Finding suitable f (R, ϕ) cosmologies

Sukanta Pandaa ,1 Trupti Patilb ,1 Rahul Thakurc ,1 and Archit Vidyarthid1

1
Department of Physics, Indian Institute of Science Education and Research Bhopal, Madhya Pradesh - 462066, India

We constrain f (R, ϕ) theories to find ones suitable for cosmological applications

I. INTRODUCTION

The universe we see today looks homogeneous, isotropic and flat at large scale(distance more than 200Mpc).
As there have been many random processes happened at early eras of the universe, so how is that possible for the
universe to look like that? All these answers were explained by Cosmological inflation theory. Inflation is a theoretical
paradigm which explains how our universe evolved to the point, we see today. It is a phenomena by which early small
patch of universe underwent a rapid exponential expansion becoming the large patch of the universe in its first few
moments. The expansion is believed to have occurred at a rate much faster than the speed of light, which solves
several fundamental problems of the standard Big Bang cosmology like flatness and horizon problem.

II. JORDAN FRAME ANALYSIS

We consider the following action in the Jordan frame[ref],

√
Z  
4 1 1
S= d x −g f (R, ϕ) − ∇µ ϕ∇µ ϕ − V (ϕ) (1)
2κ 2
We can derive corresponding Einstein field equations and the equation of motion for the scalar field to find,
 
1 1 α
F Rµν − f gµν − ∇µ ∇ν F + gµν □F = κ (∇µ ϕ∇ν ϕ − gµν ∇ ϕ∇α ϕ) − V gµν , (2)
2 2
 
1 f,ϕ
□ϕ + − 2V,ϕ = 0, (3)
2 κ2
∂f
where F = ∂R . We consider a spatially flat FLRW metric given by the line element,

ds2 = −dt2 + a2 (t)(dx2 + dy 2 + dz 2 ) (4)

where a(t) is the scale factor. Working in the unitary gauge, i.e. ϕ ≡ ϕ(t), 00 and ii components of the Einstein
equation (2) can be expressed respectively as,

6F (Ḣ + H 2 ) + (2κV − f ) − 6Ḟ H + κϕ̇2 = 0, (5)

−κϕ̇2 + 2F (Ḣ + H 2 ) + 4F H 2 + (2κV − f ) − 4H Ḟ − 2F̈ = 0, (6)

a sukanta@iiserb.ac.in
b trupti19@iiserb.ac.in
c rahul19@iiserb.ac.in
d archit17@iiserb.ac.in
 2

and equation of motion of ϕ in (3) becomes:

 
1 f,ϕ
ϕ̈ + 3H ϕ̇ − − 2V,ϕ = 0. (7)
2 κ

Using equations of motion (5), (6) and assuming only time dependence of fields, we get:

F̈ − H Ḟ + 2ḢF + κϕ̇2 = 0, (8)

F̈ + 5H Ḟ − 4ḢF + f − 6H 2 F = 2κV. (9)

As an example, assuming a potential V = λϕ4 we obtain the form of ϕ(t) as

Mp 2 
  41
2
ϕ(t) = 3H Ḟ + F̈ − 6H F + f (10)
2λ

Perturbation analysis for f (R, ϕ) has been performed extensively in several scenarios. The general equations to do the
same are provided in . The equations in their general form are quite cumbersome and require certain approximations
in order to arrive at exact analytical solutions. We omit them in this work since we’re working exclusively with
background behavior of f (R, ϕ) models.

A. Example

Assuming as an example a particular form of f (R, ϕ),

f (R, ϕ) = Reh(ϕ)R (11)

2 2
and further assuming a(t) = a0 e−t0 /t and h(ϕ)R =constant and F = 2e (essentially a rescaled Einstein-Hilbert
Lagrangian) for some constant t0 [1], we find,
 14
1 12eMp 2 2 2t20
 
ϕ= t0 −1 (12)
t λ t2

which is markedly different from [1] mostly because we’ve assumed ϕ as being dimensionful in our calculations. Using
the slow roll parameter ϵ = − HḢ2 , we can quantify the condition for inflation as ϵ < 1. When inflation ends,
r
tf 2 2
ϵf = 1 =⇒ Ḣ + H = 0 =⇒ = (13)
t0 3

Additionally, ϕ > 0 requires λ > 0 to be compatible with this condition (see (12). Number of e-folds can be calculated
from the time of horizon crossing ti to the end of inflation,
tf
t20
Z  
3
N= H dt = − (14)
ti t2i 2

q
2
For 60 e-folds, ti = t0 123 and
r  1/4
3Mp 24e
ϕf = (15)
2t0 λ
 3

As mentioned earlier, perturbations equations require certain simplifications to become analytically solvable. For the
current exponential inflationary scenario, this simplification is given by the condition F = 2e which renders all Ḟ and
F̈ as zero.
To demonstrate how results might appear for a general f (R, ϕ), we look at tensor perturbation equations here.
Perturbing the FLRW metric as follows,

(3)
ds2 = −(1 + 2θ)dt2 − a(β,α + Bα )dtdxα + a2 [gαβ (1 − 2ψ) + 2γ,α|β + 2Cα|β + 2Cαβ ]. (16)

we find that the tensor perturbations in Newtonian gauge follow:

!
α Ḟ k2
C̈β + + 3H + 2 Cβα = 0 (17)
F a

2 2
Considering the same exponential inflation case assumed earlier [1], i.e. using a = a0 e−t0 /t without assuming F = 2e,
we get
!
Ḟ 6t2 k2 α
C̈βα + + 30 + C =0 (18)
F t a2 β

Using the standard technique, we find the solution:

" Z #
1 −ikη 1 Ḟ
Cβα = √ e exp dt (19)
a 2k 2 F

For F = 2e and Ḟ = 0, the result matches that of [1].

B. Late-Time Cosmology

The allure of modified gravity and scalar-tensor theories (STTs) lies in their ability to explain a variety of phenomena
ranging from inflation to late-time acceleration. To demonstrate this, we shall try to find conditions for a general
f (R, ϕ) theory of the form (1) that could provide a stable dark energy candidate. The trace equation corresponding
to (2), without working in the unitary gauge, is:

3□F + F R − 2f = κT = −κ[∂µ ϕ∂ µ ϕ + 4V (ϕ)] (20)

where T represents trace of the energy-momentum tensor for ϕ and other non-gravitational components (that are no
longer decoupled from the system at late times). In general modified gravity theories, F can be considered as another
scalar DOF. This means, we can write,

∂Vef f 1
□F = = (κT + 2f − F R). (21)
∂F 3

Now, for the existence of a local extremum,

∂Vef f 1
= (κT + 2f − F R) = 0,
∂F 3
1
=⇒ R = (2f + κT ). (22)
F
 4

Assuming a de Sitter expansion at late time, R ≈ 12H 2 > 0 which implies,

2f + κT > 0. (23)

For this extremum to be a local minimum, i.e. for a stable late-time cosmic expansion, we require

∂ 2 Vef f
 
1 ∂f
= 2 − R >0
∂F 2 3 ∂F
 −1  −1
∂f ∂F ∂F f
=⇒ +F − >0 (24)
∂ϕ ∂ϕ ∂R R
∂F ∂f
which is true only if ∂ϕ , ∂R ̸= 0.
This condition can be further simplified assuming that f (R, ϕ) is factorizable as f = p(R)q(ϕ), for some functions
p(R) and q(ϕ). To do so, we restate below a few properties of f (R, ϕ) and their consequences for p(R) and q(ϕ):

• f (R, ϕ) has 2 mass dimensions (see (1)). Consequently, p(R) also has 2 mass dimensions while q(ϕ) is dimen-
sionless.

• f (R, ϕ) must be analytic in both R and ϕ. This is required because as ϕ → 0, f (R, ϕ) must become some
function of R. R → 0 must also be a well defined limit in order to explain transitions between various epochs
of the universe. In factorizable scenarios, analyticity of p(R) and q(ϕ) implies that as ϕ → 0, q(ϕ) returns some
finite value, i.e. there exists a non-zero minimum of q.

• In weak gravity regimes, f (R, ϕ) must reduce to the Einstein-Hilbert action coupled with some function of ϕ,
i.e. p(R) → R.

The condition (24) can then be expressed as,

pR
>0 (25)
pRR

where the subscript R represents a differentiation w.r.t. R. Also, (25) is valid as long as pRR ̸= 0. In the Einstein
frame, late-time cosmological analyses would obviously depend on the behavior of the transformed, non-gravitational
part of the action which would be highly dependent on the form of f (R, ϕ) considered in the Jordan frame. Therefore,
for the Einstein frame, we shall omit the late-time analysis in both metric and Palatini formalisms and instead carry
it out for some suitable forms of f (R, ϕ) that could give us favorable cosmologies. This analysis is carried out in
Section

III. EINSTEIN FRAME: METRIC FORMALISM

In the Jordan frame, both metric and Palatini formalisms are expected to provide similar results (excluding pertur-
bation analyses). The differences show up in the Einstein frame since R undergoes a Weyl transformation differently
in metric and Palatini formalisms.
F
We first introduce to the action (1), the term 2κ (R − χ2 ), and then we perform a Weyl transformation gµν → F gµν

4 √
Z   
R 3 µ 1 µ 1 1 2
S = d x −g − ∂µ F ∂ F − ∂ µ ϕ∂ ϕ − 2 (χ F − f ) + V (26)
2κ 2κF 2 2F F 2κ
 5

where F ≡ F (χ, ϕ). The corresponding equations of motion are:

  
1 3 1 3 α 1 α 1 1 2
Gµν = ∂µ F ∂ν F + ∂ µ ϕ∂ ν ϕ − gµν ∂α F ∂ F + ∂α ϕ∂ ϕ + (χ F − f ) + V , (27)
2κ κF 2 F 2κF 2 2F F 2 2κ
with the 00 component,
 
−3 2 κ 2 2κ 1 2
Ḟ − ϕ̇ − 2 (χ F − f ) + V = −3H 2 (28)
F2 F F 2κ
and ii components,
 
3 2 κ 2 2κ 1 2
Ḟ + ϕ̇ − 2 (χ F − f ) + V = −2Ḣ − 3H 2 (29)
F2 F F 2κ
These equations can be used to perform perturbation and late-time analyses similar to the Jordan frame. But instead
of these cumbersome equations, we can instead work with their canonicalized versions which greatly simplify the
calculations and localize the modifications compared to standard scalar-tensor theories (STTs) to certain terms,
leaving other intact. Canonicalization is also a necessary step before proceeding with any quantum field theoretical
analyses like finding unitarity violation scales or checking renormalizability of the theory [].

A. Canonicalization

Proceeding with canonicalization in a traditional manner,

∂F ∂F
∂µ F = ∂µ ϕ + ∂µ χ (30)
∂ϕ ∂χ
which gives
  −1
∂F ∂F
∂µ ϕ = ∂µ F − ∂µ χ (31)
∂χ ∂ϕ
" 2  2   #
3 3 ∂F ∂F ∂F ∂F
∂µ F ∂ µ F = µ
∂ µ ϕ∂ ϕ + µ
∂ µ χ∂ χ + 2 · ∂ µ ϕ∂ χ µ
(32)
κF 2 κF 2 ∂ϕ ∂χ ∂ϕ ∂χ

Now,
Fϕ Fχ
∂µ ln F = ∂µ ϕ + ∂µ χ (33)
F F
 2  2  
µ Fϕ µ Fχ µ Fχ Fϕ
∂µ (ln F )∂ (ln F ) = ∂ µ ϕ∂ ϕ + ∂ µ χ∂ χ + 2 ∂ µ ϕ∂ µ χ (34)
F F F2
q
2
let, κ ln F = θ
Then,
 r  r
1 µ κ µ µ κ κ
2
∂µ ϕ∂ ϕ = exp − θ ∂µ ϕ∂ ϕ = ∂µ ϕ∂ ϕ − θ∂µ ϕ∂ µ ϕ + θ2 ∂µ ϕ∂ µ ϕ + ... (35)
F 3 3 6
which gives an infinite series of terms that can be truncated based on the scale of F and used accordingly to check
renormalizability and unitarity violation scales. But this method leaves the fate of the term (χ2 F − f ) uncertain. The
canonicalization problem is better resolved assuming f (χ, ϕ) is factorizable, i.e. f (χ, ϕ) = p(χ)q(ϕ). In that case,
"  #
2
p χ2 χ 2
    
3 µ 3 qϕ µ µ p χ2 χ qϕ µ
∂µ F ∂ F = ∂ µ ϕ∂ ϕ + ∂ µ χ∂ χ + 2 ∂ µ χ∂ ϕ (36)
κF 2 κ q p χ2 p χ2 q
 6

where the subscripts represent variable with respect to which differentiation is being done. Now, considering,
r
3 p χ2 χ
∂µ χ = ∂µ Ω (37)
κ p χ2
or
  r 
3 κ
∂µ ln(pχ2 ) = ∂µ Ω =⇒ pχ2 = exp Ω (38)
κ 3
Similarly,
κ Ω2
 r  r
1 κ 1 1 κ1
∂µ ϕ∂ µ ϕ = exp − Ω ∂µ ϕ∂ µ ϕ = ∂µ ϕ∂ µ ϕ − Ω∂µ ϕ∂ µ ϕ + ∂µ ϕ∂ µ ϕ. (39)
pχ2 q 3 q q 3q 3 q
Then, the total kinetic term of ϕ is,
r !
3 qϕ 1
+√ ∂µ ϕ = ∂µ θ (40)
κ q q

Canonicalization is now complete. But the term (χ2 F − f ) still needs to be expressed in terms of canonicalized fields
Ω and θ. In factorized form,

χ2 F − f = (χ2 pχ2 − p)q. (41)

Now, (χ2 F − f ) needs to be an explicit function of only pχ2 so that the term is factorizable as separate functions of
χ and ϕ. With this condition, the system could be simplified and expressed in the form of canonicalized fields. The
forthcoming analysis is simply to discuss the forms of p(R) in the Jordan frame action that can be canonicalized in
a straightforward manner to serve as reference for future work in this direction. From this point forward, we shall
be treating p ≡ p(R) simply for typographical clarity. The valid forms of p(R) obtained at the end can simply be
rewritten as p ≡ p(χ) to obtain an Einstein frame action in terms of canonicalized fields using transformation of pχ2
mentioned earlier in (38).
Assuming, p(R) = R u(R), where u(R) is such that for small R, u(R) = constant (this is in line with our assumptions
regarding the behavior of p(R) in low gravity regimes). Then,
dp du
pR = = u(R) + R . (42)
dR dR
1
It is non homogeneous differential equation with a complementary solution u(R) = R. The factor we need to take
2
care of is (χ F − f ), which implies,
du du
RpR − p = R u(R) + R2 − Ru(R) = R2 (43)
dR dR
Since p(R) is assumed to be an analytic function of R, we can consider three general forms of pR ,

• Case 1: When pR is polynomial in R, i.e. pR = a + bR + cR3 + ..., we find,

m
u(R) = + ApR + B (44)
R
for arbitrary constants m, A, and B. Then,
du dp
RpR − p = R2 = −m + A = −m + abR2 + 2AcR3 + .... (45)
dR dR
This case encompasses some important modified gravity models like the Starobinsky inflation model.
 7

• Case 2: When pR is exponential in R, i.e. pR = aebR , we have,

m
u(R) = + AebR (46)
R

du Ab
=⇒ RpR − p = R2 = −m + AbebR = −m + pR . (47)
dR a

• Case 3: When pR = (a + bR + cR2 + dR3 + ....)eλR , i.e. a combination of polynomial and exponential functions,
m
u(R) = + (A + BR + CR2 + ....)eλR (48)
R

du
=⇒ RpR − p = R2 = −m + (Aλ + B)R2 + (Bλ + 2C)R3 + (Cλ + 3D)R4 + ... eλR
 
(49)
dR

For the aforementioned forms of pR , and consequently p(R), we can formulate an f (R, ϕ) theory that can be canonical-
ized in a well-defined manner. These conditions are expected to be a tad different when working in Palatini formalism,
however, as we shall see in the next section.

IV. EINSTEIN FRAME: PALATINI FORMALISM

Following the same Weyl transformation as done in the metric formalism in the previous section, we arrive at the
Einstein frame action as follows:
√
Z   
R 1 1 1 2
S = d4 x −g − ∂ µ ϕ∂ µ ϕ − 2 (χ F − f ) + V (50)
2κ 2F F 2κ

It is clear that χ in this case is a non-dynamical field. So, varying the action with respect to χ2 , we can obtain a
constraint equation for the system,
      
1 ∂F µ 2 ∂F 1 2 1 2 ∂F
∂µ ϕ∂ ϕ + 3 (χ F − f ) + V − F +χ −F =0 (51)
2F 2 ∂χ2 F ∂χ2 2κ 2κF 2 ∂χ2
which simplified to give,

κF ∂µ ϕ∂ µ + F χ2 − 2f + 4κV = 0. (52)

This equation can be used to eliminate the non-dynamical DOF χ from (50), which is necessary to proceed with any
physical analysis in this system [2].

A. Canonicalization

Canonicalizing ϕ in (50) is straightforward. Now, following the same protocol as the metric formalism, we’ll first
try to find valid forms of p for which the the constraint equation can be expressed in a form that could eliminate
F (χ, ϕ) and instead write it only in terms of ϕ. Taking cues from the previous section, it is simply easier to find
conditions for which the non-dynamical DOF can be eliminated from the system if we rewrite the constraint equation
(52) in terms of by factorizing f (χ, ϕ) as f (χ, ϕ) = p(χ)q(ϕ) and rewriting p ≡ p(R) for typographical ease, we have,

κpR q∂µ ϕ∂ µ ϕ + χ2 pR q − 2pq + 4κV = 0 (53)

 8

4κV
=⇒ χ2 pR − 2p = − − κpR ∂µ ϕ∂ µ ϕ (54)
q

Assuming p(R) = R u(R) for the same reasons as earlier, we find,

du
pR = u(R) + R (55)
dR

This equation has the same complimentary solution as in metric formalism. The terms we need to manage are slightly
different, however. From the LHS of (54),

du
RpR − 2p = R2 − uR (56)
dR

• Case 1: When pR is polynomial in R as pR = a + bR + cR3 + .., the complete solution has the form,

m
u(R) = + APR + B (57)
R

Substituting in (56), we get,

h m i
RpR − 2p = R2 − 2 + A(b + 2cR + ...) − m − AR(a + bR + cR3 + ...) − BR (58)
R

=⇒ RpR − 2p = −2m − (aA + B)R + acR3 + ... (59)

It is evident from this form that only by considering p = R+αR2 , i.e. a Starobinsky-like form, we can completely
eliminate R, or equivalently χ from the Einstein frame action (50). Substituting this in (54) and manipulating
the expression, we find,

4κV + κq∂µ ϕ∂ µ ϕ
χ2 = (60)
q(1 − 2κα∂µ ϕ∂ µ ϕ)

q + 8καV
=⇒ pR q = (61)
1 − 2κα∂µ ϕ∂ µ ϕ

• Case 2: When pR is an exponential function of R, pR = aeλR for which,

m
u= + AeλR (62)
R

Following the same procedure as the previous case, we arrive at the following equation:
h  p i A  p  4κV
R R
−2m + 1 + ln ln pR + κpR ∂µ ϕ∂ µ ϕ = − (63)
a a a q

Which gives no straightforward solution except the trivial case where λ = 0, i.e. pR = a or p(R) = aR. For
this, we find the relation
 
a 2κV
pR = 2m − . (64)
A q

Clearly, exponential models where pR = aeλR are not favoured in Palatini formalism when working in the
Einstein frame.
 9

Since we couldn’t find any straightforward non-trivial solution for the exponential case, we omit the third analytic
scenario where pR is a combination of polynomial and exponential functions. It appears that a Starobinsky-like model
is a special case for which one can safely remove the non-dynamical DOF in the Einstein frame. Rewriting the action
accordingly,

√
Z  
4 R 1 κα V
S= d x −g − ∂µ ϕ∂ µ ϕ + (∂ µ ϕ∂ µ ϕ)(∂ ν ϕ∂ ν ϕ) − (65)
2κ 2(q + 8καV ) 2(q + 8καV ) q(q + 8καV )
which can be canonicalized using the transformation,
dϕ p
= ± q + 8καV (66)
dφ
to give,
√
Z  
R 1 κα(q + 8καV )
S= d4 x −g − ∂µ φ∂ µ φ + (∂ µ φ∂ µ φ)(∂ ν φ∂ ν φ) − U , (67)
2κ 2 2
where
V
U= , (68)
q(q + 8καV )
and q ≡ q(φ), V ≡ V (φ) can be obtained using the transformation in (66). Compared to metric formalism, where a
wider variety of f (R, ϕ) models could give favourable conditions, Palatini formalism demands a specific form of the
gravitational part of the action in the Einstein frame. Now that we’ve isolated this form, we shall check whether it
support a proper cosmological evolution from early- to late-times through dynamical analysis in the next section.

V. DYNAMICAL ANALYSIS FOR PALATINI f (R, ϕ) MODELS

So far, we have left potential V and the factorized coupling function q arbitrary. We shall continue to do so in
this section as well and check if we can find some constraints ensuring proper cosmological evolution that can help us
narrow down specific forms for both V and q.
As mentioned in Section II B, as φ → 0 we expect to recover the pure gravitational action from the f (R, ϕ) term.
Based on this condition, we make an assumption that we can use Planck data to constrain the constant coefficient α
of the R2 (as defined in Section IV A), i.e.
α
≥ 3 × 109 . (69)
2κ
It is clear from this constraint that κα is vanishingly small. Also, note that the dynamical analysis below is performed
after the end of inflationary era. To this end, we expect that V is no longer a large enough quantity and use the
following condition to simplify the rest of the analysis without loss of generality:
8καV
≪ 1. (70)
q
We can, then, rewrite the action as,

4 √
Z  
R 1 µ καq µ ν
S = d x −g − ∂µ φ∂ φ + (∂ µ φ∂ φ)(∂ ν φ∂ φ) − U , (71)
2κ 2 2
 10

where

V
U= , (72)
q2

First, we write down Friedmann equations for the action (71) in an FLRW background (4):

κ φ̇2 3καq φ̇4

 
2
H = + +U , (73)
3 2 2
 2 
φ̇ 4
Ḣ = −κ + 2καq φ̇ . (74)
2

And, the equation of motion for φ is found to be,

3
φ̈ + 3H φ̇(1 − 4καq φ̇2 ) + καqφ φ̇4 + Uφ (1 − 6καq φ̇2 ) = 0. (75)
2

We now define dimensionless variables,

κφ̇2 κ2 αq φ̇4 κU
x2 ≡ , y2 ≡ , z2 ≡ , (76)
6H 2 2H 2 3H 2

such that (73) gives us the constraint equation:

x2 + y 2 + z 2 = 1 (77)

From the form of this equation, we also have the constraints 0 ≤ x2 , y 2 , z 2 ≤ 1. We also find expressions for energy
density ρ and pressure P as:

φ̇2 3καq φ̇4

ρ= + +U (78)
2 2
2 4
φ̇ καq φ̇
P = + − U. (79)
2 2

using which, we can express the equation of state parameter w = P/ρ as,

x2 + y 2 − z 2
w= . (80)
x2 + 3y 2 + z 2

At this stage, we define some more dimensionless parameters,

qφ2 Uφ2
β2 ≡ , γ2 ≡ , θ2 ≡ 2αqH 2 , (81)
κ κU 2

such that,

y = 3θx2 , (82)
p
x2 = yz 6καU q. (83)

Before moving forward, we would like to repeat the characteristics of q mentioned in Section II B to ensure that the
given theory is well-defined: q is dimensionless, and as φ → 0, q approaches some non-zero minimum value. Using
these conditions and (82), we need constraints such as,

y y x2
lim = some constant, lim = some constant, lim = some constant (84)
y→0 x2 y→0 θ x→0 z
x→0 θ→0 z→0
 11

Here, the first limit ensures that q ̸= 0 for all φ. In the second limit, the condition θ → 0 simply means that its
magnitude is quite small, but never zero. Now, we can safely use these dimensionless parameters to express our
dynamical system as follows:
r r
′ 2 2 3 2 4 3 2
x = −3x(1 − 12θ x ) − 9 βθ x − γz (1 − 18θ2 x2 ) + 3x3 + 18θ2 x5 , (85)
2 2
" r ! #
3 β √ √
y ′ = 3θ + 6 x3 − 3x4 + 2x2 y 2 − 6x2 − 6βxy 2 − 6γxz 2 , (86)
2q
√
z ′ = z( 6γx + 3x2 + 2y 2 ), (87)
√
′ 6
β = qφφ x, (88)
√κ  
′ 6 Uφφ 2
γ = x − κγ , (89)
κ U
r !
′ 3x 2 2
θ =θ − 3x − 2y (90)
2q

where ′ signifies a derivative with respect to the number of e-folds N which is related to time as dN = Hdt. For this
set of equations, we can find at least four fixed points that correspond to four stages in cosmological evolution:

1. x2 = y 2 = 0; z 2 = 1.
At this point, w = −1 and Ḣ = 0 =⇒ H = H0 (constant), i.e. scale factor a = a0 e±H0 t . Sticking to spacetime
expansion scenarios, we choose the a = a0 eH0 t which corresponds to a late-time de Sitter phase (since we’ve
chosen to analyse the dynamics of this system in the post-inflation era).

At this stage, since x2 = 0 =⇒ φ̇ = 0. Then,

z 2 = 1 =⇒ U = 3H02 MP2 , (91)

which is a constant value.

2. x2 = z 2 = 0; y 2 = 1.
At this point, w = 1/3 and Ḣ = −2H 2 =⇒ H = 1/2t or a = a0 t1/2 which corresponds to the radiation
dominated epoch. The phase of domination of this term is called the hyperkination era. In the presence of
relativistic matter, hyperkination is thus expected to cause some variations compared to standard cosmology.
Also,

M2 t
y 2 = 1 =⇒ φ̇ = √ P . (92)
2αq

To determine the exact form of φ with respect to time, we need q(φ).

y
3. y 2 = z 2 = 0; x2 = 1, which is only possible if θ2 → 0 such that lim y→0 θ = 3.
θ→0
Here, w = 1 and Ḣ = −3H 2 =⇒ H = 1/3t or a = a0 t1/3 . This era is called the standard kination epoch and
is expected to chronologically follow the hyperkination era. At this point,
√
x2 = 1 =⇒ φ̇ = ± 6HMP . (93)
 12

We choose the negative sign because the magnitude of φ is expected to fall in late-times. Then,
r  
2 t0
φ= MP ln , (94)
3 t
where t0 is a constant of integration.
q
4. x2 = 32 καU q ≈ 0; y 2 ≈ 12 = z 2 .
In this case, w = 0 and Ḣ = − 32 H 2 =⇒ H = 2/3t and a = a0 t3/2 . This corresponds to the matter
domination era. In the presence of non-relativistic matter, this condition may cause variations compared to
standard cosmology. Now,
r
2 1 2 MP
y = =⇒ φ̇ = ± p√ , (95)
2 3 t αq
q
3
which can only be solved to obtain φ once we fix a form for q(φ). However, using x2 = 2 καU q, one can
substitute the expression for φ̇ above and obtain the relation,
t2
αq = (96)
8
which can be substituted back to find,
2(2)1/4
 
t0
φ = √ MP ln (97)
3 t
where we have chosen the negative solution for φ̇ for decreasing φ and t0 is an integration constant. It is evident
that since φ cannot assume negative values, this phase lasts until t < t0 . Also,
1 8MP2
z2 = =⇒ U = . (98)
2 3t2
t can be eliminated from this expression to get,
 2 √
8 MP 3 φ
U= e (2)1/4 MP . (99)
3 t
Assuming that this matter domination phase is followed by the dark energy domination phase in point 1, as is
the case in standard cosmological evolution, this potential must asymptotically approach the potential in (91).
We can use the φ ≪ MP limit to find t0 as,
√
2 2
t0 ≈ . (100)
3H0

A. Stability Analysis

VI. CONCLUSION

[1] Joseph P. Johnson, Jose Mathew, and S. Shankaranarayanan. Exact inflationary solutions in exponential gravity. Gen. Rel.
Grav., 51(3):45, 2019. doi:10.1007/s10714-019-2531-4.
[2] Antonio De Felice and Shinji Tsujikawa. f(R) theories. Living Rev. Rel., 13:3, 2010. doi:10.12942/lrr-2010-3.