T2- inflation: Sourced by energy-momentum squared

gravity

Seyed Ali Hosseini Mansooria , Fereshteh Felegarya , Mahmood Roshanb,c , Özgür

Akarsud , and Mohammad Sami e,f,g
a
Faculty of Physics, Shahrood University of Technology,
P.O. Box 3619995161, Shahrood, Iran
b
Department of Physics, Faculty of Science, Ferdowsi University of Mashhad,
P.O. Box 1436, Mashhad, Iran
arXiv:2306.09181v2 [gr-qc] 19 Oct 2023

c
School of Astronomy, Institute for Research in Fundamental Sciences (IPM), Tehran, Iran, P.O.
Box 19395-5531
d
Department of Physics, Istanbul Technical University,
Maslak 34469 Istanbul, Turkey
e
Centre for Cosmology and Science Popularization (CCSP), SGT University,
Gurugram, Delhi- NCR, Haryana- 122505, India
f
Eurasian International Centre for Theoretical Physics, Astana, Kazakhstan
g
Chinese Academy of Sciences,
52 Sanlihe Rd, Xicheng District, Beijing
E-mail: [email protected], [email protected],
[email protected], [email protected], sami [email protected]

Abstract: In this paper, we examine chaotic inflation within the context of the energy-
momentum squared gravity (EMSG) focusing on the energy-momentum powered gravity
(EMPG) that incorporates the functional f (T2 ) ∝ (T2 )β in the Einstein-Hilbert action,
in which β is a constant and T2 ≡ Tµν T µν where Tµν is the energy-momentum tensor,
which we consider to represent a single scalar field with a power-law potential. We also
demonstrate that the presence of EMSG terms allows the single-field monomial chaotic
inflationary models to fall within current observational constraints, which are otherwise
disfavored by Planck and BICEP/Keck findings. We show that the use of a non-canonical
Lagrangian with chaotic potential in EMSG can lead to significantly larger values of the
equi
non-Gaussianity parameter, fNl whereas EMSG framework with canonical Lagrangian
gives rise to results similar to those of the standard single-field model.
Contents

1 Introduction 1

2 The model and background field equations 3

3 Cosmological perturbations and the stability of the model 5

4 Inflationary solutions 6
4.1 Predictions of the model with chaotic potentials 8
4.2 Numerics 13

5 Non-canonical matter Lagrangian and large non-gaussianities 14

6 Conclusions 19

1 Introduction

Over the past few years, a variety of extended theories of gravity have been discussed in the
literature, motivated by theoretical and observational considerations. For instance, some
of these schemes aim to explain the cosmic speed-up, whereas others propose to replace
the role of dark matter with a modification of gravity. There are several theories that have
been proposed to extend General Relativity (GR) by adding new gravitational scalar fields
to the Einstein-Hilbert action. Some of these theories include the Brans-Dicke scalar-tensor
theory [1] and the mimetic dark matter theory [2]. To address the dark matter problem,
some theories also incorporate an additional vector field; examples of such theories include
TeVeS [3], MOG [4] and the new relativistic theory of Modified Newtonian Dynamics
(MOND) [5]. While some theories introduce new fields to extend General Relativity (GR),
others modify the existing fields instead. An example of the latter is f (R) gravity where
a function of the Ricci scalar R is used in the action [6]. Although this theory can be
transformed into a scalar-tensor theory, the theory itself only involves the metric tensor as
the gravitational field. In non-local gravity (NLG) [7], the metric tensor is similarly the
only gravitational field present.
In addition, there exists a specific class of modified theories which permit the presence
of scalars constructed from the energy-momentum tensor Tµν in the action. An illustration
of this concept can be seen in f (R, T ) gravity, where the action involves the scalar T =
g µν Tµν , which is the trace of Tµν [8]. Similarly, the f (R, T2 ) model includes an arbitrary
function, T2 ≡ T µν Tµν (viz., the self-contraction of the energy-momentum tensor T µν ) in
the action [9–12]. This model is commonly known in the literature as Energy-Momentum-
Squared-Gravity (EMSG).

–1–
 Before we proceed further, let us briefly review the current status of EMSG within
the wider landscape of modified gravity theories. Numerous studies in the literature have
delved into the cosmological and astrophysical consequences of EMSG. For instance, in [10],
a basic model of the form f (R, T2 ) = R − αT2 is employed to investigate bouncing cosmo-
logical solutions and address the big bang singularity. Additionally, the EMSG incorporates
additional terms that introduce quadratic pressure and density terms into the Friedmann
equations, similar to the corrections found in loop-quantum gravity [13] for α > 0, and the
brane world scenarios [14] for α < 0. These terms, for α > 0, permit the possibility of
bouncing solutions. After conducting further investigations, doubts were raised about the
feasibility of the cosmological bounce in the aforementioned simple model [15]. However,
in Ref. [16], the authors successfully obtained feasible bouncing solutions in EMSG by
introducing the Palatini version of the theory.
In the framework of EMSG, it is possible to achieve late-time accelerated expansion
through conventional matter-energy sources without having to incorporate the cosmologi-
cal constant; for instance, in the EMSG of the form f (R, T2 ) = R − α(T2 )β [known also as
energy-momentum powered gravity (EMPG)], provided that β ∼ 0 [11, 12]. To explore var-
ious cosmological exact solutions in EMPG, we recommend referring to [12]. The dynamical
systems analysis of EMSG has also been performed in [10, 17], which demonstrates that
the simple R−αT2 model features a suitable series of cosmological fixed points (or epochs).
Furthermore, [18, 19] conduct a comprehensive analysis of the cosmological implications
√ of
2
the scale-independent EMSG, specifically with regard to the f (R, T ) = R − α T term; 2

which can also lead to the accelerated expansion of the universe. Last but not least, it
should be noted that most of the modifications to gravity involve extra degrees of freedom,
whereas EMSG does not resort to the same.
Several studies have also explored the astrophysical implications of EMSG, including
the post-Newtonian limit of EMSG and the bending of light experiment analyzed in [20], the
structure of compact stars in EMSG detailed in [21], the Jeans analysis in EMSG conducted
in [22], constraints on EMSG from binary pulsar observations explained in [23, 24], and
constraints on EMSG from neutron star observations discussed in [25]. Additionally, certain
cosmological observations have imposed limitations on the free parameter of EMSG, namely
α [26, 27].
On the other hand, in the framework of inflationary cosmology, the recent Planck
results put severe restriction on the inflationary parameters. For instance, using observa-
tions from Planck, WMAP, and BICEP/Keck during the 2018 observing season [28], the
tensor-to-scalar ratio parameter is limited to r < 0.036 at 95% confidence. As a result of
this limitation, chaotic inflation [29, 30] with a potential of ϕn , even for n = 2/3, has been
excluded at about 95% CL. Therefore, the focus of our paper is to examine the inflationary
parameters - such as the spectral index ns , tensor-to-scalar ratio r, and non-Gaussianity
equi
[31–40] parameter in an equilateral shape fNl - for chaotic inflation in the presence of
EMSG corrections. Our expectation is that the EMSG corrections will yield values of ns
and r that fall within the current BICEP/Keck bound [28, 41], as opposed to being ruled
out in the standard model of chaotic inflation.
The rest of the paper is organized as follows: In Section 2, we attempt to construct our

–2–
model within the EMGS framework by using the energy-momentum tensor associated with
the canonical single field Lagrangian. In this regard, this setup is a subset of K-essence
models [42]. Section 3 begins by examining the stability of the model at the level of cosmo-
logical perturbations. In particular, this finding imposes restrictions on the free parameter
of EMSG, disfavoring many works have been done in the EMSG model. In Section 4,
under the slow-roll scheme, we then obtain an inflationary solution via the background
equi
solutions in our scenario. Moreover, the inflationary parameters - such as {ns , r, fNl },
are both analyzed and discussed, especially at the end of Section 4. These findings are
compared to those of standard chaotic inflation [29, 30]. One striking feature of EMSG
corrections is that they shift the tensor-to-scalar ratio r to smaller values, which brings
them in line with the recent BICEP/Keck bound. Furthermore, in Section 5, we determine
the inflationary parameters by considering the energy-momentum tensor derived from a
equi
non-canonical Lagrangian [43]. The value of fNl is significantly larger when compared to
that in the canonical Lagrangian. Our conclusions are drawn in Section 6.

2 The model and background field equations

Let us start by taking the EMPG model described by the following action [11, 12]:
√ h
Z
1 i
S= d4 x −g Mp2 R − α Mp4(1−2β) (T2 )β + 2Lm , (2.1)
2
where Mp is the reduced Planck mass, R is the Ricci scalar associated with the spacetime
metric gµν , Lm is the Lagrangian density corresponding to the matter source described
by the energy-momentum tensor Tµν . In addition, T2 ≡ Tµν T µν is a scalar and α is a
dimensionless constant that determines the coupling strength of the EMPG modification.
It should be stressed that in [10], the specific case β = 1 is explored. So here we study a
more general case.
Unlike previous studies, such as [8–10, 44], where the perfect fluid energy-momentum
tensor Tµν = (ρ + p)uµ uν + pgµν (where ρ is the energy density, p is the thermodynamic
pressure, and uµ is the four-velocity satisfying the conditions uµ uµ = −11 ) was used, in this
work we construct Tµν by varying the canonical scalar field Lagrangian Lm = X − V (ϕ)
where X = −(∂µ ϕ∂ µ ϕ)/2 with respect to the metric. In this case, Tµν reads
√
2 δ( −gLm )
Tµν ≡ − √ = ∂µ ϕ∂ν ϕ + gµν (X − V ), (2.2)
−g δg µν
from which we obtain
" 2 #
2 2 X X
T = (2V ) − +1 . (2.3)
V V

Combining this result with Eq. (2.1), the action recasts to the K-essence [42] model de-
scribed by
α
P (X, ϕ) = X − V − Mp−4(2β−1) T2β . (2.4)
2
1
We use the metric signature, (-,+,+,+).

–3–
Making use of such a function, we are able to derive the background equations of motion
in a spatially flat FLRW spacetime,

ds2 = −dt2 + a2 δij dxi dxj . (2.5)

Here the scale factor a and the field ϕ depend only on the cosmic time, i.e., a = a(t) and
ϕ = ϕ(t). Generally, the corresponding energy-momentum tensor is characterised by the
pressure p = P (X, ϕ) and the density

ρ(X, ϕ) = 2XP,X (X, ϕ) − P (X, ϕ), (2.6)

where the comma denotes the partial derivative with respect to X. As a result, the back-
ground evolution of the scale factor of the universe and the scalar field is given by a set of
cosmological equations [45], i.e.,

ρ = 3H 2 Mp2 , (2.7)
ρ̇ = −3H (ρ + P ) . (2.8)

where H = ȧ/a is the Hubble parameter and the dot stands for derivative with respect to
the cosmic time. Hereafter, for the sake of convenience, we fix Mp2 = 1 throughout paper.
Taking advantage of Eq. (2.4), the above relations reduce to
h  X 2 X  i
3H 2 = X + V + 2αT2(β−1) V 2 (1 − 4β) + (2β − 1) +1 . (2.9)
V V
and
h   X 3  X 2 X  i
Ẋ 1 − 16αβT2β−2 V 3 2(4β − 1) + (1 − 8β) + (2β + 5) −1
V V V
h   X 3  X 2 X  i
+ ϕ̇V ′ 1 + 8αβT2β−2 V 3 (4β − 3) + (11 − 10β) − (5 − 4β) +2
V V V
h   X i
2β−2
= −6XH 1 + 2αβT V 1−2 , (2.10)
V
where the prime stands for the derivative with respect to ϕ and X = ϕ̇2 /2. We observe that
2
the term ∂g∂αβL∂gmµν that arises from the variations of the action (2.1) does not contribute
to these field equations. It can be verified that this term is identically zero for a canonical
scalar field described by Lm = X − V (ϕ), see Ref. [46, 47]. One may also verify that, for
α = 0, as expected, the field equations reduce to their canonical form, specifically the field
equations of GR in the presence of a canonical scalar field.
It’s important to note that even when α ̸= 0, the standard Einstein field equations of
GR persist, but now in the presence of a canonical scalar field being complemented by a
specific K-essence model determined by the EMSG model under consideration. This aligns
with a recent study [46] suggesting that EMSG, and more broadly, matter-type modified
gravity theories like f (Lm ), f (gµν T µν ), and f (Tµν T µν ), which modify the introduction
of the material source in the conventional Einstein-Hilbert (EH) action by incorporating
exclusively matter-related terms into the matter Lagrangian density Lm , are equivalent
to GR. In this equivalence, the usual source is accompanied by a distinct new source,
determined by the matter-type modified gravity, which typically interacts non-minimally
with the usual source.

–4–
3 Cosmological perturbations and the stability of the model

In this section, we attempt to investigate whether the model suffers from the ghost and
gradient instabilities. To do this, we review the analysis of cosmological perturbations done
in Ref. [45] in the comoving gauge.
The scalar and tensor perturbations of the metric around the background geome-
try (2.5) in the comoving gauge are given by
 
δg00 = 2A, δg0i = 2a∂i B, δgij = a2 e2R δij + hij , (3.1)

where A, B, and R are scalar perturbations, while hij describe the tensor perturbations.
Note that the action (2.1) has O(3) symmetry. Consequently, the scalar and tensor pertur-
bations decouples at the linear order of perturbations. On the other hand, because of the
isotropic symmetry, the vector perturbations decay in an expanding Universe. Therefor,
we do not consider them here.
After substituting the metric perturbations in the action (2.1), and expanding it up to
the second order and then integrating out the non-dynamical modes (A, B), one obtains
the quadratic action in terms of the dynamical modes (R, hij ) [45, 48] as follows:
c2
Z
1 hε 
H
 1 1 i
(2)
S = dtd3 xa3 2 Ṙ2 − s2 (∂R)2 + (ḣij )2 − 2 (∂hij )2 , (3.2)
2 cs a 4 a
where speed of sound cs and the standard slow roll parameter εH are defined as
P,X Ḣ XP,X
c2s = , εH ≡ − 2
= . (3.3)
P,X + 2XP,XX H H2
For the perturbations to be free from ghost and gradient instabilities, it is necessary that
both parameters, c2s and εH , are positive. Correspondingly, we require

P,X > 0, P,X + 2XP,XX > 0 (or equivalently P,XX > 0). (3.4)

Exploiting Eq. (2.4), these constraints read

 X
P,X = 1 + 2αβT2β−2 V 1 − , (3.5)
V
h  X 2 X  i
P,XX = −8αβT2β−4 V 2 2(2β − 1) − 2(2β − 1) + (1 + β) , (3.6)
V V
respectively. Interestingly, for β = 1, one requires that α < 0 to satisfy the second con-
straint, whereas the first one confirms (1 + 2αV ) > 0. In the next section, we will show the
last constraint provides us with H 2 in the slow roll approximation. During the inflation
epoch under the slow-roll limit where X ≪ V , Eq. (3.6) reduces to

P,XX ≃ −22β−1 αβ(1 + β)V 2β−2 . (3.7)

This condition implies that α < 0 in our scenario for any choice of β. As a result, in order
to satisfy the first condition, the potential must be constrained as follows:
1  1
1−2β
V < |α|β (3.8)
2

–5–
for β > 1/2. In the scale-independent EMSG model with β = 1/2, the first constraint
limits us to taking −2 < α < 0. Therefore as far as we confine ourselves to the scalar
field inflation models, the cosmological models with α > 0 in the scale-independent EMSG
are ruled out. For example, one of the interesting cosmological solutions presented in [19]
takes α = 22 and reproduces the original steady state universe in the presence of dust [49].
Based on the stability analysis discussed above, this choice of α is not a healthy one.
A similar challenge arises with the original bouncing EMSG model with β = 1 studied
in [10]. To be specific, after inflation and still deep inside the radiation dominated phase,
the Friedmann equation in EMSG is written as [10]:
 
2 ρ 1 2 4 1 2
H = −α p + ρp + ρ (3.9)
3 2 3 6

As α < 0 due to stability concerns, it is evident that a viable bounce does not exist in this
model. Specifically, the first condition for the existence of a bounce, i.e., H = 0, is not
satisfied in the early universe.
As a final remark in this section, it is interesting to note that, in the same spirit,
EMSG stabilizes (or destabilizes) a non-relativistic fluid when α < 0 (or α > 0). The Jeans
analysis in EMSG has been explored in [22]. EMSG corrections can be combined to define
the effective density ρeff and pressure peff for a fluid with a given fluid density ρ and p. The
the standard Jeans analysis reveals that the effective sound speed Cs , obtained from ρeff and
peff , appears in the dispersion relation of the perturbations instead of the standard sound
speed cs . Specifically, the effective sound speed is given by Cs2 = c2s − αρ. Therefore, for
α < 0, EMSG corrections increase the effective sound speed. If we consider the sound speed
as the representative of the pressure in the system, acting against gravitational collapse,
one may infer that EMSG induces stabilizing effects in the case of α > 0. For more details,
see Ref. [22].

4 Inflationary solutions

The aim of this section is to obtain a period of inflation in the early universe by making use
of the background equations. Under the slow-roll scheme, when ϕ̇2 ≪ V (or equivalently
X ≪ V ) and ϕ̈ ≪ H ϕ̇ (or Ẋ ≪ HX), Eqs. (2.9) and (2.10) reduce to
 
3H 2 ≃ V 1 + αṼ , (4.1)
   
ϕ̇V ′ 1 + 2βαṼ ≃ −6XH 1 + βαṼ , (4.2)

where Ṽ = (2V )γ with γ = 2β − 1. Note that Eq. (4.1) demonstrates that there is an
upper bound on the potential, i.e.,
1 1 1
V < |α| 1−2β for β> (4.3)
2 2
[19] uses a parameter α in the action of the theory to identify EMSG corrections. Let us call it α∗ .
2

The relation between α∗ and our α is α∗ = −α/2.

–6–
This is situated within the acceptable bound (3.8), necessary to evade the ghost instability.
In addition, in the particular case β = 1/2, the positiveness of the energy density modifies
the lower bound on α, namely −1 < α < 0. To sum up, to remedy the ghost and gradient
instabilities, the coupling constant α must be placed at

α<0 for β > 12 ,

(4.4)
−1 < α < 0 for β = 21 .

It is worth noting that for the EMPG model with β = 1, α > 0 is not well-behaved when
the growth of the linear matter perturbations is concerned [50]. Furthermore, this finding
raises severe criticism about the inflationary solutions obtained in [26], where the authors
assume α > 0 without considering the stability of the model.
Taking the time derivative of the both sides of Eq. (4.1) and using it together with
Eq. (4.1), we obtain the Hubble slow-roll parameter

1  Ṽ ′  ϕ̇  1 + 2αβ Ṽ 
εH ≃ − . (4.5)
2γ Ṽ H 1 + αṼ

From Eq. (4.2), it is straightforward to write the ratio ϕ̇/H as

ϕ̇  Ṽ ′  1 + 2αβ Ṽ
=−   . (4.6)
H γ Ṽ 1 + αβ Ṽ 1 + αṼ

Thus the slow roll parameter (4.5) converts to

 2
1

Ṽ ′
2 1 + 2αβ Ṽ
εH = 2 . (4.7)
2γ 2
 
Ṽ 1 + αβ Ṽ 1 + αṼ

The other Hubble slow-roll parameter ηH can be expressed as

 
′  ′ 2 2 + αṼ 4 + 3β + αβ Ṽ (9 + 2β + 6αβ Ṽ )
ε˙H ε ϕ̇ 1 Ṽ
ηH = = H = 2  2
HεH εH H γ Ṽ

1 + αṼ 1 + αβ Ṽ
  (4.8)
 ′′ 
2 Ṽ 1 + 2αβ Ṽ
−   .
γ Ṽ 1 + αṼ 1 + αβ Ṽ

At the limit α → 0, both slow roll parameters reduce to the standard form. Great care
must be taken when investigating the slow roll parameters. First, during the inflation era
these parameters must be much smaller than one, namely εH ≪ 1 and ηH ≪ 1. Second,
their evolution must take at least 50-60 number of e-folds to solve the flatness and the
horizon problems. At the end, inflation ends when either of the slow-roll parameters tends
to unity. Moreover, in the slow-roll limit, the sound speed can be estimated as

αβ(1 + β)(1 + αṼ )Ṽ  ϕ̇ 2

cs ≃ 1 + . (4.9)
6(1 + αβ Ṽ ) H

–7–
Since β > 0 and 1 + αṼ > 0, it is clear that the case of α > 0 leads to superluminal
speed of sound which naturally implies the violation of Null Energy Condition (NEC) in
regular bounce models. It should be stressed that the superluminal sound speed does not
necessarily violate the causality [51]. We leave this as a subject of study for future works.
Furthermore, using Eq. (4.6), one can express the above relation as function of the
potential and its derivatives. Following [45, 48], the scalar and tensor power spectrum3 in
the slow roll approximation are also given by

1 H2 2 2
PR ≃ |c k=aH , Ph = H |k=aH (4.12)
8π 2 εH cs s π2
Having calculated power spectra, we can also calculate the spectral index ns and the
tensor-to-scalar ratio r as follows:
d ln PR
ns − 1 ≡ ≃ −2εH − ηH − s , (4.13)
d ln k
Ph
r≡ = 16εH cs , (4.14)
PR
where s ≡ c˙s /(Hcs ). Moreover, the tensor spectral index nT is given by

d ln Ph
nT ≡ = −2εH (4.15)
d ln k
and it satisfies the consistency relation r = −8cs nT . The parameters r and ns can be
limited by current observational constraints on inflationary parameters [53, 54]. Using
Eqs. (4.7) and (4.8), one can express the above parameter in terms of the scalar potential
V (ϕ) and its derivatives, rather than with ϕ itself.

4.1 Predictions of the model with chaotic potentials

Up to now, we have obtained the equations for the slow roll parameters, the spectral
index, and tensor to scalar ratio without selecting a particular potential type. In this
subsection, we proceed with using the chaotic potentials V (ϕ) = M A n
n−4 ϕ , where A is a
p
dimensionless coefficient. For such a potential, n takes rational numbers and A stands for
the normalisation parameter given by the amplitude of the scalar power spectrum at the
CMB pivot scale.
3
An exact determination of PR is provided by solving the Mukhanov-Sasaki equation given by

d2 vk  2 2 1 d2 z 
+ cs k − vk = 0 (4.10)
dτ 2 z dτ 2
where τ is the conformal time, the variable k relates to the comoving scale by λ = 2π/k, subscript k
indicates the momentum space [52]. Additionally, the curvature perturbation R is related to v via R = v/z
2
and the background variable z is defined as z = a c(ρ+P
2 k2
)
. Finally, the power spectrum of the scalar mode
s
at the horizon crossing where aH = cs k can be expressed as

k3 vk 2
PR = | | (4.11)
2π z

–8–
 The number of e-folding (N ) measures how much inflation took place to the end of
inflation and is defined as
Z t Z ϕ Z Ṽ
H 1 dṼ 1 + 2αβ Ṽ
N =− Hdt = − dϕ = , (4.16)
te ϕe ϕ̇ Ṽe 2γεH Ṽ 1 + αṼ
where we have used Eq. (4.5). Here the subscript e indicates the value of the quantities
at the end of the inflation. By substituting Ṽ = Ãϕγn , where à = (2A)γ into the above
expression, it is straightforward to verify that
2 h
1  Ṽ  γn αṼ  2 2 i
N= 1+ 1 − γ 2 F1 (1, 1 + ,2 + , −2αβ Ṽ ) . (4.17)
2n à 2 + γn γn γn
This concurs well with [43, 55] in the α → 0 limit. Notice that the potential Ṽ in (4.17)
should be calculated at the beginning of the inflation. Moreover, for the scale-independent
EMSG (β = 1/2, or γ → 0), the above general relation reduces to
2 + α  V  n2  4nN  n
2
N= and conversely V = A . (4.18)
4n A 2+α
While for the case γ > 0, due to the existence of the hyper-geometrical function in
Eq. (4.17), it becomes difficult to express the potential function V q
as a function of N
in reverse. However, it can be done in the situation that |α|1/γ V ≤ O( εSC
H ) < 1/2 during
inflation.4 By making this assumption, and defining the expansion parameters ϵ = αA and
ϵ̃ = αÃ, one obtains
γn
Ṽ  
2
h i
≃ 2nN 1 + f1 (N )ϵ̃ + f2 (N )ϵ̃2 + O(ϵ̃3 ) , (4.19)
Ã
where the auxiliary functions f1 and f2 are defined as
γn
(2nN ) 2 (β − 1)
f1 (N ) = γn ,
2 + γn
(4.20)
(2nN )γn
f2 (N ) = γn f (β),
2(1 + γn)(2 + γn)2
and the function f (β) is given by
   
f (β) = 2 − 6β 2 + n2 γ 2 3 + (β − 5)β + nγ 5 − 3β(2 + β) . (4.21)

Although we have considered

q the above relation up to the second order, as long as |α|1/γ V
is close to the bound εSC −1
H ∼ O(10 ), one needs to take higher orders, at least the
forth order, to make an excellent agreement between analytical and numerical results as
presented in Fig. 2.
Now, by combing Eqs. (4.7), (4.8) with Eq. (4.19), we can obtain the relation between
the slow-roll parameters and N approximated to the second order of αÃ
n h f1 (N )(2 + n(3β − 2)) 2f2 (N ) i
εH ≃ 1+ ϵ̃ + (g(β) − f (β))ϵ̃2 + O(ϵ̃3 ) , (4.22)
4N n(β − 1) γnf (β)
4
The symbol SC stands for the standard chaotic inflation.

–9–
 1h γf1 (N )(2 + n(3β − 2)) f2 (N ) i
ηH ≃ 1− ϵ̃ + (p(β) − 2f (β))ϵ̃2 + O(ϵ̃3 ) , (4.23)
N 2(β − 1) nγf (β)
where the functions g(β) and p(β) are defined as
 
g(β) = (1 + nγ)(2 + nγ) 3 + 5nγ + β(nγ(4β − 11) − (4 + 3β)) , (4.24)

p(β) = (1 + γn)(2 + γn)(2 − 6β 2 + 2nβγ(7β − 3) + n2 γ 2 (β(10 + β) − 6)). (4.25)

Note that our finding confirms the relation ηH = −d ln ϵH /dN . In addition, the sound
speed cs is obtained to be
nγ
2n nγ
h 2nγ(2nN ) 2 (β 2 − 1) i
cs ≃ 1 + (2nN ) 2 β(β + 1)Ãα 1 + ϵ̃ + O(ϵ̃2 ) . (4.26)
3N (2 + nγ)(1 + β)

This allows to calculate the slow-roll parameter s ≡ c˙s /(Hcs ) = −d(ln cs )/dN . Therefore,
a formal solution for the spectral index (4.13) is given by

1 h 2 + n n (1 − γ)f1 (N )(2 + n(3β − 2)) n nγ

o
ns − 1 ≃ − + − (2nN ) 2 β(β + 1)(nγ − 2) ϵ̃ ,
N 2 2(β − 1) 24N 2
n n2 nγ

2 2

+ (2nN ) β(β + 1) − 48γN (β − 1)(nγ − 1) + β(β + 1)(n γ − 4)
288(2 + nγ)N 2
f2 (N )  o i
+ p(β) + ng(β) − (2 + n)f (β) ϵ̃2 + O(ϵ̃3 ) . (4.27)
nγf (β)

It should be noted that all the parameters mentioned above can be expressed up to arbitrary
order of αÃ. Here, for the sake of simplicity, we have written the first three leading terms.
As we can see, the result for the spectral index for the case α = 0 in [43, 55] is modified
by the orders of αà in our scenario. Note that it is easy to write down r in the terms of
N by combining Eqs. (4.14), (4.22), and (4.26) together.
For the special case of the scale-independent EMSG (β = 1/2, or γ → 0), and by
taking the potential V = Aϕn , we arrive at

nα(α + 1)
cs = 1 + , (4.28)
4N (2 + α)2
n 1
εH = , ηH = , (4.29)
4N N
4n n2 α(α + 1) 4n
r = 16εH cs = + 2 2
≃ , (4.30)
N N (2 + α) N
2+n
ns − 1 = − . (4.31)
2N
As a consequence of the above discussion, the scale-independent EMSG predicts the same
inflationary parameters as those predicted by the standard single-field power-law infla-
tion [43, 56].
Now, let us consider another special EMSG model with β = 1 (or γ → 1). By using
the chaotic potential Aϕ2 , the parameters for this model can be expressed up to fifth order

– 10 –
of ϵ as follows.
8ϵ 1024 2 3 16384 3 4
cs ≃ 1 + − N ϵ + N ϵ + O(ϵ5 ) , (4.32)
3 3 3
8  1   512 2560  2048
r≃ + 64 1 + ϵ+ − N ϵ2 + N (27N − 22)ϵ3
N 3N 3 3 9
16384 2 22
+ N ( N + 21)ϵ4 + O(ϵ5 ) , (4.33)
9 5
2 512 2 512
ns − 1 ≃ − − Nϵ + N (−4 + 27N )ϵ3
N 3 3
16384  208 
− N N( N − 9) − 1 ϵ4 + O(ϵ5 ). (4.34)
9 5
It is obvious that the EMSG corrections provides us with a clear improvement on the
value of r and ns for the standard single field inflation. Specifically, due to the negative
nature of α and ϵ, all corrections have the potential to decrease r and ns in comparison
to standard single field inflation, bringing them closer to current observational constraints.
For instance, in order to be compatible with large scale CMB observations [53, 54] at the
pivot scale k∗ = 0.05 Mpc−1 with

0.956 < ns < 0.978, r(k∗ ) ≤ 0.06 at 95% C.L., (4.35)

which in turn implies for ϵ = αA that

−0.00064 < αA ≤ −0.00052 (4.36)

for N = 70, whereas for N = 50 and N = 60 the EMSG model cannot put into the
observational constraint. In Fig. 2, the solid curves present the values of {ns , r} with
respect to ϵ by using analytical relations for the power law potential V = Aϕn with n = 2
and n = 3/2, respectively. Moreover, we compared these results with numerics that are
depicted by colored points in Fig. 2. More importantly, there is satisfactory agreement
between numerics and analytics. In the next subsection, we explain our numerical method.

Before proceeding, let us analytically investigate another crucial inflationary parame-

ter, namely the non-Gaussianity parameter fNL in our scenario. Observational limitations
on primordial non-Gaussianities have the potential to constrain the parameter space of our
model. In fact, according to the Planck observations, the amplitude of the non-Gaussianity
equi
parameter fNL in the equilateral shape is constrained to be fNL = −26±47 (68% C.L.) [53].
In the K-essence models, the non-Gaussianity parameter in the equilateral limit at leading
order is given by
equi 51 λ 35  1 
fNL ≃ − 1 − 2 − − 1 , (4.37)
81 c2s Σ 108 c2s
where
2
λ = X 2 P,XX + X 3 P,XXX , Σ = XP,X + 2X 2 P,XX . (4.38)
3
In the slow-roll approximation, one finds, from Eq. (4.9) that
λ αβ(β + 1)(1 + αṼ )Ṽ  ϕ̇ 2 1
2 ≃− ≃ 2 − 1. (4.39)
Σ 3(1 + αβ Ṽ ) H cs

– 11 –
 10−8
Numerics
CMB Normalization Analytics
10−9 ns = 0.9622
100

CMB Scale
r = 0.0899

Slow-roll parameters
10−10
PR 10−1
|ηH |
10−11

H
−12 −2
10 10
Mukhanov-Sasaki
Slow-roll Approx
10−13
0 10 20 30 40 50 60 10 20 30 40 50 60
N N

Figure 1. Left panel : The scalar power spectrum PR is plotted as a function of the number of
e-folds before the end of inflation Ne (a) by using the slow-roll approximation (4.12) (dotted blue)
and (b) by numerically solving the Mukhanov- Sasaki equation (4.11) (red dots). Interestingly,
both methods give the same results for a smoothly varying potential, in which case PR increases
monotonically with increasing Ne . Right Panel : The evolution of the slow-roll parameters during
the inflationary phase (a) by numerical solving background equations (2.9) and (2.10) (b) by using
Eqs. (4.7), (4.8), and (4.17) under slow roll approximation. Note that the slow roll conditions
εH , ηH ≪ 1 remain satisfied during inflation. Here we have considered the potential V = Aϕ2 in
which αA = −0.00046 for EMSG model with β = 1 in both panels. A similar behavior is held for
the other power law potential types.

0.16 N = 70 0.101 N = 70
0.14
N = 60 V (φ) = A φ2 0.081 V (φ) = A φ2/3 N = 60
0.12
N = 50 N = 50
0.10
r

0.061
0.08
0.041
0.06
0.04 0.021
0.02
0.001

0.97 0.99

0.96 0.98
ns

ns

0.95 0.97

0.94 0.96

0.93 0.95

0.92 0.94
0.65
1.17
−3

−3
L | × 10

× 10

0.97
0.45
0.77
L |
|fNequi

|fNequi

0.57
0.25

0.37

0.17 0.05
−0.0009 −0.0008 −0.0007 −0.0006 −0.0005 −0.0004 −0.055 −0.050 −0.045 −0.040 −0.035 −0.030 −0.025 −0.020

 

Figure 2. A comparison between analytical and numerical results for inflationary parameters in
EMSG model with β = 1 by considering a chaotic potential case, left panel : V = Aϕ2 and right
panel : V = Aϕ2/3 . Note that the bullet points represent numerical data, while the solid curves
depict analytical results. Additionally, the horizontal dashed lines in different colors indicate the
relevant observational measurements: BK15 (green) and BK18 (blue).

– 12 –
By substituting the above relation into (4.37), we obtain
equi 35  1 
fNL ≃− − 1 . (4.40)
108 c2s
Obviously, the leading order contribution in the non-Gaussianity parameter vanishes similar
to the DBI models (with a non-canonical kinetic term) where large non-Gaussianity of the
equilateral type can be generated [57].
For the EMSG model with β = 1, in the valid range of the EMSG parameter ϵ (4.36),
the non-Gaussianity parameter for a chaotic potential with n = 2 can be estimated to be
equi 70
fNL ≃ ϵ ∼ O(10−3 ) ∼ O(εSC
H ). (4.41)
81
This is consistent with the numerical results, as shown in Figure 2. Clearly, the non-
equi
Gaussianity parameter fNL falls within an acceptable range, but its value is so small
similar to the standard single field inflation [56]. Consequently, for the models which are
built of the energy momentum tensor associated with the canonical matter Lagrangian, the
equi
non-Gaussianity parameter fNL gets very small values. However, we expect that a non-
canonical matter Lagrangian leads to a large equilateral non-Gaussianity in our scenario.
Several studies, for example [45, 48, 57, 58] have shown that a suitable choice of non-
equi
canonical matter Lagrangian in K-inflation model can generate large fNL by altering the
speed of sound. We will investigate such models in the next section. Before moving on, let
us now compare analytical results with numerics.

4.2 Numerics
equi
To obtain numerical values for ns , r, and fNL , it is necessary to first obtain numerical
solutions for ϕ and H, as required by the background equations, Eqs. (2.9) and (2.10),
as a function of the e-folding number N . In our numerical calculations, the variable N
represents the number of e-foldings before the end of inflation. Thus, N = Ni denotes the
beginning of inflation, while N = Ne = 0 corresponds to the end of inflation. We also
define N ∗ as the number of e-foldings before the end of inflation when the CMB pivot scale
k ∗ = 0.05, Mpc−1 exited the comoving Hubble radius. For convenience, we consider three
values of N ∗ , namely N ∗ = {50, 60, 70}.
It is common for power law potentials of the form V = Aϕn to predict a smooth scalar
power spectrum PR that decreases monotonically from the largest scales (N ∼ N ∗ ) to
the smallest scales (N ∼ 0), as shown in the left panel of Fig. 1.It is worth noting that
the value of the coefficient A is fixed by the CMB normalization on the power spectrum,
which requires PR ≃ 2.1 × 10−9 at the CMB pivot scale k ∗ [53]. Additionally, we show
the evolution of the slow-roll parameters ϵH and ηH in the right panel of Fig. 1. As can
be seen, there is perfect agreement between the numerical results (without the slow-roll
approximation) and those obtained using the slow-roll approximation. To illustrate this,
we plot the slow-roll parameters (4.7) and (4.8) as a function of N from Eq. (4.17), by
smoothly varying the potential form from Ve = Aϕne to Vi = Aϕni . It should be noted that
Ve (or equivalently ϕe ) is determined when the slow-roll parameter εH reaches unity, i.e.,
when the inflationary phase ends.

– 13 –
 0.200 0.200
Planck TT,TE,EE + lowE+lensing N = 70, β = 1/2 Planck TT,TE,EE + lowE+lensing N = 70, β = 1/2
+BK18+BAO N = 60, β = 1/2 +BK18+BAO N = 60, β = 1/2
0.175 0.175
V (φ) = A φ2 V (φ) = A φ2/3
N = 50, β = 1/2 N = 50, β = 1/2
N = 70, β =1 N = 70, β =1
N = 60, β =1 N = 60, β =1
0.150 N = 50, β =1 0.150 N = 50, β =1

0.125 0.125

r 0.100 r 0.100

0.075 0.075

0.050 0.050

0.025 0.025

0.000 0.000
0.95 0.96 0.97 0.98 0.95 0.96 0.97 0.98
ns ns

Figure 3. Tensor-to-scalar ratio vs spectral index for EMSG model with the power law potential
V = Aϕn for n = 2 (right) n = 2/3 (left), compared to the data of Ref. [28].

In Fig. 3, we also present the tensor-to-scalar ratio as a function of the spectral index,
along with the observational constraints from the Planck 2018 data, as well as BICEP/Keck
(BK15 [41] and BK18 [28]) data and BAO data5 . Fig. 3 is interesting for several reasons.
Firstly, the numerical values of r and ns for the case of β = 1/2 are located exactly
where those predicted by the standard single-field inflation for various values of N [41].
This finding confirms the analytical relations (4.30) and (4.31) in the slow-roll regime.
Secondly, incorporating the correction of the energy-momentum tensor with β = 1 into the
standard canonical scalar field inflation improves the predicted values of {r, ns }, bringing
them into agreement with recent observational constraints. For instance, for the potential
V = Aϕ2/3 , the predicted values fall entirely within the region determined by the BK18
results [28].
equi
Furthermore, in Fig. 2, we compare the numerical and analytical results for {ns , r, fNL },
equi
revealing a noteworthy finding: the numerical values of {ns , r, fNL } are in close agreement
equil
with the analytical predictions discussed earlier. Additionally, we observe that fNL yields
small values, similar to those of standard chaotic inflation. However, to address this issue,
we consider a non-canonical Lagrangian in our scenario.

5 Non-canonical matter Lagrangian and large non-gaussianities

Now let’s consider a non-canonical Lagrangian density for a scalar field as [43]

Lm = X δ − V (ϕ) (5.1)

where δ is a positive constant, for which δ = 1 we recover the canonical scalar field La-
grangian. The corresponding Lorentz scalar reads
h i
T2 = 4 (δ(δ − 1) + 1)X 2δ + (δ − 2)V X δ + V 2 . (5.2)
5
The BK18 analysis yielded a 95% confidence constraint from BK15 as r0.05 < 0.07 improved to r0.05 <
0.036. Additionally, the BK18 simulations result in a median 95 upper limit of r0.05 < 0.019.

– 14 –
Here, we apply the slow-roll approximation to evaluate the most important inflationary
parameters for the EMSG model with β = 1. In this type of K-essence model, with
P (X, ϕ) = X δ − V − αT2 /2, we can derive equations for the scale factor a(t) and the scalar
field ϕ, in the slow-roll limit, as follows:6

3H 2 ≃ V (1 + 2αV ), (5.3)
 
V ′ ϕ̇(1 + 4αV ) ≃ −6δHX δ 1 + 2α(2 − δ)V . (5.4)

Taking advantage of Eq. (5.3), we can calculate the Hubble slow-roll parameter as

1  V ′  ϕ̇  1 + 4αV 
εH ≃ − . (5.5)
2 V H 1 + 2αV

From Eq. (5.4), we can write the ratio ϕ̇/H as

ϕ̇ h V ′ (1 + 4αV ) i 1
2δ−1
= −K(δ) δ  , (5.6)
H
 
δ
V 1 + 2αV 1 + 2α(2 − δ)V

  1
6δ−1 2δ−1
where K(δ) = δ . Now, it is straightforward to verify that

  1
2δ  2δ−1
′ (1
K(δ)  V + 4αV )   1
1−2δ
εH ≃ 1 + 2α(2 − δ)V . (5.7)

3δ−1 
2

V (1 + 2αV )

Similarly, the sound speed cs takes the following form

 δ
1 8αδ(1 − δ + δ2) V (1 + 2αV )  ϕ̇ 2δ
c2s = + (5.8)
2δ − 1 6δ (2δ − 1)2 1 + 4αV (1 − 21 δ)α H

Under the assumption of small αV , we obtain the relation between the potential and the
e-folding number N as
n 2δ 4δ 6δ 8δ
n
h i
V ≃ (κKN ) κ n κ(2δ−1) A κ(2δ−1) 1 + f3 αA κ(2δ−1) + f4 α2 A κ(2δ−1) + O(α3 A κ(2δ−1) ) , (5.9)

where the new functions f3 (N ) and f4 (N ) are defined as

δ(δ − 3)
2n
 
1+ 2n
f3 (N ) = 2(κKN ) κ n κ(2δ−1) ,
2δ + n(5δ − 3)
(5.10)
3n δ(11 + (δ − 6)δ) 1+ 3n
f4 (N ) = 8(κKN ) κ n κ(2δ−1)
3(2δ + n(7δ − 4))
2
6
During the derivation of field equations, it is crucial to recognize that the term ∂g∂µνL∂g
m
σϵ exhibits

non-zero behavior in the context of non-canonical scalar fields, in contrast to canonical scalar fields. It
is important to note that the decision to treat this term as zero remains a matter of choice. For further
reading on this topic, refer to Ref. [46].

– 15 –
with κ(δ, n) = n(δ−1)+2δ
2δ−1 . Then, using Eq. (5.7) and Eq. (5.9), we reach the final relation
for the spectral index and the tensor to scalar ratio in Eqs. (4.13) and (4.14);
1h 2n 4δ 8δ
i
ns −1 ≃ − I + (KN ) κ N α2 A κ(2δ−1) + O(α3 A κ(2δ−1) ) , (5.11)
N
8n h 2δ 4δ 8δ
i
r ≃ 16εH cs ≃ cs + f5 αA κ(2δ−1) + f6 α2 A κ(2δ−1) + O(α3 A κ(2δ−1) ) , (5.12)
Nκ
where
n(3δ − 2) + 2δ
I= (5.13)
n(δ − 1) + 2δ
where N is a complicated polynomial function of δ and n, which we cannot present here.
However, for the chaotic potential (n = 2), it simplifies to:
2(4δ−1) 6 + δ(δ − 15)
N =2 2δ−1 . (5.14)
6δ − 3
On the other hand, the functions f5 (N ) and f6 (N ) are found to be
1 n n
f5 (N ) = (κKN ) κ n κ(2δ−1) ,
8
(5.15)
1 2n 2n n(3 + δ(δ − 8)) − 2δ
f6 (N ) = (κKN ) κ n κ(2δ−1) .
4 2δ + n(5δ − 3)

Under the assumption of slow-roll, using (5.8), the sound speed squared is found to be
1 2δ 4δ
c2s = + f7 αA κ(2δ−1) + O(α2 A κ(2δ−1) ). (5.16)
2δ − 1
We find that
8δ (n+2)δ (2+n)δ 2n−2δ(2+n) (n−2)δ
f7 (N ) = δ
K κ (nN ) κ(2δ−1) (2δ − 1) κ(2δ−1) (1 + δ(δ − 1))κ κ(2δ−1) (5.17)
6
as α → 0, which is in consistent with one given in [43]. And, from (4.38), we find that

λ 1 (4δ − 1) h 1i
≃ (δ − 1) + 1 − 2δ + 2 . (5.18)
Σ 3 6(2δ − 1) cs
equi
Finally, the non-Gaussianity parameter fNL (4.37) at leading order is obtained to be

equi 85  1 40  1 (4δ − 1) h 1 i
fNL ≃ 1− 2 − (δ − 1) + 1 − 2δ + 2 . (5.19)
324 cs 324 3 6(2δ − 1) cs

Note that as one takes c−2

s = 2δ − 1, this matches with [58] for α → 0. For instance,
in EMPG model, by selecting a chaotic potential V = Aϕ2 and L = X 2 − V , the non-
Gaussianity parameter reads

equi 275 1810  2 2  13   275

fNL ≃− + A α + O (α2 A4/3 ) ≃ − ,
486 243 9 486
equi
This result is in excellent agreement with the Planck result of fNL = −26 ± 47 (68%
C.L.) [53], as confirmed by the numerical data presented in Tabels 1 and 2. Our view

– 16 –
 N = 70 Numerics Analytics
equil equil
|α|A2/3 ns r c2s |fNL | ns r c2s |fNL |
8.5 × 10−4 0.9253 0.0170 0.3331 0.5663 0.9271 0.0143 0.3332 0.5667
7.9 × 10−4 0.9391 0.0212 0.3330 0.5665 0.9405 0.0193 0.3331 0.5666
7.1 × 10−4 0.9513 0.0271 03330 0.5666 0.9520 0.0262 0.3330 0.5663
6.6 × 10−4 0.9568 0.0316 0.3329 0.5667 0.9569 0.0310 0.3330 0.5661
6.0 × 10−4 0.9612 0.0360 0.3329 0.5668 0.9610 0.0356 0.3329 0.5659
N = 60
8.5 × 10−4 0.9412 0.0311 0.3329 0.5668 0.9418 0.0297 0.3330 0.5667
7.9 × 10−4 0.9478 0.0353 0.3329 0.5669 0.9481 0.0344 0.3329 0.5668
7.1 × 10−4 0.9542 0.0411 0.3329 0.5670 0.9539 0.0405 0.3329 0.5669
6.6 × 10−4 0.9570 0.0441 0.3329 0.5670 0.9556 0.0438 0.3329 0.5670
6.0 × 10−4 0.9597 0.0483 0.3328 0.5670 0.9591 0.0481 0.3328 0.5670

Table 1. A comparison between analytical and numerical results for inflationary parametersfor
inflationary parameters in the non-canonical EMSG model with δ = 2 by considering a chaotic
potential, V = Aϕ2 .
0.200 0.200
N = 70, α = 0 N = 70, α = 0
N = 60, α = 0 N = 60, α = 0
0.175 N = 50, α = 0 0.175 N = 50, α = 0
V (φ) = A φ2 N = 70, α = −6 × 10−4 A−2/3 V (φ) = A φ2/3 N = 70, α = −0.02A−6/7
N = 60, α = −6 × 10−4 A−2/3 N = 60, α = −0.02A−6/7
0.150 N = 50, α = −6 × 10−4
A −2/3 0.150 N = 50, α = −0.02A−6/7

Planck TT,TE,EE + lowE+lensing Planck TT,TE,EE + lowE+lensing

0.125 +BK18+BAO 0.125 +BK18+BAO

r 0.100 r 0.100

0.075 0.075

0.050 0.050

0.025 0.025

0.000 0.000
0.95 0.96 0.97 0.98 0.95 0.96 0.97 0.98
ns ns

Figure 4. Tensor-to-scalar ratio vs spectral index for the non-canonical EMSG model with the
power law potential V = Aϕn for n = 2 (right) n = 2/3 (left), compared to the data of Ref. [28].

is that the results reported in the tables emphasize the validity of our analytical solu-
tions (5.16), (5.11), and (5.12) in the slow-roll limit for inflationary parameters. The most
remarkable finding in our scenario is that the inclusion of the T2 term in the non-canonical
Lagrangian leads to a shift of the data on r and ns towards the BICEP/Keck (BK15 and
BK18) plus BAO bound. In other words, as shown in Fig. 4, the values of the inflationary
parameters ns and r reported in Ref. [43] are improved in non-canonical EMSG inflation,
making them compatible with current observational constraints [28].
As a final remark, let us discuss the prediction of the inflationary parameters in the
scale-independent EMSG (β = 1/2) under the slow-roll scheme. The dynamical background

– 17 –
 N = 70 Numerics Analytics
equil equil
|α|A6/7 ns r c2s |fNL | ns r c2s |fNL |
2.5 × 10−2 0.9629 0.0086 0.3332 0.5661 0.9629 0.0081 0.3332 0.5660
2.0 × 10−2 0.9736 0.0140 0.3331 0.5662 0.9731 0.0138 0.3331 0.5662
1.5 × 10−2 0.9776 0.0185 0.3331 0.5663 0.9767 0.0184 0.3331 0.5663
1.0 × 10−2 0.9790 0.0224 0.3331 0.5662 0.9781 0.0223 0.3331 0.5662
0 0.9797 0.0280 0.3333 0.5658 0.9787 0.0279 0.3333 0.5658
N = 60
2.5 × 10−2 0.9625 0.0118 0.3332 0.5662 0.9621 0.0114 0.3332 0.5661
2.0 × 10−2 0.9708 0.0179 0.3331 0.5663 0.9700 0.0177 0.3331 0.5663
1.5 × 10−2 0.9743 0.0226 0.3331 0.5663 0.9733 0.0224 0.3331 0.5663
1.0 × 10−2 0.9756 0.0267 0.3331 0.5663 0.9746 0.0265 0.3331 0.5663
0 0.9764 0.0326 0.3333 0.5658 0.9752 0.0325 0.3333 0.5658

Table 2. A comparison between analytical and numerical results for inflationary parameters for
inflationary parameters in the non-canonical EMSG model with δ = 2 by considering the potential,
2
V = Aϕ 3 .

equations and the sound speed squared are obtained as

3H 2 ≃ (1 + α)V , (5.20)
6δHX δ  α 
V ′ ϕ̇ ≃ − 1 − (δ − 2) , (5.21)
1+α 2
1 3αδ 3 (α + 1)δ δ−1 ϕ̇
 2δ
c2s ≃ + δ V , (5.22)
2δ − 1 6 (2δ − 1)2 (2 − α(δ − 2)) H
′ ′
ε ϕ̇
1 V ϕ̇
  ε˙H
εH = − , ηH = = H . (5.23)
2 V H HεH εH H
Note that if we choose δ = 2, the last term in the sound speed relation vanishes. Using the
first two equations above, we can derive the ratio ϕ̇/H as follows:

ϕ̇ h (1 + α)1−δ  V ′ i 1
2δ−1
= −K(δ) α δ
. (5.24)
H 1 − 2 (δ − 2) V

Hence, in the power law potential case V = Aϕn , it is straightforward to confirm that

n 1 h (1 + α)1−δ i 1  1
 2δ
2δ−1 2δ−1
N= V κ(δ,n)/n ; χ(δ, n) = K(δ) α nA n . (5.25)
κ(δ, n) χ(δ, n) 1 − 2 (δ − 2)

We can establish a relationship between the sound speed, slow-roll parameters, and N by
combining Eqs. (5.22), (5.23), and the equation above;

1 nα(1 + α)(−1)2δ δ 2 1
c2s = + 2 2
, (5.26)
2δ − 1 κ(α(δ − 1) − 2) (2δ − 1) N
δ 1 1
ϵH = , ηH = . (5.27)
2κ N N

– 18 –
 equil
Accordingly, the inflationary parameters, ns , r, and fNL read
1
ns − 1 = −I(δ, n) (5.28)
N
8n h (−1)2δ α(α + 1)δ 2 i
r ≈ 16εH cs = 1+ (5.29)
κ(2δ − 1)N κ(α(δ − 2) − 2)2 (1 − 2δ)N
equil 1 h  1  i
fNL = (275 − 590δ) 2 − 1 + 80δ(δ − 1) . (5.30)
972(2δ − 1) cs
As α approaches zero, these findings align with the values presented in [43, 58]. Based on
Eq. (5.26), the negative value of α results in smaller c2s values (less than 1/3). As a result,
equi
fNL is lower compared to the α = 0 scenario [58]. While other inflationary parameters,
such as ns and r, match well with those reported in [43].

6 Conclusions

In this paper, we have studied a single scalar field inflation within the framework of a
particular form of energy momentum squared gravity (EMSG ) with an extra piece, f (T2 ) =
−αT2β added to Einstein-Hilbert action. As stated in the introduction, the EMSG theory
has been investigated in different contexts. For example, in [10], a cosmological bouncing
solution was found for the specific cases of β = 1 and α > 0. Another instance is the
scale-independent EMSG (β = 1/2) with α = 2, which can reproduce the original steady
state universe in the presence of dust, as demonstrated in [19]. It is worth noting that our
analysis reveals that both models are subject to instabilities due to the positive sign of the
coupling parameter α.
We have shown that, in case T2 is constituted by a scalar field, the EMPG is equivalent
to a specific K-essence model. Subsequently, we have examined the tensor and scalar
perturbations of the metric around the FRW background in the presence of an inflaton
scalar field ϕ. We have shown that to circumvent the ghost and gradient instabilities,
the sound speed and the slow roll parameter εH must be positive. Consequently, we have
discussed the constraints on the free parameter α. Our findings indicate that, for any value
of β, the α parameter should be negative (α < 0). And, in the particular case β = 1/2,
i.e., the scale-independent EMSG, we must have −2 < α < 0. Then, we have examined the
slow roll inflation in EMPG using two different matter Lagrangians; the canonical scalar
field case (Lm = X − V (ϕ)), and the non-canonical scalar field case (Lm = X δ − V (ϕ))
with some specific forms of the field potential. We have shown the presence of EMPG
modifications allows us to bring the parameters r and ns in consistency with the recent
BICEP/Keck constraints on these parameters in both the canonical and non-canonical
scalar field cases. Furthermore, we have found that in the case of canonical Lagrangian,
equil
the fNL takes small values similar to those of standard chaotic inflation. In contrast, for
equil
the non-canonical Lagrangian case, the fNL turns out to be larger.
As a follow-up to the present investigations, it would be interesting to examine the
possibility of large fluctuations in the framework of EMSG and associated issues related to
primary and secondary GWS [59–70]. Unfortunately, the scope of PBH formation in this
case is limited by quantum loop effects [71–85].

– 19 –
 While our primary focus in this paper has been on inflation, it is worth mentioning in
closing that our specific K-essence model might also hold significance for the study in the
context of the late universe, in relevance with dark energy and/or cold dark matter. Such
behaviors in scalar fields have been previously documented in [86]. Conducting a dynamical
system analysis could offer valuable insights into diagnosing the model and identifying the
existence of phases dominated by dark energy and dark matter in the cosmic history of
this model, as explored in [87]. We defer these for future investigations.

Acknowledgments

The authors thank Hassan Firouzjahi, Shahab Shahidi, Zahra Haghani, Sayantan Choud-
hury, Mohamad Ali Gorji, Alireza Talebian, and Phongpichit Channuie for the useful com-
ments and discussions. SAH and FF acknowledge the partial support from the “Sara-
madan” federation of Iran. The research of MR is supported by the Ferdowsi University of
Mashhad. Ö.A. acknowledges the support by the Turkish Academy of Sciences in scheme
of the Outstanding Young Scientist Award (TÜBA-GEBİP). MS is supported by Science
and Engineering Research Board (SERB), DST, Government of India under the Grant
Agreement number CRG/2022/004120 (Core Research Grant). MS is also partially sup-
ported by the Ministry of Education and Science of the Republic of Kazakhstan, Grant
No. AP14870191 and CAS President’s International Fellowship Initiative(PIFI).

References

[1] C. Brans and R. H. Dicke, Mach’s Principle and a Relativistic Theory of Gravitation,
Physical Review 124 (Nov., 1961) 925–935.
[2] A. H. Chamseddine and V. Mukhanov, Mimetic Dark Matter, JHEP 11 (2013) 135,
[1308.5410].
[3] J. D. Bekenstein, Relativistic gravitation theory for the MOND paradigm, Phys. Rev. D 70
(2004) 083509, [astro-ph/0403694].
[4] J. W. Moffat, Scalar-tensor-vector gravity theory, JCAP 03 (2006) 004, [gr-qc/0506021].
[5] C. Skordis and T. Zlosnik, New Relativistic Theory for Modified Newtonian Dynamics,
Phys. Rev. Lett. 127 (2021) 161302, [2007.00082].
[6] T. P. Sotiriou and V. Faraoni, f(R) Theories Of Gravity, Rev. Mod. Phys. 82 (2010)
451–497, [0805.1726].
[7] F. W. Hehl and B. Mashhoon,
A Formal framework for a nonlocal generalization of Einstein’s theory of gravitation, Phys.
Rev. D 79 (2009) 064028, [0902.0560].
[8] T. Harko, F. S. N. Lobo, S. Nojiri and S. D. Odintsov, f (R, T ) gravity, Phys. Rev. D 84
(2011) 024020, [1104.2669].
[9] N. Katırcı and M. Kavuk,
f (R, Tµν T µν ) gravity and Cardassian-like expansion as one of its consequences, Eur. Phys. J.
Plus 129 (2014) 163, [1302.4300].

– 20 –
[10] M. Roshan and F. Shojai, Energy-Momentum Squared Gravity, Phys. Rev. D 94 (2016)
044002, [1607.06049].
[11] O. Akarsu, N. Katırcı and S. Kumar,
Cosmic acceleration in a dust only universe via energy-momentum powered gravity, Phys.
Rev. D 97 (2018) 024011, [1709.02367].
[12] C. V. R. Board and J. D. Barrow,
Cosmological Models in Energy-Momentum-Squared Gravity, Phys. Rev. D 96 (2017)
123517, [1709.09501].
[13] A. Ashtekar, T. Pawlowski and P. Singh,
Quantum Nature of the Big Bang: Improved dynamics, Phys. Rev. D 74 (2006) 084003,
[gr-qc/0607039].
[14] P. Brax and C. van de Bruck, Cosmology and brane worlds: A Review, Class. Quant. Grav.
20 (2003) R201–R232, [hep-th/0303095].
[15] A. H. Barbar, A. M. Awad and M. T. AlFiky,
Viability of bouncing cosmology in energy-momentum-squared gravity, Phys. Rev. D 101
(2020) 044058, [1911.00556].
[16] E. Nazari, F. Sarvi and M. Roshan,
Generalized Energy-Momentum-Squared Gravity in the Palatini Formalism, Phys. Rev. D
102 (2020) 064016, [2008.06681].
[17] S. Bahamonde, M. Marciu and P. Rudra,
Dynamical system analysis of generalized energy-momentum-squared gravity, Phys. Rev. D
100 (2019) 083511, [1906.00027].
[18] O. Akarsu, N. Katirci, S. Kumar, R. C. Nunes and M. Sami,
Cosmological implications of scale-independent energy-momentum squared gravity: Pseudo nonminimal interact
Phys. Rev. D 98 (2018) 063522, [1807.01588].
[19] O. Akarsu and N. M. Uzun,
Cosmological models in scale-independent energy-momentum squared gravity, Phys. Dark
Univ. 40 (2023) 101194, [2301.11204].
[20] E. Nazari, Light bending and gravitational lensing in energy-momentum-squared gravity,
Phys. Rev. D 105 (2022) 104026, [2204.11003].
[21] N. Nari and M. Roshan, Compact stars in Energy-Momentum Squared Gravity, Phys. Rev.
D 98 (2018) 024031, [1802.02399].
[22] A. Kazemi, M. Roshan, I. De Martino and M. De Laurentis,
Jeans analysis in energy–momentum-squared gravity, Eur. Phys. J. C 80 (2020) 150,
[2001.04702].
[23] E. Nazari, M. Roshan and I. De Martino,
Constraining energy-momentum-squared gravity by binary pulsar observations, Phys. Rev.
D 105 (2022) 044014, [2201.08578].
[24] O. Akarsu, E. Nazari and M. Roshan,
Relativistic binary systems in scale-independent energy-momentum squared gravity,
2302.04682.
[25] O. Akarsu, J. D. Barrow, S. Çıkıntoğlu, K. Y. Ekşi and N. Katırcı,

– 21 –
 Constraint on energy-momentum squared gravity from neutron stars and its cosmological implications,
Phys. Rev. D 97 (2018) 124017, [1802.02093].
[26] M. Faraji, N. Rashidi and K. Nozari,
Inflation in energy-momentum squared gravity in light of Planck2018, Eur. Phys. J. Plus
137 (2022) 593, [2107.13547].
[27] C. Ranjit, P. Rudra and S. Kundu,
Constraints on Energy–Momentum Squared Gravity from cosmic chronometers and Supernovae Type Ia data,
Annals Phys. 428 (2021) 168432, [2010.02753].
[28] BICEP, Keck collaboration, P. A. R. Ade et al.,
Improved Constraints on Primordial Gravitational Waves using Planck, WMAP, and BICEP/Keck Observations
Phys. Rev. Lett. 127 (2021) 151301, [2110.00483].
[29] A. D. Linde, Chaotic inflation, Physics Letters B 129 (1983) 177–181.
[30] A. D. Linde, A new inflationary universe scenario: a possible solution of the horizon,
flatness, homogeneity, isotropy and primordial monopole problems, Physics Letters B 108
(1982) 389–393.
[31] J. M. Maldacena,
Non-Gaussian features of primordial fluctuations in single field inflationary models, JHEP
05 (2003) 013, [astro-ph/0210603].
[32] J. M. Maldacena and G. L. Pimentel, On graviton non-Gaussianities during inflation, JHEP
09 (2011) 045, [1104.2846].
[33] S. Choudhury, CMB from EFT, Universe 5 (2019) 155, [1712.04766].
[34] S. Choudhury and S. Pal, Primordial non-Gaussian features from DBI Galileon inflation,
Eur. Phys. J. C 75 (2015) 241, [1210.4478].
[35] S. Choudhury,
Constraining N = 1 supergravity inflation with non-minimal Kaehler operators using δN formalism,
JHEP 04 (2014) 105, [1402.1251].
[36] M. Celoria and S. Matarrese, Primordial Non-Gaussianity, Proc. Int. Sch. Phys. Fermi 200
(2020) 179–215, [1812.08197].
[37] X. Chen, Primordial Non-Gaussianities from Inflation Models, Adv. Astron. 2010 (2010)
638979, [1002.1416].
[38] D. Baumann, Inflation, in
Theoretical Advanced Study Institute in Elementary Particle Physics:
Physics of the Large and the Small, pp. 523–686, 2011. 0907.5424. DOI.
[39] L. Senatore, Lectures on Inflation, in
Theoretical Advanced Study Institute in Elementary Particle Physics:
New Frontiers in Fields and Strings, pp. 447–543, 2017. 1609.00716. DOI.
[40] D. Baumann, Primordial Cosmology, PoS TASI2017 (2018) 009, [1807.03098].
[41] Planck collaboration, N. Aghanim et al., Planck 2018 results. VI. Cosmological parameters,
Astron. Astrophys. 641 (2020) A6, [1807.06209].
[42] C. Armendariz-Picon, V. F. Mukhanov and P. J. Steinhardt, Essentials of k essence, Phys.
Rev. D 63 (2001) 103510, [astro-ph/0006373].

– 22 –
[43] S. Li and A. R. Liddle, Observational constraints on K-inflation models, JCAP 10 (2012)
011, [1204.6214].
[44] C. V. R. Board and J. D. Barrow,
Cosmological Models in Energy-Momentum-Squared Gravity, Phys. Rev. D 96 (2017)
123517, [1709.09501].
[45] X. Chen, M.-x. Huang, S. Kachru and G. Shiu,
Observational signatures and non-Gaussianities of general single field inflation, JCAP 01
(2007) 002, [hep-th/0605045].
[46] O. Akarsu, M. Bouhmadi-López, N. Katırcı, E. Nazari, M. Roshan and N. M. Uzun,
Equivalence of matter-type modified gravity theories to general relativity with nonminimal matter interaction,
2306.11717.
[47] C.-Y. Chen and P. Chen,
Eikonal black hole ringings in generalized energy-momentum squared gravity, Phys. Rev. D
101 (2020) 064021, [1910.12262].
[48] D. Seery and J. E. Lidsey, Primordial non-Gaussianities in single field inflation, JCAP 06
(2005) 003, [astro-ph/0503692].
[49] F. Hoyle, A New Model for the Expanding Universe, Monthly Notices of the Royal
Astronomical Society 108 (1948) 108.
[50] B. Farsi, A. Sheykhi and M. Khodadi,
Growth of Perturbations in Energy-Momentum-Squared Gravity, 2304.01571.
[51] E. Babichev, V. Mukhanov and A. Vikman,
k-Essence, superluminal propagation, causality and emergent geometry, JHEP 02 (2008)
101, [0708.0561].
[52] J. Garriga and V. F. Mukhanov, Perturbations in k-inflation, Phys. Lett. B 458 (1999)
219–225, [hep-th/9904176].
[53] Planck collaboration, Y. Akrami et al., Planck 2018 results. X. Constraints on inflation,
Astron. Astrophys. 641 (2020) A10, [1807.06211].
[54] BICEP2, Keck Array collaboration, P. A. R. Ade et al.,
BICEP2 / Keck Array x: Constraints on Primordial Gravitational Waves using Planck, WMAP, and New BICE
Phys. Rev. Lett. 121 (2018) 221301, [1810.05216].
[55] S. Unnikrishnan, V. Sahni and A. Toporensky, Refining inflation using non-canonical scalars,
JCAP 08 (2012) 018, [1205.0786].
[56] J. M. Maldacena,
Non-Gaussian features of primordial fluctuations in single field inflationary models, JHEP
05 (2003) 013, [astro-ph/0210603].
[57] M. Alishahiha, E. Silverstein and D. Tong, DBI in the sky, Phys. Rev. D 70 (2004) 123505,
[hep-th/0404084].
[58] S. Unnikrishnan and V. Sahni,
Resurrecting power law inflation in the light of Planck results, JCAP 10 (2013) 063,
[1305.5260].
[59] J.-P. Li, S. Wang, Z.-C. Zhao and K. Kohri,

– 23 –
 Primordial Non-Gaussianity and Anisotropies in Gravitational Waves induced by Scalar Perturbations,
2305.19950.
[60] A. Bodas and R. Sundrum,
Large Primordial Fluctuations in Gravitational Waves from Phase Transitions, 2211.09301.
[61] C. Chen, A. Ota, H.-Y. Zhu and Y. Zhu,
Missing one-loop contributions in secondary gravitational waves, Phys. Rev. D 107 (2023)
083518, [2210.17176].
[62] M. Cicoli, F. G. Pedro and N. Pedron,
Secondary GWs and PBHs in string inflation: formation and detectability, JCAP 08 (2022)
030, [2203.00021].
[63] J. Lin, S. Gao, Y. Gong, Y. Lu, Z. Wang and F. Zhang,
Primordial black holes and scalar induced gravitational waves from Higgs inflation with noncanonical kinetic ter
Phys. Rev. D 107 (2023) 043517, [2111.01362].
[64] K. Rezazadeh, Z. Teimoori, S. Karimi and K. Karami,
Non-Gaussianity and secondary gravitational waves from primordial black holes production in α-attractor inflati
Eur. Phys. J. C 82 (2022) 758, [2110.01482].
[65] W. Ahmed, M. Junaid and U. Zubair,
Primordial black holes and gravitational waves in hybrid inflation with chaotic potentials,
Nucl. Phys. B 984 (2022) 115968, [2109.14838].
[66] L. Wu, Y. Gong and T. Li,
Primordial black holes and secondary gravitational waves from string inspired general no-scale supergravity,
Phys. Rev. D 104 (2021) 123544, [2105.07694].
[67] S. Choudhury and A. Mazumdar,
Primordial blackholes and gravitational waves for an inflection-point model of inflation,
Phys. Lett. B 733 (2014) 270–275, [1307.5119].
[68] K. Inomata, K. Kohri, T. Nakama and T. Terada,
Enhancement of Gravitational Waves Induced by Scalar Perturbations due to a Sudden Transition from an Early
Phys. Rev. D 100 (2019) 043532, [1904.12879].
[69] S. Wang, T. Terada and K. Kohri,
Prospective constraints on the primordial black hole abundance from the stochastic gravitational-wave backgrou
Phys. Rev. D 99 (2019) 103531, [1903.05924].
[70] L. Alabidi, K. Kohri, M. Sasaki and Y. Sendouda,
Observable Spectra of Induced Gravitational Waves from Inflation, JCAP 09 (2012) 017,
[1203.4663].
[71] J. Kristiano and J. Yokoyama,
Ruling Out Primordial Black Hole Formation From Single-Field Inflation, 2211.03395.
[72] A. Riotto,
The Primordial Black Hole Formation from Single-Field Inflation is Not Ruled Out,
2301.00599.
[73] S. Choudhury, M. R. Gangopadhyay and M. Sami,
No-go for the formation of heavy mass Primordial Black Holes in Single Field Inflation,
2301.10000.

– 24 –
[74] S. Choudhury, S. Panda and M. Sami,
No-go for PBH formation in EFT of single field inflation, 2302.05655.
[75] J. Kristiano and J. Yokoyama,
Response to criticism on ”Ruling Out Primordial Black Hole Formation From Single-Field Inflation”: A note on
2303.00341.
[76] A. Riotto,
The Primordial Black Hole Formation from Single-Field Inflation is Still Not Ruled Out,
2303.01727.
[77] S. Choudhury, S. Panda and M. Sami,
Quantum loop effects on the power spectrum and constraints on primordial black holes,
2303.06066.
[78] S. Choudhury, S. Panda and M. Sami,
Galileon inflation evades the no-go for PBH formation in the single-field framework,
2304.04065.
[79] H. Firouzjahi and A. Riotto, Primordial Black Holes and Loops in Single-Field Inflation,
2304.07801.
[80] R. Kawaguchi, T. Fujita and M. Sasaki,
Highly asymmetric probability distribution from a finite-width upward step during inflation,
2305.18140.
[81] S.-L. Cheng, D.-S. Lee and K.-W. Ng,
Primordial perturbations from ultra-slow-roll single-field inflation with quantum loop effects,
2305.16810.
[82] G. Tasinato, A large |η| approach to single field inflation, 2305.11568.
[83] G. Franciolini, A. Iovino, Junior., M. Taoso and A. Urbano,
One loop to rule them all: Perturbativity in the presence of ultra slow-roll dynamics,
2305.03491.
[84] H. Motohashi and Y. Tada,
Squeezed bispectrum and one-loop corrections in transient constant-roll inflation,
2303.16035.
[85] H. Firouzjahi, One-loop Corrections in Power Spectrum in Single Field Inflation,
2303.12025.
[86] N. Bose and A. S. Majumdar, Unified Model of k-Inflation, Dark Matter & Dark Energy,
Phys. Rev. D 80 (2009) 103508, [0907.2330].
[87] T. Kashfi and M. Roshan, Cosmological dynamics of relativistic MOND, JCAP 10 (2022)
029, [2204.05672].

– 25 –