Energy-Momentum Squared Gravity

Mahmood Roshan1∗ and Fatimah Shojai2,3†

1
Department of Physics, Ferdowsi University of Mashhad, P.O. Box 1436, Mashhad, Iran
2
Department of Physics, University of Tehran, Tehran, Iran
3
Foundations of Physics Group, School of Physics,
Institute for Research in Fundamental Sciences (IPM), Tehran, Iran.

A new covariant generalization of Einstein’s general relativity is developed which allows the ex-
istence of a term proportional to Tαβ T αβ in the action functional of the theory (Tαβ is the energy-
momentum tensor). Consequently the relevant field equations are different from general relativity
only in the presence of matter sources. In the case of a charged black hole, we find exact solutions
for the field equations. Applying this theory to a homogeneous and isotropic space-time, we find
that there is a maximum energy density ρmax , and correspondingly a minimum length amin , at early
arXiv:1607.06049v2 [gr-qc] 3 Aug 2016

universe. This means that there is a bounce at early times and this theory avoids the existence of an
early time singularity. Moreover we show that this theory possesses a true sequence of cosmological
eras. Also, we argue that although in the context of the standard cosmological model the cosmo-
logical constant Λ does not play any important role in the early times and becomes important only
after the matter dominated era, in this theory the ”repulsive” nature of the cosmological constant
plays a crucial role at early times for resolving the singularity.

PACS numbers: 04.50.Kd, 98.80.-k

I. INTRODUCTION tions, we refer the reader to the review article [8]. Here
we propose a new model which , despite of its simple
appearance, possesses interesting features. Let us start
Modifying a gravitational theory dates back to late with the following action
1800s, there were some attempts modeled on Maxwell’s
electrodynamics to modify Newtonian gravity. Since Ein- 1 √
Z
−g R − 2Λ − ηT2 d4 x + SM


stein developed his general relativity (GR) in 1915, vari- S= (1)
2κ
ous attempts with different motivations have been carried
out to generalize it [1]. Some motivations have theoret- where T2 = Tαβ T αβ , Tαβ is the energy-momentum ten-
ical character and some observational. Einstein himself sor, R is the Ricci scalar, κ = 8πG, Λ is the cosmolog-
modified the original field equations by adding a term in- ical constant and SM is the matter action. Also η is a
cluding the cosmological constant. Also he proposed the coupling constant which its value can be constrained by
Palatini formulation of GR [2]. Eddington proposed an observations. For a somehow similar approach we refer
interesting alternative to GR in 1924 [3]. Brans-Dicke the reader to [9]. In general η can be negative or positive.
theory [4] and the Einstein-Cartan theory [5] are two However as we will show in this paper, a positive η leads
other examples of a very broad variety of alternatives. to a bounce at early universe and to a satisfactory cos-
Currently, observations of the dark matter and the dark mological behavior after the bounce. This bounce avoids
energy provide one of the main motivations for extend- the early time singularity. On the other hand, as we will
ing GR (for a review on modified gravity theories see e.g. see in section III, a negative η leads to unsatisfactory cos-
[6]). mological behavior. More specifically there is no stable
One of the main intriguing enigmas in GR is that it late time accelerated phase in the case of η < 0. There-
predicts the existence of space-time singularity at some fore our main purpose in this paper is to study the η > 0
finite time in the past. However it turns out that GR case.
itself is no longer valid at the singularity because of the The situation here is somehow reminiscent of the ap-
expected quantum effects. On the other hand a precise pearance of the cosmological constant in the standard
formulation for quantum gravity is still lacking. Never- cosmological model, where Λ is postulated to be positive.
theless, there are some classical models in which this kind A negative cosmological constant leads to completely dif-
of singularity can be resolved. For example Eddington- ferent consequences which are inconsistent with the cos-
inspired Born-Infeld (EiBI) theory, is a modified theory mological observations. More specifically positive Λ ac-
of gravity which is equivalent to GR only in vacuum and celerates the universe while a negative Λ decelerates it.
can resolve the singularity [7]. For other examples and The standard Einstein-Hilbert action can be recovered
also for other motivations behind this kind of modifica- by setting η = 0. Because of the correction term T2 ,
we refer to this theory as Energy-Momentum Squared
Gravity (EMSG). Throughout the paper, we use units
with c = 1 and assume the metric signature (−, +, +, +)
∗ [email protected] for the metric. It is natural to expect that this correction
† [email protected] term be important only in the high energy regimes such
 2

as the early universe or within the black holes. Therefore Finally, bearing in mind that variation of the other terms
there are no departures from GR in the low curvature lead to the convenient terms in the Einstein field equa-
regime. tions, the field equations of EMSG can be written as
The outline of the paper is as follows. In section II we
eff
derive the field equations of EMSG by varying the action Gµν + Λgµν = κTµν (7)
(1) with respect to the metric. In section III we derive the
modified Freidmann equations and show that there is a where Gµν is the Einstein tensor and the effective energy-
eff
maximum energy density and a minimum length at early momentum tensor Tµν is given by
universe (when η > 0). Also using the dynamical sys-  
eff η 1
tem method we study the cosmological consequences of Tµν = Tµν + 2 Ψµν + Tµσ Tνσ − gµν T2 (8)
EMSG. More specifically we show that this theory pos- κ 4
sesses a true sequence of cosmological epochs. In sec-
tion IV, we find an exact charged black hole solution in From (7), it is clear that ∇µ Tµν
eff
= 0 and consequently
µ
EMSG. Finally, conclusions are drawn in section V. ∇ Tµν 6= 0. We also recall that the matter action SM
can be written as follows
√
Z
II. FIELD EQUATIONS SM = Lm −gd4 x (9)

Before moving on to derive the field equations it is im- where Lm is the matter Lagrangian density. The energy-
portant to mention that the correction term ηT2 can be momentum tensor then is defined as
defined only when the Lagrangian density for the matter √
2 δ( −gLm )
content is specified. Therefore one may not immediately Tµν = − √ (10)
obtain the field equations from a first-order variation of −g δg µν
the action. In other words, one must first vary the mat- assuming that Lm depends only on the metric itself and
ter action with respect to the gravitational degrees of not on its derivatives, we obtain
freedom. Although this feature is not the case in GR,
it appears in theories which introduce correction terms δLm
including the energy-momentum tensor in the generic ac- Tµν = gµν Lm − 2 (11)
δg µν
tion, for example see [10].
Comparing the action (1) with the Einstein-Hilbert ac- see [10] for more details. As it is clear from field equations
tion, it is obvious that we need only vary the T2 term (7), we need the variation of the energy-momentum ten-
with respect to metric. In other words, the other terms sor with respect to the metric. Therefore using equation
lead to the standard terms in the Einstein field equations. (11) we can write
Thus we have
√ √ √ δTαβ δgαβ δLm ∂ 2 Lm
δ( −gTαβ T αβ ) = −gδ(Tαβ T αβ ) + Tαβ T αβ δ −g (2) = L m + g αβ − 2
δg µν δg µν δg µν ∂g µν g αβ
√ √ 1 1
we know that δ −g = − 21 gµν −gδg µν . Therefore the = −gαµ gβν Lm + gαβ gµν Lm − gαβ Tµν (12)
last term in the right hand side of (2) can be simply 2 2
written as ∂ 2 Lm
− 2 µν αβ
1 √ ∂g g
− gµν T2 −gδg µν (3)
2
where we have also used the following expression
Now let us consider the first term in the right hand side
of (2). We can write δgαβ θρ
= −gαθ gβρ δµν (13)
δg µν
δT2 = δ(g αρ g βσ Tαβ Tρσ )
θρ
= 2δg αρ Tασ Tρσ + 2T αβ δTαβ where δµν is the generalized Kronecker delta symbol.
(4) This relation can be simply obtained using the condi-
 δTαβ 
= 2 Tµσ Tνσ + T αβ µν δg µν tion gαθ g θβ = δαβ . Finally multiplying equation (12)
δg through by T αβ , we find Ψµν with respect to the matter
consequently we obtain Lagrangian density
√  1 √
1 ∂ 2 Lm
δ( −gTαβ T αβ ) = 2 Tµσ Tνσ − gµν T2 + Ψµν −gδg µν Ψµν = −Lm Sµν − T Tµν − 2T αβ αβ µν (14)
4 2 ∂g ∂g
(5)
where Sµν = Tµν − T gµν /2, and T is the trace of the
where the new tensor Ψµν is defined as
energy-momentum tensor. Therefore for a given matter
δTαβ Lagrangian density the field equations (7) are completely
Ψµν = T αβ (6) known. In the case of a perfect fluid Lm can be simply
δg µν
 3

defined as Lm = p [10, 11]. It is important to mention us call it ”bare” cosmological constant as it is called in
that the non-relativistic limit of this theory is the same as [12]. On the other hand it can be a part of the matter
the Newtonian limit of GR. Therefore the Poisson equa- action. In this case one may assign an effective perfect
tion does not change. Albeit, η needs to be small enough fluid energy-momentum tensor to Λ with energy density
Λ
to pass the the classical tests of gravity. However, in the ρΛ = Λ/κ and pressure pΛ = −Λ/κ, i.e. Tµν = pΛ gµν .
presence of matter sources, the higher post-Newtonian In this case Λ can be called as ”vacuum energy”. Albeit
corrections will be different from that of GR. the field equations are the same in both pictures.
Before moving on to discuss the cosmology of EMSG, Then the questions is: Does EMSG lead to different
let us first discuss some points about the appearance of Λ field equations when we put Λ in the geometric part or in
in the action of EMSG. As we will see in the next section, the matter action? Here we show that EMSG, unlike GR,
when η > 0 the correction terms in EMSG are important discriminates between these approaches. This is also the
only in the sufficiently early times, and do not disturb the case in other theories which include scalars constructed
late time cosmology. Therefore it is evident that without from the energy-momentum tensor, such as Rµν T µν and
any contribution from other extra fields, such as scalar gµν T µν , in their generic action, for example see [10, 13].
fields which can enter the matter Lagrangian, EMSG can- Now let us put the cosmological constant in the matter
not be considered as a dark energy model. Consequently action. Therefore the total energy-momentum tensor is
one has to retain the cosmological constant in the the- given by
ory in order to explain the accelerated expansion of the
total Λ
universe. Of course, as in GR, one may remove the cos- Tµν = Tµν + Tµν (15)
mological constant and add new energy contributions to
Tαβ in order to construct a dynamical dark energy model. and after some algebra, the field equations can be written
We mean that one can, for instance, add a quintessence as
scalar field to the theory instead of keeping the cosmo- eff
logical constant. Gµν = κTµν − Λgµν + Hµν (16)
However, comparing EMSG with GR a question nat-
eff
urally raises. We know that in GR, Λ can be a part of where Tµν is still given by (8) and depends only on Tµν .
Λ
the geometric sector. In this case Λ is written in the All the contributions from Tµν have been collected in the
left hand side of the Einstein equations. In this case, let new tensor Hµν as

Λ

Λ αβ δTαβ
δTαβ 1 
Hµν = 2η ΨΛ
µν + T + T αβ
+ T Λ
µσ T σ
ν + T σ Λ
µ T νσ + T Λ
µσ T Λ σθ
νθ g − g µν (T Λ
αβ T Λ αβ
+ 2T Λ
αβ T αβ
) (17)
δg µν δg µν 4

In following we show that Hµν 6= 0. This means that It is obvious that in this case field equations are more
by putting the cosmological constant in the matter sec- complicated than the case of a bare cosmological con-
tor, we get different field equations than (7). Thus the stant. More specifically, as expected, Λ2 term appears in
cosmological behavior, in principle, would be different. the field equations. It is worth mentioning that, in GR,
Λ
Using Tµν = pΛ gµν and equations (12)-(14), we find an effective cosmological constant can be defined as the
linear sum of a bare cosmological constant and the vac-
ΨΛ 2
µν = −pΛ gµν
uum energy contribution [12]. However, as we showed, it
is not that simple in EMSG and the gravitational effects
δTαβ
T Λ αβ µν = −2pΛ Tµν + pΛ Lm gµν of a bare cosmological constant and the vacuum energy
δg cannot be summed simply. On the other hand, in GR,
Λ
δTαβ interpreting Λ as the vacuum energy raises the so-called
T αβ µν = −pΛ Tµν cosmological constant problem. Consequently, for the
δg (18)
sake of simplicity, we adopt the geometric description,
Tµσ Tνσ + Tµσ Tνσ
Λ Λ
= 2pΛ Tµν
the bare cosmological constant, in this paper. A more
Λ Λ σθ
Tµσ Tνθ g = p2Λ gµν general case including the vacuum energy can be consid-
1 Λ 1 ered as a matter of study for future works. Therefore we
gµν (Tαβ T Λ αβ + 2Tαβ
Λ
T αβ ) = p2Λ gµν + T pΛ gµν work with the field equations (7) in what follows.
4 2
Therefore Hµν takes the following form
 1  
Hµν = 2η pΛ Lm − pΛ − T gµν − Tµν (19)
2
 4

III. COSMOLOGY OF EMSG without any scalar field. Note that the effective energy
density and pressure remain also finite and all effective
Let us start with the consequences of this theory in energy conditions fail at the bounce. The effective energy
the early universe where we expect significant deviations density and pressure, ρeff and peff , are defined using the
eff
from the ΛCDM model. We assume a flat Friedmann- effective energy-momentum tensor Tµν . At the bounce
Robertson-Walker (FRW) geometry we have ρeff = −Λ/κ < 0 and peff = ρeff /3. It is clear
that the null energy condition is violated. It is well known
ds2 = −dt2 + a(t)2 (dx2 + dy 2 + dz 2 ) (20) that in bouncing cosmologies one or more energy condi-
tions are violated [8]. For generalized energy conditions
where a(t) is the cosmic scale factor. Also we assume an in modified theories of gravity we refer the reader to [16].
ideal energy-momentum tensor Tµν = (ρ + p)uµ uν + pgµν It is clear that in this theory, existence of a positive
for the cosmic fluid. Using field equation (7), we find the cosmological constant is necessary for preventing the sin-
modified version of the Friedmann equations gularity. In fact the repulsive nature of the cosmologi-
  cal constant plays an important role here. On the other
2 κ Λ 1 2 4 1 2 hand, in the ΛCDM model, Λ does not have a significant
H = ρ+ −η p + ρp + ρ (21)
3 3 2 3 6 effect in the early universe and its role is dominant at
sufficiently the late times.
ä κ Λ

2 1
 A maximum in energy density naturally implies a min-
= − (ρ + 3p) + + η p2 + ρp + ρ2 (22) imum value for the cosmic scale factor. Here we estimate
a 6 3 3 3 this minimum length amin . As we mentioned before, in
where H = ȧ/a and the dot denotes a derivative with EMSG the perfect fluid does not satisfy the standard con-
respect to time. Equations (21) and (22) together with servation laws. For a two component cosmic fluid, the
an equation of state form a complete set to study the conservation equation ∇µ Tµν eff
= 0 includes some interac-
dynamics of the cosmic fluid and the scale factor. It is tion terms proportional to ρ˙m ρr , ρ˙r ρm and ρm ρr , where
worth mentioning that the correction terms in the equa- ρm is the matter energy density. One may equally dis-
tions (21) and (22) are somewhat reminiscent of those tribute these interaction terms between two energy com-
from quantum geometry effects in loop quantum gravity, ponents. In this case the conservation equation can be
for example see [14], or those from the braneworlds [15]. split up into two separate equations
At small energy densities, we recover the standard Fried-
9ρ̄2m + 48ρ̄r ρ̄m + 56ρ̄2r − 66ρ̄m − 152ρ̄r + 96


mann equations. However at high densities a new effect ρ̄′r
=
appears: for η > 0, there is a critical point H = 0 at ρ̄r 26ρ̄m + 62ρ̄r − 7ρ̄2m − 24ρ̄r ρ̄m − 28ρ̄2r − 24
r ! (25)
κ 2ηΛγ
ρmax = 1+ 1+ (23) ρ̄′m 21ρ̄2m + 72ρ̄r ρ̄m + 64ρ̄2r − 78ρ̄m − 148ρ̄r + 72


γη κ2 =
ρ̄m 26ρ̄m + 62ρ̄r − 7ρ̄2m − 24ρ̄r ρ̄m − 28ρ̄2r − 24
Where γ = 3w2 + 8w + 1 and we have assumed a (26)
barotropic equation of state p = wρ (w > 0). In the very where ρ̄i = 2ηρ κ
i
and prime denotes derivative with re-
early universe, ρmax is an explicit cutoff in the energy spect to ln a. For small ρ̄i we recover the conventional
density. In fact in this era w =p 1/3 and the maximum conservation equations. These conservation equations
energy density is ρmax = (1 + 1 + 8Λη/κ2 )κ/4η. In are complicated and so it is not easy to find the exact
other words, this means that the early radiation domi- value of amin . In order to obtain an estimate for amin ,
nated universe does not start from a singularity. More we neglect ρ̄m in the conservation equation of the radi-
surprisingly, the universe passes across a regular bounce ation (25). This assumption makes sense because ρ̄m is
at this point. One may easily verify that at this point very smaller than ρ̄r in the beginning and it decreases
ä = 2Λ even more after the bounce. Therefore, although eventu-
3 a > 0. Fortunately, in the radiation dominated
universe the field equations (21) and (22) can be exactly ally ρ̄r will fall bellow ρ̄m , during a long period of time
solved. The result is compared to the age of the universe this approximation
√ holds. In this case, equation (25) can be expressed as
a(t) = amin cosh αt
r ! ∞
(24)
X
κ 8ηΛ 2 ρ̄′r + 4ρ̄r ≃ − 2n ρ̄nr (27)
ρr (t) = 1 + 1 + 2 sech αt
4η κ n=2

q
4Λ
by retaining only the dominant term on the right hand
where α = 3 .
We recall that the corresponding so- side, we find
√
lution in ΛCDM model is a(t) ∼ sinh αt. The main
difference is that unlike the ΛCDM case, ä is positive a4min ρmax
ρ̄r = (28)
in EMSG. This is reminiscent of an inflationary epoch (κ/2η + ρmax )a4 − a4min ρmax
 5

0.5 P3 must start with a radiation era. Also it has to possess

a mater dominated era followed by an accelerated ex-
0.4
pansion. In principle, the correction terms in the new
0.3
Friedmann equations can destroy the standard thermal
history. In order to check this important requirement
W mp

0.2 and to determine whether EMSG provides a viable alter-

native to the standard model, we use dynamical system
0.1 analysis [18]. The advantage of the finite dynamical sys-
P4 P2 P1 tems analysis is to provide a fast and numerically reliable
0.0 integration of the modified Friedmann equations. Using
0.0 0.2 0.4 0.6 0.8 1.0 this method, one may link the early and late time evo-
W rp
lution of the model by analyzing the fixed points in the
compact phase-space of the model, for example see [17].
FIG. 1: 2-dimensional phase space trajectories in the Poincaré
projected phase space. The fixed point P1 is an unstable
The expansion history can then be easily compared to the
very early time radiation dominated fixed point appearing in standard ΛCDM model. This method has been applied
the ”infinity” of the system. P2 corresponds to a standard to several alternative theories and cosmological models,
radiation era. The fixed point P3 is a true matter era, and P4 for example see [19].
is the final late time de Sitter attractor. We use cosmic density parameters Ωm = κρ 3H 2 , Ωr and
m

Λ
ΩΛ = 3H 2 as phase space variables. Using modified
Friedmann equations (21) and (22) and also the corre-
finally one may simply show that sponding conservation equations for matter and radiation
 1/4 (25) and (26), one can find three autonomous equations.
12η
1/4
These equations take the following form
amin ≃ H02 Ωr0 a0 (29)
κ2
κρr Ω′m = f1 (Ωm , Ωr , ΩΛ )
where Ωr = 3H 2 is the radiation density parameter and

subscript 0 denotes the present value of the quantities. Ω′r = f2 (Ωm , Ωr , ΩΛ ) (30)
This is the minimum length in the cosmology of EMSG. Ω′Λ = f3 (Ωm , Ωr , ΩΛ )
Although this theory avoids the singularity, it should
possess a valid expansion history. In other words, it where functions fi are given by

−21Ωm3 + (−72Ωr + 78Ωmax ) Ωm 2 + −64Ωr 2 + 148ΩrΩmax − 72Ωmax 2 Ωm

f1 (Ωm , Ωr , ΩΛ ) =
7Ωm 2 + (24Ωr − 26Ωmax) Ωm + 28Ωr 2 − 62 (Ωr ) Ωmax + 24Ωmax2
1 (Ωm + 2Ωr − 2Ωmax ) Ωm (3Ωm + 4Ωr )
−
2 Ωmax
Ωr 56Ωr 2 + 48Ωr Ωm − 152Ωr Ωmax + 9Ωm 2 − 66Ωm (Ωmax ) + 96Ωmax 2


f2 (Ωm , Ωr , ΩΛ ) = − (31)
7Ωm 2 + (24Ωr − 26Ωmax) (Ωm ) + 28Ωr 2 − 62 (Ωr ) Ωmax + 24Ωmax 2
1 Ωr (3Ωm + 4Ωr ) (Ωm + 2Ωr − 2Ωmax )
−
2 Ωmax
3 ΩΛ (Ωm + 2Ωr − 2Ωmax ) (Ωm + 4/3Ωr )
f3 (Ωm , Ωr , ΩΛ ) = −
2 Ωmax

where tions fi and find the relevant roots. Then we can deter-
2 2 mine whether the system approaches one of the critical
κρmax 1 3Ωm + 14Ωr Ωm + 12Ωr
Ωmax = = (32) points or not by analyzing the stability around the criti-
3H 2 12 −1 + ΩΛ (t) + Ωm + Ωr cal points. Let us consider small perturbations δΩm , δΩr
It is interesting that although the field equations and con- and δΩΛ around the critical point (Ωcm , Ωcr , ΩcΛ ), namely
servation equations seem complicated, as we will show,
the cosmological behavior of the model is close to the
Ωm = Ωcm + δΩm
ΛCDM model.
In order to find the critical/fixed points of the dynam- Ωr = Ωcr + δΩr (33)
ical system (30), it is just enough to set to zero func- ΩΛ = ΩcΛ + δΩΛ
 6

1.0
a compact phase space. In order to demonstrate the
Wrp
main features of the system in a compact region (a
0.8 sphere with radius 1) including the infinity, we use the
Wmp Poincaré coordinates,
p obtained by the transformation
0.6 Ωip = Ωi /(1 + Ω2m + Ω2r + Ω2Λ ). Using these new vari-
WLp
Wip

ables it becomes clear that there is another unstable

0.4 radiation dominated fixed point at infinity, i.e. P1 :
(Ωmp , Ωrp , ΩΛp ) = (0, 1, 0). In Fig. 1 we show a 2-
0.2 dimensional phase space plot. The above mentioned criti-
cal points have been shown in this phase plot. Therefore
0.0 we see that the evolution can be started from P1 and
-12 -10 -8 -6 -4 -2 0 rapidly repels away from it and is followed by a standard
lna radiation dominated epoch P2 . Then the models enters
FIG. 2: Behavior of the cosmic density parameters for a spe- the matter era P3 and finally falls into the late time de
cific choice of initial conditions at ln a = −12.5 as Ωm = 0.01, Sitter attractor P4 . In fact after the rapid evolution from
Ωr = 1 and ΩΛ ≃ 0. Thick curves show the projected density P1 to P2 , the system follows standard trajectories sim-
parameters and the dashed curves correspond to the density ilar to ΛCDM model. In Fig. 2 we have plotted the
parameters themselves. evolution of the cosmic density parameters for a specific
choice of initial conditions. As it is clear from this figure,
after a rapid decrease in Ωrp system evolves as standard
Substituting into equations (31) yields to the linear dif-
cosmological model and the current values of the density
ferential equations
parameters can match the observed values.

δΩ′m
 
δΩm
 In order to emphasize that EMSG with a bare cos-
 δΩ′r  = M  δΩr  (34) mological constant does not disturb the standard cosmic
δΩ′Λ δΩΛ evolution after the bounce, we have plotted the behavior
of the deceleration parameter q and the effective equa-
The stability matrix M depends upon Ωcm , Ωcr and ΩcΛ , tion of state parameter ωeff in Fig. 3. These parameters
and is given by belong to the numerical solution presented in Fig 2. We
 ∂f ∂f ∂f  recall that
1 1 1
∂Ωm ∂Ωr ∂ΩΛ
∂f2 ∂f2 ∂f2 äa 2Ḣ
M= (35)
 
∂Ωm ∂Ωr ∂ΩΛ q=− , ωeff = −1 − (36)
ȧ2 3H 2

∂f3 ∂f3 ∂f3
∂Ωm ∂Ωr ∂ΩΛ (Ωcm ,Ωcr ,ΩcΛ )
These parameters can be written with respect to the
for each fixed point, this matrix possesses three eigenval- phase space variables. Since the relevant expressions are
ues. The fixed point is stable/attractor if all the eigen- too long, we have not written them. In Fig. 3, the solid
values are negative and is unstable if there is at leas one lines correspond to EMSG and the dashed curves corre-
positive eigenvalue. spond to ΛCDM model. The purple regions demonstrate
We have found the critical points and developed the approximate time intervals for which the system is
the stability matrix. Here are the results: as in the close to its critical points. It is worth mentioning that
ΛCDM model there are two repulsive critical points the autonomous equations in ΛCDM model can be ob-
P2 : (Ωcm , Ωcr , ΩcΛ ) = (0, 1, 0), P3 = (1, 0, 0). The eigen- tained from (31) by considering the limit Ωmax → ∞.
values for theses points are (1, 4, −4) and (−3, 3, −1) re- In Fig 3, the initial conditions in ΛCDM model are cho-
spectively. As it is clear, P2 corresponds to a standard sen in a way to lead the same values as in EMSG for
unstable radiation dominated era, and P3 is a true unsta- the cosmic density parameters, i.e. Ωi , at present epoch
ble matter dominated point. In fact the effective equation ln a = 0. As it is obvious, q and ωeff are exactly the same
of state parameter ωeff , see (36) for definition, for P2 is in both theories. This means that the growth rate of the
equal to 1/3 (this is also evident from Fig 3). Therefore cosmic scale factor in EMSG coincides with the standard
the cosmic scale factor grows as a(t) ∝ t1/3 . On the other case. Consequently after the bounce the thermal history
hand for P3 we have ωeff = 0 and therefore a(t) ∝ t2/3 . of the universe is the same. It should be noted that this
Also there is a late time attractor P4 = (0, 0, 1). In this behavior does not belong only to the above mentioned
case the eigenvalues are (0, −6/5, −4). This point corre- initial condition, and we have checked it for several ini-
sponds to a late time de Sitter expansion. Let us recall tial conditions.
again that Λ has a twofold task in this model, providing a Before closing this section we remind that for the whole
regular bounce and triggering the late time cosmic speed description presented in this section, we assumed η >
up. 0. However as we mentioned before, there is no a-priori
It is also necessary to check the behavior of this sys- reason why this parameter should be positive. However
tem at ”infinity”. Note that at the bounce the phase if η < 0 then using (21) one can conclude that there will
variables become infinite and thus they do not make not be a bounce in this theory. More importantly using
 7

1.0 IV. CHARGED BLACK HOLES IN EMSG

P2 P3 P4

0.5 As we mentioned before, we expect that in high den-

sity regions, for example within a black hole, EMSG to be
q & Ωeff

0.0 different form GR. On the other hand, it is obvious that

in the vacuum, where the matter energy density is zero,
EMSG is equivalent to GR. Consequently, Schwarzschild-
-0.5
de Sitter metric and Kerr metric are also solutions for
EMSG field equations. In order to show some interest-
-1.0 ing differences of these two theories, let us examine a
-12 -10 -8 -6 -4 -2 0 2 charged black hole. Outside the black hole there exists
lna matter field (Tµν 6= 0), i.e. the electromagnetic field with
Lagrangian density Lm = − 41 Fµν F µν = − 14 F2 , where
FIG. 3: Behavior of the deceleration parameter q (blue curves)
and the effective equation of state parameter ωeff (red curves) F µν = ∂µ Aν − ∂ν Aµ is the electromagnetic field strength
in ΛCDM model (dashed curves) and EMSG (solid curves). tensor. It is important to mention that EMSG not only
For EMSG The initial conditions is the same as in Fig. 2. changes the gravitational theory but somehow postulates
In the case of ΛCDM, the initial conditions have been set a universal modification to all matter field’s equations of
in such a way to recover the same values for cosmic density motion. In the case of the electromagnetic filed, EMSG
parameters as in EMSG at ln a = 0. adds some non-linear terms to Maxwell equations. In
this sense, EMSG is reminiscent of the Born-Infeld non-
linear electrodynamics [20]. Albeit the field equations
the dynamical system approach we found that there is are different. More specifically, as we will show, unlike
no stable late time accelerated expansion in this case. It the Born-Infeld theory EMSG does not lead to a non-
should be notes for η < 0 the autonomous differential singular electric field for a point charge. Using the above
equations can be simply obtained from (30) by replacing mentioned Lagrangian for the electrodynamics we have
Ωmax with Ωη . Where Ωη > 0 is
1
Tµν = F αµ Fαν − gµν F2 (39)
4
κ2 1 3Ωm 2 + 14Ωr Ωm + 12Ωr 2
Ωη = − = − (37)
6ηH 2 12 −1 + ΩΛ (t) + Ωm + Ωr one can simply show that T = 0. Therefore

then finding the fixed points and their stability is com- ∂ 2 Lm

Ψµν = −Lm Tµν − 2T αβ (40)
pletely similar to the η > 0 case. There are three fixed ∂g αβ g µν
points (Ωcm , Ωcr , ΩcΛ ) as
On the other hand
P1′ = (0, 1, 0) ∂ 2 Lm 1
= − Fαµ Fβν (41)
P2′ = (1, 0, 0) (38) ∂g αβ g µν 2
P3′ = (0, Ωr , 1 − 2Ωr )
Consequently Ψµν takes the following form
P1′ corresponds to an unstable radiation dominated 1 2
phase, and the corresponding eigenvalues are (1, 12, 4). Ψµν = F Tµν + T αβ Fαµ Fβν
4 (42)
P2′ is an unstable matter dominated phase with two pos- 1
itive eigenvalues (3, 9, −1). As expected, these points are = F γα Fγ β Fαµ Fβν − gµν (F2 )2
16
unstable.
On the other hand, P3′ is not a fixed point but a fixed Also after some algebra we find
”line”. This line exists provided that Ωr < 1/2. The 1
corresponding eigenvalues are (0, 12, −6/5). Surprisingly, Tµσ Tνσ = F γα Fγ β Fαµ Fβν + gµν (F2 )2
all the points on this curve are unstable even the point 16 (43)
1 1 1
(0, 0, 1). This means that the late time de Sitter uni- − gµν T2 = − gµν F αθ Fαρ F γθ Fγ ρ + gµν (F2 )2
verse is not stable when η < 0. It is interesting that the 4 4 16
correction terms in EMSG, which are expected to be im- Therefore the field equations (7) can be rewritten as
portant in early times, substantially disturb the standard
paradigm specially in the late times. As a final remark on  1 
the cosmology of EMSG, we should stress that these con- Gµν +Λgµν = κ F αµ Fαν − gµν F2
4
clusions are true only for a bare cosmological constant. 1
In the presence of the other fields and the vacuum en- + 2η gµν (F ) + 2F γα Fγ β Fαµ Fβν
2 2 (44)
16
ergy density, in principle, the cosmological consequences 1 
would be different. − gµν F αθ Fαρ F γθ Fγ ρ
4
 8

Although these field equations seem very complicated, the equations (50) are automatically satisfied. Also from
they can be solved analytically for a charged black hole. the field equations, it turns out straightforwardly that
Now we need the generalized form of the Maxwell equa- φ(r) = 1. On the other hand equation (49) has only one
tions. we recall again that EMSG not only changes the nonzero component as
gravitational filed equations but also modifies the matter
field equations in the high curvature regime. To find the
 6η  dE(r) 2E(r) 4η
1 + E(r)2 + + E(r)3 = 0 (52)
modified electrodynamic field equations in the vacuum, κ dr r κr
we need to vary the following part of the action (1) with
Fortunately this equation can be integrated and the so-
respect to Aµ
lution is
√ 1 η 2 4
Z  √ 
−g( F2 +

2/3
T )d x (45) r
κ  x−1/3
1 + 1 + x 4 − x
4 2κ E(x) = (53)
6η √
1/3 
x1/3 1 + 1 + x4
Variation leads to the following Euler-Lagrange equation

∂(κF2 + 2ηT2 ) where the new parameter x is defined as r =

∇µ =0 (46) (27q 2 η/2κ)1/4 x, and q is the integration constant. As
∂(∇µ Aν )
we will see q is related to the total electric charge of the
∂(κF2 +2ηT2 ) black hole.
Note that ∂Aν = 0. Also we remind that On the other hand, the field equations (44) lead to two
independent differential equations. More specifically, tt
∂Fαβ
= δαµ δβν − δβµ δαν (47) and rr components lead to a same equation and θθ and
∂(∇µ Aν ) φφ components lead to another independent equation. tt
and θθ components are respectively
Using this equation it is straightforward to obtain
df (r) 1  3η 
∂F2 r − 1 + f (r) + κr2 E(r)2 1 + E(r)2 = 0 (54)
= 4F µν dr 2 κ
∂(∇µ Aν )
(48)
∂T2
= 8Fγ ν F µρ F γρ − 2F µν F2 df (r) r2 d2 f (r) 1 2 2
 η 2

∂(∇µ Aν ) r + − κr E(r) 1 + E(r) = 0 (55)
dr 2 dr2 2 κ
Finally the modified version of the Maxwell equations in At the first glance it seems that we have two differential
the vacuum are equations for one unknown function f (r). However, one
η may easily show that equation (55) is not also indepen-
∇µ F µν = ∇µ 4Fγ ν F µρ F γρ − F µν F2
 
(49)
κ dent equation. In fact differentiating (54) with respect to
r and combining with (52), we get equation (55). Hence
we need to solve only equation (54). The solution can be
∇[µ Fνλ] = 0 (50)
expressed as
Equations (50) are geometrical equations valid indepen-
Λr2
Z  
κM κ 3η
dently of the Lagrangian chosen. Equations (49) and f (r) = 1 − − − E 2 (r) + E 4 (r) r2 dr
(50) are coupled to equations (44). In fact Fµν enters 4πr 3 2r κ
the gravitational field equations (44) and the metric gµν (56)
enters the electromagnetic field equations (49) through where M is the ”mass”. Note that the electric field is
the covariant derivative. These equations, i.e. (44), (49) singular at r = 0. One may easily verify that for small
and (50), form a complete set of differential equations to η/κ the Reissner-Nordstrom metric is recovered. In fact
obtain the metric components and the electromagnetic the expansion of f (r) and the electric field E(r) to the
fields. From equations (49) it is clear that the correction first order of η/κ is
term is proportional to ρ−1max . Therefore, as expected,
Λr2 q2 q4 η
 
M
in low energy regimes there would be no deviations from f (r) ≃ 1 − −κ − 2+
standard electrodynamics. Now, let us start with a spher- 3 4πr 2r 10r6 κ
(57)
ically symmetric space-time as q 2q 3 η
E(r) ≃ 2 − 6
r r κ
dr2
ds2 = −φ(r)f (r)dt2 + + r2 dΩ2 (51) The metric is still singular at r = 0. However, this sin-
f (r)
gularity is inside the horizon. Since this solution is an
As in GR, regarding the spherical symmetry of the met- exterior solution to a charged black hole, the r = 0 sin-
ric, one may use the following components for Fµν : gularity does not imply a singular solution. One needs to
Ftr = −Frt = E(r), where E(r) is the electric field, find interior solutions in the presence of normal matter
and the other components are all zero. With this choice, in order to check the singularity structure of EMSG. For
 9

example the singularity-free expanding/collapsing FRW Comparing to ΛCDM model, there is an extra duty for
universe that we have already explored in this paper lies the cosmological constant in this theory. In fact, a posi-
in this category. Therefore if a finite maximum density tive Λ is necessary for the existence of a regular bounce
arises in the interior solutions, more specifically in the at early universe.
gravitational collapse, then EMSG propose an entirely Also an exact solution for a charged black hole has been
singularity-free universe. found. We recall that Schwarzschild and Kerr metrics are
also solutions for EMSG field equations. However, the
charged black hole solution in EMSG is different from
V. CONCLUSION the standard Reissner-Nordström space-time.
As a further study it is needed to check the existence
In this paper a new covariant generalization of GR is of stable compact stars in EMSG; for such a study in the
developed. This theory allows the existence of a term context of EiBI see [21]. It is also necessary to inves-
proportional to Tαβ T αβ in the action. Therefore we tigate the consequences of the rapid decrease of Ωr and
referred to this theory as Energy-Momentum Squared the accelerated expansion right after the bounce. Finally
Gravity (EMSG). EMSG is different from GR only in one may expect quantum effects to become important
the presence of matter sources. In this theory the cor- at ultra-short distances and ultra-high energy densities.
rection term can be defined only when the Lagrangian Although in order to avoid these effects one may require
density for the matter content is specified. Therefore in ρmax < ρp and amin > lp where ρp is the Planck den-
order to find the field equations, one must first vary the sity and lp the Plancklength. Using the current value
matter action with respect to the gravitational degrees of the radiation energy density and scaling a0 = 1, one
of freedom. Although this feature is not the case in GR, can easily show that if η > ~G3 then both conditions are
it appears in theories which introduce correction terms satisfied. If this constraint is consistent with the cosmo-
including the energy-momentum tensor in the generic ac- logical observations, then, in EMSG, the universe may
tion not enter a quantum era during its evolution.
Applying this theory to a homogeneous and isotropic
space-time, we find that there is a maximum energy den-
sity ρmax , and correspondingly a minimum length scale VI. ACKNOWLEDGMENTS
amin , at early universe. In other words, we showed that
there is a bounce at early times and consequently the We thank the anonymous referee for useful and con-
early time singularity is avoided. We found the exact structive comments. M. Roshan is grateful to Ahmad
value of ρmax . Also we estimated the minimum value of Ghodsi and Luca Amendola for valuable discussions, and
the cosmic scale factor. Moreover, the dynamical system to Salvatore Capozziello for valuable comments. F. Sho-
method has been used to investigate the cosmological jai is grateful to University of Tehran for supporting this
behavior of EMSG. It turned out that EMSG possesses work under a grant provided by the university research
a true sequence of cosmological eras (or fixed points). council.

[1] Eric Poisson and Clifford M. Will, Gravity: Newto- Class. Quant. Grav. 10 1579 (1993); R. L. Seliger and
nian, Post-Newtonian, Relativistic (Cambridge Univer- G. B. Whitham, Proc. Roy. Soc. Lond. A 305, 1480,
sity Press, Cambridge, 2014) 1 (1968); T. P. Sotiriou and V. Faraoni, Class. Quant.
[2] A. Einstein, Sitzungber. Preuss. Akad. Wiss., 414 (1925). Grav. 25, 205002 (2008).
[3] A. S. Eddington, The Mathematical Theory of Relativity [12] S. M. Carroll, Living Rev. Rel. 4, 1 (2001).
(Cambridge University Press, Cambridge, 1924). [13] Z. Haghani, T. Harko, F. S. N. Lobo, H. R. Sepangi and
[4] C.H. Brans, R.H. Dicke, Phys. Rev. 124, 925 (1961). S. Shahidi, Phys. Rev. D 88, 044023 (2013)
[5] F. W. Hehl et al., Rev. Mod. Phys. 48, 393 (1976). [14] A. Ashtekar, T. Pawlowski and P. Singh, Phys. Rev. D
[6] S. Capozziello and M. De Laurentis, Phys. Rept. 509, 74, 084003 (2006).
167 (2011). [15] Y. Shtanov and V. Sahni, Phys. Lett. B 557, 1 (2003).
[7] M. Banados and P. G. Ferreira, Phys. Rev. Lett. 105, [16] S. Capozziello, F. S. N. Lobo and J. P. Mimoso, Phys.
011101 (2010). Rev. D 91, no. 12, 124019 (2015); S. Capozziello,
[8] M. Novello and S. E. P. Bergliaffa, Phys. Rept. 463, 127 F. S. N. Lobo and J. P. Mimoso, Phys. Lett. B 730,
(2008). 280 (2014).
[9] N. Katrc and M. Kavuk, Eur. Phys. J. Plus 129, 163 [17] L. Amendola, Phys. Rev. D 62, 043511 (2000); L. Amen-
(2014). Unfortunately this paper hasn’t been cited in our dola, D. Polarski and S. Tsujikawa, Phys. Rev. Lett. 98,
published paper in Physical review D. In fact after pub- 131302 (2007).
lication of our paper we found out about this paper. [18] L. Perko, Differential equations and dynamical systems,(
[10] T. Harko, F. S. N. Lobo, S. Nojiri and S. D. Odintsov, Springer, 1991)
Phys. Rev. D 84, 024020 (2011). [19] E. J. Copeland, A. R. Liddle and D. Wands, Phys. Rev.
[11] B. F. Schutz, Phys. Rev. D 2, 2762 (1970); J. D. Brown, D 57, 4686 (1998); C. Xu, E. N. Saridakis and G. Leon,
 10

JCAP 1207, 005 (2012); S. Capozziello and M. Roshan, 1997).

Phys. Lett. B 726, 471 (2013); M. Roshan, M. Nouri [20] M. Born and L. Infeld, Proc. Roy. Soc. Lond. A 144, 425
and F. Shojai, Phys. Lett. B 672, 197 (2009); M. Jamil, (1934).
D. Momeni and R. Myrzakulov, Eur. Phys. J. C 72, 2075 [21] P. Pani, V. Cardoso and T. Delsate, Phys. Rev. Lett.
(2012); H. Shabani and M. Farhoudi, Phys. Rev. D 88, 107, 031101 (2011).
044048 (2013); J. Wainwright, G. F. R. Ellis, Dynami-
cal Systems in Cosmology,(Cambridge University Press,