hold for all other matters composing the polytropic uid. We may argue that the same
phenomena may happen for other matters composing the polytropic matters because this is due to the geometric
eect. Therefore, for the particles previously described by a polytropic uid with > 3/2 in Eq. (4), we propose a
new EoS as
p = K

+
3/2
. (24)
Note that the correction term decreases with the radius of the star but increases with the mass of the star. In
addition, the correction term vanishes in the GR limit, 0. With the EoS in Eq. (24), for > 3/2, the second
term dominates the pressure when
<
c
=
_

K
_
2/(23)
.
In that regions, the EoS takes after that of the polytropic uid with = 3/2, in which case no singularity exists at the
star surface. Higher curvature induces higher pressure. As the pressure becomes higher, the EoS changes to decrease
the value of . When arrives at 3/2, the singularity disappears and therefore further modication of EoS will not
happen. Therefore, the form in Eq. (24) is intuitively correct once we accept the fact that the curvature modies the
EoS.
Let us calculate

again with the new EoS (24). The rst derivative

becomes

g
m
(); g
m
()
2(K +
1
+
3/2
)
4K +a
2

2
+KH() +
3/2
G()
, (25)
where
G() = 6 +
__
1
a
2
+
9
2b
2
_
+
3K(2 + 3)
2b
2

+
9
2b
2

3/2
_
. (26)
Note that, when = 0, Eq. (25) reproduces Eq. (14). The second derivative is given by

g
m
() +
2
g
m
()
2
_
1

K + (2 )
1
+ (5/2 )
3/2
K +
1
+
3/2

(3/2 )
1/2
G() +
3/2
G

() +
_
(2 )a
2

1
a
4

2
+KH

()

4K +a
2

2
+KH() +
3/2
G()
_
. (27)
For > 3/2,

2
/(18
2
). The curvature scalar at the surface becomes
R
g

16
2
9
2
0
(1 2M/R)
1
R
2
. (28)
For an O(1) number
0
, this value is acceptable. Interestingly, the curvature scalar does not contain even though
it originates from the modication of EoS by the EiBI gravity. Another interesting point is that it diverges is the
surface is located at the Schwarzschild radius. In GR, the surface value of the curvature is R
GR
M/R
3
. R
g
is
larger than R
GR
by the factor R/M. For the case of a neutron star, this ratio is not large.
For 2 < < 3, the situation is a little tricky. However, a similar treatment can be done by generating polytropic
uids with lower successively. For example, consider a uid with = 8/3. The geodesic deviation equation leads
to a frequency, f
1

1/6
. Setting = R, we get a correction term of the form p
gd1

5/3
. Note, however, that this
correction term still generates a diverging curvature. Therefore, the geodesic deviation equation provides additional
oscillations with the frequency given in Eq. (21) in addition to f
1
. This gives an additional correction term with the
form (23). Because the generation of p
gd1
is only intermediate, we may ignore it and simply use Eq.(24) as a new
EoS.
In summary, we studied the phenomena happening at the surface of a star in the EiBI theory, where the star is made
of particles which are eectively described by a polytropic uid. The presence of a surface singularity is reproduced
for a star composed of a specic kind of ideal-polytropic uid. Noting that the uid is composed of particles, we
demonstrated the possibility that a back-reaction of the spacetime curvature on the matter dynamics may modify the
eective EoS for particles. By investigating the geodesic deviation equation, we have obtained a modied EoS, which
6
appears natural and the unacceptable singularity of the theory is removed. This happens because the pressure may
not be a local quantity if the curvature is high enough to capture/modify the microscopic denition of pressure. The
non-local property is captured through the geodesic deviation equation which deals the relative motions of two distant
geodesics through the strong gravitational tidal force. The induced geodesic deviation equation resembles the Hookes
law, where its frequency is proportional to the square root of the curvature. The fast oscillation makes a particle
collide with a probing wall measuring the pressure more often, which modies of the EoS. With the modied EoS, the
surface is no longer singular and the size of the curvature at the surface is shown to be acceptable. Interestingly, the
curvature is independent of and is almost independent of the mass of the (large) star. The curvature is inversely
proportional to the area of the star surface and is larger than that in GR by the factor R/M. One of the most
important merit of this result is that the surface curvature is independent of the internal structure of the star but is
determined by the physics at the surface only.
Accepting that the EiBI gravity is sound theoretically, one of the most urgent task is to nd out observable eects
which discriminates the EiBI gravity from GR. An answer is given in Ref. [8]. There, it was shown that the EiBI
stars are more massive than their GR counterparts. They appeals that some stellar-mass blackhole candidates could
be in fact EiBI neutron or quark stars. However, this observational eect is quite dicult to be distinguished from
other eects in GR because it is related to the internal structures of the star. We propose a second observable
phenomena that can be found by the relatively large curvature at the star surface. For example, one may design
a scattering experiment of light which grazes the surface of a neutron star. The light will bears information of the
surface curvature, which can be detected from a distant place independently from other information. The value of
the curvature scalar driven by the gravitational back-reaction is independent of . Therefore, the fairly large surface
curvature will provide an opportunity to distinguish the EiBI theory from GR experimentally.
In Ref. [7], similar singularity was shown to happen during the phase transition inside a star. Because the origin
of the singularity is the same as that of the present one, this problem may also be cured with the same prescription.
Thus, the EiBI gravity can be saved from the pathology of surface singularity. Similar surface singularities were known
to exist in the Palatini f(R) gravity [21, 22]. The origin of the pathology is also the same. Therefore, we expect that
those singularities will be removed with the present prescript.
Acknowledgement
This work was supported by the National Research Foundation of Korea grants funded by the Korea government
NRF-2013R1A1A2006548.
[1] M. Banados and P. G. Ferreira, Phys. Rev. Lett. 105, 011101 (2010) [arXiv:1006.1769 [astro-ph.CO]].
[2] P. Pani, T. Delsate and V. Cardoso, Phys. Rev. D 85, 084020 (2012) [arXiv:1201.2814 [gr-qc]].
[3] T. Delsate and J. Steinho, arXiv:1201.4989 [gr-qc].
[4] I. Cho and H.-C. Kim, Phys. Rev. D 88, 064038 (2013) [arXiv:1302.3341 [gr-qc]].
[5] I. Cho, H.-C. Kim and T. Moon, Phys. Rev. D 86, 084018 (2012) [arXiv:1208.2146 [gr-qc]].
[6] P. Pani, V. Cardoso and T. Delsate, Phys. Rev. Lett. 107, 031101 (2011) [arXiv:1106.3569 [gr-qc]].
[7] Y. -H. Sham, L. -M. Lin and P. T. Leung, Phys. Rev. D 86, 064015 (2012) [arXiv:1208.1314 [gr-qc]].
[8] T. Harko, F. S. N. Lobo, M. K. Mak and S. V. Sushkov, Phys. Rev. D 88, 044032 (2013) [arXiv:1305.6770 [gr-qc]].
[9] J. Casanellas, P. Pani, I. Lopes and V. Cardoso, Astrophys. J. 745, 15 (2012) [arXiv:1109.0249 [astro-ph.SR]].
[10] A. De Felice, B. Gumjudpai and S. Jhingan, arXiv:1205.1168 [gr-qc].
[11] P. Pani and T. P. Sotiriou, Phys. Rev. Lett. 109, 251102 (2012) [arXiv:1209.2972 [gr-qc]].
[12] P. P. Avelino, Phys. Rev. D 85, 104053 (2012) [arXiv:1201.2544 [astro-ph.CO]]; P. P. Avelino, JCAP 1211, 022 (2012)
[arXiv:1207.4730 [astro-ph.CO]].
[13] C. Escamilla-Rivera, M. Banados and P. G. Ferreira, Phys. Rev. D 85, 087302 (2012) [arXiv:1204.1691 [gr-qc]].
[14] P. P. Avelino and R. Z. Ferreira, arXiv:1205.6676 [astro-ph.CO].
[15] Y.-X. Liu, K. Yang, H. Guo and Y. Zhong, Phys. Rev. D 85, 124053 (2012) [arXiv:1203.2349 [hep-th]].
[16] T. P. Sotiriou and S. Liberati, Annals Phys. 322, 935 (2007) [gr-qc/0604006].
[17] Y. -H. Sham, P. -T. Leung and L. -M. Lin, Phys. Rev. D 87, 061503 (2013) [arXiv:1304.0550 [gr-qc]].
[18] I. Cho, H.-C. Kim and T. Moon, Phys. Rev. Lett. 111, 071301 (2013) [arXiv:1305.2020 [gr-qc]].
[19] K. Yang, X. -L. Du and Y. -X. Liu, arXiv:1307.2969 [gr-qc].
[20] M. Lagos, Mx. Baados, P. G. Ferreira and S. nGarca-Senz, arXiv:1311.3828 [gr-qc].
[21] T. P. Sotiriou and V. Faraoni, Rev. Mod. Phys. 82, 451 (2010) [arXiv:0805.1726 [gr-qc]].
[22] E. Barausse, T. P. Sotiriou and J. C. Miller, EAS Publ. Ser. 30, 189 (2008) [arXiv:0801.4852 [gr-qc]].