Gauge invariant perturbations in teleparallel Horndeski gravity

Bobomurat Ahmedov∗
New Uzbekistan University, Movarounnahr str. 1, Tashkent 100000, Uzbekistan and

d
Institute of Fundamental and Applied Research, National Research
University TIIAME, Kori Niyoziy 39, Tashkent 100000, Uzbekistan

we
Maria Caruana† and Jackson Levi Said‡
Institute of Space Sciences and Astronomy, University of Malta, Malta, MSD 2080 and
Department of Physics, University of Malta, Malta

Konstantinos F. Dialektopoulos§
Department of Mathematics and Computer Science,

ie
Transilvania University of Brasov, 500091, Brasov, Romania

Abdurakhmon Nosirov¶
Center for Astronomy and Astrophysics, Center for Field Theory and Particle

ev
Physics and Department of Physics, Fudan University, 200438 Shanghai, China and
Ulugh Beg Astronomical Institute, Astronomy St. 33, Tashkent 100052, Uzbekistan

Zinovia Oikonomopoulou∗∗
Institute of Space Sciences and Astronomy, University of Malta, Malta, MSD 2080

r
Odil Yunusov††
National Research University TIIAME, Kori Niyoziy 39, Tashkent 100000, Uzbekistan and
Shahrisabz State Pedagogical Institute, Shahrisabz Str. 10, Shahrisabz 181301, Uzbekistan
er
(Dated: December 6, 2024)
We present in the form of a catalogue the cosmological perturbations within the Bahamonde-
Dialektopoulos-Levi Said (BDLS) theory, which serves as the teleparallel counterpart of Horndeski
gravity. To understand structure formation in cosmological models, it is essential to study both the
pe
background and perturbative aspects of their cosmology. While extensive analysis of both Horndeski
gravity and its teleparallel analog exists in the literature, a quantitative understanding requires a
detailed examination of their cosmological perturbations. We review here all the different gauges
for the scalar, vector and tensor perturbations of a cosmological background up to second order and
we hope this will help people who work with observations, to incorporate it in existing codes.
ot

I. INTRODUCTION

For several deacdes, the ΛCDM model understood to be the best way to unify cosmological observations at different
tn

scales as well as to accurately describe astrophysical phenomena [38, 72]. For this setting, the cosmological constant (Λ)
drives a late-time accelerated expansion [73, 74] while gravitational interactions are expressed through general relativity
(GR) whereas cold dark matter (CDM) plays a key role in the primordial formation of the seeds of large-scale structure
formation in the early Universe and as a stabilizing agent in galactic structures in the late-Universe [22, 29]. The ΛCDM
model poses many foundational questions such as the fine-tuning of the cosmological constant [77], but many others.
Putting aside these important concerns, the concordance model description of the Universe is increasingly coming
rin

into tension when observational surveys from the early and late Universe are contrasted [40–42, 75]. Simultaneously,
the prospect of direct observations of CDM appear to be less probably as further measurements continue not report
ep

∗ Electronic address: [email protected]

† Electronic address: [email protected]
‡ Electronic address: [email protected]
§ Electronic address: [email protected]
¶ Electronic address: [email protected]
∗∗ Electronic address: [email protected]
†† Electronic address: [email protected]
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 2

any direct detections [5, 52].

A large spectrum of competing modifications to the ΛCDM model have been proposed in the literature together with
the search for different physical phenomena that may resolve the most pressing open challenges to the concordance
model. The vast battery of alterations of the concordance model include new proposals for the behaviour of CDM [1,

d
44, 48, 60], adding dynamical features to dark energy [21, 25, 38, 76], as well as extensions to the GR description of
gravity [3, 6, 9, 14, 34, 37], among many others. By and large, these models consist of additional layers of complexity
on top of the baseline ΛCDM model. Another approach is to reconsider the foundations of ΛCDM. Teleparallel gravity

we
(TG) is grounded in the exchange of the curvature associated with the Levi-Civita connection with geometric torsion
as the description of gravitational interactions [14, 32, 64]. This has some advantages in foundational aspects of the
theory such as having a well defined gravitational energy momentum tensor [8], among others, and may also offer a
vehicle to meet the growing body of observational challenges.
The novel architecture of teleparallel gravity is curvature-free, satisfies metricity, and features a formulation which
is dynamically equivalent to GR called the teleparallel equivalent of general relativity (TEGR) [7, 69]. TEGR differs

ie
from GR by a boundary term which plays an important role in modifications of the model, and may also produce
new IR limit possible realizations. As in regular curvature-based formulations of gravity, TEGR can be modified in
different ways to expose various new physics. The most direct modification of TEGR is known as f (T ) gravity [18–
20, 24, 36, 47, 49, 50, 66, 71], and is a direct generalization of the so-called torsion scalar T to an arbitrary functional of

ev
this variable. However, unlike most generic modifications of the gravitational sector, f (T ) gravity produces generally
second order field equations, and agrees with a widening spectrum of observational phenomena. Building on these
toy models, there has also been growing interest in exploring possible scalar-tensor generalizations of the underlying
theory in an analogous fashion to Horndeski gravity [57].
Horndeski gravity encompasses the largest class of second order scalar-tensor models in which only one scalar field

r
is adopted and geometric curvature is retained. However, the constraints brought about by the observation of the
gravitational wave event GW170817 [2] and its electromagnetic counterpart GRB170817A [53] has placed severely
limiting constraints its most exotic branches of models which was a focal point within the literature [61]. Building on
er
this background, a teleparallel gravity formulation of the scalar-tensor Horndeski gravity was proposed in Ref. [17].
The speed of gravitational wave constraint was revisited in Ref. [15] where it was found to be circumvented as a
limiting factor in the construction of models within the class of theories, where as in Ref. [13] the gravitational wave
polarization modes were determined for various subclasses of models. The post-Newtonian parametrization framework
was investigated in Ref. [16] where the standard tests were found to be observed for most for the majority of functional
pe
models of the theory. While efforts to build models motivated by Noether symmetries were explored in Ref. [43] which
also contains the full classification of these models. Other features have been probed such as the well-temperating
of the class of theories [27, 28] as well as conditions on the stability of the system [33] and reconstructed classes of
models [26].
For models within the class of theories contained in the teleparallel analogue of Horndeski gravity to more robustly
be investigated against the latest observational measurements, its cosmological perturbations must be determined.
This is also important to fully understand the behavior of these models both in the early Universe as well as through
ot

its evolution into the late Universe. This was first investigated in Ref. [4] where the full calculation is undertaken for
a particular gauge choice. Here, the primordial power spectrum and the alpha parametrization of the perturbations is
described in detail. In the present work, we explore the gauge-invariant formulation of the cosmological perturbations
which is important for exploring different phenomena which have their own natural gauge choices in further works.
tn

To facilitate these expressions of the different perturbation equations of motion, we also show the different gauge
scenarios for the gauge-invariant cosmological perturbation equations. In Sec. II, the BDLS technical details are
discussed while the cosmological perturbations are presented in Sec. III. The expressions are analysed for potential
ghost and stability conditions in Sec. IV. The main results are summarized and discussed in Sec. V. Units in which
c = 1 are assumed unless otherwise stated.
rin

II. BDLS GRAVITY: THE TELEPARALLEL ANALOGUE OF HORNDESKI THEORY

To begin, we will provide an overview of teleparallel gravity, including its foundational principles and the dynamics
that govern its behavior in a cosmological background. This introduction will offer insight into the core concepts
ep

of teleparallel gravity and set the stage for understanding its role in cosmological evolution. We will also discuss
the basic equations and key assumptions underlying the cosmological dynamics within this framework, establishing a
basis for exploring perturbations and structure formation in teleparallel theories.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 3

A. Teleparallel Gravity

Gravity theories based on curvature, like General Relativity (GR), are built upon a geometric framework where
◦
the Levi-Civitat connection, Γσµν (with over-circles indicating quantities derived from this connection), serves as the

d
foundation for the theory’s core geometric structures, such as the Riemann tensor. Consequently, curvature-based
gravity models rely heavily on the Levi-Civita connection, as seen in the formulation of the Einstein-Hilbert action

we
via the Ricci scalar. Teleparallel Gravity (TG), however, presents an alternative framework, where the curvature-
dependent connection is replaced by the torsion-based connection, Γσµν [8, 14, 32, 64].
Practically speaking, curvature-based and torsion-based theories of gravity differ significantly in their mathematical
frameworks. Curvature-based theories like GR utilize the metric tensor gµν and its derivatives, whereas TG is
formulated using the tetrad eAµ , which defines the system’s gravitational variables, along with a flat spin connection
ω BCν . Here, greek indices represent coordinates on the general manifold, while Latin ones refer to the local Minkowski
spacetime. Although, the tetrad and spin connection also appear in GR, their roles are more complex and less practical;

ie
in TG, the spin connection serves as an inertial object. The tetrad directly related to the metric tensor by

gµν = eAµ eB ν ηAB and ηAB = EA µ EB ν gµν , (1)

ev
where EA µ represents the inverse tetrad. This relationship shows the flexibility in choosing tetrad components, with
the spin connection preserving diffeomorphism invariance across these choices.
The tetrad-spin connection pair defines the potential components for a given spacetime in TG, enabling the telepar-
allel connection to be expressed as [32, 64]

r
Γλ νµ = EA λ ∂µ eAν + EA λ ω ABµ eB ν , (2)

where the flatness of the spin connection is ensured by the condition [14]
er
∂[µ ω A|B|ν] + ω AC[µ ω C|B|ν] ≡ 0 . (3)

There are also specific frames for any spacetime where all spin connection terms vanish for certain tetrad choices. This
configuration is known as the Weitzenböck gauge [78] and is consistently applied in cases where the spin connection
pe
field equations vanish identically for these particular tetrad selections.
In TG, gravitational scalars are constructed by substituting the Levi-Civita connection with its teleparallel coun-
terpart. As a result, the Riemann tensor vanishes identically, Rαβγϵ (Γσµν ) ≡ 0, even though the standard Riemann
◦ ◦
tensor based on the Levi-Civita connection remains nonzero, Rαβγϵ (Γσµν ) ̸= 0. To proceed, we introduce a torsion
tensor that relies only on the teleparallel connection, defined as [8, 70]
ot

T Aµν := 2ΓA[νµ] , (4)

where the square brackets indicate antisymmetrization. This torsion tensor, which serves as the field strength in the
theory [14] is invariant under both local Lorentz transformations and diffeomorphisms [65].
The torsion tensor in TG can be decomposed into three fundamental irreducible components [10, 11, 55]
tn

1
aµ := ϵµνλρ T νλρ , (5)
6
vµ := T λλµ , (6)
rin

1 1 1
tλµν := (Tλµν + Tµλν ) + (gνλ vµ + gνµ vλ ) − gλµ vν , (7)
2 6 3
representing the axial, vector and tensor parts respectively. Here, ϵµνλρ is the totally antisymmetric Levi-Civita tensor
in four dimensions. Using this decomposition, distinct gravitational scalar invariants can be defined as follows [12]
ep

1
Tax := aµ aµ = − Tλµν T λµν − 2Tλµν T µλν ,


(8)
18
Tvec := vµ v µ = T λλµ Tρ ρµ , (9)
1
1
T ten := tλµν tλµν = Tλµν T λµν + Tλµν T µλν − T λλµ Tρ ρµ . (10)
Pr

2 2

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 4

These scalars encompass all general non-parity-violating scalar invariants involving up to quadratic contractions of
the torsion tensor.
A particular linear combination of the axial, vector, and purely tensorial scalar invariants yields the torsion scalar,
defined as [14]

d
 
3 2 2 1 λ ρµ ν ρ λµ ν 1 µρ νλ
T := Tax + Tten − T vec = EA g EB + 2EB g EA + ηAB g g T A µν T B ρλ . (11)
2 3 3 2 2

we
The torsion scalar is essential, as it equates to the Ricci scalar up to a total derivative term [12]
◦ 2  
R = R + T − ∂µ eT λλ µ = 0 , (12)
e
where R represents
the √
Ricci scalar computed with the teleparallel connection, which vanishes as mentioned earlier,

ie
and e = det eAµ = −g is the determinant of the tetrad. The conventional Ricci scalar computed from the
Levi-Civita connection can then be written as
◦ 2  
R = −T + ∂µ eT λλ µ := −T + B , (13)

ev
e
where B represents a total divergence term.
An action based on the linear form of the torsion scalar leads to the Teleparallel Equivalent of GR (TEGR), which
is dynamically equivalent to GR [8, 56]. By analogy to gravity theories based on curvature, the TEGR action can be
generalized to an f (T ) gravity framework [18, 24, 32, 36, 39, 46, 49, 50, 58, 66, 71] where the Lagrangian is promoted

r
from T to an arbitrary function of it, f (T ). A notable advantage of f (T ) gravity is that, unlike its curvature-analog,
f (R), the resulting field equations remain second order in derivatives of the tetrad, simplifying thus the equations of
motion. er
In this work, we explore the teleparallel version of Horndeski gravity, which studies the interactions of the metric
with a scalar field. In TG, scalar fields couple to matter in the same manner as in GR through a minimal coupling
approach, where the partial derivatives are elevated to Levi-Civita covariant derivatives, i.e. [8, 23]
pe
˚µ ,
∂µ → ∇ (14)

which applies only to the matter sector. Based on that, we can examine the recently introduced teleparallel analog of
Horndeski gravity[15–17], also known as the Bahamonde-Dialektopoulos-Levi Said (BDLS) theory. The formulation
of this theory is based on three fundamental conditions: (i) the field equations must maintain a maximum order of
second derivatives of the tetrads; (ii) the scalar invariants involved should not violate parity; and (iii) the number
of contractions with the torsion tensor must be limited to a maximum of quadratic order. Although, higher-order
ot

contractions of the torsion tensor can yield second-order field equations, BDLS theory was intentionally crafted under
these criteria, which we adhere in our analysis.
Due to the second-order nature of various extensions of TG, the resulting action serves as an extension of the con-
ventional Horndeski gravity. Consequently, the conditions outlined above yield both the standard terms of Horndeski
tn

gravity and additional terms that are linear in contractions with the torsion tensor [17]

I2 = v µ ϕ;µ , (15)

where ϕ denotes the scalar field. Furthermore, we obtain terms that are quadratic in this context, including

J1 = aµ aν ϕ;µ ϕ;ν ,
rin

(16)
J3 = vσ tσµν ϕ;µ ϕ;ν , (17)
σµν α
J5 = t tσ ν ϕ;µ ϕ;α , (18)

J6 = tσµν tσαβ ϕ;µ ϕ;ν ϕ;α ϕ;β , (19)

ep

J8 = tσµν tσµα ϕ;ν ϕ;α , (20)

J10 = ϵµνσρ aν tαρσ ϕ;µ ϕ;α , (21)

where the semicolons denote covariant derivatives associated with the Levi-Civita connection.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 5

Therefore, we can express the teleparallel analogue of Horndeski gravity in the form of an action, given by
Z 5 Z
X Z
4 4
SBDLS = d x eLTele + d x eLi + d4 x eLm , (22)

d
i=2

where the contributions from standard Horndeski gravity remain present as seen below [57]

we
L2 := G2 (ϕ, X) , (23)
˚ ,
L3 := −G3 (ϕ, X)□ϕ (24)
  
2
˚
L4 := G4 (ϕ, X) (−T + B) + G4,X (ϕ, X) □ϕ − ϕ;µν ϕ ;µν
, (25)

ie
  
1 3
˚
L5 := G5 (ϕ, X)G̊µν ϕ;µν − G5,X (ϕ, X) □ϕ ˚ .
+ 2ϕ;µν ϕ;ν α ϕ;αµ − 3ϕ;µν ϕ;µν □ϕ (26)
6

These terms are analogous to their standard Horndeski counterparts but are computed using the teleparallel quantities

ev
instead of the metric. Nevertheless, they yield the same contributions to the equations of motion for specific systems.
Here we define

LTele := GTele (ϕ, X, T, Tax , Tvec , I2 , J1 , J3 , J5 , J6 , J8 , J10 ) , (27)

r
where the kinetic term is expressed as X := − 12 ∂ µ ϕ∂µ ϕ, Lm represents the matter Lagrangian in the Jordan conformal
◦
frame and Gµν denoted the regular Einstein tensor. In this notation, commas signify standard partial derivatives.
Notably, when GTele = 0, we retrieve the original form of Horndeski gravity.
er
B. Background Cosmology
pe
By varying the action with respect to the tetrad, spin connection, and scalar field, we can derive the field equations as
detailed in Ref. [16]. The broader scope of this theory results in significantly more complex field equations. Therefore,
we will focus exclusively on the equations of motion relevant to a flat Friedmann-Lemaı̂tre-Robertson-Walker (FLRW)
cosmology, described by the metric

ds2 = −N (t)2 dt2 + a(t)2 (dx2 + dy 2 + dz 2 ) , (28)

ot

where N (t) denotes the lapse function (which we can set to one after obtaining the equations of motion), and
a(t) representes the scale factor. To derive the modified equations of motion, we adopt the tetrad choice eaµ =
diag(N (t), a(t), a(t), a(t)) which aligns with the Weitzenböck gauge [14, 64].
By varying the point-like Lagrangian with respect to the dynamical variables N (t), a(t), and ϕ(t), we derive the
equations of motion of the system for a flat homogeneous and isotropic background. This leads us to the Friedmann
tn

equation
5
X
ETele + Ei = 0 , (29)
i=2
rin

where

ETele = 6H ϕ̇G̃6,I2 + 12H 2 G̃6,T + 2X G̃6,X − G̃6 , (30)

E2 = 2XG2,X − G2 , (31)
E3 = 6X ϕ̇HG3,X − 2XG3,ϕ , (32)
ep

2 2
E4 = −6H G4 + 24H X(G4,X + XG4,XX ) − 12HX ϕ̇G4,ϕX − 6H ϕ̇G4,ϕ , (33)
3 2
E5 = 2H X ϕ̇ (5G5,X + 2XG5,XX ) − 6H X (3G5,ϕ + 2XG5,ϕX ) , (34)
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 6

and

LTele = G̃6 (ϕ, X, T, I2 ) , (35)

d
which encompasses all the nonzero scalars associated with GTele . The Hubble parameter is defined as H = ȧ/a, and
dots indicate derivatives with respect to cosmic time. The torsion scalar is expressed as T = 6H 2 , while I2 = 3H ϕ̇
and X = 21 ϕ̇2 , with commas indicating partial derivatives. Next, varying with respect to the scalar factor leads us to

we
the second Friedmann equation
5
X
PTele + Pi = 0 , (36)
i=2

where

ie
d 
PTele = −3H ϕ̇G̃6,I2 − 12H 2 G̃6,T − 4H G̃6,T + ϕ̇ G̃6,I2 + G̃6 , (37)
dt
P2 = G2 , (38)

ev
 
P3 = −2X G3,ϕ + ϕ̈G3,X , (39)
 
P4 = 2 3H 2 + 2Ḣ G4 − 12H 2 XG4,X − 4H ẊG4,X − 8ḢXG4,X
   
− 8HX ẊG4,XX + 2 ϕ̈ + 2H ϕ̇ G4,ϕ + 4XG4,ϕϕ + 4X ϕ̈ − 2H ϕ̇ G4,ϕX , (40)

r
 
P5 = −2X 2H 3 ϕ̇ + 2H Ḣ ϕ̇ + 3H 2 ϕ̈ G5,X − 4H 2 X 2 ϕ̈G5,XX
  er 
d

+ 4HX Ẋ − HX G5,ϕX + 2 2 (HX) + 3H 2 X G5,ϕ + 4HX ϕ̇G5,ϕϕ .
dt
(41)

Finally, the modified Klein-Gordon equation can be obtained by varying with respect to the scalar field, resulting in
pe
1 dh 3 i
a (J + J Tele ) = Pϕ + PTele , (42)
a3 dt
where the standard Horndeski terms appear as J and Pϕ , arising from the Lagrangian terms Li , with i = 2, .., 5 as
shown in [62]

J = ϕ̇G2,X + 6HXG3,X − 2ϕ̇G3,ϕ + 6H 2 ϕ̇ (G4,X + 2XG4,XX ) − 12HXG4,ϕX

ot

+ 2H 3 X (3G5,X + 2XG5,XX ) − 6H 2 ϕ̇ (G5,ϕ + XG5,ϕX ) , (43)

   
Pϕ = G2,ϕ − 2X G3,ϕϕ + ϕ̈G3,ϕX + 6 2H 2 + Ḣ G4,ϕ
 
+ 6H Ẋ + 2HX G4,ϕX − 6H 2 XG5,ϕϕ + 2H 3 X ϕ̇G5,ϕX ,
tn

(44)

where JTele and PTele denote additional terms associated with the teleparallel Horndeski, given by

JTele = ϕ̇G̃6,X , (45)

d  
rin

PTele = −9H 2 G̃6,I2 + G̃6,ϕ − 3 H G̃6,I2 . (46)

dt

Notably, the parameters of G̃6 are independent of Tax and Tten , as these values are zero for the flat FLRW metric.
Consequently, we can express the contributions solely in terms of Tvec , given that T = −(2/3)Tvec = 6H 2 in this
scenario. It is important to recognize that this relationship may not hold true at the perturbative level [15]. In
ep

this study, we investigate the cosmological perturbations within the teleparallel analogue of Horndeski gravity. This
approach will facilitate more thorough explorations of specific models within the framework, potentially enhancing
our understanding of which models best align with observations regarding the cosmic evolution of the Universe.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 7

III. COSMOLOGICAL PERTURBATIONS

We will now calculate the perturbations around a FLRW spacetime up to second order for the BDLS theory, which
is given by the Lagrangian (22). Since the dynamical variable in our formulation of the theory is the tetrad, we will

d
perturb it as

eA µ = ēA µ + δeA µ , (47)

we
where ēA µ is the background FLRW diagonal tetrad, i.e.

ēA µ = diag[N (t), a(t), a(t), a(t)] , (48)

and δeA µ is the tetrad perturbation. We decompose it according to the group of spatial rotations and thus it is

ie
described by
 
Φ a (∂i β + ui )
δeAµ =  (49)
 

ev

δiA ∂ i B + v i aδ Aj δij ψ + ∂i ∂j E + 2∂(i wj) + 21 hij + ϵijk (∂ k σ + V k )

The 16 components of this field are split into one traceless and transverse tensor hij , three traceless vectors ui , v i and
wi , one traceless pseudovector V k and five scalars Φ, ψ, B, β and E as well as one pseudoscalar σ. One can easily find
that at the linearized level, the (pseudo)scalar, (pseudo)vector, and tensor perturbations decouple. In the following

r
subsections, we will treat these modes separately, and study the degrees of freedom (DoFs) of the theory, but before
that, let’s see how the perturbative variables transform under diffeomorphisms.
er
A. Gauge symmetry

The BDLS action (22) is invariant under infinitesimal coordinate transformations

pe
xµ → x′µ = xµ + ξ µ (x). (50)

Under these transformations the tetrad becomes

∂xν
eA µ → e′A µ = eA ν .
∂x′µ
ot

We decompose the transformation vector as

 
1
ξ µ = ξ 0 , ξ i + δ ij ∂j ξ ,


a
tn

with ξ 0 and ξ being scalars and ξ i a divergenceless vector. The perturbative variables on FLRW transform as

σ̃ = σ , (51a)
˜
δϕ = δϕ + ϕ̇ξ 0 , (51b)
rin

Φ̃ = Φ + ξ˙0 , (51c)
1 ijk
ψ̃ = ψ + Hξ 0 , Ṽ k = V k + ϵ ∂i ξj , (51d)
2a
1
β̃ = β + ξ 0 , ũi = ui , (51e)
a
ep

B̃ = B − Hξ + ξ˙ , ṽi = vi − Hξi + ξ˙i , (51f)

1 1
Ẽ = E + ξ , w̃i = wi + ξi , h̃ij = hij , (51g)
a 2a
In what follows, we present the perturbative analysis of the different modes separately.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 8

B. Scalars

Regarding the scalar perturbations, apart from the five scalar modes discussed above, we have an additional one
coming from the perturbation of the scalar field. Specifically, we have

d
ϕ = ϕ0 + δϕ.

we
The BDLS action (22) up to second order terms becomes
!2  
−
"
Z 2
∇ (B − aĖ) B aĖ  
2 ˙
SS = dtd3 xa3 A1 + A2 ∇ 2
Φ + A3 ∇ 2
ψ̇ + A 4 ∇ 2
δϕ + A5 ∇ δϕ
a2 a2
β  ˙ + A10 ∇2 ψ + A12 ∇2 ψ̇ + β̇ A7 ∇2 Φ + A11 ∇2 ψ

+ 2 A6 ∇2 Φ + A8 ∇2 δϕ + A9 ∇2 δϕ

ie


a a2
!2 2
2 2
  
∇β̇ ∇ ψ ∇ δϕ ˙ + A19 ∇Φ
+ A13 + Φ A14 2 + A15 ψ̇ + A16 δϕ + A17 2 + A18 δϕ
a a a a

ev
 2  2
2 ∇ψ 2 ∇ψ∇δϕ ˙ ∇δϕ
+ A20 Φ + A21 + A22 ψ̇ + A23 + A24 δϕψ̇ + A25 δϕψ̇ + A26
a a2 a
#
+ A27 δϕ2 + A28 δϕ˙2 , (52)

r
where all the coefficients A1 to A28 are presented in the appendix A. Note that further integration by parts
would reduce the number of terms present in the action, but for simplicity this is omitted. Coefficients
er
{A1 , A7 , A10 , A11 , A13 , A19 } are purely teleparallel and would completely vanish in the standard Horndeski limit.
As already mentioned above, we have the gauge freedom to choose one element out of each set {δϕ, Φ, β, ψ} and
{B, E} to be set to zero. From the above action, it is easy to obtain the results expressed in different gauge choices.
In this section we will work with gauge invariant variables, but in the Appendix B we will present the same results in
pe
flat, unitary, Newtonian and synchronous gauges. In that way, people who want to emphasize on different things can
choose any gauge they want to work with.
Since {δϕ, Φ, β, ψ, B, E} are gauge dependent under (50), we define

X1 = δϕ − aϕ˙0 β , (53a)
X2 = Φ − a(Hβ + β̇) , (53b)
ot

X3 = ψ − aHβ , (53c)
X4 = −B + aĖ , (53d)

which are gauge invariant. Substituting these in (52) we get

tn

"  2  2  2
∇X2 ∇X3 ∇X1
Z
3 3 2 2 2 2
SS = dt d x a Ã1 X2 + Ã2 + Ã3 + Ã4 Ẋ3 + Ã5 X1 + Ã6 Ẋ1 + Ã7
a a a
 2 2
∇ X4 ∇X2 ∇X3 ∇X2 ∇X1
+ Ã8 2
+ Ã9 X2 Ẋ3 + Ã10 2
+ Ã11 X2 X1 + Ã12 X2 Ẋ1 + Ã13
a a a2
rin

#
∇X3 ∇X1 ∇2 X 4
+ Ã15 Ẋ3 X1 + Ã16 Ẋ3 Ẋ1 + Ã17 + (Ã14 X2 + Ã18 Ẋ3 + Ã19 X1 + Ã20 Ẋ1 ) , (54)
a2 a2

where again, the coefficients Ã1 − Ã20 are presented in the appendix A. We notice that not all the scalar modes are
ep

propagating and specifically X2 and X4 are auxiliary. In order to find the action with the propagating modes only, we
need to vary the above action with respect to the auxiliary fields, find the constraint equations and substitute them
back to the action. This will be done in the next section IV.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 9

C. Pseudoscalar perturbations

Moving on to the pseudoscalar mode σ and perturbing the BDLS action up to second order, we get

d
a3 
Z
(2)
SPS = dtd3 x B1 (∂i σ̇)2 + B2 (∆σ)2

(55)
9

we
where

B1 = 4GTele,Tax + 9XGTele,J5 − 12XGTele,J10 , (56a)

4
B2 = 2 (2XGTele,J1 − GTele,Tax ) . (56b)
a

ie
The pseudoscalar in the tetrad, introduced to account for changes in sign under parity transformations, only con-
tributes to the antisymmetric part of the tetrad. It can be seen that its contribution vanishes when calculating the
symmetric metric, which is further highlighted in the action (55), where all coefficients are from the teleparallel sector.
In some classes of teleparallel theories, like Type I, II and V New General Relativity (NGR) the pseudoscalar mode

ev
propagates [10], while in others, like f (T ) it does not because of the presence of remnant symmetries [51, 59]. Since
both NGR and f (T ) theories are subclasses of BDLS we cannot extract a unique result for all subclasses regarding
the propagation of the pseudoscalar mode. However, in case they’re propagating, in order for them to be ghost-free,
the condition

B1 > 0 , (57)

r
has to be satisfied.
er D. Vectors

Let us now study the vector modes. In the curvature-based theory, the vector sector decays immediately. Here,
pe
where the fundamental dynamical object is the tetrad, the same perturbation procedure is applied to the vector sector,
yielding

(∇v )2 (∆w)2 (∇V)2 (∇u)2

Z 
(∇v )(∇ẇ)
SV = dt d3 x a3 C1 2
+ C2 2
+ C3 (∇ẇ)2 + C2 2
+ C4 V̇2 + C5 u̇2 + C6 2
− C3
a a a a a
(∇v )(∇u) (∇u)(∇w) (∇u)(∇ẇ) (∇u̇)(∇w) (∇ × V) ∆w
ot

+ C7 2
+ C8 + C9 + C10 + 2C2
 a a a a  a2
1
+ C4 V̇ (∇ × v ) + C8 V (∇ × u) + C10 V (∇ × u̇) + C11 V̇ (∇ × u) , (58)
a
tn

where as previously, the coefficients C1 − C1 1 are shown in the appendix.

Note that the pseudovector modes couple with the vector ones already at linear order (in constrast with the rest of
the modes which couple at second order and beyond) and thus cannot be treated separately1 . While non-teleparallel
coefficients appear in the action, further integration by parts would show that any remaining contributions correspond
to a non-dynamical mode, which in this case is vi . By fixing vi , wi or Vi to zero we can obtain a gauge fixed result
(see Appendix B for further details). Nonetheless, the gauge invariant route is opted.
rin

We define the following gauge invariant variables

Yi = vi − 2aẇi , (59a)
j k
Zi = Vi + ϵijk ∂ w . (59b)
ep

1 In principle, one could set the pseudovectors to zero as a gauge choice, since one of the {vi , ui , Vi } can vanish. This will be studied in
the next section; here we want to study gauge invariant variables.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 10

and the action (58) takes the form

(∇u)2

Z (∇ × u) (∇ × u̇) Z
Z
(∇u) (∇Y)
SV = dt d3 x a3 C˜1 u̇2 + C˜2 2
+ C˜3 + C˜4 2
− C˜5
a a a a

d
2 2

Ż (∇ × Y) (∇Z) (∇Y)
+ C˜6 + C˜7 Ż2 + C˜8 + C˜9 (60)
a a2 a2

we
where again the coefficients are shown in the appendix A. The non-teleparallel terms only appear in {C˜2 , C˜4 , C˜9 } such
that if the Horndeski limit is applied to the system, all modes would be non-dynamical and give a trivial solution, as
expected. However, as can be seen from the action (60), not all modes are propagating; this will be studied in the
next section IV.

ie
E. Tensor perturbations

The quadratic action of the tensor modes reads

ev
a3
 
D2
Z
(2)
ST = dtd3 x D1 ḣ2ij − 2 (∇hij )2 , (61)
4 a

where

r
 
X
D1 = 2 G4 − 2XG4,X + XG5,ϕ0 − HX ϕ˙0 G5,X + 2XGTele,J8 + GTele,J5 − GTele,T , (62a)
2
 
D2 = 2 G4 − XG5,ϕ0 − X ϕ¨0 G5,X − GTele,T ,
er
where Gi,A denotes the derivative of Gi with respect to A. As expected, when GTele → 0, Eqs. (62a) and (62b) take
(62b)

the standard Horndeski form for the perturbed action, as presented in [62].
Variation of the action (61) yields the propagation equation of gravitational waves, which confirms the result in
pe
Ref. [15]. The standard αi parametrization of cosmological perturbations is presented in Ref. [4]. The squared speed
of the tensor modes reads
D2
c2T = , (63)
D1
which in general, is not equal to unity. Furthermore, as will be discussed in Sec. IV, we can deduce from the action
ot

(61) that the conditions

D1 > 0 and D2 > 0 , (64)

must hold in order to avoid ghost and gradient instabilities.

tn

IV. GHOST AND LAPLACIAN STABILITY CONDITIONS

In the previous section we presented the quadratic actions for all the modes of the perturbations, tensors, pseu-
rin

doscalar, scalars and vectors. The tensor modes are two, since hij is transverse and traceless, and in order for them
to propagate healthily, conditions (64) must be satisfied. Similarly for the pseudoscalar mode condition (57) should
hold. In this section, we study first of all which are the scalar and vector modes which propagate and second which
are the stability conditions for them to be healthy.
Ghosts represent extra degrees of freedom that may propagate due to a negative kinetic term, these instabilities
can have a detrimental effect on the health of the underlying theory. In order to identify the possible appearance of
ep

ghosts, the action is expanded up to second order about the FLRW background. In Fourier space, the auxiliary modes
are determined and eliminated by appropriate variations of the action with respect to those non-dynamical modes.
In practice the auxiliary field variations are substituted back into the original system and thus removed. In this way,
only dynamical modes are retained in the resulting action. This is performed separately for the scalar, vector, and
tensor sectors.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 11

Firstly, a gauge invariant action is formulated by taking its variations with respect to spatial and temporal parts
of an arbitrary vector field transformation, and imposing that this vanishes. In turn, a diagonalized kinetic matrix
K can be produced wherein a constraint can be associated with every entry. Due to the time dependence of the
background spacetime, these constraints may also express some time-dependence. At this point the gradient or

d
Laplacian instability can be defined as the imposition that the speed of propagation of these modes is positive
definite. In this way, both the ghost and gradient instabilities are confirmed to be stable. Due to the high energy
regime being the most likely region where ghosts appear, a high k limit is assumed throughout the procedure.

we
The general approach rests on an action perturbed up to second order where non-dynamical modes are identified.
Once these modes are nullified and then substituted back into the original action, it will take on a form

d3 k 3 ˙ ⊺
Z   2  
1 ˙ +χ k ˙ ,
δ2 S = 2 dt a χ
⃗ K(t, k) χ
⃗ ⃗ ⊺
G(t, k) + M(t, k) χ
⃗ + χ
⃗ ⊺
Q(t, k) χ
⃗ (65)
2κ (2π)3 a2

ie
where χ⃗ is a vector of modes, G is the gradient matrix, M is the matrix corresponding to non-dynamical portions
of the modes, and Q is an imaginary matrix. For the high-k limit, ghost modes correspond to entries of K, while
positive definite propagation speeds correspond to gradient instability corrections.

ev
A. Scalars

The scalar and pseudoscalar modes do not suffer any mixing and so can be treated separately. This approach is
identical to the Arnowitt-Deser-Misner (ADM) decomposition used in other torsion based treatments [33, 54], and

r
similarly leads to the same metric as curvature-based formulations. Taking the perturbation up to the second order
followed by integration by parts and the insertion of the background equations of motion, the gauge invariant action
(54) produces the Fourier transformed form given by er
"
d3 k 3 k2 k2
Z 
2 2
SS = dt 3 a Ã6 Ẋ1 + Ã4 Ẋ3 + 2 Ã8 X42 + Ã2 X22 + Ã3 X32 + Ã7 X12 + Ã10 X2 X3 + Ã13 X2 X1
(2π) 2 a a2

pe
+ Ã17 X3 X1 − X4 (Ã14 X2 + Ã18 Ẋ3 + Ã19 X1 + Ã20 Ẋ1 ) + Ã9 X2 Ẋ3 + Ã12 X2 Ẋ1
#
+ Ã15 Ẋ3 X1 + Ã16 Ẋ3 Ẋ1 + Ã1 X22 + Ã5 X12 + Ã11 X2 X1 , (66)

where X2 and X4 are non-dynamical modes. Taking variation with respect these two quantities results in the respective
relationships
ot

k2  
0= 2
2Ã2 X̃2 + Ã10 X3 + Ã13 X2 − Ã14 X4 + Ã9 Ẋ3 + Ã12 Ẋ1 + 2Ã1 X2 + Ã11 X1 , (67)
a
k2
0 = 2 2 Ã8 X4 − Ã19 X1 − Ã14 X2 − Ã20 Ẋ1 − Ã18 Ẋ3 , (68)
tn

a
which when substituted back in the action (66) gives

d3 k
Z
1 h
δ 2 SS = dt 3 a3 Â1 X12 + Â2 X1 Ẋ1 + Â3 Ẋ12 + Â4 X32 + Â5 X3 Ẋ3 + Â6 Ẋ32
2κ2 (2π)
rin

2
i
+Â7 X1 X3 + Â8 Ẋ1 X3 + Â9 X1 Ẋ3 + Â10 Ẋ1 Ẋ3 , (69)

where
  k2 
ep

ãÂ1 = Ã11 Ã14 Ã19 − Ã1 Ã219 − Ã5 Ã214 − Ã8 Ã211 − 4Ã1 Ã5 + 2 Ã13 Ã14 Ã19 − Ã2 Ã219
a
  k 4  
−Ã7 Ã214 + 2Ã8 −Ã11 Ã13 + 2Ã2 Ã5 + 2Ã1 Ã7 − 4 Ã8 Ã213 − 4Ã2 Ã7 , (70a)
a
    k2 
ãÂ2 = −Ã20 −Ã11 Ã14 + 2Ã1 Ã19 + Ã12 Ã14 Ã19 − 2Ã8 Ã11 − 2 2Ã2 Ã19 Ã20
a
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 12
 
+Ã13 −Ã14 Ã20 + 2Ã8 Ã12 , (70b)
  k2  
ãÂ3 = Ã12 Ã14 Ã20 − Ã1 Ã220 − Ã6 Ã214 − Ã8 Ã212 − 4Ã1 Ã6 − 2 Ã2 Ã220 − 4Ã6 Ã8 , (70c)
a

d
k2   k4  
ãÂ4 = − 2 Ã3 Ã214 − 4Ã1 Ã8 − 4 Ã8 Ã210 − 4Ã2 Ã3 , (70d)
a a
k2

we
 
ãÂ5 = 2 Ã10 Ã14 Ã18 − 2Ã8 Ã9 , (70e)
a
    k2  
ãÂ6 = −Ã4 Ã214 − Ã1 Ã218 − 4Ã4 Ã8 + Ã9 Ã14 Ã18 − Ã8 Ã9 − 2 Ã2 Ã218 − 4Ã4 Ã8 , (70f)
a
k2 h 2
i k4  
ãÂ7 = 2 −Ã14 Ã17 + Ã10 Ã14 Ã19 − 2Ã8 Ã10 Ã11 + 4Ã1 Ã8 Ã17 − 2 4 Ã8 Ã10 Ã13 − 2Ã2 Ã17 , (70g)
a a

ie
k2  
ãÂ8 = − 2 Ã10 −Ã14 Ã20 + 2Ã8 Ã12 , (70h)
a
ãÂ9 = −Ã214 Ã15 + Ã11 Ã14 Ã18 − 2Ã1 Ã18 Ã19 + 4Ã1 Ã8 Ã15 + Ã9 Ã14 Ã19 − 2Ã8 Ã9 Ã11
k2 

ev

+ 2 Ã13 Ã14 Ã18 − 2Ã2 Ã18 Ã19 + 4Ã2 Ã8 Ã15 − 2Ã8 Ã9 Ã13 , (70i)
a  
ãÂ10 = −Ã214 Ã16 + Ã12 Ã14 Ã18 + 2Ã1 Ã18 Ã20 + 4Ã1 Ã8 Ã16 − Ã9 −Ã14 Ã20 + 2Ã8 Ã12
k2  
+ 2 2 Ã2 −Ã18 Ã20 + 2Ã8 Ã16 , (70j)

r
a
 2

and ã = −Ã214 + 4Ã8 Ã1 + ka2 Ã2 . Provided that ã ̸= 0 the following analysis holds, otherwise the analysis
er would
need to be altered at the (66) stage of the calculation. Additionally, emphasizes have been made to keep the k-
dependencies. At a later stage this becomes crucial to obtain high-k limit. Through integration by parts Eq. (69)
changes to

d3 k 3
Z    
1 1 d  3  1 d  3 
pe
δ 2 SS = 2 dt 3 a Â1 − a Â2 X 2
1 + Â Ẋ
3 1
2
+ Â4 − a Â5 X32
2κ (2π) 2 2a3 dt 2a3 dt
  
2 1 d  3   
+Â6 Ẋ3 + Â7 − 3 a Â8 X1 X3 + Â9 − Â8 X1 Ẋ3 + Â10 Ẋ1 Ẋ3 . (71)
a dt

When looking at the kinetic contributions, it was noted that there is a mixing of modes of Ẋ1 Ẋ3 corresponding to
the coefficient Â10 . To obtain a diagonal kinetic matrix, as required in Eq. (65), the gauge invariant quantities are
ot

redefined as
!
1 Â10
X1 := Ψ1 , X3 := Ψ2 − Ψ1 , (72)
2 Â6
tn

where Ψ1 and Ψ2 are still gauge invariant as they are a linear combination of the previous dynamical modes. Due
to the complexity of the action upon substituting these new fields, the generalised result has been omitted. By
considering the high-k limit, the result simplifies to

d3 k 3 k2 
Z 
1
rin


δ 2 SS = 2 dt 3 a Ă Ψ̇
1 1
2
+ Ă Ψ̇
2 2
2
+ Ă Ψ
3 1
2
+ Ă Ψ
4 2
2
+ Ă Ψ Ψ
5 1 2
2κ (2π) 2 a2
i
+Ă6 Ψ21 + Ă7 Ψ22 + Ă8 Ψ1 Ψ2 + Ă9 Ψ1 Ψ̇2 . (73)

Before listing the coefficients of Ăi for i ∈ [1, 9], we consider the temporal Fourier transformation Ψ̇ → iωΨ such that
ep

the Lagrangian portion of Eq (73) can be expressed as follows:

    
2 k2 1 k2
 −ω Ă1 + a2 Ă3
+ Ă6 a2 Ă5
+ Ă8 + iω Ă9  Ψ1 
 
2
LS = Ψ1 Ψ2      . (74)
1 k2 2

2 a2 Ă5 + Ă8 + iω Ă9 −ω 2 Ă2 + ka2 Ă4 + Ă7 Ψ2

Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 13

The dispersion relation given by ω = cS ka shows that ω is of the same order of k. If we consider only the leading order
of k, coefficients {Ă6 , Ă7 , Ă8 , Ă9 } do not contribute, thus

d3 k k2 
Z  
1

d
2 3 2 2 2 2
δ SS = 2 dt 3 a Ă1 Ψ̇1 + Ă2 Ψ̇2 + 2 Ă3 Ψ1 + Ă4 Ψ2 + Ă5 Ψ1 Ψ2 , (75)
2κ (2π) 2 a

where

we
−Ã16 Ã18 Ã20 + Ã4 Ã220 + Ã8 Ã216
Ă1 = −Ã6 − , (76a)
Ã218 − 4Ã4 Ã8
Ã218
Ă2 = Ã4 − , (76b)
4Ã8

ie
1 (Ã210 − 4Ã2 Ã3 )(−Ã18 Ã20 + 2Ã8 Ã16 )2
Ă3 = (Ã10 Ã13 − 2Ã2 Ã17 )(−Ã18 Ã20 + 2Ã8 Ã16 ) +
2 4(Ã218 − 4Ã4 Ã8 )
1
+ (Ã213 − 4Ã2 Ã7 )(Ã218 − 4Ã4 Ã8 ) , (76c)

ev
4
Ã2
Ă4 = Ã3 − 10 , (76d)
4Ã2
Ã10 Ã13 (Ã210 − 4Ã2 Ã3 )(−Ã18 Ã20 + 2Ã8 Ã16 )
Ă5 = Ã17 − − . (76e)
2Ã2 2Ã2 (Ã218 − 4Ã4 Ã8 )

r
These results hold given Ã218 − 4Ã4 Ã8 ̸= 0, Ã2 ̸= 0 and Ã8 ̸= 0. Next, taking variations with respect to the dynamical
modes Ψ1 and Ψ2 results in the field equations

Ă′1
!
er k2 Ă3
!
k2 Ă5
!
0 ≈ Ψ̈1 + Ψ̇1 − 2 Ψ1 − 2 Ψ2 , (77)
Ă1 a Ă1 a 2Ă1
pe
! ! !
Ă′2 k2 Ă4 k2 Ă5
0 ≈ Ψ̈2 + Ψ̇2 − 2 Ψ2 − 2 Ψ1 . (78)
Ă2 a Ă2 a 2Ă2

respectively, indicating that a mixing of modes occurs due to Ă5 contribution. These two systems decouple for the
condition

Ă5 = 0 . (79)
ot

Under the restriction of Eq. (79), the propagating speed for each mode is given by the coefficient of the gradient term
such that
tn

Ă3 Ă4
c2Ψ1 := − > 0, and c2Ψ2 := − > 0, (80)
Ă1 Ă2
and ghost stability can be attained provided that

MΨ1 := Ă1 > 0 , and MΨ2 := Ă2 > 0 , (81)

rin

and Laplacian/gradient stability for

NΨ1 := −Ă3 > 0 , and NΨ2 := −Ă4 > 0 . (82)

By satisfying these conditions, ghost and gradient instabilities can be removed from the resulting models.
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 14

B. Vectors

We next consider the vector perturbations keeping with the gauge invariant approach. For this sector, the second-
order perturbation where integration by parts and the background equations of motion are resubstituted produces

d
d3 k 3 ˜ k2 
Z  
SV = dt 3 a C1 u̇i u̇i + C˜7 Żi Ż i + 2 C˜2 ui ui + C˜4 ui Y i + C˜8 Zi Z i + C˜9 Yi Y i
a

we
(2π) 2
ijk 
i ϵ kj ˜ 
˜ ˜
− C3 Zi uk − C5 u̇k Zi + C6 Żi Yk . (83)
a

Now, taking the variation with respect to the non-dynamical mode Yk yields the equation

k2  ˜ ijk
˜9 Yk − iϵ kj C˜6 Żi ,

ie

0= C 4 uk + 2 C (84)
a2 a
which when solved for Yk and substituted back in action (83) yields

ev
" ! ! !
d3 k 3 ˜ C˜62 k2 C˜42
Z
SV = dt 3 a
i
C1 u̇i u̇ + C7 +˜ Żi Ż i + 2 C˜8 Zi Z i − ˜
C2 − ui u i
(2π) 2 4C˜9 a 4C˜9
! !#
i ϵijk kj  ˜ C˜4 C˜6
C3 − H C˜5 + C˜˙5 Zi uk + C˜5 −

− Żi uk . (85)
a 2C˜9

r
Both the kinetic and the gradients portions of the action are a diagonal matrix when cast in the form given by Eq. (65).
The term corresponding to Zi uk can be omitted within the high-k limit as it is of a lower order of k-dependency in
er
comparison with the rest of the terms. In contrast with the scalar sector, the term Żi uk appears to have the same
2
order of k upon substituting the temporal Fourier transformation and ω 2 = c2V ka2 . To further analyse this scenario,
we take the variation with respect to the both ui and Zi , respectively
pe
C˜˙1
! ! !
k 2 C˜42 − 4C˜9 C˜2 i ϵijk k j 2C˜9 C˜5 − C˜4 C˜6
0 ≈ ük + 3H + u̇k − 2 uk + Ż i , (86)
C˜1 a 4C˜9 C˜1 a 2C˜9 C˜1
C̃62
      
d 3 ˜
C
!
1 dt a 7 + 4C̃9 k 2
4 ˜
C ˜
C i ϵ k j
2C˜9 C˜5 − C˜4 C˜6
9 8 ijk
0 ≈ Z̈k + 3H + 3  Żk −    Zk + u̇i , (87)
2 ˜ ˜ ˜
 
a ˜ C̃62 a ˜ ˜ ˜2 a 4 C C + C 2
C7 + 4C̃ 4C9 C7 + C6 9 7 6
9
ot

for which the two modes are said to decouple provided that

2C˜9 C˜5 − C˜4 C˜6 = 0 . (88)

tn

When this condition applies, the propagating speed for both modes can be determined:
 
4C˜9 C˜2 − C˜2 4 C˜9 C˜8
4
c2u := > 0, and c2Z := −    > 0 . (89)
4C˜9 C˜1 ˜ ˜
4C9 C7 + C ˜2
6
rin

Thus, ghost stability can be obtained when

C˜2
Mu := C˜1 > 0 , and MZ := C˜7 + 6 > 0 , (90)
4C˜9
ep

and Laplacian stability satisfied provided that

C˜2
Nu := C˜2 − 4 > 0 , and NZ := −C˜8 > 0 . (91)
4C˜9
which both depend on a combination of first order derivatives of the action functionals.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 15

V. CONCLUSION

The teleparallel analogue of Horndeski gravity expressed through the BDLS action in Eq. (22) offers an interesting
way to circumvent the restrictions placed on regular Horndeski gravity by the multimessenger events associated

d
with GW170817 and GRB170817A [45]. In the present work, we have explored the gauge-invariant cosmological
perturbations of this framework theory together with some initial stability and ghost conditions imposed on these
classes of models. This builds on previous work [4] where the cosmological perturbations were determined for a specific

we
gauge. The current work broadens the applicability of this analysis, and provides a practical way forward on both
the gauge invariant form of these perturbations, as well as example cases of the most popular gauge choices, which
appear in App. B.
To perform this analysis in a gauge invariant form, the tetrad inherits the perturbative form from the metric through
the choice in Eq. (49), which importantly remains within the Weitzenböck gauge in terms of the form of the spin
connection components. The general cosmological perturbations are then probed for their scalar, pseudoscalar, vector

ie
and tensor modes in Sec. III. The gauge invariant nature of these perturbations, coupled with the complexity of the
underlying model framework results in highly involved expressions for the perturbation modes. For this reason, we
provide example cases of the most popular gauges in the literature in App. B for the scalar and vector modes.
Already, at this level of analysis an assessment can be made on the propagation of modes and whether they contain

ev
ghosts or other instabilities, which are then probed in Sec. IV. The generality of this class of models means that it must
be imposed that the scalar field kinetic term be positive, while the speed of propagation of propagating modes should
also be positive, which will respectively alleviate ghost and Laplacian instabilities immediately. The gauge-invariant
approach results in fairly convoluted conditions on the underlying Lagrangian terms, which for the scalar sector give
Eqs. (81–82), while for the vector section Eqs. (89–91), and tensor sector Eqs. (64). By exploring models that satisfy

r
this combination of conditions, one can produce cosmologies that do not suffer from perturbative stability pathologies.

Theory PDoFs (Pseudo)Scalars Vectors Tensors

BDLS
er 9 Ψ1 , Ψ2 , σ u i , Zi hij

GR 2 - - hij
pe
f (T ) 2 - - hij

f (ϕ, X, T ) 3 Ψ2 - hij

Generalized Teleparallel Dark Energy 3 X1 - hij

ot

Generalized Scalar Tensor 3 Ψ2 - hij

NGR 8 X3 , σ u i , Zi hij
tn

Horndeski 3 Ψ2 - hij

TABLE I: List of propagating degrees of freedom for GR, f (T ), f (ϕ, X, T ), NGR and Horndeski gravity. In the case
of f (ϕ, X, T ) and NGR, diagionalisation of the kinetic matrix is required thus scalar modes are expressed in terms of
{Ψ1 , Ψ2 }. The scalar sector of the rest of the theories can be expressed in terms of {X1 , X3 }.
rin

The general BDLS framework can be reduced for specific subclasses which represent a smaller number of propagating
DoF. For the full treatment, it is found that there are 9 propagating DoFs which include 2 scalars, 1 pseudoscalar, 2
pairs of vectors modes, and the regular pair of tensor modes. This confirms the indications made in Ref. [13] where
the DoF were obtained through the respective dispersion relations of the different perturbative sectors. In Table I
the number of propagating DoFs is listed for several leading subclasses of the BDLS frame. By substituting these
ep

subclasses into our results, we confirm the expected 2 DoFs for GR and f (T ) gravity, as well as an extra scalar
mode for f (ϕ, X, T ), regular Horndeski gravity, and other forms. It is only NGR that also contains a large number of
propagating DoFs which in that case results in 8 DoFs.
In the case of f (T ) gravity, the number of propagating DoFs falls to 2 due to the case when the background is
Minkowski wherein the extra scalar DoF is rendered non-dynamical [13]. By exploring the longitudinal gauge with
Pr

the added assumption that β = 0 gives identical results as in Ref. [59]. However, this may yield over-fixing since

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 16

the gauge choice only imposes that β = B. As for the NGR case, the non-trivial linearised field equations for the
Minkowski limit imply that GTele,Tax = 0 which forces the psuedoscalar contribution to be non-dynamical. The other
DoFs propagate in agreement with Ref. [10]. This does not occur in the general BDLS framework since the field
equations also depend on the invariants Tax , J1 , J5 , J10 which incur contributions from both B1 and B2 .

d
The central motivation of this work was to determine the gauge-invariant cosmological perturbations of the teleparal-
lel analogue of Horndeski gravity as expressed through the BDLS formalism. It would now be interesting to investigate
the numerical strength of the propagating DoFs and their impact on changes to the standard cosmological paradigm.

we
To do this, further physically motivated models are needed, which can be analysed against a spectrum of observational
data connected to the power spectra of these perturbative modes. The cosmological perturbations can also be further
investigated in order to assess condition on producing healthy theories.

Acknowledgments

ie
This article is also based upon work from COST Action CA21136 Addressing observational tensions in cosmology
with systematics and fundamental physics (CosmoVerse) supported by COST (European Cooperation in Science and
Technology). K.F.D. was supported by the PNRR-III-C9-2022–I9 call, with project number 760016/27.01.2023. JLS

ev
would also like to acknowledge funding from “Xjenza Malta” as part of the “FUSION R&I: Research Excellence
Programme” REP-2023-019 (CosmoLearn) Project, and the “Net4Tensions” as part of the “Research Networking
Scheme”.

r
Appendix A: Coefficients of the quadratic actions

The scalar quadratic action of BDLS theory in terms of the gauge-dependent variables is given by Eq. (52). The
er
coefficients A1 − A2 8 are shown here

a2 h
A1 = − 6GTele,Tvec + 12H 2 (9GTele,Tvec Tvec − 12GTele,T Tvec + 4GTele,T T ) + 12H ϕ˙0 (−3GTele,Tvec I2
6 i
pe
+ 2GTele,T I2 ) + ϕ˙0 2 (4GTele,J8 + GTele,J5 + 3GTele,I2 I2 ) , (A1a)
h
A2 = −a − 12H 3 (9GTele,Tvec Tvec + 4(−3GTele,T Tvec + GTele,T T )) + ϕ˙0 (2G4,ϕ0 − GTele,I2
− 2X(G3,X − 2G4,ϕ0 X + GTele,XI2 )) − H 2 ϕ˙0 (−54GTele,Tvec I2 + 36GTele,T I2 + 10XG5,X + 4X 2 G5,XX )

+ 2H 2G4 + 3GTele,Tvec − 2GTele,T + 2X(−4G4,X − 2GTele,XT + 3(G5,ϕ0 − GTele,I2 I2 + GTele,XTvec ))
i
+ 4X 2 (−2G4,XX + G5,ϕ0 X ) , (A1b)
ot

h
A3 = a 4G4 + 6GTele,Tvec − 4GTele,T − 12H 2 (9GTele,Tvec Tvec − 12GTele,T Tvec + 4GTele,T T )
i
+ 2X(−4G4,X + 2G5,ϕ0 − 3GTele,I2 I2 ) − 2H ϕ˙0 (−18GTele,Tvec I2 + 12GTele,T I2 + 2XG5,X ) , (A1c)
tn

h
A4 = a H(−2G4,ϕ0 + 6XG3,X − 20XG4,ϕ0 X + 4XG5,ϕ0 ϕ0 + 3GTele,I2 + 6GTele,ϕ0 Tvec − 4GTele,ϕ0 T )
+ H 2 ϕ˙0 (6G4,X − 6G5,ϕ0 + 12XG4,XX − 8XG5,ϕ0 X ) + H 3 (6XG5,X + 4X 2 G5,XX ) + ϕ˙0 (G2,X
i
− 2G3,ϕ0 + 2G4,ϕ0 ϕ0 + GTele,X − GTele,ϕ0 I2 ) , (A1d)
rin

h
A5 = −a − 2G4,ϕ0 + GTele,I2 + 2X(G3,X − 2G4,ϕ0 X + GTele,XI2 ) + H ϕ˙0 (4G4,X − 4G5,ϕ0 + 3GTele,I2 I2
− 6GTele,XTvec + 4GTele,XT + 4X(2G4,XX − G5,ϕ0 X )) + H 2 (−18GTele,Tvec I2 + 12GTele,T I2 + 6XG5,X
i
+ 4X 2 G5,XX ) , (A1e)
ep

h
A6 = a ϕ˙0 (2G4,ϕ0 − GTele,I2 − 2X(G3,X − 2G4,ϕ0 X )) + 2H(2G4 + 3GTele,Tvec − 2GTele,T
i
+ 2X(−4G4,X + 3G5,ϕ0 ) + 4X 2 (−2G4,XX + G5,ϕ0 X )) − 2X ϕ˙0 H 2 (5G5,X + 2XG5,XX ) , (A1f)
2 h i
A7 = a 9GTele,Tvec + 2X(−2GTele,J8 − 5GTele,J5 + 3GTele,J3 + 2XGTele,J6 ) , (A1g)
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 17
h
A8 = −a 2XH 3 (3G5,X + 2XG5,XX ) + H 2 ϕ˙0 6G4,X − 6G5,ϕ0 − 2GTele,J3 + 6Ḣ(3GTele,Tvec J3 − 2GTele,T J3 )


+ 4X(3G4,XX − 2G5,ϕ0 X ) − 3ϕ¨0 GTele,I2 J3 + ϕ˙0 G2,X − 2G3,ϕ0 + 2G4,ϕ0 ϕ0 + GTele,X − GTele,ϕ0 I2
 

− Ḣ(GTele,J3 + 3GTele,I2 I2 ) − ϕ¨0 GTele,XI2 + H − 2G4,ϕ0 + GTele,I2 + 18ḢGTele,Tvec I2

 

d
− 12ḢGTele,T I2 + 2X(3G3,X − 10G4,ϕ0 X + 2G5,ϕ0 ϕ0 − GTele,ϕ0 J3 − 3ḢGTele,I2 J3 ) − ϕ¨0 (GTele,J3
i

we
+ 3GTele,I2 I2 + 2XGTele,XJ3 ) , (A1h)
h
A9 = a − 2G4,ϕ0 + 2X[G3,X + H 2 (3G5,X + 2XG5,XX ) − 2G4,ϕ0 X ] + GTele,I2 + H ϕ˙0 4(G4,X + 2XG4,XX


i
− G5,ϕ0 − XG5,ϕ0 X ) + GTele,J3 , (A1i)
h i
A10 = −2a H(−6GTele,Tvec + 4GTele,T ) + ϕ˙0 GTele,I2 , (A1j)

ie
2 h i
A11 = a 18(GTele,Tvec − GTele,T ) + X(4GTele,J8 + 10GTele,J5 + 3GTele,J3 − 4XGTele,J6 ) , (A1k)
9 h i
A12 = 4a − G4 + X(2G4,X − G5,ϕ0 + H ϕ˙0 G5,X ) , (A1l)

ev
1 h i
A13 = a2 2X(−2GTele,J8 + 2XGTele,J6 − 5GTele,J5 + 3GTele,J3 ) + 9GTele,Tvec , (A1m)
9
2 h
A14 = − 18(G4 − GTele,T + GTele,Tvec ) + X(36G4,X − 3GTele,J3 − 2(9G5,ϕ0 + 2GTele,J8 + 5GTele,J5 ))
9

r
i
+ 18ϕ˙0 XHG5,X + 4X 2 GTele,J6 , (A1n)
h
A15 = −3 12H 3 (9GTele,Tvec Tvec + 4(−3GTele,T Tvec + GTele,T T )) + ϕ˙0 (−2G4,ϕ0 + GTele,I2
er
+ 2X(G3,X − 2G4,ϕ0 X + GTele,XI2 )) + H 2 ϕ˙0 (−54GTele,Tvec I2 + 36GTele,T I2 + 10XG5,X + 4X 2 G5,XX )
+ 2H(−2G4 − 3GTele,Tvec + 2GTele,T + 2X(4G4,X + 2GTele,XT − 3(G5,ϕ0 − GTele,I2 I2 + GTele,XTvec ))
i
+ 4X 2 (2G4,XX − G5,ϕ0 X )) , (A1o)
pe
2

A16 = G2,ϕ0 + GTele,ϕ0 + 6H G4,ϕ0 + 3GTele,ϕ0 Tvec − 2GTele,ϕT + X(−4G4,ϕ0 X − 4XG4,ϕ0 XX + 3G5,ϕ0 ϕ0
+ 2XG5,ϕ ϕ X ) − 2X(G2,ϕ X − G3,ϕ ϕ + GTele,ϕ X ) − 2XH 3 ϕ˙0 (5G5,ϕ X + 2XG5,ϕ XX )

0 0 0 0 0 0 0 0

− 6H ϕ˙0 (XG3,ϕ0 X − G4,ϕ0 ϕ0 − 2XG4,ϕ0 ϕ0 X + GTele,ϕ0 I2 ) , (A1p)

2
A17 = −2G4,ϕ0 + 2X(G3,X − 2G4,ϕ0 X + H (3G5,X + 2XG5,XX )) + GTele,I2 + H ϕ˙0 (4G4,X + 8XG4,XX
− 4(G5,ϕ0 + XG5,ϕ0 X ) + GTele,J3 ) , (A1q)
ot

A18 = −2H 3 15XG5,X + 4X 2 (5G5,XX + XG5,XXX ) − 27GTele,Tvec I2 + 18GTele,T I2 − 3H − 2G4,ϕ0

  

+ GTele,I2 + 2X(3G3,X + 2XG3,XX − 8G4,ϕ0 X − 4XG4,ϕ0 XX + 3GTele,XI2 ) − 6H 2 ϕ˙0 3G4,X

 

+ X(12G4,XX + 4XG4,XXX − 7G5,ϕ0 X − 2XG5,ϕ0 XX ) − 3(G5,ϕ0 − GTele,I2 I2 + GTele,XTvec )

tn

+ 2GTele,XT − ϕ˙0 G2,X − 2G3,ϕ0 + GTele,X + 2X(G2,XX − G3,ϕ0 X + GTele,XX ) ,

  
(A1r)
1 
A19 = 2X(−2GTele,J8 + 2XGTele,J6 − 5GTele,J5 + 3GTele,J3 ) + 9GTele,Tvec , (A1s)
9
= 18H 4 (9GTele,Tvec Tvec − 12GTele,T Tvec + 4GTele,T T ) + 3H 2 − 2G4 − 3GTele,Tvec + 2GTele,T

A20
rin

+ 2X(7G4,X + X(16G4,XX + 4XG4,XXX − 9G5,ϕ0 X − 2XG5,ϕ0 XX ) − 6(G5,ϕ0 − GTele,I2 I2 + GTele,XTvec )

+ 4GTele,XT ) + X G2,X − 2G3,ϕ0 + GTele,X + 2X(G2,XX − G3,ϕ0 X + GTele,XX ) + 3H ϕ˙0 − 2G4,ϕ0
   

+ GTele,I2 + 2X(2G3,X + XG3,XX − 5G4,ϕ0 X − 2XG4,ϕ0 XX + 2GTele,XI2 ) + 2H 3 ϕ˙0 15XG5,X

 

+ 13X 2 G5,XX + 2X 3 G5,XXX − 54GTele,Tvec I2 + 36GTele,T I2 ,


(A1t)
ep

2
A21 = 9(G4 + 2GTele,Tvec − GTele,T ) + X(−9G5,ϕ0 − 2GTele,J8 + 2XGTele,J6 − 5GTele,J5 − 6GTele,J3
9
− 9ϕ¨0 G5,X ) ,

(A1u)
A22 = −6G4 + 12XG4,X − 6XG5,ϕ0 + 9XGTele,I2 I2 − 9GTele,Tvec + 6GTele,T + 18H 2 [9GTele,Tvec Tvec
Pr

− 12GTele,T T + 4GTele,T T ] + 6H ϕ˙0 [XG5,X − 9GTele,T I + 6GTele,T I ] ,

vec vec 2 2
(A1v)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 18

A23 = −4XH 2 G5,X + 4G4,ϕ0 − 2GTele,I2 + 4X(−2G4,ϕ0 X + G5,ϕ0 ϕ0 − ḢG5,X ) + 4ϕ¨0 (−G4,X − 2XG4,XX
+ G5,ϕ0 + XG5,ϕ0 X ) + H ϕ˙0 (−4G4,X + 4G5,ϕ0 − 4XG5,ϕ0 X + GTele,J3 − 4ϕ¨0 (G5,X + XG5,XX )) , (A1w)
3

A24 = −3 2XH (3G5,X + 2XG5,XX ) + H(6XG3,X − 2G4,ϕ0 + 4X(−5G4,ϕ0 X + G5,ϕ0 ϕ0 ) + 3GTele,I2

d
+ 6GTele,ϕ0 Tvec − 4GTele,ϕ0 T ) + 2H 2 ϕ˙0 (3G4,X + 6XG4,XX − 3G5,ϕ0 − 4XG5,ϕ0 X ) + ϕ˙0 (G2,X − 2G3,ϕ0

+ 2G4,ϕ0 ϕ0 + GTele,X − GTele,ϕ0 I2 ) , (A1x)

we
2 2

A25 = 3 2XG3,X − 2G4,ϕ0 − 4XG4,ϕ0 X + GTele,I2 + 2H (3XG5,X + 2X G5,XX − 9GTele,Tvec I2 + 6GTele,T I2 )
+ 2XGTele,XI2 + H ϕ˙0 (4G4,X + 8XG4,XX − 4G5,ϕ0 − 4XG5,ϕ0 X + 3GTele,I2 I2 − 6GTele,XTvec

+ 4GTele,XT ) , (A1y)
1
A26 = − (G2,X + GTele,X ) + G3,ϕ0 − (2Ḣ + 3H 2 )(G4,X − G5,ϕ0 ) − 2H 3 ϕ˙0 (G5,X + XG5,XX )(H 2 + Ḣ)
2

ie
− (ϕ¨0 + 2H ϕ˙0 )(G3,X − 3G4,ϕ0 X ) + H 2 (5X(−2G4,XX + G5,ϕ0 X ) − 2X 2 (G5,ϕ0 XX + ϕ¨0 G5,XXX ))
+ 2H ϕ˙0 (−ḢG5,X − 3ϕ¨0 G4,XX + 2ϕ¨0 G5,ϕ0 X + X(−2G4,ϕ0 XX + G5,ϕ0 ϕ0 X − ḢG5,XX + ϕ¨0 (G5,ϕ0 XX
− 2G4,XXX ))) − X(G3,ϕ0 X − 2G4,ϕ0 ϕ0 X + 2Ḣ(2G4,XX − G5,ϕ0 X ) + ϕ¨0 (G3,XX − 2G4,ϕ0 XX )) ,

(A1z)

ev
1
− 6H 4 X(3G5,ϕ0 X + 2XG5,ϕ0 XX ) + G2,ϕ0 ϕ0 + GTele,ϕ0 ϕ0 + 3Ḣ(2G4,ϕ0 ϕ0 − GTele,ϕ0 I2 ) − 3H ϕ˙0 G2,ϕ0 X

A27 =
2
− 2G3,ϕ0 ϕ0 + 2XG3,ϕ0 ϕ0 X − 4XG4,ϕ0 ϕ0 ϕ0 X + GTele,ϕ0 X + GTele,ϕ0 ϕ0 I2 + Ḣ(4G4,ϕ0 X + 8XG4,ϕ0 XX
− 4G5,ϕ ϕ − 4XG5,ϕ ϕ X + 3GTele,ϕ I I − 6GTele,ϕ XT + 4GTele,ϕ XT ) − ϕ¨0 (G2,ϕ X − 2G3,ϕ ϕ

r
0 0 0 0 0 2 2 0 vec 0 0 0 0

+ GTele,ϕ0 X ) − 6H ϕ˙0 ϕ¨0 (G3,ϕ0 X − 3G4,ϕ0 ϕ0 X + X(G3,ϕ0 XX − 2G4,ϕ0 ϕ0 XX ) + GTele,ϕ0 XI2 )

+ 3H 2 (4G4,ϕ0 ϕ0 + 2X(−3G3,ϕ0 X + 6G4,ϕ0 ϕ0 X + G5,ϕ0 ϕ0 ϕ0 + 2X(−2G4,ϕ0 ϕ0 XX + G5,ϕ0 ϕ0 ϕ0 X ))
er
− 3GTele,ϕ I − 2Ḣ(3XG5,ϕ X + 2X 2 G5,ϕ XX − 9GTele,ϕ T I + 6GTele,ϕ T I ) − 2G4,ϕ X ϕ¨0
0 2 0 0 0 vec 2 0 2 0

+ (2G5,ϕ0 ϕ0 + 2X(−8G4,ϕ0 XX + 5G5,ϕ0 ϕ0 X + 2X(−2G4,ϕ0 XXX + G5,ϕ0 ϕ0 XX )) − 3GTele,ϕ0 I2 I2 )ϕ¨0 )

+ 2X(−G2,ϕ ϕ X + G3,ϕ ϕ ϕ − GTele,ϕ ϕ X − 3Ḣ(G3,ϕ X − 2G4,ϕ ϕ X + GTele,ϕ XI ) − ϕ¨0 (G2,ϕ XX
0 0 0 0 0 0 0 0 0 0 0 2 0
pe
− G3,ϕ0 ϕ0 X + GTele,ϕ0 XX )) − 2H 3 ϕ˙0 (9G4,ϕ0 X − 9G5,ϕ0 ϕ0 + 3ϕ¨0 G5,ϕ0 X + X(18G4,ϕ0 XX − 7G5,ϕ0 ϕ0 X
+ 2XG5,ϕ0 ϕ0 XX + (7G5,ϕ0 XX + 2XG5,ϕ0 XXX )ϕ¨0 )) ,

(A1aa)
1
A28 = (G2,X + GTele,X ) − G3,ϕ0 + X(G2,XX − G3,ϕ0 X + GTele,XX ) + 3H ϕ˙0 (G3,X − 3G4,ϕ0 X
2
+ X(G3,XX − 2G4,ϕ0 X ) + GTele,XI2 ) + H 3 ϕ˙0 (3G5,X + 7XG5,XX + 2X 2 G5,XXX )
ot

3 i
+ 3H 2 (G4,X − G5,ϕ0 + X(8G4,XX − 5G5,ϕ0 X ) + 2X 2 (2G4,XXX − G5,ϕ0 XX ) + GTele,I2 I2 ) . (A1bb)
2
Furthermore, the coefficients of the gauge invariant variables for the scalar quadratic action (54) are the following
tn

Ã1 = X G2,X − 2G3,ϕ0 + XGTele,X + 2X 2 G2,XX − G3,ϕ0 X + GTele,XX − 9H 2 GTele,Tvec + 6XH ϕ˙0 2G3,X

+ XG3,XX − 5G4,ϕ0 X − 2XG4,ϕ0 XX + 2GTele,XI2 + 5H 2 G5,X + 6H 2 − G4 + GTele,T + 7XG4,X

+ 4XGTele,XT + 6X(−G5,ϕ0 + GTele,I2 I2 − GTele,XTvec ) + 6X 2 H 2 4(4G4,XX + XG4,XXX ) − 9G5,ϕ0 X

+ 2H 3 X 2 13ϕ˙0 G5,XX + 2X ϕ˙0 G5,XXX + 3H ϕ˙0 GTele,I2 − 2G4,ϕ − 6H 2 (6GTele,Tvec I2 − 4GTele,T I2 )

rin

+ 18H 4 9GTele,Tvec Tvec − 12GTele,T Tvec + 4GTele,T T − 12X 3 H 2 G5,ϕ0 XX ,

(A2a)
1
Ã2 = (9GTele,Tvec + 6XGTele,J3 − 10XGTele,J5 + 4X(XGTele,J6 − GTele,J8 )) , (A2b)
9

X
Ã3 = 2 G4 − X(ϕ¨0 G5,X + G5,ϕ0 ) + 2GTele,Tvec − GTele,T + 4(XGTele,J6 − GTele,J8 − 3GTele,J3 )
9
ep

− 10XGTele,J5 , (A2c)
Ã4 = 6 X(2G4,X − G5,ϕ0 ) − G4 + GTele,T + 6H ϕ˙0 XG5,X + 6GTele,T I2 − 9GTele,Tvec I2

+ 9 XGTele,I2 I2 − GTele,Tvec + 2H 2 (9GTele,Tvec Tvec − 12GTele,T Tvec + 4GTele,T T ) ,

(A2d)
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 19
 
1 1
G2,ϕ0 ϕ0 + GTele,ϕ0 ϕ0 − ϕ¨0 (G2,ϕ0 X + GTele,ϕ0 X ) + 3Ḣ G4,ϕ0 ϕ0 − GTele,ϕ0 I2 + ϕ¨0 G3,ϕ0 ϕ0


Ã5 =
2 2

+ X − G2,ϕ0 ϕ0 X − GTele,ϕ0 ϕ0 X + G3,ϕ0 ϕ0 ϕ0 − 3Ḣ (G3,ϕ0 X − 2G4,ϕ0 ϕ0 X + GTele,ϕ0 XI2 ) − ϕ¨0 G2,ϕ0 XX

d

 h 3
− G3,ϕ0 ϕ0 X + GTele,ϕ0 XX + H ϕ˙0 − G2,ϕ0 X − 2G3,ϕ0 ϕ0 + GTele,ϕ0 X + GTele,ϕ0 ϕ0 I2 + Ḣ − 6G4,ϕ0 X


2
9

we
+ 6G5,ϕ0 ϕ0 − GTele,ϕ0 I2 I2 − 6GTele,ϕ0 XT + 9GTele,ϕ0 XTvec − 3ϕ¨0 G3,ϕ0 X − 3G4,ϕ0 ϕ0 X + GTele,ϕ0 XI2



 2

i
+ 3X − G3,ϕ0 ϕ0 X + 2G4,ϕ0 ϕ0 ϕ0 X + Ḣ − 4G4,ϕ0 XX + 2G5,ϕ0 ϕ0 X + ϕ¨0 − G3,ϕ0 XX + 2G4,ϕ0 ϕ0 XX

h 3
+ 3H 2 2G4,ϕ0 ϕ0 − (GTele,ϕ0 I2 + ϕ¨0 GTele,ϕ0 I2 I2 ) + Ḣ − 6GTele,ϕ0 T I2 + 9GTele,ϕ0 Tvec I2 + ϕ¨0 − G4,ϕ0 X


2
+ G5,ϕ0 ϕ0 + X − 3G3,ϕ0 X + 6G4,ϕ0 ϕ0 X + G5,ϕ0 ϕ0 ϕ0 − 3ḢG5,ϕ0 X − 8ϕ¨0 G4,ϕ0 XX + 5ϕ¨0 G5,ϕ0 ϕ0 X

ie

i h
+ 2X 2 − 2G4,ϕ0 ϕ0 XX + G5,ϕ0 ϕ0 ϕ0 X − ḢG5,ϕ0 XX − 2ϕ¨0 G4,ϕ0 XXX + ϕ¨0 G5,ϕ0 ϕ0 XX − H 3 ϕ˙0 3(3G4,ϕ0 X
− 3G5,ϕ0 ϕ0 + ϕ¨0 G5,ϕ0 X ) + X 18G4,ϕ0 XX − 7(G5,ϕ0 ϕ0 X − ϕ¨0 G5,ϕ0 XX ) + 2X 2 G5,ϕ0 ϕ0 XX

ev

i
+ ϕ¨0 G5,ϕ0 XXX − 3XH 4 3G5,ϕ0 X + 2XG5,ϕ0 XX ,


(A2e)
1 h
Ã6 = G2,X − 2G3,ϕ0 + 2X(GTele,XX + G2,XX − G3,ϕ0 X ) + 6H ϕ˙0 (G3,X + H 2 G5,X − 3G4,ϕ0 X − 2XG4,ϕ0 XX
2
+ GTele,XI2 ) + 2XH ϕ˙0 (2G3,XX + 7H 2 G5,XX ) + GTele,X + 6H 2 G4,X + 4X(2G4,XX + XG4,XXX )

r
i
− G5,ϕ0 − 5XG5,ϕ0 X − 2X 2 G5,ϕ0 XX + 4X 2 H 3 ϕ˙0 G5,XXX + 9H 2 GTele,I2 I2 ,


(A2f)
1
G2,X + GTele,X + G3,ϕ0 + ϕ¨0 (2XG4,ϕ0 XX + 3G4,ϕ0 X − H 2 G5,X − G3,X − XG3,XX ) + 5XH 2 (G5,ϕ0 X

er
Ã7 = −
2
− ϕ¨0 G5,XX − 2G4,XX ) + 2XH ϕ˙0 (−ḢG5,XX − 2ϕ¨0 G4,XXX − 2G4,ϕ0 XX + ϕ̈G5,ϕ0 XX + G5,ϕ0 ϕ0 X )
− 2XH 2 (H ϕ˙0 G5,XX + X ϕ¨0 G5,XXX + XG5,ϕ XX ) + 3H 2 (G5,ϕ − G4,X ) + 2Ḣ(G5,ϕ − G4,X )
0 0
pe
+ 2X Ḣ(G5,ϕ0 X − 2G4,XX ) + 2H ϕ˙0 (−G3,X − (H 2 + Ḣ)G5,X − 3ϕ¨0 G4,XX + 3G4,ϕ0 X + 2ϕ¨0 G5,ϕX )
+ X(2G4,ϕ0 ϕ0 X − G3,ϕ0 X ) , (A2g)
1 h
Ã8 = a2 X(4GTele,J8 + GTele,J5 + 3GTele,I2 I2 ) − 3GTele,Tvec + 6H H(9GTele,Tvec Tvec − 12GTele,T Tvec
3

i
+ 4GTele,T T ) + ϕ˙0 (−3GTele,Tvec I2 + 2GTele,T I2 ) , (A2h)
Ã9 = 3 − 12H 3 [9GTele,Tvec Tvec + 4(−3GTele,T Tvec + GTele,T T )] + 2H[2G4 + 4X 2 (−2G4,XX + G5,ϕ0 X )
ot

+ 3GTele,Tvec − 2GTele,T + 2X(−4G4,X + 3G5,ϕ0 − 3GTele,I2 I2 + 3GTele,XTvec − 2GTele,XT )]

− 2H 2 ϕ˙0 (5XG5,X + 2X 2 G5,XX − 27GTele,Tvec I2 + 18GTele,T I2 ) − ϕ˙0 [−2G4,ϕ0 + GTele,I2 + 2X(G3,X
tn


− 2G4,ϕ0 X + GTele,XI2 )] , (A2i)
2
Ã10 = 18(G4 + GTele,Tvec − GTele,T ) + X(−36G4,X + 18G5,ϕ0 + 4GTele,J8 − 4XGTele,J6 + 10GTele,J5
9
+ 3GTele,J3 − 18H ϕ˙0 G5,X ) ,

(A2j)
2
Ã11 = G2,ϕ0 + GTele,ϕ0 + 6H [G4,ϕ0 + X(−4G4,ϕ0 X + 3G5,ϕ0 ϕ0 − 4XG4,ϕ0 XX + 2XG5,ϕ0 ϕ0 X ) + 3GTele,ϕ0 Tvec
rin

− 2GTele,ϕ0 T ] − 2X(G2,ϕ0 X − G3,ϕ0 ϕ0 + GTele,ϕ0 X ) − 2H 3 X ϕ˙0 (5G5,ϕ0 X + 2XG5,ϕ0 XX ) − 6H ϕ˙0 [−G4,ϕ0 ϕ0

+ X(G3,ϕ0 X − 2G4,ϕ0 ϕ0 X ) + GTele,ϕ0 I2 ] , (A2k)
Ã12 = − H (30XG5,X + 40X 2 G5,XX + 8X 3 G5,XXX − 54GTele,Tvec I2 + 36GTele,T I2 ) + 3H[−2G4,ϕ0 + GTele,I2
 3

+ 2X(3G3,X − 8G4,ϕ X + 2X(G3,XX − 2G4,ϕ XX ) + 3GTele,XI )] + 6H 2 ϕ˙0 [3G4,X + X(12G4,XX

ep

0 0 2

+ 4XG4,XXX − 7G5,ϕ0 X − 2XG5,ϕ0 XX ) − 3(G5,ϕ0 − GTele,I2 I2 + GTele,XTvec ) + 2GTele,XT ] + ϕ˙0 [G2,X


− 2G3,ϕ0 + GTele,X + 2X(G2,XX − G3,ϕ0 X + GTele,XX )] , (A2l)
= 2G4,ϕ0 − 2X G3,X + H (3G5,X + 2XG5,XX ) − 2G4,ϕ0 X − GTele,I2 − H ϕ˙0 4(G4,X − G5,ϕ0
2


Ã13


+ X(2G4,XX − G5,ϕ0 X )) + GTele,J3 , (A2m)
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 20

Ã9
Ã14 = , (A2n)
3a   
Ã15 = 3ϕ˙0 2G3,ϕ0 − G2,X − GTele,X − 2G4,ϕ0 ϕ0 + GTele,ϕ0 I2 + 6H 2 ϕ˙0 3(G5,ϕ0 − G4,X ) + 2X(2G5,ϕ0 X

d
   
− 3G4,XX ) + 3H 2G4,ϕ0 + 4GTele,ϕ0 T − 6GTele,ϕ0 Tvec − 3GTele,I2 + 6XH 2(5G4,ϕ0 X − G5,ϕ0 ϕ0 ) (A2o)

− 3(G3,X + H 2 G5,X ) − 12X 2 H 3 G5,XX ,

we
Ã16 = 3 2XG3,X − 2G4,ϕ0 + GTele,I2 + 6H ϕ˙0 2G4,X − 2G5,ϕ0 − 3GTele,XTvec + 2GTele,XT

+ 6X GTele,XI2 − 2G4,ϕ0 X + 9H ϕ˙0 GTele,I2 I2 + 12XH ϕ˙0 2G4,XX − G5,ϕ0 X

 
+ 6H 2 6GTele,T I2 − 9GTele,Tvec I2 + X(3G5,X + 2XG5,XX ) , (A2p)
Ã17 = 4H ϕ˙0 G5,ϕ0 − G4,X − ϕ¨0 G5,X − X(ϕ¨0 G5,XX + G5,ϕ0 X ) + 4ϕ¨0 XG5,ϕ0 X + G5,ϕ0 − G4,X − 2XG4,XX

ie

+ 4G4,ϕ0 − 2GTele,I2 − 4X ḢG5,X + 4X G5,ϕ0 ϕ0 − 2G4,ϕ0 X − H 2 G5,X + H ϕ˙0 GTele,J3 ,

(A2q)
h
Ã18 = a 4 G4 − GTele,T + X(G5,ϕ0 − 2G4,X ) + 4H ϕ˙0 − XG5,X + 9GTele,Tvec I2 − 6GTele,T I2

ev

i
+ 6 GTele,Tvec − XGTele,I2 I2 + 12H 2 12GTele,T Tvec − 9GTele,Tvec Tvec − 4GTele,T T ,


(A2r)
h
Ã19 = −a ϕ˙0 G2,X − 2G3,ϕ0 + 2G4,ϕ0 ϕ0 + GTele,X − GTele,ϕ0 I2 + H 3GTele,I2 − 2G4,ϕ0 + 6GTele,ϕ0 Tvec

− 4GTele,ϕ0 T + 6H 2 ϕ˙0 G4,X − G5,ϕ0 + 2XH 3G3,X + 6H ϕ˙0 G4,XX − 10G4,ϕ0 X + 2G5,ϕ0 ϕ0

r

i
+ 2XH 2 3HG5,X + 2XHG5,XX − 4ϕ˙0 G5,ϕ0 X , (A2s)
h
Ã20 = −a 2X 2G4,ϕ0 X − G3,X − GTele,XI2 − GTele,I2 + 2G4,ϕ0 + H ϕ˙0 4(G5,ϕ0 − G4,X − 2XG4,XX

er
+ XG5,ϕ0 X − GTele,XT ) + 3(2GTele,XTvec − GTele,I2 I2 ) + H 2 6(3GTele,Tvec I2 − 2GTele,T I2 − XG5,X )



i
− 4X 2 G5,XX . (A2t)
pe
For the quadratic vector action (58) we have the following coefficients
1 h i
C1 = 2GTele,Tax + 3[3(G4 − GTele,T ) + X(−6G4,X + 3G5,ϕ + 2GTele,J10 + 6GTele,J8 + 3GTele,J5 − 3H ϕ̇G5,X )] ,
18
(A3a)
X
C2 = GTele,Tvec + [−2GTele,J8 − 5GTele,J5 − 6GTele,J3 + 2XGTele,J6 ] , (A3b)
ot

18
C3 = 2(G4 − GTele,T ) + X[−4G4,X + 2G5,ϕ + 4GTele,J8 + GTele,J5 − 2H ϕ̇G5,X ] , (A3c)
1
C4 = [3X(4GTele,J10 + 3GTele,J5 ) + 4GTele,Tax ] , (A3d)
9
tn

2X
C5 = GTele,Tvec + [−2GTele,J8 − 5GTele,J5 + 3GTele,J3 + 2XGTele,J6 ] , (A3e)
9
1
C6 = [2GTele,Tax + 9(G4 − GTele,T ) − 3X(6G4,X − 3G5,ϕ + 4GTele,J10 − 6GTele,J5 + 3H ϕ̇G5,X )] , (A3f)
18
1
C7 = − [9(G4 − GTele,T ) − 2GTele,Tax + 3X(−6G4,X + 3G5,ϕ − 3H ϕ̇G5,X + GTele,J10 + 3GTele,J5 )] , (A3g)
rin

9
C8 = ϕ̇GTele,I2 + 2H (−3GTele,Tvec + 2GTele,T ) , (A3h)
  
C9 = 2 G4 + X −2G4,X + G5,ϕ − H ϕ̇G5,X , (A3i)
1
C10 = (18 (−GTele,Tvec + GTele,T ) + X (−4GTele,J8 + 4XGTele,J6 − 10GTele,J5 − 3GTele,J3 )) , (A3j)
ep

9
2
C11 = − (3X (GTele,J10 + 3GTele,J5 ) − 2GTele,Tax − 9GTele,T ) . (A3k)
9
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 21

While the coefficients C˜1 − C˜9 for the gauge invariant quadratic vector action (59) are
1
C˜1 = (9GTele,Tvec + X(6GTele,J3 − 10GTele,J5 + 4XGTele,J6 − 4GTele,J8 )) , (A4a)
9

d
1  
C˜2 = 9(G4 − GTele,T ) + 2GTele,Tax + X(9G5,ϕ − 18G4,X − 12GTele,J10 + 18GTele,J5 − 9H ϕ̇G5,X ) , (A4b)
9
4
C˜3 = −2X Ḣ ϕ̇GTele,I2 J10 − 6X Ḣ ϕ̇GTele,I2 J5 + Ḣ ϕ̇GTele,Tax I2 + 6Ḣ ϕ̇GTele,T I2 − 8XH ḢGTele,T J10

we
3
16
+ 12XH ḢGTele,Tvec J10 − 24XH ḢGTele,T J5 + 36XH ḢGTele,Tvec J5 + H ḢGTele,T Tax − 8H ḢGTele,Tax Tvec
3
4
− 36H ḢGTele,T Tvec + 24H ḢGTele,T T − 2XH ϕ̈GTele,I2 J10 − 6XH ϕ̈GTele,I2 J5 + H ϕ̈GTele,Tax ,I2
3
4 8

ie
+ 6H ϕ̈GTele,T I2 − XHGTele,J10 − 4XHGTele,J5 + HGTele,Tax − 6HGTele,Tvec + ϕ̇GTele,I2
3 9
2 2 2
− X ϕ̇GTele,ϕJ10 − X ϕ̇ϕ̈GTele,XJ10 − ϕ̇ϕ̈GTele,J10 − 2X ϕ̇GTele,ϕJ5 − 2X ϕ̇ϕ̈GTele,XJ5 − 2ϕ̇ϕ̈GTele,J5
3 3 3
4 4

ev
+ ϕ̇ϕ̈GTele,XTax + ϕ̇GTele,ϕTax + 2ϕ̇ϕ̈GTele,XT + 2ϕ̇GTele,ϕT + 8HGTele,T , (A4c)
9 9
1  
C˜4 = − 9(G4 − GTele,T ) − 2GTele,Tax + X(9G5,ϕ − 18G4,X + 3GTele,J10 + 9GTele,J5 − 9H ϕ̇G5,X ) , (A4d)
9
1
C˜5 = (18GTele,Tvec + 4GTele,Tax + X(4GTele,J8 − 4XGTele,J6 − 8GTele,J5 + 3GTele,J3 − 6GTele,J10 )) , (A4e)
9

r
1
C˜6 = (4GTele,Tax + X(9GTele,J5 + 12GTele,J10 )) , (A4f)
9
C˜7 = C˜6 ,

C˜8 =
1
er
(18GTele,Tvec − X(6GTele,J3 + 5GTele,J5 + 2GTele,J8 − 2XGTele,J6 )) ,
(A4g)

(A4h)
18
1  
C˜9 = 9(G4 − GTele,T ) + 2GTele,Tax + X(9G5,ϕ − 18G4,X + 6GTele,J10 + 9GTele,J5 + 18GTele,J8 − 9H ϕ̇G5,X ) .
pe
18
(A4i)

Appendix B: Gauge Choices

Section III provides a system of equations for an SVT decomposition without imposing any gauge choices. Fur-
ot

thermore, it was shown that the perturbation of the BDLS action can be expressed in terms of gauge invariant
quantities given by Eq. (53). Alternatively, through the gauge transformation in Eq. (51), the gauge invariant quan-
tities {σ, ui , hij } are identified but non-gauge invariant variables dependent on each other can be lumped together to
form the following groups: {δϕ, Φ, ψ, β}, {B, E}, {vi , wi , Vi }. A gauge choice is done by setting one from each group
to vanish. Here, we list some of the commonly used gauge choices within the scalar and vector sectors.
tn

All gauge choices can be obtained through the original action perturbations given by Eqs (52, 55, 58, 61). These
calculations cannot be done through the minimized gauge invariant action explored in Sec. IV, as certain gauge choices
may result in undefined regions with a null denominator. We present a simplified version of these results obtained
through integration by parts and collecting like terms.
rin

1. Scalars

In the scalar sector, we explore the flat, unitary, Newtonian and synchronous gauge. It is known that in the syn-
chronous gauge, residual gauge freedom remains due to spatial coordinate transformations that are time-independent.
ep

These correspond to arbitrary spatial translations or mode-dependent redefinitions of spatial coordinates. To fully fix
the gauge, additional conditions are required, such as aligning the perturbation center of mass with a specific fluid or
setting one metric potential to zero. Properly addressing this residual freedom is crucial to avoid spurious results and
ensure consistent interpretation of physical observables in theoretical and observational cosmology [30, 31, 63, 67, 68].
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 22

a. Flat Gauge

The flat gauge is characterized by ψ = 0 and E = 0, for which Eq. (52) simplifies to

d
" ! !
∇2 β ∇2 B ∇2 β ∇2 B ∇2 Φ Φ∇2 β
Z
Flat 3 3 ˙
SS = dt d xa δϕ F1 Φ − F2 + F3 + δϕ F4 Φ − F5 + F6 − F2 2 − F7 (B1)
a a a a a a

we

2 !2 #
2
Φ∇2 B 2 2 (∇δϕ) ∇2 B ∇Φ ˙ 2
+ F8 + F9 Φ + F10 δϕ + F11 + F12 + F13 ∇β̇ − + F14 δϕ
a a2 a2 a

where
 
F1 =ϕ˙0 X (−2GTele,XX − 2G2,XX + 2G3,ϕ0 X ) − GTele,X − G2,X + 2G3,ϕ0 + H X − 18GTele,XI2 − 18G3,X

ie
+ 48G4,ϕ0 X − 3GTele,I2 + X 2 (24G4,ϕ0 XX − 12G3,XX ) + 6G4,ϕ0 + H 2 ϕ˙0 − 18GTele,I2 I2 − 12GTele,XT

+ 18GTele,XTvec + X 2 (12G5,ϕ0 XX − 24G4,XXX ) + X (42G5,ϕ0 X − 72G4,XX ) − 18G4,X + 18G5,ϕ0

ev
+ H 3 − 36GTele,T I2 + 54GTele,Tvec I2 − 8X 3 G5,XXX − 40X 2 G5,XX − 30XG5,X ,


(B2a)
 
F2 =2G4,ϕ0 − GTele,I2 − X (2G3,X − 4G4,ϕ0 X ) − H ϕ˙0 GTele,J3 + X (8G4,XX − 4G5,ϕ0 X ) + 4G4,X − 4G5,ϕ0
− H 2 4X 2 G5,XX + 6XG5,X ,


(B2b)
˙ 2

r
F3 =F2 + H ϕ0 (−3GTele,I I + GTele,J − 4GTele,XT + 6GTele,XT ) − 6H (2GTele,T I − 3GTele,T I )
2 2 3 vec 2 vec 2

− 2XGTele,XI2 , (B2c)
F4 = − 2XGTele,ϕ0 X + GTele,ϕ0 − 2XG2,ϕ0 X + G2,ϕ0 + 2XG3,ϕ0 ϕ0
er + H ϕ˙0 X − 6GTele,ϕ0 I2 − 6XG3,ϕ0 X
+ 12XG4,ϕ0 ϕ0 X + 6G4,ϕ0 ϕ0 + H 2 X − 12GTele,ϕT + 18GTele,ϕ0 Tvec − 24X 2 G4,ϕ0 XX − 24XG4,ϕ0 X

+ 6G4,ϕ + 12X 2 G5,ϕ ϕ X + 18XG5,ϕ ϕ + H 3 ϕ˙0 X −4X 2 G5,ϕ XX − 10XG5,ϕ X ,

0 0 0 0 0 0 0
(B2d)
F5 ¨
=F6 − H Ḣ (6XGTele,I2 J3 + 12GTele,T I2 − 18GTele,Tvec I2 ) + ϕ0 (3GTele,I2 I2 + 2XGTele,XJ3 + GTele,J3 )
pe
+ 2GTele,I2 + 2XGTele,ϕ0 J3 − 4GTele,ϕ0 T + 6GTele,ϕ0 Tvec − H 2 ϕ˙0 3 4ḢGTele,T J3 − 6ḢGTele,Tvec J3


 
+ ϕ¨0 GTele,I2 J3 + 2GTele,J3 − ϕ˙0 3ḢGTele,I2 I2 + ḢGTele,J3 + ϕ¨0 GTele,XI2 ,



(B2e)
=ϕ˙0 − GTele,ϕ0 I2 + GTele,X + G2,X − 2G3,ϕ0 + 2G4,ϕ0 ϕ0 + H 3GTele,I2 − 4GTele,ϕ0 T + 6GTele,ϕ0 Tvec


F6
+ X (6G3,X − 20G4,ϕ0 X + 4G5,ϕ0 ϕ0 ) − 2G4,ϕ0 + H 2 ϕ˙0 X (12G4,XX − 8G5,ϕ0 X ) + 6G4,X − 6G5,ϕ0



ot

+ H 3 4X 2 G5,XX + 6XG5,X ,


(B2f)
=ϕ˙0 − GTele,I2 + X (4G4,ϕ0 X − 2G3,X ) + 2G4,ϕ0 + H − 4GTele,T + 6GTele,Tvec + X (8G5,ϕ0 X − 16G4,XX )
2


F7
+ X (12G5,ϕ0 − 16G4,X ) + 4G4 + H 2 ϕ˙0 −4X 2 G5,XX − 10XG5,X ,



(B2g)
tn

2 ˙


F8 =F7 + XH 12GTele,I2 I2 + 8GTele,XT − 12GTele,XTvec + H ϕ0 (36GTele,T I2 − 54GTele,Tvec I2 )
+ H 3 − 144GTele,T Tvec + 48GTele,T T + 108GTele,Tvec Tvec + 2X ϕ˙0 GTele,XI2 ,


(B2h)

F9 =X (GTele,X + G2,X − 2G3,ϕ0 ) + X 2 (2GTele,XX + 2G2,XX − 2G3,ϕ0 X ) + H ϕ˙0 X (12GTele,XI2 + 12G3,X
 
−30G4,ϕ0 X ) + 3GTele,I2 + X 2 (6G3,XX − 12G4,ϕ0 XX ) − 6G4,ϕ0 + H 2 X (36GTele,I2 I2 + 24GTele,XT
rin

−36GTele,XTvec + 42G4,X − 36G5,ϕ0 ) + 6GTele,T − 9GTele,Tvec + X 3 (24G4,XXX − 12G5,ϕ0 XX )


+ X 2 (96G4,XX − 54G5,ϕ0 X ) − 6G4 + H 3 ϕ˙0 72GTele,T I2 − 108GTele,Tvec I2 + 4X 3 G5,XXX + 26X 2 G5,XX
+30XG5,X ) + H 4 (−216GTele,T Tvec + 72GTele,T T + 162GTele,Tvec Tvec ) , (B2i)
ep

1
= − 3ḢGTele,ϕ0 I2 − ϕ¨0 GTele,ϕ0 X + GTele,ϕ0 ϕ0 − G2,ϕ0 X ϕ¨0 + G2,ϕ0 ϕ0 + G3,ϕ0 ϕ0 ϕ̈ + 6ḢG4,ϕ0 ϕ0


F10
2
+ X − 3ḢGTele,ϕ0 XI2 − ϕ¨0 GTele,ϕ0 XX − GTele,ϕ0 ϕ0 X + G3,ϕ0 ϕ0 ϕ0 − G2,ϕ0 XX ϕ¨0 − G2,ϕ0 ϕ0 X − 3G3,ϕ0 X Ḣ
9
+ G3,ϕ0 ϕ0 X ϕ¨0 + 6G4,ϕ0 ϕ0 X Ḣ + H ϕ˙0 − ḢGTele,ϕ0 I2 I2 − 6ḢGTele,ϕ0 XT + 9ḢGTele,ϕ0 XTvec


2
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 23

3 3 3G2,ϕ0 X
− 3ϕ¨0 GTele,ϕ0 XI2 − GTele,ϕ0 ϕ0 I2 − GTele,ϕ0 X − + X − 3G3,ϕ0 XX ϕ¨0 − 3G3,ϕ0 ϕ0 X
2 2 2
+ 6G4,ϕ0 ϕ0 XX ϕ¨0 − 12G4,ϕ0 XX Ḣ + 6G4,ϕ0 ϕ0 ϕ0 X + 6G5,ϕ0 ϕ0 X Ḣ − 3G3,ϕ0 X ϕ¨0 + 3G3,ϕ0 ϕ0 − 6G4,ϕ0 X Ḣ

9 9

d
+ 9G4,ϕ0 ϕ0 X ϕ¨0 + 6G5,ϕ0 ϕ0 Ḣ + H 2 − 18ḢGTele,ϕ0 T I2 + 27ḢGTele,ϕ0 Tvec I2 − ϕ¨0 GTele,ϕ0 I2 I2 − GTele,ϕ0 I2


2 2
+ X − 9G3,ϕ0 X − 24G4,ϕ0 XX ϕ¨0 + 18G4,ϕ0 ϕ0 X − 9G5,ϕ0 X Ḣ + 15G5,ϕ0 ϕ0 X ϕ¨0 + 3G5,ϕ0 ϕ0 ϕ0

we
+ X 2 − 12G4,ϕ0 XXX ϕ¨0 − 12G4,ϕ0 ϕ0 XX − 6G5,ϕ0 XX Ḣ + 6G5,ϕ0 ϕ0 XX ϕ¨0 + 6G5,ϕ0 ϕ0 ϕ0 X − 3G4,ϕ0 X ϕ¨0

+ 6G4,ϕ0 ϕ0 + 3G5,ϕ0 ϕ0 ϕ¨0 + H 3 Ḣ ϕ˙0 X − 18G4,ϕ0 XX − 7G5,ϕ0 XX ϕ¨0 + 7G5,ϕ0 ϕ0 X − 9G4,ϕ0 X

 
+ X 2 −2G5,ϕ0 XXX ϕ¨0 − 2G5,ϕ0 ϕ0 XX − 3G5,ϕ0 X ϕ¨0 + 9G5,ϕ0 ϕ0 + H 4 −6X 2 G5,ϕ0 XX − 9XG5,ϕ0 X



(B2j)
1 
= − GTele,X + G2,X − G3,X ϕ¨0 + G3,ϕ0 + X −G3,XX ϕ¨0 − G3,ϕ0 X − 4G4,XX Ḣ + 2G4,ϕ0 XX ϕ¨0


F11

ie
2  
+2G4,ϕ0 ϕ0 + 2G5,ϕ0 X Ḣ + H ϕ˙0 − 2G3,X + X −4G4,XXX ϕ¨0 − 4G4,ϕ0 XX − 2G5,XX Ḣ + 2G5,ϕ0 XX ϕ¨0
+2G5,ϕ0 ϕ0 X ) − 6G4,XX ϕ¨0 + 6G4,ϕ0 X − 2G5,X Ḣ + 4G5,ϕ0 X ϕ¨0 − 2G4,X Ḣ + 3G4,ϕ0 X ϕ¨0 + 2G5,ϕ0 Ḣ

ev
    
+ H 2 X −10G4,XX − 5G5,XX ϕ¨0 + 5G5,ϕ X − 3G4,X + X 2 −2G5,XXX ϕ¨0 − 2G5,ϕ XX − G5,X ϕ¨0
0 0

+3G5,ϕ0 ) + H (−2XG5,XX − 2G5,X ) ϕ˙0 ,

3
(B2k)
X
F12 = (3GTele,I2 I2 + GTele,J5 + 4GTele,J8 ) − GTele,Tvec + H ϕ˙0 (4GTele,T I2 − 6GTele,Tvec I2 )
3

r
+ H 2 (−24GTele,T Tvec + 8GTele,T T + 18GTele,Tvec Tvec ) , (B2l)
1  
F13 = 9GTele,Tvec + X (6GTele,J3 − 10GTele,J5 − 4GTele,J8 ) + 4X 2 GTele,J6 ,
er (B2m)
9
1 h
F14 = G2,X + 2 (X(G2,XX − G3,ϕX ) − G3,ϕ ) + 6H 2 G4,X − G5,ϕ + X(G4,XX − 5G5,ϕX ) + 2X 2 (2G4,XXX
2
−G5,ϕXX ) + 23 GTele,I2 I2 + GTele,X + 2XGTele,XX + 2H 3 ϕ̇ (3G5,X + X(7G5,XX + 2XG5,XXX ))


pe
i
+ 6H ϕ̇ (G3,X − 3G4,ϕX + GTele,XI2 + X(G3,XX − 2G4,ϕXX )) (B2n)

where F12 and F13 are purely teleparallel coefficients. Moreover, difference between coefficients of β and B quantities
stem from the teleparallel contribution of the BDLS action. For this case, only δϕ and β appear to be temporally
dynamical with Φ and B being the auxiliary modes.
Following a minimization through obtaining field equations for auxiliary modes Φ and B, and diagionalization of
the kinetic matrix, the final action can be given by
ot

"
dk 3
Z   
Flat 1 ˙ 2 1 ˙ 2

˙
  
SS = dt 3 F̃ 1 − F̃ 2 Ψ 1 + F̃ 3 − F̃ 4 Ψ2 + F̃5 − F̃7 Ψ1 Ψ2 + F̃6 − F̃7 Ψ̇1 Ψ2
(2π) 2 2 2
tn

#
+ F̃8 Ψ̇21 + F̃9 Ψ̇22 , (B3)

resulting in two propagating DoFs where coefficients F̃i are presented in Ref. [35]. The number of DoFs for subclasses
given in Table I is remains the same in the flat gauge.
rin
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 24

b. Unitary Gauge

Next, we consider the unitary gauge where the assumptions of δϕ = 0 and E = 0 are implemented in Eq. (52),
resulting in

d
" ! !
∇2 B (∇ψ)2 ∇2 B ∇2 β
Z
Unitary 3 3 2 2
SS = dt d x a U1 2ψ̇ − 3ψ̇ + U2 + U 3 Φ + 2U4 Φ 3ψ̇ − + U 5 Φ

we
a a2 a a
! !
2 #
∇2 β ∇2 ψ (∇Φ)2  2 Φ∇2 β̇ ∇2 B 2
+ U6 ψ̇ +Φ 2 + U7 + ∇β̇ + 2 + U8 + U9 ψ∇ β ,
a a a2 a a2
(B4)

where

ie
U1 =2G4 − GTele,T + 3GTele,Tvec + X (−4G4,X + 2G5,ϕ0 − 3GTele,I2 I2 ) + H ϕ˙0 (−2XG5,X − 12GTele,T I2
+18GTele,Tvec I2 ) + H 2 (−24GTele,T T + 72GTele,T Tvec − 54GTeleTvec Tvec ) , (B5a)

ev
4 X
U2 =2G4 − 2GTele,T + 4GTele,Tvec + X 2 GTele,J6 − 2G5,ϕ0 + 2G5,X ϕ¨0 + 12GTele,J3 + 10GTele,J5 + 4GTele,J8 ,


9 9
(B5b)
U3 =X (GTele,X + G2,X − 2G3,ϕ0 ) + X 2 (2GTele,XX + 2G2,XX − 2G3,ϕ0 X )
+ H ϕ˙0 X (12GTele,X,I2 + 12G3,X − 30G4,ϕ0 X ) + 3GTele,I2 + X 2 (6G3,XX − 12G4,ϕ0 XX ) − 6G4,ϕ0

r
+ H 2 (X (36GTele,I2 I2 + 24GTele,XT − 36GTele,XTvec + 42G4,X − 36G5,ϕ0 ) + 6GTele,T − 9GTele,Tvec
+X 3 (24G4,XXX − 12G5,ϕ0 XX ) + X 2 (96G4,XX − 54G5,ϕ0 X ) − 6G4 + H 3 ϕ˙0 (72GTele,T I2 − 108GTele,Tvec I2
er

+4X 3 G5,XXX + 26X 2 G5,XX + 30XG5,X + 18H 4 (−12GTele,T Tvec + 4GTele,T T + 9GTele,Tvec Tvec ) ,


(B5c)
 
1
U4 =ϕ˙0 X (−GTele,XI2 − G3,X + 2G4,ϕ0 X ) − GTele,I2 + G4,ϕ0 + H (X (−6GTele,I2 I2 − 4GTele,XT + 6GTele,XTvec
2
pe
−8G4,X + 6G5,ϕ0 ) − 2GTele,T + 3GTele,Tvec + X 2 (4G5,ϕ0 X − 8G4,XX ) + 2G4 + H 2 ϕ˙0 (−18GTele,T I2

+27GTele,Tvec I2 − 2X 2 G5,XX − 5XG5,X + H 3 (72GTele,T Tvec − 24GTele,T T − 54GTele,Tvec Tvec ) ,

(B5d)

U5 =2H 2G4 + 2X(−4G4,X + 3G5,ϕ + 2X(−2G4,XX + G5,ϕX )) + 3GTele,Tvec − 2GTele,T − HX ϕ̇(5G5,X

+ 2XG5,XX ) − ϕ̇(−2G4,ϕ + 2X(G3,X − 2G4,ϕX ) + GTele,I2 ) , (B5e)
  1
U6 =4 GTele,T − GTele,Tvec − G4 + XG5X H ϕ˙0 − X (6GTele,J3 + 20GTele,J5 + 8GTele,J8 − 72G4,X + 36G5,ϕ0 )
ot

9
8 2
+ X GTele,J6 , (B5f)
9
X 4
(6GTele,J3 − 10GTele,J5 − 4GTele,J8 ) + X 2 GTele,J6 ,
tn

U7 =GTele,Tvec + (B5g)
9 9
X
U8 = − GTele,Tvec + (3GTele,I2 I2 + GTele,J5 + 4GTele,J8 ) + H ϕ˙0 (4GTele,T I2 − 6GTele,Tvec I2 )
3
+ H 2 (−24GTele,T Tvec + 8GTele,T T + 18GTele,Tvec Ttvec ) , (B5h)
1h ˙ 
= ϕ0 − 18GTele,I2 + X ϕ¨0 (−6GTele,XJ3 − 20GTele,XJ5 + 16GTele,J6 − 8GTele,XJ8 ) − 6GTele,ϕ0 J3
rin

U9
9
−20GTele,ϕ0 J5 − 8GTele,ϕ0 J8 ) + ϕ¨0 (−6GTele,J3 − 20GTele,J5 − 8GTele,J8 + 36GTele,XT − 36GTele,XTvec )
 
+ X 2 8ϕ¨0 GTele,XJ6 + 8GTele,ϕ0 J6 + 36GTele,ϕ0 T − 36GTele,ϕ0 Tvec + H Ḣ X (−72GTele,T J3

+108GTele,Tvec J3 − 240GTele,T J5 + 360GTele,Tvec J5 − 96GTele,T J8 + 144GTele,Tvec J8 ) + X 2 (96GTele,T J6

ep

 
−144GTele,Tvec J6 ) − 1080GTele,T Tvec + 432GTele,T T + 648GTele,Tvec Tvec + X ϕ¨0 (−18GTele,I2 J3
−60GTele,I2 J5 − 24GTele,I2 J8 ) − 12GTele,J3 − 40GTele,J5 − 16GTele,J8 ) + 36GTele,Tvec
 
+ X 2 24ϕ¨0 GTele,I2 J6 + 16GTele,J6 + ϕ¨0 (108GTele,T I2 − 108GTele,Tvec I2 ) + Ḣ ϕ˙0 X (−18GTele,I2 J3


Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 25

i
−60GTele,I2 J5 − 24GTele,I2 J8 ) + 24X 2 GTele,I2 J6 + 108GTele,T I2 − 108GTele,Tvec I2 . (B5i)

for which {U7 , U8 , U9 } are coefficients arising from the purely teleparallel sector of the gravitational theory. The
unitary gauge has been favoured in scalar-tensor theories, so the standard Horndeski result given in Ref. [61] can be

d
obtained by taking the limit GTele → 0 limit. First and foremost, the coefficients {U7 , U8 , U9 } completely disappear,
eliminating in particularly the contribution proportional to k 4 (upon applying Fourier transformation). Next, in
metric-based theories, there is no distinction between the off-diagional terms since the metric is said to be symmetric.

we
The relationship between tetrad and metric given by Eq. (1) indicates that β and B can be combined to have a
single off-diagonal variable B = −B + β. Provided these modifications are implemented, the Horndeski gravity result
obtained in Ref. [61] is attained. For the general BDLS action, the final action is of the form
"
d3 k
Z   
Unitary 1 ˙ 2 1 ˙ 2

˙
  
SS = dt 3 Ũ 1 − Ũ2 Ψ 1 + Ũ3 − Ũ4 Ψ2 + Ũ5 − Ũ7 Ψ1 Ψ2 + Ũ6 − Ũ7 Ψ̇1 Ψ2
2 2

ie
(2π) 2
#
+ Ũ8 Ψ̇21 + Ũ9 Ψ̇22 , (B6)

ev
where the procedure and details of coefficients Ũi for i ∈ [1, 9] have been included in Ref. [35]. In-line with the
gauge-invariant and flat gauge, only two dynamical modes propagate following a diagonalization of the kinetic term.

c. Newtonian Gauge

r
The Newtonian / longitudinal gauge has been commonly used in teleparallel gravity, where E = 0 and β = B. In
er
Ref. [59], the gauge choice was further restricted by imposing β = 0, leading to potentially overfixing the gauge. The
action transforms to
"  2
∇2 β
Z
Newtonian 3 3 β 
SS = dt d x a A1 2
+ 2 (A2 + A6 )∇2 Φ + (A3 + A12 )∇2 ψ̇ + (A4 + A8 )∇2 δϕ
a a
pe
!2
2 ˙
 β 2 β̇ 2 2

∇β̇
+(A5 + A9 )∇ δϕ + A10 2 ∇ ψ + 2 A7 ∇ Φ + A11 ∇ ψ + A13
a a a
2
∇2 ψ ∇2 δϕ
   
+ Φ A14 2 + A15 ψ̇ + A16 δϕ + A17 2 + A18 δϕ ˙ + A19 ∇Φ + A20 Φ2
a a a
2 2
ot

 
∇ψ ∇ψ∇δϕ ˙ ψ̇ + A26 ∇δϕ
+ A21 + A22 ψ̇ 2 + A23 + A 24 δϕψ̇ + A25 δϕ
a a2 a
#
+ A27 δϕ2 + A28 δϕ ˙2 . (B7)
tn

with very minor modifications to the original action (52). Here, {δϕ, β, ψ} are all dynamical modes and only Φ is
non-dynamical. Varying with the latter mode and diagionalizing the kinetic matrix by relabelling the dynamical
modes in terms of new mode {Ψ1 , Ψ3 , Ψ3 }, the action becomes of the form
rin

dk 3
Z     
SSNewtonian = dt 3 Ñ1 Ψ21 + Ñ2 Ψ22 + Ñ3 Ψ23 + Ψ1 Ñ4 Ψ2 + Ñ6 Ψ3 + Ñ7 Ψ̇3 + Ñ9 Ψ̇2 + Ψ2 Ñ5 Ψ3 + Ñ8 Ψ̇3
(2π) 2

2 2 2
+ Ñ10 Ψ̇1 + Ñ11 Ψ̇2 + Ñ12 Ψ̇3 , (B8)
ep

where coefficients Ñi for i ∈ [1, 12] are detailed in Ref. [35]. The longitudinal mode suggests an additional dynamical
scalar mode for the general BDLS model, but further investigation would be required as discrepancy in the number
of DoFs does not occur for widely used subclasses of BDLS. For the particular case of β = 0, the number reduced to
two.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 26

d. Synchronous Gauge

The synchronous gauge is obtained by setting Φ = 0 and β = B and very popular in numerical studies while it is
the one mostly used in Boltzmann solvers like CLASS and CAMB. The action for the gauge choice reads

d
!2  
β − aĖ 
"
2
∇ −
Z
(β a Ė)  β 
SSSynchronous = dtd3 xa3 A1 + A ∇ 2
ψ̇ + A ∇ 2
δϕ + A ∇ 2 ˙
δϕ + ˙
A8 ∇2 δϕ + A9 ∇2 δϕ

we
3 4 5
a2 a2 a2
!2  2
2 2
 β̇ 2 ∇β̇ ∇ψ ∇ψ∇δϕ
+A10 ∇ ψ + A12 ∇ ψ̇ + A11 2 ∇ ψ + A13 + A21 + A22 ψ̇ 2 + A23
a a a a2
 2 #
+ A24 δϕψ̇ + A25 δϕ ˙ ψ̇ + A26 ∇δϕ + A27 δϕ2 + A28 δϕ ˙2 , (B9)

ie
a

showing very minor modifications from the action given by Eq. (52). Here, all modes {δϕ, β, ψ, E} are dynamical,
as {Φ, B} are typically the auxiliary modes in the system but have been eliminated through the gauge choice. Since

ev
mode E appears to have at least a first order derivative, a relabeling it as E = Ė allows treating E as an non-dynamical
mode. Thus, the final action is of the form
"
d3 k
Z    
Synchronous
SS = dt 3 S̃1 Ψ21 + S̃3 Ψ23 + Ψ1 S̃4 Ψ2 + S̃5 Ψ̇2 + S̃6 Ψ3 + S̃7 Ψ̇3 + Ψ3 S̃8 Ψ2 + S̃9 Ψ̇2
(2π) 2

r
#
+ S̃10 Ψ̇21 + S̃11 Ψ̇22 + S̃12 Ψ̇23 , (B10)
er
with coefficient details included in the supplementary document of Ref. [35]. Similar to the longitudinal gauge choice,
the substitution of β = B yields an extra scalar DoF from the rest of the gauge choices. While further setting β = 0,
the two dynamical modes are obtained but may lead to an overfixed gauge.
pe
2. Vectors

The vector sector contains a gauge invariant mode ui and the remaining modes {vi , wi , Vi } can be expressed as a
linear combination of each other. The gauge choice Vi = 0 is favoured as it eliminates contributions related to the
Levi-Civita tensor associated with the pseudovector. Regardless, all possible gauge choices are presented here.
ot

In the case of vi = 0, the action transforms to

  
3 C2
Z
vi =0 3 2 2 2 2

1
SV = dt d x a (∇ w) + (∇V) + 2 (∇ × V)(∇ w) + C4 V̇ V̇ + ∇ × v
a2 a
tn

2
(∇u) C8   (∇u) (∇ẇ)
+ C3 (∇ẇ)2 + C5 u̇2 + C6 + (∇u)(∇w) + V(∇ × u) + C9
a2 a a 
C10   1  1 d 2

+ (∇u̇)(∇w) + V(∇ × u̇) − V C11 (∇ × u̇) + 2 a C11 (∇ × u) , (B11)
a a a dt
rin

such that modes {ui , wi , Vi } are dynamical modes. In the other gauge choices, it was noted that vi is the only mode
that behaves as an auxiliary mode.
The gauge choice of wi = 0, followed by substitution of the equation of the auxiliary mode vi results in

(∇V)2 C72 (∇u)2

    
C4
Z
wi =0 3 3 2
SV = dt d x a C2 + C4 1 + V̇ + C6 − + C5 (u̇)2
a2 4C1 4C1 a2
ep

    
1 1 d 2
1 C4 C7
+ C8 − 2 a C10 V (∇ × u) + C11 − C10 − V̇ (∇ × u) , (B12)
a a dt a 2C1

for a total of two dynamical modes {ui , Vi } provided C1 ̸= 0, noting a difference in the number of dynamical modes
when compared to Eq. (B11).
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 27

For the gauge choice Vi = 0,

(∆w)2 C72 (∇u)2

    
C3
Z
Vi =0 3 3 2 2
SV = dt d x a C2 + C3 1 − (∇ ẇ) + C 6 − ( u̇) + C 7
a2 4C1 4C1 a2

d
    
1 d 2
(∇u) (∇w) C3 C7 (∇u) (∇ẇ)
+ C8 − 2 a C10 + C9 − C10 + , (B13)
a dt a 2C1 a

we
where now the two dynamical modes are {ui , wi } provided C1 ̸= 0. The latter two results, along with the gauge
invariant analysis, suggests that the gauge choice of vi = 0 requires further analysis within the BDLS framework.

ie
[1] K. N. Abazajian et al. Light Sterile Neutrinos: A White Paper. 4 2012.
[2] B. P. Abbott et al. GW170817: Observation of Gravitational Waves from a Binary Neutron Star Inspiral. Phys. Rev.
Lett., 119(16):161101, 2017.
[3] A. Addazi et al. Quantum gravity phenomenology at the dawn of the multi-messenger era—A review. Prog. Part. Nucl.

ev
Phys., 125:103948, 2022.
[4] Bobomurat Ahmedov, Konstantinos F. Dialektopoulos, Jackson Levi Said, Abdurakhmon Nosirov, Zinovia
Oikonomopoulou, and Odil Yunusov. Cosmological perturbations in the teleparallel analog of Horndeski gravity. JCAP,
08:074, 2023.
[5] D. S. Akerib et al. Results from a search for dark matter in the complete LUX exposure. Phys. Rev. Lett., 118(2):021303,
2017.

r
[6] Yashar Akrami et al. Modified Gravity and Cosmology. An Update by the CANTATA Network. Springer, 2021.
[7] R. Aldrovandi and J.G. Pereira. An Introduction to Geometrical Physics. World Scientific, 1995.
[8] Ruben Aldrovandi and José Geraldo Pereira. Teleparallel Gravity: An Introduction. Springer, 2013.
er
[9] R. Alves Batista et al. EuCAPT White Paper: Opportunities and Challenges for Theoretical Astroparticle Physics in the
Next Decade. 10 2021.
[10] Sebastian Bahamonde, Daniel Blixt, Konstantinos F. Dialektopoulos, and Anamaria Hell. Revisiting Stability in New
General Relativity. 4 2024.
[11] Sebastian Bahamonde, Christian G. Böhmer, and Martin Krššák. New classes of modified teleparallel gravity models.
pe
Phys. Lett. B, 775:37–43, 2017.
[12] Sebastian Bahamonde, Christian G. Böhmer, and Matthew Wright. Modified teleparallel theories of gravity. Phys. Rev.
D, 92(10):104042, 2015.
[13] Sebastian Bahamonde, Maria Caruana, Konstantinos F. Dialektopoulos, Viktor Gakis, Manuel Hohmann, Jackson
Levi Said, Emmanuel N. Saridakis, and Joseph Sultana. Gravitational-wave propagation and polarizations in the telepar-
allel analog of Horndeski gravity. Phys. Rev. D, 104(8):084082, 2021.
[14] Sebastian Bahamonde, Konstantinos F. Dialektopoulos, Celia Escamilla-Rivera, Gabriel Farrugia, Viktor Gakis, Martin
Hendry, Manuel Hohmann, Jackson Levi Said, Jurgen Mifsud, and Eleonora Di Valentino. Teleparallel gravity: from theory
ot

to cosmology. Rept. Prog. Phys., 86(2):026901, 2023.

[15] Sebastian Bahamonde, Konstantinos F. Dialektopoulos, Viktor Gakis, and Jackson Levi Said. Reviving Horndeski theory
using teleparallel gravity after GW170817. Phys. Rev. D, 101(8):084060, 2020.
[16] Sebastian Bahamonde, Konstantinos F. Dialektopoulos, Manuel Hohmann, and Jackson Levi Said. Post-Newtonian limit
tn

of Teleparallel Horndeski gravity. Class. Quant. Grav., 38(2):025006, 2020.

[17] Sebastian Bahamonde, Konstantinos F. Dialektopoulos, and Jackson Levi Said. Can Horndeski Theory be recast using
Teleparallel Gravity? Phys. Rev. D, 100(6):064018, 2019.
[18] Sebastian Bahamonde, Kai Flathmann, and Christian Pfeifer. Photon sphere and perihelion shift in weak f (T ) gravity.
Phys. Rev. D, 100(8):084064, 2019.
[19] Sebastian Bahamonde, Alexey Golovnev, Marı́a-José Guzmán, Jackson Levi Said, and Christian Pfeifer. Black holes in
rin

f(T,B) gravity: exact and perturbed solutions. JCAP, 01(01):037, 2022.

[20] Sebastian Bahamonde, Jackson Levi Said, and M. Zubair. Solar system tests in modified teleparallel gravity. JCAP,
10:024, 2020.
[21] Kazuharu Bamba, Salvatore Capozziello, Shin’ichi Nojiri, and Sergei D. Odintsov. Dark energy cosmology: the equivalent
description via different theoretical models and cosmography tests. Astrophys. Space Sci., 342:155–228, 2012.
[22] Laura Baudis. Dark matter detection. J. Phys. G, 43(4):044001, 2016.
[23] Jose Beltrán Jiménez, Lavinia Heisenberg, and Tomi Koivisto. The coupling of matter and spacetime geometry. Class.
ep

Quant. Grav., 37(19):195013, 2020.

[24] Gabriel R. Bengochea and Rafael Ferraro. Dark torsion as the cosmic speed-up. Phys. Rev. D, 79:124019, 2009.
[25] David Benisty and Denitsa Staicova. Testing late-time cosmic acceleration with uncorrelated baryon acoustic oscillation
dataset. Astron. Astrophys., 647:A38, 2021.
[26] Reginald Christian Bernardo and Jackson Levi Said. A data-driven reconstruction of Horndeski gravity via the Gaussian
processes. JCAP, 09:014, 2021.
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 28

[27] Reginald Christian Bernardo, Jackson Levi Said, Maria Caruana, and Stephen Appleby. Well-tempered teleparallel Horn-
deski cosmology: a teleparallel variation to the cosmological constant problem. JCAP, 10:078, 2021.
[28] Reginald Christian Bernardo, Jackson Levi Said, Maria Caruana, and Stephen Appleby. Well-tempered Minkowski solutions
in teleparallel Horndeski theory. Class. Quant. Grav., 39(1):015013, 2022.

d
[29] Gianfranco Bertone, Dan Hooper, and Joseph Silk. Particle dark matter: Evidence, candidates and constraints. Phys.
Rept., 405:279–390, 2005.
[30] Marco Bruni, Peter K. S. Dunsby, and George F. R. Ellis. Cosmological Perturbations and the Physical Meaning of

we
Gauge-invariant Variables. Astrophys. J. , 395:34, August 1992.
[31] Marco Bruni, Sabino Matarrese, Silvia Mollerach, and Sebastiano Sonego. Perturbations of space-time: Gauge transfor-
mations and gauge invariance at second order and beyond. Class. Quant. Grav., 14:2585–2606, 1997.
[32] Yi-Fu Cai, Salvatore Capozziello, Mariafelicia De Laurentis, and Emmanuel N. Saridakis. f(T) teleparallel gravity and
cosmology. Rept. Prog. Phys., 79(10):106901, 2016.
[33] Salvatore Capozziello, Maria Caruana, Jackson Levi Said, and Joseph Sultana. Ghost and Laplacian instabilities in
teleparallel Horndeski gravity. JCAP, 03:060, 2023.

ie
[34] Salvatore Capozziello and Mariafelicia De Laurentis. Extended Theories of Gravity. Phys. Rept., 509:167–321, 2011.
[35] Maria Caruana. BDLS Perturbations - Gauge Choices. https://github.com/MCaruana05/BDLS_Perturbations.git, 2024.
[Accessed 01-12-2024].
[36] Shih-Hung Chen, James B. Dent, Sourish Dutta, and Emmanuel N. Saridakis. Cosmological perturbations in f(T) gravity.
Phys. Rev. D, 83:023508, 2011.

ev
[37] Timothy Clifton, Pedro G. Ferreira, Antonio Padilla, and Constantinos Skordis. Modified Gravity and Cosmology. Phys.
Rept., 513:1–189, 2012.
[38] Edmund J. Copeland, M. Sami, and Shinji Tsujikawa. Dynamics of dark energy. Int. J. Mod. Phys. D, 15:1753–1936,
2006.
[39] Xue-Mei Deng. Probing f(T) gravity with gravitational time advancement. Class. Quant. Grav., 35(17):175013, 2018.

r
[40] Eleonora Di Valentino et al. Cosmology Intertwined III: f σ8 and S8 . Astropart. Phys., 131:102604, 2021.
[41] Eleonora Di Valentino et al. Snowmass2021 - Letter of interest cosmology intertwined I: Perspectives for the next decade.
Astropart. Phys., 131:102606, 2021.
[42] Eleonora Di Valentino et al. Snowmass2021 - Letter of interest cosmology intertwined II: The hubble constant tension.
er
Astropart. Phys., 131:102605, 2021.
[43] Konstantinos F. Dialektopoulos, Jackson Levi Said, and Zinovia Oikonomopoulou. Classification of teleparallel Horndeski
cosmology via Noether symmetries. Eur. Phys. J. C, 82(3):259, 2022.
[44] Scott Dodelson and Lawrence M. Widrow. Sterile-neutrinos as dark matter. Phys. Rev. Lett., 72:17–20, 1994.
[45] Jose Marı́a Ezquiaga and Miguel Zumalacárregui. Dark Energy After GW170817: Dead Ends and the Road Ahead. Phys.
pe
Rev. Lett., 119(25):251304, 2017.
[46] Gabriel Farrugia and Jackson Levi Said. Stability of the flat FLRW metric in f (T ) gravity. Phys. Rev. D, 94(12):124054,
2016.
[47] Gabriel Farrugia, Jackson Levi Said, and Andrew Finch. Gravitoelectromagnetism, Solar System Tests, and Weak-Field
Solutions in f (T, B) Gravity with Observational Constraints. Universe, 6(2):34, 2020.
[48] Jonathan L. Feng. Dark Matter Candidates from Particle Physics and Methods of Detection. Ann. Rev. Astron. Astrophys.,
48:495–545, 2010.
ot

[49] Rafael Ferraro and Franco Fiorini. Modified teleparallel gravity: Inflation without inflaton. Phys. Rev. D, 75:084031, 2007.
[50] Rafael Ferraro and Franco Fiorini. On Born-Infeld Gravity in Weitzenbock spacetime. Phys. Rev. D, 78:124019, 2008.
[51] Rafael Ferraro and Franco Fiorini. Remnant group of local Lorentz transformations in {(T ) theories. Phys. Rev. D,
91(6):064019, 2015.
[52] R. J. Gaitskell. Direct detection of dark matter. Ann. Rev. Nucl. Part. Sci., 54:315–359, 2004.
tn

[53] A. Goldstein et al. An Ordinary Short Gamma-Ray Burst with Extraordinary Implications: Fermi-GBM Detection of
GRB 170817A. Astrophys. J. Lett., 848(2):L14, 2017.
[54] Manuel Gonzalez-Espinoza, Giovanni Otalora, and Joel Saavedra. Stability of scalar perturbations in scalar-torsion f(T,ϕ)
gravity theories in the presence of a matter fluid. JCAP, 10:007, 2021.
[55] Kenji Hayashi and Takeshi Shirafuji. New general relativity. Phys. Rev. D, 19:3524–3553, 1979. [Addendum: Phys.Rev.D
24, 3312–3314 (1982)].
rin

[56] Friedrich W. Hehl, J. Dermott McCrea, Eckehard W. Mielke, and Yuval Ne’eman. Metric affine gauge theory of gravity:
Field equations, Noether identities, world spinors, and breaking of dilation invariance. Phys. Rept., 258:1–171, 1995.
[57] Gregory Walter Horndeski. Second-order scalar-tensor field equations in a four-dimensional space. Int. J. Theor. Phys.,
10:363–384, 1974.
[58] Lorenzo Iorio and Emmanuel N. Saridakis. Solar system constraints on f(T) gravity. Mon. Not. Roy. Astron. Soc., 427:1555,
2012.
ep

[59] Keisuke Izumi and Yen Chin Ong. Cosmological Perturbation in f(T) Gravity Revisited. JCAP, 06:029, 2013.
[60] Austin Joyce, Bhuvnesh Jain, Justin Khoury, and Mark Trodden. Beyond the Cosmological Standard Model. Phys. Rept.,
568:1–98, 2015.
[61] Tsutomu Kobayashi. Horndeski theory and beyond: a review. Rept. Prog. Phys., 82(8):086901, 2019.
[62] Tsutomu Kobayashi, Masahide Yamaguchi, and Jun’ichi Yokoyama. Generalized G-inflation: Inflation with the most
general second-order field equations. Prog. Theor. Phys., 126:511–529, 2011.
Pr

[63] Hideo Kodama and Misao Sasaki. Cosmological Perturbation Theory. Progress of Theoretical Physics Supplement, 78:1,

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418
 29

January 1984.
[64] M. Krssak, R. J. van den Hoogen, J. G. Pereira, C. G. Böhmer, and A. A. Coley. Teleparallel theories of gravity:
illuminating a fully invariant approach. Class. Quant. Grav., 36(18):183001, 2019.
[65] Martin Krššák and Emmanuel N. Saridakis. The covariant formulation of f(T) gravity. Class. Quant. Grav., 33(11):115009,

d
2016.
[66] Eric V. Linder. Einstein’s Other Gravity and the Acceleration of the Universe. Phys. Rev. D, 81:127301, 2010. [Erratum:
Phys.Rev.D 82, 109902 (2010)].

we
[67] Chung-Pei Ma and Edmund Bertschinger. Cosmological perturbation theory in the synchronous and conformal Newtonian
gauges. Astrophys. J., 455:7–25, 1995.
[68] Karim A. Malik and David Wands. Cosmological perturbations. Phys. Rept., 475:1–51, 2009.
[69] J. W. Maluf. The teleparallel equivalent of general relativity. Annalen Phys., 525:339–357, 2013.
[70] Tomas Ortin. Gravity and strings. Cambridge Monographs on Mathematical Physics. Cambridge Univ. Press, 3 2004.
[71] Andronikos Paliathanasis, Jackson Levi Said, and John D. Barrow. Stability of the Kasner Universe in f(T) Gravity. Phys.
Rev. D, 97(4):044008, 2018.

ie
[72] P. J. E. Peebles and Bharat Ratra. The Cosmological Constant and Dark Energy. Rev. Mod. Phys., 75:559–606, 2003.
[73] S. Perlmutter et al. Measurements of Ω and Λ from 42 High Redshift Supernovae. Astrophys. J., 517:565–586, 1999.
[74] Adam G. Riess et al. Observational evidence from supernovae for an accelerating universe and a cosmological constant.
Astron. J., 116:1009–1038, 1998.
[75] Denitsa Staicova. Hints for the H0 — rd tension in uncorrelated Baryon Acoustic Oscillations dataset. In 16th Mar-

ev
cel Grossmann Meeting on Recent Developments in Theoretical and Experimental General Relativity, Astrophysics and
Relativistic Field Theories, 11 2021.
[76] Denitsa Staicova and David Benisty. Constraining the dark energy models using baryon acoustic oscillations: An approach
independent of H0 · rd. Astron. Astrophys., 668:A135, 2022.
[77] Steven Weinberg. The Cosmological Constant Problem. Rev. Mod. Phys., 61:1–23, 1989.

r
[78] R. Weitzenböck. Invariantentheorie. P. Noordhoff, Groningen, 1923.

er
pe
ot
tn
rin
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=5051418